4 * The contents of this file are subject to the terms of the
5 * Common Development and Distribution License (the "License").
6 * You may not use this file except in compliance with the License.
8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9 * or http://www.opensolaris.org/os/licensing.
10 * See the License for the specific language governing permissions
11 * and limitations under the License.
13 * When distributing Covered Code, include this CDDL HEADER in each
14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15 * If applicable, add the following below this CDDL HEADER, with the
16 * fields enclosed by brackets "[]" replaced with your own identifying
17 * information: Portions Copyright [yyyy] [name of copyright owner]
23 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
24 * Copyright (c) 2011-2012 Pawel Jakub Dawidek <pawel@dawidek.net>.
25 * All rights reserved.
26 * Copyright 2013 Martin Matuska <mm@FreeBSD.org>. All rights reserved.
27 * Copyright 2014 Xin Li <delphij@FreeBSD.org>. All rights reserved.
28 * Copyright 2011 Nexenta Systems, Inc. All rights reserved.
29 * Copyright (c) 2014, Joyent, Inc. All rights reserved.
30 * Copyright (c) 2011, 2014 by Delphix. All rights reserved.
31 * Copyright (c) 2013 by Saso Kiselkov. All rights reserved.
32 * Copyright (c) 2013 Steven Hartland. All rights reserved.
33 * Copyright (c) 2014, Nexenta Systems, Inc. All rights reserved.
39 * This file handles the ioctls to /dev/zfs, used for configuring ZFS storage
40 * pools and filesystems, e.g. with /sbin/zfs and /sbin/zpool.
42 * There are two ways that we handle ioctls: the legacy way where almost
43 * all of the logic is in the ioctl callback, and the new way where most
44 * of the marshalling is handled in the common entry point, zfsdev_ioctl().
46 * Non-legacy ioctls should be registered by calling
47 * zfs_ioctl_register() from zfs_ioctl_init(). The ioctl is invoked
48 * from userland by lzc_ioctl().
50 * The registration arguments are as follows:
53 * The name of the ioctl. This is used for history logging. If the
54 * ioctl returns successfully (the callback returns 0), and allow_log
55 * is true, then a history log entry will be recorded with the input &
56 * output nvlists. The log entry can be printed with "zpool history -i".
59 * The ioctl request number, which userland will pass to ioctl(2).
60 * The ioctl numbers can change from release to release, because
61 * the caller (libzfs) must be matched to the kernel.
63 * zfs_secpolicy_func_t *secpolicy
64 * This function will be called before the zfs_ioc_func_t, to
65 * determine if this operation is permitted. It should return EPERM
66 * on failure, and 0 on success. Checks include determining if the
67 * dataset is visible in this zone, and if the user has either all
68 * zfs privileges in the zone (SYS_MOUNT), or has been granted permission
69 * to do this operation on this dataset with "zfs allow".
71 * zfs_ioc_namecheck_t namecheck
72 * This specifies what to expect in the zfs_cmd_t:zc_name -- a pool
73 * name, a dataset name, or nothing. If the name is not well-formed,
74 * the ioctl will fail and the callback will not be called.
75 * Therefore, the callback can assume that the name is well-formed
76 * (e.g. is null-terminated, doesn't have more than one '@' character,
77 * doesn't have invalid characters).
79 * zfs_ioc_poolcheck_t pool_check
80 * This specifies requirements on the pool state. If the pool does
81 * not meet them (is suspended or is readonly), the ioctl will fail
82 * and the callback will not be called. If any checks are specified
83 * (i.e. it is not POOL_CHECK_NONE), namecheck must not be NO_NAME.
84 * Multiple checks can be or-ed together (e.g. POOL_CHECK_SUSPENDED |
85 * POOL_CHECK_READONLY).
87 * boolean_t smush_outnvlist
88 * If smush_outnvlist is true, then the output is presumed to be a
89 * list of errors, and it will be "smushed" down to fit into the
90 * caller's buffer, by removing some entries and replacing them with a
91 * single "N_MORE_ERRORS" entry indicating how many were removed. See
92 * nvlist_smush() for details. If smush_outnvlist is false, and the
93 * outnvlist does not fit into the userland-provided buffer, then the
94 * ioctl will fail with ENOMEM.
96 * zfs_ioc_func_t *func
97 * The callback function that will perform the operation.
99 * The callback should return 0 on success, or an error number on
100 * failure. If the function fails, the userland ioctl will return -1,
101 * and errno will be set to the callback's return value. The callback
102 * will be called with the following arguments:
105 * The name of the pool or dataset to operate on, from
106 * zfs_cmd_t:zc_name. The 'namecheck' argument specifies the
107 * expected type (pool, dataset, or none).
110 * The input nvlist, deserialized from zfs_cmd_t:zc_nvlist_src. Or
111 * NULL if no input nvlist was provided. Changes to this nvlist are
112 * ignored. If the input nvlist could not be deserialized, the
113 * ioctl will fail and the callback will not be called.
116 * The output nvlist, initially empty. The callback can fill it in,
117 * and it will be returned to userland by serializing it into
118 * zfs_cmd_t:zc_nvlist_dst. If it is non-empty, and serialization
119 * fails (e.g. because the caller didn't supply a large enough
120 * buffer), then the overall ioctl will fail. See the
121 * 'smush_nvlist' argument above for additional behaviors.
123 * There are two typical uses of the output nvlist:
124 * - To return state, e.g. property values. In this case,
125 * smush_outnvlist should be false. If the buffer was not large
126 * enough, the caller will reallocate a larger buffer and try
129 * - To return multiple errors from an ioctl which makes on-disk
130 * changes. In this case, smush_outnvlist should be true.
131 * Ioctls which make on-disk modifications should generally not
132 * use the outnvl if they succeed, because the caller can not
133 * distinguish between the operation failing, and
134 * deserialization failing.
137 #include <sys/types.h>
138 #include <sys/param.h>
139 #include <sys/systm.h>
140 #include <sys/conf.h>
141 #include <sys/kernel.h>
142 #include <sys/lock.h>
143 #include <sys/malloc.h>
144 #include <sys/mutex.h>
145 #include <sys/proc.h>
146 #include <sys/errno.h>
149 #include <sys/file.h>
150 #include <sys/kmem.h>
151 #include <sys/conf.h>
152 #include <sys/cmn_err.h>
153 #include <sys/stat.h>
154 #include <sys/zfs_ioctl.h>
155 #include <sys/zfs_vfsops.h>
156 #include <sys/zfs_znode.h>
159 #include <sys/spa_impl.h>
160 #include <sys/vdev.h>
162 #include <sys/dsl_dir.h>
163 #include <sys/dsl_dataset.h>
164 #include <sys/dsl_prop.h>
165 #include <sys/dsl_deleg.h>
166 #include <sys/dmu_objset.h>
167 #include <sys/dmu_impl.h>
168 #include <sys/dmu_tx.h>
169 #include <sys/sunddi.h>
170 #include <sys/policy.h>
171 #include <sys/zone.h>
172 #include <sys/nvpair.h>
173 #include <sys/mount.h>
174 #include <sys/taskqueue.h>
176 #include <sys/varargs.h>
177 #include <sys/fs/zfs.h>
178 #include <sys/zfs_ctldir.h>
179 #include <sys/zfs_dir.h>
180 #include <sys/zfs_onexit.h>
181 #include <sys/zvol.h>
182 #include <sys/dsl_scan.h>
183 #include <sys/dmu_objset.h>
184 #include <sys/dmu_send.h>
185 #include <sys/dsl_destroy.h>
186 #include <sys/dsl_bookmark.h>
187 #include <sys/dsl_userhold.h>
188 #include <sys/zfeature.h>
190 #include "zfs_namecheck.h"
191 #include "zfs_prop.h"
192 #include "zfs_deleg.h"
193 #include "zfs_comutil.h"
194 #include "zfs_ioctl_compat.h"
196 CTASSERT(sizeof(zfs_cmd_t) < IOCPARM_MAX);
198 static int snapshot_list_prefetch;
199 SYSCTL_DECL(_vfs_zfs);
200 TUNABLE_INT("vfs.zfs.snapshot_list_prefetch", &snapshot_list_prefetch);
201 SYSCTL_INT(_vfs_zfs, OID_AUTO, snapshot_list_prefetch, CTLFLAG_RW,
202 &snapshot_list_prefetch, 0, "Prefetch data when listing snapshots");
204 static struct cdev *zfsdev;
206 extern void zfs_init(void);
207 extern void zfs_fini(void);
209 uint_t zfs_fsyncer_key;
210 extern uint_t rrw_tsd_key;
211 static uint_t zfs_allow_log_key;
213 typedef int zfs_ioc_legacy_func_t(zfs_cmd_t *);
214 typedef int zfs_ioc_func_t(const char *, nvlist_t *, nvlist_t *);
215 typedef int zfs_secpolicy_func_t(zfs_cmd_t *, nvlist_t *, cred_t *);
221 } zfs_ioc_namecheck_t;
224 POOL_CHECK_NONE = 1 << 0,
225 POOL_CHECK_SUSPENDED = 1 << 1,
226 POOL_CHECK_READONLY = 1 << 2,
227 } zfs_ioc_poolcheck_t;
229 typedef struct zfs_ioc_vec {
230 zfs_ioc_legacy_func_t *zvec_legacy_func;
231 zfs_ioc_func_t *zvec_func;
232 zfs_secpolicy_func_t *zvec_secpolicy;
233 zfs_ioc_namecheck_t zvec_namecheck;
234 boolean_t zvec_allow_log;
235 zfs_ioc_poolcheck_t zvec_pool_check;
236 boolean_t zvec_smush_outnvlist;
237 const char *zvec_name;
240 /* This array is indexed by zfs_userquota_prop_t */
241 static const char *userquota_perms[] = {
242 ZFS_DELEG_PERM_USERUSED,
243 ZFS_DELEG_PERM_USERQUOTA,
244 ZFS_DELEG_PERM_GROUPUSED,
245 ZFS_DELEG_PERM_GROUPQUOTA,
248 static int zfs_ioc_userspace_upgrade(zfs_cmd_t *zc);
249 static int zfs_check_settable(const char *name, nvpair_t *property,
251 static int zfs_check_clearable(char *dataset, nvlist_t *props,
253 static int zfs_fill_zplprops_root(uint64_t, nvlist_t *, nvlist_t *,
255 int zfs_set_prop_nvlist(const char *, zprop_source_t, nvlist_t *, nvlist_t *);
256 static int get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp);
258 static void zfsdev_close(void *data);
260 static int zfs_prop_activate_feature(spa_t *spa, spa_feature_t feature);
262 /* _NOTE(PRINTFLIKE(4)) - this is printf-like, but lint is too whiney */
264 __dprintf(const char *file, const char *func, int line, const char *fmt, ...)
271 * Get rid of annoying "../common/" prefix to filename.
273 newfile = strrchr(file, '/');
274 if (newfile != NULL) {
275 newfile = newfile + 1; /* Get rid of leading / */
281 (void) vsnprintf(buf, sizeof (buf), fmt, adx);
285 * To get this data, use the zfs-dprintf probe as so:
286 * dtrace -q -n 'zfs-dprintf \
287 * /stringof(arg0) == "dbuf.c"/ \
288 * {printf("%s: %s", stringof(arg1), stringof(arg3))}'
290 * arg1 = function name
294 DTRACE_PROBE4(zfs__dprintf,
295 char *, newfile, char *, func, int, line, char *, buf);
299 history_str_free(char *buf)
301 kmem_free(buf, HIS_MAX_RECORD_LEN);
305 history_str_get(zfs_cmd_t *zc)
309 if (zc->zc_history == 0)
312 buf = kmem_alloc(HIS_MAX_RECORD_LEN, KM_SLEEP);
313 if (copyinstr((void *)(uintptr_t)zc->zc_history,
314 buf, HIS_MAX_RECORD_LEN, NULL) != 0) {
315 history_str_free(buf);
319 buf[HIS_MAX_RECORD_LEN -1] = '\0';
325 * Check to see if the named dataset is currently defined as bootable
328 zfs_is_bootfs(const char *name)
332 if (dmu_objset_hold(name, FTAG, &os) == 0) {
334 ret = (dmu_objset_id(os) == spa_bootfs(dmu_objset_spa(os)));
335 dmu_objset_rele(os, FTAG);
342 * Return non-zero if the spa version is less than requested version.
345 zfs_earlier_version(const char *name, int version)
349 if (spa_open(name, &spa, FTAG) == 0) {
350 if (spa_version(spa) < version) {
351 spa_close(spa, FTAG);
354 spa_close(spa, FTAG);
360 * Return TRUE if the ZPL version is less than requested version.
363 zpl_earlier_version(const char *name, int version)
366 boolean_t rc = B_TRUE;
368 if (dmu_objset_hold(name, FTAG, &os) == 0) {
371 if (dmu_objset_type(os) != DMU_OST_ZFS) {
372 dmu_objset_rele(os, FTAG);
375 /* XXX reading from non-owned objset */
376 if (zfs_get_zplprop(os, ZFS_PROP_VERSION, &zplversion) == 0)
377 rc = zplversion < version;
378 dmu_objset_rele(os, FTAG);
384 zfs_log_history(zfs_cmd_t *zc)
389 if ((buf = history_str_get(zc)) == NULL)
392 if (spa_open(zc->zc_name, &spa, FTAG) == 0) {
393 if (spa_version(spa) >= SPA_VERSION_ZPOOL_HISTORY)
394 (void) spa_history_log(spa, buf);
395 spa_close(spa, FTAG);
397 history_str_free(buf);
401 * Policy for top-level read operations (list pools). Requires no privileges,
402 * and can be used in the local zone, as there is no associated dataset.
406 zfs_secpolicy_none(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
412 * Policy for dataset read operations (list children, get statistics). Requires
413 * no privileges, but must be visible in the local zone.
417 zfs_secpolicy_read(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
419 if (INGLOBALZONE(curthread) ||
420 zone_dataset_visible(zc->zc_name, NULL))
423 return (SET_ERROR(ENOENT));
427 zfs_dozonecheck_impl(const char *dataset, uint64_t zoned, cred_t *cr)
432 * The dataset must be visible by this zone -- check this first
433 * so they don't see EPERM on something they shouldn't know about.
435 if (!INGLOBALZONE(curthread) &&
436 !zone_dataset_visible(dataset, &writable))
437 return (SET_ERROR(ENOENT));
439 if (INGLOBALZONE(curthread)) {
441 * If the fs is zoned, only root can access it from the
444 if (secpolicy_zfs(cr) && zoned)
445 return (SET_ERROR(EPERM));
448 * If we are in a local zone, the 'zoned' property must be set.
451 return (SET_ERROR(EPERM));
453 /* must be writable by this zone */
455 return (SET_ERROR(EPERM));
461 zfs_dozonecheck(const char *dataset, cred_t *cr)
465 if (dsl_prop_get_integer(dataset, "jailed", &zoned, NULL))
466 return (SET_ERROR(ENOENT));
468 return (zfs_dozonecheck_impl(dataset, zoned, cr));
472 zfs_dozonecheck_ds(const char *dataset, dsl_dataset_t *ds, cred_t *cr)
476 if (dsl_prop_get_int_ds(ds, "jailed", &zoned))
477 return (SET_ERROR(ENOENT));
479 return (zfs_dozonecheck_impl(dataset, zoned, cr));
483 zfs_secpolicy_write_perms_ds(const char *name, dsl_dataset_t *ds,
484 const char *perm, cred_t *cr)
488 error = zfs_dozonecheck_ds(name, ds, cr);
490 error = secpolicy_zfs(cr);
492 error = dsl_deleg_access_impl(ds, perm, cr);
498 zfs_secpolicy_write_perms(const char *name, const char *perm, cred_t *cr)
504 error = dsl_pool_hold(name, FTAG, &dp);
508 error = dsl_dataset_hold(dp, name, FTAG, &ds);
510 dsl_pool_rele(dp, FTAG);
514 error = zfs_secpolicy_write_perms_ds(name, ds, perm, cr);
516 dsl_dataset_rele(ds, FTAG);
517 dsl_pool_rele(dp, FTAG);
523 * Policy for setting the security label property.
525 * Returns 0 for success, non-zero for access and other errors.
528 zfs_set_slabel_policy(const char *name, char *strval, cred_t *cr)
530 char ds_hexsl[MAXNAMELEN];
531 bslabel_t ds_sl, new_sl;
532 boolean_t new_default = FALSE;
534 int needed_priv = -1;
537 /* First get the existing dataset label. */
538 error = dsl_prop_get(name, zfs_prop_to_name(ZFS_PROP_MLSLABEL),
539 1, sizeof (ds_hexsl), &ds_hexsl, NULL);
541 return (SET_ERROR(EPERM));
543 if (strcasecmp(strval, ZFS_MLSLABEL_DEFAULT) == 0)
546 /* The label must be translatable */
547 if (!new_default && (hexstr_to_label(strval, &new_sl) != 0))
548 return (SET_ERROR(EINVAL));
551 * In a non-global zone, disallow attempts to set a label that
552 * doesn't match that of the zone; otherwise no other checks
555 if (!INGLOBALZONE(curproc)) {
556 if (new_default || !blequal(&new_sl, CR_SL(CRED())))
557 return (SET_ERROR(EPERM));
562 * For global-zone datasets (i.e., those whose zoned property is
563 * "off", verify that the specified new label is valid for the
566 if (dsl_prop_get_integer(name,
567 zfs_prop_to_name(ZFS_PROP_ZONED), &zoned, NULL))
568 return (SET_ERROR(EPERM));
570 if (zfs_check_global_label(name, strval) != 0)
571 return (SET_ERROR(EPERM));
575 * If the existing dataset label is nondefault, check if the
576 * dataset is mounted (label cannot be changed while mounted).
577 * Get the zfsvfs; if there isn't one, then the dataset isn't
578 * mounted (or isn't a dataset, doesn't exist, ...).
580 if (strcasecmp(ds_hexsl, ZFS_MLSLABEL_DEFAULT) != 0) {
582 static char *setsl_tag = "setsl_tag";
585 * Try to own the dataset; abort if there is any error,
586 * (e.g., already mounted, in use, or other error).
588 error = dmu_objset_own(name, DMU_OST_ZFS, B_TRUE,
591 return (SET_ERROR(EPERM));
593 dmu_objset_disown(os, setsl_tag);
596 needed_priv = PRIV_FILE_DOWNGRADE_SL;
600 if (hexstr_to_label(strval, &new_sl) != 0)
601 return (SET_ERROR(EPERM));
603 if (blstrictdom(&ds_sl, &new_sl))
604 needed_priv = PRIV_FILE_DOWNGRADE_SL;
605 else if (blstrictdom(&new_sl, &ds_sl))
606 needed_priv = PRIV_FILE_UPGRADE_SL;
608 /* dataset currently has a default label */
610 needed_priv = PRIV_FILE_UPGRADE_SL;
614 if (needed_priv != -1)
615 return (PRIV_POLICY(cr, needed_priv, B_FALSE, EPERM, NULL));
618 #endif /* SECLABEL */
621 zfs_secpolicy_setprop(const char *dsname, zfs_prop_t prop, nvpair_t *propval,
627 * Check permissions for special properties.
632 * Disallow setting of 'zoned' from within a local zone.
634 if (!INGLOBALZONE(curthread))
635 return (SET_ERROR(EPERM));
639 case ZFS_PROP_FILESYSTEM_LIMIT:
640 case ZFS_PROP_SNAPSHOT_LIMIT:
641 if (!INGLOBALZONE(curthread)) {
643 char setpoint[MAXNAMELEN];
645 * Unprivileged users are allowed to modify the
646 * limit on things *under* (ie. contained by)
647 * the thing they own.
649 if (dsl_prop_get_integer(dsname, "jailed", &zoned,
651 return (SET_ERROR(EPERM));
652 if (!zoned || strlen(dsname) <= strlen(setpoint))
653 return (SET_ERROR(EPERM));
657 case ZFS_PROP_MLSLABEL:
659 if (!is_system_labeled())
660 return (SET_ERROR(EPERM));
662 if (nvpair_value_string(propval, &strval) == 0) {
665 err = zfs_set_slabel_policy(dsname, strval, CRED());
675 return (zfs_secpolicy_write_perms(dsname, zfs_prop_to_name(prop), cr));
680 zfs_secpolicy_set_fsacl(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
684 error = zfs_dozonecheck(zc->zc_name, cr);
689 * permission to set permissions will be evaluated later in
690 * dsl_deleg_can_allow()
697 zfs_secpolicy_rollback(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
699 return (zfs_secpolicy_write_perms(zc->zc_name,
700 ZFS_DELEG_PERM_ROLLBACK, cr));
705 zfs_secpolicy_send(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
713 * Generate the current snapshot name from the given objsetid, then
714 * use that name for the secpolicy/zone checks.
716 cp = strchr(zc->zc_name, '@');
718 return (SET_ERROR(EINVAL));
719 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
723 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &ds);
725 dsl_pool_rele(dp, FTAG);
729 dsl_dataset_name(ds, zc->zc_name);
731 error = zfs_secpolicy_write_perms_ds(zc->zc_name, ds,
732 ZFS_DELEG_PERM_SEND, cr);
733 dsl_dataset_rele(ds, FTAG);
734 dsl_pool_rele(dp, FTAG);
741 zfs_secpolicy_send_new(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
743 return (zfs_secpolicy_write_perms(zc->zc_name,
744 ZFS_DELEG_PERM_SEND, cr));
749 zfs_secpolicy_deleg_share(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
754 if ((error = lookupname(zc->zc_value, UIO_SYSSPACE,
755 NO_FOLLOW, NULL, &vp)) != 0)
758 /* Now make sure mntpnt and dataset are ZFS */
760 if (strcmp(vp->v_vfsp->mnt_stat.f_fstypename, "zfs") != 0 ||
761 (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource),
762 zc->zc_name) != 0)) {
764 return (SET_ERROR(EPERM));
768 return (dsl_deleg_access(zc->zc_name,
769 ZFS_DELEG_PERM_SHARE, cr));
773 zfs_secpolicy_share(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
775 if (!INGLOBALZONE(curthread))
776 return (SET_ERROR(EPERM));
778 if (secpolicy_nfs(cr) == 0) {
781 return (zfs_secpolicy_deleg_share(zc, innvl, cr));
786 zfs_secpolicy_smb_acl(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
788 if (!INGLOBALZONE(curthread))
789 return (SET_ERROR(EPERM));
791 if (secpolicy_smb(cr) == 0) {
794 return (zfs_secpolicy_deleg_share(zc, innvl, cr));
799 zfs_get_parent(const char *datasetname, char *parent, int parentsize)
804 * Remove the @bla or /bla from the end of the name to get the parent.
806 (void) strncpy(parent, datasetname, parentsize);
807 cp = strrchr(parent, '@');
811 cp = strrchr(parent, '/');
813 return (SET_ERROR(ENOENT));
821 zfs_secpolicy_destroy_perms(const char *name, cred_t *cr)
825 if ((error = zfs_secpolicy_write_perms(name,
826 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
829 return (zfs_secpolicy_write_perms(name, ZFS_DELEG_PERM_DESTROY, cr));
834 zfs_secpolicy_destroy(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
836 return (zfs_secpolicy_destroy_perms(zc->zc_name, cr));
840 * Destroying snapshots with delegated permissions requires
841 * descendant mount and destroy permissions.
845 zfs_secpolicy_destroy_snaps(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
848 nvpair_t *pair, *nextpair;
851 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
852 return (SET_ERROR(EINVAL));
853 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
855 nextpair = nvlist_next_nvpair(snaps, pair);
856 error = zfs_secpolicy_destroy_perms(nvpair_name(pair), cr);
857 if (error == ENOENT) {
859 * Ignore any snapshots that don't exist (we consider
860 * them "already destroyed"). Remove the name from the
861 * nvl here in case the snapshot is created between
862 * now and when we try to destroy it (in which case
863 * we don't want to destroy it since we haven't
864 * checked for permission).
866 fnvlist_remove_nvpair(snaps, pair);
877 zfs_secpolicy_rename_perms(const char *from, const char *to, cred_t *cr)
879 char parentname[MAXNAMELEN];
882 if ((error = zfs_secpolicy_write_perms(from,
883 ZFS_DELEG_PERM_RENAME, cr)) != 0)
886 if ((error = zfs_secpolicy_write_perms(from,
887 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
890 if ((error = zfs_get_parent(to, parentname,
891 sizeof (parentname))) != 0)
894 if ((error = zfs_secpolicy_write_perms(parentname,
895 ZFS_DELEG_PERM_CREATE, cr)) != 0)
898 if ((error = zfs_secpolicy_write_perms(parentname,
899 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
907 zfs_secpolicy_rename(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
912 if ((zc->zc_cookie & 1) != 0) {
914 * This is recursive rename, so the starting snapshot might
915 * not exist. Check file system or volume permission instead.
917 at = strchr(zc->zc_name, '@');
923 error = zfs_secpolicy_rename_perms(zc->zc_name, zc->zc_value, cr);
933 zfs_secpolicy_promote(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
936 dsl_dataset_t *clone;
939 error = zfs_secpolicy_write_perms(zc->zc_name,
940 ZFS_DELEG_PERM_PROMOTE, cr);
944 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
948 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &clone);
951 char parentname[MAXNAMELEN];
952 dsl_dataset_t *origin = NULL;
956 error = dsl_dataset_hold_obj(dd->dd_pool,
957 dd->dd_phys->dd_origin_obj, FTAG, &origin);
959 dsl_dataset_rele(clone, FTAG);
960 dsl_pool_rele(dp, FTAG);
964 error = zfs_secpolicy_write_perms_ds(zc->zc_name, clone,
965 ZFS_DELEG_PERM_MOUNT, cr);
967 dsl_dataset_name(origin, parentname);
969 error = zfs_secpolicy_write_perms_ds(parentname, origin,
970 ZFS_DELEG_PERM_PROMOTE, cr);
972 dsl_dataset_rele(clone, FTAG);
973 dsl_dataset_rele(origin, FTAG);
975 dsl_pool_rele(dp, FTAG);
981 zfs_secpolicy_recv(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
985 if ((error = zfs_secpolicy_write_perms(zc->zc_name,
986 ZFS_DELEG_PERM_RECEIVE, cr)) != 0)
989 if ((error = zfs_secpolicy_write_perms(zc->zc_name,
990 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
993 return (zfs_secpolicy_write_perms(zc->zc_name,
994 ZFS_DELEG_PERM_CREATE, cr));
998 zfs_secpolicy_snapshot_perms(const char *name, cred_t *cr)
1000 return (zfs_secpolicy_write_perms(name,
1001 ZFS_DELEG_PERM_SNAPSHOT, cr));
1005 * Check for permission to create each snapshot in the nvlist.
1009 zfs_secpolicy_snapshot(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1015 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
1016 return (SET_ERROR(EINVAL));
1017 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
1018 pair = nvlist_next_nvpair(snaps, pair)) {
1019 char *name = nvpair_name(pair);
1020 char *atp = strchr(name, '@');
1023 error = SET_ERROR(EINVAL);
1027 error = zfs_secpolicy_snapshot_perms(name, cr);
1036 * Check for permission to create each snapshot in the nvlist.
1040 zfs_secpolicy_bookmark(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1044 for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL);
1045 pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) {
1046 char *name = nvpair_name(pair);
1047 char *hashp = strchr(name, '#');
1049 if (hashp == NULL) {
1050 error = SET_ERROR(EINVAL);
1054 error = zfs_secpolicy_write_perms(name,
1055 ZFS_DELEG_PERM_BOOKMARK, cr);
1065 zfs_secpolicy_destroy_bookmarks(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1067 nvpair_t *pair, *nextpair;
1070 for (pair = nvlist_next_nvpair(innvl, NULL); pair != NULL;
1072 char *name = nvpair_name(pair);
1073 char *hashp = strchr(name, '#');
1074 nextpair = nvlist_next_nvpair(innvl, pair);
1076 if (hashp == NULL) {
1077 error = SET_ERROR(EINVAL);
1082 error = zfs_secpolicy_write_perms(name,
1083 ZFS_DELEG_PERM_DESTROY, cr);
1085 if (error == ENOENT) {
1087 * Ignore any filesystems that don't exist (we consider
1088 * their bookmarks "already destroyed"). Remove
1089 * the name from the nvl here in case the filesystem
1090 * is created between now and when we try to destroy
1091 * the bookmark (in which case we don't want to
1092 * destroy it since we haven't checked for permission).
1094 fnvlist_remove_nvpair(innvl, pair);
1106 zfs_secpolicy_log_history(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1109 * Even root must have a proper TSD so that we know what pool
1112 if (tsd_get(zfs_allow_log_key) == NULL)
1113 return (SET_ERROR(EPERM));
1118 zfs_secpolicy_create_clone(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1120 char parentname[MAXNAMELEN];
1124 if ((error = zfs_get_parent(zc->zc_name, parentname,
1125 sizeof (parentname))) != 0)
1128 if (nvlist_lookup_string(innvl, "origin", &origin) == 0 &&
1129 (error = zfs_secpolicy_write_perms(origin,
1130 ZFS_DELEG_PERM_CLONE, cr)) != 0)
1133 if ((error = zfs_secpolicy_write_perms(parentname,
1134 ZFS_DELEG_PERM_CREATE, cr)) != 0)
1137 return (zfs_secpolicy_write_perms(parentname,
1138 ZFS_DELEG_PERM_MOUNT, cr));
1142 * Policy for pool operations - create/destroy pools, add vdevs, etc. Requires
1143 * SYS_CONFIG privilege, which is not available in a local zone.
1147 zfs_secpolicy_config(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1149 if (secpolicy_sys_config(cr, B_FALSE) != 0)
1150 return (SET_ERROR(EPERM));
1156 * Policy for object to name lookups.
1160 zfs_secpolicy_diff(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1164 if ((error = secpolicy_sys_config(cr, B_FALSE)) == 0)
1167 error = zfs_secpolicy_write_perms(zc->zc_name, ZFS_DELEG_PERM_DIFF, cr);
1172 * Policy for fault injection. Requires all privileges.
1176 zfs_secpolicy_inject(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1178 return (secpolicy_zinject(cr));
1183 zfs_secpolicy_inherit_prop(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1185 zfs_prop_t prop = zfs_name_to_prop(zc->zc_value);
1187 if (prop == ZPROP_INVAL) {
1188 if (!zfs_prop_user(zc->zc_value))
1189 return (SET_ERROR(EINVAL));
1190 return (zfs_secpolicy_write_perms(zc->zc_name,
1191 ZFS_DELEG_PERM_USERPROP, cr));
1193 return (zfs_secpolicy_setprop(zc->zc_name, prop,
1199 zfs_secpolicy_userspace_one(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1201 int err = zfs_secpolicy_read(zc, innvl, cr);
1205 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
1206 return (SET_ERROR(EINVAL));
1208 if (zc->zc_value[0] == 0) {
1210 * They are asking about a posix uid/gid. If it's
1211 * themself, allow it.
1213 if (zc->zc_objset_type == ZFS_PROP_USERUSED ||
1214 zc->zc_objset_type == ZFS_PROP_USERQUOTA) {
1215 if (zc->zc_guid == crgetuid(cr))
1218 if (groupmember(zc->zc_guid, cr))
1223 return (zfs_secpolicy_write_perms(zc->zc_name,
1224 userquota_perms[zc->zc_objset_type], cr));
1228 zfs_secpolicy_userspace_many(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1230 int err = zfs_secpolicy_read(zc, innvl, cr);
1234 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
1235 return (SET_ERROR(EINVAL));
1237 return (zfs_secpolicy_write_perms(zc->zc_name,
1238 userquota_perms[zc->zc_objset_type], cr));
1243 zfs_secpolicy_userspace_upgrade(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1245 return (zfs_secpolicy_setprop(zc->zc_name, ZFS_PROP_VERSION,
1251 zfs_secpolicy_hold(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1257 error = nvlist_lookup_nvlist(innvl, "holds", &holds);
1259 return (SET_ERROR(EINVAL));
1261 for (pair = nvlist_next_nvpair(holds, NULL); pair != NULL;
1262 pair = nvlist_next_nvpair(holds, pair)) {
1263 char fsname[MAXNAMELEN];
1264 error = dmu_fsname(nvpair_name(pair), fsname);
1267 error = zfs_secpolicy_write_perms(fsname,
1268 ZFS_DELEG_PERM_HOLD, cr);
1277 zfs_secpolicy_release(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1282 for (pair = nvlist_next_nvpair(innvl, NULL); pair != NULL;
1283 pair = nvlist_next_nvpair(innvl, pair)) {
1284 char fsname[MAXNAMELEN];
1285 error = dmu_fsname(nvpair_name(pair), fsname);
1288 error = zfs_secpolicy_write_perms(fsname,
1289 ZFS_DELEG_PERM_RELEASE, cr);
1297 * Policy for allowing temporary snapshots to be taken or released
1300 zfs_secpolicy_tmp_snapshot(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1303 * A temporary snapshot is the same as a snapshot,
1304 * hold, destroy and release all rolled into one.
1305 * Delegated diff alone is sufficient that we allow this.
1309 if ((error = zfs_secpolicy_write_perms(zc->zc_name,
1310 ZFS_DELEG_PERM_DIFF, cr)) == 0)
1313 error = zfs_secpolicy_snapshot_perms(zc->zc_name, cr);
1315 error = zfs_secpolicy_hold(zc, innvl, cr);
1317 error = zfs_secpolicy_release(zc, innvl, cr);
1319 error = zfs_secpolicy_destroy(zc, innvl, cr);
1324 * Returns the nvlist as specified by the user in the zfs_cmd_t.
1327 get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp)
1331 nvlist_t *list = NULL;
1334 * Read in and unpack the user-supplied nvlist.
1337 return (SET_ERROR(EINVAL));
1339 packed = kmem_alloc(size, KM_SLEEP);
1341 if ((error = ddi_copyin((void *)(uintptr_t)nvl, packed, size,
1343 kmem_free(packed, size);
1347 if ((error = nvlist_unpack(packed, size, &list, 0)) != 0) {
1348 kmem_free(packed, size);
1352 kmem_free(packed, size);
1359 * Reduce the size of this nvlist until it can be serialized in 'max' bytes.
1360 * Entries will be removed from the end of the nvlist, and one int32 entry
1361 * named "N_MORE_ERRORS" will be added indicating how many entries were
1365 nvlist_smush(nvlist_t *errors, size_t max)
1369 size = fnvlist_size(errors);
1372 nvpair_t *more_errors;
1376 return (SET_ERROR(ENOMEM));
1378 fnvlist_add_int32(errors, ZPROP_N_MORE_ERRORS, 0);
1379 more_errors = nvlist_prev_nvpair(errors, NULL);
1382 nvpair_t *pair = nvlist_prev_nvpair(errors,
1384 fnvlist_remove_nvpair(errors, pair);
1386 size = fnvlist_size(errors);
1387 } while (size > max);
1389 fnvlist_remove_nvpair(errors, more_errors);
1390 fnvlist_add_int32(errors, ZPROP_N_MORE_ERRORS, n);
1391 ASSERT3U(fnvlist_size(errors), <=, max);
1398 put_nvlist(zfs_cmd_t *zc, nvlist_t *nvl)
1400 char *packed = NULL;
1404 size = fnvlist_size(nvl);
1406 if (size > zc->zc_nvlist_dst_size) {
1408 * Solaris returns ENOMEM here, because even if an error is
1409 * returned from an ioctl(2), new zc_nvlist_dst_size will be
1410 * passed to the userland. This is not the case for FreeBSD.
1411 * We need to return 0, so the kernel will copy the
1412 * zc_nvlist_dst_size back and the userland can discover that a
1413 * bigger buffer is needed.
1417 packed = fnvlist_pack(nvl, &size);
1418 if (ddi_copyout(packed, (void *)(uintptr_t)zc->zc_nvlist_dst,
1419 size, zc->zc_iflags) != 0)
1420 error = SET_ERROR(EFAULT);
1421 fnvlist_pack_free(packed, size);
1424 zc->zc_nvlist_dst_size = size;
1425 zc->zc_nvlist_dst_filled = B_TRUE;
1430 getzfsvfs(const char *dsname, zfsvfs_t **zfvp)
1435 error = dmu_objset_hold(dsname, FTAG, &os);
1438 if (dmu_objset_type(os) != DMU_OST_ZFS) {
1439 dmu_objset_rele(os, FTAG);
1440 return (SET_ERROR(EINVAL));
1443 mutex_enter(&os->os_user_ptr_lock);
1444 *zfvp = dmu_objset_get_user(os);
1446 VFS_HOLD((*zfvp)->z_vfs);
1448 error = SET_ERROR(ESRCH);
1450 mutex_exit(&os->os_user_ptr_lock);
1451 dmu_objset_rele(os, FTAG);
1456 * Find a zfsvfs_t for a mounted filesystem, or create our own, in which
1457 * case its z_vfs will be NULL, and it will be opened as the owner.
1458 * If 'writer' is set, the z_teardown_lock will be held for RW_WRITER,
1459 * which prevents all vnode ops from running.
1462 zfsvfs_hold(const char *name, void *tag, zfsvfs_t **zfvp, boolean_t writer)
1466 if (getzfsvfs(name, zfvp) != 0)
1467 error = zfsvfs_create(name, zfvp);
1469 rrm_enter(&(*zfvp)->z_teardown_lock, (writer) ? RW_WRITER :
1471 if ((*zfvp)->z_unmounted) {
1473 * XXX we could probably try again, since the unmounting
1474 * thread should be just about to disassociate the
1475 * objset from the zfsvfs.
1477 rrm_exit(&(*zfvp)->z_teardown_lock, tag);
1478 return (SET_ERROR(EBUSY));
1485 zfsvfs_rele(zfsvfs_t *zfsvfs, void *tag)
1487 rrm_exit(&zfsvfs->z_teardown_lock, tag);
1489 if (zfsvfs->z_vfs) {
1490 VFS_RELE(zfsvfs->z_vfs);
1492 dmu_objset_disown(zfsvfs->z_os, zfsvfs);
1493 zfsvfs_free(zfsvfs);
1498 zfs_ioc_pool_create(zfs_cmd_t *zc)
1501 nvlist_t *config, *props = NULL;
1502 nvlist_t *rootprops = NULL;
1503 nvlist_t *zplprops = NULL;
1505 if (error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1506 zc->zc_iflags, &config))
1509 if (zc->zc_nvlist_src_size != 0 && (error =
1510 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1511 zc->zc_iflags, &props))) {
1512 nvlist_free(config);
1517 nvlist_t *nvl = NULL;
1518 uint64_t version = SPA_VERSION;
1520 (void) nvlist_lookup_uint64(props,
1521 zpool_prop_to_name(ZPOOL_PROP_VERSION), &version);
1522 if (!SPA_VERSION_IS_SUPPORTED(version)) {
1523 error = SET_ERROR(EINVAL);
1524 goto pool_props_bad;
1526 (void) nvlist_lookup_nvlist(props, ZPOOL_ROOTFS_PROPS, &nvl);
1528 error = nvlist_dup(nvl, &rootprops, KM_SLEEP);
1530 nvlist_free(config);
1534 (void) nvlist_remove_all(props, ZPOOL_ROOTFS_PROPS);
1536 VERIFY(nvlist_alloc(&zplprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
1537 error = zfs_fill_zplprops_root(version, rootprops,
1540 goto pool_props_bad;
1543 error = spa_create(zc->zc_name, config, props, zplprops);
1546 * Set the remaining root properties
1548 if (!error && (error = zfs_set_prop_nvlist(zc->zc_name,
1549 ZPROP_SRC_LOCAL, rootprops, NULL)) != 0)
1550 (void) spa_destroy(zc->zc_name);
1553 nvlist_free(rootprops);
1554 nvlist_free(zplprops);
1555 nvlist_free(config);
1562 zfs_ioc_pool_destroy(zfs_cmd_t *zc)
1565 zfs_log_history(zc);
1566 error = spa_destroy(zc->zc_name);
1568 zvol_remove_minors(zc->zc_name);
1573 zfs_ioc_pool_import(zfs_cmd_t *zc)
1575 nvlist_t *config, *props = NULL;
1579 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1580 zc->zc_iflags, &config)) != 0)
1583 if (zc->zc_nvlist_src_size != 0 && (error =
1584 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1585 zc->zc_iflags, &props))) {
1586 nvlist_free(config);
1590 if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, &guid) != 0 ||
1591 guid != zc->zc_guid)
1592 error = SET_ERROR(EINVAL);
1594 error = spa_import(zc->zc_name, config, props, zc->zc_cookie);
1596 if (zc->zc_nvlist_dst != 0) {
1599 if ((err = put_nvlist(zc, config)) != 0)
1603 nvlist_free(config);
1612 zfs_ioc_pool_export(zfs_cmd_t *zc)
1615 boolean_t force = (boolean_t)zc->zc_cookie;
1616 boolean_t hardforce = (boolean_t)zc->zc_guid;
1618 zfs_log_history(zc);
1619 error = spa_export(zc->zc_name, NULL, force, hardforce);
1621 zvol_remove_minors(zc->zc_name);
1626 zfs_ioc_pool_configs(zfs_cmd_t *zc)
1631 if ((configs = spa_all_configs(&zc->zc_cookie)) == NULL)
1632 return (SET_ERROR(EEXIST));
1634 error = put_nvlist(zc, configs);
1636 nvlist_free(configs);
1643 * zc_name name of the pool
1646 * zc_cookie real errno
1647 * zc_nvlist_dst config nvlist
1648 * zc_nvlist_dst_size size of config nvlist
1651 zfs_ioc_pool_stats(zfs_cmd_t *zc)
1657 error = spa_get_stats(zc->zc_name, &config, zc->zc_value,
1658 sizeof (zc->zc_value));
1660 if (config != NULL) {
1661 ret = put_nvlist(zc, config);
1662 nvlist_free(config);
1665 * The config may be present even if 'error' is non-zero.
1666 * In this case we return success, and preserve the real errno
1669 zc->zc_cookie = error;
1678 * Try to import the given pool, returning pool stats as appropriate so that
1679 * user land knows which devices are available and overall pool health.
1682 zfs_ioc_pool_tryimport(zfs_cmd_t *zc)
1684 nvlist_t *tryconfig, *config;
1687 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1688 zc->zc_iflags, &tryconfig)) != 0)
1691 config = spa_tryimport(tryconfig);
1693 nvlist_free(tryconfig);
1696 return (SET_ERROR(EINVAL));
1698 error = put_nvlist(zc, config);
1699 nvlist_free(config);
1706 * zc_name name of the pool
1707 * zc_cookie scan func (pool_scan_func_t)
1710 zfs_ioc_pool_scan(zfs_cmd_t *zc)
1715 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1718 if (zc->zc_cookie == POOL_SCAN_NONE)
1719 error = spa_scan_stop(spa);
1721 error = spa_scan(spa, zc->zc_cookie);
1723 spa_close(spa, FTAG);
1729 zfs_ioc_pool_freeze(zfs_cmd_t *zc)
1734 error = spa_open(zc->zc_name, &spa, FTAG);
1737 spa_close(spa, FTAG);
1743 zfs_ioc_pool_upgrade(zfs_cmd_t *zc)
1748 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1751 if (zc->zc_cookie < spa_version(spa) ||
1752 !SPA_VERSION_IS_SUPPORTED(zc->zc_cookie)) {
1753 spa_close(spa, FTAG);
1754 return (SET_ERROR(EINVAL));
1757 spa_upgrade(spa, zc->zc_cookie);
1758 spa_close(spa, FTAG);
1764 zfs_ioc_pool_get_history(zfs_cmd_t *zc)
1771 if ((size = zc->zc_history_len) == 0)
1772 return (SET_ERROR(EINVAL));
1774 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1777 if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) {
1778 spa_close(spa, FTAG);
1779 return (SET_ERROR(ENOTSUP));
1782 hist_buf = kmem_alloc(size, KM_SLEEP);
1783 if ((error = spa_history_get(spa, &zc->zc_history_offset,
1784 &zc->zc_history_len, hist_buf)) == 0) {
1785 error = ddi_copyout(hist_buf,
1786 (void *)(uintptr_t)zc->zc_history,
1787 zc->zc_history_len, zc->zc_iflags);
1790 spa_close(spa, FTAG);
1791 kmem_free(hist_buf, size);
1796 zfs_ioc_pool_reguid(zfs_cmd_t *zc)
1801 error = spa_open(zc->zc_name, &spa, FTAG);
1803 error = spa_change_guid(spa);
1804 spa_close(spa, FTAG);
1810 zfs_ioc_dsobj_to_dsname(zfs_cmd_t *zc)
1812 return (dsl_dsobj_to_dsname(zc->zc_name, zc->zc_obj, zc->zc_value));
1817 * zc_name name of filesystem
1818 * zc_obj object to find
1821 * zc_value name of object
1824 zfs_ioc_obj_to_path(zfs_cmd_t *zc)
1829 /* XXX reading from objset not owned */
1830 if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os)) != 0)
1832 if (dmu_objset_type(os) != DMU_OST_ZFS) {
1833 dmu_objset_rele(os, FTAG);
1834 return (SET_ERROR(EINVAL));
1836 error = zfs_obj_to_path(os, zc->zc_obj, zc->zc_value,
1837 sizeof (zc->zc_value));
1838 dmu_objset_rele(os, FTAG);
1845 * zc_name name of filesystem
1846 * zc_obj object to find
1849 * zc_stat stats on object
1850 * zc_value path to object
1853 zfs_ioc_obj_to_stats(zfs_cmd_t *zc)
1858 /* XXX reading from objset not owned */
1859 if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os)) != 0)
1861 if (dmu_objset_type(os) != DMU_OST_ZFS) {
1862 dmu_objset_rele(os, FTAG);
1863 return (SET_ERROR(EINVAL));
1865 error = zfs_obj_to_stats(os, zc->zc_obj, &zc->zc_stat, zc->zc_value,
1866 sizeof (zc->zc_value));
1867 dmu_objset_rele(os, FTAG);
1873 zfs_ioc_vdev_add(zfs_cmd_t *zc)
1877 nvlist_t *config, **l2cache, **spares;
1878 uint_t nl2cache = 0, nspares = 0;
1880 error = spa_open(zc->zc_name, &spa, FTAG);
1884 error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1885 zc->zc_iflags, &config);
1886 (void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_L2CACHE,
1887 &l2cache, &nl2cache);
1889 (void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_SPARES,
1894 * A root pool with concatenated devices is not supported.
1895 * Thus, can not add a device to a root pool.
1897 * Intent log device can not be added to a rootpool because
1898 * during mountroot, zil is replayed, a seperated log device
1899 * can not be accessed during the mountroot time.
1901 * l2cache and spare devices are ok to be added to a rootpool.
1903 if (spa_bootfs(spa) != 0 && nl2cache == 0 && nspares == 0) {
1904 nvlist_free(config);
1905 spa_close(spa, FTAG);
1906 return (SET_ERROR(EDOM));
1908 #endif /* illumos */
1911 error = spa_vdev_add(spa, config);
1912 nvlist_free(config);
1914 spa_close(spa, FTAG);
1920 * zc_name name of the pool
1921 * zc_nvlist_conf nvlist of devices to remove
1922 * zc_cookie to stop the remove?
1925 zfs_ioc_vdev_remove(zfs_cmd_t *zc)
1930 error = spa_open(zc->zc_name, &spa, FTAG);
1933 error = spa_vdev_remove(spa, zc->zc_guid, B_FALSE);
1934 spa_close(spa, FTAG);
1939 zfs_ioc_vdev_set_state(zfs_cmd_t *zc)
1943 vdev_state_t newstate = VDEV_STATE_UNKNOWN;
1945 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1947 switch (zc->zc_cookie) {
1948 case VDEV_STATE_ONLINE:
1949 error = vdev_online(spa, zc->zc_guid, zc->zc_obj, &newstate);
1952 case VDEV_STATE_OFFLINE:
1953 error = vdev_offline(spa, zc->zc_guid, zc->zc_obj);
1956 case VDEV_STATE_FAULTED:
1957 if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED &&
1958 zc->zc_obj != VDEV_AUX_EXTERNAL)
1959 zc->zc_obj = VDEV_AUX_ERR_EXCEEDED;
1961 error = vdev_fault(spa, zc->zc_guid, zc->zc_obj);
1964 case VDEV_STATE_DEGRADED:
1965 if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED &&
1966 zc->zc_obj != VDEV_AUX_EXTERNAL)
1967 zc->zc_obj = VDEV_AUX_ERR_EXCEEDED;
1969 error = vdev_degrade(spa, zc->zc_guid, zc->zc_obj);
1973 error = SET_ERROR(EINVAL);
1975 zc->zc_cookie = newstate;
1976 spa_close(spa, FTAG);
1981 zfs_ioc_vdev_attach(zfs_cmd_t *zc)
1984 int replacing = zc->zc_cookie;
1988 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1991 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1992 zc->zc_iflags, &config)) == 0) {
1993 error = spa_vdev_attach(spa, zc->zc_guid, config, replacing);
1994 nvlist_free(config);
1997 spa_close(spa, FTAG);
2002 zfs_ioc_vdev_detach(zfs_cmd_t *zc)
2007 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
2010 error = spa_vdev_detach(spa, zc->zc_guid, 0, B_FALSE);
2012 spa_close(spa, FTAG);
2017 zfs_ioc_vdev_split(zfs_cmd_t *zc)
2020 nvlist_t *config, *props = NULL;
2022 boolean_t exp = !!(zc->zc_cookie & ZPOOL_EXPORT_AFTER_SPLIT);
2024 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
2027 if (error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
2028 zc->zc_iflags, &config)) {
2029 spa_close(spa, FTAG);
2033 if (zc->zc_nvlist_src_size != 0 && (error =
2034 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2035 zc->zc_iflags, &props))) {
2036 spa_close(spa, FTAG);
2037 nvlist_free(config);
2041 error = spa_vdev_split_mirror(spa, zc->zc_string, config, props, exp);
2043 spa_close(spa, FTAG);
2045 nvlist_free(config);
2052 zfs_ioc_vdev_setpath(zfs_cmd_t *zc)
2055 char *path = zc->zc_value;
2056 uint64_t guid = zc->zc_guid;
2059 error = spa_open(zc->zc_name, &spa, FTAG);
2063 error = spa_vdev_setpath(spa, guid, path);
2064 spa_close(spa, FTAG);
2069 zfs_ioc_vdev_setfru(zfs_cmd_t *zc)
2072 char *fru = zc->zc_value;
2073 uint64_t guid = zc->zc_guid;
2076 error = spa_open(zc->zc_name, &spa, FTAG);
2080 error = spa_vdev_setfru(spa, guid, fru);
2081 spa_close(spa, FTAG);
2086 zfs_ioc_objset_stats_impl(zfs_cmd_t *zc, objset_t *os)
2091 dmu_objset_fast_stat(os, &zc->zc_objset_stats);
2093 if (zc->zc_nvlist_dst != 0 &&
2094 (error = dsl_prop_get_all(os, &nv)) == 0) {
2095 dmu_objset_stats(os, nv);
2097 * NB: zvol_get_stats() will read the objset contents,
2098 * which we aren't supposed to do with a
2099 * DS_MODE_USER hold, because it could be
2100 * inconsistent. So this is a bit of a workaround...
2101 * XXX reading with out owning
2103 if (!zc->zc_objset_stats.dds_inconsistent &&
2104 dmu_objset_type(os) == DMU_OST_ZVOL) {
2105 error = zvol_get_stats(os, nv);
2110 error = put_nvlist(zc, nv);
2119 * zc_name name of filesystem
2120 * zc_nvlist_dst_size size of buffer for property nvlist
2123 * zc_objset_stats stats
2124 * zc_nvlist_dst property nvlist
2125 * zc_nvlist_dst_size size of property nvlist
2128 zfs_ioc_objset_stats(zfs_cmd_t *zc)
2133 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
2135 error = zfs_ioc_objset_stats_impl(zc, os);
2136 dmu_objset_rele(os, FTAG);
2139 if (error == ENOMEM)
2146 * zc_name name of filesystem
2147 * zc_nvlist_dst_size size of buffer for property nvlist
2150 * zc_nvlist_dst received property nvlist
2151 * zc_nvlist_dst_size size of received property nvlist
2153 * Gets received properties (distinct from local properties on or after
2154 * SPA_VERSION_RECVD_PROPS) for callers who want to differentiate received from
2155 * local property values.
2158 zfs_ioc_objset_recvd_props(zfs_cmd_t *zc)
2164 * Without this check, we would return local property values if the
2165 * caller has not already received properties on or after
2166 * SPA_VERSION_RECVD_PROPS.
2168 if (!dsl_prop_get_hasrecvd(zc->zc_name))
2169 return (SET_ERROR(ENOTSUP));
2171 if (zc->zc_nvlist_dst != 0 &&
2172 (error = dsl_prop_get_received(zc->zc_name, &nv)) == 0) {
2173 error = put_nvlist(zc, nv);
2181 nvl_add_zplprop(objset_t *os, nvlist_t *props, zfs_prop_t prop)
2187 * zfs_get_zplprop() will either find a value or give us
2188 * the default value (if there is one).
2190 if ((error = zfs_get_zplprop(os, prop, &value)) != 0)
2192 VERIFY(nvlist_add_uint64(props, zfs_prop_to_name(prop), value) == 0);
2198 * zc_name name of filesystem
2199 * zc_nvlist_dst_size size of buffer for zpl property nvlist
2202 * zc_nvlist_dst zpl property nvlist
2203 * zc_nvlist_dst_size size of zpl property nvlist
2206 zfs_ioc_objset_zplprops(zfs_cmd_t *zc)
2211 /* XXX reading without owning */
2212 if (err = dmu_objset_hold(zc->zc_name, FTAG, &os))
2215 dmu_objset_fast_stat(os, &zc->zc_objset_stats);
2218 * NB: nvl_add_zplprop() will read the objset contents,
2219 * which we aren't supposed to do with a DS_MODE_USER
2220 * hold, because it could be inconsistent.
2222 if (zc->zc_nvlist_dst != 0 &&
2223 !zc->zc_objset_stats.dds_inconsistent &&
2224 dmu_objset_type(os) == DMU_OST_ZFS) {
2227 VERIFY(nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2228 if ((err = nvl_add_zplprop(os, nv, ZFS_PROP_VERSION)) == 0 &&
2229 (err = nvl_add_zplprop(os, nv, ZFS_PROP_NORMALIZE)) == 0 &&
2230 (err = nvl_add_zplprop(os, nv, ZFS_PROP_UTF8ONLY)) == 0 &&
2231 (err = nvl_add_zplprop(os, nv, ZFS_PROP_CASE)) == 0)
2232 err = put_nvlist(zc, nv);
2235 err = SET_ERROR(ENOENT);
2237 dmu_objset_rele(os, FTAG);
2242 dataset_name_hidden(const char *name)
2245 * Skip over datasets that are not visible in this zone,
2246 * internal datasets (which have a $ in their name), and
2247 * temporary datasets (which have a % in their name).
2249 if (strchr(name, '$') != NULL)
2251 if (strchr(name, '%') != NULL)
2253 if (!INGLOBALZONE(curthread) && !zone_dataset_visible(name, NULL))
2260 * zc_name name of filesystem
2261 * zc_cookie zap cursor
2262 * zc_nvlist_dst_size size of buffer for property nvlist
2265 * zc_name name of next filesystem
2266 * zc_cookie zap cursor
2267 * zc_objset_stats stats
2268 * zc_nvlist_dst property nvlist
2269 * zc_nvlist_dst_size size of property nvlist
2272 zfs_ioc_dataset_list_next(zfs_cmd_t *zc)
2277 size_t orig_len = strlen(zc->zc_name);
2280 if (error = dmu_objset_hold(zc->zc_name, FTAG, &os)) {
2281 if (error == ENOENT)
2282 error = SET_ERROR(ESRCH);
2286 p = strrchr(zc->zc_name, '/');
2287 if (p == NULL || p[1] != '\0')
2288 (void) strlcat(zc->zc_name, "/", sizeof (zc->zc_name));
2289 p = zc->zc_name + strlen(zc->zc_name);
2292 error = dmu_dir_list_next(os,
2293 sizeof (zc->zc_name) - (p - zc->zc_name), p,
2294 NULL, &zc->zc_cookie);
2295 if (error == ENOENT)
2296 error = SET_ERROR(ESRCH);
2297 } while (error == 0 && dataset_name_hidden(zc->zc_name));
2298 dmu_objset_rele(os, FTAG);
2301 * If it's an internal dataset (ie. with a '$' in its name),
2302 * don't try to get stats for it, otherwise we'll return ENOENT.
2304 if (error == 0 && strchr(zc->zc_name, '$') == NULL) {
2305 error = zfs_ioc_objset_stats(zc); /* fill in the stats */
2306 if (error == ENOENT) {
2307 /* We lost a race with destroy, get the next one. */
2308 zc->zc_name[orig_len] = '\0';
2317 * zc_name name of filesystem
2318 * zc_cookie zap cursor
2319 * zc_nvlist_dst_size size of buffer for property nvlist
2320 * zc_simple when set, only name is requested
2323 * zc_name name of next snapshot
2324 * zc_objset_stats stats
2325 * zc_nvlist_dst property nvlist
2326 * zc_nvlist_dst_size size of property nvlist
2329 zfs_ioc_snapshot_list_next(zfs_cmd_t *zc)
2334 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
2336 return (error == ENOENT ? ESRCH : error);
2340 * A dataset name of maximum length cannot have any snapshots,
2341 * so exit immediately.
2343 if (strlcat(zc->zc_name, "@", sizeof (zc->zc_name)) >= MAXNAMELEN) {
2344 dmu_objset_rele(os, FTAG);
2345 return (SET_ERROR(ESRCH));
2348 error = dmu_snapshot_list_next(os,
2349 sizeof (zc->zc_name) - strlen(zc->zc_name),
2350 zc->zc_name + strlen(zc->zc_name), &zc->zc_obj, &zc->zc_cookie,
2353 if (error == 0 && !zc->zc_simple) {
2355 dsl_pool_t *dp = os->os_dsl_dataset->ds_dir->dd_pool;
2357 error = dsl_dataset_hold_obj(dp, zc->zc_obj, FTAG, &ds);
2361 error = dmu_objset_from_ds(ds, &ossnap);
2363 error = zfs_ioc_objset_stats_impl(zc, ossnap);
2364 dsl_dataset_rele(ds, FTAG);
2366 } else if (error == ENOENT) {
2367 error = SET_ERROR(ESRCH);
2370 dmu_objset_rele(os, FTAG);
2371 /* if we failed, undo the @ that we tacked on to zc_name */
2373 *strchr(zc->zc_name, '@') = '\0';
2378 zfs_prop_set_userquota(const char *dsname, nvpair_t *pair)
2380 const char *propname = nvpair_name(pair);
2382 unsigned int vallen;
2385 zfs_userquota_prop_t type;
2391 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2393 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
2394 if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2396 return (SET_ERROR(EINVAL));
2400 * A correctly constructed propname is encoded as
2401 * userquota@<rid>-<domain>.
2403 if ((dash = strchr(propname, '-')) == NULL ||
2404 nvpair_value_uint64_array(pair, &valary, &vallen) != 0 ||
2406 return (SET_ERROR(EINVAL));
2413 err = zfsvfs_hold(dsname, FTAG, &zfsvfs, B_FALSE);
2415 err = zfs_set_userquota(zfsvfs, type, domain, rid, quota);
2416 zfsvfs_rele(zfsvfs, FTAG);
2423 * If the named property is one that has a special function to set its value,
2424 * return 0 on success and a positive error code on failure; otherwise if it is
2425 * not one of the special properties handled by this function, return -1.
2427 * XXX: It would be better for callers of the property interface if we handled
2428 * these special cases in dsl_prop.c (in the dsl layer).
2431 zfs_prop_set_special(const char *dsname, zprop_source_t source,
2434 const char *propname = nvpair_name(pair);
2435 zfs_prop_t prop = zfs_name_to_prop(propname);
2439 if (prop == ZPROP_INVAL) {
2440 if (zfs_prop_userquota(propname))
2441 return (zfs_prop_set_userquota(dsname, pair));
2445 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2447 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
2448 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2452 if (zfs_prop_get_type(prop) == PROP_TYPE_STRING)
2455 VERIFY(0 == nvpair_value_uint64(pair, &intval));
2458 case ZFS_PROP_QUOTA:
2459 err = dsl_dir_set_quota(dsname, source, intval);
2461 case ZFS_PROP_REFQUOTA:
2462 err = dsl_dataset_set_refquota(dsname, source, intval);
2464 case ZFS_PROP_FILESYSTEM_LIMIT:
2465 case ZFS_PROP_SNAPSHOT_LIMIT:
2466 if (intval == UINT64_MAX) {
2467 /* clearing the limit, just do it */
2470 err = dsl_dir_activate_fs_ss_limit(dsname);
2473 * Set err to -1 to force the zfs_set_prop_nvlist code down the
2474 * default path to set the value in the nvlist.
2479 case ZFS_PROP_RESERVATION:
2480 err = dsl_dir_set_reservation(dsname, source, intval);
2482 case ZFS_PROP_REFRESERVATION:
2483 err = dsl_dataset_set_refreservation(dsname, source, intval);
2485 case ZFS_PROP_VOLSIZE:
2486 err = zvol_set_volsize(dsname, ddi_driver_major(zfs_dip),
2489 case ZFS_PROP_VERSION:
2493 if ((err = zfsvfs_hold(dsname, FTAG, &zfsvfs, B_TRUE)) != 0)
2496 err = zfs_set_version(zfsvfs, intval);
2497 zfsvfs_rele(zfsvfs, FTAG);
2499 if (err == 0 && intval >= ZPL_VERSION_USERSPACE) {
2502 zc = kmem_zalloc(sizeof (zfs_cmd_t), KM_SLEEP);
2503 (void) strcpy(zc->zc_name, dsname);
2504 (void) zfs_ioc_userspace_upgrade(zc);
2505 kmem_free(zc, sizeof (zfs_cmd_t));
2517 * This function is best effort. If it fails to set any of the given properties,
2518 * it continues to set as many as it can and returns the last error
2519 * encountered. If the caller provides a non-NULL errlist, it will be filled in
2520 * with the list of names of all the properties that failed along with the
2521 * corresponding error numbers.
2523 * If every property is set successfully, zero is returned and errlist is not
2527 zfs_set_prop_nvlist(const char *dsname, zprop_source_t source, nvlist_t *nvl,
2535 nvlist_t *genericnvl = fnvlist_alloc();
2536 nvlist_t *retrynvl = fnvlist_alloc();
2540 while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) {
2541 const char *propname = nvpair_name(pair);
2542 zfs_prop_t prop = zfs_name_to_prop(propname);
2545 /* decode the property value */
2547 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2549 attrs = fnvpair_value_nvlist(pair);
2550 if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2552 err = SET_ERROR(EINVAL);
2555 /* Validate value type */
2556 if (err == 0 && prop == ZPROP_INVAL) {
2557 if (zfs_prop_user(propname)) {
2558 if (nvpair_type(propval) != DATA_TYPE_STRING)
2559 err = SET_ERROR(EINVAL);
2560 } else if (zfs_prop_userquota(propname)) {
2561 if (nvpair_type(propval) !=
2562 DATA_TYPE_UINT64_ARRAY)
2563 err = SET_ERROR(EINVAL);
2565 err = SET_ERROR(EINVAL);
2567 } else if (err == 0) {
2568 if (nvpair_type(propval) == DATA_TYPE_STRING) {
2569 if (zfs_prop_get_type(prop) != PROP_TYPE_STRING)
2570 err = SET_ERROR(EINVAL);
2571 } else if (nvpair_type(propval) == DATA_TYPE_UINT64) {
2574 intval = fnvpair_value_uint64(propval);
2576 switch (zfs_prop_get_type(prop)) {
2577 case PROP_TYPE_NUMBER:
2579 case PROP_TYPE_STRING:
2580 err = SET_ERROR(EINVAL);
2582 case PROP_TYPE_INDEX:
2583 if (zfs_prop_index_to_string(prop,
2584 intval, &unused) != 0)
2585 err = SET_ERROR(EINVAL);
2589 "unknown property type");
2592 err = SET_ERROR(EINVAL);
2596 /* Validate permissions */
2598 err = zfs_check_settable(dsname, pair, CRED());
2601 err = zfs_prop_set_special(dsname, source, pair);
2604 * For better performance we build up a list of
2605 * properties to set in a single transaction.
2607 err = nvlist_add_nvpair(genericnvl, pair);
2608 } else if (err != 0 && nvl != retrynvl) {
2610 * This may be a spurious error caused by
2611 * receiving quota and reservation out of order.
2612 * Try again in a second pass.
2614 err = nvlist_add_nvpair(retrynvl, pair);
2619 if (errlist != NULL)
2620 fnvlist_add_int32(errlist, propname, err);
2625 if (nvl != retrynvl && !nvlist_empty(retrynvl)) {
2630 if (!nvlist_empty(genericnvl) &&
2631 dsl_props_set(dsname, source, genericnvl) != 0) {
2633 * If this fails, we still want to set as many properties as we
2634 * can, so try setting them individually.
2637 while ((pair = nvlist_next_nvpair(genericnvl, pair)) != NULL) {
2638 const char *propname = nvpair_name(pair);
2642 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2644 attrs = fnvpair_value_nvlist(pair);
2645 propval = fnvlist_lookup_nvpair(attrs,
2649 if (nvpair_type(propval) == DATA_TYPE_STRING) {
2650 strval = fnvpair_value_string(propval);
2651 err = dsl_prop_set_string(dsname, propname,
2654 intval = fnvpair_value_uint64(propval);
2655 err = dsl_prop_set_int(dsname, propname, source,
2660 if (errlist != NULL) {
2661 fnvlist_add_int32(errlist, propname,
2668 nvlist_free(genericnvl);
2669 nvlist_free(retrynvl);
2675 * Check that all the properties are valid user properties.
2678 zfs_check_userprops(const char *fsname, nvlist_t *nvl)
2680 nvpair_t *pair = NULL;
2683 while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) {
2684 const char *propname = nvpair_name(pair);
2686 if (!zfs_prop_user(propname) ||
2687 nvpair_type(pair) != DATA_TYPE_STRING)
2688 return (SET_ERROR(EINVAL));
2690 if (error = zfs_secpolicy_write_perms(fsname,
2691 ZFS_DELEG_PERM_USERPROP, CRED()))
2694 if (strlen(propname) >= ZAP_MAXNAMELEN)
2695 return (SET_ERROR(ENAMETOOLONG));
2697 if (strlen(fnvpair_value_string(pair)) >= ZAP_MAXVALUELEN)
2704 props_skip(nvlist_t *props, nvlist_t *skipped, nvlist_t **newprops)
2708 VERIFY(nvlist_alloc(newprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2711 while ((pair = nvlist_next_nvpair(props, pair)) != NULL) {
2712 if (nvlist_exists(skipped, nvpair_name(pair)))
2715 VERIFY(nvlist_add_nvpair(*newprops, pair) == 0);
2720 clear_received_props(const char *dsname, nvlist_t *props,
2724 nvlist_t *cleared_props = NULL;
2725 props_skip(props, skipped, &cleared_props);
2726 if (!nvlist_empty(cleared_props)) {
2728 * Acts on local properties until the dataset has received
2729 * properties at least once on or after SPA_VERSION_RECVD_PROPS.
2731 zprop_source_t flags = (ZPROP_SRC_NONE |
2732 (dsl_prop_get_hasrecvd(dsname) ? ZPROP_SRC_RECEIVED : 0));
2733 err = zfs_set_prop_nvlist(dsname, flags, cleared_props, NULL);
2735 nvlist_free(cleared_props);
2741 * zc_name name of filesystem
2742 * zc_value name of property to set
2743 * zc_nvlist_src{_size} nvlist of properties to apply
2744 * zc_cookie received properties flag
2747 * zc_nvlist_dst{_size} error for each unapplied received property
2750 zfs_ioc_set_prop(zfs_cmd_t *zc)
2753 boolean_t received = zc->zc_cookie;
2754 zprop_source_t source = (received ? ZPROP_SRC_RECEIVED :
2759 if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2760 zc->zc_iflags, &nvl)) != 0)
2764 nvlist_t *origprops;
2766 if (dsl_prop_get_received(zc->zc_name, &origprops) == 0) {
2767 (void) clear_received_props(zc->zc_name,
2769 nvlist_free(origprops);
2772 error = dsl_prop_set_hasrecvd(zc->zc_name);
2775 errors = fnvlist_alloc();
2777 error = zfs_set_prop_nvlist(zc->zc_name, source, nvl, errors);
2779 if (zc->zc_nvlist_dst != 0 && errors != NULL) {
2780 (void) put_nvlist(zc, errors);
2783 nvlist_free(errors);
2790 * zc_name name of filesystem
2791 * zc_value name of property to inherit
2792 * zc_cookie revert to received value if TRUE
2797 zfs_ioc_inherit_prop(zfs_cmd_t *zc)
2799 const char *propname = zc->zc_value;
2800 zfs_prop_t prop = zfs_name_to_prop(propname);
2801 boolean_t received = zc->zc_cookie;
2802 zprop_source_t source = (received
2803 ? ZPROP_SRC_NONE /* revert to received value, if any */
2804 : ZPROP_SRC_INHERITED); /* explicitly inherit */
2813 * zfs_prop_set_special() expects properties in the form of an
2814 * nvpair with type info.
2816 if (prop == ZPROP_INVAL) {
2817 if (!zfs_prop_user(propname))
2818 return (SET_ERROR(EINVAL));
2820 type = PROP_TYPE_STRING;
2821 } else if (prop == ZFS_PROP_VOLSIZE ||
2822 prop == ZFS_PROP_VERSION) {
2823 return (SET_ERROR(EINVAL));
2825 type = zfs_prop_get_type(prop);
2828 VERIFY(nvlist_alloc(&dummy, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2831 case PROP_TYPE_STRING:
2832 VERIFY(0 == nvlist_add_string(dummy, propname, ""));
2834 case PROP_TYPE_NUMBER:
2835 case PROP_TYPE_INDEX:
2836 VERIFY(0 == nvlist_add_uint64(dummy, propname, 0));
2840 return (SET_ERROR(EINVAL));
2843 pair = nvlist_next_nvpair(dummy, NULL);
2844 err = zfs_prop_set_special(zc->zc_name, source, pair);
2847 return (err); /* special property already handled */
2850 * Only check this in the non-received case. We want to allow
2851 * 'inherit -S' to revert non-inheritable properties like quota
2852 * and reservation to the received or default values even though
2853 * they are not considered inheritable.
2855 if (prop != ZPROP_INVAL && !zfs_prop_inheritable(prop))
2856 return (SET_ERROR(EINVAL));
2859 /* property name has been validated by zfs_secpolicy_inherit_prop() */
2860 return (dsl_prop_inherit(zc->zc_name, zc->zc_value, source));
2864 zfs_ioc_pool_set_props(zfs_cmd_t *zc)
2871 if (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2872 zc->zc_iflags, &props))
2876 * If the only property is the configfile, then just do a spa_lookup()
2877 * to handle the faulted case.
2879 pair = nvlist_next_nvpair(props, NULL);
2880 if (pair != NULL && strcmp(nvpair_name(pair),
2881 zpool_prop_to_name(ZPOOL_PROP_CACHEFILE)) == 0 &&
2882 nvlist_next_nvpair(props, pair) == NULL) {
2883 mutex_enter(&spa_namespace_lock);
2884 if ((spa = spa_lookup(zc->zc_name)) != NULL) {
2885 spa_configfile_set(spa, props, B_FALSE);
2886 spa_config_sync(spa, B_FALSE, B_TRUE);
2888 mutex_exit(&spa_namespace_lock);
2895 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) {
2900 error = spa_prop_set(spa, props);
2903 spa_close(spa, FTAG);
2909 zfs_ioc_pool_get_props(zfs_cmd_t *zc)
2913 nvlist_t *nvp = NULL;
2915 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) {
2917 * If the pool is faulted, there may be properties we can still
2918 * get (such as altroot and cachefile), so attempt to get them
2921 mutex_enter(&spa_namespace_lock);
2922 if ((spa = spa_lookup(zc->zc_name)) != NULL)
2923 error = spa_prop_get(spa, &nvp);
2924 mutex_exit(&spa_namespace_lock);
2926 error = spa_prop_get(spa, &nvp);
2927 spa_close(spa, FTAG);
2930 if (error == 0 && zc->zc_nvlist_dst != 0)
2931 error = put_nvlist(zc, nvp);
2933 error = SET_ERROR(EFAULT);
2941 * zc_name name of filesystem
2942 * zc_nvlist_src{_size} nvlist of delegated permissions
2943 * zc_perm_action allow/unallow flag
2948 zfs_ioc_set_fsacl(zfs_cmd_t *zc)
2951 nvlist_t *fsaclnv = NULL;
2953 if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2954 zc->zc_iflags, &fsaclnv)) != 0)
2958 * Verify nvlist is constructed correctly
2960 if ((error = zfs_deleg_verify_nvlist(fsaclnv)) != 0) {
2961 nvlist_free(fsaclnv);
2962 return (SET_ERROR(EINVAL));
2966 * If we don't have PRIV_SYS_MOUNT, then validate
2967 * that user is allowed to hand out each permission in
2971 error = secpolicy_zfs(CRED());
2973 if (zc->zc_perm_action == B_FALSE) {
2974 error = dsl_deleg_can_allow(zc->zc_name,
2977 error = dsl_deleg_can_unallow(zc->zc_name,
2983 error = dsl_deleg_set(zc->zc_name, fsaclnv, zc->zc_perm_action);
2985 nvlist_free(fsaclnv);
2991 * zc_name name of filesystem
2994 * zc_nvlist_src{_size} nvlist of delegated permissions
2997 zfs_ioc_get_fsacl(zfs_cmd_t *zc)
3002 if ((error = dsl_deleg_get(zc->zc_name, &nvp)) == 0) {
3003 error = put_nvlist(zc, nvp);
3011 * Search the vfs list for a specified resource. Returns a pointer to it
3012 * or NULL if no suitable entry is found. The caller of this routine
3013 * is responsible for releasing the returned vfs pointer.
3016 zfs_get_vfs(const char *resource)
3020 mtx_lock(&mountlist_mtx);
3021 TAILQ_FOREACH(vfsp, &mountlist, mnt_list) {
3022 if (strcmp(refstr_value(vfsp->vfs_resource), resource) == 0) {
3027 mtx_unlock(&mountlist_mtx);
3033 zfs_create_cb(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx)
3035 zfs_creat_t *zct = arg;
3037 zfs_create_fs(os, cr, zct->zct_zplprops, tx);
3040 #define ZFS_PROP_UNDEFINED ((uint64_t)-1)
3044 * os parent objset pointer (NULL if root fs)
3045 * fuids_ok fuids allowed in this version of the spa?
3046 * sa_ok SAs allowed in this version of the spa?
3047 * createprops list of properties requested by creator
3050 * zplprops values for the zplprops we attach to the master node object
3051 * is_ci true if requested file system will be purely case-insensitive
3053 * Determine the settings for utf8only, normalization and
3054 * casesensitivity. Specific values may have been requested by the
3055 * creator and/or we can inherit values from the parent dataset. If
3056 * the file system is of too early a vintage, a creator can not
3057 * request settings for these properties, even if the requested
3058 * setting is the default value. We don't actually want to create dsl
3059 * properties for these, so remove them from the source nvlist after
3063 zfs_fill_zplprops_impl(objset_t *os, uint64_t zplver,
3064 boolean_t fuids_ok, boolean_t sa_ok, nvlist_t *createprops,
3065 nvlist_t *zplprops, boolean_t *is_ci)
3067 uint64_t sense = ZFS_PROP_UNDEFINED;
3068 uint64_t norm = ZFS_PROP_UNDEFINED;
3069 uint64_t u8 = ZFS_PROP_UNDEFINED;
3071 ASSERT(zplprops != NULL);
3074 * Pull out creator prop choices, if any.
3077 (void) nvlist_lookup_uint64(createprops,
3078 zfs_prop_to_name(ZFS_PROP_VERSION), &zplver);
3079 (void) nvlist_lookup_uint64(createprops,
3080 zfs_prop_to_name(ZFS_PROP_NORMALIZE), &norm);
3081 (void) nvlist_remove_all(createprops,
3082 zfs_prop_to_name(ZFS_PROP_NORMALIZE));
3083 (void) nvlist_lookup_uint64(createprops,
3084 zfs_prop_to_name(ZFS_PROP_UTF8ONLY), &u8);
3085 (void) nvlist_remove_all(createprops,
3086 zfs_prop_to_name(ZFS_PROP_UTF8ONLY));
3087 (void) nvlist_lookup_uint64(createprops,
3088 zfs_prop_to_name(ZFS_PROP_CASE), &sense);
3089 (void) nvlist_remove_all(createprops,
3090 zfs_prop_to_name(ZFS_PROP_CASE));
3094 * If the zpl version requested is whacky or the file system
3095 * or pool is version is too "young" to support normalization
3096 * and the creator tried to set a value for one of the props,
3099 if ((zplver < ZPL_VERSION_INITIAL || zplver > ZPL_VERSION) ||
3100 (zplver >= ZPL_VERSION_FUID && !fuids_ok) ||
3101 (zplver >= ZPL_VERSION_SA && !sa_ok) ||
3102 (zplver < ZPL_VERSION_NORMALIZATION &&
3103 (norm != ZFS_PROP_UNDEFINED || u8 != ZFS_PROP_UNDEFINED ||
3104 sense != ZFS_PROP_UNDEFINED)))
3105 return (SET_ERROR(ENOTSUP));
3108 * Put the version in the zplprops
3110 VERIFY(nvlist_add_uint64(zplprops,
3111 zfs_prop_to_name(ZFS_PROP_VERSION), zplver) == 0);
3113 if (norm == ZFS_PROP_UNDEFINED)
3114 VERIFY(zfs_get_zplprop(os, ZFS_PROP_NORMALIZE, &norm) == 0);
3115 VERIFY(nvlist_add_uint64(zplprops,
3116 zfs_prop_to_name(ZFS_PROP_NORMALIZE), norm) == 0);
3119 * If we're normalizing, names must always be valid UTF-8 strings.
3123 if (u8 == ZFS_PROP_UNDEFINED)
3124 VERIFY(zfs_get_zplprop(os, ZFS_PROP_UTF8ONLY, &u8) == 0);
3125 VERIFY(nvlist_add_uint64(zplprops,
3126 zfs_prop_to_name(ZFS_PROP_UTF8ONLY), u8) == 0);
3128 if (sense == ZFS_PROP_UNDEFINED)
3129 VERIFY(zfs_get_zplprop(os, ZFS_PROP_CASE, &sense) == 0);
3130 VERIFY(nvlist_add_uint64(zplprops,
3131 zfs_prop_to_name(ZFS_PROP_CASE), sense) == 0);
3134 *is_ci = (sense == ZFS_CASE_INSENSITIVE);
3140 zfs_fill_zplprops(const char *dataset, nvlist_t *createprops,
3141 nvlist_t *zplprops, boolean_t *is_ci)
3143 boolean_t fuids_ok, sa_ok;
3144 uint64_t zplver = ZPL_VERSION;
3145 objset_t *os = NULL;
3146 char parentname[MAXNAMELEN];
3152 (void) strlcpy(parentname, dataset, sizeof (parentname));
3153 cp = strrchr(parentname, '/');
3157 if ((error = spa_open(dataset, &spa, FTAG)) != 0)
3160 spa_vers = spa_version(spa);
3161 spa_close(spa, FTAG);
3163 zplver = zfs_zpl_version_map(spa_vers);
3164 fuids_ok = (zplver >= ZPL_VERSION_FUID);
3165 sa_ok = (zplver >= ZPL_VERSION_SA);
3168 * Open parent object set so we can inherit zplprop values.
3170 if ((error = dmu_objset_hold(parentname, FTAG, &os)) != 0)
3173 error = zfs_fill_zplprops_impl(os, zplver, fuids_ok, sa_ok, createprops,
3175 dmu_objset_rele(os, FTAG);
3180 zfs_fill_zplprops_root(uint64_t spa_vers, nvlist_t *createprops,
3181 nvlist_t *zplprops, boolean_t *is_ci)
3185 uint64_t zplver = ZPL_VERSION;
3188 zplver = zfs_zpl_version_map(spa_vers);
3189 fuids_ok = (zplver >= ZPL_VERSION_FUID);
3190 sa_ok = (zplver >= ZPL_VERSION_SA);
3192 error = zfs_fill_zplprops_impl(NULL, zplver, fuids_ok, sa_ok,
3193 createprops, zplprops, is_ci);
3199 * "type" -> dmu_objset_type_t (int32)
3200 * (optional) "props" -> { prop -> value }
3203 * outnvl: propname -> error code (int32)
3206 zfs_ioc_create(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3209 zfs_creat_t zct = { 0 };
3210 nvlist_t *nvprops = NULL;
3211 void (*cbfunc)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx);
3213 dmu_objset_type_t type;
3214 boolean_t is_insensitive = B_FALSE;
3216 if (nvlist_lookup_int32(innvl, "type", &type32) != 0)
3217 return (SET_ERROR(EINVAL));
3219 (void) nvlist_lookup_nvlist(innvl, "props", &nvprops);
3223 cbfunc = zfs_create_cb;
3227 cbfunc = zvol_create_cb;
3234 if (strchr(fsname, '@') ||
3235 strchr(fsname, '%'))
3236 return (SET_ERROR(EINVAL));
3238 zct.zct_props = nvprops;
3241 return (SET_ERROR(EINVAL));
3243 if (type == DMU_OST_ZVOL) {
3244 uint64_t volsize, volblocksize;
3246 if (nvprops == NULL)
3247 return (SET_ERROR(EINVAL));
3248 if (nvlist_lookup_uint64(nvprops,
3249 zfs_prop_to_name(ZFS_PROP_VOLSIZE), &volsize) != 0)
3250 return (SET_ERROR(EINVAL));
3252 if ((error = nvlist_lookup_uint64(nvprops,
3253 zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE),
3254 &volblocksize)) != 0 && error != ENOENT)
3255 return (SET_ERROR(EINVAL));
3258 volblocksize = zfs_prop_default_numeric(
3259 ZFS_PROP_VOLBLOCKSIZE);
3261 if ((error = zvol_check_volblocksize(
3262 volblocksize)) != 0 ||
3263 (error = zvol_check_volsize(volsize,
3264 volblocksize)) != 0)
3266 } else if (type == DMU_OST_ZFS) {
3270 * We have to have normalization and
3271 * case-folding flags correct when we do the
3272 * file system creation, so go figure them out
3275 VERIFY(nvlist_alloc(&zct.zct_zplprops,
3276 NV_UNIQUE_NAME, KM_SLEEP) == 0);
3277 error = zfs_fill_zplprops(fsname, nvprops,
3278 zct.zct_zplprops, &is_insensitive);
3280 nvlist_free(zct.zct_zplprops);
3285 error = dmu_objset_create(fsname, type,
3286 is_insensitive ? DS_FLAG_CI_DATASET : 0, cbfunc, &zct);
3287 nvlist_free(zct.zct_zplprops);
3290 * It would be nice to do this atomically.
3293 error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL,
3296 (void) dsl_destroy_head(fsname);
3299 if (error == 0 && type == DMU_OST_ZVOL)
3300 zvol_create_minors(fsname);
3307 * "origin" -> name of origin snapshot
3308 * (optional) "props" -> { prop -> value }
3311 * outnvl: propname -> error code (int32)
3314 zfs_ioc_clone(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3317 nvlist_t *nvprops = NULL;
3320 if (nvlist_lookup_string(innvl, "origin", &origin_name) != 0)
3321 return (SET_ERROR(EINVAL));
3322 (void) nvlist_lookup_nvlist(innvl, "props", &nvprops);
3324 if (strchr(fsname, '@') ||
3325 strchr(fsname, '%'))
3326 return (SET_ERROR(EINVAL));
3328 if (dataset_namecheck(origin_name, NULL, NULL) != 0)
3329 return (SET_ERROR(EINVAL));
3330 error = dmu_objset_clone(fsname, origin_name);
3335 * It would be nice to do this atomically.
3338 error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL,
3341 (void) dsl_destroy_head(fsname);
3345 zvol_create_minors(fsname);
3352 * "snaps" -> { snapshot1, snapshot2 }
3353 * (optional) "props" -> { prop -> value (string) }
3356 * outnvl: snapshot -> error code (int32)
3359 zfs_ioc_snapshot(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3362 nvlist_t *props = NULL;
3366 (void) nvlist_lookup_nvlist(innvl, "props", &props);
3367 if ((error = zfs_check_userprops(poolname, props)) != 0)
3370 if (!nvlist_empty(props) &&
3371 zfs_earlier_version(poolname, SPA_VERSION_SNAP_PROPS))
3372 return (SET_ERROR(ENOTSUP));
3374 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
3375 return (SET_ERROR(EINVAL));
3376 poollen = strlen(poolname);
3377 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
3378 pair = nvlist_next_nvpair(snaps, pair)) {
3379 const char *name = nvpair_name(pair);
3380 const char *cp = strchr(name, '@');
3383 * The snap name must contain an @, and the part after it must
3384 * contain only valid characters.
3387 zfs_component_namecheck(cp + 1, NULL, NULL) != 0)
3388 return (SET_ERROR(EINVAL));
3391 * The snap must be in the specified pool.
3393 if (strncmp(name, poolname, poollen) != 0 ||
3394 (name[poollen] != '/' && name[poollen] != '@'))
3395 return (SET_ERROR(EXDEV));
3397 /* This must be the only snap of this fs. */
3398 for (nvpair_t *pair2 = nvlist_next_nvpair(snaps, pair);
3399 pair2 != NULL; pair2 = nvlist_next_nvpair(snaps, pair2)) {
3400 if (strncmp(name, nvpair_name(pair2), cp - name + 1)
3402 return (SET_ERROR(EXDEV));
3407 error = dsl_dataset_snapshot(snaps, props, outnvl);
3412 * innvl: "message" -> string
3416 zfs_ioc_log_history(const char *unused, nvlist_t *innvl, nvlist_t *outnvl)
3424 * The poolname in the ioctl is not set, we get it from the TSD,
3425 * which was set at the end of the last successful ioctl that allows
3426 * logging. The secpolicy func already checked that it is set.
3427 * Only one log ioctl is allowed after each successful ioctl, so
3428 * we clear the TSD here.
3430 poolname = tsd_get(zfs_allow_log_key);
3431 (void) tsd_set(zfs_allow_log_key, NULL);
3432 error = spa_open(poolname, &spa, FTAG);
3437 if (nvlist_lookup_string(innvl, "message", &message) != 0) {
3438 spa_close(spa, FTAG);
3439 return (SET_ERROR(EINVAL));
3442 if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) {
3443 spa_close(spa, FTAG);
3444 return (SET_ERROR(ENOTSUP));
3447 error = spa_history_log(spa, message);
3448 spa_close(spa, FTAG);
3453 * The dp_config_rwlock must not be held when calling this, because the
3454 * unmount may need to write out data.
3456 * This function is best-effort. Callers must deal gracefully if it
3457 * remains mounted (or is remounted after this call).
3459 * Returns 0 if the argument is not a snapshot, or it is not currently a
3460 * filesystem, or we were able to unmount it. Returns error code otherwise.
3463 zfs_unmount_snap(const char *snapname)
3469 if (strchr(snapname, '@') == NULL)
3472 vfsp = zfs_get_vfs(snapname);
3476 zfsvfs = vfsp->vfs_data;
3477 ASSERT(!dsl_pool_config_held(dmu_objset_pool(zfsvfs->z_os)));
3479 err = vn_vfswlock(vfsp->vfs_vnodecovered);
3482 return (SET_ERROR(err));
3485 * Always force the unmount for snapshots.
3489 (void) dounmount(vfsp, MS_FORCE, kcred);
3491 mtx_lock(&Giant); /* dounmount() */
3492 (void) dounmount(vfsp, MS_FORCE, curthread);
3493 mtx_unlock(&Giant); /* dounmount() */
3500 zfs_unmount_snap_cb(const char *snapname, void *arg)
3502 return (zfs_unmount_snap(snapname));
3506 * When a clone is destroyed, its origin may also need to be destroyed,
3507 * in which case it must be unmounted. This routine will do that unmount
3511 zfs_destroy_unmount_origin(const char *fsname)
3517 error = dmu_objset_hold(fsname, FTAG, &os);
3520 ds = dmu_objset_ds(os);
3521 if (dsl_dir_is_clone(ds->ds_dir) && DS_IS_DEFER_DESTROY(ds->ds_prev)) {
3522 char originname[MAXNAMELEN];
3523 dsl_dataset_name(ds->ds_prev, originname);
3524 dmu_objset_rele(os, FTAG);
3525 (void) zfs_unmount_snap(originname);
3527 dmu_objset_rele(os, FTAG);
3533 * "snaps" -> { snapshot1, snapshot2 }
3534 * (optional boolean) "defer"
3537 * outnvl: snapshot -> error code (int32)
3542 zfs_ioc_destroy_snaps(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3549 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
3550 return (SET_ERROR(EINVAL));
3551 defer = nvlist_exists(innvl, "defer");
3553 poollen = strlen(poolname);
3554 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
3555 pair = nvlist_next_nvpair(snaps, pair)) {
3556 const char *name = nvpair_name(pair);
3559 * The snap must be in the specified pool to prevent the
3560 * invalid removal of zvol minors below.
3562 if (strncmp(name, poolname, poollen) != 0 ||
3563 (name[poollen] != '/' && name[poollen] != '@'))
3564 return (SET_ERROR(EXDEV));
3566 error = zfs_unmount_snap(name);
3569 #if defined(__FreeBSD__)
3570 zvol_remove_minors(name);
3574 return (dsl_destroy_snapshots_nvl(snaps, defer, outnvl));
3578 * Create bookmarks. Bookmark names are of the form <fs>#<bmark>.
3579 * All bookmarks must be in the same pool.
3582 * bookmark1 -> snapshot1, bookmark2 -> snapshot2
3585 * outnvl: bookmark -> error code (int32)
3590 zfs_ioc_bookmark(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3592 for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL);
3593 pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) {
3597 * Verify the snapshot argument.
3599 if (nvpair_value_string(pair, &snap_name) != 0)
3600 return (SET_ERROR(EINVAL));
3603 /* Verify that the keys (bookmarks) are unique */
3604 for (nvpair_t *pair2 = nvlist_next_nvpair(innvl, pair);
3605 pair2 != NULL; pair2 = nvlist_next_nvpair(innvl, pair2)) {
3606 if (strcmp(nvpair_name(pair), nvpair_name(pair2)) == 0)
3607 return (SET_ERROR(EINVAL));
3611 return (dsl_bookmark_create(innvl, outnvl));
3616 * property 1, property 2, ...
3620 * bookmark name 1 -> { property 1, property 2, ... },
3621 * bookmark name 2 -> { property 1, property 2, ... }
3626 zfs_ioc_get_bookmarks(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3628 return (dsl_get_bookmarks(fsname, innvl, outnvl));
3633 * bookmark name 1, bookmark name 2
3636 * outnvl: bookmark -> error code (int32)
3640 zfs_ioc_destroy_bookmarks(const char *poolname, nvlist_t *innvl,
3645 poollen = strlen(poolname);
3646 for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL);
3647 pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) {
3648 const char *name = nvpair_name(pair);
3649 const char *cp = strchr(name, '#');
3652 * The bookmark name must contain an #, and the part after it
3653 * must contain only valid characters.
3656 zfs_component_namecheck(cp + 1, NULL, NULL) != 0)
3657 return (SET_ERROR(EINVAL));
3660 * The bookmark must be in the specified pool.
3662 if (strncmp(name, poolname, poollen) != 0 ||
3663 (name[poollen] != '/' && name[poollen] != '#'))
3664 return (SET_ERROR(EXDEV));
3667 error = dsl_bookmark_destroy(innvl, outnvl);
3673 * zc_name name of dataset to destroy
3674 * zc_objset_type type of objset
3675 * zc_defer_destroy mark for deferred destroy
3680 zfs_ioc_destroy(zfs_cmd_t *zc)
3684 if (zc->zc_objset_type == DMU_OST_ZFS) {
3685 err = zfs_unmount_snap(zc->zc_name);
3690 if (strchr(zc->zc_name, '@'))
3691 err = dsl_destroy_snapshot(zc->zc_name, zc->zc_defer_destroy);
3693 err = dsl_destroy_head(zc->zc_name);
3694 if (zc->zc_objset_type == DMU_OST_ZVOL && err == 0)
3696 zvol_remove_minors(zc->zc_name);
3698 (void) zvol_remove_minor(zc->zc_name);
3704 * fsname is name of dataset to rollback (to most recent snapshot)
3706 * innvl is not used.
3708 * outnvl: "target" -> name of most recent snapshot
3713 zfs_ioc_rollback(const char *fsname, nvlist_t *args, nvlist_t *outnvl)
3718 if (getzfsvfs(fsname, &zfsvfs) == 0) {
3719 error = zfs_suspend_fs(zfsvfs);
3723 error = dsl_dataset_rollback(fsname, zfsvfs, outnvl);
3724 resume_err = zfs_resume_fs(zfsvfs, fsname);
3725 error = error ? error : resume_err;
3727 VFS_RELE(zfsvfs->z_vfs);
3729 error = dsl_dataset_rollback(fsname, NULL, outnvl);
3735 recursive_unmount(const char *fsname, void *arg)
3737 const char *snapname = arg;
3738 char fullname[MAXNAMELEN];
3740 (void) snprintf(fullname, sizeof (fullname), "%s@%s", fsname, snapname);
3741 return (zfs_unmount_snap(fullname));
3746 * zc_name old name of dataset
3747 * zc_value new name of dataset
3748 * zc_cookie recursive flag (only valid for snapshots)
3753 zfs_ioc_rename(zfs_cmd_t *zc)
3755 boolean_t recursive = zc->zc_cookie & 1;
3757 boolean_t allow_mounted = zc->zc_cookie & 2;
3761 zc->zc_value[sizeof (zc->zc_value) - 1] = '\0';
3762 if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
3763 strchr(zc->zc_value, '%'))
3764 return (SET_ERROR(EINVAL));
3766 at = strchr(zc->zc_name, '@');
3768 /* snaps must be in same fs */
3771 if (strncmp(zc->zc_name, zc->zc_value, at - zc->zc_name + 1))
3772 return (SET_ERROR(EXDEV));
3775 if (zc->zc_objset_type == DMU_OST_ZFS) {
3777 if (zc->zc_objset_type == DMU_OST_ZFS && allow_mounted) {
3779 error = dmu_objset_find(zc->zc_name,
3780 recursive_unmount, at + 1,
3781 recursive ? DS_FIND_CHILDREN : 0);
3787 error = dsl_dataset_rename_snapshot(zc->zc_name,
3788 at + 1, strchr(zc->zc_value, '@') + 1, recursive);
3794 if (zc->zc_objset_type == DMU_OST_ZVOL)
3795 (void) zvol_remove_minor(zc->zc_name);
3797 return (dsl_dir_rename(zc->zc_name, zc->zc_value));
3802 zfs_check_settable(const char *dsname, nvpair_t *pair, cred_t *cr)
3804 const char *propname = nvpair_name(pair);
3805 boolean_t issnap = (strchr(dsname, '@') != NULL);
3806 zfs_prop_t prop = zfs_name_to_prop(propname);
3810 if (prop == ZPROP_INVAL) {
3811 if (zfs_prop_user(propname)) {
3812 if (err = zfs_secpolicy_write_perms(dsname,
3813 ZFS_DELEG_PERM_USERPROP, cr))
3818 if (!issnap && zfs_prop_userquota(propname)) {
3819 const char *perm = NULL;
3820 const char *uq_prefix =
3821 zfs_userquota_prop_prefixes[ZFS_PROP_USERQUOTA];
3822 const char *gq_prefix =
3823 zfs_userquota_prop_prefixes[ZFS_PROP_GROUPQUOTA];
3825 if (strncmp(propname, uq_prefix,
3826 strlen(uq_prefix)) == 0) {
3827 perm = ZFS_DELEG_PERM_USERQUOTA;
3828 } else if (strncmp(propname, gq_prefix,
3829 strlen(gq_prefix)) == 0) {
3830 perm = ZFS_DELEG_PERM_GROUPQUOTA;
3832 /* USERUSED and GROUPUSED are read-only */
3833 return (SET_ERROR(EINVAL));
3836 if (err = zfs_secpolicy_write_perms(dsname, perm, cr))
3841 return (SET_ERROR(EINVAL));
3845 return (SET_ERROR(EINVAL));
3847 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
3849 * dsl_prop_get_all_impl() returns properties in this
3853 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
3854 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
3859 * Check that this value is valid for this pool version
3862 case ZFS_PROP_COMPRESSION:
3864 * If the user specified gzip compression, make sure
3865 * the SPA supports it. We ignore any errors here since
3866 * we'll catch them later.
3868 if (nvpair_type(pair) == DATA_TYPE_UINT64 &&
3869 nvpair_value_uint64(pair, &intval) == 0) {
3870 if (intval >= ZIO_COMPRESS_GZIP_1 &&
3871 intval <= ZIO_COMPRESS_GZIP_9 &&
3872 zfs_earlier_version(dsname,
3873 SPA_VERSION_GZIP_COMPRESSION)) {
3874 return (SET_ERROR(ENOTSUP));
3877 if (intval == ZIO_COMPRESS_ZLE &&
3878 zfs_earlier_version(dsname,
3879 SPA_VERSION_ZLE_COMPRESSION))
3880 return (SET_ERROR(ENOTSUP));
3882 if (intval == ZIO_COMPRESS_LZ4) {
3885 if ((err = spa_open(dsname, &spa, FTAG)) != 0)
3888 if (!spa_feature_is_enabled(spa,
3889 SPA_FEATURE_LZ4_COMPRESS)) {
3890 spa_close(spa, FTAG);
3891 return (SET_ERROR(ENOTSUP));
3893 spa_close(spa, FTAG);
3897 * If this is a bootable dataset then
3898 * verify that the compression algorithm
3899 * is supported for booting. We must return
3900 * something other than ENOTSUP since it
3901 * implies a downrev pool version.
3903 if (zfs_is_bootfs(dsname) &&
3904 !BOOTFS_COMPRESS_VALID(intval)) {
3905 return (SET_ERROR(ERANGE));
3910 case ZFS_PROP_COPIES:
3911 if (zfs_earlier_version(dsname, SPA_VERSION_DITTO_BLOCKS))
3912 return (SET_ERROR(ENOTSUP));
3915 case ZFS_PROP_DEDUP:
3916 if (zfs_earlier_version(dsname, SPA_VERSION_DEDUP))
3917 return (SET_ERROR(ENOTSUP));
3920 case ZFS_PROP_SHARESMB:
3921 if (zpl_earlier_version(dsname, ZPL_VERSION_FUID))
3922 return (SET_ERROR(ENOTSUP));
3925 case ZFS_PROP_ACLINHERIT:
3926 if (nvpair_type(pair) == DATA_TYPE_UINT64 &&
3927 nvpair_value_uint64(pair, &intval) == 0) {
3928 if (intval == ZFS_ACL_PASSTHROUGH_X &&
3929 zfs_earlier_version(dsname,
3930 SPA_VERSION_PASSTHROUGH_X))
3931 return (SET_ERROR(ENOTSUP));
3936 return (zfs_secpolicy_setprop(dsname, prop, pair, CRED()));
3940 * Checks for a race condition to make sure we don't increment a feature flag
3944 zfs_prop_activate_feature_check(void *arg, dmu_tx_t *tx)
3946 spa_t *spa = dmu_tx_pool(tx)->dp_spa;
3947 spa_feature_t *featurep = arg;
3949 if (!spa_feature_is_active(spa, *featurep))
3952 return (SET_ERROR(EBUSY));
3956 * The callback invoked on feature activation in the sync task caused by
3957 * zfs_prop_activate_feature.
3960 zfs_prop_activate_feature_sync(void *arg, dmu_tx_t *tx)
3962 spa_t *spa = dmu_tx_pool(tx)->dp_spa;
3963 spa_feature_t *featurep = arg;
3965 spa_feature_incr(spa, *featurep, tx);
3969 * Activates a feature on a pool in response to a property setting. This
3970 * creates a new sync task which modifies the pool to reflect the feature
3974 zfs_prop_activate_feature(spa_t *spa, spa_feature_t feature)
3978 /* EBUSY here indicates that the feature is already active */
3979 err = dsl_sync_task(spa_name(spa),
3980 zfs_prop_activate_feature_check, zfs_prop_activate_feature_sync,
3981 &feature, 2, ZFS_SPACE_CHECK_RESERVED);
3983 if (err != 0 && err != EBUSY)
3990 * Removes properties from the given props list that fail permission checks
3991 * needed to clear them and to restore them in case of a receive error. For each
3992 * property, make sure we have both set and inherit permissions.
3994 * Returns the first error encountered if any permission checks fail. If the
3995 * caller provides a non-NULL errlist, it also gives the complete list of names
3996 * of all the properties that failed a permission check along with the
3997 * corresponding error numbers. The caller is responsible for freeing the
4000 * If every property checks out successfully, zero is returned and the list
4001 * pointed at by errlist is NULL.
4004 zfs_check_clearable(char *dataset, nvlist_t *props, nvlist_t **errlist)
4007 nvpair_t *pair, *next_pair;
4014 VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0);
4016 zc = kmem_alloc(sizeof (zfs_cmd_t), KM_SLEEP);
4017 (void) strcpy(zc->zc_name, dataset);
4018 pair = nvlist_next_nvpair(props, NULL);
4019 while (pair != NULL) {
4020 next_pair = nvlist_next_nvpair(props, pair);
4022 (void) strcpy(zc->zc_value, nvpair_name(pair));
4023 if ((err = zfs_check_settable(dataset, pair, CRED())) != 0 ||
4024 (err = zfs_secpolicy_inherit_prop(zc, NULL, CRED())) != 0) {
4025 VERIFY(nvlist_remove_nvpair(props, pair) == 0);
4026 VERIFY(nvlist_add_int32(errors,
4027 zc->zc_value, err) == 0);
4031 kmem_free(zc, sizeof (zfs_cmd_t));
4033 if ((pair = nvlist_next_nvpair(errors, NULL)) == NULL) {
4034 nvlist_free(errors);
4037 VERIFY(nvpair_value_int32(pair, &rv) == 0);
4040 if (errlist == NULL)
4041 nvlist_free(errors);
4049 propval_equals(nvpair_t *p1, nvpair_t *p2)
4051 if (nvpair_type(p1) == DATA_TYPE_NVLIST) {
4052 /* dsl_prop_get_all_impl() format */
4054 VERIFY(nvpair_value_nvlist(p1, &attrs) == 0);
4055 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
4059 if (nvpair_type(p2) == DATA_TYPE_NVLIST) {
4061 VERIFY(nvpair_value_nvlist(p2, &attrs) == 0);
4062 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
4066 if (nvpair_type(p1) != nvpair_type(p2))
4069 if (nvpair_type(p1) == DATA_TYPE_STRING) {
4070 char *valstr1, *valstr2;
4072 VERIFY(nvpair_value_string(p1, (char **)&valstr1) == 0);
4073 VERIFY(nvpair_value_string(p2, (char **)&valstr2) == 0);
4074 return (strcmp(valstr1, valstr2) == 0);
4076 uint64_t intval1, intval2;
4078 VERIFY(nvpair_value_uint64(p1, &intval1) == 0);
4079 VERIFY(nvpair_value_uint64(p2, &intval2) == 0);
4080 return (intval1 == intval2);
4085 * Remove properties from props if they are not going to change (as determined
4086 * by comparison with origprops). Remove them from origprops as well, since we
4087 * do not need to clear or restore properties that won't change.
4090 props_reduce(nvlist_t *props, nvlist_t *origprops)
4092 nvpair_t *pair, *next_pair;
4094 if (origprops == NULL)
4095 return; /* all props need to be received */
4097 pair = nvlist_next_nvpair(props, NULL);
4098 while (pair != NULL) {
4099 const char *propname = nvpair_name(pair);
4102 next_pair = nvlist_next_nvpair(props, pair);
4104 if ((nvlist_lookup_nvpair(origprops, propname,
4105 &match) != 0) || !propval_equals(pair, match))
4106 goto next; /* need to set received value */
4108 /* don't clear the existing received value */
4109 (void) nvlist_remove_nvpair(origprops, match);
4110 /* don't bother receiving the property */
4111 (void) nvlist_remove_nvpair(props, pair);
4118 static boolean_t zfs_ioc_recv_inject_err;
4123 * zc_name name of containing filesystem
4124 * zc_nvlist_src{_size} nvlist of properties to apply
4125 * zc_value name of snapshot to create
4126 * zc_string name of clone origin (if DRR_FLAG_CLONE)
4127 * zc_cookie file descriptor to recv from
4128 * zc_begin_record the BEGIN record of the stream (not byteswapped)
4129 * zc_guid force flag
4130 * zc_cleanup_fd cleanup-on-exit file descriptor
4131 * zc_action_handle handle for this guid/ds mapping (or zero on first call)
4134 * zc_cookie number of bytes read
4135 * zc_nvlist_dst{_size} error for each unapplied received property
4136 * zc_obj zprop_errflags_t
4137 * zc_action_handle handle for this guid/ds mapping
4140 zfs_ioc_recv(zfs_cmd_t *zc)
4143 dmu_recv_cookie_t drc;
4144 boolean_t force = (boolean_t)zc->zc_guid;
4147 int props_error = 0;
4150 nvlist_t *props = NULL; /* sent properties */
4151 nvlist_t *origprops = NULL; /* existing properties */
4152 char *origin = NULL;
4154 char tofs[ZFS_MAXNAMELEN];
4155 cap_rights_t rights;
4156 boolean_t first_recvd_props = B_FALSE;
4158 if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
4159 strchr(zc->zc_value, '@') == NULL ||
4160 strchr(zc->zc_value, '%'))
4161 return (SET_ERROR(EINVAL));
4163 (void) strcpy(tofs, zc->zc_value);
4164 tosnap = strchr(tofs, '@');
4167 if (zc->zc_nvlist_src != 0 &&
4168 (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
4169 zc->zc_iflags, &props)) != 0)
4173 fp = getf(fd, cap_rights_init(&rights, CAP_PREAD));
4176 return (SET_ERROR(EBADF));
4179 VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0);
4181 if (zc->zc_string[0])
4182 origin = zc->zc_string;
4184 error = dmu_recv_begin(tofs, tosnap,
4185 &zc->zc_begin_record, force, origin, &drc);
4190 * Set properties before we receive the stream so that they are applied
4191 * to the new data. Note that we must call dmu_recv_stream() if
4192 * dmu_recv_begin() succeeds.
4194 if (props != NULL && !drc.drc_newfs) {
4195 if (spa_version(dsl_dataset_get_spa(drc.drc_ds)) >=
4196 SPA_VERSION_RECVD_PROPS &&
4197 !dsl_prop_get_hasrecvd(tofs))
4198 first_recvd_props = B_TRUE;
4201 * If new received properties are supplied, they are to
4202 * completely replace the existing received properties, so stash
4203 * away the existing ones.
4205 if (dsl_prop_get_received(tofs, &origprops) == 0) {
4206 nvlist_t *errlist = NULL;
4208 * Don't bother writing a property if its value won't
4209 * change (and avoid the unnecessary security checks).
4211 * The first receive after SPA_VERSION_RECVD_PROPS is a
4212 * special case where we blow away all local properties
4215 if (!first_recvd_props)
4216 props_reduce(props, origprops);
4217 if (zfs_check_clearable(tofs, origprops, &errlist) != 0)
4218 (void) nvlist_merge(errors, errlist, 0);
4219 nvlist_free(errlist);
4221 if (clear_received_props(tofs, origprops,
4222 first_recvd_props ? NULL : props) != 0)
4223 zc->zc_obj |= ZPROP_ERR_NOCLEAR;
4225 zc->zc_obj |= ZPROP_ERR_NOCLEAR;
4229 if (props != NULL) {
4230 props_error = dsl_prop_set_hasrecvd(tofs);
4232 if (props_error == 0) {
4233 (void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_RECEIVED,
4238 if (zc->zc_nvlist_dst_size != 0 &&
4239 (nvlist_smush(errors, zc->zc_nvlist_dst_size) != 0 ||
4240 put_nvlist(zc, errors) != 0)) {
4242 * Caller made zc->zc_nvlist_dst less than the minimum expected
4243 * size or supplied an invalid address.
4245 props_error = SET_ERROR(EINVAL);
4249 error = dmu_recv_stream(&drc, fp, &off, zc->zc_cleanup_fd,
4250 &zc->zc_action_handle);
4253 zfsvfs_t *zfsvfs = NULL;
4255 if (getzfsvfs(tofs, &zfsvfs) == 0) {
4259 error = zfs_suspend_fs(zfsvfs);
4261 * If the suspend fails, then the recv_end will
4262 * likely also fail, and clean up after itself.
4264 end_err = dmu_recv_end(&drc, zfsvfs);
4266 error = zfs_resume_fs(zfsvfs, tofs);
4267 error = error ? error : end_err;
4268 VFS_RELE(zfsvfs->z_vfs);
4270 error = dmu_recv_end(&drc, NULL);
4274 zc->zc_cookie = off - fp->f_offset;
4275 if (off >= 0 && off <= MAXOFFSET_T)
4279 if (zfs_ioc_recv_inject_err) {
4280 zfs_ioc_recv_inject_err = B_FALSE;
4287 zvol_create_minors(tofs);
4291 * On error, restore the original props.
4293 if (error != 0 && props != NULL && !drc.drc_newfs) {
4294 if (clear_received_props(tofs, props, NULL) != 0) {
4296 * We failed to clear the received properties.
4297 * Since we may have left a $recvd value on the
4298 * system, we can't clear the $hasrecvd flag.
4300 zc->zc_obj |= ZPROP_ERR_NORESTORE;
4301 } else if (first_recvd_props) {
4302 dsl_prop_unset_hasrecvd(tofs);
4305 if (origprops == NULL && !drc.drc_newfs) {
4306 /* We failed to stash the original properties. */
4307 zc->zc_obj |= ZPROP_ERR_NORESTORE;
4311 * dsl_props_set() will not convert RECEIVED to LOCAL on or
4312 * after SPA_VERSION_RECVD_PROPS, so we need to specify LOCAL
4313 * explictly if we're restoring local properties cleared in the
4314 * first new-style receive.
4316 if (origprops != NULL &&
4317 zfs_set_prop_nvlist(tofs, (first_recvd_props ?
4318 ZPROP_SRC_LOCAL : ZPROP_SRC_RECEIVED),
4319 origprops, NULL) != 0) {
4321 * We stashed the original properties but failed to
4324 zc->zc_obj |= ZPROP_ERR_NORESTORE;
4329 nvlist_free(origprops);
4330 nvlist_free(errors);
4334 error = props_error;
4341 * zc_name name of snapshot to send
4342 * zc_cookie file descriptor to send stream to
4343 * zc_obj fromorigin flag (mutually exclusive with zc_fromobj)
4344 * zc_sendobj objsetid of snapshot to send
4345 * zc_fromobj objsetid of incremental fromsnap (may be zero)
4346 * zc_guid if set, estimate size of stream only. zc_cookie is ignored.
4347 * output size in zc_objset_type.
4348 * zc_flags if =1, WRITE_EMBEDDED records are permitted
4351 * zc_objset_type estimated size, if zc_guid is set
4354 zfs_ioc_send(zfs_cmd_t *zc)
4358 boolean_t estimate = (zc->zc_guid != 0);
4359 boolean_t embedok = (zc->zc_flags & 0x1);
4361 if (zc->zc_obj != 0) {
4363 dsl_dataset_t *tosnap;
4365 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
4369 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &tosnap);
4371 dsl_pool_rele(dp, FTAG);
4375 if (dsl_dir_is_clone(tosnap->ds_dir))
4376 zc->zc_fromobj = tosnap->ds_dir->dd_phys->dd_origin_obj;
4377 dsl_dataset_rele(tosnap, FTAG);
4378 dsl_pool_rele(dp, FTAG);
4383 dsl_dataset_t *tosnap;
4384 dsl_dataset_t *fromsnap = NULL;
4386 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
4390 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &tosnap);
4392 dsl_pool_rele(dp, FTAG);
4396 if (zc->zc_fromobj != 0) {
4397 error = dsl_dataset_hold_obj(dp, zc->zc_fromobj,
4400 dsl_dataset_rele(tosnap, FTAG);
4401 dsl_pool_rele(dp, FTAG);
4406 error = dmu_send_estimate(tosnap, fromsnap,
4407 &zc->zc_objset_type);
4409 if (fromsnap != NULL)
4410 dsl_dataset_rele(fromsnap, FTAG);
4411 dsl_dataset_rele(tosnap, FTAG);
4412 dsl_pool_rele(dp, FTAG);
4415 cap_rights_t rights;
4417 fp = getf(zc->zc_cookie,
4418 cap_rights_init(&rights, CAP_WRITE));
4420 return (SET_ERROR(EBADF));
4423 error = dmu_send_obj(zc->zc_name, zc->zc_sendobj,
4425 zc->zc_fromobj, embedok, zc->zc_cookie, fp->f_vnode, &off);
4427 zc->zc_fromobj, embedok, zc->zc_cookie, fp, &off);
4430 if (off >= 0 && off <= MAXOFFSET_T)
4432 releasef(zc->zc_cookie);
4439 * zc_name name of snapshot on which to report progress
4440 * zc_cookie file descriptor of send stream
4443 * zc_cookie number of bytes written in send stream thus far
4446 zfs_ioc_send_progress(zfs_cmd_t *zc)
4450 dmu_sendarg_t *dsp = NULL;
4453 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
4457 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &ds);
4459 dsl_pool_rele(dp, FTAG);
4463 mutex_enter(&ds->ds_sendstream_lock);
4466 * Iterate over all the send streams currently active on this dataset.
4467 * If there's one which matches the specified file descriptor _and_ the
4468 * stream was started by the current process, return the progress of
4471 for (dsp = list_head(&ds->ds_sendstreams); dsp != NULL;
4472 dsp = list_next(&ds->ds_sendstreams, dsp)) {
4473 if (dsp->dsa_outfd == zc->zc_cookie &&
4474 dsp->dsa_proc == curproc)
4479 zc->zc_cookie = *(dsp->dsa_off);
4481 error = SET_ERROR(ENOENT);
4483 mutex_exit(&ds->ds_sendstream_lock);
4484 dsl_dataset_rele(ds, FTAG);
4485 dsl_pool_rele(dp, FTAG);
4490 zfs_ioc_inject_fault(zfs_cmd_t *zc)
4494 error = zio_inject_fault(zc->zc_name, (int)zc->zc_guid, &id,
4495 &zc->zc_inject_record);
4498 zc->zc_guid = (uint64_t)id;
4504 zfs_ioc_clear_fault(zfs_cmd_t *zc)
4506 return (zio_clear_fault((int)zc->zc_guid));
4510 zfs_ioc_inject_list_next(zfs_cmd_t *zc)
4512 int id = (int)zc->zc_guid;
4515 error = zio_inject_list_next(&id, zc->zc_name, sizeof (zc->zc_name),
4516 &zc->zc_inject_record);
4524 zfs_ioc_error_log(zfs_cmd_t *zc)
4528 size_t count = (size_t)zc->zc_nvlist_dst_size;
4530 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
4533 error = spa_get_errlog(spa, (void *)(uintptr_t)zc->zc_nvlist_dst,
4536 zc->zc_nvlist_dst_size = count;
4538 zc->zc_nvlist_dst_size = spa_get_errlog_size(spa);
4540 spa_close(spa, FTAG);
4546 zfs_ioc_clear(zfs_cmd_t *zc)
4553 * On zpool clear we also fix up missing slogs
4555 mutex_enter(&spa_namespace_lock);
4556 spa = spa_lookup(zc->zc_name);
4558 mutex_exit(&spa_namespace_lock);
4559 return (SET_ERROR(EIO));
4561 if (spa_get_log_state(spa) == SPA_LOG_MISSING) {
4562 /* we need to let spa_open/spa_load clear the chains */
4563 spa_set_log_state(spa, SPA_LOG_CLEAR);
4565 spa->spa_last_open_failed = 0;
4566 mutex_exit(&spa_namespace_lock);
4568 if (zc->zc_cookie & ZPOOL_NO_REWIND) {
4569 error = spa_open(zc->zc_name, &spa, FTAG);
4572 nvlist_t *config = NULL;
4574 if (zc->zc_nvlist_src == 0)
4575 return (SET_ERROR(EINVAL));
4577 if ((error = get_nvlist(zc->zc_nvlist_src,
4578 zc->zc_nvlist_src_size, zc->zc_iflags, &policy)) == 0) {
4579 error = spa_open_rewind(zc->zc_name, &spa, FTAG,
4581 if (config != NULL) {
4584 if ((err = put_nvlist(zc, config)) != 0)
4586 nvlist_free(config);
4588 nvlist_free(policy);
4595 spa_vdev_state_enter(spa, SCL_NONE);
4597 if (zc->zc_guid == 0) {
4600 vd = spa_lookup_by_guid(spa, zc->zc_guid, B_TRUE);
4602 (void) spa_vdev_state_exit(spa, NULL, ENODEV);
4603 spa_close(spa, FTAG);
4604 return (SET_ERROR(ENODEV));
4608 vdev_clear(spa, vd);
4610 (void) spa_vdev_state_exit(spa, NULL, 0);
4613 * Resume any suspended I/Os.
4615 if (zio_resume(spa) != 0)
4616 error = SET_ERROR(EIO);
4618 spa_close(spa, FTAG);
4624 zfs_ioc_pool_reopen(zfs_cmd_t *zc)
4629 error = spa_open(zc->zc_name, &spa, FTAG);
4633 spa_vdev_state_enter(spa, SCL_NONE);
4636 * If a resilver is already in progress then set the
4637 * spa_scrub_reopen flag to B_TRUE so that we don't restart
4638 * the scan as a side effect of the reopen. Otherwise, let
4639 * vdev_open() decided if a resilver is required.
4641 spa->spa_scrub_reopen = dsl_scan_resilvering(spa->spa_dsl_pool);
4642 vdev_reopen(spa->spa_root_vdev);
4643 spa->spa_scrub_reopen = B_FALSE;
4645 (void) spa_vdev_state_exit(spa, NULL, 0);
4646 spa_close(spa, FTAG);
4651 * zc_name name of filesystem
4652 * zc_value name of origin snapshot
4655 * zc_string name of conflicting snapshot, if there is one
4658 zfs_ioc_promote(zfs_cmd_t *zc)
4663 * We don't need to unmount *all* the origin fs's snapshots, but
4666 cp = strchr(zc->zc_value, '@');
4669 (void) dmu_objset_find(zc->zc_value,
4670 zfs_unmount_snap_cb, NULL, DS_FIND_SNAPSHOTS);
4671 return (dsl_dataset_promote(zc->zc_name, zc->zc_string));
4675 * Retrieve a single {user|group}{used|quota}@... property.
4678 * zc_name name of filesystem
4679 * zc_objset_type zfs_userquota_prop_t
4680 * zc_value domain name (eg. "S-1-234-567-89")
4681 * zc_guid RID/UID/GID
4684 * zc_cookie property value
4687 zfs_ioc_userspace_one(zfs_cmd_t *zc)
4692 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
4693 return (SET_ERROR(EINVAL));
4695 error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE);
4699 error = zfs_userspace_one(zfsvfs,
4700 zc->zc_objset_type, zc->zc_value, zc->zc_guid, &zc->zc_cookie);
4701 zfsvfs_rele(zfsvfs, FTAG);
4708 * zc_name name of filesystem
4709 * zc_cookie zap cursor
4710 * zc_objset_type zfs_userquota_prop_t
4711 * zc_nvlist_dst[_size] buffer to fill (not really an nvlist)
4714 * zc_nvlist_dst[_size] data buffer (array of zfs_useracct_t)
4715 * zc_cookie zap cursor
4718 zfs_ioc_userspace_many(zfs_cmd_t *zc)
4721 int bufsize = zc->zc_nvlist_dst_size;
4724 return (SET_ERROR(ENOMEM));
4726 int error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE);
4730 void *buf = kmem_alloc(bufsize, KM_SLEEP);
4732 error = zfs_userspace_many(zfsvfs, zc->zc_objset_type, &zc->zc_cookie,
4733 buf, &zc->zc_nvlist_dst_size);
4736 error = ddi_copyout(buf,
4737 (void *)(uintptr_t)zc->zc_nvlist_dst,
4738 zc->zc_nvlist_dst_size, zc->zc_iflags);
4740 kmem_free(buf, bufsize);
4741 zfsvfs_rele(zfsvfs, FTAG);
4748 * zc_name name of filesystem
4754 zfs_ioc_userspace_upgrade(zfs_cmd_t *zc)
4760 if (getzfsvfs(zc->zc_name, &zfsvfs) == 0) {
4761 if (!dmu_objset_userused_enabled(zfsvfs->z_os)) {
4763 * If userused is not enabled, it may be because the
4764 * objset needs to be closed & reopened (to grow the
4765 * objset_phys_t). Suspend/resume the fs will do that.
4767 error = zfs_suspend_fs(zfsvfs);
4769 dmu_objset_refresh_ownership(zfsvfs->z_os,
4771 error = zfs_resume_fs(zfsvfs, zc->zc_name);
4775 error = dmu_objset_userspace_upgrade(zfsvfs->z_os);
4776 VFS_RELE(zfsvfs->z_vfs);
4778 /* XXX kind of reading contents without owning */
4779 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
4783 error = dmu_objset_userspace_upgrade(os);
4784 dmu_objset_rele(os, FTAG);
4792 * We don't want to have a hard dependency
4793 * against some special symbols in sharefs
4794 * nfs, and smbsrv. Determine them if needed when
4795 * the first file system is shared.
4796 * Neither sharefs, nfs or smbsrv are unloadable modules.
4798 int (*znfsexport_fs)(void *arg);
4799 int (*zshare_fs)(enum sharefs_sys_op, share_t *, uint32_t);
4800 int (*zsmbexport_fs)(void *arg, boolean_t add_share);
4802 int zfs_nfsshare_inited;
4803 int zfs_smbshare_inited;
4805 ddi_modhandle_t nfs_mod;
4806 ddi_modhandle_t sharefs_mod;
4807 ddi_modhandle_t smbsrv_mod;
4809 kmutex_t zfs_share_lock;
4817 ASSERT(MUTEX_HELD(&zfs_share_lock));
4818 /* Both NFS and SMB shares also require sharetab support. */
4819 if (sharefs_mod == NULL && ((sharefs_mod =
4820 ddi_modopen("fs/sharefs",
4821 KRTLD_MODE_FIRST, &error)) == NULL)) {
4822 return (SET_ERROR(ENOSYS));
4824 if (zshare_fs == NULL && ((zshare_fs =
4825 (int (*)(enum sharefs_sys_op, share_t *, uint32_t))
4826 ddi_modsym(sharefs_mod, "sharefs_impl", &error)) == NULL)) {
4827 return (SET_ERROR(ENOSYS));
4834 zfs_ioc_share(zfs_cmd_t *zc)
4840 switch (zc->zc_share.z_sharetype) {
4842 case ZFS_UNSHARE_NFS:
4843 if (zfs_nfsshare_inited == 0) {
4844 mutex_enter(&zfs_share_lock);
4845 if (nfs_mod == NULL && ((nfs_mod = ddi_modopen("fs/nfs",
4846 KRTLD_MODE_FIRST, &error)) == NULL)) {
4847 mutex_exit(&zfs_share_lock);
4848 return (SET_ERROR(ENOSYS));
4850 if (znfsexport_fs == NULL &&
4851 ((znfsexport_fs = (int (*)(void *))
4853 "nfs_export", &error)) == NULL)) {
4854 mutex_exit(&zfs_share_lock);
4855 return (SET_ERROR(ENOSYS));
4857 error = zfs_init_sharefs();
4859 mutex_exit(&zfs_share_lock);
4860 return (SET_ERROR(ENOSYS));
4862 zfs_nfsshare_inited = 1;
4863 mutex_exit(&zfs_share_lock);
4867 case ZFS_UNSHARE_SMB:
4868 if (zfs_smbshare_inited == 0) {
4869 mutex_enter(&zfs_share_lock);
4870 if (smbsrv_mod == NULL && ((smbsrv_mod =
4871 ddi_modopen("drv/smbsrv",
4872 KRTLD_MODE_FIRST, &error)) == NULL)) {
4873 mutex_exit(&zfs_share_lock);
4874 return (SET_ERROR(ENOSYS));
4876 if (zsmbexport_fs == NULL && ((zsmbexport_fs =
4877 (int (*)(void *, boolean_t))ddi_modsym(smbsrv_mod,
4878 "smb_server_share", &error)) == NULL)) {
4879 mutex_exit(&zfs_share_lock);
4880 return (SET_ERROR(ENOSYS));
4882 error = zfs_init_sharefs();
4884 mutex_exit(&zfs_share_lock);
4885 return (SET_ERROR(ENOSYS));
4887 zfs_smbshare_inited = 1;
4888 mutex_exit(&zfs_share_lock);
4892 return (SET_ERROR(EINVAL));
4895 switch (zc->zc_share.z_sharetype) {
4897 case ZFS_UNSHARE_NFS:
4899 znfsexport_fs((void *)
4900 (uintptr_t)zc->zc_share.z_exportdata))
4904 case ZFS_UNSHARE_SMB:
4905 if (error = zsmbexport_fs((void *)
4906 (uintptr_t)zc->zc_share.z_exportdata,
4907 zc->zc_share.z_sharetype == ZFS_SHARE_SMB ?
4914 opcode = (zc->zc_share.z_sharetype == ZFS_SHARE_NFS ||
4915 zc->zc_share.z_sharetype == ZFS_SHARE_SMB) ?
4916 SHAREFS_ADD : SHAREFS_REMOVE;
4919 * Add or remove share from sharetab
4921 error = zshare_fs(opcode,
4922 (void *)(uintptr_t)zc->zc_share.z_sharedata,
4923 zc->zc_share.z_sharemax);
4932 ace_t full_access[] = {
4933 {(uid_t)-1, ACE_ALL_PERMS, ACE_EVERYONE, 0}
4938 * zc_name name of containing filesystem
4939 * zc_obj object # beyond which we want next in-use object #
4942 * zc_obj next in-use object #
4945 zfs_ioc_next_obj(zfs_cmd_t *zc)
4947 objset_t *os = NULL;
4950 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
4954 error = dmu_object_next(os, &zc->zc_obj, B_FALSE,
4955 os->os_dsl_dataset->ds_phys->ds_prev_snap_txg);
4957 dmu_objset_rele(os, FTAG);
4963 * zc_name name of filesystem
4964 * zc_value prefix name for snapshot
4965 * zc_cleanup_fd cleanup-on-exit file descriptor for calling process
4968 * zc_value short name of new snapshot
4971 zfs_ioc_tmp_snapshot(zfs_cmd_t *zc)
4978 error = zfs_onexit_fd_hold(zc->zc_cleanup_fd, &minor);
4982 snap_name = kmem_asprintf("%s-%016llx", zc->zc_value,
4983 (u_longlong_t)ddi_get_lbolt64());
4984 hold_name = kmem_asprintf("%%%s", zc->zc_value);
4986 error = dsl_dataset_snapshot_tmp(zc->zc_name, snap_name, minor,
4989 (void) strcpy(zc->zc_value, snap_name);
4992 zfs_onexit_fd_rele(zc->zc_cleanup_fd);
4998 * zc_name name of "to" snapshot
4999 * zc_value name of "from" snapshot
5000 * zc_cookie file descriptor to write diff data on
5003 * dmu_diff_record_t's to the file descriptor
5006 zfs_ioc_diff(zfs_cmd_t *zc)
5009 cap_rights_t rights;
5013 fp = getf(zc->zc_cookie, cap_rights_init(&rights, CAP_WRITE));
5015 return (SET_ERROR(EBADF));
5020 error = dmu_diff(zc->zc_name, zc->zc_value, fp->f_vnode, &off);
5022 error = dmu_diff(zc->zc_name, zc->zc_value, fp, &off);
5025 if (off >= 0 && off <= MAXOFFSET_T)
5027 releasef(zc->zc_cookie);
5034 * Remove all ACL files in shares dir
5037 zfs_smb_acl_purge(znode_t *dzp)
5040 zap_attribute_t zap;
5041 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
5044 for (zap_cursor_init(&zc, zfsvfs->z_os, dzp->z_id);
5045 (error = zap_cursor_retrieve(&zc, &zap)) == 0;
5046 zap_cursor_advance(&zc)) {
5047 if ((error = VOP_REMOVE(ZTOV(dzp), zap.za_name, kcred,
5051 zap_cursor_fini(&zc);
5057 zfs_ioc_smb_acl(zfs_cmd_t *zc)
5062 vnode_t *resourcevp = NULL;
5071 if ((error = lookupname(zc->zc_value, UIO_SYSSPACE,
5072 NO_FOLLOW, NULL, &vp)) != 0)
5075 /* Now make sure mntpnt and dataset are ZFS */
5077 if (strcmp(vp->v_vfsp->mnt_stat.f_fstypename, "zfs") != 0 ||
5078 (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource),
5079 zc->zc_name) != 0)) {
5081 return (SET_ERROR(EINVAL));
5085 zfsvfs = dzp->z_zfsvfs;
5089 * Create share dir if its missing.
5091 mutex_enter(&zfsvfs->z_lock);
5092 if (zfsvfs->z_shares_dir == 0) {
5095 tx = dmu_tx_create(zfsvfs->z_os);
5096 dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, TRUE,
5098 dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL);
5099 error = dmu_tx_assign(tx, TXG_WAIT);
5103 error = zfs_create_share_dir(zfsvfs, tx);
5107 mutex_exit(&zfsvfs->z_lock);
5113 mutex_exit(&zfsvfs->z_lock);
5115 ASSERT(zfsvfs->z_shares_dir);
5116 if ((error = zfs_zget(zfsvfs, zfsvfs->z_shares_dir, &sharedir)) != 0) {
5122 switch (zc->zc_cookie) {
5123 case ZFS_SMB_ACL_ADD:
5124 vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE;
5125 vattr.va_type = VREG;
5126 vattr.va_mode = S_IFREG|0777;
5130 vsec.vsa_mask = VSA_ACE;
5131 vsec.vsa_aclentp = &full_access;
5132 vsec.vsa_aclentsz = sizeof (full_access);
5133 vsec.vsa_aclcnt = 1;
5135 error = VOP_CREATE(ZTOV(sharedir), zc->zc_string,
5136 &vattr, EXCL, 0, &resourcevp, kcred, 0, NULL, &vsec);
5138 VN_RELE(resourcevp);
5141 case ZFS_SMB_ACL_REMOVE:
5142 error = VOP_REMOVE(ZTOV(sharedir), zc->zc_string, kcred,
5146 case ZFS_SMB_ACL_RENAME:
5147 if ((error = get_nvlist(zc->zc_nvlist_src,
5148 zc->zc_nvlist_src_size, zc->zc_iflags, &nvlist)) != 0) {
5153 if (nvlist_lookup_string(nvlist, ZFS_SMB_ACL_SRC, &src) ||
5154 nvlist_lookup_string(nvlist, ZFS_SMB_ACL_TARGET,
5157 VN_RELE(ZTOV(sharedir));
5159 nvlist_free(nvlist);
5162 error = VOP_RENAME(ZTOV(sharedir), src, ZTOV(sharedir), target,
5164 nvlist_free(nvlist);
5167 case ZFS_SMB_ACL_PURGE:
5168 error = zfs_smb_acl_purge(sharedir);
5172 error = SET_ERROR(EINVAL);
5177 VN_RELE(ZTOV(sharedir));
5183 return (EOPNOTSUPP);
5189 * "holds" -> { snapname -> holdname (string), ... }
5190 * (optional) "cleanup_fd" -> fd (int32)
5194 * snapname -> error value (int32)
5200 zfs_ioc_hold(const char *pool, nvlist_t *args, nvlist_t *errlist)
5203 int cleanup_fd = -1;
5207 error = nvlist_lookup_nvlist(args, "holds", &holds);
5209 return (SET_ERROR(EINVAL));
5211 if (nvlist_lookup_int32(args, "cleanup_fd", &cleanup_fd) == 0) {
5212 error = zfs_onexit_fd_hold(cleanup_fd, &minor);
5217 error = dsl_dataset_user_hold(holds, minor, errlist);
5219 zfs_onexit_fd_rele(cleanup_fd);
5224 * innvl is not used.
5227 * holdname -> time added (uint64 seconds since epoch)
5233 zfs_ioc_get_holds(const char *snapname, nvlist_t *args, nvlist_t *outnvl)
5235 return (dsl_dataset_get_holds(snapname, outnvl));
5240 * snapname -> { holdname, ... }
5245 * snapname -> error value (int32)
5251 zfs_ioc_release(const char *pool, nvlist_t *holds, nvlist_t *errlist)
5253 return (dsl_dataset_user_release(holds, errlist));
5258 * zc_name name of new filesystem or snapshot
5259 * zc_value full name of old snapshot
5262 * zc_cookie space in bytes
5263 * zc_objset_type compressed space in bytes
5264 * zc_perm_action uncompressed space in bytes
5267 zfs_ioc_space_written(zfs_cmd_t *zc)
5271 dsl_dataset_t *new, *old;
5273 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
5276 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &new);
5278 dsl_pool_rele(dp, FTAG);
5281 error = dsl_dataset_hold(dp, zc->zc_value, FTAG, &old);
5283 dsl_dataset_rele(new, FTAG);
5284 dsl_pool_rele(dp, FTAG);
5288 error = dsl_dataset_space_written(old, new, &zc->zc_cookie,
5289 &zc->zc_objset_type, &zc->zc_perm_action);
5290 dsl_dataset_rele(old, FTAG);
5291 dsl_dataset_rele(new, FTAG);
5292 dsl_pool_rele(dp, FTAG);
5298 * "firstsnap" -> snapshot name
5302 * "used" -> space in bytes
5303 * "compressed" -> compressed space in bytes
5304 * "uncompressed" -> uncompressed space in bytes
5308 zfs_ioc_space_snaps(const char *lastsnap, nvlist_t *innvl, nvlist_t *outnvl)
5312 dsl_dataset_t *new, *old;
5314 uint64_t used, comp, uncomp;
5316 if (nvlist_lookup_string(innvl, "firstsnap", &firstsnap) != 0)
5317 return (SET_ERROR(EINVAL));
5319 error = dsl_pool_hold(lastsnap, FTAG, &dp);
5323 error = dsl_dataset_hold(dp, lastsnap, FTAG, &new);
5325 dsl_pool_rele(dp, FTAG);
5328 error = dsl_dataset_hold(dp, firstsnap, FTAG, &old);
5330 dsl_dataset_rele(new, FTAG);
5331 dsl_pool_rele(dp, FTAG);
5335 error = dsl_dataset_space_wouldfree(old, new, &used, &comp, &uncomp);
5336 dsl_dataset_rele(old, FTAG);
5337 dsl_dataset_rele(new, FTAG);
5338 dsl_pool_rele(dp, FTAG);
5339 fnvlist_add_uint64(outnvl, "used", used);
5340 fnvlist_add_uint64(outnvl, "compressed", comp);
5341 fnvlist_add_uint64(outnvl, "uncompressed", uncomp);
5346 zfs_ioc_jail(zfs_cmd_t *zc)
5349 return (zone_dataset_attach(curthread->td_ucred, zc->zc_name,
5350 (int)zc->zc_jailid));
5354 zfs_ioc_unjail(zfs_cmd_t *zc)
5357 return (zone_dataset_detach(curthread->td_ucred, zc->zc_name,
5358 (int)zc->zc_jailid));
5363 * "fd" -> file descriptor to write stream to (int32)
5364 * (optional) "fromsnap" -> full snap name to send an incremental from
5365 * (optional) "embedok" -> (value ignored)
5366 * presence indicates DRR_WRITE_EMBEDDED records are permitted
5373 zfs_ioc_send_new(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl)
5375 cap_rights_t rights;
5378 char *fromname = NULL;
5382 error = nvlist_lookup_int32(innvl, "fd", &fd);
5384 return (SET_ERROR(EINVAL));
5386 (void) nvlist_lookup_string(innvl, "fromsnap", &fromname);
5388 embedok = nvlist_exists(innvl, "embedok");
5390 file_t *fp = getf(fd, cap_rights_init(&rights, CAP_READ));
5392 return (SET_ERROR(EBADF));
5396 error = dmu_send(snapname, fromname, embedok, fd, fp->f_vnode, &off);
5398 error = dmu_send(snapname, fromname, embedok, fd, fp, &off);
5402 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
5413 * Determine approximately how large a zfs send stream will be -- the number
5414 * of bytes that will be written to the fd supplied to zfs_ioc_send_new().
5417 * (optional) "fromsnap" -> full snap name to send an incremental from
5421 * "space" -> bytes of space (uint64)
5425 zfs_ioc_send_space(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl)
5428 dsl_dataset_t *fromsnap = NULL;
5429 dsl_dataset_t *tosnap;
5434 error = dsl_pool_hold(snapname, FTAG, &dp);
5438 error = dsl_dataset_hold(dp, snapname, FTAG, &tosnap);
5440 dsl_pool_rele(dp, FTAG);
5444 error = nvlist_lookup_string(innvl, "fromsnap", &fromname);
5446 error = dsl_dataset_hold(dp, fromname, FTAG, &fromsnap);
5448 dsl_dataset_rele(tosnap, FTAG);
5449 dsl_pool_rele(dp, FTAG);
5454 error = dmu_send_estimate(tosnap, fromsnap, &space);
5455 fnvlist_add_uint64(outnvl, "space", space);
5457 if (fromsnap != NULL)
5458 dsl_dataset_rele(fromsnap, FTAG);
5459 dsl_dataset_rele(tosnap, FTAG);
5460 dsl_pool_rele(dp, FTAG);
5465 static zfs_ioc_vec_t zfs_ioc_vec[ZFS_IOC_LAST - ZFS_IOC_FIRST];
5468 zfs_ioctl_register_legacy(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5469 zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck,
5470 boolean_t log_history, zfs_ioc_poolcheck_t pool_check)
5472 zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST];
5474 ASSERT3U(ioc, >=, ZFS_IOC_FIRST);
5475 ASSERT3U(ioc, <, ZFS_IOC_LAST);
5476 ASSERT3P(vec->zvec_legacy_func, ==, NULL);
5477 ASSERT3P(vec->zvec_func, ==, NULL);
5479 vec->zvec_legacy_func = func;
5480 vec->zvec_secpolicy = secpolicy;
5481 vec->zvec_namecheck = namecheck;
5482 vec->zvec_allow_log = log_history;
5483 vec->zvec_pool_check = pool_check;
5487 * See the block comment at the beginning of this file for details on
5488 * each argument to this function.
5491 zfs_ioctl_register(const char *name, zfs_ioc_t ioc, zfs_ioc_func_t *func,
5492 zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck,
5493 zfs_ioc_poolcheck_t pool_check, boolean_t smush_outnvlist,
5494 boolean_t allow_log)
5496 zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST];
5498 ASSERT3U(ioc, >=, ZFS_IOC_FIRST);
5499 ASSERT3U(ioc, <, ZFS_IOC_LAST);
5500 ASSERT3P(vec->zvec_legacy_func, ==, NULL);
5501 ASSERT3P(vec->zvec_func, ==, NULL);
5503 /* if we are logging, the name must be valid */
5504 ASSERT(!allow_log || namecheck != NO_NAME);
5506 vec->zvec_name = name;
5507 vec->zvec_func = func;
5508 vec->zvec_secpolicy = secpolicy;
5509 vec->zvec_namecheck = namecheck;
5510 vec->zvec_pool_check = pool_check;
5511 vec->zvec_smush_outnvlist = smush_outnvlist;
5512 vec->zvec_allow_log = allow_log;
5516 zfs_ioctl_register_pool(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5517 zfs_secpolicy_func_t *secpolicy, boolean_t log_history,
5518 zfs_ioc_poolcheck_t pool_check)
5520 zfs_ioctl_register_legacy(ioc, func, secpolicy,
5521 POOL_NAME, log_history, pool_check);
5525 zfs_ioctl_register_dataset_nolog(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5526 zfs_secpolicy_func_t *secpolicy, zfs_ioc_poolcheck_t pool_check)
5528 zfs_ioctl_register_legacy(ioc, func, secpolicy,
5529 DATASET_NAME, B_FALSE, pool_check);
5533 zfs_ioctl_register_pool_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func)
5535 zfs_ioctl_register_legacy(ioc, func, zfs_secpolicy_config,
5536 POOL_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5540 zfs_ioctl_register_pool_meta(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5541 zfs_secpolicy_func_t *secpolicy)
5543 zfs_ioctl_register_legacy(ioc, func, secpolicy,
5544 NO_NAME, B_FALSE, POOL_CHECK_NONE);
5548 zfs_ioctl_register_dataset_read_secpolicy(zfs_ioc_t ioc,
5549 zfs_ioc_legacy_func_t *func, zfs_secpolicy_func_t *secpolicy)
5551 zfs_ioctl_register_legacy(ioc, func, secpolicy,
5552 DATASET_NAME, B_FALSE, POOL_CHECK_SUSPENDED);
5556 zfs_ioctl_register_dataset_read(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func)
5558 zfs_ioctl_register_dataset_read_secpolicy(ioc, func,
5559 zfs_secpolicy_read);
5563 zfs_ioctl_register_dataset_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5564 zfs_secpolicy_func_t *secpolicy)
5566 zfs_ioctl_register_legacy(ioc, func, secpolicy,
5567 DATASET_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5571 zfs_ioctl_init(void)
5573 zfs_ioctl_register("snapshot", ZFS_IOC_SNAPSHOT,
5574 zfs_ioc_snapshot, zfs_secpolicy_snapshot, POOL_NAME,
5575 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5577 zfs_ioctl_register("log_history", ZFS_IOC_LOG_HISTORY,
5578 zfs_ioc_log_history, zfs_secpolicy_log_history, NO_NAME,
5579 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_FALSE);
5581 zfs_ioctl_register("space_snaps", ZFS_IOC_SPACE_SNAPS,
5582 zfs_ioc_space_snaps, zfs_secpolicy_read, DATASET_NAME,
5583 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5585 zfs_ioctl_register("send", ZFS_IOC_SEND_NEW,
5586 zfs_ioc_send_new, zfs_secpolicy_send_new, DATASET_NAME,
5587 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5589 zfs_ioctl_register("send_space", ZFS_IOC_SEND_SPACE,
5590 zfs_ioc_send_space, zfs_secpolicy_read, DATASET_NAME,
5591 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5593 zfs_ioctl_register("create", ZFS_IOC_CREATE,
5594 zfs_ioc_create, zfs_secpolicy_create_clone, DATASET_NAME,
5595 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5597 zfs_ioctl_register("clone", ZFS_IOC_CLONE,
5598 zfs_ioc_clone, zfs_secpolicy_create_clone, DATASET_NAME,
5599 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5601 zfs_ioctl_register("destroy_snaps", ZFS_IOC_DESTROY_SNAPS,
5602 zfs_ioc_destroy_snaps, zfs_secpolicy_destroy_snaps, POOL_NAME,
5603 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5605 zfs_ioctl_register("hold", ZFS_IOC_HOLD,
5606 zfs_ioc_hold, zfs_secpolicy_hold, POOL_NAME,
5607 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5608 zfs_ioctl_register("release", ZFS_IOC_RELEASE,
5609 zfs_ioc_release, zfs_secpolicy_release, POOL_NAME,
5610 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5612 zfs_ioctl_register("get_holds", ZFS_IOC_GET_HOLDS,
5613 zfs_ioc_get_holds, zfs_secpolicy_read, DATASET_NAME,
5614 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5616 zfs_ioctl_register("rollback", ZFS_IOC_ROLLBACK,
5617 zfs_ioc_rollback, zfs_secpolicy_rollback, DATASET_NAME,
5618 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_TRUE);
5620 zfs_ioctl_register("bookmark", ZFS_IOC_BOOKMARK,
5621 zfs_ioc_bookmark, zfs_secpolicy_bookmark, POOL_NAME,
5622 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5624 zfs_ioctl_register("get_bookmarks", ZFS_IOC_GET_BOOKMARKS,
5625 zfs_ioc_get_bookmarks, zfs_secpolicy_read, DATASET_NAME,
5626 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5628 zfs_ioctl_register("destroy_bookmarks", ZFS_IOC_DESTROY_BOOKMARKS,
5629 zfs_ioc_destroy_bookmarks, zfs_secpolicy_destroy_bookmarks,
5631 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5633 /* IOCTLS that use the legacy function signature */
5635 zfs_ioctl_register_legacy(ZFS_IOC_POOL_FREEZE, zfs_ioc_pool_freeze,
5636 zfs_secpolicy_config, NO_NAME, B_FALSE, POOL_CHECK_READONLY);
5638 zfs_ioctl_register_pool(ZFS_IOC_POOL_CREATE, zfs_ioc_pool_create,
5639 zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE);
5640 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SCAN,
5642 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_UPGRADE,
5643 zfs_ioc_pool_upgrade);
5644 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ADD,
5646 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_REMOVE,
5647 zfs_ioc_vdev_remove);
5648 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SET_STATE,
5649 zfs_ioc_vdev_set_state);
5650 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ATTACH,
5651 zfs_ioc_vdev_attach);
5652 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_DETACH,
5653 zfs_ioc_vdev_detach);
5654 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETPATH,
5655 zfs_ioc_vdev_setpath);
5656 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETFRU,
5657 zfs_ioc_vdev_setfru);
5658 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SET_PROPS,
5659 zfs_ioc_pool_set_props);
5660 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SPLIT,
5661 zfs_ioc_vdev_split);
5662 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_REGUID,
5663 zfs_ioc_pool_reguid);
5665 zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_CONFIGS,
5666 zfs_ioc_pool_configs, zfs_secpolicy_none);
5667 zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_TRYIMPORT,
5668 zfs_ioc_pool_tryimport, zfs_secpolicy_config);
5669 zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_FAULT,
5670 zfs_ioc_inject_fault, zfs_secpolicy_inject);
5671 zfs_ioctl_register_pool_meta(ZFS_IOC_CLEAR_FAULT,
5672 zfs_ioc_clear_fault, zfs_secpolicy_inject);
5673 zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_LIST_NEXT,
5674 zfs_ioc_inject_list_next, zfs_secpolicy_inject);
5677 * pool destroy, and export don't log the history as part of
5678 * zfsdev_ioctl, but rather zfs_ioc_pool_export
5679 * does the logging of those commands.
5681 zfs_ioctl_register_pool(ZFS_IOC_POOL_DESTROY, zfs_ioc_pool_destroy,
5682 zfs_secpolicy_config, B_FALSE, POOL_CHECK_NONE);
5683 zfs_ioctl_register_pool(ZFS_IOC_POOL_EXPORT, zfs_ioc_pool_export,
5684 zfs_secpolicy_config, B_FALSE, POOL_CHECK_NONE);
5686 zfs_ioctl_register_pool(ZFS_IOC_POOL_STATS, zfs_ioc_pool_stats,
5687 zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE);
5688 zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_PROPS, zfs_ioc_pool_get_props,
5689 zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE);
5691 zfs_ioctl_register_pool(ZFS_IOC_ERROR_LOG, zfs_ioc_error_log,
5692 zfs_secpolicy_inject, B_FALSE, POOL_CHECK_NONE);
5693 zfs_ioctl_register_pool(ZFS_IOC_DSOBJ_TO_DSNAME,
5694 zfs_ioc_dsobj_to_dsname,
5695 zfs_secpolicy_diff, B_FALSE, POOL_CHECK_NONE);
5696 zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_HISTORY,
5697 zfs_ioc_pool_get_history,
5698 zfs_secpolicy_config, B_FALSE, POOL_CHECK_SUSPENDED);
5700 zfs_ioctl_register_pool(ZFS_IOC_POOL_IMPORT, zfs_ioc_pool_import,
5701 zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE);
5703 zfs_ioctl_register_pool(ZFS_IOC_CLEAR, zfs_ioc_clear,
5704 zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE);
5705 zfs_ioctl_register_pool(ZFS_IOC_POOL_REOPEN, zfs_ioc_pool_reopen,
5706 zfs_secpolicy_config, B_TRUE, POOL_CHECK_SUSPENDED);
5708 zfs_ioctl_register_dataset_read(ZFS_IOC_SPACE_WRITTEN,
5709 zfs_ioc_space_written);
5710 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_RECVD_PROPS,
5711 zfs_ioc_objset_recvd_props);
5712 zfs_ioctl_register_dataset_read(ZFS_IOC_NEXT_OBJ,
5714 zfs_ioctl_register_dataset_read(ZFS_IOC_GET_FSACL,
5716 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_STATS,
5717 zfs_ioc_objset_stats);
5718 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_ZPLPROPS,
5719 zfs_ioc_objset_zplprops);
5720 zfs_ioctl_register_dataset_read(ZFS_IOC_DATASET_LIST_NEXT,
5721 zfs_ioc_dataset_list_next);
5722 zfs_ioctl_register_dataset_read(ZFS_IOC_SNAPSHOT_LIST_NEXT,
5723 zfs_ioc_snapshot_list_next);
5724 zfs_ioctl_register_dataset_read(ZFS_IOC_SEND_PROGRESS,
5725 zfs_ioc_send_progress);
5727 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_DIFF,
5728 zfs_ioc_diff, zfs_secpolicy_diff);
5729 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_STATS,
5730 zfs_ioc_obj_to_stats, zfs_secpolicy_diff);
5731 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_PATH,
5732 zfs_ioc_obj_to_path, zfs_secpolicy_diff);
5733 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_ONE,
5734 zfs_ioc_userspace_one, zfs_secpolicy_userspace_one);
5735 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_MANY,
5736 zfs_ioc_userspace_many, zfs_secpolicy_userspace_many);
5737 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_SEND,
5738 zfs_ioc_send, zfs_secpolicy_send);
5740 zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_PROP, zfs_ioc_set_prop,
5741 zfs_secpolicy_none);
5742 zfs_ioctl_register_dataset_modify(ZFS_IOC_DESTROY, zfs_ioc_destroy,
5743 zfs_secpolicy_destroy);
5744 zfs_ioctl_register_dataset_modify(ZFS_IOC_RENAME, zfs_ioc_rename,
5745 zfs_secpolicy_rename);
5746 zfs_ioctl_register_dataset_modify(ZFS_IOC_RECV, zfs_ioc_recv,
5747 zfs_secpolicy_recv);
5748 zfs_ioctl_register_dataset_modify(ZFS_IOC_PROMOTE, zfs_ioc_promote,
5749 zfs_secpolicy_promote);
5750 zfs_ioctl_register_dataset_modify(ZFS_IOC_INHERIT_PROP,
5751 zfs_ioc_inherit_prop, zfs_secpolicy_inherit_prop);
5752 zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_FSACL, zfs_ioc_set_fsacl,
5753 zfs_secpolicy_set_fsacl);
5755 zfs_ioctl_register_dataset_nolog(ZFS_IOC_SHARE, zfs_ioc_share,
5756 zfs_secpolicy_share, POOL_CHECK_NONE);
5757 zfs_ioctl_register_dataset_nolog(ZFS_IOC_SMB_ACL, zfs_ioc_smb_acl,
5758 zfs_secpolicy_smb_acl, POOL_CHECK_NONE);
5759 zfs_ioctl_register_dataset_nolog(ZFS_IOC_USERSPACE_UPGRADE,
5760 zfs_ioc_userspace_upgrade, zfs_secpolicy_userspace_upgrade,
5761 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5762 zfs_ioctl_register_dataset_nolog(ZFS_IOC_TMP_SNAPSHOT,
5763 zfs_ioc_tmp_snapshot, zfs_secpolicy_tmp_snapshot,
5764 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5767 zfs_ioctl_register_dataset_nolog(ZFS_IOC_JAIL, zfs_ioc_jail,
5768 zfs_secpolicy_config, POOL_CHECK_NONE);
5769 zfs_ioctl_register_dataset_nolog(ZFS_IOC_UNJAIL, zfs_ioc_unjail,
5770 zfs_secpolicy_config, POOL_CHECK_NONE);
5775 pool_status_check(const char *name, zfs_ioc_namecheck_t type,
5776 zfs_ioc_poolcheck_t check)
5781 ASSERT(type == POOL_NAME || type == DATASET_NAME);
5783 if (check & POOL_CHECK_NONE)
5786 error = spa_open(name, &spa, FTAG);
5788 if ((check & POOL_CHECK_SUSPENDED) && spa_suspended(spa))
5789 error = SET_ERROR(EAGAIN);
5790 else if ((check & POOL_CHECK_READONLY) && !spa_writeable(spa))
5791 error = SET_ERROR(EROFS);
5792 spa_close(spa, FTAG);
5798 * Find a free minor number.
5801 zfsdev_minor_alloc(void)
5803 static minor_t last_minor;
5806 ASSERT(MUTEX_HELD(&spa_namespace_lock));
5808 for (m = last_minor + 1; m != last_minor; m++) {
5809 if (m > ZFSDEV_MAX_MINOR)
5811 if (ddi_get_soft_state(zfsdev_state, m) == NULL) {
5821 zfs_ctldev_init(struct cdev *devp)
5824 zfs_soft_state_t *zs;
5826 ASSERT(MUTEX_HELD(&spa_namespace_lock));
5828 minor = zfsdev_minor_alloc();
5830 return (SET_ERROR(ENXIO));
5832 if (ddi_soft_state_zalloc(zfsdev_state, minor) != DDI_SUCCESS)
5833 return (SET_ERROR(EAGAIN));
5835 devfs_set_cdevpriv((void *)(uintptr_t)minor, zfsdev_close);
5837 zs = ddi_get_soft_state(zfsdev_state, minor);
5838 zs->zss_type = ZSST_CTLDEV;
5839 zfs_onexit_init((zfs_onexit_t **)&zs->zss_data);
5845 zfs_ctldev_destroy(zfs_onexit_t *zo, minor_t minor)
5847 ASSERT(MUTEX_HELD(&spa_namespace_lock));
5849 zfs_onexit_destroy(zo);
5850 ddi_soft_state_free(zfsdev_state, minor);
5854 zfsdev_get_soft_state(minor_t minor, enum zfs_soft_state_type which)
5856 zfs_soft_state_t *zp;
5858 zp = ddi_get_soft_state(zfsdev_state, minor);
5859 if (zp == NULL || zp->zss_type != which)
5862 return (zp->zss_data);
5866 zfsdev_open(struct cdev *devp, int flag, int mode, struct thread *td)
5871 if (getminor(*devp) != 0)
5872 return (zvol_open(devp, flag, otyp, cr));
5875 /* This is the control device. Allocate a new minor if requested. */
5877 mutex_enter(&spa_namespace_lock);
5878 error = zfs_ctldev_init(devp);
5879 mutex_exit(&spa_namespace_lock);
5886 zfsdev_close(void *data)
5889 minor_t minor = (minor_t)(uintptr_t)data;
5894 mutex_enter(&spa_namespace_lock);
5895 zo = zfsdev_get_soft_state(minor, ZSST_CTLDEV);
5897 mutex_exit(&spa_namespace_lock);
5900 zfs_ctldev_destroy(zo, minor);
5901 mutex_exit(&spa_namespace_lock);
5905 zfsdev_ioctl(struct cdev *dev, u_long zcmd, caddr_t arg, int flag,
5912 minor_t minor = getminor(dev);
5914 zfs_iocparm_t *zc_iocparm;
5915 int cflag, cmd, oldvecnum;
5916 boolean_t newioc, compat;
5917 void *compat_zc = NULL;
5918 cred_t *cr = td->td_ucred;
5920 const zfs_ioc_vec_t *vec;
5921 char *saved_poolname = NULL;
5922 nvlist_t *innvl = NULL;
5924 cflag = ZFS_CMD_COMPAT_NONE;
5926 newioc = B_TRUE; /* "new" style (zfs_iocparm_t) ioctl */
5928 len = IOCPARM_LEN(zcmd);
5929 vecnum = cmd = zcmd & 0xff;
5932 * Check if we are talking to supported older binaries
5933 * and translate zfs_cmd if necessary
5935 if (len != sizeof(zfs_iocparm_t)) {
5942 case sizeof(zfs_cmd_zcmd_t):
5943 cflag = ZFS_CMD_COMPAT_LZC;
5945 case sizeof(zfs_cmd_deadman_t):
5946 cflag = ZFS_CMD_COMPAT_DEADMAN;
5948 case sizeof(zfs_cmd_v28_t):
5949 cflag = ZFS_CMD_COMPAT_V28;
5951 case sizeof(zfs_cmd_v15_t):
5952 cflag = ZFS_CMD_COMPAT_V15;
5953 vecnum = zfs_ioctl_v15_to_v28[cmd];
5956 * Return without further handling
5957 * if the command is blacklisted.
5959 if (vecnum == ZFS_IOC_COMPAT_PASS)
5961 else if (vecnum == ZFS_IOC_COMPAT_FAIL)
5970 vecnum = cmd - ZFS_IOC_FIRST;
5971 ASSERT3U(getmajor(dev), ==, ddi_driver_major(zfs_dip));
5974 if (vecnum >= sizeof (zfs_ioc_vec) / sizeof (zfs_ioc_vec[0]))
5975 return (SET_ERROR(EINVAL));
5976 vec = &zfs_ioc_vec[vecnum];
5978 zc = kmem_zalloc(sizeof(zfs_cmd_t), KM_SLEEP);
5981 error = ddi_copyin((void *)arg, zc, sizeof (zfs_cmd_t), flag);
5983 error = SET_ERROR(EFAULT);
5986 #else /* !illumos */
5987 bzero(zc, sizeof(zfs_cmd_t));
5990 zc_iocparm = (void *)arg;
5992 switch (zc_iocparm->zfs_ioctl_version) {
5993 case ZFS_IOCVER_CURRENT:
5994 if (zc_iocparm->zfs_cmd_size != sizeof(zfs_cmd_t)) {
5995 error = SET_ERROR(EINVAL);
5999 case ZFS_IOCVER_ZCMD:
6000 if (zc_iocparm->zfs_cmd_size > sizeof(zfs_cmd_t) ||
6001 zc_iocparm->zfs_cmd_size < sizeof(zfs_cmd_zcmd_t)) {
6002 error = SET_ERROR(EFAULT);
6006 cflag = ZFS_CMD_COMPAT_ZCMD;
6009 error = SET_ERROR(EINVAL);
6015 ASSERT(sizeof(zfs_cmd_t) >= zc_iocparm->zfs_cmd_size);
6016 compat_zc = kmem_zalloc(sizeof(zfs_cmd_t), KM_SLEEP);
6017 bzero(compat_zc, sizeof(zfs_cmd_t));
6019 error = ddi_copyin((void *)(uintptr_t)zc_iocparm->zfs_cmd,
6020 compat_zc, zc_iocparm->zfs_cmd_size, flag);
6022 error = SET_ERROR(EFAULT);
6026 error = ddi_copyin((void *)(uintptr_t)zc_iocparm->zfs_cmd,
6027 zc, zc_iocparm->zfs_cmd_size, flag);
6029 error = SET_ERROR(EFAULT);
6037 ASSERT(compat_zc != NULL);
6038 zfs_cmd_compat_get(zc, compat_zc, cflag);
6040 ASSERT(compat_zc == NULL);
6041 zfs_cmd_compat_get(zc, arg, cflag);
6044 error = zfs_ioctl_compat_pre(zc, &vecnum, cflag);
6047 if (oldvecnum != vecnum)
6048 vec = &zfs_ioc_vec[vecnum];
6050 #endif /* !illumos */
6052 zc->zc_iflags = flag & FKIOCTL;
6053 if (zc->zc_nvlist_src_size != 0) {
6054 error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
6055 zc->zc_iflags, &innvl);
6060 /* rewrite innvl for backwards compatibility */
6062 innvl = zfs_ioctl_compat_innvl(zc, innvl, vecnum, cflag);
6065 * Ensure that all pool/dataset names are valid before we pass down to
6068 zc->zc_name[sizeof (zc->zc_name) - 1] = '\0';
6069 switch (vec->zvec_namecheck) {
6071 if (pool_namecheck(zc->zc_name, NULL, NULL) != 0)
6072 error = SET_ERROR(EINVAL);
6074 error = pool_status_check(zc->zc_name,
6075 vec->zvec_namecheck, vec->zvec_pool_check);
6079 if (dataset_namecheck(zc->zc_name, NULL, NULL) != 0)
6080 error = SET_ERROR(EINVAL);
6082 error = pool_status_check(zc->zc_name,
6083 vec->zvec_namecheck, vec->zvec_pool_check);
6090 if (error == 0 && !(flag & FKIOCTL))
6091 error = vec->zvec_secpolicy(zc, innvl, cr);
6096 /* legacy ioctls can modify zc_name */
6097 len = strcspn(zc->zc_name, "/@#") + 1;
6098 saved_poolname = kmem_alloc(len, KM_SLEEP);
6099 (void) strlcpy(saved_poolname, zc->zc_name, len);
6101 if (vec->zvec_func != NULL) {
6105 nvlist_t *lognv = NULL;
6107 ASSERT(vec->zvec_legacy_func == NULL);
6110 * Add the innvl to the lognv before calling the func,
6111 * in case the func changes the innvl.
6113 if (vec->zvec_allow_log) {
6114 lognv = fnvlist_alloc();
6115 fnvlist_add_string(lognv, ZPOOL_HIST_IOCTL,
6117 if (!nvlist_empty(innvl)) {
6118 fnvlist_add_nvlist(lognv, ZPOOL_HIST_INPUT_NVL,
6123 outnvl = fnvlist_alloc();
6124 error = vec->zvec_func(zc->zc_name, innvl, outnvl);
6126 if (error == 0 && vec->zvec_allow_log &&
6127 spa_open(zc->zc_name, &spa, FTAG) == 0) {
6128 if (!nvlist_empty(outnvl)) {
6129 fnvlist_add_nvlist(lognv, ZPOOL_HIST_OUTPUT_NVL,
6132 (void) spa_history_log_nvl(spa, lognv);
6133 spa_close(spa, FTAG);
6135 fnvlist_free(lognv);
6137 /* rewrite outnvl for backwards compatibility */
6139 outnvl = zfs_ioctl_compat_outnvl(zc, outnvl, vecnum,
6142 if (!nvlist_empty(outnvl) || zc->zc_nvlist_dst_size != 0) {
6144 if (vec->zvec_smush_outnvlist) {
6145 smusherror = nvlist_smush(outnvl,
6146 zc->zc_nvlist_dst_size);
6148 if (smusherror == 0)
6149 puterror = put_nvlist(zc, outnvl);
6155 nvlist_free(outnvl);
6157 error = vec->zvec_legacy_func(zc);
6164 rc = ddi_copyout(zc, (void *)arg, sizeof (zfs_cmd_t), flag);
6165 if (error == 0 && rc != 0)
6166 error = SET_ERROR(EFAULT);
6169 zfs_ioctl_compat_post(zc, cmd, cflag);
6171 ASSERT(compat_zc != NULL);
6172 ASSERT(sizeof(zfs_cmd_t) >= zc_iocparm->zfs_cmd_size);
6174 zfs_cmd_compat_put(zc, compat_zc, vecnum, cflag);
6175 rc = ddi_copyout(compat_zc,
6176 (void *)(uintptr_t)zc_iocparm->zfs_cmd,
6177 zc_iocparm->zfs_cmd_size, flag);
6178 if (error == 0 && rc != 0)
6179 error = SET_ERROR(EFAULT);
6180 kmem_free(compat_zc, sizeof (zfs_cmd_t));
6182 zfs_cmd_compat_put(zc, arg, vecnum, cflag);
6187 rc = ddi_copyout(zc, (void *)(uintptr_t)zc_iocparm->zfs_cmd,
6188 sizeof (zfs_cmd_t), flag);
6189 if (error == 0 && rc != 0)
6190 error = SET_ERROR(EFAULT);
6193 if (error == 0 && vec->zvec_allow_log) {
6194 char *s = tsd_get(zfs_allow_log_key);
6197 (void) tsd_set(zfs_allow_log_key, saved_poolname);
6199 if (saved_poolname != NULL)
6200 strfree(saved_poolname);
6203 kmem_free(zc, sizeof (zfs_cmd_t));
6209 zfs_attach(dev_info_t *dip, ddi_attach_cmd_t cmd)
6211 if (cmd != DDI_ATTACH)
6212 return (DDI_FAILURE);
6214 if (ddi_create_minor_node(dip, "zfs", S_IFCHR, 0,
6215 DDI_PSEUDO, 0) == DDI_FAILURE)
6216 return (DDI_FAILURE);
6220 ddi_report_dev(dip);
6222 return (DDI_SUCCESS);
6226 zfs_detach(dev_info_t *dip, ddi_detach_cmd_t cmd)
6228 if (spa_busy() || zfs_busy() || zvol_busy())
6229 return (DDI_FAILURE);
6231 if (cmd != DDI_DETACH)
6232 return (DDI_FAILURE);
6236 ddi_prop_remove_all(dip);
6237 ddi_remove_minor_node(dip, NULL);
6239 return (DDI_SUCCESS);
6244 zfs_info(dev_info_t *dip, ddi_info_cmd_t infocmd, void *arg, void **result)
6247 case DDI_INFO_DEVT2DEVINFO:
6249 return (DDI_SUCCESS);
6251 case DDI_INFO_DEVT2INSTANCE:
6252 *result = (void *)0;
6253 return (DDI_SUCCESS);
6256 return (DDI_FAILURE);
6261 * OK, so this is a little weird.
6263 * /dev/zfs is the control node, i.e. minor 0.
6264 * /dev/zvol/[r]dsk/pool/dataset are the zvols, minor > 0.
6266 * /dev/zfs has basically nothing to do except serve up ioctls,
6267 * so most of the standard driver entry points are in zvol.c.
6270 static struct cb_ops zfs_cb_ops = {
6271 zfsdev_open, /* open */
6272 zfsdev_close, /* close */
6273 zvol_strategy, /* strategy */
6275 zvol_dump, /* dump */
6276 zvol_read, /* read */
6277 zvol_write, /* write */
6278 zfsdev_ioctl, /* ioctl */
6282 nochpoll, /* poll */
6283 ddi_prop_op, /* prop_op */
6284 NULL, /* streamtab */
6285 D_NEW | D_MP | D_64BIT, /* Driver compatibility flag */
6286 CB_REV, /* version */
6287 nodev, /* async read */
6288 nodev, /* async write */
6291 static struct dev_ops zfs_dev_ops = {
6292 DEVO_REV, /* version */
6294 zfs_info, /* info */
6295 nulldev, /* identify */
6296 nulldev, /* probe */
6297 zfs_attach, /* attach */
6298 zfs_detach, /* detach */
6300 &zfs_cb_ops, /* driver operations */
6301 NULL, /* no bus operations */
6303 ddi_quiesce_not_needed, /* quiesce */
6306 static struct modldrv zfs_modldrv = {
6312 static struct modlinkage modlinkage = {
6314 (void *)&zfs_modlfs,
6315 (void *)&zfs_modldrv,
6320 static struct cdevsw zfs_cdevsw = {
6321 .d_version = D_VERSION,
6322 .d_open = zfsdev_open,
6323 .d_ioctl = zfsdev_ioctl,
6324 .d_name = ZFS_DEV_NAME
6328 zfs_allow_log_destroy(void *arg)
6330 char *poolname = arg;
6337 zfsdev = make_dev(&zfs_cdevsw, 0x0, UID_ROOT, GID_OPERATOR, 0666,
6345 destroy_dev(zfsdev);
6348 static struct root_hold_token *zfs_root_token;
6349 struct proc *zfsproc;
6357 spa_init(FREAD | FWRITE);
6362 if ((error = mod_install(&modlinkage)) != 0) {
6369 tsd_create(&zfs_fsyncer_key, NULL);
6370 tsd_create(&rrw_tsd_key, rrw_tsd_destroy);
6371 tsd_create(&zfs_allow_log_key, zfs_allow_log_destroy);
6373 error = ldi_ident_from_mod(&modlinkage, &zfs_li);
6375 mutex_init(&zfs_share_lock, NULL, MUTEX_DEFAULT, NULL);
6385 if (spa_busy() || zfs_busy() || zvol_busy() || zio_injection_enabled)
6386 return (SET_ERROR(EBUSY));
6388 if ((error = mod_remove(&modlinkage)) != 0)
6394 if (zfs_nfsshare_inited)
6395 (void) ddi_modclose(nfs_mod);
6396 if (zfs_smbshare_inited)
6397 (void) ddi_modclose(smbsrv_mod);
6398 if (zfs_nfsshare_inited || zfs_smbshare_inited)
6399 (void) ddi_modclose(sharefs_mod);
6401 tsd_destroy(&zfs_fsyncer_key);
6402 ldi_ident_release(zfs_li);
6404 mutex_destroy(&zfs_share_lock);
6410 _info(struct modinfo *modinfop)
6412 return (mod_info(&modlinkage, modinfop));
6416 static int zfs__init(void);
6417 static int zfs__fini(void);
6418 static void zfs_shutdown(void *, int);
6420 static eventhandler_tag zfs_shutdown_event_tag;
6426 zfs_root_token = root_mount_hold("ZFS");
6428 mutex_init(&zfs_share_lock, NULL, MUTEX_DEFAULT, NULL);
6430 spa_init(FREAD | FWRITE);
6435 tsd_create(&zfs_fsyncer_key, NULL);
6436 tsd_create(&rrw_tsd_key, rrw_tsd_destroy);
6437 tsd_create(&zfs_allow_log_key, zfs_allow_log_destroy);
6439 printf("ZFS storage pool version: features support (" SPA_VERSION_STRING ")\n");
6440 root_mount_rel(zfs_root_token);
6450 if (spa_busy() || zfs_busy() || zvol_busy() ||
6451 zio_injection_enabled) {
6460 tsd_destroy(&zfs_fsyncer_key);
6461 tsd_destroy(&rrw_tsd_key);
6462 tsd_destroy(&zfs_allow_log_key);
6464 mutex_destroy(&zfs_share_lock);
6470 zfs_shutdown(void *arg __unused, int howto __unused)
6474 * ZFS fini routines can not properly work in a panic-ed system.
6476 if (panicstr == NULL)
6482 zfs_modevent(module_t mod, int type, void *unused __unused)
6490 zfs_shutdown_event_tag = EVENTHANDLER_REGISTER(
6491 shutdown_post_sync, zfs_shutdown, NULL,
6492 SHUTDOWN_PRI_FIRST);
6496 if (err == 0 && zfs_shutdown_event_tag != NULL)
6497 EVENTHANDLER_DEREGISTER(shutdown_post_sync,
6498 zfs_shutdown_event_tag);
6505 return (EOPNOTSUPP);
6508 static moduledata_t zfs_mod = {
6513 DECLARE_MODULE(zfsctrl, zfs_mod, SI_SUB_VFS, SI_ORDER_ANY);
6514 MODULE_VERSION(zfsctrl, 1);
6515 MODULE_DEPEND(zfsctrl, opensolaris, 1, 1, 1);
6516 MODULE_DEPEND(zfsctrl, krpc, 1, 1, 1);
6517 MODULE_DEPEND(zfsctrl, acl_nfs4, 1, 1, 1);