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 * Portions Copyright 2011 Martin Matuska
25 * Copyright 2015, OmniTI Computer Consulting, Inc. All rights reserved.
26 * Portions Copyright 2012 Pawel Jakub Dawidek <pawel@dawidek.net>
27 * Copyright (c) 2014, 2016 Joyent, Inc. All rights reserved.
28 * Copyright 2016 Nexenta Systems, Inc. All rights reserved.
29 * Copyright (c) 2014, Joyent, Inc. All rights reserved.
30 * Copyright (c) 2011, 2015 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 Integros [integros.com]
34 * Copyright 2016 Toomas Soome <tsoome@me.com>
35 * Copyright (c) 2016 Actifio, Inc. All rights reserved.
36 * Copyright (c) 2017, loli10K <ezomori.nozomu@gmail.com>. All rights reserved.
37 * Copyright (c) 2017 Datto Inc.
38 * Copyright 2017 RackTop Systems.
44 * This file handles the ioctls to /dev/zfs, used for configuring ZFS storage
45 * pools and filesystems, e.g. with /sbin/zfs and /sbin/zpool.
47 * There are two ways that we handle ioctls: the legacy way where almost
48 * all of the logic is in the ioctl callback, and the new way where most
49 * of the marshalling is handled in the common entry point, zfsdev_ioctl().
51 * Non-legacy ioctls should be registered by calling
52 * zfs_ioctl_register() from zfs_ioctl_init(). The ioctl is invoked
53 * from userland by lzc_ioctl().
55 * The registration arguments are as follows:
58 * The name of the ioctl. This is used for history logging. If the
59 * ioctl returns successfully (the callback returns 0), and allow_log
60 * is true, then a history log entry will be recorded with the input &
61 * output nvlists. The log entry can be printed with "zpool history -i".
64 * The ioctl request number, which userland will pass to ioctl(2).
65 * The ioctl numbers can change from release to release, because
66 * the caller (libzfs) must be matched to the kernel.
68 * zfs_secpolicy_func_t *secpolicy
69 * This function will be called before the zfs_ioc_func_t, to
70 * determine if this operation is permitted. It should return EPERM
71 * on failure, and 0 on success. Checks include determining if the
72 * dataset is visible in this zone, and if the user has either all
73 * zfs privileges in the zone (SYS_MOUNT), or has been granted permission
74 * to do this operation on this dataset with "zfs allow".
76 * zfs_ioc_namecheck_t namecheck
77 * This specifies what to expect in the zfs_cmd_t:zc_name -- a pool
78 * name, a dataset name, or nothing. If the name is not well-formed,
79 * the ioctl will fail and the callback will not be called.
80 * Therefore, the callback can assume that the name is well-formed
81 * (e.g. is null-terminated, doesn't have more than one '@' character,
82 * doesn't have invalid characters).
84 * zfs_ioc_poolcheck_t pool_check
85 * This specifies requirements on the pool state. If the pool does
86 * not meet them (is suspended or is readonly), the ioctl will fail
87 * and the callback will not be called. If any checks are specified
88 * (i.e. it is not POOL_CHECK_NONE), namecheck must not be NO_NAME.
89 * Multiple checks can be or-ed together (e.g. POOL_CHECK_SUSPENDED |
90 * POOL_CHECK_READONLY).
92 * boolean_t smush_outnvlist
93 * If smush_outnvlist is true, then the output is presumed to be a
94 * list of errors, and it will be "smushed" down to fit into the
95 * caller's buffer, by removing some entries and replacing them with a
96 * single "N_MORE_ERRORS" entry indicating how many were removed. See
97 * nvlist_smush() for details. If smush_outnvlist is false, and the
98 * outnvlist does not fit into the userland-provided buffer, then the
99 * ioctl will fail with ENOMEM.
101 * zfs_ioc_func_t *func
102 * The callback function that will perform the operation.
104 * The callback should return 0 on success, or an error number on
105 * failure. If the function fails, the userland ioctl will return -1,
106 * and errno will be set to the callback's return value. The callback
107 * will be called with the following arguments:
110 * The name of the pool or dataset to operate on, from
111 * zfs_cmd_t:zc_name. The 'namecheck' argument specifies the
112 * expected type (pool, dataset, or none).
115 * The input nvlist, deserialized from zfs_cmd_t:zc_nvlist_src. Or
116 * NULL if no input nvlist was provided. Changes to this nvlist are
117 * ignored. If the input nvlist could not be deserialized, the
118 * ioctl will fail and the callback will not be called.
121 * The output nvlist, initially empty. The callback can fill it in,
122 * and it will be returned to userland by serializing it into
123 * zfs_cmd_t:zc_nvlist_dst. If it is non-empty, and serialization
124 * fails (e.g. because the caller didn't supply a large enough
125 * buffer), then the overall ioctl will fail. See the
126 * 'smush_nvlist' argument above for additional behaviors.
128 * There are two typical uses of the output nvlist:
129 * - To return state, e.g. property values. In this case,
130 * smush_outnvlist should be false. If the buffer was not large
131 * enough, the caller will reallocate a larger buffer and try
134 * - To return multiple errors from an ioctl which makes on-disk
135 * changes. In this case, smush_outnvlist should be true.
136 * Ioctls which make on-disk modifications should generally not
137 * use the outnvl if they succeed, because the caller can not
138 * distinguish between the operation failing, and
139 * deserialization failing.
142 #include <sys/types.h>
143 #include <sys/param.h>
144 #include <sys/errno.h>
147 #include <sys/modctl.h>
148 #include <sys/open.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>
161 #include <sys/vdev_impl.h>
162 #include <sys/priv_impl.h>
164 #include <sys/dsl_dir.h>
165 #include <sys/dsl_dataset.h>
166 #include <sys/dsl_prop.h>
167 #include <sys/dsl_deleg.h>
168 #include <sys/dmu_objset.h>
169 #include <sys/dmu_impl.h>
170 #include <sys/dmu_tx.h>
172 #include <sys/sunddi.h>
173 #include <sys/sunldi.h>
174 #include <sys/policy.h>
175 #include <sys/zone.h>
176 #include <sys/nvpair.h>
177 #include <sys/pathname.h>
178 #include <sys/mount.h>
180 #include <sys/fs/zfs.h>
181 #include <sys/zfs_ctldir.h>
182 #include <sys/zfs_dir.h>
183 #include <sys/zfs_onexit.h>
184 #include <sys/zvol.h>
185 #include <sys/dsl_scan.h>
186 #include <sharefs/share.h>
187 #include <sys/fm/util.h>
189 #include <sys/dmu_send.h>
190 #include <sys/dsl_destroy.h>
191 #include <sys/dsl_bookmark.h>
192 #include <sys/dsl_userhold.h>
193 #include <sys/zfeature.h>
194 #include <sys/zio_checksum.h>
196 #include <linux/miscdevice.h>
197 #include <linux/slab.h>
199 #include "zfs_namecheck.h"
200 #include "zfs_prop.h"
201 #include "zfs_deleg.h"
202 #include "zfs_comutil.h"
205 * Limit maximum nvlist size. We don't want users passing in insane values
206 * for zc->zc_nvlist_src_size, since we will need to allocate that much memory.
208 #define MAX_NVLIST_SRC_SIZE KMALLOC_MAX_SIZE
210 kmutex_t zfsdev_state_lock;
211 zfsdev_state_t *zfsdev_state_list;
213 extern void zfs_init(void);
214 extern void zfs_fini(void);
216 uint_t zfs_fsyncer_key;
217 extern uint_t rrw_tsd_key;
218 static uint_t zfs_allow_log_key;
220 typedef int zfs_ioc_legacy_func_t(zfs_cmd_t *);
221 typedef int zfs_ioc_func_t(const char *, nvlist_t *, nvlist_t *);
222 typedef int zfs_secpolicy_func_t(zfs_cmd_t *, nvlist_t *, cred_t *);
228 } zfs_ioc_namecheck_t;
231 POOL_CHECK_NONE = 1 << 0,
232 POOL_CHECK_SUSPENDED = 1 << 1,
233 POOL_CHECK_READONLY = 1 << 2,
234 } zfs_ioc_poolcheck_t;
236 typedef struct zfs_ioc_vec {
237 zfs_ioc_legacy_func_t *zvec_legacy_func;
238 zfs_ioc_func_t *zvec_func;
239 zfs_secpolicy_func_t *zvec_secpolicy;
240 zfs_ioc_namecheck_t zvec_namecheck;
241 boolean_t zvec_allow_log;
242 zfs_ioc_poolcheck_t zvec_pool_check;
243 boolean_t zvec_smush_outnvlist;
244 const char *zvec_name;
247 /* This array is indexed by zfs_userquota_prop_t */
248 static const char *userquota_perms[] = {
249 ZFS_DELEG_PERM_USERUSED,
250 ZFS_DELEG_PERM_USERQUOTA,
251 ZFS_DELEG_PERM_GROUPUSED,
252 ZFS_DELEG_PERM_GROUPQUOTA,
253 ZFS_DELEG_PERM_USEROBJUSED,
254 ZFS_DELEG_PERM_USEROBJQUOTA,
255 ZFS_DELEG_PERM_GROUPOBJUSED,
256 ZFS_DELEG_PERM_GROUPOBJQUOTA,
259 static int zfs_ioc_userspace_upgrade(zfs_cmd_t *zc);
260 static int zfs_ioc_userobjspace_upgrade(zfs_cmd_t *zc);
261 static int zfs_check_settable(const char *name, nvpair_t *property,
263 static int zfs_check_clearable(char *dataset, nvlist_t *props,
265 static int zfs_fill_zplprops_root(uint64_t, nvlist_t *, nvlist_t *,
267 int zfs_set_prop_nvlist(const char *, zprop_source_t, nvlist_t *, nvlist_t *);
268 static int get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp);
271 history_str_free(char *buf)
273 kmem_free(buf, HIS_MAX_RECORD_LEN);
277 history_str_get(zfs_cmd_t *zc)
281 if (zc->zc_history == 0)
284 buf = kmem_alloc(HIS_MAX_RECORD_LEN, KM_SLEEP);
285 if (copyinstr((void *)(uintptr_t)zc->zc_history,
286 buf, HIS_MAX_RECORD_LEN, NULL) != 0) {
287 history_str_free(buf);
291 buf[HIS_MAX_RECORD_LEN -1] = '\0';
297 * Check to see if the named dataset is currently defined as bootable
300 zfs_is_bootfs(const char *name)
304 if (dmu_objset_hold(name, FTAG, &os) == 0) {
306 ret = (dmu_objset_id(os) == spa_bootfs(dmu_objset_spa(os)));
307 dmu_objset_rele(os, FTAG);
314 * Return non-zero if the spa version is less than requested version.
317 zfs_earlier_version(const char *name, int version)
321 if (spa_open(name, &spa, FTAG) == 0) {
322 if (spa_version(spa) < version) {
323 spa_close(spa, FTAG);
326 spa_close(spa, FTAG);
332 * Return TRUE if the ZPL version is less than requested version.
335 zpl_earlier_version(const char *name, int version)
338 boolean_t rc = B_TRUE;
340 if (dmu_objset_hold(name, FTAG, &os) == 0) {
343 if (dmu_objset_type(os) != DMU_OST_ZFS) {
344 dmu_objset_rele(os, FTAG);
347 /* XXX reading from non-owned objset */
348 if (zfs_get_zplprop(os, ZFS_PROP_VERSION, &zplversion) == 0)
349 rc = zplversion < version;
350 dmu_objset_rele(os, FTAG);
356 zfs_log_history(zfs_cmd_t *zc)
361 if ((buf = history_str_get(zc)) == NULL)
364 if (spa_open(zc->zc_name, &spa, FTAG) == 0) {
365 if (spa_version(spa) >= SPA_VERSION_ZPOOL_HISTORY)
366 (void) spa_history_log(spa, buf);
367 spa_close(spa, FTAG);
369 history_str_free(buf);
373 * Policy for top-level read operations (list pools). Requires no privileges,
374 * and can be used in the local zone, as there is no associated dataset.
378 zfs_secpolicy_none(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
384 * Policy for dataset read operations (list children, get statistics). Requires
385 * no privileges, but must be visible in the local zone.
389 zfs_secpolicy_read(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
391 if (INGLOBALZONE(curproc) ||
392 zone_dataset_visible(zc->zc_name, NULL))
395 return (SET_ERROR(ENOENT));
399 zfs_dozonecheck_impl(const char *dataset, uint64_t zoned, cred_t *cr)
404 * The dataset must be visible by this zone -- check this first
405 * so they don't see EPERM on something they shouldn't know about.
407 if (!INGLOBALZONE(curproc) &&
408 !zone_dataset_visible(dataset, &writable))
409 return (SET_ERROR(ENOENT));
411 if (INGLOBALZONE(curproc)) {
413 * If the fs is zoned, only root can access it from the
416 if (secpolicy_zfs(cr) && zoned)
417 return (SET_ERROR(EPERM));
420 * If we are in a local zone, the 'zoned' property must be set.
423 return (SET_ERROR(EPERM));
425 /* must be writable by this zone */
427 return (SET_ERROR(EPERM));
433 zfs_dozonecheck(const char *dataset, cred_t *cr)
437 if (dsl_prop_get_integer(dataset, "zoned", &zoned, NULL))
438 return (SET_ERROR(ENOENT));
440 return (zfs_dozonecheck_impl(dataset, zoned, cr));
444 zfs_dozonecheck_ds(const char *dataset, dsl_dataset_t *ds, cred_t *cr)
448 if (dsl_prop_get_int_ds(ds, "zoned", &zoned))
449 return (SET_ERROR(ENOENT));
451 return (zfs_dozonecheck_impl(dataset, zoned, cr));
455 zfs_secpolicy_write_perms_ds(const char *name, dsl_dataset_t *ds,
456 const char *perm, cred_t *cr)
460 error = zfs_dozonecheck_ds(name, ds, cr);
462 error = secpolicy_zfs(cr);
464 error = dsl_deleg_access_impl(ds, perm, cr);
470 zfs_secpolicy_write_perms(const char *name, const char *perm, cred_t *cr)
477 * First do a quick check for root in the global zone, which
478 * is allowed to do all write_perms. This ensures that zfs_ioc_*
479 * will get to handle nonexistent datasets.
481 if (INGLOBALZONE(curproc) && secpolicy_zfs(cr) == 0)
484 error = dsl_pool_hold(name, FTAG, &dp);
488 error = dsl_dataset_hold(dp, name, FTAG, &ds);
490 dsl_pool_rele(dp, FTAG);
494 error = zfs_secpolicy_write_perms_ds(name, ds, perm, cr);
496 dsl_dataset_rele(ds, FTAG);
497 dsl_pool_rele(dp, FTAG);
502 * Policy for setting the security label property.
504 * Returns 0 for success, non-zero for access and other errors.
507 zfs_set_slabel_policy(const char *name, char *strval, cred_t *cr)
510 char ds_hexsl[MAXNAMELEN];
511 bslabel_t ds_sl, new_sl;
512 boolean_t new_default = FALSE;
514 int needed_priv = -1;
517 /* First get the existing dataset label. */
518 error = dsl_prop_get(name, zfs_prop_to_name(ZFS_PROP_MLSLABEL),
519 1, sizeof (ds_hexsl), &ds_hexsl, NULL);
521 return (SET_ERROR(EPERM));
523 if (strcasecmp(strval, ZFS_MLSLABEL_DEFAULT) == 0)
526 /* The label must be translatable */
527 if (!new_default && (hexstr_to_label(strval, &new_sl) != 0))
528 return (SET_ERROR(EINVAL));
531 * In a non-global zone, disallow attempts to set a label that
532 * doesn't match that of the zone; otherwise no other checks
535 if (!INGLOBALZONE(curproc)) {
536 if (new_default || !blequal(&new_sl, CR_SL(CRED())))
537 return (SET_ERROR(EPERM));
542 * For global-zone datasets (i.e., those whose zoned property is
543 * "off", verify that the specified new label is valid for the
546 if (dsl_prop_get_integer(name,
547 zfs_prop_to_name(ZFS_PROP_ZONED), &zoned, NULL))
548 return (SET_ERROR(EPERM));
550 if (zfs_check_global_label(name, strval) != 0)
551 return (SET_ERROR(EPERM));
555 * If the existing dataset label is nondefault, check if the
556 * dataset is mounted (label cannot be changed while mounted).
557 * Get the zfsvfs_t; if there isn't one, then the dataset isn't
558 * mounted (or isn't a dataset, doesn't exist, ...).
560 if (strcasecmp(ds_hexsl, ZFS_MLSLABEL_DEFAULT) != 0) {
562 static char *setsl_tag = "setsl_tag";
565 * Try to own the dataset; abort if there is any error,
566 * (e.g., already mounted, in use, or other error).
568 error = dmu_objset_own(name, DMU_OST_ZFS, B_TRUE,
571 return (SET_ERROR(EPERM));
573 dmu_objset_disown(os, setsl_tag);
576 needed_priv = PRIV_FILE_DOWNGRADE_SL;
580 if (hexstr_to_label(strval, &new_sl) != 0)
581 return (SET_ERROR(EPERM));
583 if (blstrictdom(&ds_sl, &new_sl))
584 needed_priv = PRIV_FILE_DOWNGRADE_SL;
585 else if (blstrictdom(&new_sl, &ds_sl))
586 needed_priv = PRIV_FILE_UPGRADE_SL;
588 /* dataset currently has a default label */
590 needed_priv = PRIV_FILE_UPGRADE_SL;
594 if (needed_priv != -1)
595 return (PRIV_POLICY(cr, needed_priv, B_FALSE, EPERM, NULL));
599 #endif /* HAVE_MLSLABEL */
603 zfs_secpolicy_setprop(const char *dsname, zfs_prop_t prop, nvpair_t *propval,
609 * Check permissions for special properties.
616 * Disallow setting of 'zoned' from within a local zone.
618 if (!INGLOBALZONE(curproc))
619 return (SET_ERROR(EPERM));
623 case ZFS_PROP_FILESYSTEM_LIMIT:
624 case ZFS_PROP_SNAPSHOT_LIMIT:
625 if (!INGLOBALZONE(curproc)) {
627 char setpoint[ZFS_MAX_DATASET_NAME_LEN];
629 * Unprivileged users are allowed to modify the
630 * limit on things *under* (ie. contained by)
631 * the thing they own.
633 if (dsl_prop_get_integer(dsname, "zoned", &zoned,
635 return (SET_ERROR(EPERM));
636 if (!zoned || strlen(dsname) <= strlen(setpoint))
637 return (SET_ERROR(EPERM));
641 case ZFS_PROP_MLSLABEL:
642 if (!is_system_labeled())
643 return (SET_ERROR(EPERM));
645 if (nvpair_value_string(propval, &strval) == 0) {
648 err = zfs_set_slabel_policy(dsname, strval, CRED());
655 return (zfs_secpolicy_write_perms(dsname, zfs_prop_to_name(prop), cr));
660 zfs_secpolicy_set_fsacl(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
664 error = zfs_dozonecheck(zc->zc_name, cr);
669 * permission to set permissions will be evaluated later in
670 * dsl_deleg_can_allow()
677 zfs_secpolicy_rollback(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
679 return (zfs_secpolicy_write_perms(zc->zc_name,
680 ZFS_DELEG_PERM_ROLLBACK, cr));
685 zfs_secpolicy_send(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
693 * Generate the current snapshot name from the given objsetid, then
694 * use that name for the secpolicy/zone checks.
696 cp = strchr(zc->zc_name, '@');
698 return (SET_ERROR(EINVAL));
699 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
703 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &ds);
705 dsl_pool_rele(dp, FTAG);
709 dsl_dataset_name(ds, zc->zc_name);
711 error = zfs_secpolicy_write_perms_ds(zc->zc_name, ds,
712 ZFS_DELEG_PERM_SEND, cr);
713 dsl_dataset_rele(ds, FTAG);
714 dsl_pool_rele(dp, FTAG);
721 zfs_secpolicy_send_new(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
723 return (zfs_secpolicy_write_perms(zc->zc_name,
724 ZFS_DELEG_PERM_SEND, cr));
727 #ifdef HAVE_SMB_SHARE
730 zfs_secpolicy_deleg_share(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
735 if ((error = lookupname(zc->zc_value, UIO_SYSSPACE,
736 NO_FOLLOW, NULL, &vp)) != 0)
739 /* Now make sure mntpnt and dataset are ZFS */
741 if (vp->v_vfsp->vfs_fstype != zfsfstype ||
742 (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource),
743 zc->zc_name) != 0)) {
745 return (SET_ERROR(EPERM));
749 return (dsl_deleg_access(zc->zc_name,
750 ZFS_DELEG_PERM_SHARE, cr));
752 #endif /* HAVE_SMB_SHARE */
755 zfs_secpolicy_share(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
757 #ifdef HAVE_SMB_SHARE
758 if (!INGLOBALZONE(curproc))
759 return (SET_ERROR(EPERM));
761 if (secpolicy_nfs(cr) == 0) {
764 return (zfs_secpolicy_deleg_share(zc, innvl, cr));
767 return (SET_ERROR(ENOTSUP));
768 #endif /* HAVE_SMB_SHARE */
772 zfs_secpolicy_smb_acl(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
774 #ifdef HAVE_SMB_SHARE
775 if (!INGLOBALZONE(curproc))
776 return (SET_ERROR(EPERM));
778 if (secpolicy_smb(cr) == 0) {
781 return (zfs_secpolicy_deleg_share(zc, innvl, cr));
784 return (SET_ERROR(ENOTSUP));
785 #endif /* HAVE_SMB_SHARE */
789 zfs_get_parent(const char *datasetname, char *parent, int parentsize)
794 * Remove the @bla or /bla from the end of the name to get the parent.
796 (void) strncpy(parent, datasetname, parentsize);
797 cp = strrchr(parent, '@');
801 cp = strrchr(parent, '/');
803 return (SET_ERROR(ENOENT));
811 zfs_secpolicy_destroy_perms(const char *name, cred_t *cr)
815 if ((error = zfs_secpolicy_write_perms(name,
816 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
819 return (zfs_secpolicy_write_perms(name, ZFS_DELEG_PERM_DESTROY, cr));
824 zfs_secpolicy_destroy(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
826 return (zfs_secpolicy_destroy_perms(zc->zc_name, cr));
830 * Destroying snapshots with delegated permissions requires
831 * descendant mount and destroy permissions.
835 zfs_secpolicy_destroy_snaps(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
838 nvpair_t *pair, *nextpair;
841 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
842 return (SET_ERROR(EINVAL));
843 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
845 nextpair = nvlist_next_nvpair(snaps, pair);
846 error = zfs_secpolicy_destroy_perms(nvpair_name(pair), cr);
847 if (error == ENOENT) {
849 * Ignore any snapshots that don't exist (we consider
850 * them "already destroyed"). Remove the name from the
851 * nvl here in case the snapshot is created between
852 * now and when we try to destroy it (in which case
853 * we don't want to destroy it since we haven't
854 * checked for permission).
856 fnvlist_remove_nvpair(snaps, pair);
867 zfs_secpolicy_rename_perms(const char *from, const char *to, cred_t *cr)
869 char parentname[ZFS_MAX_DATASET_NAME_LEN];
872 if ((error = zfs_secpolicy_write_perms(from,
873 ZFS_DELEG_PERM_RENAME, cr)) != 0)
876 if ((error = zfs_secpolicy_write_perms(from,
877 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
880 if ((error = zfs_get_parent(to, parentname,
881 sizeof (parentname))) != 0)
884 if ((error = zfs_secpolicy_write_perms(parentname,
885 ZFS_DELEG_PERM_CREATE, cr)) != 0)
888 if ((error = zfs_secpolicy_write_perms(parentname,
889 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
897 zfs_secpolicy_rename(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
899 return (zfs_secpolicy_rename_perms(zc->zc_name, zc->zc_value, cr));
904 zfs_secpolicy_promote(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
907 dsl_dataset_t *clone;
910 error = zfs_secpolicy_write_perms(zc->zc_name,
911 ZFS_DELEG_PERM_PROMOTE, cr);
915 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
919 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &clone);
922 char parentname[ZFS_MAX_DATASET_NAME_LEN];
923 dsl_dataset_t *origin = NULL;
927 error = dsl_dataset_hold_obj(dd->dd_pool,
928 dsl_dir_phys(dd)->dd_origin_obj, FTAG, &origin);
930 dsl_dataset_rele(clone, FTAG);
931 dsl_pool_rele(dp, FTAG);
935 error = zfs_secpolicy_write_perms_ds(zc->zc_name, clone,
936 ZFS_DELEG_PERM_MOUNT, cr);
938 dsl_dataset_name(origin, parentname);
940 error = zfs_secpolicy_write_perms_ds(parentname, origin,
941 ZFS_DELEG_PERM_PROMOTE, cr);
943 dsl_dataset_rele(clone, FTAG);
944 dsl_dataset_rele(origin, FTAG);
946 dsl_pool_rele(dp, FTAG);
952 zfs_secpolicy_recv(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
956 if ((error = zfs_secpolicy_write_perms(zc->zc_name,
957 ZFS_DELEG_PERM_RECEIVE, cr)) != 0)
960 if ((error = zfs_secpolicy_write_perms(zc->zc_name,
961 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
964 return (zfs_secpolicy_write_perms(zc->zc_name,
965 ZFS_DELEG_PERM_CREATE, cr));
970 zfs_secpolicy_recv_new(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
972 return (zfs_secpolicy_recv(zc, innvl, cr));
976 zfs_secpolicy_snapshot_perms(const char *name, cred_t *cr)
978 return (zfs_secpolicy_write_perms(name,
979 ZFS_DELEG_PERM_SNAPSHOT, cr));
983 * Check for permission to create each snapshot in the nvlist.
987 zfs_secpolicy_snapshot(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
993 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
994 return (SET_ERROR(EINVAL));
995 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
996 pair = nvlist_next_nvpair(snaps, pair)) {
997 char *name = nvpair_name(pair);
998 char *atp = strchr(name, '@');
1001 error = SET_ERROR(EINVAL);
1005 error = zfs_secpolicy_snapshot_perms(name, cr);
1014 * Check for permission to create each snapshot in the nvlist.
1018 zfs_secpolicy_bookmark(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1023 for (pair = nvlist_next_nvpair(innvl, NULL);
1024 pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) {
1025 char *name = nvpair_name(pair);
1026 char *hashp = strchr(name, '#');
1028 if (hashp == NULL) {
1029 error = SET_ERROR(EINVAL);
1033 error = zfs_secpolicy_write_perms(name,
1034 ZFS_DELEG_PERM_BOOKMARK, cr);
1044 zfs_secpolicy_destroy_bookmarks(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1046 nvpair_t *pair, *nextpair;
1049 for (pair = nvlist_next_nvpair(innvl, NULL); pair != NULL;
1051 char *name = nvpair_name(pair);
1052 char *hashp = strchr(name, '#');
1053 nextpair = nvlist_next_nvpair(innvl, pair);
1055 if (hashp == NULL) {
1056 error = SET_ERROR(EINVAL);
1061 error = zfs_secpolicy_write_perms(name,
1062 ZFS_DELEG_PERM_DESTROY, cr);
1064 if (error == ENOENT) {
1066 * Ignore any filesystems that don't exist (we consider
1067 * their bookmarks "already destroyed"). Remove
1068 * the name from the nvl here in case the filesystem
1069 * is created between now and when we try to destroy
1070 * the bookmark (in which case we don't want to
1071 * destroy it since we haven't checked for permission).
1073 fnvlist_remove_nvpair(innvl, pair);
1085 zfs_secpolicy_log_history(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1088 * Even root must have a proper TSD so that we know what pool
1091 if (tsd_get(zfs_allow_log_key) == NULL)
1092 return (SET_ERROR(EPERM));
1097 zfs_secpolicy_create_clone(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1099 char parentname[ZFS_MAX_DATASET_NAME_LEN];
1103 if ((error = zfs_get_parent(zc->zc_name, parentname,
1104 sizeof (parentname))) != 0)
1107 if (nvlist_lookup_string(innvl, "origin", &origin) == 0 &&
1108 (error = zfs_secpolicy_write_perms(origin,
1109 ZFS_DELEG_PERM_CLONE, cr)) != 0)
1112 if ((error = zfs_secpolicy_write_perms(parentname,
1113 ZFS_DELEG_PERM_CREATE, cr)) != 0)
1116 return (zfs_secpolicy_write_perms(parentname,
1117 ZFS_DELEG_PERM_MOUNT, cr));
1121 * Policy for pool operations - create/destroy pools, add vdevs, etc. Requires
1122 * SYS_CONFIG privilege, which is not available in a local zone.
1126 zfs_secpolicy_config(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1128 if (secpolicy_sys_config(cr, B_FALSE) != 0)
1129 return (SET_ERROR(EPERM));
1135 * Policy for object to name lookups.
1139 zfs_secpolicy_diff(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1143 if ((error = secpolicy_sys_config(cr, B_FALSE)) == 0)
1146 error = zfs_secpolicy_write_perms(zc->zc_name, ZFS_DELEG_PERM_DIFF, cr);
1151 * Policy for fault injection. Requires all privileges.
1155 zfs_secpolicy_inject(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1157 return (secpolicy_zinject(cr));
1162 zfs_secpolicy_inherit_prop(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1164 zfs_prop_t prop = zfs_name_to_prop(zc->zc_value);
1166 if (prop == ZPROP_INVAL) {
1167 if (!zfs_prop_user(zc->zc_value))
1168 return (SET_ERROR(EINVAL));
1169 return (zfs_secpolicy_write_perms(zc->zc_name,
1170 ZFS_DELEG_PERM_USERPROP, cr));
1172 return (zfs_secpolicy_setprop(zc->zc_name, prop,
1178 zfs_secpolicy_userspace_one(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1180 int err = zfs_secpolicy_read(zc, innvl, cr);
1184 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
1185 return (SET_ERROR(EINVAL));
1187 if (zc->zc_value[0] == 0) {
1189 * They are asking about a posix uid/gid. If it's
1190 * themself, allow it.
1192 if (zc->zc_objset_type == ZFS_PROP_USERUSED ||
1193 zc->zc_objset_type == ZFS_PROP_USERQUOTA ||
1194 zc->zc_objset_type == ZFS_PROP_USEROBJUSED ||
1195 zc->zc_objset_type == ZFS_PROP_USEROBJQUOTA) {
1196 if (zc->zc_guid == crgetuid(cr))
1199 if (groupmember(zc->zc_guid, cr))
1204 return (zfs_secpolicy_write_perms(zc->zc_name,
1205 userquota_perms[zc->zc_objset_type], cr));
1209 zfs_secpolicy_userspace_many(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1211 int err = zfs_secpolicy_read(zc, innvl, cr);
1215 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
1216 return (SET_ERROR(EINVAL));
1218 return (zfs_secpolicy_write_perms(zc->zc_name,
1219 userquota_perms[zc->zc_objset_type], cr));
1224 zfs_secpolicy_userspace_upgrade(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1226 return (zfs_secpolicy_setprop(zc->zc_name, ZFS_PROP_VERSION,
1232 zfs_secpolicy_hold(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1238 error = nvlist_lookup_nvlist(innvl, "holds", &holds);
1240 return (SET_ERROR(EINVAL));
1242 for (pair = nvlist_next_nvpair(holds, NULL); pair != NULL;
1243 pair = nvlist_next_nvpair(holds, pair)) {
1244 char fsname[ZFS_MAX_DATASET_NAME_LEN];
1245 error = dmu_fsname(nvpair_name(pair), fsname);
1248 error = zfs_secpolicy_write_perms(fsname,
1249 ZFS_DELEG_PERM_HOLD, cr);
1258 zfs_secpolicy_release(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1263 for (pair = nvlist_next_nvpair(innvl, NULL); pair != NULL;
1264 pair = nvlist_next_nvpair(innvl, pair)) {
1265 char fsname[ZFS_MAX_DATASET_NAME_LEN];
1266 error = dmu_fsname(nvpair_name(pair), fsname);
1269 error = zfs_secpolicy_write_perms(fsname,
1270 ZFS_DELEG_PERM_RELEASE, cr);
1278 * Policy for allowing temporary snapshots to be taken or released
1281 zfs_secpolicy_tmp_snapshot(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1284 * A temporary snapshot is the same as a snapshot,
1285 * hold, destroy and release all rolled into one.
1286 * Delegated diff alone is sufficient that we allow this.
1290 if ((error = zfs_secpolicy_write_perms(zc->zc_name,
1291 ZFS_DELEG_PERM_DIFF, cr)) == 0)
1294 error = zfs_secpolicy_snapshot_perms(zc->zc_name, cr);
1296 error = zfs_secpolicy_hold(zc, innvl, cr);
1298 error = zfs_secpolicy_release(zc, innvl, cr);
1300 error = zfs_secpolicy_destroy(zc, innvl, cr);
1305 * Returns the nvlist as specified by the user in the zfs_cmd_t.
1308 get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp)
1312 nvlist_t *list = NULL;
1315 * Read in and unpack the user-supplied nvlist.
1318 return (SET_ERROR(EINVAL));
1320 packed = vmem_alloc(size, KM_SLEEP);
1322 if ((error = ddi_copyin((void *)(uintptr_t)nvl, packed, size,
1324 vmem_free(packed, size);
1325 return (SET_ERROR(EFAULT));
1328 if ((error = nvlist_unpack(packed, size, &list, 0)) != 0) {
1329 vmem_free(packed, size);
1333 vmem_free(packed, size);
1340 * Reduce the size of this nvlist until it can be serialized in 'max' bytes.
1341 * Entries will be removed from the end of the nvlist, and one int32 entry
1342 * named "N_MORE_ERRORS" will be added indicating how many entries were
1346 nvlist_smush(nvlist_t *errors, size_t max)
1350 size = fnvlist_size(errors);
1353 nvpair_t *more_errors;
1357 return (SET_ERROR(ENOMEM));
1359 fnvlist_add_int32(errors, ZPROP_N_MORE_ERRORS, 0);
1360 more_errors = nvlist_prev_nvpair(errors, NULL);
1363 nvpair_t *pair = nvlist_prev_nvpair(errors,
1365 fnvlist_remove_nvpair(errors, pair);
1367 size = fnvlist_size(errors);
1368 } while (size > max);
1370 fnvlist_remove_nvpair(errors, more_errors);
1371 fnvlist_add_int32(errors, ZPROP_N_MORE_ERRORS, n);
1372 ASSERT3U(fnvlist_size(errors), <=, max);
1379 put_nvlist(zfs_cmd_t *zc, nvlist_t *nvl)
1381 char *packed = NULL;
1385 size = fnvlist_size(nvl);
1387 if (size > zc->zc_nvlist_dst_size) {
1388 error = SET_ERROR(ENOMEM);
1390 packed = fnvlist_pack(nvl, &size);
1391 if (ddi_copyout(packed, (void *)(uintptr_t)zc->zc_nvlist_dst,
1392 size, zc->zc_iflags) != 0)
1393 error = SET_ERROR(EFAULT);
1394 fnvlist_pack_free(packed, size);
1397 zc->zc_nvlist_dst_size = size;
1398 zc->zc_nvlist_dst_filled = B_TRUE;
1403 getzfsvfs(const char *dsname, zfsvfs_t **zfvp)
1408 error = dmu_objset_hold(dsname, FTAG, &os);
1411 if (dmu_objset_type(os) != DMU_OST_ZFS) {
1412 dmu_objset_rele(os, FTAG);
1413 return (SET_ERROR(EINVAL));
1416 mutex_enter(&os->os_user_ptr_lock);
1417 *zfvp = dmu_objset_get_user(os);
1418 /* bump s_active only when non-zero to prevent umount race */
1419 if (*zfvp == NULL || (*zfvp)->z_sb == NULL ||
1420 !atomic_inc_not_zero(&((*zfvp)->z_sb->s_active))) {
1421 error = SET_ERROR(ESRCH);
1423 mutex_exit(&os->os_user_ptr_lock);
1424 dmu_objset_rele(os, FTAG);
1429 * Find a zfsvfs_t for a mounted filesystem, or create our own, in which
1430 * case its z_sb will be NULL, and it will be opened as the owner.
1431 * If 'writer' is set, the z_teardown_lock will be held for RW_WRITER,
1432 * which prevents all inode ops from running.
1435 zfsvfs_hold(const char *name, void *tag, zfsvfs_t **zfvp, boolean_t writer)
1439 if (getzfsvfs(name, zfvp) != 0)
1440 error = zfsvfs_create(name, zfvp);
1442 rrm_enter(&(*zfvp)->z_teardown_lock, (writer) ? RW_WRITER :
1444 if ((*zfvp)->z_unmounted) {
1446 * XXX we could probably try again, since the unmounting
1447 * thread should be just about to disassociate the
1448 * objset from the zfsvfs.
1450 rrm_exit(&(*zfvp)->z_teardown_lock, tag);
1451 return (SET_ERROR(EBUSY));
1458 zfsvfs_rele(zfsvfs_t *zfsvfs, void *tag)
1460 rrm_exit(&zfsvfs->z_teardown_lock, tag);
1463 deactivate_super(zfsvfs->z_sb);
1465 dmu_objset_disown(zfsvfs->z_os, zfsvfs);
1466 zfsvfs_free(zfsvfs);
1471 zfs_ioc_pool_create(zfs_cmd_t *zc)
1474 nvlist_t *config, *props = NULL;
1475 nvlist_t *rootprops = NULL;
1476 nvlist_t *zplprops = NULL;
1478 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1479 zc->zc_iflags, &config)))
1482 if (zc->zc_nvlist_src_size != 0 && (error =
1483 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1484 zc->zc_iflags, &props))) {
1485 nvlist_free(config);
1490 nvlist_t *nvl = NULL;
1491 uint64_t version = SPA_VERSION;
1493 (void) nvlist_lookup_uint64(props,
1494 zpool_prop_to_name(ZPOOL_PROP_VERSION), &version);
1495 if (!SPA_VERSION_IS_SUPPORTED(version)) {
1496 error = SET_ERROR(EINVAL);
1497 goto pool_props_bad;
1499 (void) nvlist_lookup_nvlist(props, ZPOOL_ROOTFS_PROPS, &nvl);
1501 error = nvlist_dup(nvl, &rootprops, KM_SLEEP);
1503 nvlist_free(config);
1507 (void) nvlist_remove_all(props, ZPOOL_ROOTFS_PROPS);
1509 VERIFY(nvlist_alloc(&zplprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
1510 error = zfs_fill_zplprops_root(version, rootprops,
1513 goto pool_props_bad;
1516 error = spa_create(zc->zc_name, config, props, zplprops);
1519 * Set the remaining root properties
1521 if (!error && (error = zfs_set_prop_nvlist(zc->zc_name,
1522 ZPROP_SRC_LOCAL, rootprops, NULL)) != 0)
1523 (void) spa_destroy(zc->zc_name);
1526 nvlist_free(rootprops);
1527 nvlist_free(zplprops);
1528 nvlist_free(config);
1535 zfs_ioc_pool_destroy(zfs_cmd_t *zc)
1538 zfs_log_history(zc);
1539 error = spa_destroy(zc->zc_name);
1545 zfs_ioc_pool_import(zfs_cmd_t *zc)
1547 nvlist_t *config, *props = NULL;
1551 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1552 zc->zc_iflags, &config)) != 0)
1555 if (zc->zc_nvlist_src_size != 0 && (error =
1556 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1557 zc->zc_iflags, &props))) {
1558 nvlist_free(config);
1562 if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, &guid) != 0 ||
1563 guid != zc->zc_guid)
1564 error = SET_ERROR(EINVAL);
1566 error = spa_import(zc->zc_name, config, props, zc->zc_cookie);
1568 if (zc->zc_nvlist_dst != 0) {
1571 if ((err = put_nvlist(zc, config)) != 0)
1575 nvlist_free(config);
1582 zfs_ioc_pool_export(zfs_cmd_t *zc)
1585 boolean_t force = (boolean_t)zc->zc_cookie;
1586 boolean_t hardforce = (boolean_t)zc->zc_guid;
1588 zfs_log_history(zc);
1589 error = spa_export(zc->zc_name, NULL, force, hardforce);
1595 zfs_ioc_pool_configs(zfs_cmd_t *zc)
1600 if ((configs = spa_all_configs(&zc->zc_cookie)) == NULL)
1601 return (SET_ERROR(EEXIST));
1603 error = put_nvlist(zc, configs);
1605 nvlist_free(configs);
1612 * zc_name name of the pool
1615 * zc_cookie real errno
1616 * zc_nvlist_dst config nvlist
1617 * zc_nvlist_dst_size size of config nvlist
1620 zfs_ioc_pool_stats(zfs_cmd_t *zc)
1626 error = spa_get_stats(zc->zc_name, &config, zc->zc_value,
1627 sizeof (zc->zc_value));
1629 if (config != NULL) {
1630 ret = put_nvlist(zc, config);
1631 nvlist_free(config);
1634 * The config may be present even if 'error' is non-zero.
1635 * In this case we return success, and preserve the real errno
1638 zc->zc_cookie = error;
1647 * Try to import the given pool, returning pool stats as appropriate so that
1648 * user land knows which devices are available and overall pool health.
1651 zfs_ioc_pool_tryimport(zfs_cmd_t *zc)
1653 nvlist_t *tryconfig, *config = NULL;
1656 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1657 zc->zc_iflags, &tryconfig)) != 0)
1660 config = spa_tryimport(tryconfig);
1662 nvlist_free(tryconfig);
1665 return (SET_ERROR(EINVAL));
1667 error = put_nvlist(zc, config);
1668 nvlist_free(config);
1675 * zc_name name of the pool
1676 * zc_cookie scan func (pool_scan_func_t)
1677 * zc_flags scrub pause/resume flag (pool_scrub_cmd_t)
1680 zfs_ioc_pool_scan(zfs_cmd_t *zc)
1685 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1688 if (zc->zc_flags >= POOL_SCRUB_FLAGS_END)
1689 return (SET_ERROR(EINVAL));
1691 if (zc->zc_flags == POOL_SCRUB_PAUSE)
1692 error = spa_scrub_pause_resume(spa, POOL_SCRUB_PAUSE);
1693 else if (zc->zc_cookie == POOL_SCAN_NONE)
1694 error = spa_scan_stop(spa);
1696 error = spa_scan(spa, zc->zc_cookie);
1698 spa_close(spa, FTAG);
1704 zfs_ioc_pool_freeze(zfs_cmd_t *zc)
1709 error = spa_open(zc->zc_name, &spa, FTAG);
1712 spa_close(spa, FTAG);
1718 zfs_ioc_pool_upgrade(zfs_cmd_t *zc)
1723 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1726 if (zc->zc_cookie < spa_version(spa) ||
1727 !SPA_VERSION_IS_SUPPORTED(zc->zc_cookie)) {
1728 spa_close(spa, FTAG);
1729 return (SET_ERROR(EINVAL));
1732 spa_upgrade(spa, zc->zc_cookie);
1733 spa_close(spa, FTAG);
1739 zfs_ioc_pool_get_history(zfs_cmd_t *zc)
1746 if ((size = zc->zc_history_len) == 0)
1747 return (SET_ERROR(EINVAL));
1749 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1752 if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) {
1753 spa_close(spa, FTAG);
1754 return (SET_ERROR(ENOTSUP));
1757 hist_buf = vmem_alloc(size, KM_SLEEP);
1758 if ((error = spa_history_get(spa, &zc->zc_history_offset,
1759 &zc->zc_history_len, hist_buf)) == 0) {
1760 error = ddi_copyout(hist_buf,
1761 (void *)(uintptr_t)zc->zc_history,
1762 zc->zc_history_len, zc->zc_iflags);
1765 spa_close(spa, FTAG);
1766 vmem_free(hist_buf, size);
1771 zfs_ioc_pool_reguid(zfs_cmd_t *zc)
1776 error = spa_open(zc->zc_name, &spa, FTAG);
1778 error = spa_change_guid(spa);
1779 spa_close(spa, FTAG);
1785 zfs_ioc_dsobj_to_dsname(zfs_cmd_t *zc)
1787 return (dsl_dsobj_to_dsname(zc->zc_name, zc->zc_obj, zc->zc_value));
1792 * zc_name name of filesystem
1793 * zc_obj object to find
1796 * zc_value name of object
1799 zfs_ioc_obj_to_path(zfs_cmd_t *zc)
1804 /* XXX reading from objset not owned */
1805 if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os)) != 0)
1807 if (dmu_objset_type(os) != DMU_OST_ZFS) {
1808 dmu_objset_rele(os, FTAG);
1809 return (SET_ERROR(EINVAL));
1811 error = zfs_obj_to_path(os, zc->zc_obj, zc->zc_value,
1812 sizeof (zc->zc_value));
1813 dmu_objset_rele(os, FTAG);
1820 * zc_name name of filesystem
1821 * zc_obj object to find
1824 * zc_stat stats on object
1825 * zc_value path to object
1828 zfs_ioc_obj_to_stats(zfs_cmd_t *zc)
1833 /* XXX reading from objset not owned */
1834 if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os)) != 0)
1836 if (dmu_objset_type(os) != DMU_OST_ZFS) {
1837 dmu_objset_rele(os, FTAG);
1838 return (SET_ERROR(EINVAL));
1840 error = zfs_obj_to_stats(os, zc->zc_obj, &zc->zc_stat, zc->zc_value,
1841 sizeof (zc->zc_value));
1842 dmu_objset_rele(os, FTAG);
1848 zfs_ioc_vdev_add(zfs_cmd_t *zc)
1854 error = spa_open(zc->zc_name, &spa, FTAG);
1858 error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1859 zc->zc_iflags, &config);
1861 error = spa_vdev_add(spa, config);
1862 nvlist_free(config);
1864 spa_close(spa, FTAG);
1870 * zc_name name of the pool
1871 * zc_nvlist_conf nvlist of devices to remove
1872 * zc_cookie to stop the remove?
1875 zfs_ioc_vdev_remove(zfs_cmd_t *zc)
1880 error = spa_open(zc->zc_name, &spa, FTAG);
1883 error = spa_vdev_remove(spa, zc->zc_guid, B_FALSE);
1884 spa_close(spa, FTAG);
1889 zfs_ioc_vdev_set_state(zfs_cmd_t *zc)
1893 vdev_state_t newstate = VDEV_STATE_UNKNOWN;
1895 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1897 switch (zc->zc_cookie) {
1898 case VDEV_STATE_ONLINE:
1899 error = vdev_online(spa, zc->zc_guid, zc->zc_obj, &newstate);
1902 case VDEV_STATE_OFFLINE:
1903 error = vdev_offline(spa, zc->zc_guid, zc->zc_obj);
1906 case VDEV_STATE_FAULTED:
1907 if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED &&
1908 zc->zc_obj != VDEV_AUX_EXTERNAL &&
1909 zc->zc_obj != VDEV_AUX_EXTERNAL_PERSIST)
1910 zc->zc_obj = VDEV_AUX_ERR_EXCEEDED;
1912 error = vdev_fault(spa, zc->zc_guid, zc->zc_obj);
1915 case VDEV_STATE_DEGRADED:
1916 if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED &&
1917 zc->zc_obj != VDEV_AUX_EXTERNAL)
1918 zc->zc_obj = VDEV_AUX_ERR_EXCEEDED;
1920 error = vdev_degrade(spa, zc->zc_guid, zc->zc_obj);
1924 error = SET_ERROR(EINVAL);
1926 zc->zc_cookie = newstate;
1927 spa_close(spa, FTAG);
1932 zfs_ioc_vdev_attach(zfs_cmd_t *zc)
1935 int replacing = zc->zc_cookie;
1939 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1942 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1943 zc->zc_iflags, &config)) == 0) {
1944 error = spa_vdev_attach(spa, zc->zc_guid, config, replacing);
1945 nvlist_free(config);
1948 spa_close(spa, FTAG);
1953 zfs_ioc_vdev_detach(zfs_cmd_t *zc)
1958 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1961 error = spa_vdev_detach(spa, zc->zc_guid, 0, B_FALSE);
1963 spa_close(spa, FTAG);
1968 zfs_ioc_vdev_split(zfs_cmd_t *zc)
1971 nvlist_t *config, *props = NULL;
1973 boolean_t exp = !!(zc->zc_cookie & ZPOOL_EXPORT_AFTER_SPLIT);
1975 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1978 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1979 zc->zc_iflags, &config))) {
1980 spa_close(spa, FTAG);
1984 if (zc->zc_nvlist_src_size != 0 && (error =
1985 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1986 zc->zc_iflags, &props))) {
1987 spa_close(spa, FTAG);
1988 nvlist_free(config);
1992 error = spa_vdev_split_mirror(spa, zc->zc_string, config, props, exp);
1994 spa_close(spa, FTAG);
1996 nvlist_free(config);
2003 zfs_ioc_vdev_setpath(zfs_cmd_t *zc)
2006 char *path = zc->zc_value;
2007 uint64_t guid = zc->zc_guid;
2010 error = spa_open(zc->zc_name, &spa, FTAG);
2014 error = spa_vdev_setpath(spa, guid, path);
2015 spa_close(spa, FTAG);
2020 zfs_ioc_vdev_setfru(zfs_cmd_t *zc)
2023 char *fru = zc->zc_value;
2024 uint64_t guid = zc->zc_guid;
2027 error = spa_open(zc->zc_name, &spa, FTAG);
2031 error = spa_vdev_setfru(spa, guid, fru);
2032 spa_close(spa, FTAG);
2037 zfs_ioc_objset_stats_impl(zfs_cmd_t *zc, objset_t *os)
2042 dmu_objset_fast_stat(os, &zc->zc_objset_stats);
2044 if (zc->zc_nvlist_dst != 0 &&
2045 (error = dsl_prop_get_all(os, &nv)) == 0) {
2046 dmu_objset_stats(os, nv);
2048 * NB: zvol_get_stats() will read the objset contents,
2049 * which we aren't supposed to do with a
2050 * DS_MODE_USER hold, because it could be
2051 * inconsistent. So this is a bit of a workaround...
2052 * XXX reading with out owning
2054 if (!zc->zc_objset_stats.dds_inconsistent &&
2055 dmu_objset_type(os) == DMU_OST_ZVOL) {
2056 error = zvol_get_stats(os, nv);
2064 error = put_nvlist(zc, nv);
2073 * zc_name name of filesystem
2074 * zc_nvlist_dst_size size of buffer for property nvlist
2077 * zc_objset_stats stats
2078 * zc_nvlist_dst property nvlist
2079 * zc_nvlist_dst_size size of property nvlist
2082 zfs_ioc_objset_stats(zfs_cmd_t *zc)
2087 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
2089 error = zfs_ioc_objset_stats_impl(zc, os);
2090 dmu_objset_rele(os, FTAG);
2098 * zc_name name of filesystem
2099 * zc_nvlist_dst_size size of buffer for property nvlist
2102 * zc_nvlist_dst received property nvlist
2103 * zc_nvlist_dst_size size of received property nvlist
2105 * Gets received properties (distinct from local properties on or after
2106 * SPA_VERSION_RECVD_PROPS) for callers who want to differentiate received from
2107 * local property values.
2110 zfs_ioc_objset_recvd_props(zfs_cmd_t *zc)
2116 * Without this check, we would return local property values if the
2117 * caller has not already received properties on or after
2118 * SPA_VERSION_RECVD_PROPS.
2120 if (!dsl_prop_get_hasrecvd(zc->zc_name))
2121 return (SET_ERROR(ENOTSUP));
2123 if (zc->zc_nvlist_dst != 0 &&
2124 (error = dsl_prop_get_received(zc->zc_name, &nv)) == 0) {
2125 error = put_nvlist(zc, nv);
2133 nvl_add_zplprop(objset_t *os, nvlist_t *props, zfs_prop_t prop)
2139 * zfs_get_zplprop() will either find a value or give us
2140 * the default value (if there is one).
2142 if ((error = zfs_get_zplprop(os, prop, &value)) != 0)
2144 VERIFY(nvlist_add_uint64(props, zfs_prop_to_name(prop), value) == 0);
2150 * zc_name name of filesystem
2151 * zc_nvlist_dst_size size of buffer for zpl property nvlist
2154 * zc_nvlist_dst zpl property nvlist
2155 * zc_nvlist_dst_size size of zpl property nvlist
2158 zfs_ioc_objset_zplprops(zfs_cmd_t *zc)
2163 /* XXX reading without owning */
2164 if ((err = dmu_objset_hold(zc->zc_name, FTAG, &os)))
2167 dmu_objset_fast_stat(os, &zc->zc_objset_stats);
2170 * NB: nvl_add_zplprop() will read the objset contents,
2171 * which we aren't supposed to do with a DS_MODE_USER
2172 * hold, because it could be inconsistent.
2174 if (zc->zc_nvlist_dst != 0 &&
2175 !zc->zc_objset_stats.dds_inconsistent &&
2176 dmu_objset_type(os) == DMU_OST_ZFS) {
2179 VERIFY(nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2180 if ((err = nvl_add_zplprop(os, nv, ZFS_PROP_VERSION)) == 0 &&
2181 (err = nvl_add_zplprop(os, nv, ZFS_PROP_NORMALIZE)) == 0 &&
2182 (err = nvl_add_zplprop(os, nv, ZFS_PROP_UTF8ONLY)) == 0 &&
2183 (err = nvl_add_zplprop(os, nv, ZFS_PROP_CASE)) == 0)
2184 err = put_nvlist(zc, nv);
2187 err = SET_ERROR(ENOENT);
2189 dmu_objset_rele(os, FTAG);
2194 dataset_name_hidden(const char *name)
2197 * Skip over datasets that are not visible in this zone,
2198 * internal datasets (which have a $ in their name), and
2199 * temporary datasets (which have a % in their name).
2201 if (strchr(name, '$') != NULL)
2203 if (strchr(name, '%') != NULL)
2205 if (!INGLOBALZONE(curproc) && !zone_dataset_visible(name, NULL))
2212 * zc_name name of filesystem
2213 * zc_cookie zap cursor
2214 * zc_nvlist_dst_size size of buffer for property nvlist
2217 * zc_name name of next filesystem
2218 * zc_cookie zap cursor
2219 * zc_objset_stats stats
2220 * zc_nvlist_dst property nvlist
2221 * zc_nvlist_dst_size size of property nvlist
2224 zfs_ioc_dataset_list_next(zfs_cmd_t *zc)
2229 size_t orig_len = strlen(zc->zc_name);
2232 if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os))) {
2233 if (error == ENOENT)
2234 error = SET_ERROR(ESRCH);
2238 p = strrchr(zc->zc_name, '/');
2239 if (p == NULL || p[1] != '\0')
2240 (void) strlcat(zc->zc_name, "/", sizeof (zc->zc_name));
2241 p = zc->zc_name + strlen(zc->zc_name);
2244 error = dmu_dir_list_next(os,
2245 sizeof (zc->zc_name) - (p - zc->zc_name), p,
2246 NULL, &zc->zc_cookie);
2247 if (error == ENOENT)
2248 error = SET_ERROR(ESRCH);
2249 } while (error == 0 && dataset_name_hidden(zc->zc_name));
2250 dmu_objset_rele(os, FTAG);
2253 * If it's an internal dataset (ie. with a '$' in its name),
2254 * don't try to get stats for it, otherwise we'll return ENOENT.
2256 if (error == 0 && strchr(zc->zc_name, '$') == NULL) {
2257 error = zfs_ioc_objset_stats(zc); /* fill in the stats */
2258 if (error == ENOENT) {
2259 /* We lost a race with destroy, get the next one. */
2260 zc->zc_name[orig_len] = '\0';
2269 * zc_name name of filesystem
2270 * zc_cookie zap cursor
2271 * zc_nvlist_dst_size size of buffer for property nvlist
2274 * zc_name name of next snapshot
2275 * zc_objset_stats stats
2276 * zc_nvlist_dst property nvlist
2277 * zc_nvlist_dst_size size of property nvlist
2280 zfs_ioc_snapshot_list_next(zfs_cmd_t *zc)
2285 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
2287 return (error == ENOENT ? ESRCH : error);
2291 * A dataset name of maximum length cannot have any snapshots,
2292 * so exit immediately.
2294 if (strlcat(zc->zc_name, "@", sizeof (zc->zc_name)) >=
2295 ZFS_MAX_DATASET_NAME_LEN) {
2296 dmu_objset_rele(os, FTAG);
2297 return (SET_ERROR(ESRCH));
2300 error = dmu_snapshot_list_next(os,
2301 sizeof (zc->zc_name) - strlen(zc->zc_name),
2302 zc->zc_name + strlen(zc->zc_name), &zc->zc_obj, &zc->zc_cookie,
2305 if (error == 0 && !zc->zc_simple) {
2307 dsl_pool_t *dp = os->os_dsl_dataset->ds_dir->dd_pool;
2309 error = dsl_dataset_hold_obj(dp, zc->zc_obj, FTAG, &ds);
2313 error = dmu_objset_from_ds(ds, &ossnap);
2315 error = zfs_ioc_objset_stats_impl(zc, ossnap);
2316 dsl_dataset_rele(ds, FTAG);
2318 } else if (error == ENOENT) {
2319 error = SET_ERROR(ESRCH);
2322 dmu_objset_rele(os, FTAG);
2323 /* if we failed, undo the @ that we tacked on to zc_name */
2325 *strchr(zc->zc_name, '@') = '\0';
2330 zfs_prop_set_userquota(const char *dsname, nvpair_t *pair)
2332 const char *propname = nvpair_name(pair);
2334 unsigned int vallen;
2337 zfs_userquota_prop_t type;
2343 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2345 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
2346 if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2348 return (SET_ERROR(EINVAL));
2352 * A correctly constructed propname is encoded as
2353 * userquota@<rid>-<domain>.
2355 if ((dash = strchr(propname, '-')) == NULL ||
2356 nvpair_value_uint64_array(pair, &valary, &vallen) != 0 ||
2358 return (SET_ERROR(EINVAL));
2365 err = zfsvfs_hold(dsname, FTAG, &zfsvfs, B_FALSE);
2367 err = zfs_set_userquota(zfsvfs, type, domain, rid, quota);
2368 zfsvfs_rele(zfsvfs, FTAG);
2375 * If the named property is one that has a special function to set its value,
2376 * return 0 on success and a positive error code on failure; otherwise if it is
2377 * not one of the special properties handled by this function, return -1.
2379 * XXX: It would be better for callers of the property interface if we handled
2380 * these special cases in dsl_prop.c (in the dsl layer).
2383 zfs_prop_set_special(const char *dsname, zprop_source_t source,
2386 const char *propname = nvpair_name(pair);
2387 zfs_prop_t prop = zfs_name_to_prop(propname);
2391 if (prop == ZPROP_INVAL) {
2392 if (zfs_prop_userquota(propname))
2393 return (zfs_prop_set_userquota(dsname, pair));
2397 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2399 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
2400 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2404 if (zfs_prop_get_type(prop) == PROP_TYPE_STRING)
2407 VERIFY(0 == nvpair_value_uint64(pair, &intval));
2410 case ZFS_PROP_QUOTA:
2411 err = dsl_dir_set_quota(dsname, source, intval);
2413 case ZFS_PROP_REFQUOTA:
2414 err = dsl_dataset_set_refquota(dsname, source, intval);
2416 case ZFS_PROP_FILESYSTEM_LIMIT:
2417 case ZFS_PROP_SNAPSHOT_LIMIT:
2418 if (intval == UINT64_MAX) {
2419 /* clearing the limit, just do it */
2422 err = dsl_dir_activate_fs_ss_limit(dsname);
2425 * Set err to -1 to force the zfs_set_prop_nvlist code down the
2426 * default path to set the value in the nvlist.
2431 case ZFS_PROP_RESERVATION:
2432 err = dsl_dir_set_reservation(dsname, source, intval);
2434 case ZFS_PROP_REFRESERVATION:
2435 err = dsl_dataset_set_refreservation(dsname, source, intval);
2437 case ZFS_PROP_VOLSIZE:
2438 err = zvol_set_volsize(dsname, intval);
2440 case ZFS_PROP_SNAPDEV:
2441 err = zvol_set_snapdev(dsname, source, intval);
2443 case ZFS_PROP_VOLMODE:
2444 err = zvol_set_volmode(dsname, source, intval);
2446 case ZFS_PROP_VERSION:
2450 if ((err = zfsvfs_hold(dsname, FTAG, &zfsvfs, B_TRUE)) != 0)
2453 err = zfs_set_version(zfsvfs, intval);
2454 zfsvfs_rele(zfsvfs, FTAG);
2456 if (err == 0 && intval >= ZPL_VERSION_USERSPACE) {
2459 zc = kmem_zalloc(sizeof (zfs_cmd_t), KM_SLEEP);
2460 (void) strcpy(zc->zc_name, dsname);
2461 (void) zfs_ioc_userspace_upgrade(zc);
2462 (void) zfs_ioc_userobjspace_upgrade(zc);
2463 kmem_free(zc, sizeof (zfs_cmd_t));
2475 * This function is best effort. If it fails to set any of the given properties,
2476 * it continues to set as many as it can and returns the last error
2477 * encountered. If the caller provides a non-NULL errlist, it will be filled in
2478 * with the list of names of all the properties that failed along with the
2479 * corresponding error numbers.
2481 * If every property is set successfully, zero is returned and errlist is not
2485 zfs_set_prop_nvlist(const char *dsname, zprop_source_t source, nvlist_t *nvl,
2494 nvlist_t *genericnvl = fnvlist_alloc();
2495 nvlist_t *retrynvl = fnvlist_alloc();
2498 while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) {
2499 const char *propname = nvpair_name(pair);
2500 zfs_prop_t prop = zfs_name_to_prop(propname);
2503 /* decode the property value */
2505 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2507 attrs = fnvpair_value_nvlist(pair);
2508 if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2510 err = SET_ERROR(EINVAL);
2513 /* Validate value type */
2514 if (err == 0 && source == ZPROP_SRC_INHERITED) {
2515 /* inherited properties are expected to be booleans */
2516 if (nvpair_type(propval) != DATA_TYPE_BOOLEAN)
2517 err = SET_ERROR(EINVAL);
2518 } else if (err == 0 && prop == ZPROP_INVAL) {
2519 if (zfs_prop_user(propname)) {
2520 if (nvpair_type(propval) != DATA_TYPE_STRING)
2521 err = SET_ERROR(EINVAL);
2522 } else if (zfs_prop_userquota(propname)) {
2523 if (nvpair_type(propval) !=
2524 DATA_TYPE_UINT64_ARRAY)
2525 err = SET_ERROR(EINVAL);
2527 err = SET_ERROR(EINVAL);
2529 } else if (err == 0) {
2530 if (nvpair_type(propval) == DATA_TYPE_STRING) {
2531 if (zfs_prop_get_type(prop) != PROP_TYPE_STRING)
2532 err = SET_ERROR(EINVAL);
2533 } else if (nvpair_type(propval) == DATA_TYPE_UINT64) {
2536 intval = fnvpair_value_uint64(propval);
2538 switch (zfs_prop_get_type(prop)) {
2539 case PROP_TYPE_NUMBER:
2541 case PROP_TYPE_STRING:
2542 err = SET_ERROR(EINVAL);
2544 case PROP_TYPE_INDEX:
2545 if (zfs_prop_index_to_string(prop,
2546 intval, &unused) != 0)
2547 err = SET_ERROR(EINVAL);
2551 "unknown property type");
2554 err = SET_ERROR(EINVAL);
2558 /* Validate permissions */
2560 err = zfs_check_settable(dsname, pair, CRED());
2563 if (source == ZPROP_SRC_INHERITED)
2564 err = -1; /* does not need special handling */
2566 err = zfs_prop_set_special(dsname, source,
2570 * For better performance we build up a list of
2571 * properties to set in a single transaction.
2573 err = nvlist_add_nvpair(genericnvl, pair);
2574 } else if (err != 0 && nvl != retrynvl) {
2576 * This may be a spurious error caused by
2577 * receiving quota and reservation out of order.
2578 * Try again in a second pass.
2580 err = nvlist_add_nvpair(retrynvl, pair);
2585 if (errlist != NULL)
2586 fnvlist_add_int32(errlist, propname, err);
2591 if (nvl != retrynvl && !nvlist_empty(retrynvl)) {
2596 if (!nvlist_empty(genericnvl) &&
2597 dsl_props_set(dsname, source, genericnvl) != 0) {
2599 * If this fails, we still want to set as many properties as we
2600 * can, so try setting them individually.
2603 while ((pair = nvlist_next_nvpair(genericnvl, pair)) != NULL) {
2604 const char *propname = nvpair_name(pair);
2608 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2610 attrs = fnvpair_value_nvlist(pair);
2611 propval = fnvlist_lookup_nvpair(attrs,
2615 if (nvpair_type(propval) == DATA_TYPE_STRING) {
2616 strval = fnvpair_value_string(propval);
2617 err = dsl_prop_set_string(dsname, propname,
2619 } else if (nvpair_type(propval) == DATA_TYPE_BOOLEAN) {
2620 err = dsl_prop_inherit(dsname, propname,
2623 intval = fnvpair_value_uint64(propval);
2624 err = dsl_prop_set_int(dsname, propname, source,
2629 if (errlist != NULL) {
2630 fnvlist_add_int32(errlist, propname,
2637 nvlist_free(genericnvl);
2638 nvlist_free(retrynvl);
2644 * Check that all the properties are valid user properties.
2647 zfs_check_userprops(const char *fsname, nvlist_t *nvl)
2649 nvpair_t *pair = NULL;
2652 while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) {
2653 const char *propname = nvpair_name(pair);
2655 if (!zfs_prop_user(propname) ||
2656 nvpair_type(pair) != DATA_TYPE_STRING)
2657 return (SET_ERROR(EINVAL));
2659 if ((error = zfs_secpolicy_write_perms(fsname,
2660 ZFS_DELEG_PERM_USERPROP, CRED())))
2663 if (strlen(propname) >= ZAP_MAXNAMELEN)
2664 return (SET_ERROR(ENAMETOOLONG));
2666 if (strlen(fnvpair_value_string(pair)) >= ZAP_MAXVALUELEN)
2667 return (SET_ERROR(E2BIG));
2673 props_skip(nvlist_t *props, nvlist_t *skipped, nvlist_t **newprops)
2677 VERIFY(nvlist_alloc(newprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2680 while ((pair = nvlist_next_nvpair(props, pair)) != NULL) {
2681 if (nvlist_exists(skipped, nvpair_name(pair)))
2684 VERIFY(nvlist_add_nvpair(*newprops, pair) == 0);
2689 clear_received_props(const char *dsname, nvlist_t *props,
2693 nvlist_t *cleared_props = NULL;
2694 props_skip(props, skipped, &cleared_props);
2695 if (!nvlist_empty(cleared_props)) {
2697 * Acts on local properties until the dataset has received
2698 * properties at least once on or after SPA_VERSION_RECVD_PROPS.
2700 zprop_source_t flags = (ZPROP_SRC_NONE |
2701 (dsl_prop_get_hasrecvd(dsname) ? ZPROP_SRC_RECEIVED : 0));
2702 err = zfs_set_prop_nvlist(dsname, flags, cleared_props, NULL);
2704 nvlist_free(cleared_props);
2710 * zc_name name of filesystem
2711 * zc_value name of property to set
2712 * zc_nvlist_src{_size} nvlist of properties to apply
2713 * zc_cookie received properties flag
2716 * zc_nvlist_dst{_size} error for each unapplied received property
2719 zfs_ioc_set_prop(zfs_cmd_t *zc)
2722 boolean_t received = zc->zc_cookie;
2723 zprop_source_t source = (received ? ZPROP_SRC_RECEIVED :
2728 if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2729 zc->zc_iflags, &nvl)) != 0)
2733 nvlist_t *origprops;
2735 if (dsl_prop_get_received(zc->zc_name, &origprops) == 0) {
2736 (void) clear_received_props(zc->zc_name,
2738 nvlist_free(origprops);
2741 error = dsl_prop_set_hasrecvd(zc->zc_name);
2744 errors = fnvlist_alloc();
2746 error = zfs_set_prop_nvlist(zc->zc_name, source, nvl, errors);
2748 if (zc->zc_nvlist_dst != 0 && errors != NULL) {
2749 (void) put_nvlist(zc, errors);
2752 nvlist_free(errors);
2759 * zc_name name of filesystem
2760 * zc_value name of property to inherit
2761 * zc_cookie revert to received value if TRUE
2766 zfs_ioc_inherit_prop(zfs_cmd_t *zc)
2768 const char *propname = zc->zc_value;
2769 zfs_prop_t prop = zfs_name_to_prop(propname);
2770 boolean_t received = zc->zc_cookie;
2771 zprop_source_t source = (received
2772 ? ZPROP_SRC_NONE /* revert to received value, if any */
2773 : ZPROP_SRC_INHERITED); /* explicitly inherit */
2781 * Only check this in the non-received case. We want to allow
2782 * 'inherit -S' to revert non-inheritable properties like quota
2783 * and reservation to the received or default values even though
2784 * they are not considered inheritable.
2786 if (prop != ZPROP_INVAL && !zfs_prop_inheritable(prop))
2787 return (SET_ERROR(EINVAL));
2790 if (prop == ZPROP_INVAL) {
2791 if (!zfs_prop_user(propname))
2792 return (SET_ERROR(EINVAL));
2794 type = PROP_TYPE_STRING;
2795 } else if (prop == ZFS_PROP_VOLSIZE || prop == ZFS_PROP_VERSION) {
2796 return (SET_ERROR(EINVAL));
2798 type = zfs_prop_get_type(prop);
2802 * zfs_prop_set_special() expects properties in the form of an
2803 * nvpair with type info.
2805 dummy = fnvlist_alloc();
2808 case PROP_TYPE_STRING:
2809 VERIFY(0 == nvlist_add_string(dummy, propname, ""));
2811 case PROP_TYPE_NUMBER:
2812 case PROP_TYPE_INDEX:
2813 VERIFY(0 == nvlist_add_uint64(dummy, propname, 0));
2816 err = SET_ERROR(EINVAL);
2820 pair = nvlist_next_nvpair(dummy, NULL);
2822 err = SET_ERROR(EINVAL);
2824 err = zfs_prop_set_special(zc->zc_name, source, pair);
2825 if (err == -1) /* property is not "special", needs handling */
2826 err = dsl_prop_inherit(zc->zc_name, zc->zc_value,
2836 zfs_ioc_pool_set_props(zfs_cmd_t *zc)
2843 if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2844 zc->zc_iflags, &props)))
2848 * If the only property is the configfile, then just do a spa_lookup()
2849 * to handle the faulted case.
2851 pair = nvlist_next_nvpair(props, NULL);
2852 if (pair != NULL && strcmp(nvpair_name(pair),
2853 zpool_prop_to_name(ZPOOL_PROP_CACHEFILE)) == 0 &&
2854 nvlist_next_nvpair(props, pair) == NULL) {
2855 mutex_enter(&spa_namespace_lock);
2856 if ((spa = spa_lookup(zc->zc_name)) != NULL) {
2857 spa_configfile_set(spa, props, B_FALSE);
2858 spa_config_sync(spa, B_FALSE, B_TRUE);
2860 mutex_exit(&spa_namespace_lock);
2867 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) {
2872 error = spa_prop_set(spa, props);
2875 spa_close(spa, FTAG);
2881 zfs_ioc_pool_get_props(zfs_cmd_t *zc)
2885 nvlist_t *nvp = NULL;
2887 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) {
2889 * If the pool is faulted, there may be properties we can still
2890 * get (such as altroot and cachefile), so attempt to get them
2893 mutex_enter(&spa_namespace_lock);
2894 if ((spa = spa_lookup(zc->zc_name)) != NULL)
2895 error = spa_prop_get(spa, &nvp);
2896 mutex_exit(&spa_namespace_lock);
2898 error = spa_prop_get(spa, &nvp);
2899 spa_close(spa, FTAG);
2902 if (error == 0 && zc->zc_nvlist_dst != 0)
2903 error = put_nvlist(zc, nvp);
2905 error = SET_ERROR(EFAULT);
2913 * zc_name name of filesystem
2914 * zc_nvlist_src{_size} nvlist of delegated permissions
2915 * zc_perm_action allow/unallow flag
2920 zfs_ioc_set_fsacl(zfs_cmd_t *zc)
2923 nvlist_t *fsaclnv = NULL;
2925 if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2926 zc->zc_iflags, &fsaclnv)) != 0)
2930 * Verify nvlist is constructed correctly
2932 if ((error = zfs_deleg_verify_nvlist(fsaclnv)) != 0) {
2933 nvlist_free(fsaclnv);
2934 return (SET_ERROR(EINVAL));
2938 * If we don't have PRIV_SYS_MOUNT, then validate
2939 * that user is allowed to hand out each permission in
2943 error = secpolicy_zfs(CRED());
2945 if (zc->zc_perm_action == B_FALSE) {
2946 error = dsl_deleg_can_allow(zc->zc_name,
2949 error = dsl_deleg_can_unallow(zc->zc_name,
2955 error = dsl_deleg_set(zc->zc_name, fsaclnv, zc->zc_perm_action);
2957 nvlist_free(fsaclnv);
2963 * zc_name name of filesystem
2966 * zc_nvlist_src{_size} nvlist of delegated permissions
2969 zfs_ioc_get_fsacl(zfs_cmd_t *zc)
2974 if ((error = dsl_deleg_get(zc->zc_name, &nvp)) == 0) {
2975 error = put_nvlist(zc, nvp);
2984 zfs_create_cb(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx)
2986 zfs_creat_t *zct = arg;
2988 zfs_create_fs(os, cr, zct->zct_zplprops, tx);
2991 #define ZFS_PROP_UNDEFINED ((uint64_t)-1)
2995 * os parent objset pointer (NULL if root fs)
2996 * fuids_ok fuids allowed in this version of the spa?
2997 * sa_ok SAs allowed in this version of the spa?
2998 * createprops list of properties requested by creator
3001 * zplprops values for the zplprops we attach to the master node object
3002 * is_ci true if requested file system will be purely case-insensitive
3004 * Determine the settings for utf8only, normalization and
3005 * casesensitivity. Specific values may have been requested by the
3006 * creator and/or we can inherit values from the parent dataset. If
3007 * the file system is of too early a vintage, a creator can not
3008 * request settings for these properties, even if the requested
3009 * setting is the default value. We don't actually want to create dsl
3010 * properties for these, so remove them from the source nvlist after
3014 zfs_fill_zplprops_impl(objset_t *os, uint64_t zplver,
3015 boolean_t fuids_ok, boolean_t sa_ok, nvlist_t *createprops,
3016 nvlist_t *zplprops, boolean_t *is_ci)
3018 uint64_t sense = ZFS_PROP_UNDEFINED;
3019 uint64_t norm = ZFS_PROP_UNDEFINED;
3020 uint64_t u8 = ZFS_PROP_UNDEFINED;
3023 ASSERT(zplprops != NULL);
3025 if (os != NULL && os->os_phys->os_type != DMU_OST_ZFS)
3026 return (SET_ERROR(EINVAL));
3029 * Pull out creator prop choices, if any.
3032 (void) nvlist_lookup_uint64(createprops,
3033 zfs_prop_to_name(ZFS_PROP_VERSION), &zplver);
3034 (void) nvlist_lookup_uint64(createprops,
3035 zfs_prop_to_name(ZFS_PROP_NORMALIZE), &norm);
3036 (void) nvlist_remove_all(createprops,
3037 zfs_prop_to_name(ZFS_PROP_NORMALIZE));
3038 (void) nvlist_lookup_uint64(createprops,
3039 zfs_prop_to_name(ZFS_PROP_UTF8ONLY), &u8);
3040 (void) nvlist_remove_all(createprops,
3041 zfs_prop_to_name(ZFS_PROP_UTF8ONLY));
3042 (void) nvlist_lookup_uint64(createprops,
3043 zfs_prop_to_name(ZFS_PROP_CASE), &sense);
3044 (void) nvlist_remove_all(createprops,
3045 zfs_prop_to_name(ZFS_PROP_CASE));
3049 * If the zpl version requested is whacky or the file system
3050 * or pool is version is too "young" to support normalization
3051 * and the creator tried to set a value for one of the props,
3054 if ((zplver < ZPL_VERSION_INITIAL || zplver > ZPL_VERSION) ||
3055 (zplver >= ZPL_VERSION_FUID && !fuids_ok) ||
3056 (zplver >= ZPL_VERSION_SA && !sa_ok) ||
3057 (zplver < ZPL_VERSION_NORMALIZATION &&
3058 (norm != ZFS_PROP_UNDEFINED || u8 != ZFS_PROP_UNDEFINED ||
3059 sense != ZFS_PROP_UNDEFINED)))
3060 return (SET_ERROR(ENOTSUP));
3063 * Put the version in the zplprops
3065 VERIFY(nvlist_add_uint64(zplprops,
3066 zfs_prop_to_name(ZFS_PROP_VERSION), zplver) == 0);
3068 if (norm == ZFS_PROP_UNDEFINED &&
3069 (error = zfs_get_zplprop(os, ZFS_PROP_NORMALIZE, &norm)) != 0)
3071 VERIFY(nvlist_add_uint64(zplprops,
3072 zfs_prop_to_name(ZFS_PROP_NORMALIZE), norm) == 0);
3075 * If we're normalizing, names must always be valid UTF-8 strings.
3079 if (u8 == ZFS_PROP_UNDEFINED &&
3080 (error = zfs_get_zplprop(os, ZFS_PROP_UTF8ONLY, &u8)) != 0)
3082 VERIFY(nvlist_add_uint64(zplprops,
3083 zfs_prop_to_name(ZFS_PROP_UTF8ONLY), u8) == 0);
3085 if (sense == ZFS_PROP_UNDEFINED &&
3086 (error = zfs_get_zplprop(os, ZFS_PROP_CASE, &sense)) != 0)
3088 VERIFY(nvlist_add_uint64(zplprops,
3089 zfs_prop_to_name(ZFS_PROP_CASE), sense) == 0);
3092 *is_ci = (sense == ZFS_CASE_INSENSITIVE);
3098 zfs_fill_zplprops(const char *dataset, nvlist_t *createprops,
3099 nvlist_t *zplprops, boolean_t *is_ci)
3101 boolean_t fuids_ok, sa_ok;
3102 uint64_t zplver = ZPL_VERSION;
3103 objset_t *os = NULL;
3104 char parentname[ZFS_MAX_DATASET_NAME_LEN];
3110 (void) strlcpy(parentname, dataset, sizeof (parentname));
3111 cp = strrchr(parentname, '/');
3115 if ((error = spa_open(dataset, &spa, FTAG)) != 0)
3118 spa_vers = spa_version(spa);
3119 spa_close(spa, FTAG);
3121 zplver = zfs_zpl_version_map(spa_vers);
3122 fuids_ok = (zplver >= ZPL_VERSION_FUID);
3123 sa_ok = (zplver >= ZPL_VERSION_SA);
3126 * Open parent object set so we can inherit zplprop values.
3128 if ((error = dmu_objset_hold(parentname, FTAG, &os)) != 0)
3131 error = zfs_fill_zplprops_impl(os, zplver, fuids_ok, sa_ok, createprops,
3133 dmu_objset_rele(os, FTAG);
3138 zfs_fill_zplprops_root(uint64_t spa_vers, nvlist_t *createprops,
3139 nvlist_t *zplprops, boolean_t *is_ci)
3143 uint64_t zplver = ZPL_VERSION;
3146 zplver = zfs_zpl_version_map(spa_vers);
3147 fuids_ok = (zplver >= ZPL_VERSION_FUID);
3148 sa_ok = (zplver >= ZPL_VERSION_SA);
3150 error = zfs_fill_zplprops_impl(NULL, zplver, fuids_ok, sa_ok,
3151 createprops, zplprops, is_ci);
3157 * "type" -> dmu_objset_type_t (int32)
3158 * (optional) "props" -> { prop -> value }
3161 * outnvl: propname -> error code (int32)
3164 zfs_ioc_create(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3167 zfs_creat_t zct = { 0 };
3168 nvlist_t *nvprops = NULL;
3169 void (*cbfunc)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx);
3171 dmu_objset_type_t type;
3172 boolean_t is_insensitive = B_FALSE;
3174 if (nvlist_lookup_int32(innvl, "type", &type32) != 0)
3175 return (SET_ERROR(EINVAL));
3177 (void) nvlist_lookup_nvlist(innvl, "props", &nvprops);
3181 cbfunc = zfs_create_cb;
3185 cbfunc = zvol_create_cb;
3192 if (strchr(fsname, '@') ||
3193 strchr(fsname, '%'))
3194 return (SET_ERROR(EINVAL));
3196 zct.zct_props = nvprops;
3199 return (SET_ERROR(EINVAL));
3201 if (type == DMU_OST_ZVOL) {
3202 uint64_t volsize, volblocksize;
3204 if (nvprops == NULL)
3205 return (SET_ERROR(EINVAL));
3206 if (nvlist_lookup_uint64(nvprops,
3207 zfs_prop_to_name(ZFS_PROP_VOLSIZE), &volsize) != 0)
3208 return (SET_ERROR(EINVAL));
3210 if ((error = nvlist_lookup_uint64(nvprops,
3211 zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE),
3212 &volblocksize)) != 0 && error != ENOENT)
3213 return (SET_ERROR(EINVAL));
3216 volblocksize = zfs_prop_default_numeric(
3217 ZFS_PROP_VOLBLOCKSIZE);
3219 if ((error = zvol_check_volblocksize(fsname,
3220 volblocksize)) != 0 ||
3221 (error = zvol_check_volsize(volsize,
3222 volblocksize)) != 0)
3224 } else if (type == DMU_OST_ZFS) {
3228 * We have to have normalization and
3229 * case-folding flags correct when we do the
3230 * file system creation, so go figure them out
3233 VERIFY(nvlist_alloc(&zct.zct_zplprops,
3234 NV_UNIQUE_NAME, KM_SLEEP) == 0);
3235 error = zfs_fill_zplprops(fsname, nvprops,
3236 zct.zct_zplprops, &is_insensitive);
3238 nvlist_free(zct.zct_zplprops);
3243 error = dmu_objset_create(fsname, type,
3244 is_insensitive ? DS_FLAG_CI_DATASET : 0, cbfunc, &zct);
3245 nvlist_free(zct.zct_zplprops);
3248 * It would be nice to do this atomically.
3251 error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL,
3258 * Volumes will return EBUSY and cannot be destroyed
3259 * until all asynchronous minor handling has completed.
3260 * Wait for the spa_zvol_taskq to drain then retry.
3262 error2 = dsl_destroy_head(fsname);
3263 while ((error2 == EBUSY) && (type == DMU_OST_ZVOL)) {
3264 error2 = spa_open(fsname, &spa, FTAG);
3266 taskq_wait(spa->spa_zvol_taskq);
3267 spa_close(spa, FTAG);
3269 error2 = dsl_destroy_head(fsname);
3278 * "origin" -> name of origin snapshot
3279 * (optional) "props" -> { prop -> value }
3283 * outnvl: propname -> error code (int32)
3286 zfs_ioc_clone(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3289 nvlist_t *nvprops = NULL;
3292 if (nvlist_lookup_string(innvl, "origin", &origin_name) != 0)
3293 return (SET_ERROR(EINVAL));
3294 (void) nvlist_lookup_nvlist(innvl, "props", &nvprops);
3296 if (strchr(fsname, '@') ||
3297 strchr(fsname, '%'))
3298 return (SET_ERROR(EINVAL));
3300 if (dataset_namecheck(origin_name, NULL, NULL) != 0)
3301 return (SET_ERROR(EINVAL));
3302 error = dmu_objset_clone(fsname, origin_name);
3307 * It would be nice to do this atomically.
3310 error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL,
3313 (void) dsl_destroy_head(fsname);
3320 * "snaps" -> { snapshot1, snapshot2 }
3321 * (optional) "props" -> { prop -> value (string) }
3324 * outnvl: snapshot -> error code (int32)
3327 zfs_ioc_snapshot(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3330 nvlist_t *props = NULL;
3332 nvpair_t *pair, *pair2;
3334 (void) nvlist_lookup_nvlist(innvl, "props", &props);
3335 if ((error = zfs_check_userprops(poolname, props)) != 0)
3338 if (!nvlist_empty(props) &&
3339 zfs_earlier_version(poolname, SPA_VERSION_SNAP_PROPS))
3340 return (SET_ERROR(ENOTSUP));
3342 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
3343 return (SET_ERROR(EINVAL));
3344 poollen = strlen(poolname);
3345 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
3346 pair = nvlist_next_nvpair(snaps, pair)) {
3347 const char *name = nvpair_name(pair);
3348 const char *cp = strchr(name, '@');
3351 * The snap name must contain an @, and the part after it must
3352 * contain only valid characters.
3355 zfs_component_namecheck(cp + 1, NULL, NULL) != 0)
3356 return (SET_ERROR(EINVAL));
3359 * The snap must be in the specified pool.
3361 if (strncmp(name, poolname, poollen) != 0 ||
3362 (name[poollen] != '/' && name[poollen] != '@'))
3363 return (SET_ERROR(EXDEV));
3365 /* This must be the only snap of this fs. */
3366 for (pair2 = nvlist_next_nvpair(snaps, pair);
3367 pair2 != NULL; pair2 = nvlist_next_nvpair(snaps, pair2)) {
3368 if (strncmp(name, nvpair_name(pair2), cp - name + 1)
3370 return (SET_ERROR(EXDEV));
3375 error = dsl_dataset_snapshot(snaps, props, outnvl);
3381 * innvl: "message" -> string
3385 zfs_ioc_log_history(const char *unused, nvlist_t *innvl, nvlist_t *outnvl)
3393 * The poolname in the ioctl is not set, we get it from the TSD,
3394 * which was set at the end of the last successful ioctl that allows
3395 * logging. The secpolicy func already checked that it is set.
3396 * Only one log ioctl is allowed after each successful ioctl, so
3397 * we clear the TSD here.
3399 poolname = tsd_get(zfs_allow_log_key);
3400 if (poolname == NULL)
3401 return (SET_ERROR(EINVAL));
3402 (void) tsd_set(zfs_allow_log_key, NULL);
3403 error = spa_open(poolname, &spa, FTAG);
3408 if (nvlist_lookup_string(innvl, "message", &message) != 0) {
3409 spa_close(spa, FTAG);
3410 return (SET_ERROR(EINVAL));
3413 if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) {
3414 spa_close(spa, FTAG);
3415 return (SET_ERROR(ENOTSUP));
3418 error = spa_history_log(spa, message);
3419 spa_close(spa, FTAG);
3424 * The dp_config_rwlock must not be held when calling this, because the
3425 * unmount may need to write out data.
3427 * This function is best-effort. Callers must deal gracefully if it
3428 * remains mounted (or is remounted after this call).
3430 * Returns 0 if the argument is not a snapshot, or it is not currently a
3431 * filesystem, or we were able to unmount it. Returns error code otherwise.
3434 zfs_unmount_snap(const char *snapname)
3438 if (strchr(snapname, '@') == NULL)
3441 err = zfsctl_snapshot_unmount((char *)snapname, MNT_FORCE);
3442 if (err != 0 && err != ENOENT)
3443 return (SET_ERROR(err));
3450 zfs_unmount_snap_cb(const char *snapname, void *arg)
3452 return (zfs_unmount_snap(snapname));
3456 * When a clone is destroyed, its origin may also need to be destroyed,
3457 * in which case it must be unmounted. This routine will do that unmount
3461 zfs_destroy_unmount_origin(const char *fsname)
3467 error = dmu_objset_hold(fsname, FTAG, &os);
3470 ds = dmu_objset_ds(os);
3471 if (dsl_dir_is_clone(ds->ds_dir) && DS_IS_DEFER_DESTROY(ds->ds_prev)) {
3472 char originname[ZFS_MAX_DATASET_NAME_LEN];
3473 dsl_dataset_name(ds->ds_prev, originname);
3474 dmu_objset_rele(os, FTAG);
3475 (void) zfs_unmount_snap(originname);
3477 dmu_objset_rele(os, FTAG);
3483 * "snaps" -> { snapshot1, snapshot2 }
3484 * (optional boolean) "defer"
3487 * outnvl: snapshot -> error code (int32)
3491 zfs_ioc_destroy_snaps(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3497 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
3498 return (SET_ERROR(EINVAL));
3499 defer = nvlist_exists(innvl, "defer");
3501 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
3502 pair = nvlist_next_nvpair(snaps, pair)) {
3503 (void) zfs_unmount_snap(nvpair_name(pair));
3506 return (dsl_destroy_snapshots_nvl(snaps, defer, outnvl));
3510 * Create bookmarks. Bookmark names are of the form <fs>#<bmark>.
3511 * All bookmarks must be in the same pool.
3514 * bookmark1 -> snapshot1, bookmark2 -> snapshot2
3517 * outnvl: bookmark -> error code (int32)
3522 zfs_ioc_bookmark(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3524 nvpair_t *pair, *pair2;
3526 for (pair = nvlist_next_nvpair(innvl, NULL);
3527 pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) {
3531 * Verify the snapshot argument.
3533 if (nvpair_value_string(pair, &snap_name) != 0)
3534 return (SET_ERROR(EINVAL));
3537 /* Verify that the keys (bookmarks) are unique */
3538 for (pair2 = nvlist_next_nvpair(innvl, pair);
3539 pair2 != NULL; pair2 = nvlist_next_nvpair(innvl, pair2)) {
3540 if (strcmp(nvpair_name(pair), nvpair_name(pair2)) == 0)
3541 return (SET_ERROR(EINVAL));
3545 return (dsl_bookmark_create(innvl, outnvl));
3550 * property 1, property 2, ...
3554 * bookmark name 1 -> { property 1, property 2, ... },
3555 * bookmark name 2 -> { property 1, property 2, ... }
3560 zfs_ioc_get_bookmarks(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3562 return (dsl_get_bookmarks(fsname, innvl, outnvl));
3567 * bookmark name 1, bookmark name 2
3570 * outnvl: bookmark -> error code (int32)
3574 zfs_ioc_destroy_bookmarks(const char *poolname, nvlist_t *innvl,
3580 poollen = strlen(poolname);
3581 for (pair = nvlist_next_nvpair(innvl, NULL);
3582 pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) {
3583 const char *name = nvpair_name(pair);
3584 const char *cp = strchr(name, '#');
3587 * The bookmark name must contain an #, and the part after it
3588 * must contain only valid characters.
3591 zfs_component_namecheck(cp + 1, NULL, NULL) != 0)
3592 return (SET_ERROR(EINVAL));
3595 * The bookmark must be in the specified pool.
3597 if (strncmp(name, poolname, poollen) != 0 ||
3598 (name[poollen] != '/' && name[poollen] != '#'))
3599 return (SET_ERROR(EXDEV));
3602 error = dsl_bookmark_destroy(innvl, outnvl);
3608 * zc_name name of dataset to destroy
3609 * zc_objset_type type of objset
3610 * zc_defer_destroy mark for deferred destroy
3615 zfs_ioc_destroy(zfs_cmd_t *zc)
3619 if (zc->zc_objset_type == DMU_OST_ZFS) {
3620 err = zfs_unmount_snap(zc->zc_name);
3625 if (strchr(zc->zc_name, '@')) {
3626 err = dsl_destroy_snapshot(zc->zc_name, zc->zc_defer_destroy);
3628 err = dsl_destroy_head(zc->zc_name);
3629 if (err == EEXIST) {
3631 * It is possible that the given DS may have
3632 * hidden child (%recv) datasets - "leftovers"
3633 * resulting from the previously interrupted
3636 * 6 extra bytes for /%recv
3638 char namebuf[ZFS_MAX_DATASET_NAME_LEN + 6];
3640 if (snprintf(namebuf, sizeof (namebuf), "%s/%s",
3641 zc->zc_name, recv_clone_name) >=
3643 return (SET_ERROR(EINVAL));
3646 * Try to remove the hidden child (%recv) and after
3647 * that try to remove the target dataset.
3648 * If the hidden child (%recv) does not exist
3649 * the original error (EEXIST) will be returned
3651 err = dsl_destroy_head(namebuf);
3653 err = dsl_destroy_head(zc->zc_name);
3654 else if (err == ENOENT)
3663 * fsname is name of dataset to rollback (to most recent snapshot)
3665 * innvl may contain name of expected target snapshot
3667 * outnvl: "target" -> name of most recent snapshot
3672 zfs_ioc_rollback(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3676 char *target = NULL;
3679 (void) nvlist_lookup_string(innvl, "target", &target);
3680 if (target != NULL) {
3681 int fslen = strlen(fsname);
3683 if (strncmp(fsname, target, fslen) != 0)
3684 return (SET_ERROR(EINVAL));
3685 if (target[fslen] != '@')
3686 return (SET_ERROR(EINVAL));
3689 if (getzfsvfs(fsname, &zfsvfs) == 0) {
3692 ds = dmu_objset_ds(zfsvfs->z_os);
3693 error = zfs_suspend_fs(zfsvfs);
3697 error = dsl_dataset_rollback(fsname, target, zfsvfs,
3699 resume_err = zfs_resume_fs(zfsvfs, ds);
3700 error = error ? error : resume_err;
3702 deactivate_super(zfsvfs->z_sb);
3703 } else if ((zv = zvol_suspend(fsname)) != NULL) {
3704 error = dsl_dataset_rollback(fsname, target, zvol_tag(zv),
3708 error = dsl_dataset_rollback(fsname, target, NULL, outnvl);
3714 recursive_unmount(const char *fsname, void *arg)
3716 const char *snapname = arg;
3720 fullname = kmem_asprintf("%s@%s", fsname, snapname);
3721 error = zfs_unmount_snap(fullname);
3729 * zc_name old name of dataset
3730 * zc_value new name of dataset
3731 * zc_cookie recursive flag (only valid for snapshots)
3736 zfs_ioc_rename(zfs_cmd_t *zc)
3738 boolean_t recursive = zc->zc_cookie & 1;
3741 zc->zc_value[sizeof (zc->zc_value) - 1] = '\0';
3742 if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
3743 strchr(zc->zc_value, '%'))
3744 return (SET_ERROR(EINVAL));
3746 at = strchr(zc->zc_name, '@');
3748 /* snaps must be in same fs */
3751 if (strncmp(zc->zc_name, zc->zc_value, at - zc->zc_name + 1))
3752 return (SET_ERROR(EXDEV));
3754 if (zc->zc_objset_type == DMU_OST_ZFS) {
3755 error = dmu_objset_find(zc->zc_name,
3756 recursive_unmount, at + 1,
3757 recursive ? DS_FIND_CHILDREN : 0);
3763 error = dsl_dataset_rename_snapshot(zc->zc_name,
3764 at + 1, strchr(zc->zc_value, '@') + 1, recursive);
3769 return (dsl_dir_rename(zc->zc_name, zc->zc_value));
3774 zfs_check_settable(const char *dsname, nvpair_t *pair, cred_t *cr)
3776 const char *propname = nvpair_name(pair);
3777 boolean_t issnap = (strchr(dsname, '@') != NULL);
3778 zfs_prop_t prop = zfs_name_to_prop(propname);
3782 if (prop == ZPROP_INVAL) {
3783 if (zfs_prop_user(propname)) {
3784 if ((err = zfs_secpolicy_write_perms(dsname,
3785 ZFS_DELEG_PERM_USERPROP, cr)))
3790 if (!issnap && zfs_prop_userquota(propname)) {
3791 const char *perm = NULL;
3792 const char *uq_prefix =
3793 zfs_userquota_prop_prefixes[ZFS_PROP_USERQUOTA];
3794 const char *gq_prefix =
3795 zfs_userquota_prop_prefixes[ZFS_PROP_GROUPQUOTA];
3796 const char *uiq_prefix =
3797 zfs_userquota_prop_prefixes[ZFS_PROP_USEROBJQUOTA];
3798 const char *giq_prefix =
3799 zfs_userquota_prop_prefixes[ZFS_PROP_GROUPOBJQUOTA];
3801 if (strncmp(propname, uq_prefix,
3802 strlen(uq_prefix)) == 0) {
3803 perm = ZFS_DELEG_PERM_USERQUOTA;
3804 } else if (strncmp(propname, uiq_prefix,
3805 strlen(uiq_prefix)) == 0) {
3806 perm = ZFS_DELEG_PERM_USEROBJQUOTA;
3807 } else if (strncmp(propname, gq_prefix,
3808 strlen(gq_prefix)) == 0) {
3809 perm = ZFS_DELEG_PERM_GROUPQUOTA;
3810 } else if (strncmp(propname, giq_prefix,
3811 strlen(giq_prefix)) == 0) {
3812 perm = ZFS_DELEG_PERM_GROUPOBJQUOTA;
3814 /* USERUSED and GROUPUSED are read-only */
3815 return (SET_ERROR(EINVAL));
3818 if ((err = zfs_secpolicy_write_perms(dsname, perm, cr)))
3823 return (SET_ERROR(EINVAL));
3827 return (SET_ERROR(EINVAL));
3829 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
3831 * dsl_prop_get_all_impl() returns properties in this
3835 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
3836 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
3841 * Check that this value is valid for this pool version
3844 case ZFS_PROP_COMPRESSION:
3846 * If the user specified gzip compression, make sure
3847 * the SPA supports it. We ignore any errors here since
3848 * we'll catch them later.
3850 if (nvpair_value_uint64(pair, &intval) == 0) {
3851 if (intval >= ZIO_COMPRESS_GZIP_1 &&
3852 intval <= ZIO_COMPRESS_GZIP_9 &&
3853 zfs_earlier_version(dsname,
3854 SPA_VERSION_GZIP_COMPRESSION)) {
3855 return (SET_ERROR(ENOTSUP));
3858 if (intval == ZIO_COMPRESS_ZLE &&
3859 zfs_earlier_version(dsname,
3860 SPA_VERSION_ZLE_COMPRESSION))
3861 return (SET_ERROR(ENOTSUP));
3863 if (intval == ZIO_COMPRESS_LZ4) {
3866 if ((err = spa_open(dsname, &spa, FTAG)) != 0)
3869 if (!spa_feature_is_enabled(spa,
3870 SPA_FEATURE_LZ4_COMPRESS)) {
3871 spa_close(spa, FTAG);
3872 return (SET_ERROR(ENOTSUP));
3874 spa_close(spa, FTAG);
3878 * If this is a bootable dataset then
3879 * verify that the compression algorithm
3880 * is supported for booting. We must return
3881 * something other than ENOTSUP since it
3882 * implies a downrev pool version.
3884 if (zfs_is_bootfs(dsname) &&
3885 !BOOTFS_COMPRESS_VALID(intval)) {
3886 return (SET_ERROR(ERANGE));
3891 case ZFS_PROP_COPIES:
3892 if (zfs_earlier_version(dsname, SPA_VERSION_DITTO_BLOCKS))
3893 return (SET_ERROR(ENOTSUP));
3896 case ZFS_PROP_VOLBLOCKSIZE:
3897 case ZFS_PROP_RECORDSIZE:
3898 /* Record sizes above 128k need the feature to be enabled */
3899 if (nvpair_value_uint64(pair, &intval) == 0 &&
3900 intval > SPA_OLD_MAXBLOCKSIZE) {
3904 * We don't allow setting the property above 1MB,
3905 * unless the tunable has been changed.
3907 if (intval > zfs_max_recordsize ||
3908 intval > SPA_MAXBLOCKSIZE)
3909 return (SET_ERROR(ERANGE));
3911 if ((err = spa_open(dsname, &spa, FTAG)) != 0)
3914 if (!spa_feature_is_enabled(spa,
3915 SPA_FEATURE_LARGE_BLOCKS)) {
3916 spa_close(spa, FTAG);
3917 return (SET_ERROR(ENOTSUP));
3919 spa_close(spa, FTAG);
3923 case ZFS_PROP_DNODESIZE:
3924 /* Dnode sizes above 512 need the feature to be enabled */
3925 if (nvpair_value_uint64(pair, &intval) == 0 &&
3926 intval != ZFS_DNSIZE_LEGACY) {
3930 * If this is a bootable dataset then
3931 * we don't allow large (>512B) dnodes,
3932 * because GRUB doesn't support them.
3934 if (zfs_is_bootfs(dsname) &&
3935 intval != ZFS_DNSIZE_LEGACY) {
3936 return (SET_ERROR(EDOM));
3939 if ((err = spa_open(dsname, &spa, FTAG)) != 0)
3942 if (!spa_feature_is_enabled(spa,
3943 SPA_FEATURE_LARGE_DNODE)) {
3944 spa_close(spa, FTAG);
3945 return (SET_ERROR(ENOTSUP));
3947 spa_close(spa, FTAG);
3951 case ZFS_PROP_SHARESMB:
3952 if (zpl_earlier_version(dsname, ZPL_VERSION_FUID))
3953 return (SET_ERROR(ENOTSUP));
3956 case ZFS_PROP_ACLINHERIT:
3957 if (nvpair_type(pair) == DATA_TYPE_UINT64 &&
3958 nvpair_value_uint64(pair, &intval) == 0) {
3959 if (intval == ZFS_ACL_PASSTHROUGH_X &&
3960 zfs_earlier_version(dsname,
3961 SPA_VERSION_PASSTHROUGH_X))
3962 return (SET_ERROR(ENOTSUP));
3965 case ZFS_PROP_CHECKSUM:
3966 case ZFS_PROP_DEDUP:
3968 spa_feature_t feature;
3973 /* dedup feature version checks */
3974 if (prop == ZFS_PROP_DEDUP &&
3975 zfs_earlier_version(dsname, SPA_VERSION_DEDUP))
3976 return (SET_ERROR(ENOTSUP));
3978 if (nvpair_value_uint64(pair, &intval) != 0)
3979 return (SET_ERROR(EINVAL));
3981 /* check prop value is enabled in features */
3982 feature = zio_checksum_to_feature(intval & ZIO_CHECKSUM_MASK);
3983 if (feature == SPA_FEATURE_NONE)
3986 if ((err = spa_open(dsname, &spa, FTAG)) != 0)
3989 * Salted checksums are not supported on root pools.
3991 if (spa_bootfs(spa) != 0 &&
3992 intval < ZIO_CHECKSUM_FUNCTIONS &&
3993 (zio_checksum_table[intval].ci_flags &
3994 ZCHECKSUM_FLAG_SALTED)) {
3995 spa_close(spa, FTAG);
3996 return (SET_ERROR(ERANGE));
3998 if (!spa_feature_is_enabled(spa, feature)) {
3999 spa_close(spa, FTAG);
4000 return (SET_ERROR(ENOTSUP));
4002 spa_close(spa, FTAG);
4010 return (zfs_secpolicy_setprop(dsname, prop, pair, CRED()));
4014 * Removes properties from the given props list that fail permission checks
4015 * needed to clear them and to restore them in case of a receive error. For each
4016 * property, make sure we have both set and inherit permissions.
4018 * Returns the first error encountered if any permission checks fail. If the
4019 * caller provides a non-NULL errlist, it also gives the complete list of names
4020 * of all the properties that failed a permission check along with the
4021 * corresponding error numbers. The caller is responsible for freeing the
4024 * If every property checks out successfully, zero is returned and the list
4025 * pointed at by errlist is NULL.
4028 zfs_check_clearable(char *dataset, nvlist_t *props, nvlist_t **errlist)
4031 nvpair_t *pair, *next_pair;
4038 VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0);
4040 zc = kmem_alloc(sizeof (zfs_cmd_t), KM_SLEEP);
4041 (void) strlcpy(zc->zc_name, dataset, sizeof (zc->zc_name));
4042 pair = nvlist_next_nvpair(props, NULL);
4043 while (pair != NULL) {
4044 next_pair = nvlist_next_nvpair(props, pair);
4046 (void) strlcpy(zc->zc_value, nvpair_name(pair),
4047 sizeof (zc->zc_value));
4048 if ((err = zfs_check_settable(dataset, pair, CRED())) != 0 ||
4049 (err = zfs_secpolicy_inherit_prop(zc, NULL, CRED())) != 0) {
4050 VERIFY(nvlist_remove_nvpair(props, pair) == 0);
4051 VERIFY(nvlist_add_int32(errors,
4052 zc->zc_value, err) == 0);
4056 kmem_free(zc, sizeof (zfs_cmd_t));
4058 if ((pair = nvlist_next_nvpair(errors, NULL)) == NULL) {
4059 nvlist_free(errors);
4062 VERIFY(nvpair_value_int32(pair, &rv) == 0);
4065 if (errlist == NULL)
4066 nvlist_free(errors);
4074 propval_equals(nvpair_t *p1, nvpair_t *p2)
4076 if (nvpair_type(p1) == DATA_TYPE_NVLIST) {
4077 /* dsl_prop_get_all_impl() format */
4079 VERIFY(nvpair_value_nvlist(p1, &attrs) == 0);
4080 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
4084 if (nvpair_type(p2) == DATA_TYPE_NVLIST) {
4086 VERIFY(nvpair_value_nvlist(p2, &attrs) == 0);
4087 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
4091 if (nvpair_type(p1) != nvpair_type(p2))
4094 if (nvpair_type(p1) == DATA_TYPE_STRING) {
4095 char *valstr1, *valstr2;
4097 VERIFY(nvpair_value_string(p1, (char **)&valstr1) == 0);
4098 VERIFY(nvpair_value_string(p2, (char **)&valstr2) == 0);
4099 return (strcmp(valstr1, valstr2) == 0);
4101 uint64_t intval1, intval2;
4103 VERIFY(nvpair_value_uint64(p1, &intval1) == 0);
4104 VERIFY(nvpair_value_uint64(p2, &intval2) == 0);
4105 return (intval1 == intval2);
4110 * Remove properties from props if they are not going to change (as determined
4111 * by comparison with origprops). Remove them from origprops as well, since we
4112 * do not need to clear or restore properties that won't change.
4115 props_reduce(nvlist_t *props, nvlist_t *origprops)
4117 nvpair_t *pair, *next_pair;
4119 if (origprops == NULL)
4120 return; /* all props need to be received */
4122 pair = nvlist_next_nvpair(props, NULL);
4123 while (pair != NULL) {
4124 const char *propname = nvpair_name(pair);
4127 next_pair = nvlist_next_nvpair(props, pair);
4129 if ((nvlist_lookup_nvpair(origprops, propname,
4130 &match) != 0) || !propval_equals(pair, match))
4131 goto next; /* need to set received value */
4133 /* don't clear the existing received value */
4134 (void) nvlist_remove_nvpair(origprops, match);
4135 /* don't bother receiving the property */
4136 (void) nvlist_remove_nvpair(props, pair);
4143 * Extract properties that cannot be set PRIOR to the receipt of a dataset.
4144 * For example, refquota cannot be set until after the receipt of a dataset,
4145 * because in replication streams, an older/earlier snapshot may exceed the
4146 * refquota. We want to receive the older/earlier snapshot, but setting
4147 * refquota pre-receipt will set the dsl's ACTUAL quota, which will prevent
4148 * the older/earlier snapshot from being received (with EDQUOT).
4150 * The ZFS test "zfs_receive_011_pos" demonstrates such a scenario.
4152 * libzfs will need to be judicious handling errors encountered by props
4153 * extracted by this function.
4156 extract_delay_props(nvlist_t *props)
4158 nvlist_t *delayprops;
4159 nvpair_t *nvp, *tmp;
4160 static const zfs_prop_t delayable[] = { ZFS_PROP_REFQUOTA, 0 };
4163 VERIFY(nvlist_alloc(&delayprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
4165 for (nvp = nvlist_next_nvpair(props, NULL); nvp != NULL;
4166 nvp = nvlist_next_nvpair(props, nvp)) {
4168 * strcmp() is safe because zfs_prop_to_name() always returns
4171 for (i = 0; delayable[i] != 0; i++) {
4172 if (strcmp(zfs_prop_to_name(delayable[i]),
4173 nvpair_name(nvp)) == 0) {
4177 if (delayable[i] != 0) {
4178 tmp = nvlist_prev_nvpair(props, nvp);
4179 VERIFY(nvlist_add_nvpair(delayprops, nvp) == 0);
4180 VERIFY(nvlist_remove_nvpair(props, nvp) == 0);
4185 if (nvlist_empty(delayprops)) {
4186 nvlist_free(delayprops);
4189 return (delayprops);
4193 static boolean_t zfs_ioc_recv_inject_err;
4197 * nvlist 'errors' is always allocated. It will contain descriptions of
4198 * encountered errors, if any. It's the callers responsibility to free.
4201 zfs_ioc_recv_impl(char *tofs, char *tosnap, char *origin, nvlist_t *recvprops,
4202 nvlist_t *localprops, boolean_t force, boolean_t resumable, int input_fd,
4203 dmu_replay_record_t *begin_record, int cleanup_fd, uint64_t *read_bytes,
4204 uint64_t *errflags, uint64_t *action_handle, nvlist_t **errors)
4206 dmu_recv_cookie_t drc;
4208 int props_error = 0;
4210 nvlist_t *delayprops = NULL; /* sent properties applied post-receive */
4211 nvlist_t *origprops = NULL; /* existing properties */
4212 nvlist_t *origrecvd = NULL; /* existing received properties */
4213 boolean_t first_recvd_props = B_FALSE;
4218 *errors = fnvlist_alloc();
4220 input_fp = getf(input_fd);
4221 if (input_fp == NULL)
4222 return (SET_ERROR(EBADF));
4224 error = dmu_recv_begin(tofs, tosnap,
4225 begin_record, force, resumable, origin, &drc);
4230 * Set properties before we receive the stream so that they are applied
4231 * to the new data. Note that we must call dmu_recv_stream() if
4232 * dmu_recv_begin() succeeds.
4234 if (recvprops != NULL && !drc.drc_newfs) {
4235 if (spa_version(dsl_dataset_get_spa(drc.drc_ds)) >=
4236 SPA_VERSION_RECVD_PROPS &&
4237 !dsl_prop_get_hasrecvd(tofs))
4238 first_recvd_props = B_TRUE;
4241 * If new received properties are supplied, they are to
4242 * completely replace the existing received properties, so stash
4243 * away the existing ones.
4245 if (dsl_prop_get_received(tofs, &origrecvd) == 0) {
4246 nvlist_t *errlist = NULL;
4248 * Don't bother writing a property if its value won't
4249 * change (and avoid the unnecessary security checks).
4251 * The first receive after SPA_VERSION_RECVD_PROPS is a
4252 * special case where we blow away all local properties
4255 if (!first_recvd_props)
4256 props_reduce(recvprops, origrecvd);
4257 if (zfs_check_clearable(tofs, origrecvd, &errlist) != 0)
4258 (void) nvlist_merge(*errors, errlist, 0);
4259 nvlist_free(errlist);
4261 if (clear_received_props(tofs, origrecvd,
4262 first_recvd_props ? NULL : recvprops) != 0)
4263 *errflags |= ZPROP_ERR_NOCLEAR;
4265 *errflags |= ZPROP_ERR_NOCLEAR;
4270 * Stash away existing properties so we can restore them on error unless
4271 * we're doing the first receive after SPA_VERSION_RECVD_PROPS, in which
4272 * case "origrecvd" will take care of that.
4274 if (localprops != NULL && !drc.drc_newfs && !first_recvd_props) {
4276 if (dmu_objset_hold(tofs, FTAG, &os) == 0) {
4277 if (dsl_prop_get_all(os, &origprops) != 0) {
4278 *errflags |= ZPROP_ERR_NOCLEAR;
4280 dmu_objset_rele(os, FTAG);
4282 *errflags |= ZPROP_ERR_NOCLEAR;
4286 if (recvprops != NULL) {
4287 props_error = dsl_prop_set_hasrecvd(tofs);
4289 if (props_error == 0) {
4290 delayprops = extract_delay_props(recvprops);
4291 (void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_RECEIVED,
4292 recvprops, *errors);
4296 if (localprops != NULL) {
4297 nvlist_t *oprops = fnvlist_alloc();
4298 nvlist_t *xprops = fnvlist_alloc();
4299 nvpair_t *nvp = NULL;
4301 while ((nvp = nvlist_next_nvpair(localprops, nvp)) != NULL) {
4302 if (nvpair_type(nvp) == DATA_TYPE_BOOLEAN) {
4304 const char *name = nvpair_name(nvp);
4305 zfs_prop_t prop = zfs_name_to_prop(name);
4306 if (prop != ZPROP_INVAL) {
4307 if (!zfs_prop_inheritable(prop))
4309 } else if (!zfs_prop_user(name))
4311 fnvlist_add_boolean(xprops, name);
4313 /* -o property=value */
4314 fnvlist_add_nvpair(oprops, nvp);
4317 (void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_LOCAL,
4319 (void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_INHERITED,
4322 nvlist_free(oprops);
4323 nvlist_free(xprops);
4326 off = input_fp->f_offset;
4327 error = dmu_recv_stream(&drc, input_fp->f_vnode, &off, cleanup_fd,
4331 zfsvfs_t *zfsvfs = NULL;
4332 zvol_state_t *zv = NULL;
4334 if (getzfsvfs(tofs, &zfsvfs) == 0) {
4339 ds = dmu_objset_ds(zfsvfs->z_os);
4340 error = zfs_suspend_fs(zfsvfs);
4342 * If the suspend fails, then the recv_end will
4343 * likely also fail, and clean up after itself.
4345 end_err = dmu_recv_end(&drc, zfsvfs);
4347 error = zfs_resume_fs(zfsvfs, ds);
4348 error = error ? error : end_err;
4349 deactivate_super(zfsvfs->z_sb);
4350 } else if ((zv = zvol_suspend(tofs)) != NULL) {
4351 error = dmu_recv_end(&drc, zvol_tag(zv));
4354 error = dmu_recv_end(&drc, NULL);
4357 /* Set delayed properties now, after we're done receiving. */
4358 if (delayprops != NULL && error == 0) {
4359 (void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_RECEIVED,
4360 delayprops, *errors);
4364 if (delayprops != NULL) {
4366 * Merge delayed props back in with initial props, in case
4367 * we're DEBUG and zfs_ioc_recv_inject_err is set (which means
4368 * we have to make sure clear_received_props() includes
4369 * the delayed properties).
4371 * Since zfs_ioc_recv_inject_err is only in DEBUG kernels,
4372 * using ASSERT() will be just like a VERIFY.
4374 ASSERT(nvlist_merge(recvprops, delayprops, 0) == 0);
4375 nvlist_free(delayprops);
4379 *read_bytes = off - input_fp->f_offset;
4380 if (VOP_SEEK(input_fp->f_vnode, input_fp->f_offset, &off, NULL) == 0)
4381 input_fp->f_offset = off;
4384 if (zfs_ioc_recv_inject_err) {
4385 zfs_ioc_recv_inject_err = B_FALSE;
4391 * On error, restore the original props.
4393 if (error != 0 && recvprops != NULL && !drc.drc_newfs) {
4394 if (clear_received_props(tofs, recvprops, NULL) != 0) {
4396 * We failed to clear the received properties.
4397 * Since we may have left a $recvd value on the
4398 * system, we can't clear the $hasrecvd flag.
4400 *errflags |= ZPROP_ERR_NORESTORE;
4401 } else if (first_recvd_props) {
4402 dsl_prop_unset_hasrecvd(tofs);
4405 if (origrecvd == NULL && !drc.drc_newfs) {
4406 /* We failed to stash the original properties. */
4407 *errflags |= ZPROP_ERR_NORESTORE;
4411 * dsl_props_set() will not convert RECEIVED to LOCAL on or
4412 * after SPA_VERSION_RECVD_PROPS, so we need to specify LOCAL
4413 * explicitly if we're restoring local properties cleared in the
4414 * first new-style receive.
4416 if (origrecvd != NULL &&
4417 zfs_set_prop_nvlist(tofs, (first_recvd_props ?
4418 ZPROP_SRC_LOCAL : ZPROP_SRC_RECEIVED),
4419 origrecvd, NULL) != 0) {
4421 * We stashed the original properties but failed to
4424 *errflags |= ZPROP_ERR_NORESTORE;
4427 if (error != 0 && localprops != NULL && !drc.drc_newfs &&
4428 !first_recvd_props) {
4430 nvlist_t *inheritprops;
4433 if (origprops == NULL) {
4434 /* We failed to stash the original properties. */
4435 *errflags |= ZPROP_ERR_NORESTORE;
4439 /* Restore original props */
4440 setprops = fnvlist_alloc();
4441 inheritprops = fnvlist_alloc();
4443 while ((nvp = nvlist_next_nvpair(localprops, nvp)) != NULL) {
4444 const char *name = nvpair_name(nvp);
4448 if (!nvlist_exists(origprops, name)) {
4450 * Property was not present or was explicitly
4451 * inherited before the receive, restore this.
4453 fnvlist_add_boolean(inheritprops, name);
4456 attrs = fnvlist_lookup_nvlist(origprops, name);
4457 source = fnvlist_lookup_string(attrs, ZPROP_SOURCE);
4459 /* Skip received properties */
4460 if (strcmp(source, ZPROP_SOURCE_VAL_RECVD) == 0)
4463 if (strcmp(source, tofs) == 0) {
4464 /* Property was locally set */
4465 fnvlist_add_nvlist(setprops, name, attrs);
4467 /* Property was implicitly inherited */
4468 fnvlist_add_boolean(inheritprops, name);
4472 if (zfs_set_prop_nvlist(tofs, ZPROP_SRC_LOCAL, setprops,
4474 *errflags |= ZPROP_ERR_NORESTORE;
4475 if (zfs_set_prop_nvlist(tofs, ZPROP_SRC_INHERITED, inheritprops,
4477 *errflags |= ZPROP_ERR_NORESTORE;
4479 nvlist_free(setprops);
4480 nvlist_free(inheritprops);
4484 nvlist_free(origrecvd);
4485 nvlist_free(origprops);
4488 error = props_error;
4495 * zc_name name of containing filesystem (unused)
4496 * zc_nvlist_src{_size} nvlist of properties to apply
4497 * zc_nvlist_conf{_size} nvlist of properties to exclude
4498 * (DATA_TYPE_BOOLEAN) and override (everything else)
4499 * zc_value name of snapshot to create
4500 * zc_string name of clone origin (if DRR_FLAG_CLONE)
4501 * zc_cookie file descriptor to recv from
4502 * zc_begin_record the BEGIN record of the stream (not byteswapped)
4503 * zc_guid force flag
4504 * zc_cleanup_fd cleanup-on-exit file descriptor
4505 * zc_action_handle handle for this guid/ds mapping (or zero on first call)
4508 * zc_cookie number of bytes read
4509 * zc_obj zprop_errflags_t
4510 * zc_action_handle handle for this guid/ds mapping
4511 * zc_nvlist_dst{_size} error for each unapplied received property
4514 zfs_ioc_recv(zfs_cmd_t *zc)
4516 dmu_replay_record_t begin_record;
4517 nvlist_t *errors = NULL;
4518 nvlist_t *recvdprops = NULL;
4519 nvlist_t *localprops = NULL;
4520 char *origin = NULL;
4522 char tofs[ZFS_MAX_DATASET_NAME_LEN];
4525 if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
4526 strchr(zc->zc_value, '@') == NULL ||
4527 strchr(zc->zc_value, '%'))
4528 return (SET_ERROR(EINVAL));
4530 (void) strlcpy(tofs, zc->zc_value, sizeof (tofs));
4531 tosnap = strchr(tofs, '@');
4534 if (zc->zc_nvlist_src != 0 &&
4535 (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
4536 zc->zc_iflags, &recvdprops)) != 0)
4539 if (zc->zc_nvlist_conf != 0 &&
4540 (error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
4541 zc->zc_iflags, &localprops)) != 0)
4544 if (zc->zc_string[0])
4545 origin = zc->zc_string;
4547 begin_record.drr_type = DRR_BEGIN;
4548 begin_record.drr_payloadlen = 0;
4549 begin_record.drr_u.drr_begin = zc->zc_begin_record;
4551 error = zfs_ioc_recv_impl(tofs, tosnap, origin, recvdprops, localprops,
4552 zc->zc_guid, B_FALSE, zc->zc_cookie, &begin_record,
4553 zc->zc_cleanup_fd, &zc->zc_cookie, &zc->zc_obj,
4554 &zc->zc_action_handle, &errors);
4555 nvlist_free(recvdprops);
4556 nvlist_free(localprops);
4559 * Now that all props, initial and delayed, are set, report the prop
4560 * errors to the caller.
4562 if (zc->zc_nvlist_dst_size != 0 && errors != NULL &&
4563 (nvlist_smush(errors, zc->zc_nvlist_dst_size) != 0 ||
4564 put_nvlist(zc, errors) != 0)) {
4566 * Caller made zc->zc_nvlist_dst less than the minimum expected
4567 * size or supplied an invalid address.
4569 error = SET_ERROR(EINVAL);
4572 nvlist_free(errors);
4579 * "snapname" -> full name of the snapshot to create
4580 * (optional) "props" -> received properties to set (nvlist)
4581 * (optional) "localprops" -> override and exclude properties (nvlist)
4582 * (optional) "origin" -> name of clone origin (DRR_FLAG_CLONE)
4583 * "begin_record" -> non-byteswapped dmu_replay_record_t
4584 * "input_fd" -> file descriptor to read stream from (int32)
4585 * (optional) "force" -> force flag (value ignored)
4586 * (optional) "resumable" -> resumable flag (value ignored)
4587 * (optional) "cleanup_fd" -> cleanup-on-exit file descriptor
4588 * (optional) "action_handle" -> handle for this guid/ds mapping
4592 * "read_bytes" -> number of bytes read
4593 * "error_flags" -> zprop_errflags_t
4594 * "action_handle" -> handle for this guid/ds mapping
4595 * "errors" -> error for each unapplied received property (nvlist)
4599 zfs_ioc_recv_new(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
4601 dmu_replay_record_t *begin_record;
4602 uint_t begin_record_size;
4603 nvlist_t *errors = NULL;
4604 nvlist_t *recvprops = NULL;
4605 nvlist_t *localprops = NULL;
4606 char *snapname = NULL;
4607 char *origin = NULL;
4609 char tofs[ZFS_MAX_DATASET_NAME_LEN];
4611 boolean_t resumable;
4612 uint64_t action_handle = 0;
4613 uint64_t read_bytes = 0;
4614 uint64_t errflags = 0;
4616 int cleanup_fd = -1;
4619 error = nvlist_lookup_string(innvl, "snapname", &snapname);
4621 return (SET_ERROR(EINVAL));
4623 if (dataset_namecheck(snapname, NULL, NULL) != 0 ||
4624 strchr(snapname, '@') == NULL ||
4625 strchr(snapname, '%'))
4626 return (SET_ERROR(EINVAL));
4628 (void) strcpy(tofs, snapname);
4629 tosnap = strchr(tofs, '@');
4632 error = nvlist_lookup_string(innvl, "origin", &origin);
4633 if (error && error != ENOENT)
4636 error = nvlist_lookup_byte_array(innvl, "begin_record",
4637 (uchar_t **)&begin_record, &begin_record_size);
4638 if (error != 0 || begin_record_size != sizeof (*begin_record))
4639 return (SET_ERROR(EINVAL));
4641 error = nvlist_lookup_int32(innvl, "input_fd", &input_fd);
4643 return (SET_ERROR(EINVAL));
4645 force = nvlist_exists(innvl, "force");
4646 resumable = nvlist_exists(innvl, "resumable");
4648 error = nvlist_lookup_int32(innvl, "cleanup_fd", &cleanup_fd);
4649 if (error && error != ENOENT)
4652 error = nvlist_lookup_uint64(innvl, "action_handle", &action_handle);
4653 if (error && error != ENOENT)
4656 /* we still use "props" here for backwards compatibility */
4657 error = nvlist_lookup_nvlist(innvl, "props", &recvprops);
4658 if (error && error != ENOENT)
4661 error = nvlist_lookup_nvlist(innvl, "localprops", &localprops);
4662 if (error && error != ENOENT)
4665 error = zfs_ioc_recv_impl(tofs, tosnap, origin, recvprops, localprops,
4666 force, resumable, input_fd, begin_record, cleanup_fd, &read_bytes,
4667 &errflags, &action_handle, &errors);
4669 fnvlist_add_uint64(outnvl, "read_bytes", read_bytes);
4670 fnvlist_add_uint64(outnvl, "error_flags", errflags);
4671 fnvlist_add_uint64(outnvl, "action_handle", action_handle);
4672 fnvlist_add_nvlist(outnvl, "errors", errors);
4674 nvlist_free(errors);
4675 nvlist_free(recvprops);
4676 nvlist_free(localprops);
4683 * zc_name name of snapshot to send
4684 * zc_cookie file descriptor to send stream to
4685 * zc_obj fromorigin flag (mutually exclusive with zc_fromobj)
4686 * zc_sendobj objsetid of snapshot to send
4687 * zc_fromobj objsetid of incremental fromsnap (may be zero)
4688 * zc_guid if set, estimate size of stream only. zc_cookie is ignored.
4689 * output size in zc_objset_type.
4690 * zc_flags lzc_send_flags
4693 * zc_objset_type estimated size, if zc_guid is set
4696 zfs_ioc_send(zfs_cmd_t *zc)
4700 boolean_t estimate = (zc->zc_guid != 0);
4701 boolean_t embedok = (zc->zc_flags & 0x1);
4702 boolean_t large_block_ok = (zc->zc_flags & 0x2);
4703 boolean_t compressok = (zc->zc_flags & 0x4);
4705 if (zc->zc_obj != 0) {
4707 dsl_dataset_t *tosnap;
4709 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
4713 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &tosnap);
4715 dsl_pool_rele(dp, FTAG);
4719 if (dsl_dir_is_clone(tosnap->ds_dir))
4721 dsl_dir_phys(tosnap->ds_dir)->dd_origin_obj;
4722 dsl_dataset_rele(tosnap, FTAG);
4723 dsl_pool_rele(dp, FTAG);
4728 dsl_dataset_t *tosnap;
4729 dsl_dataset_t *fromsnap = NULL;
4731 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
4735 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &tosnap);
4737 dsl_pool_rele(dp, FTAG);
4741 if (zc->zc_fromobj != 0) {
4742 error = dsl_dataset_hold_obj(dp, zc->zc_fromobj,
4745 dsl_dataset_rele(tosnap, FTAG);
4746 dsl_pool_rele(dp, FTAG);
4751 error = dmu_send_estimate(tosnap, fromsnap, compressok,
4752 &zc->zc_objset_type);
4754 if (fromsnap != NULL)
4755 dsl_dataset_rele(fromsnap, FTAG);
4756 dsl_dataset_rele(tosnap, FTAG);
4757 dsl_pool_rele(dp, FTAG);
4759 file_t *fp = getf(zc->zc_cookie);
4761 return (SET_ERROR(EBADF));
4764 error = dmu_send_obj(zc->zc_name, zc->zc_sendobj,
4765 zc->zc_fromobj, embedok, large_block_ok, compressok,
4766 zc->zc_cookie, fp->f_vnode, &off);
4768 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
4770 releasef(zc->zc_cookie);
4777 * zc_name name of snapshot on which to report progress
4778 * zc_cookie file descriptor of send stream
4781 * zc_cookie number of bytes written in send stream thus far
4784 zfs_ioc_send_progress(zfs_cmd_t *zc)
4788 dmu_sendarg_t *dsp = NULL;
4791 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
4795 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &ds);
4797 dsl_pool_rele(dp, FTAG);
4801 mutex_enter(&ds->ds_sendstream_lock);
4804 * Iterate over all the send streams currently active on this dataset.
4805 * If there's one which matches the specified file descriptor _and_ the
4806 * stream was started by the current process, return the progress of
4810 for (dsp = list_head(&ds->ds_sendstreams); dsp != NULL;
4811 dsp = list_next(&ds->ds_sendstreams, dsp)) {
4812 if (dsp->dsa_outfd == zc->zc_cookie &&
4813 dsp->dsa_proc->group_leader == curproc->group_leader)
4818 zc->zc_cookie = *(dsp->dsa_off);
4820 error = SET_ERROR(ENOENT);
4822 mutex_exit(&ds->ds_sendstream_lock);
4823 dsl_dataset_rele(ds, FTAG);
4824 dsl_pool_rele(dp, FTAG);
4829 zfs_ioc_inject_fault(zfs_cmd_t *zc)
4833 error = zio_inject_fault(zc->zc_name, (int)zc->zc_guid, &id,
4834 &zc->zc_inject_record);
4837 zc->zc_guid = (uint64_t)id;
4843 zfs_ioc_clear_fault(zfs_cmd_t *zc)
4845 return (zio_clear_fault((int)zc->zc_guid));
4849 zfs_ioc_inject_list_next(zfs_cmd_t *zc)
4851 int id = (int)zc->zc_guid;
4854 error = zio_inject_list_next(&id, zc->zc_name, sizeof (zc->zc_name),
4855 &zc->zc_inject_record);
4863 zfs_ioc_error_log(zfs_cmd_t *zc)
4867 size_t count = (size_t)zc->zc_nvlist_dst_size;
4869 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
4872 error = spa_get_errlog(spa, (void *)(uintptr_t)zc->zc_nvlist_dst,
4875 zc->zc_nvlist_dst_size = count;
4877 zc->zc_nvlist_dst_size = spa_get_errlog_size(spa);
4879 spa_close(spa, FTAG);
4885 zfs_ioc_clear(zfs_cmd_t *zc)
4892 * On zpool clear we also fix up missing slogs
4894 mutex_enter(&spa_namespace_lock);
4895 spa = spa_lookup(zc->zc_name);
4897 mutex_exit(&spa_namespace_lock);
4898 return (SET_ERROR(EIO));
4900 if (spa_get_log_state(spa) == SPA_LOG_MISSING) {
4901 /* we need to let spa_open/spa_load clear the chains */
4902 spa_set_log_state(spa, SPA_LOG_CLEAR);
4904 spa->spa_last_open_failed = 0;
4905 mutex_exit(&spa_namespace_lock);
4907 if (zc->zc_cookie & ZPOOL_NO_REWIND) {
4908 error = spa_open(zc->zc_name, &spa, FTAG);
4911 nvlist_t *config = NULL;
4913 if (zc->zc_nvlist_src == 0)
4914 return (SET_ERROR(EINVAL));
4916 if ((error = get_nvlist(zc->zc_nvlist_src,
4917 zc->zc_nvlist_src_size, zc->zc_iflags, &policy)) == 0) {
4918 error = spa_open_rewind(zc->zc_name, &spa, FTAG,
4920 if (config != NULL) {
4923 if ((err = put_nvlist(zc, config)) != 0)
4925 nvlist_free(config);
4927 nvlist_free(policy);
4934 spa_vdev_state_enter(spa, SCL_NONE);
4936 if (zc->zc_guid == 0) {
4939 vd = spa_lookup_by_guid(spa, zc->zc_guid, B_TRUE);
4941 (void) spa_vdev_state_exit(spa, NULL, ENODEV);
4942 spa_close(spa, FTAG);
4943 return (SET_ERROR(ENODEV));
4947 vdev_clear(spa, vd);
4949 (void) spa_vdev_state_exit(spa, spa->spa_root_vdev, 0);
4952 * Resume any suspended I/Os.
4954 if (zio_resume(spa) != 0)
4955 error = SET_ERROR(EIO);
4957 spa_close(spa, FTAG);
4963 zfs_ioc_pool_reopen(zfs_cmd_t *zc)
4968 error = spa_open(zc->zc_name, &spa, FTAG);
4972 spa_vdev_state_enter(spa, SCL_NONE);
4975 * If a resilver is already in progress then set the
4976 * spa_scrub_reopen flag to B_TRUE so that we don't restart
4977 * the scan as a side effect of the reopen. Otherwise, let
4978 * vdev_open() decided if a resilver is required.
4980 spa->spa_scrub_reopen = dsl_scan_resilvering(spa->spa_dsl_pool);
4981 vdev_reopen(spa->spa_root_vdev);
4982 spa->spa_scrub_reopen = B_FALSE;
4984 (void) spa_vdev_state_exit(spa, NULL, 0);
4985 spa_close(spa, FTAG);
4990 * zc_name name of filesystem
4993 * zc_string name of conflicting snapshot, if there is one
4996 zfs_ioc_promote(zfs_cmd_t *zc)
4999 dsl_dataset_t *ds, *ods;
5000 char origin[ZFS_MAX_DATASET_NAME_LEN];
5004 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
5008 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &ds);
5010 dsl_pool_rele(dp, FTAG);
5014 if (!dsl_dir_is_clone(ds->ds_dir)) {
5015 dsl_dataset_rele(ds, FTAG);
5016 dsl_pool_rele(dp, FTAG);
5017 return (SET_ERROR(EINVAL));
5020 error = dsl_dataset_hold_obj(dp,
5021 dsl_dir_phys(ds->ds_dir)->dd_origin_obj, FTAG, &ods);
5023 dsl_dataset_rele(ds, FTAG);
5024 dsl_pool_rele(dp, FTAG);
5028 dsl_dataset_name(ods, origin);
5029 dsl_dataset_rele(ods, FTAG);
5030 dsl_dataset_rele(ds, FTAG);
5031 dsl_pool_rele(dp, FTAG);
5034 * We don't need to unmount *all* the origin fs's snapshots, but
5037 cp = strchr(origin, '@');
5040 (void) dmu_objset_find(origin,
5041 zfs_unmount_snap_cb, NULL, DS_FIND_SNAPSHOTS);
5042 return (dsl_dataset_promote(zc->zc_name, zc->zc_string));
5046 * Retrieve a single {user|group}{used|quota}@... property.
5049 * zc_name name of filesystem
5050 * zc_objset_type zfs_userquota_prop_t
5051 * zc_value domain name (eg. "S-1-234-567-89")
5052 * zc_guid RID/UID/GID
5055 * zc_cookie property value
5058 zfs_ioc_userspace_one(zfs_cmd_t *zc)
5063 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
5064 return (SET_ERROR(EINVAL));
5066 error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE);
5070 error = zfs_userspace_one(zfsvfs,
5071 zc->zc_objset_type, zc->zc_value, zc->zc_guid, &zc->zc_cookie);
5072 zfsvfs_rele(zfsvfs, FTAG);
5079 * zc_name name of filesystem
5080 * zc_cookie zap cursor
5081 * zc_objset_type zfs_userquota_prop_t
5082 * zc_nvlist_dst[_size] buffer to fill (not really an nvlist)
5085 * zc_nvlist_dst[_size] data buffer (array of zfs_useracct_t)
5086 * zc_cookie zap cursor
5089 zfs_ioc_userspace_many(zfs_cmd_t *zc)
5092 int bufsize = zc->zc_nvlist_dst_size;
5097 return (SET_ERROR(ENOMEM));
5099 error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE);
5103 buf = vmem_alloc(bufsize, KM_SLEEP);
5105 error = zfs_userspace_many(zfsvfs, zc->zc_objset_type, &zc->zc_cookie,
5106 buf, &zc->zc_nvlist_dst_size);
5109 error = xcopyout(buf,
5110 (void *)(uintptr_t)zc->zc_nvlist_dst,
5111 zc->zc_nvlist_dst_size);
5113 vmem_free(buf, bufsize);
5114 zfsvfs_rele(zfsvfs, FTAG);
5121 * zc_name name of filesystem
5127 zfs_ioc_userspace_upgrade(zfs_cmd_t *zc)
5133 if (getzfsvfs(zc->zc_name, &zfsvfs) == 0) {
5134 if (!dmu_objset_userused_enabled(zfsvfs->z_os)) {
5136 * If userused is not enabled, it may be because the
5137 * objset needs to be closed & reopened (to grow the
5138 * objset_phys_t). Suspend/resume the fs will do that.
5142 ds = dmu_objset_ds(zfsvfs->z_os);
5143 error = zfs_suspend_fs(zfsvfs);
5145 dmu_objset_refresh_ownership(zfsvfs->z_os,
5147 error = zfs_resume_fs(zfsvfs, ds);
5151 error = dmu_objset_userspace_upgrade(zfsvfs->z_os);
5152 deactivate_super(zfsvfs->z_sb);
5154 /* XXX kind of reading contents without owning */
5155 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
5159 error = dmu_objset_userspace_upgrade(os);
5160 dmu_objset_rele(os, FTAG);
5168 * zc_name name of filesystem
5174 zfs_ioc_userobjspace_upgrade(zfs_cmd_t *zc)
5179 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
5183 dsl_dataset_long_hold(dmu_objset_ds(os), FTAG);
5184 dsl_pool_rele(dmu_objset_pool(os), FTAG);
5186 if (dmu_objset_userobjspace_upgradable(os)) {
5187 mutex_enter(&os->os_upgrade_lock);
5188 if (os->os_upgrade_id == 0) {
5189 /* clear potential error code and retry */
5190 os->os_upgrade_status = 0;
5191 mutex_exit(&os->os_upgrade_lock);
5193 dmu_objset_userobjspace_upgrade(os);
5195 mutex_exit(&os->os_upgrade_lock);
5198 taskq_wait_id(os->os_spa->spa_upgrade_taskq, os->os_upgrade_id);
5199 error = os->os_upgrade_status;
5202 dsl_dataset_long_rele(dmu_objset_ds(os), FTAG);
5203 dsl_dataset_rele(dmu_objset_ds(os), FTAG);
5209 zfs_ioc_share(zfs_cmd_t *zc)
5211 return (SET_ERROR(ENOSYS));
5214 ace_t full_access[] = {
5215 {(uid_t)-1, ACE_ALL_PERMS, ACE_EVERYONE, 0}
5220 * zc_name name of containing filesystem
5221 * zc_obj object # beyond which we want next in-use object #
5224 * zc_obj next in-use object #
5227 zfs_ioc_next_obj(zfs_cmd_t *zc)
5229 objset_t *os = NULL;
5232 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
5236 error = dmu_object_next(os, &zc->zc_obj, B_FALSE, 0);
5238 dmu_objset_rele(os, FTAG);
5244 * zc_name name of filesystem
5245 * zc_value prefix name for snapshot
5246 * zc_cleanup_fd cleanup-on-exit file descriptor for calling process
5249 * zc_value short name of new snapshot
5252 zfs_ioc_tmp_snapshot(zfs_cmd_t *zc)
5259 error = zfs_onexit_fd_hold(zc->zc_cleanup_fd, &minor);
5263 snap_name = kmem_asprintf("%s-%016llx", zc->zc_value,
5264 (u_longlong_t)ddi_get_lbolt64());
5265 hold_name = kmem_asprintf("%%%s", zc->zc_value);
5267 error = dsl_dataset_snapshot_tmp(zc->zc_name, snap_name, minor,
5270 (void) strlcpy(zc->zc_value, snap_name,
5271 sizeof (zc->zc_value));
5274 zfs_onexit_fd_rele(zc->zc_cleanup_fd);
5280 * zc_name name of "to" snapshot
5281 * zc_value name of "from" snapshot
5282 * zc_cookie file descriptor to write diff data on
5285 * dmu_diff_record_t's to the file descriptor
5288 zfs_ioc_diff(zfs_cmd_t *zc)
5294 fp = getf(zc->zc_cookie);
5296 return (SET_ERROR(EBADF));
5300 error = dmu_diff(zc->zc_name, zc->zc_value, fp->f_vnode, &off);
5302 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
5304 releasef(zc->zc_cookie);
5310 * Remove all ACL files in shares dir
5312 #ifdef HAVE_SMB_SHARE
5314 zfs_smb_acl_purge(znode_t *dzp)
5317 zap_attribute_t zap;
5318 zfsvfs_t *zfsvfs = ZTOZSB(dzp);
5321 for (zap_cursor_init(&zc, zfsvfs->z_os, dzp->z_id);
5322 (error = zap_cursor_retrieve(&zc, &zap)) == 0;
5323 zap_cursor_advance(&zc)) {
5324 if ((error = VOP_REMOVE(ZTOV(dzp), zap.za_name, kcred,
5328 zap_cursor_fini(&zc);
5331 #endif /* HAVE_SMB_SHARE */
5334 zfs_ioc_smb_acl(zfs_cmd_t *zc)
5336 #ifdef HAVE_SMB_SHARE
5339 vnode_t *resourcevp = NULL;
5348 if ((error = lookupname(zc->zc_value, UIO_SYSSPACE,
5349 NO_FOLLOW, NULL, &vp)) != 0)
5352 /* Now make sure mntpnt and dataset are ZFS */
5354 if (vp->v_vfsp->vfs_fstype != zfsfstype ||
5355 (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource),
5356 zc->zc_name) != 0)) {
5358 return (SET_ERROR(EINVAL));
5362 zfsvfs = ZTOZSB(dzp);
5366 * Create share dir if its missing.
5368 mutex_enter(&zfsvfs->z_lock);
5369 if (zfsvfs->z_shares_dir == 0) {
5372 tx = dmu_tx_create(zfsvfs->z_os);
5373 dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, TRUE,
5375 dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL);
5376 error = dmu_tx_assign(tx, TXG_WAIT);
5380 error = zfs_create_share_dir(zfsvfs, tx);
5384 mutex_exit(&zfsvfs->z_lock);
5390 mutex_exit(&zfsvfs->z_lock);
5392 ASSERT(zfsvfs->z_shares_dir);
5393 if ((error = zfs_zget(zfsvfs, zfsvfs->z_shares_dir, &sharedir)) != 0) {
5399 switch (zc->zc_cookie) {
5400 case ZFS_SMB_ACL_ADD:
5401 vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE;
5402 vattr.va_mode = S_IFREG|0777;
5406 vsec.vsa_mask = VSA_ACE;
5407 vsec.vsa_aclentp = &full_access;
5408 vsec.vsa_aclentsz = sizeof (full_access);
5409 vsec.vsa_aclcnt = 1;
5411 error = VOP_CREATE(ZTOV(sharedir), zc->zc_string,
5412 &vattr, EXCL, 0, &resourcevp, kcred, 0, NULL, &vsec);
5414 VN_RELE(resourcevp);
5417 case ZFS_SMB_ACL_REMOVE:
5418 error = VOP_REMOVE(ZTOV(sharedir), zc->zc_string, kcred,
5422 case ZFS_SMB_ACL_RENAME:
5423 if ((error = get_nvlist(zc->zc_nvlist_src,
5424 zc->zc_nvlist_src_size, zc->zc_iflags, &nvlist)) != 0) {
5426 VN_RELE(ZTOV(sharedir));
5430 if (nvlist_lookup_string(nvlist, ZFS_SMB_ACL_SRC, &src) ||
5431 nvlist_lookup_string(nvlist, ZFS_SMB_ACL_TARGET,
5434 VN_RELE(ZTOV(sharedir));
5436 nvlist_free(nvlist);
5439 error = VOP_RENAME(ZTOV(sharedir), src, ZTOV(sharedir), target,
5441 nvlist_free(nvlist);
5444 case ZFS_SMB_ACL_PURGE:
5445 error = zfs_smb_acl_purge(sharedir);
5449 error = SET_ERROR(EINVAL);
5454 VN_RELE(ZTOV(sharedir));
5460 return (SET_ERROR(ENOTSUP));
5461 #endif /* HAVE_SMB_SHARE */
5466 * "holds" -> { snapname -> holdname (string), ... }
5467 * (optional) "cleanup_fd" -> fd (int32)
5471 * snapname -> error value (int32)
5477 zfs_ioc_hold(const char *pool, nvlist_t *args, nvlist_t *errlist)
5481 int cleanup_fd = -1;
5485 error = nvlist_lookup_nvlist(args, "holds", &holds);
5487 return (SET_ERROR(EINVAL));
5489 /* make sure the user didn't pass us any invalid (empty) tags */
5490 for (pair = nvlist_next_nvpair(holds, NULL); pair != NULL;
5491 pair = nvlist_next_nvpair(holds, pair)) {
5494 error = nvpair_value_string(pair, &htag);
5496 return (SET_ERROR(error));
5498 if (strlen(htag) == 0)
5499 return (SET_ERROR(EINVAL));
5502 if (nvlist_lookup_int32(args, "cleanup_fd", &cleanup_fd) == 0) {
5503 error = zfs_onexit_fd_hold(cleanup_fd, &minor);
5508 error = dsl_dataset_user_hold(holds, minor, errlist);
5510 zfs_onexit_fd_rele(cleanup_fd);
5515 * innvl is not used.
5518 * holdname -> time added (uint64 seconds since epoch)
5524 zfs_ioc_get_holds(const char *snapname, nvlist_t *args, nvlist_t *outnvl)
5526 ASSERT3P(args, ==, NULL);
5527 return (dsl_dataset_get_holds(snapname, outnvl));
5532 * snapname -> { holdname, ... }
5537 * snapname -> error value (int32)
5543 zfs_ioc_release(const char *pool, nvlist_t *holds, nvlist_t *errlist)
5545 return (dsl_dataset_user_release(holds, errlist));
5550 * zc_guid flags (ZEVENT_NONBLOCK)
5551 * zc_cleanup_fd zevent file descriptor
5554 * zc_nvlist_dst next nvlist event
5555 * zc_cookie dropped events since last get
5558 zfs_ioc_events_next(zfs_cmd_t *zc)
5561 nvlist_t *event = NULL;
5563 uint64_t dropped = 0;
5566 error = zfs_zevent_fd_hold(zc->zc_cleanup_fd, &minor, &ze);
5571 error = zfs_zevent_next(ze, &event,
5572 &zc->zc_nvlist_dst_size, &dropped);
5573 if (event != NULL) {
5574 zc->zc_cookie = dropped;
5575 error = put_nvlist(zc, event);
5579 if (zc->zc_guid & ZEVENT_NONBLOCK)
5582 if ((error == 0) || (error != ENOENT))
5585 error = zfs_zevent_wait(ze);
5590 zfs_zevent_fd_rele(zc->zc_cleanup_fd);
5597 * zc_cookie cleared events count
5600 zfs_ioc_events_clear(zfs_cmd_t *zc)
5604 zfs_zevent_drain_all(&count);
5605 zc->zc_cookie = count;
5612 * zc_guid eid | ZEVENT_SEEK_START | ZEVENT_SEEK_END
5613 * zc_cleanup zevent file descriptor
5616 zfs_ioc_events_seek(zfs_cmd_t *zc)
5622 error = zfs_zevent_fd_hold(zc->zc_cleanup_fd, &minor, &ze);
5626 error = zfs_zevent_seek(ze, zc->zc_guid);
5627 zfs_zevent_fd_rele(zc->zc_cleanup_fd);
5634 * zc_name name of new filesystem or snapshot
5635 * zc_value full name of old snapshot
5638 * zc_cookie space in bytes
5639 * zc_objset_type compressed space in bytes
5640 * zc_perm_action uncompressed space in bytes
5643 zfs_ioc_space_written(zfs_cmd_t *zc)
5647 dsl_dataset_t *new, *old;
5649 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
5652 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &new);
5654 dsl_pool_rele(dp, FTAG);
5657 error = dsl_dataset_hold(dp, zc->zc_value, FTAG, &old);
5659 dsl_dataset_rele(new, FTAG);
5660 dsl_pool_rele(dp, FTAG);
5664 error = dsl_dataset_space_written(old, new, &zc->zc_cookie,
5665 &zc->zc_objset_type, &zc->zc_perm_action);
5666 dsl_dataset_rele(old, FTAG);
5667 dsl_dataset_rele(new, FTAG);
5668 dsl_pool_rele(dp, FTAG);
5674 * "firstsnap" -> snapshot name
5678 * "used" -> space in bytes
5679 * "compressed" -> compressed space in bytes
5680 * "uncompressed" -> uncompressed space in bytes
5684 zfs_ioc_space_snaps(const char *lastsnap, nvlist_t *innvl, nvlist_t *outnvl)
5688 dsl_dataset_t *new, *old;
5690 uint64_t used, comp, uncomp;
5692 if (nvlist_lookup_string(innvl, "firstsnap", &firstsnap) != 0)
5693 return (SET_ERROR(EINVAL));
5695 error = dsl_pool_hold(lastsnap, FTAG, &dp);
5699 error = dsl_dataset_hold(dp, lastsnap, FTAG, &new);
5700 if (error == 0 && !new->ds_is_snapshot) {
5701 dsl_dataset_rele(new, FTAG);
5702 error = SET_ERROR(EINVAL);
5705 dsl_pool_rele(dp, FTAG);
5708 error = dsl_dataset_hold(dp, firstsnap, FTAG, &old);
5709 if (error == 0 && !old->ds_is_snapshot) {
5710 dsl_dataset_rele(old, FTAG);
5711 error = SET_ERROR(EINVAL);
5714 dsl_dataset_rele(new, FTAG);
5715 dsl_pool_rele(dp, FTAG);
5719 error = dsl_dataset_space_wouldfree(old, new, &used, &comp, &uncomp);
5720 dsl_dataset_rele(old, FTAG);
5721 dsl_dataset_rele(new, FTAG);
5722 dsl_pool_rele(dp, FTAG);
5723 fnvlist_add_uint64(outnvl, "used", used);
5724 fnvlist_add_uint64(outnvl, "compressed", comp);
5725 fnvlist_add_uint64(outnvl, "uncompressed", uncomp);
5731 * "fd" -> file descriptor to write stream to (int32)
5732 * (optional) "fromsnap" -> full snap name to send an incremental from
5733 * (optional) "largeblockok" -> (value ignored)
5734 * indicates that blocks > 128KB are permitted
5735 * (optional) "embedok" -> (value ignored)
5736 * presence indicates DRR_WRITE_EMBEDDED records are permitted
5737 * (optional) "compressok" -> (value ignored)
5738 * presence indicates compressed DRR_WRITE records are permitted
5739 * (optional) "resume_object" and "resume_offset" -> (uint64)
5740 * if present, resume send stream from specified object and offset.
5747 zfs_ioc_send_new(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl)
5751 char *fromname = NULL;
5754 boolean_t largeblockok;
5756 boolean_t compressok;
5757 uint64_t resumeobj = 0;
5758 uint64_t resumeoff = 0;
5760 error = nvlist_lookup_int32(innvl, "fd", &fd);
5762 return (SET_ERROR(EINVAL));
5764 (void) nvlist_lookup_string(innvl, "fromsnap", &fromname);
5766 largeblockok = nvlist_exists(innvl, "largeblockok");
5767 embedok = nvlist_exists(innvl, "embedok");
5768 compressok = nvlist_exists(innvl, "compressok");
5770 (void) nvlist_lookup_uint64(innvl, "resume_object", &resumeobj);
5771 (void) nvlist_lookup_uint64(innvl, "resume_offset", &resumeoff);
5773 if ((fp = getf(fd)) == NULL)
5774 return (SET_ERROR(EBADF));
5777 error = dmu_send(snapname, fromname, embedok, largeblockok, compressok,
5778 fd, resumeobj, resumeoff, fp->f_vnode, &off);
5780 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
5788 * Determine approximately how large a zfs send stream will be -- the number
5789 * of bytes that will be written to the fd supplied to zfs_ioc_send_new().
5792 * (optional) "from" -> full snap or bookmark name to send an incremental
5794 * (optional) "largeblockok" -> (value ignored)
5795 * indicates that blocks > 128KB are permitted
5796 * (optional) "embedok" -> (value ignored)
5797 * presence indicates DRR_WRITE_EMBEDDED records are permitted
5798 * (optional) "compressok" -> (value ignored)
5799 * presence indicates compressed DRR_WRITE records are permitted
5803 * "space" -> bytes of space (uint64)
5807 zfs_ioc_send_space(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl)
5810 dsl_dataset_t *tosnap;
5813 /* LINTED E_FUNC_SET_NOT_USED */
5814 boolean_t largeblockok;
5815 /* LINTED E_FUNC_SET_NOT_USED */
5817 boolean_t compressok;
5820 error = dsl_pool_hold(snapname, FTAG, &dp);
5824 error = dsl_dataset_hold(dp, snapname, FTAG, &tosnap);
5826 dsl_pool_rele(dp, FTAG);
5830 largeblockok = nvlist_exists(innvl, "largeblockok");
5831 embedok = nvlist_exists(innvl, "embedok");
5832 compressok = nvlist_exists(innvl, "compressok");
5834 error = nvlist_lookup_string(innvl, "from", &fromname);
5836 if (strchr(fromname, '@') != NULL) {
5838 * If from is a snapshot, hold it and use the more
5839 * efficient dmu_send_estimate to estimate send space
5840 * size using deadlists.
5842 dsl_dataset_t *fromsnap;
5843 error = dsl_dataset_hold(dp, fromname, FTAG, &fromsnap);
5846 error = dmu_send_estimate(tosnap, fromsnap, compressok,
5848 dsl_dataset_rele(fromsnap, FTAG);
5849 } else if (strchr(fromname, '#') != NULL) {
5851 * If from is a bookmark, fetch the creation TXG of the
5852 * snapshot it was created from and use that to find
5853 * blocks that were born after it.
5855 zfs_bookmark_phys_t frombm;
5857 error = dsl_bookmark_lookup(dp, fromname, tosnap,
5861 error = dmu_send_estimate_from_txg(tosnap,
5862 frombm.zbm_creation_txg, compressok, &space);
5865 * from is not properly formatted as a snapshot or
5868 error = SET_ERROR(EINVAL);
5872 // If estimating the size of a full send, use dmu_send_estimate
5873 error = dmu_send_estimate(tosnap, NULL, compressok, &space);
5876 fnvlist_add_uint64(outnvl, "space", space);
5879 dsl_dataset_rele(tosnap, FTAG);
5880 dsl_pool_rele(dp, FTAG);
5885 * Sync the currently open TXG to disk for the specified pool.
5886 * This is somewhat similar to 'zfs_sync()'.
5887 * For cases that do not result in error this ioctl will wait for
5888 * the currently open TXG to commit before returning back to the caller.
5891 * "force" -> when true, force uberblock update even if there is no dirty data.
5892 * In addition this will cause the vdev configuration to be written
5893 * out including updating the zpool cache file. (boolean_t)
5900 zfs_ioc_pool_sync(const char *pool, nvlist_t *innvl, nvlist_t *onvl)
5906 if ((err = spa_open(pool, &spa, FTAG)) != 0)
5909 force = fnvlist_lookup_boolean_value(innvl, "force");
5911 spa_config_enter(spa, SCL_CONFIG, FTAG, RW_WRITER);
5912 vdev_config_dirty(spa->spa_root_vdev);
5913 spa_config_exit(spa, SCL_CONFIG, FTAG);
5915 txg_wait_synced(spa_get_dsl(spa), 0);
5917 spa_close(spa, FTAG);
5922 static zfs_ioc_vec_t zfs_ioc_vec[ZFS_IOC_LAST - ZFS_IOC_FIRST];
5925 zfs_ioctl_register_legacy(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5926 zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck,
5927 boolean_t log_history, zfs_ioc_poolcheck_t pool_check)
5929 zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST];
5931 ASSERT3U(ioc, >=, ZFS_IOC_FIRST);
5932 ASSERT3U(ioc, <, ZFS_IOC_LAST);
5933 ASSERT3P(vec->zvec_legacy_func, ==, NULL);
5934 ASSERT3P(vec->zvec_func, ==, NULL);
5936 vec->zvec_legacy_func = func;
5937 vec->zvec_secpolicy = secpolicy;
5938 vec->zvec_namecheck = namecheck;
5939 vec->zvec_allow_log = log_history;
5940 vec->zvec_pool_check = pool_check;
5944 * See the block comment at the beginning of this file for details on
5945 * each argument to this function.
5948 zfs_ioctl_register(const char *name, zfs_ioc_t ioc, zfs_ioc_func_t *func,
5949 zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck,
5950 zfs_ioc_poolcheck_t pool_check, boolean_t smush_outnvlist,
5951 boolean_t allow_log)
5953 zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST];
5955 ASSERT3U(ioc, >=, ZFS_IOC_FIRST);
5956 ASSERT3U(ioc, <, ZFS_IOC_LAST);
5957 ASSERT3P(vec->zvec_legacy_func, ==, NULL);
5958 ASSERT3P(vec->zvec_func, ==, NULL);
5960 /* if we are logging, the name must be valid */
5961 ASSERT(!allow_log || namecheck != NO_NAME);
5963 vec->zvec_name = name;
5964 vec->zvec_func = func;
5965 vec->zvec_secpolicy = secpolicy;
5966 vec->zvec_namecheck = namecheck;
5967 vec->zvec_pool_check = pool_check;
5968 vec->zvec_smush_outnvlist = smush_outnvlist;
5969 vec->zvec_allow_log = allow_log;
5973 zfs_ioctl_register_pool(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5974 zfs_secpolicy_func_t *secpolicy, boolean_t log_history,
5975 zfs_ioc_poolcheck_t pool_check)
5977 zfs_ioctl_register_legacy(ioc, func, secpolicy,
5978 POOL_NAME, log_history, pool_check);
5982 zfs_ioctl_register_dataset_nolog(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5983 zfs_secpolicy_func_t *secpolicy, zfs_ioc_poolcheck_t pool_check)
5985 zfs_ioctl_register_legacy(ioc, func, secpolicy,
5986 DATASET_NAME, B_FALSE, pool_check);
5990 zfs_ioctl_register_pool_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func)
5992 zfs_ioctl_register_legacy(ioc, func, zfs_secpolicy_config,
5993 POOL_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5997 zfs_ioctl_register_pool_meta(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5998 zfs_secpolicy_func_t *secpolicy)
6000 zfs_ioctl_register_legacy(ioc, func, secpolicy,
6001 NO_NAME, B_FALSE, POOL_CHECK_NONE);
6005 zfs_ioctl_register_dataset_read_secpolicy(zfs_ioc_t ioc,
6006 zfs_ioc_legacy_func_t *func, zfs_secpolicy_func_t *secpolicy)
6008 zfs_ioctl_register_legacy(ioc, func, secpolicy,
6009 DATASET_NAME, B_FALSE, POOL_CHECK_SUSPENDED);
6013 zfs_ioctl_register_dataset_read(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func)
6015 zfs_ioctl_register_dataset_read_secpolicy(ioc, func,
6016 zfs_secpolicy_read);
6020 zfs_ioctl_register_dataset_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
6021 zfs_secpolicy_func_t *secpolicy)
6023 zfs_ioctl_register_legacy(ioc, func, secpolicy,
6024 DATASET_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
6028 zfs_ioctl_init(void)
6030 zfs_ioctl_register("snapshot", ZFS_IOC_SNAPSHOT,
6031 zfs_ioc_snapshot, zfs_secpolicy_snapshot, POOL_NAME,
6032 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
6034 zfs_ioctl_register("log_history", ZFS_IOC_LOG_HISTORY,
6035 zfs_ioc_log_history, zfs_secpolicy_log_history, NO_NAME,
6036 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_FALSE);
6038 zfs_ioctl_register("space_snaps", ZFS_IOC_SPACE_SNAPS,
6039 zfs_ioc_space_snaps, zfs_secpolicy_read, DATASET_NAME,
6040 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
6042 zfs_ioctl_register("send", ZFS_IOC_SEND_NEW,
6043 zfs_ioc_send_new, zfs_secpolicy_send_new, DATASET_NAME,
6044 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
6046 zfs_ioctl_register("send_space", ZFS_IOC_SEND_SPACE,
6047 zfs_ioc_send_space, zfs_secpolicy_read, DATASET_NAME,
6048 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
6050 zfs_ioctl_register("create", ZFS_IOC_CREATE,
6051 zfs_ioc_create, zfs_secpolicy_create_clone, DATASET_NAME,
6052 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
6054 zfs_ioctl_register("clone", ZFS_IOC_CLONE,
6055 zfs_ioc_clone, zfs_secpolicy_create_clone, DATASET_NAME,
6056 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
6058 zfs_ioctl_register("destroy_snaps", ZFS_IOC_DESTROY_SNAPS,
6059 zfs_ioc_destroy_snaps, zfs_secpolicy_destroy_snaps, POOL_NAME,
6060 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
6062 zfs_ioctl_register("hold", ZFS_IOC_HOLD,
6063 zfs_ioc_hold, zfs_secpolicy_hold, POOL_NAME,
6064 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
6065 zfs_ioctl_register("release", ZFS_IOC_RELEASE,
6066 zfs_ioc_release, zfs_secpolicy_release, POOL_NAME,
6067 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
6069 zfs_ioctl_register("get_holds", ZFS_IOC_GET_HOLDS,
6070 zfs_ioc_get_holds, zfs_secpolicy_read, DATASET_NAME,
6071 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
6073 zfs_ioctl_register("rollback", ZFS_IOC_ROLLBACK,
6074 zfs_ioc_rollback, zfs_secpolicy_rollback, DATASET_NAME,
6075 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_TRUE);
6077 zfs_ioctl_register("bookmark", ZFS_IOC_BOOKMARK,
6078 zfs_ioc_bookmark, zfs_secpolicy_bookmark, POOL_NAME,
6079 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
6081 zfs_ioctl_register("get_bookmarks", ZFS_IOC_GET_BOOKMARKS,
6082 zfs_ioc_get_bookmarks, zfs_secpolicy_read, DATASET_NAME,
6083 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
6085 zfs_ioctl_register("destroy_bookmarks", ZFS_IOC_DESTROY_BOOKMARKS,
6086 zfs_ioc_destroy_bookmarks, zfs_secpolicy_destroy_bookmarks,
6088 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
6090 zfs_ioctl_register("receive", ZFS_IOC_RECV_NEW,
6091 zfs_ioc_recv_new, zfs_secpolicy_recv_new, DATASET_NAME,
6092 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
6094 zfs_ioctl_register("sync", ZFS_IOC_POOL_SYNC,
6095 zfs_ioc_pool_sync, zfs_secpolicy_none, POOL_NAME,
6096 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_FALSE);
6098 /* IOCTLS that use the legacy function signature */
6100 zfs_ioctl_register_legacy(ZFS_IOC_POOL_FREEZE, zfs_ioc_pool_freeze,
6101 zfs_secpolicy_config, NO_NAME, B_FALSE, POOL_CHECK_READONLY);
6103 zfs_ioctl_register_pool(ZFS_IOC_POOL_CREATE, zfs_ioc_pool_create,
6104 zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE);
6105 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SCAN,
6107 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_UPGRADE,
6108 zfs_ioc_pool_upgrade);
6109 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ADD,
6111 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_REMOVE,
6112 zfs_ioc_vdev_remove);
6113 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SET_STATE,
6114 zfs_ioc_vdev_set_state);
6115 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ATTACH,
6116 zfs_ioc_vdev_attach);
6117 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_DETACH,
6118 zfs_ioc_vdev_detach);
6119 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETPATH,
6120 zfs_ioc_vdev_setpath);
6121 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETFRU,
6122 zfs_ioc_vdev_setfru);
6123 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SET_PROPS,
6124 zfs_ioc_pool_set_props);
6125 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SPLIT,
6126 zfs_ioc_vdev_split);
6127 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_REGUID,
6128 zfs_ioc_pool_reguid);
6130 zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_CONFIGS,
6131 zfs_ioc_pool_configs, zfs_secpolicy_none);
6132 zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_TRYIMPORT,
6133 zfs_ioc_pool_tryimport, zfs_secpolicy_config);
6134 zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_FAULT,
6135 zfs_ioc_inject_fault, zfs_secpolicy_inject);
6136 zfs_ioctl_register_pool_meta(ZFS_IOC_CLEAR_FAULT,
6137 zfs_ioc_clear_fault, zfs_secpolicy_inject);
6138 zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_LIST_NEXT,
6139 zfs_ioc_inject_list_next, zfs_secpolicy_inject);
6142 * pool destroy, and export don't log the history as part of
6143 * zfsdev_ioctl, but rather zfs_ioc_pool_export
6144 * does the logging of those commands.
6146 zfs_ioctl_register_pool(ZFS_IOC_POOL_DESTROY, zfs_ioc_pool_destroy,
6147 zfs_secpolicy_config, B_FALSE, POOL_CHECK_SUSPENDED);
6148 zfs_ioctl_register_pool(ZFS_IOC_POOL_EXPORT, zfs_ioc_pool_export,
6149 zfs_secpolicy_config, B_FALSE, POOL_CHECK_SUSPENDED);
6151 zfs_ioctl_register_pool(ZFS_IOC_POOL_STATS, zfs_ioc_pool_stats,
6152 zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE);
6153 zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_PROPS, zfs_ioc_pool_get_props,
6154 zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE);
6156 zfs_ioctl_register_pool(ZFS_IOC_ERROR_LOG, zfs_ioc_error_log,
6157 zfs_secpolicy_inject, B_FALSE, POOL_CHECK_SUSPENDED);
6158 zfs_ioctl_register_pool(ZFS_IOC_DSOBJ_TO_DSNAME,
6159 zfs_ioc_dsobj_to_dsname,
6160 zfs_secpolicy_diff, B_FALSE, POOL_CHECK_SUSPENDED);
6161 zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_HISTORY,
6162 zfs_ioc_pool_get_history,
6163 zfs_secpolicy_config, B_FALSE, POOL_CHECK_SUSPENDED);
6165 zfs_ioctl_register_pool(ZFS_IOC_POOL_IMPORT, zfs_ioc_pool_import,
6166 zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE);
6168 zfs_ioctl_register_pool(ZFS_IOC_CLEAR, zfs_ioc_clear,
6169 zfs_secpolicy_config, B_TRUE, POOL_CHECK_READONLY);
6170 zfs_ioctl_register_pool(ZFS_IOC_POOL_REOPEN, zfs_ioc_pool_reopen,
6171 zfs_secpolicy_config, B_TRUE, POOL_CHECK_SUSPENDED);
6173 zfs_ioctl_register_dataset_read(ZFS_IOC_SPACE_WRITTEN,
6174 zfs_ioc_space_written);
6175 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_RECVD_PROPS,
6176 zfs_ioc_objset_recvd_props);
6177 zfs_ioctl_register_dataset_read(ZFS_IOC_NEXT_OBJ,
6179 zfs_ioctl_register_dataset_read(ZFS_IOC_GET_FSACL,
6181 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_STATS,
6182 zfs_ioc_objset_stats);
6183 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_ZPLPROPS,
6184 zfs_ioc_objset_zplprops);
6185 zfs_ioctl_register_dataset_read(ZFS_IOC_DATASET_LIST_NEXT,
6186 zfs_ioc_dataset_list_next);
6187 zfs_ioctl_register_dataset_read(ZFS_IOC_SNAPSHOT_LIST_NEXT,
6188 zfs_ioc_snapshot_list_next);
6189 zfs_ioctl_register_dataset_read(ZFS_IOC_SEND_PROGRESS,
6190 zfs_ioc_send_progress);
6192 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_DIFF,
6193 zfs_ioc_diff, zfs_secpolicy_diff);
6194 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_STATS,
6195 zfs_ioc_obj_to_stats, zfs_secpolicy_diff);
6196 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_PATH,
6197 zfs_ioc_obj_to_path, zfs_secpolicy_diff);
6198 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_ONE,
6199 zfs_ioc_userspace_one, zfs_secpolicy_userspace_one);
6200 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_MANY,
6201 zfs_ioc_userspace_many, zfs_secpolicy_userspace_many);
6202 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_SEND,
6203 zfs_ioc_send, zfs_secpolicy_send);
6205 zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_PROP, zfs_ioc_set_prop,
6206 zfs_secpolicy_none);
6207 zfs_ioctl_register_dataset_modify(ZFS_IOC_DESTROY, zfs_ioc_destroy,
6208 zfs_secpolicy_destroy);
6209 zfs_ioctl_register_dataset_modify(ZFS_IOC_RENAME, zfs_ioc_rename,
6210 zfs_secpolicy_rename);
6211 zfs_ioctl_register_dataset_modify(ZFS_IOC_RECV, zfs_ioc_recv,
6212 zfs_secpolicy_recv);
6213 zfs_ioctl_register_dataset_modify(ZFS_IOC_PROMOTE, zfs_ioc_promote,
6214 zfs_secpolicy_promote);
6215 zfs_ioctl_register_dataset_modify(ZFS_IOC_INHERIT_PROP,
6216 zfs_ioc_inherit_prop, zfs_secpolicy_inherit_prop);
6217 zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_FSACL, zfs_ioc_set_fsacl,
6218 zfs_secpolicy_set_fsacl);
6220 zfs_ioctl_register_dataset_nolog(ZFS_IOC_SHARE, zfs_ioc_share,
6221 zfs_secpolicy_share, POOL_CHECK_NONE);
6222 zfs_ioctl_register_dataset_nolog(ZFS_IOC_SMB_ACL, zfs_ioc_smb_acl,
6223 zfs_secpolicy_smb_acl, POOL_CHECK_NONE);
6224 zfs_ioctl_register_dataset_nolog(ZFS_IOC_USERSPACE_UPGRADE,
6225 zfs_ioc_userspace_upgrade, zfs_secpolicy_userspace_upgrade,
6226 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
6227 zfs_ioctl_register_dataset_nolog(ZFS_IOC_TMP_SNAPSHOT,
6228 zfs_ioc_tmp_snapshot, zfs_secpolicy_tmp_snapshot,
6229 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
6234 zfs_ioctl_register_legacy(ZFS_IOC_EVENTS_NEXT, zfs_ioc_events_next,
6235 zfs_secpolicy_config, NO_NAME, B_FALSE, POOL_CHECK_NONE);
6236 zfs_ioctl_register_legacy(ZFS_IOC_EVENTS_CLEAR, zfs_ioc_events_clear,
6237 zfs_secpolicy_config, NO_NAME, B_FALSE, POOL_CHECK_NONE);
6238 zfs_ioctl_register_legacy(ZFS_IOC_EVENTS_SEEK, zfs_ioc_events_seek,
6239 zfs_secpolicy_config, NO_NAME, B_FALSE, POOL_CHECK_NONE);
6243 pool_status_check(const char *name, zfs_ioc_namecheck_t type,
6244 zfs_ioc_poolcheck_t check)
6249 ASSERT(type == POOL_NAME || type == DATASET_NAME);
6251 if (check & POOL_CHECK_NONE)
6254 error = spa_open(name, &spa, FTAG);
6256 if ((check & POOL_CHECK_SUSPENDED) && spa_suspended(spa))
6257 error = SET_ERROR(EAGAIN);
6258 else if ((check & POOL_CHECK_READONLY) && !spa_writeable(spa))
6259 error = SET_ERROR(EROFS);
6260 spa_close(spa, FTAG);
6266 zfsdev_get_state_impl(minor_t minor, enum zfsdev_state_type which)
6270 for (zs = zfsdev_state_list; zs != NULL; zs = zs->zs_next) {
6271 if (zs->zs_minor == minor) {
6275 return (zs->zs_onexit);
6277 return (zs->zs_zevent);
6288 zfsdev_get_state(minor_t minor, enum zfsdev_state_type which)
6292 ptr = zfsdev_get_state_impl(minor, which);
6298 zfsdev_getminor(struct file *filp, minor_t *minorp)
6300 zfsdev_state_t *zs, *fpd;
6302 ASSERT(filp != NULL);
6303 ASSERT(!MUTEX_HELD(&zfsdev_state_lock));
6305 fpd = filp->private_data;
6309 mutex_enter(&zfsdev_state_lock);
6311 for (zs = zfsdev_state_list; zs != NULL; zs = zs->zs_next) {
6313 if (zs->zs_minor == -1)
6317 *minorp = fpd->zs_minor;
6318 mutex_exit(&zfsdev_state_lock);
6323 mutex_exit(&zfsdev_state_lock);
6329 * Find a free minor number. The zfsdev_state_list is expected to
6330 * be short since it is only a list of currently open file handles.
6333 zfsdev_minor_alloc(void)
6335 static minor_t last_minor = 0;
6338 ASSERT(MUTEX_HELD(&zfsdev_state_lock));
6340 for (m = last_minor + 1; m != last_minor; m++) {
6341 if (m > ZFSDEV_MAX_MINOR)
6343 if (zfsdev_get_state_impl(m, ZST_ALL) == NULL) {
6353 zfsdev_state_init(struct file *filp)
6355 zfsdev_state_t *zs, *zsprev = NULL;
6357 boolean_t newzs = B_FALSE;
6359 ASSERT(MUTEX_HELD(&zfsdev_state_lock));
6361 minor = zfsdev_minor_alloc();
6363 return (SET_ERROR(ENXIO));
6365 for (zs = zfsdev_state_list; zs != NULL; zs = zs->zs_next) {
6366 if (zs->zs_minor == -1)
6372 zs = kmem_zalloc(sizeof (zfsdev_state_t), KM_SLEEP);
6377 filp->private_data = zs;
6379 zfs_onexit_init((zfs_onexit_t **)&zs->zs_onexit);
6380 zfs_zevent_init((zfs_zevent_t **)&zs->zs_zevent);
6384 * In order to provide for lock-free concurrent read access
6385 * to the minor list in zfsdev_get_state_impl(), new entries
6386 * must be completely written before linking them into the
6387 * list whereas existing entries are already linked; the last
6388 * operation must be updating zs_minor (from -1 to the new
6392 zs->zs_minor = minor;
6394 zsprev->zs_next = zs;
6397 zs->zs_minor = minor;
6404 zfsdev_state_destroy(struct file *filp)
6408 ASSERT(MUTEX_HELD(&zfsdev_state_lock));
6409 ASSERT(filp->private_data != NULL);
6411 zs = filp->private_data;
6413 zfs_onexit_destroy(zs->zs_onexit);
6414 zfs_zevent_destroy(zs->zs_zevent);
6420 zfsdev_open(struct inode *ino, struct file *filp)
6424 mutex_enter(&zfsdev_state_lock);
6425 error = zfsdev_state_init(filp);
6426 mutex_exit(&zfsdev_state_lock);
6432 zfsdev_release(struct inode *ino, struct file *filp)
6436 mutex_enter(&zfsdev_state_lock);
6437 error = zfsdev_state_destroy(filp);
6438 mutex_exit(&zfsdev_state_lock);
6444 zfsdev_ioctl(struct file *filp, unsigned cmd, unsigned long arg)
6448 int error, rc, flag = 0;
6449 const zfs_ioc_vec_t *vec;
6450 char *saved_poolname = NULL;
6451 nvlist_t *innvl = NULL;
6452 fstrans_cookie_t cookie;
6454 vecnum = cmd - ZFS_IOC_FIRST;
6455 if (vecnum >= sizeof (zfs_ioc_vec) / sizeof (zfs_ioc_vec[0]))
6456 return (-SET_ERROR(EINVAL));
6457 vec = &zfs_ioc_vec[vecnum];
6460 * The registered ioctl list may be sparse, verify that either
6461 * a normal or legacy handler are registered.
6463 if (vec->zvec_func == NULL && vec->zvec_legacy_func == NULL)
6464 return (-SET_ERROR(EINVAL));
6466 zc = kmem_zalloc(sizeof (zfs_cmd_t), KM_SLEEP);
6468 error = ddi_copyin((void *)arg, zc, sizeof (zfs_cmd_t), flag);
6470 error = SET_ERROR(EFAULT);
6474 zc->zc_iflags = flag & FKIOCTL;
6475 if (zc->zc_nvlist_src_size > MAX_NVLIST_SRC_SIZE) {
6477 * Make sure the user doesn't pass in an insane value for
6478 * zc_nvlist_src_size. We have to check, since we will end
6479 * up allocating that much memory inside of get_nvlist(). This
6480 * prevents a nefarious user from allocating tons of kernel
6483 * Also, we return EINVAL instead of ENOMEM here. The reason
6484 * being that returning ENOMEM from an ioctl() has a special
6485 * connotation; that the user's size value is too small and
6486 * needs to be expanded to hold the nvlist. See
6487 * zcmd_expand_dst_nvlist() for details.
6489 error = SET_ERROR(EINVAL); /* User's size too big */
6491 } else if (zc->zc_nvlist_src_size != 0) {
6492 error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
6493 zc->zc_iflags, &innvl);
6499 * Ensure that all pool/dataset names are valid before we pass down to
6502 zc->zc_name[sizeof (zc->zc_name) - 1] = '\0';
6503 switch (vec->zvec_namecheck) {
6505 if (pool_namecheck(zc->zc_name, NULL, NULL) != 0)
6506 error = SET_ERROR(EINVAL);
6508 error = pool_status_check(zc->zc_name,
6509 vec->zvec_namecheck, vec->zvec_pool_check);
6513 if (dataset_namecheck(zc->zc_name, NULL, NULL) != 0)
6514 error = SET_ERROR(EINVAL);
6516 error = pool_status_check(zc->zc_name,
6517 vec->zvec_namecheck, vec->zvec_pool_check);
6526 cookie = spl_fstrans_mark();
6527 error = vec->zvec_secpolicy(zc, innvl, CRED());
6528 spl_fstrans_unmark(cookie);
6534 /* legacy ioctls can modify zc_name */
6535 saved_poolname = strdup(zc->zc_name);
6536 if (saved_poolname == NULL) {
6537 error = SET_ERROR(ENOMEM);
6540 saved_poolname[strcspn(saved_poolname, "/@#")] = '\0';
6543 if (vec->zvec_func != NULL) {
6547 nvlist_t *lognv = NULL;
6549 ASSERT(vec->zvec_legacy_func == NULL);
6552 * Add the innvl to the lognv before calling the func,
6553 * in case the func changes the innvl.
6555 if (vec->zvec_allow_log) {
6556 lognv = fnvlist_alloc();
6557 fnvlist_add_string(lognv, ZPOOL_HIST_IOCTL,
6559 if (!nvlist_empty(innvl)) {
6560 fnvlist_add_nvlist(lognv, ZPOOL_HIST_INPUT_NVL,
6565 outnvl = fnvlist_alloc();
6566 cookie = spl_fstrans_mark();
6567 error = vec->zvec_func(zc->zc_name, innvl, outnvl);
6568 spl_fstrans_unmark(cookie);
6570 if (error == 0 && vec->zvec_allow_log &&
6571 spa_open(zc->zc_name, &spa, FTAG) == 0) {
6572 if (!nvlist_empty(outnvl)) {
6573 fnvlist_add_nvlist(lognv, ZPOOL_HIST_OUTPUT_NVL,
6576 (void) spa_history_log_nvl(spa, lognv);
6577 spa_close(spa, FTAG);
6579 fnvlist_free(lognv);
6581 if (!nvlist_empty(outnvl) || zc->zc_nvlist_dst_size != 0) {
6583 if (vec->zvec_smush_outnvlist) {
6584 smusherror = nvlist_smush(outnvl,
6585 zc->zc_nvlist_dst_size);
6587 if (smusherror == 0)
6588 puterror = put_nvlist(zc, outnvl);
6594 nvlist_free(outnvl);
6596 cookie = spl_fstrans_mark();
6597 error = vec->zvec_legacy_func(zc);
6598 spl_fstrans_unmark(cookie);
6603 rc = ddi_copyout(zc, (void *)arg, sizeof (zfs_cmd_t), flag);
6604 if (error == 0 && rc != 0)
6605 error = SET_ERROR(EFAULT);
6606 if (error == 0 && vec->zvec_allow_log) {
6607 char *s = tsd_get(zfs_allow_log_key);
6610 (void) tsd_set(zfs_allow_log_key, saved_poolname);
6612 if (saved_poolname != NULL)
6613 strfree(saved_poolname);
6616 kmem_free(zc, sizeof (zfs_cmd_t));
6620 #ifdef CONFIG_COMPAT
6622 zfsdev_compat_ioctl(struct file *filp, unsigned cmd, unsigned long arg)
6624 return (zfsdev_ioctl(filp, cmd, arg));
6627 #define zfsdev_compat_ioctl NULL
6630 static const struct file_operations zfsdev_fops = {
6631 .open = zfsdev_open,
6632 .release = zfsdev_release,
6633 .unlocked_ioctl = zfsdev_ioctl,
6634 .compat_ioctl = zfsdev_compat_ioctl,
6635 .owner = THIS_MODULE,
6638 static struct miscdevice zfs_misc = {
6639 .minor = MISC_DYNAMIC_MINOR,
6641 .fops = &zfsdev_fops,
6649 mutex_init(&zfsdev_state_lock, NULL, MUTEX_DEFAULT, NULL);
6650 zfsdev_state_list = kmem_zalloc(sizeof (zfsdev_state_t), KM_SLEEP);
6651 zfsdev_state_list->zs_minor = -1;
6653 error = misc_register(&zfs_misc);
6655 printk(KERN_INFO "ZFS: misc_register() failed %d\n", error);
6665 zfsdev_state_t *zs, *zsprev = NULL;
6667 misc_deregister(&zfs_misc);
6668 mutex_destroy(&zfsdev_state_lock);
6670 for (zs = zfsdev_state_list; zs != NULL; zs = zs->zs_next) {
6672 kmem_free(zsprev, sizeof (zfsdev_state_t));
6676 kmem_free(zsprev, sizeof (zfsdev_state_t));
6680 zfs_allow_log_destroy(void *arg)
6682 char *poolname = arg;
6684 if (poolname != NULL)
6689 #define ZFS_DEBUG_STR " (DEBUG mode)"
6691 #define ZFS_DEBUG_STR ""
6699 error = -vn_set_pwd("/");
6702 "ZFS: Warning unable to set pwd to '/': %d\n", error);
6706 if ((error = -zvol_init()) != 0)
6709 spa_init(FREAD | FWRITE);
6714 if ((error = zfs_attach()) != 0)
6717 tsd_create(&zfs_fsyncer_key, NULL);
6718 tsd_create(&rrw_tsd_key, rrw_tsd_destroy);
6719 tsd_create(&zfs_allow_log_key, zfs_allow_log_destroy);
6721 printk(KERN_NOTICE "ZFS: Loaded module v%s-%s%s, "
6722 "ZFS pool version %s, ZFS filesystem version %s\n",
6723 ZFS_META_VERSION, ZFS_META_RELEASE, ZFS_DEBUG_STR,
6724 SPA_VERSION_STRING, ZPL_VERSION_STRING);
6725 #ifndef CONFIG_FS_POSIX_ACL
6726 printk(KERN_NOTICE "ZFS: Posix ACLs disabled by kernel\n");
6727 #endif /* CONFIG_FS_POSIX_ACL */
6735 printk(KERN_NOTICE "ZFS: Failed to Load ZFS Filesystem v%s-%s%s"
6736 ", rc = %d\n", ZFS_META_VERSION, ZFS_META_RELEASE,
6737 ZFS_DEBUG_STR, error);
6750 tsd_destroy(&zfs_fsyncer_key);
6751 tsd_destroy(&rrw_tsd_key);
6752 tsd_destroy(&zfs_allow_log_key);
6754 printk(KERN_NOTICE "ZFS: Unloaded module v%s-%s%s\n",
6755 ZFS_META_VERSION, ZFS_META_RELEASE, ZFS_DEBUG_STR);
6762 MODULE_DESCRIPTION("ZFS");
6763 MODULE_AUTHOR(ZFS_META_AUTHOR);
6764 MODULE_LICENSE(ZFS_META_LICENSE);
6765 MODULE_VERSION(ZFS_META_VERSION "-" ZFS_META_RELEASE);
6766 #endif /* HAVE_SPL */