4 * The contents of this file are subject to the terms of the
5 * Common Development and Distribution License (the "License").
6 * You may not use this file except in compliance with the License.
8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9 * or http://www.opensolaris.org/os/licensing.
10 * See the License for the specific language governing permissions
11 * and limitations under the License.
13 * When distributing Covered Code, include this CDDL HEADER in each
14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15 * If applicable, add the following below this CDDL HEADER, with the
16 * fields enclosed by brackets "[]" replaced with your own identifying
17 * information: Portions Copyright [yyyy] [name of copyright owner]
23 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
24 * Copyright (c) 2011-2012 Pawel Jakub Dawidek. All rights reserved.
25 * Copyright 2013 Martin Matuska <mm@FreeBSD.org>. All rights reserved.
26 * Copyright 2014 Xin Li <delphij@FreeBSD.org>. All rights reserved.
27 * Copyright 2015, OmniTI Computer Consulting, Inc. All rights reserved.
28 * Copyright 2015 Nexenta Systems, Inc. All rights reserved.
29 * Copyright (c) 2014, 2016 Joyent, Inc. All rights reserved.
30 * Copyright (c) 2011, 2017 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 2017 RackTop Systems.
36 * Copyright (c) 2017 Datto Inc.
42 * This file handles the ioctls to /dev/zfs, used for configuring ZFS storage
43 * pools and filesystems, e.g. with /sbin/zfs and /sbin/zpool.
45 * There are two ways that we handle ioctls: the legacy way where almost
46 * all of the logic is in the ioctl callback, and the new way where most
47 * of the marshalling is handled in the common entry point, zfsdev_ioctl().
49 * Non-legacy ioctls should be registered by calling
50 * zfs_ioctl_register() from zfs_ioctl_init(). The ioctl is invoked
51 * from userland by lzc_ioctl().
53 * The registration arguments are as follows:
56 * The name of the ioctl. This is used for history logging. If the
57 * ioctl returns successfully (the callback returns 0), and allow_log
58 * is true, then a history log entry will be recorded with the input &
59 * output nvlists. The log entry can be printed with "zpool history -i".
62 * The ioctl request number, which userland will pass to ioctl(2).
63 * The ioctl numbers can change from release to release, because
64 * the caller (libzfs) must be matched to the kernel.
66 * zfs_secpolicy_func_t *secpolicy
67 * This function will be called before the zfs_ioc_func_t, to
68 * determine if this operation is permitted. It should return EPERM
69 * on failure, and 0 on success. Checks include determining if the
70 * dataset is visible in this zone, and if the user has either all
71 * zfs privileges in the zone (SYS_MOUNT), or has been granted permission
72 * to do this operation on this dataset with "zfs allow".
74 * zfs_ioc_namecheck_t namecheck
75 * This specifies what to expect in the zfs_cmd_t:zc_name -- a pool
76 * name, a dataset name, or nothing. If the name is not well-formed,
77 * the ioctl will fail and the callback will not be called.
78 * Therefore, the callback can assume that the name is well-formed
79 * (e.g. is null-terminated, doesn't have more than one '@' character,
80 * doesn't have invalid characters).
82 * zfs_ioc_poolcheck_t pool_check
83 * This specifies requirements on the pool state. If the pool does
84 * not meet them (is suspended or is readonly), the ioctl will fail
85 * and the callback will not be called. If any checks are specified
86 * (i.e. it is not POOL_CHECK_NONE), namecheck must not be NO_NAME.
87 * Multiple checks can be or-ed together (e.g. POOL_CHECK_SUSPENDED |
88 * POOL_CHECK_READONLY).
90 * boolean_t smush_outnvlist
91 * If smush_outnvlist is true, then the output is presumed to be a
92 * list of errors, and it will be "smushed" down to fit into the
93 * caller's buffer, by removing some entries and replacing them with a
94 * single "N_MORE_ERRORS" entry indicating how many were removed. See
95 * nvlist_smush() for details. If smush_outnvlist is false, and the
96 * outnvlist does not fit into the userland-provided buffer, then the
97 * ioctl will fail with ENOMEM.
99 * zfs_ioc_func_t *func
100 * The callback function that will perform the operation.
102 * The callback should return 0 on success, or an error number on
103 * failure. If the function fails, the userland ioctl will return -1,
104 * and errno will be set to the callback's return value. The callback
105 * will be called with the following arguments:
108 * The name of the pool or dataset to operate on, from
109 * zfs_cmd_t:zc_name. The 'namecheck' argument specifies the
110 * expected type (pool, dataset, or none).
113 * The input nvlist, deserialized from zfs_cmd_t:zc_nvlist_src. Or
114 * NULL if no input nvlist was provided. Changes to this nvlist are
115 * ignored. If the input nvlist could not be deserialized, the
116 * ioctl will fail and the callback will not be called.
119 * The output nvlist, initially empty. The callback can fill it in,
120 * and it will be returned to userland by serializing it into
121 * zfs_cmd_t:zc_nvlist_dst. If it is non-empty, and serialization
122 * fails (e.g. because the caller didn't supply a large enough
123 * buffer), then the overall ioctl will fail. See the
124 * 'smush_nvlist' argument above for additional behaviors.
126 * There are two typical uses of the output nvlist:
127 * - To return state, e.g. property values. In this case,
128 * smush_outnvlist should be false. If the buffer was not large
129 * enough, the caller will reallocate a larger buffer and try
132 * - To return multiple errors from an ioctl which makes on-disk
133 * changes. In this case, smush_outnvlist should be true.
134 * Ioctls which make on-disk modifications should generally not
135 * use the outnvl if they succeed, because the caller can not
136 * distinguish between the operation failing, and
137 * deserialization failing.
140 #include "opt_kstack_pages.h"
143 #include <sys/types.h>
144 #include <sys/param.h>
145 #include <sys/systm.h>
146 #include <sys/conf.h>
147 #include <sys/kernel.h>
148 #include <sys/lock.h>
149 #include <sys/malloc.h>
150 #include <sys/mutex.h>
151 #include <sys/proc.h>
152 #include <sys/errno.h>
155 #include <sys/file.h>
156 #include <sys/kmem.h>
157 #include <sys/conf.h>
158 #include <sys/cmn_err.h>
159 #include <sys/stat.h>
160 #include <sys/zfs_ioctl.h>
161 #include <sys/zfs_vfsops.h>
162 #include <sys/zfs_znode.h>
165 #include <sys/spa_impl.h>
166 #include <sys/vdev.h>
168 #include <sys/dsl_dir.h>
169 #include <sys/dsl_dataset.h>
170 #include <sys/dsl_prop.h>
171 #include <sys/dsl_deleg.h>
172 #include <sys/dmu_objset.h>
173 #include <sys/dmu_impl.h>
174 #include <sys/dmu_tx.h>
175 #include <sys/sunddi.h>
176 #include <sys/policy.h>
177 #include <sys/zone.h>
178 #include <sys/nvpair.h>
179 #include <sys/mount.h>
180 #include <sys/taskqueue.h>
182 #include <sys/varargs.h>
183 #include <sys/fs/zfs.h>
184 #include <sys/zfs_ctldir.h>
185 #include <sys/zfs_dir.h>
186 #include <sys/zfs_onexit.h>
187 #include <sys/zvol.h>
188 #include <sys/dsl_scan.h>
189 #include <sys/dmu_objset.h>
190 #include <sys/dmu_send.h>
191 #include <sys/dsl_destroy.h>
192 #include <sys/dsl_bookmark.h>
193 #include <sys/dsl_userhold.h>
194 #include <sys/zfeature.h>
196 #include <sys/zio_checksum.h>
198 #include "zfs_namecheck.h"
199 #include "zfs_prop.h"
200 #include "zfs_deleg.h"
201 #include "zfs_comutil.h"
202 #include "zfs_ioctl_compat.h"
207 CTASSERT(sizeof(zfs_cmd_t) < IOCPARM_MAX);
208 static struct cdev *zfsdev;
210 extern void zfs_init(void);
211 extern void zfs_fini(void);
213 uint_t zfs_fsyncer_key;
214 extern uint_t rrw_tsd_key;
215 static uint_t zfs_allow_log_key;
216 extern uint_t zfs_geom_probe_vdev_key;
218 typedef int zfs_ioc_legacy_func_t(zfs_cmd_t *);
219 typedef int zfs_ioc_func_t(const char *, nvlist_t *, nvlist_t *);
220 typedef int zfs_secpolicy_func_t(zfs_cmd_t *, nvlist_t *, cred_t *);
226 } zfs_ioc_namecheck_t;
229 POOL_CHECK_NONE = 1 << 0,
230 POOL_CHECK_SUSPENDED = 1 << 1,
231 POOL_CHECK_READONLY = 1 << 2,
232 } zfs_ioc_poolcheck_t;
234 typedef struct zfs_ioc_vec {
235 zfs_ioc_legacy_func_t *zvec_legacy_func;
236 zfs_ioc_func_t *zvec_func;
237 zfs_secpolicy_func_t *zvec_secpolicy;
238 zfs_ioc_namecheck_t zvec_namecheck;
239 boolean_t zvec_allow_log;
240 zfs_ioc_poolcheck_t zvec_pool_check;
241 boolean_t zvec_smush_outnvlist;
242 const char *zvec_name;
245 /* This array is indexed by zfs_userquota_prop_t */
246 static const char *userquota_perms[] = {
247 ZFS_DELEG_PERM_USERUSED,
248 ZFS_DELEG_PERM_USERQUOTA,
249 ZFS_DELEG_PERM_GROUPUSED,
250 ZFS_DELEG_PERM_GROUPQUOTA,
253 static int zfs_ioc_userspace_upgrade(zfs_cmd_t *zc);
254 static int zfs_check_settable(const char *name, nvpair_t *property,
256 static int zfs_check_clearable(char *dataset, nvlist_t *props,
258 static int zfs_fill_zplprops_root(uint64_t, nvlist_t *, nvlist_t *,
260 int zfs_set_prop_nvlist(const char *, zprop_source_t, nvlist_t *, nvlist_t *);
261 static int get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp);
263 static void zfsdev_close(void *data);
265 static int zfs_prop_activate_feature(spa_t *spa, spa_feature_t feature);
267 /* _NOTE(PRINTFLIKE(4)) - this is printf-like, but lint is too whiney */
269 __dprintf(const char *file, const char *func, int line, const char *fmt, ...)
276 * Get rid of annoying "../common/" prefix to filename.
278 newfile = strrchr(file, '/');
279 if (newfile != NULL) {
280 newfile = newfile + 1; /* Get rid of leading / */
286 (void) vsnprintf(buf, sizeof (buf), fmt, adx);
290 * To get this data, use the zfs-dprintf probe as so:
291 * dtrace -q -n 'zfs-dprintf \
292 * /stringof(arg0) == "dbuf.c"/ \
293 * {printf("%s: %s", stringof(arg1), stringof(arg3))}'
295 * arg1 = function name
299 DTRACE_PROBE4(zfs__dprintf,
300 char *, newfile, char *, func, int, line, char *, buf);
304 history_str_free(char *buf)
306 kmem_free(buf, HIS_MAX_RECORD_LEN);
310 history_str_get(zfs_cmd_t *zc)
314 if (zc->zc_history == 0)
317 buf = kmem_alloc(HIS_MAX_RECORD_LEN, KM_SLEEP);
318 if (copyinstr((void *)(uintptr_t)zc->zc_history,
319 buf, HIS_MAX_RECORD_LEN, NULL) != 0) {
320 history_str_free(buf);
324 buf[HIS_MAX_RECORD_LEN -1] = '\0';
330 * Check to see if the named dataset is currently defined as bootable
333 zfs_is_bootfs(const char *name)
337 if (dmu_objset_hold(name, FTAG, &os) == 0) {
339 ret = (dmu_objset_id(os) == spa_bootfs(dmu_objset_spa(os)));
340 dmu_objset_rele(os, FTAG);
347 * Return non-zero if the spa version is less than requested version.
350 zfs_earlier_version(const char *name, int version)
354 if (spa_open(name, &spa, FTAG) == 0) {
355 if (spa_version(spa) < version) {
356 spa_close(spa, FTAG);
359 spa_close(spa, FTAG);
365 * Return TRUE if the ZPL version is less than requested version.
368 zpl_earlier_version(const char *name, int version)
371 boolean_t rc = B_TRUE;
373 if (dmu_objset_hold(name, FTAG, &os) == 0) {
376 if (dmu_objset_type(os) != DMU_OST_ZFS) {
377 dmu_objset_rele(os, FTAG);
380 /* XXX reading from non-owned objset */
381 if (zfs_get_zplprop(os, ZFS_PROP_VERSION, &zplversion) == 0)
382 rc = zplversion < version;
383 dmu_objset_rele(os, FTAG);
389 zfs_log_history(zfs_cmd_t *zc)
394 if ((buf = history_str_get(zc)) == NULL)
397 if (spa_open(zc->zc_name, &spa, FTAG) == 0) {
398 if (spa_version(spa) >= SPA_VERSION_ZPOOL_HISTORY)
399 (void) spa_history_log(spa, buf);
400 spa_close(spa, FTAG);
402 history_str_free(buf);
406 * Policy for top-level read operations (list pools). Requires no privileges,
407 * and can be used in the local zone, as there is no associated dataset.
411 zfs_secpolicy_none(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
417 * Policy for dataset read operations (list children, get statistics). Requires
418 * no privileges, but must be visible in the local zone.
422 zfs_secpolicy_read(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
424 if (INGLOBALZONE(curthread) ||
425 zone_dataset_visible(zc->zc_name, NULL))
428 return (SET_ERROR(ENOENT));
432 zfs_dozonecheck_impl(const char *dataset, uint64_t zoned, cred_t *cr)
437 * The dataset must be visible by this zone -- check this first
438 * so they don't see EPERM on something they shouldn't know about.
440 if (!INGLOBALZONE(curthread) &&
441 !zone_dataset_visible(dataset, &writable))
442 return (SET_ERROR(ENOENT));
444 if (INGLOBALZONE(curthread)) {
446 * If the fs is zoned, only root can access it from the
449 if (secpolicy_zfs(cr) && zoned)
450 return (SET_ERROR(EPERM));
453 * If we are in a local zone, the 'zoned' property must be set.
456 return (SET_ERROR(EPERM));
458 /* must be writable by this zone */
460 return (SET_ERROR(EPERM));
466 zfs_dozonecheck(const char *dataset, cred_t *cr)
470 if (dsl_prop_get_integer(dataset, "jailed", &zoned, NULL))
471 return (SET_ERROR(ENOENT));
473 return (zfs_dozonecheck_impl(dataset, zoned, cr));
477 zfs_dozonecheck_ds(const char *dataset, dsl_dataset_t *ds, cred_t *cr)
481 if (dsl_prop_get_int_ds(ds, "jailed", &zoned))
482 return (SET_ERROR(ENOENT));
484 return (zfs_dozonecheck_impl(dataset, zoned, cr));
488 zfs_secpolicy_write_perms_ds(const char *name, dsl_dataset_t *ds,
489 const char *perm, cred_t *cr)
493 error = zfs_dozonecheck_ds(name, ds, cr);
495 error = secpolicy_zfs(cr);
497 error = dsl_deleg_access_impl(ds, perm, cr);
503 zfs_secpolicy_write_perms(const char *name, const char *perm, cred_t *cr)
510 * First do a quick check for root in the global zone, which
511 * is allowed to do all write_perms. This ensures that zfs_ioc_*
512 * will get to handle nonexistent datasets.
514 if (INGLOBALZONE(curthread) && secpolicy_zfs(cr) == 0)
517 error = dsl_pool_hold(name, FTAG, &dp);
521 error = dsl_dataset_hold(dp, name, FTAG, &ds);
523 dsl_pool_rele(dp, FTAG);
527 error = zfs_secpolicy_write_perms_ds(name, ds, perm, cr);
529 dsl_dataset_rele(ds, FTAG);
530 dsl_pool_rele(dp, FTAG);
536 * Policy for setting the security label property.
538 * Returns 0 for success, non-zero for access and other errors.
541 zfs_set_slabel_policy(const char *name, char *strval, cred_t *cr)
543 char ds_hexsl[MAXNAMELEN];
544 bslabel_t ds_sl, new_sl;
545 boolean_t new_default = FALSE;
547 int needed_priv = -1;
550 /* First get the existing dataset label. */
551 error = dsl_prop_get(name, zfs_prop_to_name(ZFS_PROP_MLSLABEL),
552 1, sizeof (ds_hexsl), &ds_hexsl, NULL);
554 return (SET_ERROR(EPERM));
556 if (strcasecmp(strval, ZFS_MLSLABEL_DEFAULT) == 0)
559 /* The label must be translatable */
560 if (!new_default && (hexstr_to_label(strval, &new_sl) != 0))
561 return (SET_ERROR(EINVAL));
564 * In a non-global zone, disallow attempts to set a label that
565 * doesn't match that of the zone; otherwise no other checks
568 if (!INGLOBALZONE(curproc)) {
569 if (new_default || !blequal(&new_sl, CR_SL(CRED())))
570 return (SET_ERROR(EPERM));
575 * For global-zone datasets (i.e., those whose zoned property is
576 * "off", verify that the specified new label is valid for the
579 if (dsl_prop_get_integer(name,
580 zfs_prop_to_name(ZFS_PROP_ZONED), &zoned, NULL))
581 return (SET_ERROR(EPERM));
583 if (zfs_check_global_label(name, strval) != 0)
584 return (SET_ERROR(EPERM));
588 * If the existing dataset label is nondefault, check if the
589 * dataset is mounted (label cannot be changed while mounted).
590 * Get the zfsvfs; if there isn't one, then the dataset isn't
591 * mounted (or isn't a dataset, doesn't exist, ...).
593 if (strcasecmp(ds_hexsl, ZFS_MLSLABEL_DEFAULT) != 0) {
595 static char *setsl_tag = "setsl_tag";
598 * Try to own the dataset; abort if there is any error,
599 * (e.g., already mounted, in use, or other error).
601 error = dmu_objset_own(name, DMU_OST_ZFS, B_TRUE,
604 return (SET_ERROR(EPERM));
606 dmu_objset_disown(os, setsl_tag);
609 needed_priv = PRIV_FILE_DOWNGRADE_SL;
613 if (hexstr_to_label(strval, &new_sl) != 0)
614 return (SET_ERROR(EPERM));
616 if (blstrictdom(&ds_sl, &new_sl))
617 needed_priv = PRIV_FILE_DOWNGRADE_SL;
618 else if (blstrictdom(&new_sl, &ds_sl))
619 needed_priv = PRIV_FILE_UPGRADE_SL;
621 /* dataset currently has a default label */
623 needed_priv = PRIV_FILE_UPGRADE_SL;
627 if (needed_priv != -1)
628 return (PRIV_POLICY(cr, needed_priv, B_FALSE, EPERM, NULL));
631 #endif /* SECLABEL */
634 zfs_secpolicy_setprop(const char *dsname, zfs_prop_t prop, nvpair_t *propval,
640 * Check permissions for special properties.
645 * Disallow setting of 'zoned' from within a local zone.
647 if (!INGLOBALZONE(curthread))
648 return (SET_ERROR(EPERM));
652 case ZFS_PROP_FILESYSTEM_LIMIT:
653 case ZFS_PROP_SNAPSHOT_LIMIT:
654 if (!INGLOBALZONE(curthread)) {
656 char setpoint[ZFS_MAX_DATASET_NAME_LEN];
658 * Unprivileged users are allowed to modify the
659 * limit on things *under* (ie. contained by)
660 * the thing they own.
662 if (dsl_prop_get_integer(dsname, "jailed", &zoned,
664 return (SET_ERROR(EPERM));
665 if (!zoned || strlen(dsname) <= strlen(setpoint))
666 return (SET_ERROR(EPERM));
670 case ZFS_PROP_MLSLABEL:
672 if (!is_system_labeled())
673 return (SET_ERROR(EPERM));
675 if (nvpair_value_string(propval, &strval) == 0) {
678 err = zfs_set_slabel_policy(dsname, strval, CRED());
688 return (zfs_secpolicy_write_perms(dsname, zfs_prop_to_name(prop), cr));
693 zfs_secpolicy_set_fsacl(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
697 error = zfs_dozonecheck(zc->zc_name, cr);
702 * permission to set permissions will be evaluated later in
703 * dsl_deleg_can_allow()
710 zfs_secpolicy_rollback(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
712 return (zfs_secpolicy_write_perms(zc->zc_name,
713 ZFS_DELEG_PERM_ROLLBACK, cr));
718 zfs_secpolicy_send(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
726 * Generate the current snapshot name from the given objsetid, then
727 * use that name for the secpolicy/zone checks.
729 cp = strchr(zc->zc_name, '@');
731 return (SET_ERROR(EINVAL));
732 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
736 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &ds);
738 dsl_pool_rele(dp, FTAG);
742 dsl_dataset_name(ds, zc->zc_name);
744 error = zfs_secpolicy_write_perms_ds(zc->zc_name, ds,
745 ZFS_DELEG_PERM_SEND, cr);
746 dsl_dataset_rele(ds, FTAG);
747 dsl_pool_rele(dp, FTAG);
754 zfs_secpolicy_send_new(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
756 return (zfs_secpolicy_write_perms(zc->zc_name,
757 ZFS_DELEG_PERM_SEND, cr));
762 zfs_secpolicy_deleg_share(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
767 if ((error = lookupname(zc->zc_value, UIO_SYSSPACE,
768 NO_FOLLOW, NULL, &vp)) != 0)
771 /* Now make sure mntpnt and dataset are ZFS */
773 if (strcmp(vp->v_vfsp->mnt_stat.f_fstypename, "zfs") != 0 ||
774 (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource),
775 zc->zc_name) != 0)) {
777 return (SET_ERROR(EPERM));
781 return (dsl_deleg_access(zc->zc_name,
782 ZFS_DELEG_PERM_SHARE, cr));
786 zfs_secpolicy_share(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
788 if (!INGLOBALZONE(curthread))
789 return (SET_ERROR(EPERM));
791 if (secpolicy_nfs(cr) == 0) {
794 return (zfs_secpolicy_deleg_share(zc, innvl, cr));
799 zfs_secpolicy_smb_acl(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
801 if (!INGLOBALZONE(curthread))
802 return (SET_ERROR(EPERM));
804 if (secpolicy_smb(cr) == 0) {
807 return (zfs_secpolicy_deleg_share(zc, innvl, cr));
812 zfs_get_parent(const char *datasetname, char *parent, int parentsize)
817 * Remove the @bla or /bla from the end of the name to get the parent.
819 (void) strncpy(parent, datasetname, parentsize);
820 cp = strrchr(parent, '@');
824 cp = strrchr(parent, '/');
826 return (SET_ERROR(ENOENT));
834 zfs_secpolicy_destroy_perms(const char *name, cred_t *cr)
838 if ((error = zfs_secpolicy_write_perms(name,
839 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
842 return (zfs_secpolicy_write_perms(name, ZFS_DELEG_PERM_DESTROY, cr));
847 zfs_secpolicy_destroy(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
849 return (zfs_secpolicy_destroy_perms(zc->zc_name, cr));
853 * Destroying snapshots with delegated permissions requires
854 * descendant mount and destroy permissions.
858 zfs_secpolicy_destroy_snaps(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
861 nvpair_t *pair, *nextpair;
864 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
865 return (SET_ERROR(EINVAL));
866 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
868 nextpair = nvlist_next_nvpair(snaps, pair);
869 error = zfs_secpolicy_destroy_perms(nvpair_name(pair), cr);
870 if (error == ENOENT) {
872 * Ignore any snapshots that don't exist (we consider
873 * them "already destroyed"). Remove the name from the
874 * nvl here in case the snapshot is created between
875 * now and when we try to destroy it (in which case
876 * we don't want to destroy it since we haven't
877 * checked for permission).
879 fnvlist_remove_nvpair(snaps, pair);
890 zfs_secpolicy_rename_perms(const char *from, const char *to, cred_t *cr)
892 char parentname[ZFS_MAX_DATASET_NAME_LEN];
895 if ((error = zfs_secpolicy_write_perms(from,
896 ZFS_DELEG_PERM_RENAME, cr)) != 0)
899 if ((error = zfs_secpolicy_write_perms(from,
900 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
903 if ((error = zfs_get_parent(to, parentname,
904 sizeof (parentname))) != 0)
907 if ((error = zfs_secpolicy_write_perms(parentname,
908 ZFS_DELEG_PERM_CREATE, cr)) != 0)
911 if ((error = zfs_secpolicy_write_perms(parentname,
912 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
920 zfs_secpolicy_rename(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
925 if ((zc->zc_cookie & 1) != 0) {
927 * This is recursive rename, so the starting snapshot might
928 * not exist. Check file system or volume permission instead.
930 at = strchr(zc->zc_name, '@');
936 error = zfs_secpolicy_rename_perms(zc->zc_name, zc->zc_value, cr);
946 zfs_secpolicy_promote(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
949 dsl_dataset_t *clone;
952 error = zfs_secpolicy_write_perms(zc->zc_name,
953 ZFS_DELEG_PERM_PROMOTE, cr);
957 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
961 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &clone);
964 char parentname[ZFS_MAX_DATASET_NAME_LEN];
965 dsl_dataset_t *origin = NULL;
969 error = dsl_dataset_hold_obj(dd->dd_pool,
970 dsl_dir_phys(dd)->dd_origin_obj, FTAG, &origin);
972 dsl_dataset_rele(clone, FTAG);
973 dsl_pool_rele(dp, FTAG);
977 error = zfs_secpolicy_write_perms_ds(zc->zc_name, clone,
978 ZFS_DELEG_PERM_MOUNT, cr);
980 dsl_dataset_name(origin, parentname);
982 error = zfs_secpolicy_write_perms_ds(parentname, origin,
983 ZFS_DELEG_PERM_PROMOTE, cr);
985 dsl_dataset_rele(clone, FTAG);
986 dsl_dataset_rele(origin, FTAG);
988 dsl_pool_rele(dp, FTAG);
994 zfs_secpolicy_recv(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
998 if ((error = zfs_secpolicy_write_perms(zc->zc_name,
999 ZFS_DELEG_PERM_RECEIVE, cr)) != 0)
1002 if ((error = zfs_secpolicy_write_perms(zc->zc_name,
1003 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
1006 return (zfs_secpolicy_write_perms(zc->zc_name,
1007 ZFS_DELEG_PERM_CREATE, cr));
1011 zfs_secpolicy_snapshot_perms(const char *name, cred_t *cr)
1013 return (zfs_secpolicy_write_perms(name,
1014 ZFS_DELEG_PERM_SNAPSHOT, cr));
1018 * Check for permission to create each snapshot in the nvlist.
1022 zfs_secpolicy_snapshot(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1028 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
1029 return (SET_ERROR(EINVAL));
1030 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
1031 pair = nvlist_next_nvpair(snaps, pair)) {
1032 char *name = nvpair_name(pair);
1033 char *atp = strchr(name, '@');
1036 error = SET_ERROR(EINVAL);
1040 error = zfs_secpolicy_snapshot_perms(name, cr);
1049 * Check for permission to create each snapshot in the nvlist.
1053 zfs_secpolicy_bookmark(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1057 for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL);
1058 pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) {
1059 char *name = nvpair_name(pair);
1060 char *hashp = strchr(name, '#');
1062 if (hashp == NULL) {
1063 error = SET_ERROR(EINVAL);
1067 error = zfs_secpolicy_write_perms(name,
1068 ZFS_DELEG_PERM_BOOKMARK, cr);
1078 zfs_secpolicy_destroy_bookmarks(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1080 nvpair_t *pair, *nextpair;
1083 for (pair = nvlist_next_nvpair(innvl, NULL); pair != NULL;
1085 char *name = nvpair_name(pair);
1086 char *hashp = strchr(name, '#');
1087 nextpair = nvlist_next_nvpair(innvl, pair);
1089 if (hashp == NULL) {
1090 error = SET_ERROR(EINVAL);
1095 error = zfs_secpolicy_write_perms(name,
1096 ZFS_DELEG_PERM_DESTROY, cr);
1098 if (error == ENOENT) {
1100 * Ignore any filesystems that don't exist (we consider
1101 * their bookmarks "already destroyed"). Remove
1102 * the name from the nvl here in case the filesystem
1103 * is created between now and when we try to destroy
1104 * the bookmark (in which case we don't want to
1105 * destroy it since we haven't checked for permission).
1107 fnvlist_remove_nvpair(innvl, pair);
1119 zfs_secpolicy_log_history(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1122 * Even root must have a proper TSD so that we know what pool
1125 if (tsd_get(zfs_allow_log_key) == NULL)
1126 return (SET_ERROR(EPERM));
1131 zfs_secpolicy_create_clone(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1133 char parentname[ZFS_MAX_DATASET_NAME_LEN];
1137 if ((error = zfs_get_parent(zc->zc_name, parentname,
1138 sizeof (parentname))) != 0)
1141 if (nvlist_lookup_string(innvl, "origin", &origin) == 0 &&
1142 (error = zfs_secpolicy_write_perms(origin,
1143 ZFS_DELEG_PERM_CLONE, cr)) != 0)
1146 if ((error = zfs_secpolicy_write_perms(parentname,
1147 ZFS_DELEG_PERM_CREATE, cr)) != 0)
1150 return (zfs_secpolicy_write_perms(parentname,
1151 ZFS_DELEG_PERM_MOUNT, cr));
1155 * Policy for pool operations - create/destroy pools, add vdevs, etc. Requires
1156 * SYS_CONFIG privilege, which is not available in a local zone.
1160 zfs_secpolicy_config(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1162 if (secpolicy_sys_config(cr, B_FALSE) != 0)
1163 return (SET_ERROR(EPERM));
1169 * Policy for object to name lookups.
1173 zfs_secpolicy_diff(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1177 if ((error = secpolicy_sys_config(cr, B_FALSE)) == 0)
1180 error = zfs_secpolicy_write_perms(zc->zc_name, ZFS_DELEG_PERM_DIFF, cr);
1185 * Policy for fault injection. Requires all privileges.
1189 zfs_secpolicy_inject(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1191 return (secpolicy_zinject(cr));
1196 zfs_secpolicy_inherit_prop(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1198 zfs_prop_t prop = zfs_name_to_prop(zc->zc_value);
1200 if (prop == ZPROP_INVAL) {
1201 if (!zfs_prop_user(zc->zc_value))
1202 return (SET_ERROR(EINVAL));
1203 return (zfs_secpolicy_write_perms(zc->zc_name,
1204 ZFS_DELEG_PERM_USERPROP, cr));
1206 return (zfs_secpolicy_setprop(zc->zc_name, prop,
1212 zfs_secpolicy_userspace_one(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1214 int err = zfs_secpolicy_read(zc, innvl, cr);
1218 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
1219 return (SET_ERROR(EINVAL));
1221 if (zc->zc_value[0] == 0) {
1223 * They are asking about a posix uid/gid. If it's
1224 * themself, allow it.
1226 if (zc->zc_objset_type == ZFS_PROP_USERUSED ||
1227 zc->zc_objset_type == ZFS_PROP_USERQUOTA) {
1228 if (zc->zc_guid == crgetuid(cr))
1231 if (groupmember(zc->zc_guid, cr))
1236 return (zfs_secpolicy_write_perms(zc->zc_name,
1237 userquota_perms[zc->zc_objset_type], cr));
1241 zfs_secpolicy_userspace_many(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1243 int err = zfs_secpolicy_read(zc, innvl, cr);
1247 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
1248 return (SET_ERROR(EINVAL));
1250 return (zfs_secpolicy_write_perms(zc->zc_name,
1251 userquota_perms[zc->zc_objset_type], cr));
1256 zfs_secpolicy_userspace_upgrade(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1258 return (zfs_secpolicy_setprop(zc->zc_name, ZFS_PROP_VERSION,
1264 zfs_secpolicy_hold(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1270 error = nvlist_lookup_nvlist(innvl, "holds", &holds);
1272 return (SET_ERROR(EINVAL));
1274 for (pair = nvlist_next_nvpair(holds, NULL); pair != NULL;
1275 pair = nvlist_next_nvpair(holds, pair)) {
1276 char fsname[ZFS_MAX_DATASET_NAME_LEN];
1277 error = dmu_fsname(nvpair_name(pair), fsname);
1280 error = zfs_secpolicy_write_perms(fsname,
1281 ZFS_DELEG_PERM_HOLD, cr);
1290 zfs_secpolicy_release(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1295 for (pair = nvlist_next_nvpair(innvl, NULL); pair != NULL;
1296 pair = nvlist_next_nvpair(innvl, pair)) {
1297 char fsname[ZFS_MAX_DATASET_NAME_LEN];
1298 error = dmu_fsname(nvpair_name(pair), fsname);
1301 error = zfs_secpolicy_write_perms(fsname,
1302 ZFS_DELEG_PERM_RELEASE, cr);
1310 * Policy for allowing temporary snapshots to be taken or released
1313 zfs_secpolicy_tmp_snapshot(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1316 * A temporary snapshot is the same as a snapshot,
1317 * hold, destroy and release all rolled into one.
1318 * Delegated diff alone is sufficient that we allow this.
1322 if ((error = zfs_secpolicy_write_perms(zc->zc_name,
1323 ZFS_DELEG_PERM_DIFF, cr)) == 0)
1326 error = zfs_secpolicy_snapshot_perms(zc->zc_name, cr);
1328 error = zfs_secpolicy_hold(zc, innvl, cr);
1330 error = zfs_secpolicy_release(zc, innvl, cr);
1332 error = zfs_secpolicy_destroy(zc, innvl, cr);
1337 * Returns the nvlist as specified by the user in the zfs_cmd_t.
1340 get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp)
1344 nvlist_t *list = NULL;
1347 * Read in and unpack the user-supplied nvlist.
1350 return (SET_ERROR(EINVAL));
1352 packed = kmem_alloc(size, KM_SLEEP);
1354 if ((error = ddi_copyin((void *)(uintptr_t)nvl, packed, size,
1356 kmem_free(packed, size);
1357 return (SET_ERROR(EFAULT));
1360 if ((error = nvlist_unpack(packed, size, &list, 0)) != 0) {
1361 kmem_free(packed, size);
1365 kmem_free(packed, size);
1372 * Reduce the size of this nvlist until it can be serialized in 'max' bytes.
1373 * Entries will be removed from the end of the nvlist, and one int32 entry
1374 * named "N_MORE_ERRORS" will be added indicating how many entries were
1378 nvlist_smush(nvlist_t *errors, size_t max)
1382 size = fnvlist_size(errors);
1385 nvpair_t *more_errors;
1389 return (SET_ERROR(ENOMEM));
1391 fnvlist_add_int32(errors, ZPROP_N_MORE_ERRORS, 0);
1392 more_errors = nvlist_prev_nvpair(errors, NULL);
1395 nvpair_t *pair = nvlist_prev_nvpair(errors,
1397 fnvlist_remove_nvpair(errors, pair);
1399 size = fnvlist_size(errors);
1400 } while (size > max);
1402 fnvlist_remove_nvpair(errors, more_errors);
1403 fnvlist_add_int32(errors, ZPROP_N_MORE_ERRORS, n);
1404 ASSERT3U(fnvlist_size(errors), <=, max);
1411 put_nvlist(zfs_cmd_t *zc, nvlist_t *nvl)
1413 char *packed = NULL;
1417 size = fnvlist_size(nvl);
1419 if (size > zc->zc_nvlist_dst_size) {
1421 * Solaris returns ENOMEM here, because even if an error is
1422 * returned from an ioctl(2), new zc_nvlist_dst_size will be
1423 * passed to the userland. This is not the case for FreeBSD.
1424 * We need to return 0, so the kernel will copy the
1425 * zc_nvlist_dst_size back and the userland can discover that a
1426 * bigger buffer is needed.
1430 packed = fnvlist_pack(nvl, &size);
1431 if (ddi_copyout(packed, (void *)(uintptr_t)zc->zc_nvlist_dst,
1432 size, zc->zc_iflags) != 0)
1433 error = SET_ERROR(EFAULT);
1434 fnvlist_pack_free(packed, size);
1437 zc->zc_nvlist_dst_size = size;
1438 zc->zc_nvlist_dst_filled = B_TRUE;
1443 getzfsvfs_impl(objset_t *os, zfsvfs_t **zfvp)
1448 if (dmu_objset_type(os) != DMU_OST_ZFS) {
1449 return (SET_ERROR(EINVAL));
1452 mutex_enter(&os->os_user_ptr_lock);
1453 *zfvp = dmu_objset_get_user(os);
1455 vfs_ref((*zfvp)->z_vfs);
1457 error = SET_ERROR(ESRCH);
1459 mutex_exit(&os->os_user_ptr_lock);
1465 getzfsvfs(const char *dsname, zfsvfs_t **zfvp)
1470 error = dmu_objset_hold(dsname, FTAG, &os);
1474 error = getzfsvfs_impl(os, zfvp);
1475 dmu_objset_rele(os, FTAG);
1482 getzfsvfs_ref(const char *dsname, zfsvfs_t **zfvp)
1487 error = dmu_objset_hold(dsname, FTAG, &os);
1491 error = getzfsvfs_impl(os, zfvp);
1492 dmu_objset_rele(os, FTAG);
1497 getzfsvfs(const char *dsname, zfsvfs_t **zfvp)
1502 error = getzfsvfs_ref(dsname, zfvp);
1505 error = vfs_busy((*zfvp)->z_vfs, 0);
1506 vfs_rel((*zfvp)->z_vfs);
1509 error = SET_ERROR(ESRCH);
1516 * Find a zfsvfs_t for a mounted filesystem, or create our own, in which
1517 * case its z_vfs will be NULL, and it will be opened as the owner.
1518 * If 'writer' is set, the z_teardown_lock will be held for RW_WRITER,
1519 * which prevents all vnode ops from running.
1522 zfsvfs_hold(const char *name, void *tag, zfsvfs_t **zfvp, boolean_t writer)
1526 if (getzfsvfs(name, zfvp) != 0)
1527 error = zfsvfs_create(name, zfvp);
1529 rrm_enter(&(*zfvp)->z_teardown_lock, (writer) ? RW_WRITER :
1532 if ((*zfvp)->z_unmounted) {
1534 * XXX we could probably try again, since the unmounting
1535 * thread should be just about to disassociate the
1536 * objset from the zfsvfs.
1538 rrm_exit(&(*zfvp)->z_teardown_lock, tag);
1539 return (SET_ERROR(EBUSY));
1543 * vfs_busy() ensures that the filesystem is not and
1544 * can not be unmounted.
1546 ASSERT(!(*zfvp)->z_unmounted);
1553 zfsvfs_rele(zfsvfs_t *zfsvfs, void *tag)
1555 rrm_exit(&zfsvfs->z_teardown_lock, tag);
1557 if (zfsvfs->z_vfs) {
1559 VFS_RELE(zfsvfs->z_vfs);
1561 vfs_unbusy(zfsvfs->z_vfs);
1564 dmu_objset_disown(zfsvfs->z_os, zfsvfs);
1565 zfsvfs_free(zfsvfs);
1570 zfs_ioc_pool_create(zfs_cmd_t *zc)
1573 nvlist_t *config, *props = NULL;
1574 nvlist_t *rootprops = NULL;
1575 nvlist_t *zplprops = NULL;
1577 if (error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1578 zc->zc_iflags, &config))
1581 if (zc->zc_nvlist_src_size != 0 && (error =
1582 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1583 zc->zc_iflags, &props))) {
1584 nvlist_free(config);
1589 nvlist_t *nvl = NULL;
1590 uint64_t version = SPA_VERSION;
1592 (void) nvlist_lookup_uint64(props,
1593 zpool_prop_to_name(ZPOOL_PROP_VERSION), &version);
1594 if (!SPA_VERSION_IS_SUPPORTED(version)) {
1595 error = SET_ERROR(EINVAL);
1596 goto pool_props_bad;
1598 (void) nvlist_lookup_nvlist(props, ZPOOL_ROOTFS_PROPS, &nvl);
1600 error = nvlist_dup(nvl, &rootprops, KM_SLEEP);
1602 nvlist_free(config);
1606 (void) nvlist_remove_all(props, ZPOOL_ROOTFS_PROPS);
1608 VERIFY(nvlist_alloc(&zplprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
1609 error = zfs_fill_zplprops_root(version, rootprops,
1612 goto pool_props_bad;
1615 error = spa_create(zc->zc_name, config, props, zplprops);
1618 * Set the remaining root properties
1620 if (!error && (error = zfs_set_prop_nvlist(zc->zc_name,
1621 ZPROP_SRC_LOCAL, rootprops, NULL)) != 0)
1622 (void) spa_destroy(zc->zc_name);
1625 nvlist_free(rootprops);
1626 nvlist_free(zplprops);
1627 nvlist_free(config);
1634 zfs_ioc_pool_destroy(zfs_cmd_t *zc)
1637 zfs_log_history(zc);
1638 error = spa_destroy(zc->zc_name);
1640 zvol_remove_minors(zc->zc_name);
1645 zfs_ioc_pool_import(zfs_cmd_t *zc)
1647 nvlist_t *config, *props = NULL;
1651 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1652 zc->zc_iflags, &config)) != 0)
1655 if (zc->zc_nvlist_src_size != 0 && (error =
1656 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1657 zc->zc_iflags, &props))) {
1658 nvlist_free(config);
1662 if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, &guid) != 0 ||
1663 guid != zc->zc_guid)
1664 error = SET_ERROR(EINVAL);
1666 error = spa_import(zc->zc_name, config, props, zc->zc_cookie);
1668 if (zc->zc_nvlist_dst != 0) {
1671 if ((err = put_nvlist(zc, config)) != 0)
1675 nvlist_free(config);
1683 zfs_ioc_pool_export(zfs_cmd_t *zc)
1686 boolean_t force = (boolean_t)zc->zc_cookie;
1687 boolean_t hardforce = (boolean_t)zc->zc_guid;
1689 zfs_log_history(zc);
1690 error = spa_export(zc->zc_name, NULL, force, hardforce);
1692 zvol_remove_minors(zc->zc_name);
1697 zfs_ioc_pool_configs(zfs_cmd_t *zc)
1702 if ((configs = spa_all_configs(&zc->zc_cookie)) == NULL)
1703 return (SET_ERROR(EEXIST));
1705 error = put_nvlist(zc, configs);
1707 nvlist_free(configs);
1714 * zc_name name of the pool
1717 * zc_cookie real errno
1718 * zc_nvlist_dst config nvlist
1719 * zc_nvlist_dst_size size of config nvlist
1722 zfs_ioc_pool_stats(zfs_cmd_t *zc)
1728 error = spa_get_stats(zc->zc_name, &config, zc->zc_value,
1729 sizeof (zc->zc_value));
1731 if (config != NULL) {
1732 ret = put_nvlist(zc, config);
1733 nvlist_free(config);
1736 * The config may be present even if 'error' is non-zero.
1737 * In this case we return success, and preserve the real errno
1740 zc->zc_cookie = error;
1749 * Try to import the given pool, returning pool stats as appropriate so that
1750 * user land knows which devices are available and overall pool health.
1753 zfs_ioc_pool_tryimport(zfs_cmd_t *zc)
1755 nvlist_t *tryconfig, *config;
1758 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1759 zc->zc_iflags, &tryconfig)) != 0)
1762 config = spa_tryimport(tryconfig);
1764 nvlist_free(tryconfig);
1767 return (SET_ERROR(EINVAL));
1769 error = put_nvlist(zc, config);
1770 nvlist_free(config);
1777 * zc_name name of the pool
1778 * zc_cookie scan func (pool_scan_func_t)
1779 * zc_flags scrub pause/resume flag (pool_scrub_cmd_t)
1782 zfs_ioc_pool_scan(zfs_cmd_t *zc)
1787 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1790 if (zc->zc_flags >= POOL_SCRUB_FLAGS_END)
1791 return (SET_ERROR(EINVAL));
1793 if (zc->zc_flags == POOL_SCRUB_PAUSE)
1794 error = spa_scrub_pause_resume(spa, POOL_SCRUB_PAUSE);
1795 else if (zc->zc_cookie == POOL_SCAN_NONE)
1796 error = spa_scan_stop(spa);
1798 error = spa_scan(spa, zc->zc_cookie);
1800 spa_close(spa, FTAG);
1806 zfs_ioc_pool_freeze(zfs_cmd_t *zc)
1811 error = spa_open(zc->zc_name, &spa, FTAG);
1814 spa_close(spa, FTAG);
1820 zfs_ioc_pool_upgrade(zfs_cmd_t *zc)
1825 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1828 if (zc->zc_cookie < spa_version(spa) ||
1829 !SPA_VERSION_IS_SUPPORTED(zc->zc_cookie)) {
1830 spa_close(spa, FTAG);
1831 return (SET_ERROR(EINVAL));
1834 spa_upgrade(spa, zc->zc_cookie);
1835 spa_close(spa, FTAG);
1841 zfs_ioc_pool_get_history(zfs_cmd_t *zc)
1848 if ((size = zc->zc_history_len) == 0)
1849 return (SET_ERROR(EINVAL));
1851 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1854 if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) {
1855 spa_close(spa, FTAG);
1856 return (SET_ERROR(ENOTSUP));
1859 hist_buf = kmem_alloc(size, KM_SLEEP);
1860 if ((error = spa_history_get(spa, &zc->zc_history_offset,
1861 &zc->zc_history_len, hist_buf)) == 0) {
1862 error = ddi_copyout(hist_buf,
1863 (void *)(uintptr_t)zc->zc_history,
1864 zc->zc_history_len, zc->zc_iflags);
1867 spa_close(spa, FTAG);
1868 kmem_free(hist_buf, size);
1873 zfs_ioc_pool_reguid(zfs_cmd_t *zc)
1878 error = spa_open(zc->zc_name, &spa, FTAG);
1880 error = spa_change_guid(spa);
1881 spa_close(spa, FTAG);
1887 zfs_ioc_dsobj_to_dsname(zfs_cmd_t *zc)
1889 return (dsl_dsobj_to_dsname(zc->zc_name, zc->zc_obj, zc->zc_value));
1894 * zc_name name of filesystem
1895 * zc_obj object to find
1898 * zc_value name of object
1901 zfs_ioc_obj_to_path(zfs_cmd_t *zc)
1906 /* XXX reading from objset not owned */
1907 if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os)) != 0)
1909 if (dmu_objset_type(os) != DMU_OST_ZFS) {
1910 dmu_objset_rele(os, FTAG);
1911 return (SET_ERROR(EINVAL));
1913 error = zfs_obj_to_path(os, zc->zc_obj, zc->zc_value,
1914 sizeof (zc->zc_value));
1915 dmu_objset_rele(os, FTAG);
1922 * zc_name name of filesystem
1923 * zc_obj object to find
1926 * zc_stat stats on object
1927 * zc_value path to object
1930 zfs_ioc_obj_to_stats(zfs_cmd_t *zc)
1935 /* XXX reading from objset not owned */
1936 if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os)) != 0)
1938 if (dmu_objset_type(os) != DMU_OST_ZFS) {
1939 dmu_objset_rele(os, FTAG);
1940 return (SET_ERROR(EINVAL));
1942 error = zfs_obj_to_stats(os, zc->zc_obj, &zc->zc_stat, zc->zc_value,
1943 sizeof (zc->zc_value));
1944 dmu_objset_rele(os, FTAG);
1950 zfs_ioc_vdev_add(zfs_cmd_t *zc)
1954 nvlist_t *config, **l2cache, **spares;
1955 uint_t nl2cache = 0, nspares = 0;
1957 error = spa_open(zc->zc_name, &spa, FTAG);
1961 error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1962 zc->zc_iflags, &config);
1963 (void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_L2CACHE,
1964 &l2cache, &nl2cache);
1966 (void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_SPARES,
1971 * A root pool with concatenated devices is not supported.
1972 * Thus, can not add a device to a root pool.
1974 * Intent log device can not be added to a rootpool because
1975 * during mountroot, zil is replayed, a seperated log device
1976 * can not be accessed during the mountroot time.
1978 * l2cache and spare devices are ok to be added to a rootpool.
1980 if (spa_bootfs(spa) != 0 && nl2cache == 0 && nspares == 0) {
1981 nvlist_free(config);
1982 spa_close(spa, FTAG);
1983 return (SET_ERROR(EDOM));
1985 #endif /* illumos */
1988 error = spa_vdev_add(spa, config);
1989 nvlist_free(config);
1991 spa_close(spa, FTAG);
1997 * zc_name name of the pool
1998 * zc_nvlist_conf nvlist of devices to remove
1999 * zc_cookie to stop the remove?
2002 zfs_ioc_vdev_remove(zfs_cmd_t *zc)
2007 error = spa_open(zc->zc_name, &spa, FTAG);
2010 error = spa_vdev_remove(spa, zc->zc_guid, B_FALSE);
2011 spa_close(spa, FTAG);
2016 zfs_ioc_vdev_set_state(zfs_cmd_t *zc)
2020 vdev_state_t newstate = VDEV_STATE_UNKNOWN;
2022 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
2024 switch (zc->zc_cookie) {
2025 case VDEV_STATE_ONLINE:
2026 error = vdev_online(spa, zc->zc_guid, zc->zc_obj, &newstate);
2029 case VDEV_STATE_OFFLINE:
2030 error = vdev_offline(spa, zc->zc_guid, zc->zc_obj);
2033 case VDEV_STATE_FAULTED:
2034 if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED &&
2035 zc->zc_obj != VDEV_AUX_EXTERNAL)
2036 zc->zc_obj = VDEV_AUX_ERR_EXCEEDED;
2038 error = vdev_fault(spa, zc->zc_guid, zc->zc_obj);
2041 case VDEV_STATE_DEGRADED:
2042 if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED &&
2043 zc->zc_obj != VDEV_AUX_EXTERNAL)
2044 zc->zc_obj = VDEV_AUX_ERR_EXCEEDED;
2046 error = vdev_degrade(spa, zc->zc_guid, zc->zc_obj);
2050 error = SET_ERROR(EINVAL);
2052 zc->zc_cookie = newstate;
2053 spa_close(spa, FTAG);
2058 zfs_ioc_vdev_attach(zfs_cmd_t *zc)
2061 int replacing = zc->zc_cookie;
2065 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
2068 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
2069 zc->zc_iflags, &config)) == 0) {
2070 error = spa_vdev_attach(spa, zc->zc_guid, config, replacing);
2071 nvlist_free(config);
2074 spa_close(spa, FTAG);
2079 zfs_ioc_vdev_detach(zfs_cmd_t *zc)
2084 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
2087 error = spa_vdev_detach(spa, zc->zc_guid, 0, B_FALSE);
2089 spa_close(spa, FTAG);
2094 zfs_ioc_vdev_split(zfs_cmd_t *zc)
2097 nvlist_t *config, *props = NULL;
2099 boolean_t exp = !!(zc->zc_cookie & ZPOOL_EXPORT_AFTER_SPLIT);
2101 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
2104 if (error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
2105 zc->zc_iflags, &config)) {
2106 spa_close(spa, FTAG);
2110 if (zc->zc_nvlist_src_size != 0 && (error =
2111 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2112 zc->zc_iflags, &props))) {
2113 spa_close(spa, FTAG);
2114 nvlist_free(config);
2118 error = spa_vdev_split_mirror(spa, zc->zc_string, config, props, exp);
2120 spa_close(spa, FTAG);
2122 nvlist_free(config);
2129 zfs_ioc_vdev_setpath(zfs_cmd_t *zc)
2132 char *path = zc->zc_value;
2133 uint64_t guid = zc->zc_guid;
2136 error = spa_open(zc->zc_name, &spa, FTAG);
2140 error = spa_vdev_setpath(spa, guid, path);
2141 spa_close(spa, FTAG);
2146 zfs_ioc_vdev_setfru(zfs_cmd_t *zc)
2149 char *fru = zc->zc_value;
2150 uint64_t guid = zc->zc_guid;
2153 error = spa_open(zc->zc_name, &spa, FTAG);
2157 error = spa_vdev_setfru(spa, guid, fru);
2158 spa_close(spa, FTAG);
2163 zfs_ioc_objset_stats_impl(zfs_cmd_t *zc, objset_t *os)
2168 dmu_objset_fast_stat(os, &zc->zc_objset_stats);
2170 if (zc->zc_nvlist_dst != 0 &&
2171 (error = dsl_prop_get_all(os, &nv)) == 0) {
2172 dmu_objset_stats(os, nv);
2174 * NB: zvol_get_stats() will read the objset contents,
2175 * which we aren't supposed to do with a
2176 * DS_MODE_USER hold, because it could be
2177 * inconsistent. So this is a bit of a workaround...
2178 * XXX reading with out owning
2180 if (!zc->zc_objset_stats.dds_inconsistent &&
2181 dmu_objset_type(os) == DMU_OST_ZVOL) {
2182 error = zvol_get_stats(os, nv);
2187 error = put_nvlist(zc, nv);
2196 * zc_name name of filesystem
2197 * zc_nvlist_dst_size size of buffer for property nvlist
2200 * zc_objset_stats stats
2201 * zc_nvlist_dst property nvlist
2202 * zc_nvlist_dst_size size of property nvlist
2205 zfs_ioc_objset_stats(zfs_cmd_t *zc)
2210 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
2212 error = zfs_ioc_objset_stats_impl(zc, os);
2213 dmu_objset_rele(os, FTAG);
2216 if (error == ENOMEM)
2223 * zc_name name of filesystem
2224 * zc_nvlist_dst_size size of buffer for property nvlist
2227 * zc_nvlist_dst received property nvlist
2228 * zc_nvlist_dst_size size of received property nvlist
2230 * Gets received properties (distinct from local properties on or after
2231 * SPA_VERSION_RECVD_PROPS) for callers who want to differentiate received from
2232 * local property values.
2235 zfs_ioc_objset_recvd_props(zfs_cmd_t *zc)
2241 * Without this check, we would return local property values if the
2242 * caller has not already received properties on or after
2243 * SPA_VERSION_RECVD_PROPS.
2245 if (!dsl_prop_get_hasrecvd(zc->zc_name))
2246 return (SET_ERROR(ENOTSUP));
2248 if (zc->zc_nvlist_dst != 0 &&
2249 (error = dsl_prop_get_received(zc->zc_name, &nv)) == 0) {
2250 error = put_nvlist(zc, nv);
2258 nvl_add_zplprop(objset_t *os, nvlist_t *props, zfs_prop_t prop)
2264 * zfs_get_zplprop() will either find a value or give us
2265 * the default value (if there is one).
2267 if ((error = zfs_get_zplprop(os, prop, &value)) != 0)
2269 VERIFY(nvlist_add_uint64(props, zfs_prop_to_name(prop), value) == 0);
2275 * zc_name name of filesystem
2276 * zc_nvlist_dst_size size of buffer for zpl property nvlist
2279 * zc_nvlist_dst zpl property nvlist
2280 * zc_nvlist_dst_size size of zpl property nvlist
2283 zfs_ioc_objset_zplprops(zfs_cmd_t *zc)
2288 /* XXX reading without owning */
2289 if (err = dmu_objset_hold(zc->zc_name, FTAG, &os))
2292 dmu_objset_fast_stat(os, &zc->zc_objset_stats);
2295 * NB: nvl_add_zplprop() will read the objset contents,
2296 * which we aren't supposed to do with a DS_MODE_USER
2297 * hold, because it could be inconsistent.
2299 if (zc->zc_nvlist_dst != 0 &&
2300 !zc->zc_objset_stats.dds_inconsistent &&
2301 dmu_objset_type(os) == DMU_OST_ZFS) {
2304 VERIFY(nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2305 if ((err = nvl_add_zplprop(os, nv, ZFS_PROP_VERSION)) == 0 &&
2306 (err = nvl_add_zplprop(os, nv, ZFS_PROP_NORMALIZE)) == 0 &&
2307 (err = nvl_add_zplprop(os, nv, ZFS_PROP_UTF8ONLY)) == 0 &&
2308 (err = nvl_add_zplprop(os, nv, ZFS_PROP_CASE)) == 0)
2309 err = put_nvlist(zc, nv);
2312 err = SET_ERROR(ENOENT);
2314 dmu_objset_rele(os, FTAG);
2319 dataset_name_hidden(const char *name)
2322 * Skip over datasets that are not visible in this zone,
2323 * internal datasets (which have a $ in their name), and
2324 * temporary datasets (which have a % in their name).
2326 if (strchr(name, '$') != NULL)
2328 if (strchr(name, '%') != NULL)
2330 if (!INGLOBALZONE(curthread) && !zone_dataset_visible(name, NULL))
2337 * zc_name name of filesystem
2338 * zc_cookie zap cursor
2339 * zc_nvlist_dst_size size of buffer for property nvlist
2342 * zc_name name of next filesystem
2343 * zc_cookie zap cursor
2344 * zc_objset_stats stats
2345 * zc_nvlist_dst property nvlist
2346 * zc_nvlist_dst_size size of property nvlist
2349 zfs_ioc_dataset_list_next(zfs_cmd_t *zc)
2354 size_t orig_len = strlen(zc->zc_name);
2357 if (error = dmu_objset_hold(zc->zc_name, FTAG, &os)) {
2358 if (error == ENOENT)
2359 error = SET_ERROR(ESRCH);
2363 p = strrchr(zc->zc_name, '/');
2364 if (p == NULL || p[1] != '\0')
2365 (void) strlcat(zc->zc_name, "/", sizeof (zc->zc_name));
2366 p = zc->zc_name + strlen(zc->zc_name);
2369 error = dmu_dir_list_next(os,
2370 sizeof (zc->zc_name) - (p - zc->zc_name), p,
2371 NULL, &zc->zc_cookie);
2372 if (error == ENOENT)
2373 error = SET_ERROR(ESRCH);
2374 } while (error == 0 && dataset_name_hidden(zc->zc_name));
2375 dmu_objset_rele(os, FTAG);
2378 * If it's an internal dataset (ie. with a '$' in its name),
2379 * don't try to get stats for it, otherwise we'll return ENOENT.
2381 if (error == 0 && strchr(zc->zc_name, '$') == NULL) {
2382 error = zfs_ioc_objset_stats(zc); /* fill in the stats */
2383 if (error == ENOENT) {
2384 /* We lost a race with destroy, get the next one. */
2385 zc->zc_name[orig_len] = '\0';
2394 * zc_name name of filesystem
2395 * zc_cookie zap cursor
2396 * zc_nvlist_dst_size size of buffer for property nvlist
2397 * zc_simple when set, only name is requested
2400 * zc_name name of next snapshot
2401 * zc_objset_stats stats
2402 * zc_nvlist_dst property nvlist
2403 * zc_nvlist_dst_size size of property nvlist
2406 zfs_ioc_snapshot_list_next(zfs_cmd_t *zc)
2411 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
2413 return (error == ENOENT ? ESRCH : error);
2417 * A dataset name of maximum length cannot have any snapshots,
2418 * so exit immediately.
2420 if (strlcat(zc->zc_name, "@", sizeof (zc->zc_name)) >=
2421 ZFS_MAX_DATASET_NAME_LEN) {
2422 dmu_objset_rele(os, FTAG);
2423 return (SET_ERROR(ESRCH));
2426 error = dmu_snapshot_list_next(os,
2427 sizeof (zc->zc_name) - strlen(zc->zc_name),
2428 zc->zc_name + strlen(zc->zc_name), &zc->zc_obj, &zc->zc_cookie,
2431 if (error == 0 && !zc->zc_simple) {
2433 dsl_pool_t *dp = os->os_dsl_dataset->ds_dir->dd_pool;
2435 error = dsl_dataset_hold_obj(dp, zc->zc_obj, FTAG, &ds);
2439 error = dmu_objset_from_ds(ds, &ossnap);
2441 error = zfs_ioc_objset_stats_impl(zc, ossnap);
2442 dsl_dataset_rele(ds, FTAG);
2444 } else if (error == ENOENT) {
2445 error = SET_ERROR(ESRCH);
2448 dmu_objset_rele(os, FTAG);
2449 /* if we failed, undo the @ that we tacked on to zc_name */
2451 *strchr(zc->zc_name, '@') = '\0';
2456 zfs_prop_set_userquota(const char *dsname, nvpair_t *pair)
2458 const char *propname = nvpair_name(pair);
2460 unsigned int vallen;
2463 zfs_userquota_prop_t type;
2469 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2471 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
2472 if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2474 return (SET_ERROR(EINVAL));
2478 * A correctly constructed propname is encoded as
2479 * userquota@<rid>-<domain>.
2481 if ((dash = strchr(propname, '-')) == NULL ||
2482 nvpair_value_uint64_array(pair, &valary, &vallen) != 0 ||
2484 return (SET_ERROR(EINVAL));
2491 err = zfsvfs_hold(dsname, FTAG, &zfsvfs, B_FALSE);
2493 err = zfs_set_userquota(zfsvfs, type, domain, rid, quota);
2494 zfsvfs_rele(zfsvfs, FTAG);
2501 * If the named property is one that has a special function to set its value,
2502 * return 0 on success and a positive error code on failure; otherwise if it is
2503 * not one of the special properties handled by this function, return -1.
2505 * XXX: It would be better for callers of the property interface if we handled
2506 * these special cases in dsl_prop.c (in the dsl layer).
2509 zfs_prop_set_special(const char *dsname, zprop_source_t source,
2512 const char *propname = nvpair_name(pair);
2513 zfs_prop_t prop = zfs_name_to_prop(propname);
2517 if (prop == ZPROP_INVAL) {
2518 if (zfs_prop_userquota(propname))
2519 return (zfs_prop_set_userquota(dsname, pair));
2523 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2525 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
2526 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2530 if (zfs_prop_get_type(prop) == PROP_TYPE_STRING)
2533 VERIFY(0 == nvpair_value_uint64(pair, &intval));
2536 case ZFS_PROP_QUOTA:
2537 err = dsl_dir_set_quota(dsname, source, intval);
2539 case ZFS_PROP_REFQUOTA:
2540 err = dsl_dataset_set_refquota(dsname, source, intval);
2542 case ZFS_PROP_FILESYSTEM_LIMIT:
2543 case ZFS_PROP_SNAPSHOT_LIMIT:
2544 if (intval == UINT64_MAX) {
2545 /* clearing the limit, just do it */
2548 err = dsl_dir_activate_fs_ss_limit(dsname);
2551 * Set err to -1 to force the zfs_set_prop_nvlist code down the
2552 * default path to set the value in the nvlist.
2557 case ZFS_PROP_RESERVATION:
2558 err = dsl_dir_set_reservation(dsname, source, intval);
2560 case ZFS_PROP_REFRESERVATION:
2561 err = dsl_dataset_set_refreservation(dsname, source, intval);
2563 case ZFS_PROP_VOLSIZE:
2564 err = zvol_set_volsize(dsname, intval);
2566 case ZFS_PROP_VERSION:
2570 if ((err = zfsvfs_hold(dsname, FTAG, &zfsvfs, B_TRUE)) != 0)
2573 err = zfs_set_version(zfsvfs, intval);
2574 zfsvfs_rele(zfsvfs, FTAG);
2576 if (err == 0 && intval >= ZPL_VERSION_USERSPACE) {
2579 zc = kmem_zalloc(sizeof (zfs_cmd_t), KM_SLEEP);
2580 (void) strcpy(zc->zc_name, dsname);
2581 (void) zfs_ioc_userspace_upgrade(zc);
2582 kmem_free(zc, sizeof (zfs_cmd_t));
2594 * This function is best effort. If it fails to set any of the given properties,
2595 * it continues to set as many as it can and returns the last error
2596 * encountered. If the caller provides a non-NULL errlist, it will be filled in
2597 * with the list of names of all the properties that failed along with the
2598 * corresponding error numbers.
2600 * If every property is set successfully, zero is returned and errlist is not
2604 zfs_set_prop_nvlist(const char *dsname, zprop_source_t source, nvlist_t *nvl,
2612 nvlist_t *genericnvl = fnvlist_alloc();
2613 nvlist_t *retrynvl = fnvlist_alloc();
2617 while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) {
2618 const char *propname = nvpair_name(pair);
2619 zfs_prop_t prop = zfs_name_to_prop(propname);
2622 /* decode the property value */
2624 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2626 attrs = fnvpair_value_nvlist(pair);
2627 if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2629 err = SET_ERROR(EINVAL);
2632 /* Validate value type */
2633 if (err == 0 && prop == ZPROP_INVAL) {
2634 if (zfs_prop_user(propname)) {
2635 if (nvpair_type(propval) != DATA_TYPE_STRING)
2636 err = SET_ERROR(EINVAL);
2637 } else if (zfs_prop_userquota(propname)) {
2638 if (nvpair_type(propval) !=
2639 DATA_TYPE_UINT64_ARRAY)
2640 err = SET_ERROR(EINVAL);
2642 err = SET_ERROR(EINVAL);
2644 } else if (err == 0) {
2645 if (nvpair_type(propval) == DATA_TYPE_STRING) {
2646 if (zfs_prop_get_type(prop) != PROP_TYPE_STRING)
2647 err = SET_ERROR(EINVAL);
2648 } else if (nvpair_type(propval) == DATA_TYPE_UINT64) {
2651 intval = fnvpair_value_uint64(propval);
2653 switch (zfs_prop_get_type(prop)) {
2654 case PROP_TYPE_NUMBER:
2656 case PROP_TYPE_STRING:
2657 err = SET_ERROR(EINVAL);
2659 case PROP_TYPE_INDEX:
2660 if (zfs_prop_index_to_string(prop,
2661 intval, &unused) != 0)
2662 err = SET_ERROR(EINVAL);
2666 "unknown property type");
2669 err = SET_ERROR(EINVAL);
2673 /* Validate permissions */
2675 err = zfs_check_settable(dsname, pair, CRED());
2678 err = zfs_prop_set_special(dsname, source, pair);
2681 * For better performance we build up a list of
2682 * properties to set in a single transaction.
2684 err = nvlist_add_nvpair(genericnvl, pair);
2685 } else if (err != 0 && nvl != retrynvl) {
2687 * This may be a spurious error caused by
2688 * receiving quota and reservation out of order.
2689 * Try again in a second pass.
2691 err = nvlist_add_nvpair(retrynvl, pair);
2696 if (errlist != NULL)
2697 fnvlist_add_int32(errlist, propname, err);
2702 if (nvl != retrynvl && !nvlist_empty(retrynvl)) {
2707 if (!nvlist_empty(genericnvl) &&
2708 dsl_props_set(dsname, source, genericnvl) != 0) {
2710 * If this fails, we still want to set as many properties as we
2711 * can, so try setting them individually.
2714 while ((pair = nvlist_next_nvpair(genericnvl, pair)) != NULL) {
2715 const char *propname = nvpair_name(pair);
2719 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2721 attrs = fnvpair_value_nvlist(pair);
2722 propval = fnvlist_lookup_nvpair(attrs,
2726 if (nvpair_type(propval) == DATA_TYPE_STRING) {
2727 strval = fnvpair_value_string(propval);
2728 err = dsl_prop_set_string(dsname, propname,
2731 intval = fnvpair_value_uint64(propval);
2732 err = dsl_prop_set_int(dsname, propname, source,
2737 if (errlist != NULL) {
2738 fnvlist_add_int32(errlist, propname,
2745 nvlist_free(genericnvl);
2746 nvlist_free(retrynvl);
2752 * Check that all the properties are valid user properties.
2755 zfs_check_userprops(const char *fsname, nvlist_t *nvl)
2757 nvpair_t *pair = NULL;
2760 while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) {
2761 const char *propname = nvpair_name(pair);
2763 if (!zfs_prop_user(propname) ||
2764 nvpair_type(pair) != DATA_TYPE_STRING)
2765 return (SET_ERROR(EINVAL));
2767 if (error = zfs_secpolicy_write_perms(fsname,
2768 ZFS_DELEG_PERM_USERPROP, CRED()))
2771 if (strlen(propname) >= ZAP_MAXNAMELEN)
2772 return (SET_ERROR(ENAMETOOLONG));
2774 if (strlen(fnvpair_value_string(pair)) >= ZAP_MAXVALUELEN)
2781 props_skip(nvlist_t *props, nvlist_t *skipped, nvlist_t **newprops)
2785 VERIFY(nvlist_alloc(newprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2788 while ((pair = nvlist_next_nvpair(props, pair)) != NULL) {
2789 if (nvlist_exists(skipped, nvpair_name(pair)))
2792 VERIFY(nvlist_add_nvpair(*newprops, pair) == 0);
2797 clear_received_props(const char *dsname, nvlist_t *props,
2801 nvlist_t *cleared_props = NULL;
2802 props_skip(props, skipped, &cleared_props);
2803 if (!nvlist_empty(cleared_props)) {
2805 * Acts on local properties until the dataset has received
2806 * properties at least once on or after SPA_VERSION_RECVD_PROPS.
2808 zprop_source_t flags = (ZPROP_SRC_NONE |
2809 (dsl_prop_get_hasrecvd(dsname) ? ZPROP_SRC_RECEIVED : 0));
2810 err = zfs_set_prop_nvlist(dsname, flags, cleared_props, NULL);
2812 nvlist_free(cleared_props);
2818 * zc_name name of filesystem
2819 * zc_value name of property to set
2820 * zc_nvlist_src{_size} nvlist of properties to apply
2821 * zc_cookie received properties flag
2824 * zc_nvlist_dst{_size} error for each unapplied received property
2827 zfs_ioc_set_prop(zfs_cmd_t *zc)
2830 boolean_t received = zc->zc_cookie;
2831 zprop_source_t source = (received ? ZPROP_SRC_RECEIVED :
2836 if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2837 zc->zc_iflags, &nvl)) != 0)
2841 nvlist_t *origprops;
2843 if (dsl_prop_get_received(zc->zc_name, &origprops) == 0) {
2844 (void) clear_received_props(zc->zc_name,
2846 nvlist_free(origprops);
2849 error = dsl_prop_set_hasrecvd(zc->zc_name);
2852 errors = fnvlist_alloc();
2854 error = zfs_set_prop_nvlist(zc->zc_name, source, nvl, errors);
2856 if (zc->zc_nvlist_dst != 0 && errors != NULL) {
2857 (void) put_nvlist(zc, errors);
2860 nvlist_free(errors);
2867 * zc_name name of filesystem
2868 * zc_value name of property to inherit
2869 * zc_cookie revert to received value if TRUE
2874 zfs_ioc_inherit_prop(zfs_cmd_t *zc)
2876 const char *propname = zc->zc_value;
2877 zfs_prop_t prop = zfs_name_to_prop(propname);
2878 boolean_t received = zc->zc_cookie;
2879 zprop_source_t source = (received
2880 ? ZPROP_SRC_NONE /* revert to received value, if any */
2881 : ZPROP_SRC_INHERITED); /* explicitly inherit */
2890 * zfs_prop_set_special() expects properties in the form of an
2891 * nvpair with type info.
2893 if (prop == ZPROP_INVAL) {
2894 if (!zfs_prop_user(propname))
2895 return (SET_ERROR(EINVAL));
2897 type = PROP_TYPE_STRING;
2898 } else if (prop == ZFS_PROP_VOLSIZE ||
2899 prop == ZFS_PROP_VERSION) {
2900 return (SET_ERROR(EINVAL));
2902 type = zfs_prop_get_type(prop);
2905 VERIFY(nvlist_alloc(&dummy, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2908 case PROP_TYPE_STRING:
2909 VERIFY(0 == nvlist_add_string(dummy, propname, ""));
2911 case PROP_TYPE_NUMBER:
2912 case PROP_TYPE_INDEX:
2913 VERIFY(0 == nvlist_add_uint64(dummy, propname, 0));
2917 return (SET_ERROR(EINVAL));
2920 pair = nvlist_next_nvpair(dummy, NULL);
2921 err = zfs_prop_set_special(zc->zc_name, source, pair);
2924 return (err); /* special property already handled */
2927 * Only check this in the non-received case. We want to allow
2928 * 'inherit -S' to revert non-inheritable properties like quota
2929 * and reservation to the received or default values even though
2930 * they are not considered inheritable.
2932 if (prop != ZPROP_INVAL && !zfs_prop_inheritable(prop))
2933 return (SET_ERROR(EINVAL));
2936 /* property name has been validated by zfs_secpolicy_inherit_prop() */
2937 return (dsl_prop_inherit(zc->zc_name, zc->zc_value, source));
2941 zfs_ioc_pool_set_props(zfs_cmd_t *zc)
2948 if (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2949 zc->zc_iflags, &props))
2953 * If the only property is the configfile, then just do a spa_lookup()
2954 * to handle the faulted case.
2956 pair = nvlist_next_nvpair(props, NULL);
2957 if (pair != NULL && strcmp(nvpair_name(pair),
2958 zpool_prop_to_name(ZPOOL_PROP_CACHEFILE)) == 0 &&
2959 nvlist_next_nvpair(props, pair) == NULL) {
2960 mutex_enter(&spa_namespace_lock);
2961 if ((spa = spa_lookup(zc->zc_name)) != NULL) {
2962 spa_configfile_set(spa, props, B_FALSE);
2963 spa_config_sync(spa, B_FALSE, B_TRUE);
2965 mutex_exit(&spa_namespace_lock);
2972 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) {
2977 error = spa_prop_set(spa, props);
2980 spa_close(spa, FTAG);
2986 zfs_ioc_pool_get_props(zfs_cmd_t *zc)
2990 nvlist_t *nvp = NULL;
2992 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) {
2994 * If the pool is faulted, there may be properties we can still
2995 * get (such as altroot and cachefile), so attempt to get them
2998 mutex_enter(&spa_namespace_lock);
2999 if ((spa = spa_lookup(zc->zc_name)) != NULL)
3000 error = spa_prop_get(spa, &nvp);
3001 mutex_exit(&spa_namespace_lock);
3003 error = spa_prop_get(spa, &nvp);
3004 spa_close(spa, FTAG);
3007 if (error == 0 && zc->zc_nvlist_dst != 0)
3008 error = put_nvlist(zc, nvp);
3010 error = SET_ERROR(EFAULT);
3018 * zc_name name of filesystem
3019 * zc_nvlist_src{_size} nvlist of delegated permissions
3020 * zc_perm_action allow/unallow flag
3025 zfs_ioc_set_fsacl(zfs_cmd_t *zc)
3028 nvlist_t *fsaclnv = NULL;
3030 if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
3031 zc->zc_iflags, &fsaclnv)) != 0)
3035 * Verify nvlist is constructed correctly
3037 if ((error = zfs_deleg_verify_nvlist(fsaclnv)) != 0) {
3038 nvlist_free(fsaclnv);
3039 return (SET_ERROR(EINVAL));
3043 * If we don't have PRIV_SYS_MOUNT, then validate
3044 * that user is allowed to hand out each permission in
3048 error = secpolicy_zfs(CRED());
3050 if (zc->zc_perm_action == B_FALSE) {
3051 error = dsl_deleg_can_allow(zc->zc_name,
3054 error = dsl_deleg_can_unallow(zc->zc_name,
3060 error = dsl_deleg_set(zc->zc_name, fsaclnv, zc->zc_perm_action);
3062 nvlist_free(fsaclnv);
3068 * zc_name name of filesystem
3071 * zc_nvlist_src{_size} nvlist of delegated permissions
3074 zfs_ioc_get_fsacl(zfs_cmd_t *zc)
3079 if ((error = dsl_deleg_get(zc->zc_name, &nvp)) == 0) {
3080 error = put_nvlist(zc, nvp);
3089 zfs_create_cb(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx)
3091 zfs_creat_t *zct = arg;
3093 zfs_create_fs(os, cr, zct->zct_zplprops, tx);
3096 #define ZFS_PROP_UNDEFINED ((uint64_t)-1)
3100 * os parent objset pointer (NULL if root fs)
3101 * fuids_ok fuids allowed in this version of the spa?
3102 * sa_ok SAs allowed in this version of the spa?
3103 * createprops list of properties requested by creator
3106 * zplprops values for the zplprops we attach to the master node object
3107 * is_ci true if requested file system will be purely case-insensitive
3109 * Determine the settings for utf8only, normalization and
3110 * casesensitivity. Specific values may have been requested by the
3111 * creator and/or we can inherit values from the parent dataset. If
3112 * the file system is of too early a vintage, a creator can not
3113 * request settings for these properties, even if the requested
3114 * setting is the default value. We don't actually want to create dsl
3115 * properties for these, so remove them from the source nvlist after
3119 zfs_fill_zplprops_impl(objset_t *os, uint64_t zplver,
3120 boolean_t fuids_ok, boolean_t sa_ok, nvlist_t *createprops,
3121 nvlist_t *zplprops, boolean_t *is_ci)
3123 uint64_t sense = ZFS_PROP_UNDEFINED;
3124 uint64_t norm = ZFS_PROP_UNDEFINED;
3125 uint64_t u8 = ZFS_PROP_UNDEFINED;
3127 ASSERT(zplprops != NULL);
3129 if (os != NULL && os->os_phys->os_type != DMU_OST_ZFS)
3130 return (SET_ERROR(EINVAL));
3133 * Pull out creator prop choices, if any.
3136 (void) nvlist_lookup_uint64(createprops,
3137 zfs_prop_to_name(ZFS_PROP_VERSION), &zplver);
3138 (void) nvlist_lookup_uint64(createprops,
3139 zfs_prop_to_name(ZFS_PROP_NORMALIZE), &norm);
3140 (void) nvlist_remove_all(createprops,
3141 zfs_prop_to_name(ZFS_PROP_NORMALIZE));
3142 (void) nvlist_lookup_uint64(createprops,
3143 zfs_prop_to_name(ZFS_PROP_UTF8ONLY), &u8);
3144 (void) nvlist_remove_all(createprops,
3145 zfs_prop_to_name(ZFS_PROP_UTF8ONLY));
3146 (void) nvlist_lookup_uint64(createprops,
3147 zfs_prop_to_name(ZFS_PROP_CASE), &sense);
3148 (void) nvlist_remove_all(createprops,
3149 zfs_prop_to_name(ZFS_PROP_CASE));
3153 * If the zpl version requested is whacky or the file system
3154 * or pool is version is too "young" to support normalization
3155 * and the creator tried to set a value for one of the props,
3158 if ((zplver < ZPL_VERSION_INITIAL || zplver > ZPL_VERSION) ||
3159 (zplver >= ZPL_VERSION_FUID && !fuids_ok) ||
3160 (zplver >= ZPL_VERSION_SA && !sa_ok) ||
3161 (zplver < ZPL_VERSION_NORMALIZATION &&
3162 (norm != ZFS_PROP_UNDEFINED || u8 != ZFS_PROP_UNDEFINED ||
3163 sense != ZFS_PROP_UNDEFINED)))
3164 return (SET_ERROR(ENOTSUP));
3167 * Put the version in the zplprops
3169 VERIFY(nvlist_add_uint64(zplprops,
3170 zfs_prop_to_name(ZFS_PROP_VERSION), zplver) == 0);
3172 if (norm == ZFS_PROP_UNDEFINED)
3173 VERIFY(zfs_get_zplprop(os, ZFS_PROP_NORMALIZE, &norm) == 0);
3174 VERIFY(nvlist_add_uint64(zplprops,
3175 zfs_prop_to_name(ZFS_PROP_NORMALIZE), norm) == 0);
3178 * If we're normalizing, names must always be valid UTF-8 strings.
3182 if (u8 == ZFS_PROP_UNDEFINED)
3183 VERIFY(zfs_get_zplprop(os, ZFS_PROP_UTF8ONLY, &u8) == 0);
3184 VERIFY(nvlist_add_uint64(zplprops,
3185 zfs_prop_to_name(ZFS_PROP_UTF8ONLY), u8) == 0);
3187 if (sense == ZFS_PROP_UNDEFINED)
3188 VERIFY(zfs_get_zplprop(os, ZFS_PROP_CASE, &sense) == 0);
3189 VERIFY(nvlist_add_uint64(zplprops,
3190 zfs_prop_to_name(ZFS_PROP_CASE), sense) == 0);
3193 *is_ci = (sense == ZFS_CASE_INSENSITIVE);
3199 zfs_fill_zplprops(const char *dataset, nvlist_t *createprops,
3200 nvlist_t *zplprops, boolean_t *is_ci)
3202 boolean_t fuids_ok, sa_ok;
3203 uint64_t zplver = ZPL_VERSION;
3204 objset_t *os = NULL;
3205 char parentname[ZFS_MAX_DATASET_NAME_LEN];
3211 (void) strlcpy(parentname, dataset, sizeof (parentname));
3212 cp = strrchr(parentname, '/');
3216 if ((error = spa_open(dataset, &spa, FTAG)) != 0)
3219 spa_vers = spa_version(spa);
3220 spa_close(spa, FTAG);
3222 zplver = zfs_zpl_version_map(spa_vers);
3223 fuids_ok = (zplver >= ZPL_VERSION_FUID);
3224 sa_ok = (zplver >= ZPL_VERSION_SA);
3227 * Open parent object set so we can inherit zplprop values.
3229 if ((error = dmu_objset_hold(parentname, FTAG, &os)) != 0)
3232 error = zfs_fill_zplprops_impl(os, zplver, fuids_ok, sa_ok, createprops,
3234 dmu_objset_rele(os, FTAG);
3239 zfs_fill_zplprops_root(uint64_t spa_vers, nvlist_t *createprops,
3240 nvlist_t *zplprops, boolean_t *is_ci)
3244 uint64_t zplver = ZPL_VERSION;
3247 zplver = zfs_zpl_version_map(spa_vers);
3248 fuids_ok = (zplver >= ZPL_VERSION_FUID);
3249 sa_ok = (zplver >= ZPL_VERSION_SA);
3251 error = zfs_fill_zplprops_impl(NULL, zplver, fuids_ok, sa_ok,
3252 createprops, zplprops, is_ci);
3258 * "type" -> dmu_objset_type_t (int32)
3259 * (optional) "props" -> { prop -> value }
3262 * outnvl: propname -> error code (int32)
3265 zfs_ioc_create(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3268 zfs_creat_t zct = { 0 };
3269 nvlist_t *nvprops = NULL;
3270 void (*cbfunc)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx);
3272 dmu_objset_type_t type;
3273 boolean_t is_insensitive = B_FALSE;
3275 if (nvlist_lookup_int32(innvl, "type", &type32) != 0)
3276 return (SET_ERROR(EINVAL));
3278 (void) nvlist_lookup_nvlist(innvl, "props", &nvprops);
3282 cbfunc = zfs_create_cb;
3286 cbfunc = zvol_create_cb;
3293 if (strchr(fsname, '@') ||
3294 strchr(fsname, '%'))
3295 return (SET_ERROR(EINVAL));
3297 zct.zct_props = nvprops;
3300 return (SET_ERROR(EINVAL));
3302 if (type == DMU_OST_ZVOL) {
3303 uint64_t volsize, volblocksize;
3305 if (nvprops == NULL)
3306 return (SET_ERROR(EINVAL));
3307 if (nvlist_lookup_uint64(nvprops,
3308 zfs_prop_to_name(ZFS_PROP_VOLSIZE), &volsize) != 0)
3309 return (SET_ERROR(EINVAL));
3311 if ((error = nvlist_lookup_uint64(nvprops,
3312 zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE),
3313 &volblocksize)) != 0 && error != ENOENT)
3314 return (SET_ERROR(EINVAL));
3317 volblocksize = zfs_prop_default_numeric(
3318 ZFS_PROP_VOLBLOCKSIZE);
3320 if ((error = zvol_check_volblocksize(
3321 volblocksize)) != 0 ||
3322 (error = zvol_check_volsize(volsize,
3323 volblocksize)) != 0)
3325 } else if (type == DMU_OST_ZFS) {
3329 * We have to have normalization and
3330 * case-folding flags correct when we do the
3331 * file system creation, so go figure them out
3334 VERIFY(nvlist_alloc(&zct.zct_zplprops,
3335 NV_UNIQUE_NAME, KM_SLEEP) == 0);
3336 error = zfs_fill_zplprops(fsname, nvprops,
3337 zct.zct_zplprops, &is_insensitive);
3339 nvlist_free(zct.zct_zplprops);
3344 error = dmu_objset_create(fsname, type,
3345 is_insensitive ? DS_FLAG_CI_DATASET : 0, cbfunc, &zct);
3346 nvlist_free(zct.zct_zplprops);
3349 * It would be nice to do this atomically.
3352 error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL,
3355 (void) dsl_destroy_head(fsname);
3358 if (error == 0 && type == DMU_OST_ZVOL)
3359 zvol_create_minors(fsname);
3366 * "origin" -> name of origin snapshot
3367 * (optional) "props" -> { prop -> value }
3370 * outnvl: propname -> error code (int32)
3373 zfs_ioc_clone(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3376 nvlist_t *nvprops = NULL;
3379 if (nvlist_lookup_string(innvl, "origin", &origin_name) != 0)
3380 return (SET_ERROR(EINVAL));
3381 (void) nvlist_lookup_nvlist(innvl, "props", &nvprops);
3383 if (strchr(fsname, '@') ||
3384 strchr(fsname, '%'))
3385 return (SET_ERROR(EINVAL));
3387 if (dataset_namecheck(origin_name, NULL, NULL) != 0)
3388 return (SET_ERROR(EINVAL));
3389 error = dmu_objset_clone(fsname, origin_name);
3394 * It would be nice to do this atomically.
3397 error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL,
3400 (void) dsl_destroy_head(fsname);
3404 zvol_create_minors(fsname);
3411 * "snaps" -> { snapshot1, snapshot2 }
3412 * (optional) "props" -> { prop -> value (string) }
3415 * outnvl: snapshot -> error code (int32)
3418 zfs_ioc_snapshot(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3421 nvlist_t *props = NULL;
3425 (void) nvlist_lookup_nvlist(innvl, "props", &props);
3426 if ((error = zfs_check_userprops(poolname, props)) != 0)
3429 if (!nvlist_empty(props) &&
3430 zfs_earlier_version(poolname, SPA_VERSION_SNAP_PROPS))
3431 return (SET_ERROR(ENOTSUP));
3433 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
3434 return (SET_ERROR(EINVAL));
3435 poollen = strlen(poolname);
3436 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
3437 pair = nvlist_next_nvpair(snaps, pair)) {
3438 const char *name = nvpair_name(pair);
3439 const char *cp = strchr(name, '@');
3442 * The snap name must contain an @, and the part after it must
3443 * contain only valid characters.
3446 zfs_component_namecheck(cp + 1, NULL, NULL) != 0)
3447 return (SET_ERROR(EINVAL));
3450 * The snap must be in the specified pool.
3452 if (strncmp(name, poolname, poollen) != 0 ||
3453 (name[poollen] != '/' && name[poollen] != '@'))
3454 return (SET_ERROR(EXDEV));
3456 /* This must be the only snap of this fs. */
3457 for (nvpair_t *pair2 = nvlist_next_nvpair(snaps, pair);
3458 pair2 != NULL; pair2 = nvlist_next_nvpair(snaps, pair2)) {
3459 if (strncmp(name, nvpair_name(pair2), cp - name + 1)
3461 return (SET_ERROR(EXDEV));
3466 error = dsl_dataset_snapshot(snaps, props, outnvl);
3471 * innvl: "message" -> string
3475 zfs_ioc_log_history(const char *unused, nvlist_t *innvl, nvlist_t *outnvl)
3483 * The poolname in the ioctl is not set, we get it from the TSD,
3484 * which was set at the end of the last successful ioctl that allows
3485 * logging. The secpolicy func already checked that it is set.
3486 * Only one log ioctl is allowed after each successful ioctl, so
3487 * we clear the TSD here.
3489 poolname = tsd_get(zfs_allow_log_key);
3490 (void) tsd_set(zfs_allow_log_key, NULL);
3491 error = spa_open(poolname, &spa, FTAG);
3496 if (nvlist_lookup_string(innvl, "message", &message) != 0) {
3497 spa_close(spa, FTAG);
3498 return (SET_ERROR(EINVAL));
3501 if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) {
3502 spa_close(spa, FTAG);
3503 return (SET_ERROR(ENOTSUP));
3506 error = spa_history_log(spa, message);
3507 spa_close(spa, FTAG);
3513 zfs_ioc_nextboot(const char *unused, nvlist_t *innvl, nvlist_t *outnvl)
3515 char name[MAXNAMELEN];
3523 if (nvlist_lookup_uint64(innvl,
3524 ZPOOL_CONFIG_POOL_GUID, &pool_guid) != 0)
3526 if (nvlist_lookup_uint64(innvl,
3527 ZPOOL_CONFIG_GUID, &vdev_guid) != 0)
3529 if (nvlist_lookup_string(innvl,
3530 "command", &command) != 0)
3533 mutex_enter(&spa_namespace_lock);
3534 spa = spa_by_guid(pool_guid, vdev_guid);
3536 strcpy(name, spa_name(spa));
3537 mutex_exit(&spa_namespace_lock);
3541 if ((error = spa_open(name, &spa, FTAG)) != 0)
3543 spa_vdev_state_enter(spa, SCL_ALL);
3544 vd = spa_lookup_by_guid(spa, vdev_guid, B_TRUE);
3546 (void) spa_vdev_state_exit(spa, NULL, ENXIO);
3547 spa_close(spa, FTAG);
3550 error = vdev_label_write_pad2(vd, command, strlen(command));
3551 (void) spa_vdev_state_exit(spa, NULL, 0);
3552 txg_wait_synced(spa->spa_dsl_pool, 0);
3553 spa_close(spa, FTAG);
3559 * The dp_config_rwlock must not be held when calling this, because the
3560 * unmount may need to write out data.
3562 * This function is best-effort. Callers must deal gracefully if it
3563 * remains mounted (or is remounted after this call).
3565 * Returns 0 if the argument is not a snapshot, or it is not currently a
3566 * filesystem, or we were able to unmount it. Returns error code otherwise.
3569 zfs_unmount_snap(const char *snapname)
3572 zfsvfs_t *zfsvfs = NULL;
3574 if (strchr(snapname, '@') == NULL)
3577 int err = getzfsvfs_ref(snapname, &zfsvfs);
3579 ASSERT3P(zfsvfs, ==, NULL);
3582 vfsp = zfsvfs->z_vfs;
3584 ASSERT(!dsl_pool_config_held(dmu_objset_pool(zfsvfs->z_os)));
3587 err = vn_vfswlock(vfsp->vfs_vnodecovered);
3594 * Always force the unmount for snapshots.
3597 (void) dounmount(vfsp, MS_FORCE, kcred);
3599 (void) dounmount(vfsp, MS_FORCE, curthread);
3605 zfs_unmount_snap_cb(const char *snapname, void *arg)
3607 zfs_unmount_snap(snapname);
3612 * When a clone is destroyed, its origin may also need to be destroyed,
3613 * in which case it must be unmounted. This routine will do that unmount
3617 zfs_destroy_unmount_origin(const char *fsname)
3623 error = dmu_objset_hold(fsname, FTAG, &os);
3626 ds = dmu_objset_ds(os);
3627 if (dsl_dir_is_clone(ds->ds_dir) && DS_IS_DEFER_DESTROY(ds->ds_prev)) {
3628 char originname[ZFS_MAX_DATASET_NAME_LEN];
3629 dsl_dataset_name(ds->ds_prev, originname);
3630 dmu_objset_rele(os, FTAG);
3631 zfs_unmount_snap(originname);
3633 dmu_objset_rele(os, FTAG);
3639 * "snaps" -> { snapshot1, snapshot2 }
3640 * (optional boolean) "defer"
3643 * outnvl: snapshot -> error code (int32)
3648 zfs_ioc_destroy_snaps(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3655 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
3656 return (SET_ERROR(EINVAL));
3657 defer = nvlist_exists(innvl, "defer");
3659 poollen = strlen(poolname);
3660 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
3661 pair = nvlist_next_nvpair(snaps, pair)) {
3662 const char *name = nvpair_name(pair);
3665 * The snap must be in the specified pool to prevent the
3666 * invalid removal of zvol minors below.
3668 if (strncmp(name, poolname, poollen) != 0 ||
3669 (name[poollen] != '/' && name[poollen] != '@'))
3670 return (SET_ERROR(EXDEV));
3672 zfs_unmount_snap(nvpair_name(pair));
3673 #if defined(__FreeBSD__)
3674 zvol_remove_minors(name);
3678 return (dsl_destroy_snapshots_nvl(snaps, defer, outnvl));
3682 * Create bookmarks. Bookmark names are of the form <fs>#<bmark>.
3683 * All bookmarks must be in the same pool.
3686 * bookmark1 -> snapshot1, bookmark2 -> snapshot2
3689 * outnvl: bookmark -> error code (int32)
3694 zfs_ioc_bookmark(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3696 for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL);
3697 pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) {
3701 * Verify the snapshot argument.
3703 if (nvpair_value_string(pair, &snap_name) != 0)
3704 return (SET_ERROR(EINVAL));
3707 /* Verify that the keys (bookmarks) are unique */
3708 for (nvpair_t *pair2 = nvlist_next_nvpair(innvl, pair);
3709 pair2 != NULL; pair2 = nvlist_next_nvpair(innvl, pair2)) {
3710 if (strcmp(nvpair_name(pair), nvpair_name(pair2)) == 0)
3711 return (SET_ERROR(EINVAL));
3715 return (dsl_bookmark_create(innvl, outnvl));
3720 * property 1, property 2, ...
3724 * bookmark name 1 -> { property 1, property 2, ... },
3725 * bookmark name 2 -> { property 1, property 2, ... }
3730 zfs_ioc_get_bookmarks(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3732 return (dsl_get_bookmarks(fsname, innvl, outnvl));
3737 * bookmark name 1, bookmark name 2
3740 * outnvl: bookmark -> error code (int32)
3744 zfs_ioc_destroy_bookmarks(const char *poolname, nvlist_t *innvl,
3749 poollen = strlen(poolname);
3750 for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL);
3751 pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) {
3752 const char *name = nvpair_name(pair);
3753 const char *cp = strchr(name, '#');
3756 * The bookmark name must contain an #, and the part after it
3757 * must contain only valid characters.
3760 zfs_component_namecheck(cp + 1, NULL, NULL) != 0)
3761 return (SET_ERROR(EINVAL));
3764 * The bookmark must be in the specified pool.
3766 if (strncmp(name, poolname, poollen) != 0 ||
3767 (name[poollen] != '/' && name[poollen] != '#'))
3768 return (SET_ERROR(EXDEV));
3771 error = dsl_bookmark_destroy(innvl, outnvl);
3776 zfs_ioc_channel_program(const char *poolname, nvlist_t *innvl,
3780 uint64_t instrlimit, memlimit;
3781 nvpair_t *nvarg = NULL;
3783 if (0 != nvlist_lookup_string(innvl, ZCP_ARG_PROGRAM, &program)) {
3786 if (0 != nvlist_lookup_uint64(innvl, ZCP_ARG_INSTRLIMIT, &instrlimit)) {
3787 instrlimit = ZCP_DEFAULT_INSTRLIMIT;
3789 if (0 != nvlist_lookup_uint64(innvl, ZCP_ARG_MEMLIMIT, &memlimit)) {
3790 memlimit = ZCP_DEFAULT_MEMLIMIT;
3792 if (0 != nvlist_lookup_nvpair(innvl, ZCP_ARG_ARGLIST, &nvarg)) {
3796 if (instrlimit == 0 || instrlimit > zfs_lua_max_instrlimit)
3798 if (memlimit == 0 || memlimit > zfs_lua_max_memlimit)
3801 return (zcp_eval(poolname, program, instrlimit, memlimit,
3807 * zc_name name of dataset to destroy
3808 * zc_objset_type type of objset
3809 * zc_defer_destroy mark for deferred destroy
3814 zfs_ioc_destroy(zfs_cmd_t *zc)
3818 if (zc->zc_objset_type == DMU_OST_ZFS)
3819 zfs_unmount_snap(zc->zc_name);
3821 if (strchr(zc->zc_name, '@'))
3822 err = dsl_destroy_snapshot(zc->zc_name, zc->zc_defer_destroy);
3824 err = dsl_destroy_head(zc->zc_name);
3825 if (zc->zc_objset_type == DMU_OST_ZVOL && err == 0)
3827 zvol_remove_minors(zc->zc_name);
3829 (void) zvol_remove_minor(zc->zc_name);
3835 * fsname is name of dataset to rollback (to most recent snapshot)
3837 * innvl may contain name of expected target snapshot
3839 * outnvl: "target" -> name of most recent snapshot
3844 zfs_ioc_rollback(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3847 char *target = NULL;
3850 (void) nvlist_lookup_string(innvl, "target", &target);
3851 if (target != NULL) {
3852 int fslen = strlen(fsname);
3854 if (strncmp(fsname, target, fslen) != 0)
3855 return (SET_ERROR(EINVAL));
3856 if (target[fslen] != '@')
3857 return (SET_ERROR(EINVAL));
3860 if (getzfsvfs(fsname, &zfsvfs) == 0) {
3863 ds = dmu_objset_ds(zfsvfs->z_os);
3864 error = zfs_suspend_fs(zfsvfs);
3868 error = dsl_dataset_rollback(fsname, target, zfsvfs,
3870 resume_err = zfs_resume_fs(zfsvfs, ds);
3871 error = error ? error : resume_err;
3874 VFS_RELE(zfsvfs->z_vfs);
3876 vfs_unbusy(zfsvfs->z_vfs);
3879 error = dsl_dataset_rollback(fsname, target, NULL, outnvl);
3885 recursive_unmount(const char *fsname, void *arg)
3887 const char *snapname = arg;
3888 char fullname[ZFS_MAX_DATASET_NAME_LEN];
3890 (void) snprintf(fullname, sizeof (fullname), "%s@%s", fsname, snapname);
3891 zfs_unmount_snap(fullname);
3898 * zc_name old name of dataset
3899 * zc_value new name of dataset
3900 * zc_cookie recursive flag (only valid for snapshots)
3905 zfs_ioc_rename(zfs_cmd_t *zc)
3907 boolean_t recursive = zc->zc_cookie & 1;
3909 boolean_t allow_mounted = B_TRUE;
3912 allow_mounted = (zc->zc_cookie & 2) != 0;
3915 zc->zc_value[sizeof (zc->zc_value) - 1] = '\0';
3916 if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
3917 strchr(zc->zc_value, '%'))
3918 return (SET_ERROR(EINVAL));
3920 at = strchr(zc->zc_name, '@');
3922 /* snaps must be in same fs */
3925 if (strncmp(zc->zc_name, zc->zc_value, at - zc->zc_name + 1))
3926 return (SET_ERROR(EXDEV));
3928 if (zc->zc_objset_type == DMU_OST_ZFS && !allow_mounted) {
3929 error = dmu_objset_find(zc->zc_name,
3930 recursive_unmount, at + 1,
3931 recursive ? DS_FIND_CHILDREN : 0);
3937 error = dsl_dataset_rename_snapshot(zc->zc_name,
3938 at + 1, strchr(zc->zc_value, '@') + 1, recursive);
3944 if (zc->zc_objset_type == DMU_OST_ZVOL)
3945 (void) zvol_remove_minor(zc->zc_name);
3947 return (dsl_dir_rename(zc->zc_name, zc->zc_value));
3952 zfs_check_settable(const char *dsname, nvpair_t *pair, cred_t *cr)
3954 const char *propname = nvpair_name(pair);
3955 boolean_t issnap = (strchr(dsname, '@') != NULL);
3956 zfs_prop_t prop = zfs_name_to_prop(propname);
3960 if (prop == ZPROP_INVAL) {
3961 if (zfs_prop_user(propname)) {
3962 if (err = zfs_secpolicy_write_perms(dsname,
3963 ZFS_DELEG_PERM_USERPROP, cr))
3968 if (!issnap && zfs_prop_userquota(propname)) {
3969 const char *perm = NULL;
3970 const char *uq_prefix =
3971 zfs_userquota_prop_prefixes[ZFS_PROP_USERQUOTA];
3972 const char *gq_prefix =
3973 zfs_userquota_prop_prefixes[ZFS_PROP_GROUPQUOTA];
3975 if (strncmp(propname, uq_prefix,
3976 strlen(uq_prefix)) == 0) {
3977 perm = ZFS_DELEG_PERM_USERQUOTA;
3978 } else if (strncmp(propname, gq_prefix,
3979 strlen(gq_prefix)) == 0) {
3980 perm = ZFS_DELEG_PERM_GROUPQUOTA;
3982 /* USERUSED and GROUPUSED are read-only */
3983 return (SET_ERROR(EINVAL));
3986 if (err = zfs_secpolicy_write_perms(dsname, perm, cr))
3991 return (SET_ERROR(EINVAL));
3995 return (SET_ERROR(EINVAL));
3997 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
3999 * dsl_prop_get_all_impl() returns properties in this
4003 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
4004 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
4009 * Check that this value is valid for this pool version
4012 case ZFS_PROP_COMPRESSION:
4014 * If the user specified gzip compression, make sure
4015 * the SPA supports it. We ignore any errors here since
4016 * we'll catch them later.
4018 if (nvpair_value_uint64(pair, &intval) == 0) {
4019 if (intval >= ZIO_COMPRESS_GZIP_1 &&
4020 intval <= ZIO_COMPRESS_GZIP_9 &&
4021 zfs_earlier_version(dsname,
4022 SPA_VERSION_GZIP_COMPRESSION)) {
4023 return (SET_ERROR(ENOTSUP));
4026 if (intval == ZIO_COMPRESS_ZLE &&
4027 zfs_earlier_version(dsname,
4028 SPA_VERSION_ZLE_COMPRESSION))
4029 return (SET_ERROR(ENOTSUP));
4031 if (intval == ZIO_COMPRESS_LZ4) {
4034 if ((err = spa_open(dsname, &spa, FTAG)) != 0)
4037 if (!spa_feature_is_enabled(spa,
4038 SPA_FEATURE_LZ4_COMPRESS)) {
4039 spa_close(spa, FTAG);
4040 return (SET_ERROR(ENOTSUP));
4042 spa_close(spa, FTAG);
4046 * If this is a bootable dataset then
4047 * verify that the compression algorithm
4048 * is supported for booting. We must return
4049 * something other than ENOTSUP since it
4050 * implies a downrev pool version.
4052 if (zfs_is_bootfs(dsname) &&
4053 !BOOTFS_COMPRESS_VALID(intval)) {
4054 return (SET_ERROR(ERANGE));
4059 case ZFS_PROP_COPIES:
4060 if (zfs_earlier_version(dsname, SPA_VERSION_DITTO_BLOCKS))
4061 return (SET_ERROR(ENOTSUP));
4064 case ZFS_PROP_RECORDSIZE:
4065 /* Record sizes above 128k need the feature to be enabled */
4066 if (nvpair_value_uint64(pair, &intval) == 0 &&
4067 intval > SPA_OLD_MAXBLOCKSIZE) {
4071 * We don't allow setting the property above 1MB,
4072 * unless the tunable has been changed.
4074 if (intval > zfs_max_recordsize ||
4075 intval > SPA_MAXBLOCKSIZE)
4076 return (SET_ERROR(ERANGE));
4078 if ((err = spa_open(dsname, &spa, FTAG)) != 0)
4081 if (!spa_feature_is_enabled(spa,
4082 SPA_FEATURE_LARGE_BLOCKS)) {
4083 spa_close(spa, FTAG);
4084 return (SET_ERROR(ENOTSUP));
4086 spa_close(spa, FTAG);
4090 case ZFS_PROP_SHARESMB:
4091 if (zpl_earlier_version(dsname, ZPL_VERSION_FUID))
4092 return (SET_ERROR(ENOTSUP));
4095 case ZFS_PROP_ACLINHERIT:
4096 if (nvpair_type(pair) == DATA_TYPE_UINT64 &&
4097 nvpair_value_uint64(pair, &intval) == 0) {
4098 if (intval == ZFS_ACL_PASSTHROUGH_X &&
4099 zfs_earlier_version(dsname,
4100 SPA_VERSION_PASSTHROUGH_X))
4101 return (SET_ERROR(ENOTSUP));
4105 case ZFS_PROP_CHECKSUM:
4106 case ZFS_PROP_DEDUP:
4108 spa_feature_t feature;
4111 /* dedup feature version checks */
4112 if (prop == ZFS_PROP_DEDUP &&
4113 zfs_earlier_version(dsname, SPA_VERSION_DEDUP))
4114 return (SET_ERROR(ENOTSUP));
4116 if (nvpair_value_uint64(pair, &intval) != 0)
4117 return (SET_ERROR(EINVAL));
4119 /* check prop value is enabled in features */
4120 feature = zio_checksum_to_feature(intval & ZIO_CHECKSUM_MASK);
4121 if (feature == SPA_FEATURE_NONE)
4124 if ((err = spa_open(dsname, &spa, FTAG)) != 0)
4127 * Salted checksums are not supported on root pools.
4129 if (spa_bootfs(spa) != 0 &&
4130 intval < ZIO_CHECKSUM_FUNCTIONS &&
4131 (zio_checksum_table[intval].ci_flags &
4132 ZCHECKSUM_FLAG_SALTED)) {
4133 spa_close(spa, FTAG);
4134 return (SET_ERROR(ERANGE));
4136 if (!spa_feature_is_enabled(spa, feature)) {
4137 spa_close(spa, FTAG);
4138 return (SET_ERROR(ENOTSUP));
4140 spa_close(spa, FTAG);
4145 return (zfs_secpolicy_setprop(dsname, prop, pair, CRED()));
4149 * Checks for a race condition to make sure we don't increment a feature flag
4153 zfs_prop_activate_feature_check(void *arg, dmu_tx_t *tx)
4155 spa_t *spa = dmu_tx_pool(tx)->dp_spa;
4156 spa_feature_t *featurep = arg;
4158 if (!spa_feature_is_active(spa, *featurep))
4161 return (SET_ERROR(EBUSY));
4165 * The callback invoked on feature activation in the sync task caused by
4166 * zfs_prop_activate_feature.
4169 zfs_prop_activate_feature_sync(void *arg, dmu_tx_t *tx)
4171 spa_t *spa = dmu_tx_pool(tx)->dp_spa;
4172 spa_feature_t *featurep = arg;
4174 spa_feature_incr(spa, *featurep, tx);
4178 * Activates a feature on a pool in response to a property setting. This
4179 * creates a new sync task which modifies the pool to reflect the feature
4183 zfs_prop_activate_feature(spa_t *spa, spa_feature_t feature)
4187 /* EBUSY here indicates that the feature is already active */
4188 err = dsl_sync_task(spa_name(spa),
4189 zfs_prop_activate_feature_check, zfs_prop_activate_feature_sync,
4190 &feature, 2, ZFS_SPACE_CHECK_RESERVED);
4192 if (err != 0 && err != EBUSY)
4199 * Removes properties from the given props list that fail permission checks
4200 * needed to clear them and to restore them in case of a receive error. For each
4201 * property, make sure we have both set and inherit permissions.
4203 * Returns the first error encountered if any permission checks fail. If the
4204 * caller provides a non-NULL errlist, it also gives the complete list of names
4205 * of all the properties that failed a permission check along with the
4206 * corresponding error numbers. The caller is responsible for freeing the
4209 * If every property checks out successfully, zero is returned and the list
4210 * pointed at by errlist is NULL.
4213 zfs_check_clearable(char *dataset, nvlist_t *props, nvlist_t **errlist)
4216 nvpair_t *pair, *next_pair;
4223 VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0);
4225 zc = kmem_alloc(sizeof (zfs_cmd_t), KM_SLEEP);
4226 (void) strcpy(zc->zc_name, dataset);
4227 pair = nvlist_next_nvpair(props, NULL);
4228 while (pair != NULL) {
4229 next_pair = nvlist_next_nvpair(props, pair);
4231 (void) strcpy(zc->zc_value, nvpair_name(pair));
4232 if ((err = zfs_check_settable(dataset, pair, CRED())) != 0 ||
4233 (err = zfs_secpolicy_inherit_prop(zc, NULL, CRED())) != 0) {
4234 VERIFY(nvlist_remove_nvpair(props, pair) == 0);
4235 VERIFY(nvlist_add_int32(errors,
4236 zc->zc_value, err) == 0);
4240 kmem_free(zc, sizeof (zfs_cmd_t));
4242 if ((pair = nvlist_next_nvpair(errors, NULL)) == NULL) {
4243 nvlist_free(errors);
4246 VERIFY(nvpair_value_int32(pair, &rv) == 0);
4249 if (errlist == NULL)
4250 nvlist_free(errors);
4258 propval_equals(nvpair_t *p1, nvpair_t *p2)
4260 if (nvpair_type(p1) == DATA_TYPE_NVLIST) {
4261 /* dsl_prop_get_all_impl() format */
4263 VERIFY(nvpair_value_nvlist(p1, &attrs) == 0);
4264 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
4268 if (nvpair_type(p2) == DATA_TYPE_NVLIST) {
4270 VERIFY(nvpair_value_nvlist(p2, &attrs) == 0);
4271 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
4275 if (nvpair_type(p1) != nvpair_type(p2))
4278 if (nvpair_type(p1) == DATA_TYPE_STRING) {
4279 char *valstr1, *valstr2;
4281 VERIFY(nvpair_value_string(p1, (char **)&valstr1) == 0);
4282 VERIFY(nvpair_value_string(p2, (char **)&valstr2) == 0);
4283 return (strcmp(valstr1, valstr2) == 0);
4285 uint64_t intval1, intval2;
4287 VERIFY(nvpair_value_uint64(p1, &intval1) == 0);
4288 VERIFY(nvpair_value_uint64(p2, &intval2) == 0);
4289 return (intval1 == intval2);
4294 * Remove properties from props if they are not going to change (as determined
4295 * by comparison with origprops). Remove them from origprops as well, since we
4296 * do not need to clear or restore properties that won't change.
4299 props_reduce(nvlist_t *props, nvlist_t *origprops)
4301 nvpair_t *pair, *next_pair;
4303 if (origprops == NULL)
4304 return; /* all props need to be received */
4306 pair = nvlist_next_nvpair(props, NULL);
4307 while (pair != NULL) {
4308 const char *propname = nvpair_name(pair);
4311 next_pair = nvlist_next_nvpair(props, pair);
4313 if ((nvlist_lookup_nvpair(origprops, propname,
4314 &match) != 0) || !propval_equals(pair, match))
4315 goto next; /* need to set received value */
4317 /* don't clear the existing received value */
4318 (void) nvlist_remove_nvpair(origprops, match);
4319 /* don't bother receiving the property */
4320 (void) nvlist_remove_nvpair(props, pair);
4327 * Extract properties that cannot be set PRIOR to the receipt of a dataset.
4328 * For example, refquota cannot be set until after the receipt of a dataset,
4329 * because in replication streams, an older/earlier snapshot may exceed the
4330 * refquota. We want to receive the older/earlier snapshot, but setting
4331 * refquota pre-receipt will set the dsl's ACTUAL quota, which will prevent
4332 * the older/earlier snapshot from being received (with EDQUOT).
4334 * The ZFS test "zfs_receive_011_pos" demonstrates such a scenario.
4336 * libzfs will need to be judicious handling errors encountered by props
4337 * extracted by this function.
4340 extract_delay_props(nvlist_t *props)
4342 nvlist_t *delayprops;
4343 nvpair_t *nvp, *tmp;
4344 static const zfs_prop_t delayable[] = { ZFS_PROP_REFQUOTA, 0 };
4347 VERIFY(nvlist_alloc(&delayprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
4349 for (nvp = nvlist_next_nvpair(props, NULL); nvp != NULL;
4350 nvp = nvlist_next_nvpair(props, nvp)) {
4352 * strcmp() is safe because zfs_prop_to_name() always returns
4355 for (i = 0; delayable[i] != 0; i++) {
4356 if (strcmp(zfs_prop_to_name(delayable[i]),
4357 nvpair_name(nvp)) == 0) {
4361 if (delayable[i] != 0) {
4362 tmp = nvlist_prev_nvpair(props, nvp);
4363 VERIFY(nvlist_add_nvpair(delayprops, nvp) == 0);
4364 VERIFY(nvlist_remove_nvpair(props, nvp) == 0);
4369 if (nvlist_empty(delayprops)) {
4370 nvlist_free(delayprops);
4373 return (delayprops);
4377 static boolean_t zfs_ioc_recv_inject_err;
4382 * zc_name name of containing filesystem
4383 * zc_nvlist_src{_size} nvlist of properties to apply
4384 * zc_value name of snapshot to create
4385 * zc_string name of clone origin (if DRR_FLAG_CLONE)
4386 * zc_cookie file descriptor to recv from
4387 * zc_begin_record the BEGIN record of the stream (not byteswapped)
4388 * zc_guid force flag
4389 * zc_cleanup_fd cleanup-on-exit file descriptor
4390 * zc_action_handle handle for this guid/ds mapping (or zero on first call)
4391 * zc_resumable if data is incomplete assume sender will resume
4394 * zc_cookie number of bytes read
4395 * zc_nvlist_dst{_size} error for each unapplied received property
4396 * zc_obj zprop_errflags_t
4397 * zc_action_handle handle for this guid/ds mapping
4400 zfs_ioc_recv(zfs_cmd_t *zc)
4403 dmu_recv_cookie_t drc;
4404 boolean_t force = (boolean_t)zc->zc_guid;
4407 int props_error = 0;
4410 nvlist_t *props = NULL; /* sent properties */
4411 nvlist_t *origprops = NULL; /* existing properties */
4412 nvlist_t *delayprops = NULL; /* sent properties applied post-receive */
4413 char *origin = NULL;
4415 char tofs[ZFS_MAX_DATASET_NAME_LEN];
4416 cap_rights_t rights;
4417 boolean_t first_recvd_props = B_FALSE;
4419 if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
4420 strchr(zc->zc_value, '@') == NULL ||
4421 strchr(zc->zc_value, '%'))
4422 return (SET_ERROR(EINVAL));
4424 (void) strcpy(tofs, zc->zc_value);
4425 tosnap = strchr(tofs, '@');
4428 if (zc->zc_nvlist_src != 0 &&
4429 (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
4430 zc->zc_iflags, &props)) != 0)
4437 fget_read(curthread, fd, cap_rights_init(&rights, CAP_PREAD), &fp);
4441 return (SET_ERROR(EBADF));
4444 errors = fnvlist_alloc();
4446 if (zc->zc_string[0])
4447 origin = zc->zc_string;
4449 error = dmu_recv_begin(tofs, tosnap,
4450 &zc->zc_begin_record, force, zc->zc_resumable, origin, &drc);
4455 * Set properties before we receive the stream so that they are applied
4456 * to the new data. Note that we must call dmu_recv_stream() if
4457 * dmu_recv_begin() succeeds.
4459 if (props != NULL && !drc.drc_newfs) {
4460 if (spa_version(dsl_dataset_get_spa(drc.drc_ds)) >=
4461 SPA_VERSION_RECVD_PROPS &&
4462 !dsl_prop_get_hasrecvd(tofs))
4463 first_recvd_props = B_TRUE;
4466 * If new received properties are supplied, they are to
4467 * completely replace the existing received properties, so stash
4468 * away the existing ones.
4470 if (dsl_prop_get_received(tofs, &origprops) == 0) {
4471 nvlist_t *errlist = NULL;
4473 * Don't bother writing a property if its value won't
4474 * change (and avoid the unnecessary security checks).
4476 * The first receive after SPA_VERSION_RECVD_PROPS is a
4477 * special case where we blow away all local properties
4480 if (!first_recvd_props)
4481 props_reduce(props, origprops);
4482 if (zfs_check_clearable(tofs, origprops, &errlist) != 0)
4483 (void) nvlist_merge(errors, errlist, 0);
4484 nvlist_free(errlist);
4486 if (clear_received_props(tofs, origprops,
4487 first_recvd_props ? NULL : props) != 0)
4488 zc->zc_obj |= ZPROP_ERR_NOCLEAR;
4490 zc->zc_obj |= ZPROP_ERR_NOCLEAR;
4494 if (props != NULL) {
4495 props_error = dsl_prop_set_hasrecvd(tofs);
4497 if (props_error == 0) {
4498 delayprops = extract_delay_props(props);
4499 (void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_RECEIVED,
4505 error = dmu_recv_stream(&drc, fp, &off, zc->zc_cleanup_fd,
4506 &zc->zc_action_handle);
4509 zfsvfs_t *zfsvfs = NULL;
4511 if (getzfsvfs(tofs, &zfsvfs) == 0) {
4516 ds = dmu_objset_ds(zfsvfs->z_os);
4517 error = zfs_suspend_fs(zfsvfs);
4519 * If the suspend fails, then the recv_end will
4520 * likely also fail, and clean up after itself.
4522 end_err = dmu_recv_end(&drc, zfsvfs);
4524 error = zfs_resume_fs(zfsvfs, ds);
4525 error = error ? error : end_err;
4527 VFS_RELE(zfsvfs->z_vfs);
4529 vfs_unbusy(zfsvfs->z_vfs);
4532 error = dmu_recv_end(&drc, NULL);
4535 /* Set delayed properties now, after we're done receiving. */
4536 if (delayprops != NULL && error == 0) {
4537 (void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_RECEIVED,
4538 delayprops, errors);
4542 if (delayprops != NULL) {
4544 * Merge delayed props back in with initial props, in case
4545 * we're DEBUG and zfs_ioc_recv_inject_err is set (which means
4546 * we have to make sure clear_received_props() includes
4547 * the delayed properties).
4549 * Since zfs_ioc_recv_inject_err is only in DEBUG kernels,
4550 * using ASSERT() will be just like a VERIFY.
4552 ASSERT(nvlist_merge(props, delayprops, 0) == 0);
4553 nvlist_free(delayprops);
4557 * Now that all props, initial and delayed, are set, report the prop
4558 * errors to the caller.
4560 if (zc->zc_nvlist_dst_size != 0 &&
4561 (nvlist_smush(errors, zc->zc_nvlist_dst_size) != 0 ||
4562 put_nvlist(zc, errors) != 0)) {
4564 * Caller made zc->zc_nvlist_dst less than the minimum expected
4565 * size or supplied an invalid address.
4567 props_error = SET_ERROR(EINVAL);
4570 zc->zc_cookie = off - fp->f_offset;
4571 if (off >= 0 && off <= MAXOFFSET_T)
4575 if (zfs_ioc_recv_inject_err) {
4576 zfs_ioc_recv_inject_err = B_FALSE;
4583 zvol_create_minors(tofs);
4587 * On error, restore the original props.
4589 if (error != 0 && props != NULL && !drc.drc_newfs) {
4590 if (clear_received_props(tofs, props, NULL) != 0) {
4592 * We failed to clear the received properties.
4593 * Since we may have left a $recvd value on the
4594 * system, we can't clear the $hasrecvd flag.
4596 zc->zc_obj |= ZPROP_ERR_NORESTORE;
4597 } else if (first_recvd_props) {
4598 dsl_prop_unset_hasrecvd(tofs);
4601 if (origprops == NULL && !drc.drc_newfs) {
4602 /* We failed to stash the original properties. */
4603 zc->zc_obj |= ZPROP_ERR_NORESTORE;
4607 * dsl_props_set() will not convert RECEIVED to LOCAL on or
4608 * after SPA_VERSION_RECVD_PROPS, so we need to specify LOCAL
4609 * explictly if we're restoring local properties cleared in the
4610 * first new-style receive.
4612 if (origprops != NULL &&
4613 zfs_set_prop_nvlist(tofs, (first_recvd_props ?
4614 ZPROP_SRC_LOCAL : ZPROP_SRC_RECEIVED),
4615 origprops, NULL) != 0) {
4617 * We stashed the original properties but failed to
4620 zc->zc_obj |= ZPROP_ERR_NORESTORE;
4625 nvlist_free(origprops);
4626 nvlist_free(errors);
4630 error = props_error;
4637 * zc_name name of snapshot to send
4638 * zc_cookie file descriptor to send stream to
4639 * zc_obj fromorigin flag (mutually exclusive with zc_fromobj)
4640 * zc_sendobj objsetid of snapshot to send
4641 * zc_fromobj objsetid of incremental fromsnap (may be zero)
4642 * zc_guid if set, estimate size of stream only. zc_cookie is ignored.
4643 * output size in zc_objset_type.
4644 * zc_flags lzc_send_flags
4647 * zc_objset_type estimated size, if zc_guid is set
4650 zfs_ioc_send(zfs_cmd_t *zc)
4654 boolean_t estimate = (zc->zc_guid != 0);
4655 boolean_t embedok = (zc->zc_flags & 0x1);
4656 boolean_t large_block_ok = (zc->zc_flags & 0x2);
4657 boolean_t compressok = (zc->zc_flags & 0x4);
4659 if (zc->zc_obj != 0) {
4661 dsl_dataset_t *tosnap;
4663 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
4667 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &tosnap);
4669 dsl_pool_rele(dp, FTAG);
4673 if (dsl_dir_is_clone(tosnap->ds_dir))
4675 dsl_dir_phys(tosnap->ds_dir)->dd_origin_obj;
4676 dsl_dataset_rele(tosnap, FTAG);
4677 dsl_pool_rele(dp, FTAG);
4682 dsl_dataset_t *tosnap;
4683 dsl_dataset_t *fromsnap = NULL;
4685 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
4689 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &tosnap);
4691 dsl_pool_rele(dp, FTAG);
4695 if (zc->zc_fromobj != 0) {
4696 error = dsl_dataset_hold_obj(dp, zc->zc_fromobj,
4699 dsl_dataset_rele(tosnap, FTAG);
4700 dsl_pool_rele(dp, FTAG);
4705 error = dmu_send_estimate(tosnap, fromsnap, compressok,
4706 &zc->zc_objset_type);
4708 if (fromsnap != NULL)
4709 dsl_dataset_rele(fromsnap, FTAG);
4710 dsl_dataset_rele(tosnap, FTAG);
4711 dsl_pool_rele(dp, FTAG);
4714 cap_rights_t rights;
4717 fp = getf(zc->zc_cookie);
4719 fget_write(curthread, zc->zc_cookie,
4720 cap_rights_init(&rights, CAP_WRITE), &fp);
4723 return (SET_ERROR(EBADF));
4726 error = dmu_send_obj(zc->zc_name, zc->zc_sendobj,
4727 zc->zc_fromobj, embedok, large_block_ok, compressok,
4729 zc->zc_cookie, fp->f_vnode, &off);
4731 zc->zc_cookie, fp, &off);
4734 if (off >= 0 && off <= MAXOFFSET_T)
4736 releasef(zc->zc_cookie);
4743 * zc_name name of snapshot on which to report progress
4744 * zc_cookie file descriptor of send stream
4747 * zc_cookie number of bytes written in send stream thus far
4750 zfs_ioc_send_progress(zfs_cmd_t *zc)
4754 dmu_sendarg_t *dsp = NULL;
4757 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
4761 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &ds);
4763 dsl_pool_rele(dp, FTAG);
4767 mutex_enter(&ds->ds_sendstream_lock);
4770 * Iterate over all the send streams currently active on this dataset.
4771 * If there's one which matches the specified file descriptor _and_ the
4772 * stream was started by the current process, return the progress of
4775 for (dsp = list_head(&ds->ds_sendstreams); dsp != NULL;
4776 dsp = list_next(&ds->ds_sendstreams, dsp)) {
4777 if (dsp->dsa_outfd == zc->zc_cookie &&
4778 dsp->dsa_proc == curproc)
4783 zc->zc_cookie = *(dsp->dsa_off);
4785 error = SET_ERROR(ENOENT);
4787 mutex_exit(&ds->ds_sendstream_lock);
4788 dsl_dataset_rele(ds, FTAG);
4789 dsl_pool_rele(dp, FTAG);
4794 zfs_ioc_inject_fault(zfs_cmd_t *zc)
4798 error = zio_inject_fault(zc->zc_name, (int)zc->zc_guid, &id,
4799 &zc->zc_inject_record);
4802 zc->zc_guid = (uint64_t)id;
4808 zfs_ioc_clear_fault(zfs_cmd_t *zc)
4810 return (zio_clear_fault((int)zc->zc_guid));
4814 zfs_ioc_inject_list_next(zfs_cmd_t *zc)
4816 int id = (int)zc->zc_guid;
4819 error = zio_inject_list_next(&id, zc->zc_name, sizeof (zc->zc_name),
4820 &zc->zc_inject_record);
4828 zfs_ioc_error_log(zfs_cmd_t *zc)
4832 size_t count = (size_t)zc->zc_nvlist_dst_size;
4834 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
4837 error = spa_get_errlog(spa, (void *)(uintptr_t)zc->zc_nvlist_dst,
4840 zc->zc_nvlist_dst_size = count;
4842 zc->zc_nvlist_dst_size = spa_get_errlog_size(spa);
4844 spa_close(spa, FTAG);
4850 zfs_ioc_clear(zfs_cmd_t *zc)
4857 * On zpool clear we also fix up missing slogs
4859 mutex_enter(&spa_namespace_lock);
4860 spa = spa_lookup(zc->zc_name);
4862 mutex_exit(&spa_namespace_lock);
4863 return (SET_ERROR(EIO));
4865 if (spa_get_log_state(spa) == SPA_LOG_MISSING) {
4866 /* we need to let spa_open/spa_load clear the chains */
4867 spa_set_log_state(spa, SPA_LOG_CLEAR);
4869 spa->spa_last_open_failed = 0;
4870 mutex_exit(&spa_namespace_lock);
4872 if (zc->zc_cookie & ZPOOL_NO_REWIND) {
4873 error = spa_open(zc->zc_name, &spa, FTAG);
4876 nvlist_t *config = NULL;
4878 if (zc->zc_nvlist_src == 0)
4879 return (SET_ERROR(EINVAL));
4881 if ((error = get_nvlist(zc->zc_nvlist_src,
4882 zc->zc_nvlist_src_size, zc->zc_iflags, &policy)) == 0) {
4883 error = spa_open_rewind(zc->zc_name, &spa, FTAG,
4885 if (config != NULL) {
4888 if ((err = put_nvlist(zc, config)) != 0)
4890 nvlist_free(config);
4892 nvlist_free(policy);
4899 spa_vdev_state_enter(spa, SCL_NONE);
4901 if (zc->zc_guid == 0) {
4904 vd = spa_lookup_by_guid(spa, zc->zc_guid, B_TRUE);
4906 (void) spa_vdev_state_exit(spa, NULL, ENODEV);
4907 spa_close(spa, FTAG);
4908 return (SET_ERROR(ENODEV));
4912 vdev_clear(spa, vd);
4914 (void) spa_vdev_state_exit(spa, NULL, 0);
4917 * Resume any suspended I/Os.
4919 if (zio_resume(spa) != 0)
4920 error = SET_ERROR(EIO);
4922 spa_close(spa, FTAG);
4928 zfs_ioc_pool_reopen(zfs_cmd_t *zc)
4933 error = spa_open(zc->zc_name, &spa, FTAG);
4937 spa_vdev_state_enter(spa, SCL_NONE);
4940 * If a resilver is already in progress then set the
4941 * spa_scrub_reopen flag to B_TRUE so that we don't restart
4942 * the scan as a side effect of the reopen. Otherwise, let
4943 * vdev_open() decided if a resilver is required.
4945 spa->spa_scrub_reopen = dsl_scan_resilvering(spa->spa_dsl_pool);
4946 vdev_reopen(spa->spa_root_vdev);
4947 spa->spa_scrub_reopen = B_FALSE;
4949 (void) spa_vdev_state_exit(spa, NULL, 0);
4950 spa_close(spa, FTAG);
4955 * zc_name name of filesystem
4958 * zc_string name of conflicting snapshot, if there is one
4961 zfs_ioc_promote(zfs_cmd_t *zc)
4964 dsl_dataset_t *ds, *ods;
4965 char origin[ZFS_MAX_DATASET_NAME_LEN];
4969 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
4973 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &ds);
4975 dsl_pool_rele(dp, FTAG);
4979 if (!dsl_dir_is_clone(ds->ds_dir)) {
4980 dsl_dataset_rele(ds, FTAG);
4981 dsl_pool_rele(dp, FTAG);
4982 return (SET_ERROR(EINVAL));
4985 error = dsl_dataset_hold_obj(dp,
4986 dsl_dir_phys(ds->ds_dir)->dd_origin_obj, FTAG, &ods);
4988 dsl_dataset_rele(ds, FTAG);
4989 dsl_pool_rele(dp, FTAG);
4993 dsl_dataset_name(ods, origin);
4994 dsl_dataset_rele(ods, FTAG);
4995 dsl_dataset_rele(ds, FTAG);
4996 dsl_pool_rele(dp, FTAG);
4999 * We don't need to unmount *all* the origin fs's snapshots, but
5002 cp = strchr(origin, '@');
5005 (void) dmu_objset_find(origin,
5006 zfs_unmount_snap_cb, NULL, DS_FIND_SNAPSHOTS);
5007 return (dsl_dataset_promote(zc->zc_name, zc->zc_string));
5011 * Retrieve a single {user|group}{used|quota}@... property.
5014 * zc_name name of filesystem
5015 * zc_objset_type zfs_userquota_prop_t
5016 * zc_value domain name (eg. "S-1-234-567-89")
5017 * zc_guid RID/UID/GID
5020 * zc_cookie property value
5023 zfs_ioc_userspace_one(zfs_cmd_t *zc)
5028 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
5029 return (SET_ERROR(EINVAL));
5031 error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE);
5035 error = zfs_userspace_one(zfsvfs,
5036 zc->zc_objset_type, zc->zc_value, zc->zc_guid, &zc->zc_cookie);
5037 zfsvfs_rele(zfsvfs, FTAG);
5044 * zc_name name of filesystem
5045 * zc_cookie zap cursor
5046 * zc_objset_type zfs_userquota_prop_t
5047 * zc_nvlist_dst[_size] buffer to fill (not really an nvlist)
5050 * zc_nvlist_dst[_size] data buffer (array of zfs_useracct_t)
5051 * zc_cookie zap cursor
5054 zfs_ioc_userspace_many(zfs_cmd_t *zc)
5057 int bufsize = zc->zc_nvlist_dst_size;
5060 return (SET_ERROR(ENOMEM));
5062 int error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE);
5066 void *buf = kmem_alloc(bufsize, KM_SLEEP);
5068 error = zfs_userspace_many(zfsvfs, zc->zc_objset_type, &zc->zc_cookie,
5069 buf, &zc->zc_nvlist_dst_size);
5072 error = ddi_copyout(buf,
5073 (void *)(uintptr_t)zc->zc_nvlist_dst,
5074 zc->zc_nvlist_dst_size, zc->zc_iflags);
5076 kmem_free(buf, bufsize);
5077 zfsvfs_rele(zfsvfs, FTAG);
5084 * zc_name name of filesystem
5090 zfs_ioc_userspace_upgrade(zfs_cmd_t *zc)
5096 if (getzfsvfs(zc->zc_name, &zfsvfs) == 0) {
5097 if (!dmu_objset_userused_enabled(zfsvfs->z_os)) {
5099 * If userused is not enabled, it may be because the
5100 * objset needs to be closed & reopened (to grow the
5101 * objset_phys_t). Suspend/resume the fs will do that.
5105 ds = dmu_objset_ds(zfsvfs->z_os);
5106 error = zfs_suspend_fs(zfsvfs);
5108 dmu_objset_refresh_ownership(zfsvfs->z_os,
5110 error = zfs_resume_fs(zfsvfs, ds);
5114 error = dmu_objset_userspace_upgrade(zfsvfs->z_os);
5116 VFS_RELE(zfsvfs->z_vfs);
5118 vfs_unbusy(zfsvfs->z_vfs);
5121 /* XXX kind of reading contents without owning */
5122 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
5126 error = dmu_objset_userspace_upgrade(os);
5127 dmu_objset_rele(os, FTAG);
5135 * We don't want to have a hard dependency
5136 * against some special symbols in sharefs
5137 * nfs, and smbsrv. Determine them if needed when
5138 * the first file system is shared.
5139 * Neither sharefs, nfs or smbsrv are unloadable modules.
5141 int (*znfsexport_fs)(void *arg);
5142 int (*zshare_fs)(enum sharefs_sys_op, share_t *, uint32_t);
5143 int (*zsmbexport_fs)(void *arg, boolean_t add_share);
5145 int zfs_nfsshare_inited;
5146 int zfs_smbshare_inited;
5148 ddi_modhandle_t nfs_mod;
5149 ddi_modhandle_t sharefs_mod;
5150 ddi_modhandle_t smbsrv_mod;
5151 #endif /* illumos */
5152 kmutex_t zfs_share_lock;
5160 ASSERT(MUTEX_HELD(&zfs_share_lock));
5161 /* Both NFS and SMB shares also require sharetab support. */
5162 if (sharefs_mod == NULL && ((sharefs_mod =
5163 ddi_modopen("fs/sharefs",
5164 KRTLD_MODE_FIRST, &error)) == NULL)) {
5165 return (SET_ERROR(ENOSYS));
5167 if (zshare_fs == NULL && ((zshare_fs =
5168 (int (*)(enum sharefs_sys_op, share_t *, uint32_t))
5169 ddi_modsym(sharefs_mod, "sharefs_impl", &error)) == NULL)) {
5170 return (SET_ERROR(ENOSYS));
5174 #endif /* illumos */
5177 zfs_ioc_share(zfs_cmd_t *zc)
5183 switch (zc->zc_share.z_sharetype) {
5185 case ZFS_UNSHARE_NFS:
5186 if (zfs_nfsshare_inited == 0) {
5187 mutex_enter(&zfs_share_lock);
5188 if (nfs_mod == NULL && ((nfs_mod = ddi_modopen("fs/nfs",
5189 KRTLD_MODE_FIRST, &error)) == NULL)) {
5190 mutex_exit(&zfs_share_lock);
5191 return (SET_ERROR(ENOSYS));
5193 if (znfsexport_fs == NULL &&
5194 ((znfsexport_fs = (int (*)(void *))
5196 "nfs_export", &error)) == NULL)) {
5197 mutex_exit(&zfs_share_lock);
5198 return (SET_ERROR(ENOSYS));
5200 error = zfs_init_sharefs();
5202 mutex_exit(&zfs_share_lock);
5203 return (SET_ERROR(ENOSYS));
5205 zfs_nfsshare_inited = 1;
5206 mutex_exit(&zfs_share_lock);
5210 case ZFS_UNSHARE_SMB:
5211 if (zfs_smbshare_inited == 0) {
5212 mutex_enter(&zfs_share_lock);
5213 if (smbsrv_mod == NULL && ((smbsrv_mod =
5214 ddi_modopen("drv/smbsrv",
5215 KRTLD_MODE_FIRST, &error)) == NULL)) {
5216 mutex_exit(&zfs_share_lock);
5217 return (SET_ERROR(ENOSYS));
5219 if (zsmbexport_fs == NULL && ((zsmbexport_fs =
5220 (int (*)(void *, boolean_t))ddi_modsym(smbsrv_mod,
5221 "smb_server_share", &error)) == NULL)) {
5222 mutex_exit(&zfs_share_lock);
5223 return (SET_ERROR(ENOSYS));
5225 error = zfs_init_sharefs();
5227 mutex_exit(&zfs_share_lock);
5228 return (SET_ERROR(ENOSYS));
5230 zfs_smbshare_inited = 1;
5231 mutex_exit(&zfs_share_lock);
5235 return (SET_ERROR(EINVAL));
5238 switch (zc->zc_share.z_sharetype) {
5240 case ZFS_UNSHARE_NFS:
5242 znfsexport_fs((void *)
5243 (uintptr_t)zc->zc_share.z_exportdata))
5247 case ZFS_UNSHARE_SMB:
5248 if (error = zsmbexport_fs((void *)
5249 (uintptr_t)zc->zc_share.z_exportdata,
5250 zc->zc_share.z_sharetype == ZFS_SHARE_SMB ?
5257 opcode = (zc->zc_share.z_sharetype == ZFS_SHARE_NFS ||
5258 zc->zc_share.z_sharetype == ZFS_SHARE_SMB) ?
5259 SHAREFS_ADD : SHAREFS_REMOVE;
5262 * Add or remove share from sharetab
5264 error = zshare_fs(opcode,
5265 (void *)(uintptr_t)zc->zc_share.z_sharedata,
5266 zc->zc_share.z_sharemax);
5270 #else /* !illumos */
5272 #endif /* illumos */
5275 ace_t full_access[] = {
5276 {(uid_t)-1, ACE_ALL_PERMS, ACE_EVERYONE, 0}
5281 * zc_name name of containing filesystem
5282 * zc_obj object # beyond which we want next in-use object #
5285 * zc_obj next in-use object #
5288 zfs_ioc_next_obj(zfs_cmd_t *zc)
5290 objset_t *os = NULL;
5293 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
5297 error = dmu_object_next(os, &zc->zc_obj, B_FALSE,
5298 dsl_dataset_phys(os->os_dsl_dataset)->ds_prev_snap_txg);
5300 dmu_objset_rele(os, FTAG);
5306 * zc_name name of filesystem
5307 * zc_value prefix name for snapshot
5308 * zc_cleanup_fd cleanup-on-exit file descriptor for calling process
5311 * zc_value short name of new snapshot
5314 zfs_ioc_tmp_snapshot(zfs_cmd_t *zc)
5321 error = zfs_onexit_fd_hold(zc->zc_cleanup_fd, &minor);
5325 snap_name = kmem_asprintf("%s-%016llx", zc->zc_value,
5326 (u_longlong_t)ddi_get_lbolt64());
5327 hold_name = kmem_asprintf("%%%s", zc->zc_value);
5329 error = dsl_dataset_snapshot_tmp(zc->zc_name, snap_name, minor,
5332 (void) strcpy(zc->zc_value, snap_name);
5335 zfs_onexit_fd_rele(zc->zc_cleanup_fd);
5341 * zc_name name of "to" snapshot
5342 * zc_value name of "from" snapshot
5343 * zc_cookie file descriptor to write diff data on
5346 * dmu_diff_record_t's to the file descriptor
5349 zfs_ioc_diff(zfs_cmd_t *zc)
5352 cap_rights_t rights;
5357 fp = getf(zc->zc_cookie);
5359 fget_write(curthread, zc->zc_cookie,
5360 cap_rights_init(&rights, CAP_WRITE), &fp);
5363 return (SET_ERROR(EBADF));
5368 error = dmu_diff(zc->zc_name, zc->zc_value, fp->f_vnode, &off);
5370 error = dmu_diff(zc->zc_name, zc->zc_value, fp, &off);
5373 if (off >= 0 && off <= MAXOFFSET_T)
5375 releasef(zc->zc_cookie);
5382 * Remove all ACL files in shares dir
5385 zfs_smb_acl_purge(znode_t *dzp)
5388 zap_attribute_t zap;
5389 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
5392 for (zap_cursor_init(&zc, zfsvfs->z_os, dzp->z_id);
5393 (error = zap_cursor_retrieve(&zc, &zap)) == 0;
5394 zap_cursor_advance(&zc)) {
5395 if ((error = VOP_REMOVE(ZTOV(dzp), zap.za_name, kcred,
5399 zap_cursor_fini(&zc);
5402 #endif /* illumos */
5405 zfs_ioc_smb_acl(zfs_cmd_t *zc)
5410 vnode_t *resourcevp = NULL;
5419 if ((error = lookupname(zc->zc_value, UIO_SYSSPACE,
5420 NO_FOLLOW, NULL, &vp)) != 0)
5423 /* Now make sure mntpnt and dataset are ZFS */
5425 if (strcmp(vp->v_vfsp->mnt_stat.f_fstypename, "zfs") != 0 ||
5426 (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource),
5427 zc->zc_name) != 0)) {
5429 return (SET_ERROR(EINVAL));
5433 zfsvfs = dzp->z_zfsvfs;
5437 * Create share dir if its missing.
5439 mutex_enter(&zfsvfs->z_lock);
5440 if (zfsvfs->z_shares_dir == 0) {
5443 tx = dmu_tx_create(zfsvfs->z_os);
5444 dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, TRUE,
5446 dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL);
5447 error = dmu_tx_assign(tx, TXG_WAIT);
5451 error = zfs_create_share_dir(zfsvfs, tx);
5455 mutex_exit(&zfsvfs->z_lock);
5461 mutex_exit(&zfsvfs->z_lock);
5463 ASSERT(zfsvfs->z_shares_dir);
5464 if ((error = zfs_zget(zfsvfs, zfsvfs->z_shares_dir, &sharedir)) != 0) {
5470 switch (zc->zc_cookie) {
5471 case ZFS_SMB_ACL_ADD:
5472 vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE;
5473 vattr.va_type = VREG;
5474 vattr.va_mode = S_IFREG|0777;
5478 vsec.vsa_mask = VSA_ACE;
5479 vsec.vsa_aclentp = &full_access;
5480 vsec.vsa_aclentsz = sizeof (full_access);
5481 vsec.vsa_aclcnt = 1;
5483 error = VOP_CREATE(ZTOV(sharedir), zc->zc_string,
5484 &vattr, EXCL, 0, &resourcevp, kcred, 0, NULL, &vsec);
5486 VN_RELE(resourcevp);
5489 case ZFS_SMB_ACL_REMOVE:
5490 error = VOP_REMOVE(ZTOV(sharedir), zc->zc_string, kcred,
5494 case ZFS_SMB_ACL_RENAME:
5495 if ((error = get_nvlist(zc->zc_nvlist_src,
5496 zc->zc_nvlist_src_size, zc->zc_iflags, &nvlist)) != 0) {
5498 VN_RELE(ZTOV(sharedir));
5502 if (nvlist_lookup_string(nvlist, ZFS_SMB_ACL_SRC, &src) ||
5503 nvlist_lookup_string(nvlist, ZFS_SMB_ACL_TARGET,
5506 VN_RELE(ZTOV(sharedir));
5508 nvlist_free(nvlist);
5511 error = VOP_RENAME(ZTOV(sharedir), src, ZTOV(sharedir), target,
5513 nvlist_free(nvlist);
5516 case ZFS_SMB_ACL_PURGE:
5517 error = zfs_smb_acl_purge(sharedir);
5521 error = SET_ERROR(EINVAL);
5526 VN_RELE(ZTOV(sharedir));
5531 #else /* !illumos */
5532 return (EOPNOTSUPP);
5533 #endif /* illumos */
5538 * "holds" -> { snapname -> holdname (string), ... }
5539 * (optional) "cleanup_fd" -> fd (int32)
5543 * snapname -> error value (int32)
5549 zfs_ioc_hold(const char *pool, nvlist_t *args, nvlist_t *errlist)
5553 int cleanup_fd = -1;
5557 error = nvlist_lookup_nvlist(args, "holds", &holds);
5559 return (SET_ERROR(EINVAL));
5561 /* make sure the user didn't pass us any invalid (empty) tags */
5562 for (pair = nvlist_next_nvpair(holds, NULL); pair != NULL;
5563 pair = nvlist_next_nvpair(holds, pair)) {
5566 error = nvpair_value_string(pair, &htag);
5568 return (SET_ERROR(error));
5570 if (strlen(htag) == 0)
5571 return (SET_ERROR(EINVAL));
5574 if (nvlist_lookup_int32(args, "cleanup_fd", &cleanup_fd) == 0) {
5575 error = zfs_onexit_fd_hold(cleanup_fd, &minor);
5580 error = dsl_dataset_user_hold(holds, minor, errlist);
5582 zfs_onexit_fd_rele(cleanup_fd);
5587 * innvl is not used.
5590 * holdname -> time added (uint64 seconds since epoch)
5596 zfs_ioc_get_holds(const char *snapname, nvlist_t *args, nvlist_t *outnvl)
5598 return (dsl_dataset_get_holds(snapname, outnvl));
5603 * snapname -> { holdname, ... }
5608 * snapname -> error value (int32)
5614 zfs_ioc_release(const char *pool, nvlist_t *holds, nvlist_t *errlist)
5616 return (dsl_dataset_user_release(holds, errlist));
5621 * zc_name name of new filesystem or snapshot
5622 * zc_value full name of old snapshot
5625 * zc_cookie space in bytes
5626 * zc_objset_type compressed space in bytes
5627 * zc_perm_action uncompressed space in bytes
5630 zfs_ioc_space_written(zfs_cmd_t *zc)
5634 dsl_dataset_t *new, *old;
5636 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
5639 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &new);
5641 dsl_pool_rele(dp, FTAG);
5644 error = dsl_dataset_hold(dp, zc->zc_value, FTAG, &old);
5646 dsl_dataset_rele(new, FTAG);
5647 dsl_pool_rele(dp, FTAG);
5651 error = dsl_dataset_space_written(old, new, &zc->zc_cookie,
5652 &zc->zc_objset_type, &zc->zc_perm_action);
5653 dsl_dataset_rele(old, FTAG);
5654 dsl_dataset_rele(new, FTAG);
5655 dsl_pool_rele(dp, FTAG);
5661 * "firstsnap" -> snapshot name
5665 * "used" -> space in bytes
5666 * "compressed" -> compressed space in bytes
5667 * "uncompressed" -> uncompressed space in bytes
5671 zfs_ioc_space_snaps(const char *lastsnap, nvlist_t *innvl, nvlist_t *outnvl)
5675 dsl_dataset_t *new, *old;
5677 uint64_t used, comp, uncomp;
5679 if (nvlist_lookup_string(innvl, "firstsnap", &firstsnap) != 0)
5680 return (SET_ERROR(EINVAL));
5682 error = dsl_pool_hold(lastsnap, FTAG, &dp);
5686 error = dsl_dataset_hold(dp, lastsnap, FTAG, &new);
5687 if (error == 0 && !new->ds_is_snapshot) {
5688 dsl_dataset_rele(new, FTAG);
5689 error = SET_ERROR(EINVAL);
5692 dsl_pool_rele(dp, FTAG);
5695 error = dsl_dataset_hold(dp, firstsnap, FTAG, &old);
5696 if (error == 0 && !old->ds_is_snapshot) {
5697 dsl_dataset_rele(old, FTAG);
5698 error = SET_ERROR(EINVAL);
5701 dsl_dataset_rele(new, FTAG);
5702 dsl_pool_rele(dp, FTAG);
5706 error = dsl_dataset_space_wouldfree(old, new, &used, &comp, &uncomp);
5707 dsl_dataset_rele(old, FTAG);
5708 dsl_dataset_rele(new, FTAG);
5709 dsl_pool_rele(dp, FTAG);
5710 fnvlist_add_uint64(outnvl, "used", used);
5711 fnvlist_add_uint64(outnvl, "compressed", comp);
5712 fnvlist_add_uint64(outnvl, "uncompressed", uncomp);
5717 zfs_ioc_jail(zfs_cmd_t *zc)
5720 return (zone_dataset_attach(curthread->td_ucred, zc->zc_name,
5721 (int)zc->zc_jailid));
5725 zfs_ioc_unjail(zfs_cmd_t *zc)
5728 return (zone_dataset_detach(curthread->td_ucred, zc->zc_name,
5729 (int)zc->zc_jailid));
5734 * "fd" -> file descriptor to write stream to (int32)
5735 * (optional) "fromsnap" -> full snap name to send an incremental from
5736 * (optional) "largeblockok" -> (value ignored)
5737 * indicates that blocks > 128KB are permitted
5738 * (optional) "embedok" -> (value ignored)
5739 * presence indicates DRR_WRITE_EMBEDDED records are permitted
5740 * (optional) "compressok" -> (value ignored)
5741 * presence indicates compressed DRR_WRITE records are permitted
5742 * (optional) "resume_object" and "resume_offset" -> (uint64)
5743 * if present, resume send stream from specified object and offset.
5750 zfs_ioc_send_new(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl)
5752 cap_rights_t rights;
5756 char *fromname = NULL;
5758 boolean_t largeblockok;
5760 boolean_t compressok;
5761 uint64_t resumeobj = 0;
5762 uint64_t resumeoff = 0;
5764 error = nvlist_lookup_int32(innvl, "fd", &fd);
5766 return (SET_ERROR(EINVAL));
5768 (void) nvlist_lookup_string(innvl, "fromsnap", &fromname);
5770 largeblockok = nvlist_exists(innvl, "largeblockok");
5771 embedok = nvlist_exists(innvl, "embedok");
5772 compressok = nvlist_exists(innvl, "compressok");
5774 (void) nvlist_lookup_uint64(innvl, "resume_object", &resumeobj);
5775 (void) nvlist_lookup_uint64(innvl, "resume_offset", &resumeoff);
5778 file_t *fp = getf(fd);
5780 fget_write(curthread, fd, cap_rights_init(&rights, CAP_WRITE), &fp);
5783 return (SET_ERROR(EBADF));
5786 error = dmu_send(snapname, fromname, embedok, largeblockok, compressok,
5788 fd, resumeobj, resumeoff, fp->f_vnode, &off);
5790 fd, resumeobj, resumeoff, fp, &off);
5794 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
5805 * Determine approximately how large a zfs send stream will be -- the number
5806 * of bytes that will be written to the fd supplied to zfs_ioc_send_new().
5809 * (optional) "from" -> full snap or bookmark name to send an incremental
5811 * (optional) "largeblockok" -> (value ignored)
5812 * indicates that blocks > 128KB are permitted
5813 * (optional) "embedok" -> (value ignored)
5814 * presence indicates DRR_WRITE_EMBEDDED records are permitted
5815 * (optional) "compressok" -> (value ignored)
5816 * presence indicates compressed DRR_WRITE records are permitted
5820 * "space" -> bytes of space (uint64)
5824 zfs_ioc_send_space(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl)
5827 dsl_dataset_t *tosnap;
5830 boolean_t compressok;
5833 error = dsl_pool_hold(snapname, FTAG, &dp);
5837 error = dsl_dataset_hold(dp, snapname, FTAG, &tosnap);
5839 dsl_pool_rele(dp, FTAG);
5843 compressok = nvlist_exists(innvl, "compressok");
5845 error = nvlist_lookup_string(innvl, "from", &fromname);
5847 if (strchr(fromname, '@') != NULL) {
5849 * If from is a snapshot, hold it and use the more
5850 * efficient dmu_send_estimate to estimate send space
5851 * size using deadlists.
5853 dsl_dataset_t *fromsnap;
5854 error = dsl_dataset_hold(dp, fromname, FTAG, &fromsnap);
5857 error = dmu_send_estimate(tosnap, fromsnap, compressok,
5859 dsl_dataset_rele(fromsnap, FTAG);
5860 } else if (strchr(fromname, '#') != NULL) {
5862 * If from is a bookmark, fetch the creation TXG of the
5863 * snapshot it was created from and use that to find
5864 * blocks that were born after it.
5866 zfs_bookmark_phys_t frombm;
5868 error = dsl_bookmark_lookup(dp, fromname, tosnap,
5872 error = dmu_send_estimate_from_txg(tosnap,
5873 frombm.zbm_creation_txg, compressok, &space);
5876 * from is not properly formatted as a snapshot or
5879 error = SET_ERROR(EINVAL);
5884 * If estimating the size of a full send, use dmu_send_estimate.
5886 error = dmu_send_estimate(tosnap, NULL, compressok, &space);
5889 fnvlist_add_uint64(outnvl, "space", space);
5892 dsl_dataset_rele(tosnap, FTAG);
5893 dsl_pool_rele(dp, FTAG);
5897 static zfs_ioc_vec_t zfs_ioc_vec[ZFS_IOC_LAST - ZFS_IOC_FIRST];
5900 zfs_ioctl_register_legacy(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5901 zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck,
5902 boolean_t log_history, zfs_ioc_poolcheck_t pool_check)
5904 zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST];
5906 ASSERT3U(ioc, >=, ZFS_IOC_FIRST);
5907 ASSERT3U(ioc, <, ZFS_IOC_LAST);
5908 ASSERT3P(vec->zvec_legacy_func, ==, NULL);
5909 ASSERT3P(vec->zvec_func, ==, NULL);
5911 vec->zvec_legacy_func = func;
5912 vec->zvec_secpolicy = secpolicy;
5913 vec->zvec_namecheck = namecheck;
5914 vec->zvec_allow_log = log_history;
5915 vec->zvec_pool_check = pool_check;
5919 * See the block comment at the beginning of this file for details on
5920 * each argument to this function.
5923 zfs_ioctl_register(const char *name, zfs_ioc_t ioc, zfs_ioc_func_t *func,
5924 zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck,
5925 zfs_ioc_poolcheck_t pool_check, boolean_t smush_outnvlist,
5926 boolean_t allow_log)
5928 zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST];
5930 ASSERT3U(ioc, >=, ZFS_IOC_FIRST);
5931 ASSERT3U(ioc, <, ZFS_IOC_LAST);
5932 ASSERT3P(vec->zvec_legacy_func, ==, NULL);
5933 ASSERT3P(vec->zvec_func, ==, NULL);
5935 /* if we are logging, the name must be valid */
5936 ASSERT(!allow_log || namecheck != NO_NAME);
5938 vec->zvec_name = name;
5939 vec->zvec_func = func;
5940 vec->zvec_secpolicy = secpolicy;
5941 vec->zvec_namecheck = namecheck;
5942 vec->zvec_pool_check = pool_check;
5943 vec->zvec_smush_outnvlist = smush_outnvlist;
5944 vec->zvec_allow_log = allow_log;
5948 zfs_ioctl_register_pool(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5949 zfs_secpolicy_func_t *secpolicy, boolean_t log_history,
5950 zfs_ioc_poolcheck_t pool_check)
5952 zfs_ioctl_register_legacy(ioc, func, secpolicy,
5953 POOL_NAME, log_history, pool_check);
5957 zfs_ioctl_register_dataset_nolog(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5958 zfs_secpolicy_func_t *secpolicy, zfs_ioc_poolcheck_t pool_check)
5960 zfs_ioctl_register_legacy(ioc, func, secpolicy,
5961 DATASET_NAME, B_FALSE, pool_check);
5965 zfs_ioctl_register_pool_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func)
5967 zfs_ioctl_register_legacy(ioc, func, zfs_secpolicy_config,
5968 POOL_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5972 zfs_ioctl_register_pool_meta(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5973 zfs_secpolicy_func_t *secpolicy)
5975 zfs_ioctl_register_legacy(ioc, func, secpolicy,
5976 NO_NAME, B_FALSE, POOL_CHECK_NONE);
5980 zfs_ioctl_register_dataset_read_secpolicy(zfs_ioc_t ioc,
5981 zfs_ioc_legacy_func_t *func, zfs_secpolicy_func_t *secpolicy)
5983 zfs_ioctl_register_legacy(ioc, func, secpolicy,
5984 DATASET_NAME, B_FALSE, POOL_CHECK_SUSPENDED);
5988 zfs_ioctl_register_dataset_read(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func)
5990 zfs_ioctl_register_dataset_read_secpolicy(ioc, func,
5991 zfs_secpolicy_read);
5995 zfs_ioctl_register_dataset_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5996 zfs_secpolicy_func_t *secpolicy)
5998 zfs_ioctl_register_legacy(ioc, func, secpolicy,
5999 DATASET_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
6003 zfs_ioctl_init(void)
6005 zfs_ioctl_register("snapshot", ZFS_IOC_SNAPSHOT,
6006 zfs_ioc_snapshot, zfs_secpolicy_snapshot, POOL_NAME,
6007 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
6009 zfs_ioctl_register("log_history", ZFS_IOC_LOG_HISTORY,
6010 zfs_ioc_log_history, zfs_secpolicy_log_history, NO_NAME,
6011 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_FALSE);
6013 zfs_ioctl_register("space_snaps", ZFS_IOC_SPACE_SNAPS,
6014 zfs_ioc_space_snaps, zfs_secpolicy_read, DATASET_NAME,
6015 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
6017 zfs_ioctl_register("send", ZFS_IOC_SEND_NEW,
6018 zfs_ioc_send_new, zfs_secpolicy_send_new, DATASET_NAME,
6019 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
6021 zfs_ioctl_register("send_space", ZFS_IOC_SEND_SPACE,
6022 zfs_ioc_send_space, zfs_secpolicy_read, DATASET_NAME,
6023 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
6025 zfs_ioctl_register("create", ZFS_IOC_CREATE,
6026 zfs_ioc_create, zfs_secpolicy_create_clone, DATASET_NAME,
6027 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
6029 zfs_ioctl_register("clone", ZFS_IOC_CLONE,
6030 zfs_ioc_clone, zfs_secpolicy_create_clone, DATASET_NAME,
6031 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
6033 zfs_ioctl_register("destroy_snaps", ZFS_IOC_DESTROY_SNAPS,
6034 zfs_ioc_destroy_snaps, zfs_secpolicy_destroy_snaps, POOL_NAME,
6035 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
6037 zfs_ioctl_register("hold", ZFS_IOC_HOLD,
6038 zfs_ioc_hold, zfs_secpolicy_hold, POOL_NAME,
6039 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
6040 zfs_ioctl_register("release", ZFS_IOC_RELEASE,
6041 zfs_ioc_release, zfs_secpolicy_release, POOL_NAME,
6042 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
6044 zfs_ioctl_register("get_holds", ZFS_IOC_GET_HOLDS,
6045 zfs_ioc_get_holds, zfs_secpolicy_read, DATASET_NAME,
6046 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
6048 zfs_ioctl_register("rollback", ZFS_IOC_ROLLBACK,
6049 zfs_ioc_rollback, zfs_secpolicy_rollback, DATASET_NAME,
6050 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_TRUE);
6052 zfs_ioctl_register("bookmark", ZFS_IOC_BOOKMARK,
6053 zfs_ioc_bookmark, zfs_secpolicy_bookmark, POOL_NAME,
6054 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
6056 zfs_ioctl_register("get_bookmarks", ZFS_IOC_GET_BOOKMARKS,
6057 zfs_ioc_get_bookmarks, zfs_secpolicy_read, DATASET_NAME,
6058 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
6060 zfs_ioctl_register("destroy_bookmarks", ZFS_IOC_DESTROY_BOOKMARKS,
6061 zfs_ioc_destroy_bookmarks, zfs_secpolicy_destroy_bookmarks,
6063 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
6065 zfs_ioctl_register("channel_program", ZFS_IOC_CHANNEL_PROGRAM,
6066 zfs_ioc_channel_program, zfs_secpolicy_config,
6067 POOL_NAME, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE,
6070 /* IOCTLS that use the legacy function signature */
6072 zfs_ioctl_register_legacy(ZFS_IOC_POOL_FREEZE, zfs_ioc_pool_freeze,
6073 zfs_secpolicy_config, NO_NAME, B_FALSE, POOL_CHECK_READONLY);
6075 zfs_ioctl_register_pool(ZFS_IOC_POOL_CREATE, zfs_ioc_pool_create,
6076 zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE);
6077 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SCAN,
6079 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_UPGRADE,
6080 zfs_ioc_pool_upgrade);
6081 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ADD,
6083 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_REMOVE,
6084 zfs_ioc_vdev_remove);
6085 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SET_STATE,
6086 zfs_ioc_vdev_set_state);
6087 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ATTACH,
6088 zfs_ioc_vdev_attach);
6089 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_DETACH,
6090 zfs_ioc_vdev_detach);
6091 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETPATH,
6092 zfs_ioc_vdev_setpath);
6093 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETFRU,
6094 zfs_ioc_vdev_setfru);
6095 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SET_PROPS,
6096 zfs_ioc_pool_set_props);
6097 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SPLIT,
6098 zfs_ioc_vdev_split);
6099 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_REGUID,
6100 zfs_ioc_pool_reguid);
6102 zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_CONFIGS,
6103 zfs_ioc_pool_configs, zfs_secpolicy_none);
6104 zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_TRYIMPORT,
6105 zfs_ioc_pool_tryimport, zfs_secpolicy_config);
6106 zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_FAULT,
6107 zfs_ioc_inject_fault, zfs_secpolicy_inject);
6108 zfs_ioctl_register_pool_meta(ZFS_IOC_CLEAR_FAULT,
6109 zfs_ioc_clear_fault, zfs_secpolicy_inject);
6110 zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_LIST_NEXT,
6111 zfs_ioc_inject_list_next, zfs_secpolicy_inject);
6114 * pool destroy, and export don't log the history as part of
6115 * zfsdev_ioctl, but rather zfs_ioc_pool_export
6116 * does the logging of those commands.
6118 zfs_ioctl_register_pool(ZFS_IOC_POOL_DESTROY, zfs_ioc_pool_destroy,
6119 zfs_secpolicy_config, B_FALSE, POOL_CHECK_NONE);
6120 zfs_ioctl_register_pool(ZFS_IOC_POOL_EXPORT, zfs_ioc_pool_export,
6121 zfs_secpolicy_config, B_FALSE, POOL_CHECK_NONE);
6123 zfs_ioctl_register_pool(ZFS_IOC_POOL_STATS, zfs_ioc_pool_stats,
6124 zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE);
6125 zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_PROPS, zfs_ioc_pool_get_props,
6126 zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE);
6128 zfs_ioctl_register_pool(ZFS_IOC_ERROR_LOG, zfs_ioc_error_log,
6129 zfs_secpolicy_inject, B_FALSE, POOL_CHECK_NONE);
6130 zfs_ioctl_register_pool(ZFS_IOC_DSOBJ_TO_DSNAME,
6131 zfs_ioc_dsobj_to_dsname,
6132 zfs_secpolicy_diff, B_FALSE, POOL_CHECK_NONE);
6133 zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_HISTORY,
6134 zfs_ioc_pool_get_history,
6135 zfs_secpolicy_config, B_FALSE, POOL_CHECK_SUSPENDED);
6137 zfs_ioctl_register_pool(ZFS_IOC_POOL_IMPORT, zfs_ioc_pool_import,
6138 zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE);
6140 zfs_ioctl_register_pool(ZFS_IOC_CLEAR, zfs_ioc_clear,
6141 zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE);
6142 zfs_ioctl_register_pool(ZFS_IOC_POOL_REOPEN, zfs_ioc_pool_reopen,
6143 zfs_secpolicy_config, B_TRUE, POOL_CHECK_SUSPENDED);
6145 zfs_ioctl_register_dataset_read(ZFS_IOC_SPACE_WRITTEN,
6146 zfs_ioc_space_written);
6147 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_RECVD_PROPS,
6148 zfs_ioc_objset_recvd_props);
6149 zfs_ioctl_register_dataset_read(ZFS_IOC_NEXT_OBJ,
6151 zfs_ioctl_register_dataset_read(ZFS_IOC_GET_FSACL,
6153 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_STATS,
6154 zfs_ioc_objset_stats);
6155 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_ZPLPROPS,
6156 zfs_ioc_objset_zplprops);
6157 zfs_ioctl_register_dataset_read(ZFS_IOC_DATASET_LIST_NEXT,
6158 zfs_ioc_dataset_list_next);
6159 zfs_ioctl_register_dataset_read(ZFS_IOC_SNAPSHOT_LIST_NEXT,
6160 zfs_ioc_snapshot_list_next);
6161 zfs_ioctl_register_dataset_read(ZFS_IOC_SEND_PROGRESS,
6162 zfs_ioc_send_progress);
6164 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_DIFF,
6165 zfs_ioc_diff, zfs_secpolicy_diff);
6166 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_STATS,
6167 zfs_ioc_obj_to_stats, zfs_secpolicy_diff);
6168 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_PATH,
6169 zfs_ioc_obj_to_path, zfs_secpolicy_diff);
6170 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_ONE,
6171 zfs_ioc_userspace_one, zfs_secpolicy_userspace_one);
6172 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_MANY,
6173 zfs_ioc_userspace_many, zfs_secpolicy_userspace_many);
6174 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_SEND,
6175 zfs_ioc_send, zfs_secpolicy_send);
6177 zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_PROP, zfs_ioc_set_prop,
6178 zfs_secpolicy_none);
6179 zfs_ioctl_register_dataset_modify(ZFS_IOC_DESTROY, zfs_ioc_destroy,
6180 zfs_secpolicy_destroy);
6181 zfs_ioctl_register_dataset_modify(ZFS_IOC_RENAME, zfs_ioc_rename,
6182 zfs_secpolicy_rename);
6183 zfs_ioctl_register_dataset_modify(ZFS_IOC_RECV, zfs_ioc_recv,
6184 zfs_secpolicy_recv);
6185 zfs_ioctl_register_dataset_modify(ZFS_IOC_PROMOTE, zfs_ioc_promote,
6186 zfs_secpolicy_promote);
6187 zfs_ioctl_register_dataset_modify(ZFS_IOC_INHERIT_PROP,
6188 zfs_ioc_inherit_prop, zfs_secpolicy_inherit_prop);
6189 zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_FSACL, zfs_ioc_set_fsacl,
6190 zfs_secpolicy_set_fsacl);
6192 zfs_ioctl_register_dataset_nolog(ZFS_IOC_SHARE, zfs_ioc_share,
6193 zfs_secpolicy_share, POOL_CHECK_NONE);
6194 zfs_ioctl_register_dataset_nolog(ZFS_IOC_SMB_ACL, zfs_ioc_smb_acl,
6195 zfs_secpolicy_smb_acl, POOL_CHECK_NONE);
6196 zfs_ioctl_register_dataset_nolog(ZFS_IOC_USERSPACE_UPGRADE,
6197 zfs_ioc_userspace_upgrade, zfs_secpolicy_userspace_upgrade,
6198 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
6199 zfs_ioctl_register_dataset_nolog(ZFS_IOC_TMP_SNAPSHOT,
6200 zfs_ioc_tmp_snapshot, zfs_secpolicy_tmp_snapshot,
6201 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
6204 zfs_ioctl_register_dataset_nolog(ZFS_IOC_JAIL, zfs_ioc_jail,
6205 zfs_secpolicy_config, POOL_CHECK_NONE);
6206 zfs_ioctl_register_dataset_nolog(ZFS_IOC_UNJAIL, zfs_ioc_unjail,
6207 zfs_secpolicy_config, POOL_CHECK_NONE);
6208 zfs_ioctl_register("fbsd_nextboot", ZFS_IOC_NEXTBOOT,
6209 zfs_ioc_nextboot, zfs_secpolicy_config, NO_NAME,
6210 POOL_CHECK_NONE, B_FALSE, B_FALSE);
6215 pool_status_check(const char *name, zfs_ioc_namecheck_t type,
6216 zfs_ioc_poolcheck_t check)
6221 ASSERT(type == POOL_NAME || type == DATASET_NAME);
6223 if (check & POOL_CHECK_NONE)
6226 error = spa_open(name, &spa, FTAG);
6228 if ((check & POOL_CHECK_SUSPENDED) && spa_suspended(spa))
6229 error = SET_ERROR(EAGAIN);
6230 else if ((check & POOL_CHECK_READONLY) && !spa_writeable(spa))
6231 error = SET_ERROR(EROFS);
6232 spa_close(spa, FTAG);
6238 * Find a free minor number.
6241 zfsdev_minor_alloc(void)
6243 static minor_t last_minor;
6246 ASSERT(MUTEX_HELD(&spa_namespace_lock));
6248 for (m = last_minor + 1; m != last_minor; m++) {
6249 if (m > ZFSDEV_MAX_MINOR)
6251 if (ddi_get_soft_state(zfsdev_state, m) == NULL) {
6261 zfs_ctldev_init(struct cdev *devp)
6264 zfs_soft_state_t *zs;
6266 ASSERT(MUTEX_HELD(&spa_namespace_lock));
6268 minor = zfsdev_minor_alloc();
6270 return (SET_ERROR(ENXIO));
6272 if (ddi_soft_state_zalloc(zfsdev_state, minor) != DDI_SUCCESS)
6273 return (SET_ERROR(EAGAIN));
6275 devfs_set_cdevpriv((void *)(uintptr_t)minor, zfsdev_close);
6277 zs = ddi_get_soft_state(zfsdev_state, minor);
6278 zs->zss_type = ZSST_CTLDEV;
6279 zfs_onexit_init((zfs_onexit_t **)&zs->zss_data);
6285 zfs_ctldev_destroy(zfs_onexit_t *zo, minor_t minor)
6287 ASSERT(MUTEX_HELD(&spa_namespace_lock));
6289 zfs_onexit_destroy(zo);
6290 ddi_soft_state_free(zfsdev_state, minor);
6294 zfsdev_get_soft_state(minor_t minor, enum zfs_soft_state_type which)
6296 zfs_soft_state_t *zp;
6298 zp = ddi_get_soft_state(zfsdev_state, minor);
6299 if (zp == NULL || zp->zss_type != which)
6302 return (zp->zss_data);
6306 zfsdev_open(struct cdev *devp, int flag, int mode, struct thread *td)
6311 if (getminor(*devp) != 0)
6312 return (zvol_open(devp, flag, otyp, cr));
6315 /* This is the control device. Allocate a new minor if requested. */
6317 mutex_enter(&spa_namespace_lock);
6318 error = zfs_ctldev_init(devp);
6319 mutex_exit(&spa_namespace_lock);
6326 zfsdev_close(void *data)
6329 minor_t minor = (minor_t)(uintptr_t)data;
6334 mutex_enter(&spa_namespace_lock);
6335 zo = zfsdev_get_soft_state(minor, ZSST_CTLDEV);
6337 mutex_exit(&spa_namespace_lock);
6340 zfs_ctldev_destroy(zo, minor);
6341 mutex_exit(&spa_namespace_lock);
6345 zfsdev_ioctl(struct cdev *dev, u_long zcmd, caddr_t arg, int flag,
6352 minor_t minor = getminor(dev);
6354 zfs_iocparm_t *zc_iocparm;
6355 int cflag, cmd, oldvecnum;
6356 boolean_t newioc, compat;
6357 void *compat_zc = NULL;
6358 cred_t *cr = td->td_ucred;
6360 const zfs_ioc_vec_t *vec;
6361 char *saved_poolname = NULL;
6362 nvlist_t *innvl = NULL;
6364 cflag = ZFS_CMD_COMPAT_NONE;
6366 newioc = B_TRUE; /* "new" style (zfs_iocparm_t) ioctl */
6368 len = IOCPARM_LEN(zcmd);
6369 vecnum = cmd = zcmd & 0xff;
6372 * Check if we are talking to supported older binaries
6373 * and translate zfs_cmd if necessary
6375 if (len != sizeof(zfs_iocparm_t)) {
6382 case sizeof(zfs_cmd_zcmd_t):
6383 cflag = ZFS_CMD_COMPAT_LZC;
6385 case sizeof(zfs_cmd_deadman_t):
6386 cflag = ZFS_CMD_COMPAT_DEADMAN;
6388 case sizeof(zfs_cmd_v28_t):
6389 cflag = ZFS_CMD_COMPAT_V28;
6391 case sizeof(zfs_cmd_v15_t):
6392 cflag = ZFS_CMD_COMPAT_V15;
6393 vecnum = zfs_ioctl_v15_to_v28[cmd];
6396 * Return without further handling
6397 * if the command is blacklisted.
6399 if (vecnum == ZFS_IOC_COMPAT_PASS)
6401 else if (vecnum == ZFS_IOC_COMPAT_FAIL)
6410 vecnum = cmd - ZFS_IOC_FIRST;
6411 ASSERT3U(getmajor(dev), ==, ddi_driver_major(zfs_dip));
6414 if (vecnum >= sizeof (zfs_ioc_vec) / sizeof (zfs_ioc_vec[0]))
6415 return (SET_ERROR(EINVAL));
6416 vec = &zfs_ioc_vec[vecnum];
6418 zc = kmem_zalloc(sizeof(zfs_cmd_t), KM_SLEEP);
6421 error = ddi_copyin((void *)arg, zc, sizeof (zfs_cmd_t), flag);
6423 error = SET_ERROR(EFAULT);
6426 #else /* !illumos */
6427 bzero(zc, sizeof(zfs_cmd_t));
6430 zc_iocparm = (void *)arg;
6432 switch (zc_iocparm->zfs_ioctl_version) {
6433 case ZFS_IOCVER_CURRENT:
6434 if (zc_iocparm->zfs_cmd_size != sizeof(zfs_cmd_t)) {
6435 error = SET_ERROR(EINVAL);
6439 case ZFS_IOCVER_INLANES:
6440 if (zc_iocparm->zfs_cmd_size != sizeof(zfs_cmd_inlanes_t)) {
6441 error = SET_ERROR(EFAULT);
6445 cflag = ZFS_CMD_COMPAT_INLANES;
6447 case ZFS_IOCVER_RESUME:
6448 if (zc_iocparm->zfs_cmd_size != sizeof(zfs_cmd_resume_t)) {
6449 error = SET_ERROR(EFAULT);
6453 cflag = ZFS_CMD_COMPAT_RESUME;
6455 case ZFS_IOCVER_EDBP:
6456 if (zc_iocparm->zfs_cmd_size != sizeof(zfs_cmd_edbp_t)) {
6457 error = SET_ERROR(EFAULT);
6461 cflag = ZFS_CMD_COMPAT_EDBP;
6463 case ZFS_IOCVER_ZCMD:
6464 if (zc_iocparm->zfs_cmd_size > sizeof(zfs_cmd_t) ||
6465 zc_iocparm->zfs_cmd_size < sizeof(zfs_cmd_zcmd_t)) {
6466 error = SET_ERROR(EFAULT);
6470 cflag = ZFS_CMD_COMPAT_ZCMD;
6473 error = SET_ERROR(EINVAL);
6479 ASSERT(sizeof(zfs_cmd_t) >= zc_iocparm->zfs_cmd_size);
6480 compat_zc = kmem_zalloc(sizeof(zfs_cmd_t), KM_SLEEP);
6481 bzero(compat_zc, sizeof(zfs_cmd_t));
6483 error = ddi_copyin((void *)(uintptr_t)zc_iocparm->zfs_cmd,
6484 compat_zc, zc_iocparm->zfs_cmd_size, flag);
6486 error = SET_ERROR(EFAULT);
6490 error = ddi_copyin((void *)(uintptr_t)zc_iocparm->zfs_cmd,
6491 zc, zc_iocparm->zfs_cmd_size, flag);
6493 error = SET_ERROR(EFAULT);
6501 ASSERT(compat_zc != NULL);
6502 zfs_cmd_compat_get(zc, compat_zc, cflag);
6504 ASSERT(compat_zc == NULL);
6505 zfs_cmd_compat_get(zc, arg, cflag);
6508 error = zfs_ioctl_compat_pre(zc, &vecnum, cflag);
6511 if (oldvecnum != vecnum)
6512 vec = &zfs_ioc_vec[vecnum];
6514 #endif /* !illumos */
6516 zc->zc_iflags = flag & FKIOCTL;
6517 if (zc->zc_nvlist_src_size != 0) {
6518 error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
6519 zc->zc_iflags, &innvl);
6524 /* rewrite innvl for backwards compatibility */
6526 innvl = zfs_ioctl_compat_innvl(zc, innvl, vecnum, cflag);
6529 * Ensure that all pool/dataset names are valid before we pass down to
6532 zc->zc_name[sizeof (zc->zc_name) - 1] = '\0';
6533 switch (vec->zvec_namecheck) {
6535 if (pool_namecheck(zc->zc_name, NULL, NULL) != 0)
6536 error = SET_ERROR(EINVAL);
6538 error = pool_status_check(zc->zc_name,
6539 vec->zvec_namecheck, vec->zvec_pool_check);
6543 if (dataset_namecheck(zc->zc_name, NULL, NULL) != 0)
6544 error = SET_ERROR(EINVAL);
6546 error = pool_status_check(zc->zc_name,
6547 vec->zvec_namecheck, vec->zvec_pool_check);
6555 error = vec->zvec_secpolicy(zc, innvl, cr);
6560 /* legacy ioctls can modify zc_name */
6561 len = strcspn(zc->zc_name, "/@#") + 1;
6562 saved_poolname = kmem_alloc(len, KM_SLEEP);
6563 (void) strlcpy(saved_poolname, zc->zc_name, len);
6565 if (vec->zvec_func != NULL) {
6569 nvlist_t *lognv = NULL;
6571 ASSERT(vec->zvec_legacy_func == NULL);
6574 * Add the innvl to the lognv before calling the func,
6575 * in case the func changes the innvl.
6577 if (vec->zvec_allow_log) {
6578 lognv = fnvlist_alloc();
6579 fnvlist_add_string(lognv, ZPOOL_HIST_IOCTL,
6581 if (!nvlist_empty(innvl)) {
6582 fnvlist_add_nvlist(lognv, ZPOOL_HIST_INPUT_NVL,
6587 outnvl = fnvlist_alloc();
6588 error = vec->zvec_func(zc->zc_name, innvl, outnvl);
6591 * Some commands can partially execute, modfiy state, and still
6592 * return an error. In these cases, attempt to record what
6596 (cmd == ZFS_IOC_CHANNEL_PROGRAM && error != EINVAL)) &&
6597 vec->zvec_allow_log &&
6598 spa_open(zc->zc_name, &spa, FTAG) == 0) {
6599 if (!nvlist_empty(outnvl)) {
6600 fnvlist_add_nvlist(lognv, ZPOOL_HIST_OUTPUT_NVL,
6604 fnvlist_add_int64(lognv, ZPOOL_HIST_ERRNO,
6607 (void) spa_history_log_nvl(spa, lognv);
6608 spa_close(spa, FTAG);
6610 fnvlist_free(lognv);
6612 /* rewrite outnvl for backwards compatibility */
6614 outnvl = zfs_ioctl_compat_outnvl(zc, outnvl, vecnum,
6617 if (!nvlist_empty(outnvl) || zc->zc_nvlist_dst_size != 0) {
6619 if (vec->zvec_smush_outnvlist) {
6620 smusherror = nvlist_smush(outnvl,
6621 zc->zc_nvlist_dst_size);
6623 if (smusherror == 0)
6624 puterror = put_nvlist(zc, outnvl);
6630 nvlist_free(outnvl);
6632 error = vec->zvec_legacy_func(zc);
6639 rc = ddi_copyout(zc, (void *)arg, sizeof (zfs_cmd_t), flag);
6640 if (error == 0 && rc != 0)
6641 error = SET_ERROR(EFAULT);
6644 zfs_ioctl_compat_post(zc, cmd, cflag);
6646 ASSERT(compat_zc != NULL);
6647 ASSERT(sizeof(zfs_cmd_t) >= zc_iocparm->zfs_cmd_size);
6649 zfs_cmd_compat_put(zc, compat_zc, vecnum, cflag);
6650 rc = ddi_copyout(compat_zc,
6651 (void *)(uintptr_t)zc_iocparm->zfs_cmd,
6652 zc_iocparm->zfs_cmd_size, flag);
6653 if (error == 0 && rc != 0)
6654 error = SET_ERROR(EFAULT);
6655 kmem_free(compat_zc, sizeof (zfs_cmd_t));
6657 zfs_cmd_compat_put(zc, arg, vecnum, cflag);
6662 rc = ddi_copyout(zc, (void *)(uintptr_t)zc_iocparm->zfs_cmd,
6663 sizeof (zfs_cmd_t), flag);
6664 if (error == 0 && rc != 0)
6665 error = SET_ERROR(EFAULT);
6668 if (error == 0 && vec->zvec_allow_log) {
6669 char *s = tsd_get(zfs_allow_log_key);
6672 (void) tsd_set(zfs_allow_log_key, saved_poolname);
6674 if (saved_poolname != NULL)
6675 strfree(saved_poolname);
6678 kmem_free(zc, sizeof (zfs_cmd_t));
6684 zfs_attach(dev_info_t *dip, ddi_attach_cmd_t cmd)
6686 if (cmd != DDI_ATTACH)
6687 return (DDI_FAILURE);
6689 if (ddi_create_minor_node(dip, "zfs", S_IFCHR, 0,
6690 DDI_PSEUDO, 0) == DDI_FAILURE)
6691 return (DDI_FAILURE);
6695 ddi_report_dev(dip);
6697 return (DDI_SUCCESS);
6701 zfs_detach(dev_info_t *dip, ddi_detach_cmd_t cmd)
6703 if (spa_busy() || zfs_busy() || zvol_busy())
6704 return (DDI_FAILURE);
6706 if (cmd != DDI_DETACH)
6707 return (DDI_FAILURE);
6711 ddi_prop_remove_all(dip);
6712 ddi_remove_minor_node(dip, NULL);
6714 return (DDI_SUCCESS);
6719 zfs_info(dev_info_t *dip, ddi_info_cmd_t infocmd, void *arg, void **result)
6722 case DDI_INFO_DEVT2DEVINFO:
6724 return (DDI_SUCCESS);
6726 case DDI_INFO_DEVT2INSTANCE:
6727 *result = (void *)0;
6728 return (DDI_SUCCESS);
6731 return (DDI_FAILURE);
6733 #endif /* illumos */
6736 * OK, so this is a little weird.
6738 * /dev/zfs is the control node, i.e. minor 0.
6739 * /dev/zvol/[r]dsk/pool/dataset are the zvols, minor > 0.
6741 * /dev/zfs has basically nothing to do except serve up ioctls,
6742 * so most of the standard driver entry points are in zvol.c.
6745 static struct cb_ops zfs_cb_ops = {
6746 zfsdev_open, /* open */
6747 zfsdev_close, /* close */
6748 zvol_strategy, /* strategy */
6750 zvol_dump, /* dump */
6751 zvol_read, /* read */
6752 zvol_write, /* write */
6753 zfsdev_ioctl, /* ioctl */
6757 nochpoll, /* poll */
6758 ddi_prop_op, /* prop_op */
6759 NULL, /* streamtab */
6760 D_NEW | D_MP | D_64BIT, /* Driver compatibility flag */
6761 CB_REV, /* version */
6762 nodev, /* async read */
6763 nodev, /* async write */
6766 static struct dev_ops zfs_dev_ops = {
6767 DEVO_REV, /* version */
6769 zfs_info, /* info */
6770 nulldev, /* identify */
6771 nulldev, /* probe */
6772 zfs_attach, /* attach */
6773 zfs_detach, /* detach */
6775 &zfs_cb_ops, /* driver operations */
6776 NULL, /* no bus operations */
6778 ddi_quiesce_not_needed, /* quiesce */
6781 static struct modldrv zfs_modldrv = {
6787 static struct modlinkage modlinkage = {
6789 (void *)&zfs_modlfs,
6790 (void *)&zfs_modldrv,
6793 #endif /* illumos */
6795 static struct cdevsw zfs_cdevsw = {
6796 .d_version = D_VERSION,
6797 .d_open = zfsdev_open,
6798 .d_ioctl = zfsdev_ioctl,
6799 .d_name = ZFS_DEV_NAME
6803 zfs_allow_log_destroy(void *arg)
6805 char *poolname = arg;
6812 zfsdev = make_dev(&zfs_cdevsw, 0x0, UID_ROOT, GID_OPERATOR, 0666,
6820 destroy_dev(zfsdev);
6823 static struct root_hold_token *zfs_root_token;
6824 struct proc *zfsproc;
6832 spa_init(FREAD | FWRITE);
6837 if ((error = mod_install(&modlinkage)) != 0) {
6844 tsd_create(&zfs_fsyncer_key, NULL);
6845 tsd_create(&rrw_tsd_key, rrw_tsd_destroy);
6846 tsd_create(&zfs_allow_log_key, zfs_allow_log_destroy);
6848 error = ldi_ident_from_mod(&modlinkage, &zfs_li);
6850 mutex_init(&zfs_share_lock, NULL, MUTEX_DEFAULT, NULL);
6860 if (spa_busy() || zfs_busy() || zvol_busy() || zio_injection_enabled)
6861 return (SET_ERROR(EBUSY));
6863 if ((error = mod_remove(&modlinkage)) != 0)
6869 if (zfs_nfsshare_inited)
6870 (void) ddi_modclose(nfs_mod);
6871 if (zfs_smbshare_inited)
6872 (void) ddi_modclose(smbsrv_mod);
6873 if (zfs_nfsshare_inited || zfs_smbshare_inited)
6874 (void) ddi_modclose(sharefs_mod);
6876 tsd_destroy(&zfs_fsyncer_key);
6877 ldi_ident_release(zfs_li);
6879 mutex_destroy(&zfs_share_lock);
6885 _info(struct modinfo *modinfop)
6887 return (mod_info(&modlinkage, modinfop));
6889 #endif /* illumos */
6891 static int zfs__init(void);
6892 static int zfs__fini(void);
6893 static void zfs_shutdown(void *, int);
6895 static eventhandler_tag zfs_shutdown_event_tag;
6898 #define ZFS_MIN_KSTACK_PAGES 4
6906 #if KSTACK_PAGES < ZFS_MIN_KSTACK_PAGES
6907 printf("ZFS NOTICE: KSTACK_PAGES is %d which could result in stack "
6908 "overflow panic!\nPlease consider adding "
6909 "'options KSTACK_PAGES=%d' to your kernel config\n", KSTACK_PAGES,
6910 ZFS_MIN_KSTACK_PAGES);
6913 zfs_root_token = root_mount_hold("ZFS");
6915 mutex_init(&zfs_share_lock, NULL, MUTEX_DEFAULT, NULL);
6917 spa_init(FREAD | FWRITE);
6922 tsd_create(&zfs_fsyncer_key, NULL);
6923 tsd_create(&rrw_tsd_key, rrw_tsd_destroy);
6924 tsd_create(&zfs_allow_log_key, zfs_allow_log_destroy);
6925 tsd_create(&zfs_geom_probe_vdev_key, NULL);
6927 printf("ZFS storage pool version: features support (" SPA_VERSION_STRING ")\n");
6928 root_mount_rel(zfs_root_token);
6938 if (spa_busy() || zfs_busy() || zvol_busy() ||
6939 zio_injection_enabled) {
6948 tsd_destroy(&zfs_fsyncer_key);
6949 tsd_destroy(&rrw_tsd_key);
6950 tsd_destroy(&zfs_allow_log_key);
6952 mutex_destroy(&zfs_share_lock);
6958 zfs_shutdown(void *arg __unused, int howto __unused)
6962 * ZFS fini routines can not properly work in a panic-ed system.
6964 if (panicstr == NULL)
6970 zfs_modevent(module_t mod, int type, void *unused __unused)
6978 zfs_shutdown_event_tag = EVENTHANDLER_REGISTER(
6979 shutdown_post_sync, zfs_shutdown, NULL,
6980 SHUTDOWN_PRI_FIRST);
6984 if (err == 0 && zfs_shutdown_event_tag != NULL)
6985 EVENTHANDLER_DEREGISTER(shutdown_post_sync,
6986 zfs_shutdown_event_tag);
6993 return (EOPNOTSUPP);
6996 static moduledata_t zfs_mod = {
7001 DECLARE_MODULE(zfsctrl, zfs_mod, SI_SUB_VFS, SI_ORDER_ANY);
7002 MODULE_VERSION(zfsctrl, 1);
7003 MODULE_DEPEND(zfsctrl, opensolaris, 1, 1, 1);
7004 MODULE_DEPEND(zfsctrl, krpc, 1, 1, 1);
7005 MODULE_DEPEND(zfsctrl, acl_nfs4, 1, 1, 1);