zfs_ioctl.c revision 264729
1/*
2 * CDDL HEADER START
3 *
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.
7 *
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.
12 *
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]
18 *
19 * CDDL HEADER END
20 */
21
22/*
23 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
24 * Copyright (c) 2011-2012 Pawel Jakub Dawidek <pawel@dawidek.net>.
25 * All rights reserved.
26 * Copyright 2013 Martin Matuska <mm@FreeBSD.org>. All rights reserved.
27 * Copyright 2011 Nexenta Systems, Inc.  All rights reserved.
28 * Copyright (c) 2012, Joyent, Inc. All rights reserved.
29 * Copyright (c) 2013 by Delphix. All rights reserved.
30 * Copyright (c) 2013 by Saso Kiselkov. All rights reserved.
31 * Copyright (c) 2013 Steven Hartland. All rights reserved.
32 */
33
34/*
35 * ZFS ioctls.
36 *
37 * This file handles the ioctls to /dev/zfs, used for configuring ZFS storage
38 * pools and filesystems, e.g. with /sbin/zfs and /sbin/zpool.
39 *
40 * There are two ways that we handle ioctls: the legacy way where almost
41 * all of the logic is in the ioctl callback, and the new way where most
42 * of the marshalling is handled in the common entry point, zfsdev_ioctl().
43 *
44 * Non-legacy ioctls should be registered by calling
45 * zfs_ioctl_register() from zfs_ioctl_init().  The ioctl is invoked
46 * from userland by lzc_ioctl().
47 *
48 * The registration arguments are as follows:
49 *
50 * const char *name
51 *   The name of the ioctl.  This is used for history logging.  If the
52 *   ioctl returns successfully (the callback returns 0), and allow_log
53 *   is true, then a history log entry will be recorded with the input &
54 *   output nvlists.  The log entry can be printed with "zpool history -i".
55 *
56 * zfs_ioc_t ioc
57 *   The ioctl request number, which userland will pass to ioctl(2).
58 *   The ioctl numbers can change from release to release, because
59 *   the caller (libzfs) must be matched to the kernel.
60 *
61 * zfs_secpolicy_func_t *secpolicy
62 *   This function will be called before the zfs_ioc_func_t, to
63 *   determine if this operation is permitted.  It should return EPERM
64 *   on failure, and 0 on success.  Checks include determining if the
65 *   dataset is visible in this zone, and if the user has either all
66 *   zfs privileges in the zone (SYS_MOUNT), or has been granted permission
67 *   to do this operation on this dataset with "zfs allow".
68 *
69 * zfs_ioc_namecheck_t namecheck
70 *   This specifies what to expect in the zfs_cmd_t:zc_name -- a pool
71 *   name, a dataset name, or nothing.  If the name is not well-formed,
72 *   the ioctl will fail and the callback will not be called.
73 *   Therefore, the callback can assume that the name is well-formed
74 *   (e.g. is null-terminated, doesn't have more than one '@' character,
75 *   doesn't have invalid characters).
76 *
77 * zfs_ioc_poolcheck_t pool_check
78 *   This specifies requirements on the pool state.  If the pool does
79 *   not meet them (is suspended or is readonly), the ioctl will fail
80 *   and the callback will not be called.  If any checks are specified
81 *   (i.e. it is not POOL_CHECK_NONE), namecheck must not be NO_NAME.
82 *   Multiple checks can be or-ed together (e.g. POOL_CHECK_SUSPENDED |
83 *   POOL_CHECK_READONLY).
84 *
85 * boolean_t smush_outnvlist
86 *   If smush_outnvlist is true, then the output is presumed to be a
87 *   list of errors, and it will be "smushed" down to fit into the
88 *   caller's buffer, by removing some entries and replacing them with a
89 *   single "N_MORE_ERRORS" entry indicating how many were removed.  See
90 *   nvlist_smush() for details.  If smush_outnvlist is false, and the
91 *   outnvlist does not fit into the userland-provided buffer, then the
92 *   ioctl will fail with ENOMEM.
93 *
94 * zfs_ioc_func_t *func
95 *   The callback function that will perform the operation.
96 *
97 *   The callback should return 0 on success, or an error number on
98 *   failure.  If the function fails, the userland ioctl will return -1,
99 *   and errno will be set to the callback's return value.  The callback
100 *   will be called with the following arguments:
101 *
102 *   const char *name
103 *     The name of the pool or dataset to operate on, from
104 *     zfs_cmd_t:zc_name.  The 'namecheck' argument specifies the
105 *     expected type (pool, dataset, or none).
106 *
107 *   nvlist_t *innvl
108 *     The input nvlist, deserialized from zfs_cmd_t:zc_nvlist_src.  Or
109 *     NULL if no input nvlist was provided.  Changes to this nvlist are
110 *     ignored.  If the input nvlist could not be deserialized, the
111 *     ioctl will fail and the callback will not be called.
112 *
113 *   nvlist_t *outnvl
114 *     The output nvlist, initially empty.  The callback can fill it in,
115 *     and it will be returned to userland by serializing it into
116 *     zfs_cmd_t:zc_nvlist_dst.  If it is non-empty, and serialization
117 *     fails (e.g. because the caller didn't supply a large enough
118 *     buffer), then the overall ioctl will fail.  See the
119 *     'smush_nvlist' argument above for additional behaviors.
120 *
121 *     There are two typical uses of the output nvlist:
122 *       - To return state, e.g. property values.  In this case,
123 *         smush_outnvlist should be false.  If the buffer was not large
124 *         enough, the caller will reallocate a larger buffer and try
125 *         the ioctl again.
126 *
127 *       - To return multiple errors from an ioctl which makes on-disk
128 *         changes.  In this case, smush_outnvlist should be true.
129 *         Ioctls which make on-disk modifications should generally not
130 *         use the outnvl if they succeed, because the caller can not
131 *         distinguish between the operation failing, and
132 *         deserialization failing.
133 */
134
135#include <sys/types.h>
136#include <sys/param.h>
137#include <sys/systm.h>
138#include <sys/conf.h>
139#include <sys/kernel.h>
140#include <sys/lock.h>
141#include <sys/malloc.h>
142#include <sys/mutex.h>
143#include <sys/proc.h>
144#include <sys/errno.h>
145#include <sys/uio.h>
146#include <sys/buf.h>
147#include <sys/file.h>
148#include <sys/kmem.h>
149#include <sys/conf.h>
150#include <sys/cmn_err.h>
151#include <sys/stat.h>
152#include <sys/zfs_ioctl.h>
153#include <sys/zfs_vfsops.h>
154#include <sys/zfs_znode.h>
155#include <sys/zap.h>
156#include <sys/spa.h>
157#include <sys/spa_impl.h>
158#include <sys/vdev.h>
159#include <sys/dmu.h>
160#include <sys/dsl_dir.h>
161#include <sys/dsl_dataset.h>
162#include <sys/dsl_prop.h>
163#include <sys/dsl_deleg.h>
164#include <sys/dmu_objset.h>
165#include <sys/dmu_impl.h>
166#include <sys/dmu_tx.h>
167#include <sys/sunddi.h>
168#include <sys/policy.h>
169#include <sys/zone.h>
170#include <sys/nvpair.h>
171#include <sys/mount.h>
172#include <sys/taskqueue.h>
173#include <sys/sdt.h>
174#include <sys/varargs.h>
175#include <sys/fs/zfs.h>
176#include <sys/zfs_ctldir.h>
177#include <sys/zfs_dir.h>
178#include <sys/zfs_onexit.h>
179#include <sys/zvol.h>
180#include <sys/dsl_scan.h>
181#include <sys/dmu_objset.h>
182#include <sys/dmu_send.h>
183#include <sys/dsl_destroy.h>
184#include <sys/dsl_bookmark.h>
185#include <sys/dsl_userhold.h>
186#include <sys/zfeature.h>
187
188#include "zfs_namecheck.h"
189#include "zfs_prop.h"
190#include "zfs_deleg.h"
191#include "zfs_comutil.h"
192#include "zfs_ioctl_compat.h"
193
194CTASSERT(sizeof(zfs_cmd_t) < IOCPARM_MAX);
195
196static int snapshot_list_prefetch;
197SYSCTL_DECL(_vfs_zfs);
198TUNABLE_INT("vfs.zfs.snapshot_list_prefetch", &snapshot_list_prefetch);
199SYSCTL_INT(_vfs_zfs, OID_AUTO, snapshot_list_prefetch, CTLFLAG_RW,
200    &snapshot_list_prefetch, 0, "Prefetch data when listing snapshots");
201
202static struct cdev *zfsdev;
203
204extern void zfs_init(void);
205extern void zfs_fini(void);
206
207uint_t zfs_fsyncer_key;
208extern uint_t rrw_tsd_key;
209static uint_t zfs_allow_log_key;
210
211typedef int zfs_ioc_legacy_func_t(zfs_cmd_t *);
212typedef int zfs_ioc_func_t(const char *, nvlist_t *, nvlist_t *);
213typedef int zfs_secpolicy_func_t(zfs_cmd_t *, nvlist_t *, cred_t *);
214
215typedef enum {
216	NO_NAME,
217	POOL_NAME,
218	DATASET_NAME
219} zfs_ioc_namecheck_t;
220
221typedef enum {
222	POOL_CHECK_NONE		= 1 << 0,
223	POOL_CHECK_SUSPENDED	= 1 << 1,
224	POOL_CHECK_READONLY	= 1 << 2,
225} zfs_ioc_poolcheck_t;
226
227typedef struct zfs_ioc_vec {
228	zfs_ioc_legacy_func_t	*zvec_legacy_func;
229	zfs_ioc_func_t		*zvec_func;
230	zfs_secpolicy_func_t	*zvec_secpolicy;
231	zfs_ioc_namecheck_t	zvec_namecheck;
232	boolean_t		zvec_allow_log;
233	zfs_ioc_poolcheck_t	zvec_pool_check;
234	boolean_t		zvec_smush_outnvlist;
235	const char		*zvec_name;
236} zfs_ioc_vec_t;
237
238/* This array is indexed by zfs_userquota_prop_t */
239static const char *userquota_perms[] = {
240	ZFS_DELEG_PERM_USERUSED,
241	ZFS_DELEG_PERM_USERQUOTA,
242	ZFS_DELEG_PERM_GROUPUSED,
243	ZFS_DELEG_PERM_GROUPQUOTA,
244};
245
246static int zfs_ioc_userspace_upgrade(zfs_cmd_t *zc);
247static int zfs_check_settable(const char *name, nvpair_t *property,
248    cred_t *cr);
249static int zfs_check_clearable(char *dataset, nvlist_t *props,
250    nvlist_t **errors);
251static int zfs_fill_zplprops_root(uint64_t, nvlist_t *, nvlist_t *,
252    boolean_t *);
253int zfs_set_prop_nvlist(const char *, zprop_source_t, nvlist_t *, nvlist_t *);
254static int get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp);
255
256static void zfsdev_close(void *data);
257
258static int zfs_prop_activate_feature(spa_t *spa, spa_feature_t feature);
259
260/* _NOTE(PRINTFLIKE(4)) - this is printf-like, but lint is too whiney */
261void
262__dprintf(const char *file, const char *func, int line, const char *fmt, ...)
263{
264	const char *newfile;
265	char buf[512];
266	va_list adx;
267
268	/*
269	 * Get rid of annoying "../common/" prefix to filename.
270	 */
271	newfile = strrchr(file, '/');
272	if (newfile != NULL) {
273		newfile = newfile + 1; /* Get rid of leading / */
274	} else {
275		newfile = file;
276	}
277
278	va_start(adx, fmt);
279	(void) vsnprintf(buf, sizeof (buf), fmt, adx);
280	va_end(adx);
281
282	/*
283	 * To get this data, use the zfs-dprintf probe as so:
284	 * dtrace -q -n 'zfs-dprintf \
285	 *	/stringof(arg0) == "dbuf.c"/ \
286	 *	{printf("%s: %s", stringof(arg1), stringof(arg3))}'
287	 * arg0 = file name
288	 * arg1 = function name
289	 * arg2 = line number
290	 * arg3 = message
291	 */
292	DTRACE_PROBE4(zfs__dprintf,
293	    char *, newfile, char *, func, int, line, char *, buf);
294}
295
296static void
297history_str_free(char *buf)
298{
299	kmem_free(buf, HIS_MAX_RECORD_LEN);
300}
301
302static char *
303history_str_get(zfs_cmd_t *zc)
304{
305	char *buf;
306
307	if (zc->zc_history == 0)
308		return (NULL);
309
310	buf = kmem_alloc(HIS_MAX_RECORD_LEN, KM_SLEEP);
311	if (copyinstr((void *)(uintptr_t)zc->zc_history,
312	    buf, HIS_MAX_RECORD_LEN, NULL) != 0) {
313		history_str_free(buf);
314		return (NULL);
315	}
316
317	buf[HIS_MAX_RECORD_LEN -1] = '\0';
318
319	return (buf);
320}
321
322/*
323 * Check to see if the named dataset is currently defined as bootable
324 */
325static boolean_t
326zfs_is_bootfs(const char *name)
327{
328	objset_t *os;
329
330	if (dmu_objset_hold(name, FTAG, &os) == 0) {
331		boolean_t ret;
332		ret = (dmu_objset_id(os) == spa_bootfs(dmu_objset_spa(os)));
333		dmu_objset_rele(os, FTAG);
334		return (ret);
335	}
336	return (B_FALSE);
337}
338
339/*
340 * Return non-zero if the spa version is less than requested version.
341 */
342static int
343zfs_earlier_version(const char *name, int version)
344{
345	spa_t *spa;
346
347	if (spa_open(name, &spa, FTAG) == 0) {
348		if (spa_version(spa) < version) {
349			spa_close(spa, FTAG);
350			return (1);
351		}
352		spa_close(spa, FTAG);
353	}
354	return (0);
355}
356
357/*
358 * Return TRUE if the ZPL version is less than requested version.
359 */
360static boolean_t
361zpl_earlier_version(const char *name, int version)
362{
363	objset_t *os;
364	boolean_t rc = B_TRUE;
365
366	if (dmu_objset_hold(name, FTAG, &os) == 0) {
367		uint64_t zplversion;
368
369		if (dmu_objset_type(os) != DMU_OST_ZFS) {
370			dmu_objset_rele(os, FTAG);
371			return (B_TRUE);
372		}
373		/* XXX reading from non-owned objset */
374		if (zfs_get_zplprop(os, ZFS_PROP_VERSION, &zplversion) == 0)
375			rc = zplversion < version;
376		dmu_objset_rele(os, FTAG);
377	}
378	return (rc);
379}
380
381static void
382zfs_log_history(zfs_cmd_t *zc)
383{
384	spa_t *spa;
385	char *buf;
386
387	if ((buf = history_str_get(zc)) == NULL)
388		return;
389
390	if (spa_open(zc->zc_name, &spa, FTAG) == 0) {
391		if (spa_version(spa) >= SPA_VERSION_ZPOOL_HISTORY)
392			(void) spa_history_log(spa, buf);
393		spa_close(spa, FTAG);
394	}
395	history_str_free(buf);
396}
397
398/*
399 * Policy for top-level read operations (list pools).  Requires no privileges,
400 * and can be used in the local zone, as there is no associated dataset.
401 */
402/* ARGSUSED */
403static int
404zfs_secpolicy_none(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
405{
406	return (0);
407}
408
409/*
410 * Policy for dataset read operations (list children, get statistics).  Requires
411 * no privileges, but must be visible in the local zone.
412 */
413/* ARGSUSED */
414static int
415zfs_secpolicy_read(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
416{
417	if (INGLOBALZONE(curthread) ||
418	    zone_dataset_visible(zc->zc_name, NULL))
419		return (0);
420
421	return (SET_ERROR(ENOENT));
422}
423
424static int
425zfs_dozonecheck_impl(const char *dataset, uint64_t zoned, cred_t *cr)
426{
427	int writable = 1;
428
429	/*
430	 * The dataset must be visible by this zone -- check this first
431	 * so they don't see EPERM on something they shouldn't know about.
432	 */
433	if (!INGLOBALZONE(curthread) &&
434	    !zone_dataset_visible(dataset, &writable))
435		return (SET_ERROR(ENOENT));
436
437	if (INGLOBALZONE(curthread)) {
438		/*
439		 * If the fs is zoned, only root can access it from the
440		 * global zone.
441		 */
442		if (secpolicy_zfs(cr) && zoned)
443			return (SET_ERROR(EPERM));
444	} else {
445		/*
446		 * If we are in a local zone, the 'zoned' property must be set.
447		 */
448		if (!zoned)
449			return (SET_ERROR(EPERM));
450
451		/* must be writable by this zone */
452		if (!writable)
453			return (SET_ERROR(EPERM));
454	}
455	return (0);
456}
457
458static int
459zfs_dozonecheck(const char *dataset, cred_t *cr)
460{
461	uint64_t zoned;
462
463	if (dsl_prop_get_integer(dataset, "jailed", &zoned, NULL))
464		return (SET_ERROR(ENOENT));
465
466	return (zfs_dozonecheck_impl(dataset, zoned, cr));
467}
468
469static int
470zfs_dozonecheck_ds(const char *dataset, dsl_dataset_t *ds, cred_t *cr)
471{
472	uint64_t zoned;
473
474	if (dsl_prop_get_int_ds(ds, "jailed", &zoned))
475		return (SET_ERROR(ENOENT));
476
477	return (zfs_dozonecheck_impl(dataset, zoned, cr));
478}
479
480static int
481zfs_secpolicy_write_perms_ds(const char *name, dsl_dataset_t *ds,
482    const char *perm, cred_t *cr)
483{
484	int error;
485
486	error = zfs_dozonecheck_ds(name, ds, cr);
487	if (error == 0) {
488		error = secpolicy_zfs(cr);
489		if (error != 0)
490			error = dsl_deleg_access_impl(ds, perm, cr);
491	}
492	return (error);
493}
494
495static int
496zfs_secpolicy_write_perms(const char *name, const char *perm, cred_t *cr)
497{
498	int error;
499	dsl_dataset_t *ds;
500	dsl_pool_t *dp;
501
502	error = dsl_pool_hold(name, FTAG, &dp);
503	if (error != 0)
504		return (error);
505
506	error = dsl_dataset_hold(dp, name, FTAG, &ds);
507	if (error != 0) {
508		dsl_pool_rele(dp, FTAG);
509		return (error);
510	}
511
512	error = zfs_secpolicy_write_perms_ds(name, ds, perm, cr);
513
514	dsl_dataset_rele(ds, FTAG);
515	dsl_pool_rele(dp, FTAG);
516	return (error);
517}
518
519#ifdef SECLABEL
520/*
521 * Policy for setting the security label property.
522 *
523 * Returns 0 for success, non-zero for access and other errors.
524 */
525static int
526zfs_set_slabel_policy(const char *name, char *strval, cred_t *cr)
527{
528	char		ds_hexsl[MAXNAMELEN];
529	bslabel_t	ds_sl, new_sl;
530	boolean_t	new_default = FALSE;
531	uint64_t	zoned;
532	int		needed_priv = -1;
533	int		error;
534
535	/* First get the existing dataset label. */
536	error = dsl_prop_get(name, zfs_prop_to_name(ZFS_PROP_MLSLABEL),
537	    1, sizeof (ds_hexsl), &ds_hexsl, NULL);
538	if (error != 0)
539		return (SET_ERROR(EPERM));
540
541	if (strcasecmp(strval, ZFS_MLSLABEL_DEFAULT) == 0)
542		new_default = TRUE;
543
544	/* The label must be translatable */
545	if (!new_default && (hexstr_to_label(strval, &new_sl) != 0))
546		return (SET_ERROR(EINVAL));
547
548	/*
549	 * In a non-global zone, disallow attempts to set a label that
550	 * doesn't match that of the zone; otherwise no other checks
551	 * are needed.
552	 */
553	if (!INGLOBALZONE(curproc)) {
554		if (new_default || !blequal(&new_sl, CR_SL(CRED())))
555			return (SET_ERROR(EPERM));
556		return (0);
557	}
558
559	/*
560	 * For global-zone datasets (i.e., those whose zoned property is
561	 * "off", verify that the specified new label is valid for the
562	 * global zone.
563	 */
564	if (dsl_prop_get_integer(name,
565	    zfs_prop_to_name(ZFS_PROP_ZONED), &zoned, NULL))
566		return (SET_ERROR(EPERM));
567	if (!zoned) {
568		if (zfs_check_global_label(name, strval) != 0)
569			return (SET_ERROR(EPERM));
570	}
571
572	/*
573	 * If the existing dataset label is nondefault, check if the
574	 * dataset is mounted (label cannot be changed while mounted).
575	 * Get the zfsvfs; if there isn't one, then the dataset isn't
576	 * mounted (or isn't a dataset, doesn't exist, ...).
577	 */
578	if (strcasecmp(ds_hexsl, ZFS_MLSLABEL_DEFAULT) != 0) {
579		objset_t *os;
580		static char *setsl_tag = "setsl_tag";
581
582		/*
583		 * Try to own the dataset; abort if there is any error,
584		 * (e.g., already mounted, in use, or other error).
585		 */
586		error = dmu_objset_own(name, DMU_OST_ZFS, B_TRUE,
587		    setsl_tag, &os);
588		if (error != 0)
589			return (SET_ERROR(EPERM));
590
591		dmu_objset_disown(os, setsl_tag);
592
593		if (new_default) {
594			needed_priv = PRIV_FILE_DOWNGRADE_SL;
595			goto out_check;
596		}
597
598		if (hexstr_to_label(strval, &new_sl) != 0)
599			return (SET_ERROR(EPERM));
600
601		if (blstrictdom(&ds_sl, &new_sl))
602			needed_priv = PRIV_FILE_DOWNGRADE_SL;
603		else if (blstrictdom(&new_sl, &ds_sl))
604			needed_priv = PRIV_FILE_UPGRADE_SL;
605	} else {
606		/* dataset currently has a default label */
607		if (!new_default)
608			needed_priv = PRIV_FILE_UPGRADE_SL;
609	}
610
611out_check:
612	if (needed_priv != -1)
613		return (PRIV_POLICY(cr, needed_priv, B_FALSE, EPERM, NULL));
614	return (0);
615}
616#endif	/* SECLABEL */
617
618static int
619zfs_secpolicy_setprop(const char *dsname, zfs_prop_t prop, nvpair_t *propval,
620    cred_t *cr)
621{
622	char *strval;
623
624	/*
625	 * Check permissions for special properties.
626	 */
627	switch (prop) {
628	case ZFS_PROP_ZONED:
629		/*
630		 * Disallow setting of 'zoned' from within a local zone.
631		 */
632		if (!INGLOBALZONE(curthread))
633			return (SET_ERROR(EPERM));
634		break;
635
636	case ZFS_PROP_QUOTA:
637		if (!INGLOBALZONE(curthread)) {
638			uint64_t zoned;
639			char setpoint[MAXNAMELEN];
640			/*
641			 * Unprivileged users are allowed to modify the
642			 * quota on things *under* (ie. contained by)
643			 * the thing they own.
644			 */
645			if (dsl_prop_get_integer(dsname, "jailed", &zoned,
646			    setpoint))
647				return (SET_ERROR(EPERM));
648			if (!zoned || strlen(dsname) <= strlen(setpoint))
649				return (SET_ERROR(EPERM));
650		}
651		break;
652
653	case ZFS_PROP_MLSLABEL:
654#ifdef SECLABEL
655		if (!is_system_labeled())
656			return (SET_ERROR(EPERM));
657
658		if (nvpair_value_string(propval, &strval) == 0) {
659			int err;
660
661			err = zfs_set_slabel_policy(dsname, strval, CRED());
662			if (err != 0)
663				return (err);
664		}
665#else
666		return (EOPNOTSUPP);
667#endif
668		break;
669	}
670
671	return (zfs_secpolicy_write_perms(dsname, zfs_prop_to_name(prop), cr));
672}
673
674/* ARGSUSED */
675static int
676zfs_secpolicy_set_fsacl(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
677{
678	int error;
679
680	error = zfs_dozonecheck(zc->zc_name, cr);
681	if (error != 0)
682		return (error);
683
684	/*
685	 * permission to set permissions will be evaluated later in
686	 * dsl_deleg_can_allow()
687	 */
688	return (0);
689}
690
691/* ARGSUSED */
692static int
693zfs_secpolicy_rollback(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
694{
695	return (zfs_secpolicy_write_perms(zc->zc_name,
696	    ZFS_DELEG_PERM_ROLLBACK, cr));
697}
698
699/* ARGSUSED */
700static int
701zfs_secpolicy_send(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
702{
703	dsl_pool_t *dp;
704	dsl_dataset_t *ds;
705	char *cp;
706	int error;
707
708	/*
709	 * Generate the current snapshot name from the given objsetid, then
710	 * use that name for the secpolicy/zone checks.
711	 */
712	cp = strchr(zc->zc_name, '@');
713	if (cp == NULL)
714		return (SET_ERROR(EINVAL));
715	error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
716	if (error != 0)
717		return (error);
718
719	error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &ds);
720	if (error != 0) {
721		dsl_pool_rele(dp, FTAG);
722		return (error);
723	}
724
725	dsl_dataset_name(ds, zc->zc_name);
726
727	error = zfs_secpolicy_write_perms_ds(zc->zc_name, ds,
728	    ZFS_DELEG_PERM_SEND, cr);
729	dsl_dataset_rele(ds, FTAG);
730	dsl_pool_rele(dp, FTAG);
731
732	return (error);
733}
734
735/* ARGSUSED */
736static int
737zfs_secpolicy_send_new(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
738{
739	return (zfs_secpolicy_write_perms(zc->zc_name,
740	    ZFS_DELEG_PERM_SEND, cr));
741}
742
743/* ARGSUSED */
744static int
745zfs_secpolicy_deleg_share(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
746{
747	vnode_t *vp;
748	int error;
749
750	if ((error = lookupname(zc->zc_value, UIO_SYSSPACE,
751	    NO_FOLLOW, NULL, &vp)) != 0)
752		return (error);
753
754	/* Now make sure mntpnt and dataset are ZFS */
755
756	if (strcmp(vp->v_vfsp->mnt_stat.f_fstypename, "zfs") != 0 ||
757	    (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource),
758	    zc->zc_name) != 0)) {
759		VN_RELE(vp);
760		return (SET_ERROR(EPERM));
761	}
762
763	VN_RELE(vp);
764	return (dsl_deleg_access(zc->zc_name,
765	    ZFS_DELEG_PERM_SHARE, cr));
766}
767
768int
769zfs_secpolicy_share(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
770{
771	if (!INGLOBALZONE(curthread))
772		return (SET_ERROR(EPERM));
773
774	if (secpolicy_nfs(cr) == 0) {
775		return (0);
776	} else {
777		return (zfs_secpolicy_deleg_share(zc, innvl, cr));
778	}
779}
780
781int
782zfs_secpolicy_smb_acl(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
783{
784	if (!INGLOBALZONE(curthread))
785		return (SET_ERROR(EPERM));
786
787	if (secpolicy_smb(cr) == 0) {
788		return (0);
789	} else {
790		return (zfs_secpolicy_deleg_share(zc, innvl, cr));
791	}
792}
793
794static int
795zfs_get_parent(const char *datasetname, char *parent, int parentsize)
796{
797	char *cp;
798
799	/*
800	 * Remove the @bla or /bla from the end of the name to get the parent.
801	 */
802	(void) strncpy(parent, datasetname, parentsize);
803	cp = strrchr(parent, '@');
804	if (cp != NULL) {
805		cp[0] = '\0';
806	} else {
807		cp = strrchr(parent, '/');
808		if (cp == NULL)
809			return (SET_ERROR(ENOENT));
810		cp[0] = '\0';
811	}
812
813	return (0);
814}
815
816int
817zfs_secpolicy_destroy_perms(const char *name, cred_t *cr)
818{
819	int error;
820
821	if ((error = zfs_secpolicy_write_perms(name,
822	    ZFS_DELEG_PERM_MOUNT, cr)) != 0)
823		return (error);
824
825	return (zfs_secpolicy_write_perms(name, ZFS_DELEG_PERM_DESTROY, cr));
826}
827
828/* ARGSUSED */
829static int
830zfs_secpolicy_destroy(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
831{
832	return (zfs_secpolicy_destroy_perms(zc->zc_name, cr));
833}
834
835/*
836 * Destroying snapshots with delegated permissions requires
837 * descendant mount and destroy permissions.
838 */
839/* ARGSUSED */
840static int
841zfs_secpolicy_destroy_snaps(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
842{
843	nvlist_t *snaps;
844	nvpair_t *pair, *nextpair;
845	int error = 0;
846
847	if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
848		return (SET_ERROR(EINVAL));
849	for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
850	    pair = nextpair) {
851		nextpair = nvlist_next_nvpair(snaps, pair);
852		error = zfs_secpolicy_destroy_perms(nvpair_name(pair), cr);
853		if (error == ENOENT) {
854			/*
855			 * Ignore any snapshots that don't exist (we consider
856			 * them "already destroyed").  Remove the name from the
857			 * nvl here in case the snapshot is created between
858			 * now and when we try to destroy it (in which case
859			 * we don't want to destroy it since we haven't
860			 * checked for permission).
861			 */
862			fnvlist_remove_nvpair(snaps, pair);
863			error = 0;
864		}
865		if (error != 0)
866			break;
867	}
868
869	return (error);
870}
871
872int
873zfs_secpolicy_rename_perms(const char *from, const char *to, cred_t *cr)
874{
875	char	parentname[MAXNAMELEN];
876	int	error;
877
878	if ((error = zfs_secpolicy_write_perms(from,
879	    ZFS_DELEG_PERM_RENAME, cr)) != 0)
880		return (error);
881
882	if ((error = zfs_secpolicy_write_perms(from,
883	    ZFS_DELEG_PERM_MOUNT, cr)) != 0)
884		return (error);
885
886	if ((error = zfs_get_parent(to, parentname,
887	    sizeof (parentname))) != 0)
888		return (error);
889
890	if ((error = zfs_secpolicy_write_perms(parentname,
891	    ZFS_DELEG_PERM_CREATE, cr)) != 0)
892		return (error);
893
894	if ((error = zfs_secpolicy_write_perms(parentname,
895	    ZFS_DELEG_PERM_MOUNT, cr)) != 0)
896		return (error);
897
898	return (error);
899}
900
901/* ARGSUSED */
902static int
903zfs_secpolicy_rename(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
904{
905	char *at = NULL;
906	int error;
907
908	if ((zc->zc_cookie & 1) != 0) {
909		/*
910		 * This is recursive rename, so the starting snapshot might
911		 * not exist. Check file system or volume permission instead.
912		 */
913		at = strchr(zc->zc_name, '@');
914		if (at == NULL)
915			return (EINVAL);
916		*at = '\0';
917	}
918
919	error = zfs_secpolicy_rename_perms(zc->zc_name, zc->zc_value, cr);
920
921	if (at != NULL)
922		*at = '@';
923
924	return (error);
925}
926
927/* ARGSUSED */
928static int
929zfs_secpolicy_promote(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
930{
931	dsl_pool_t *dp;
932	dsl_dataset_t *clone;
933	int error;
934
935	error = zfs_secpolicy_write_perms(zc->zc_name,
936	    ZFS_DELEG_PERM_PROMOTE, cr);
937	if (error != 0)
938		return (error);
939
940	error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
941	if (error != 0)
942		return (error);
943
944	error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &clone);
945
946	if (error == 0) {
947		char parentname[MAXNAMELEN];
948		dsl_dataset_t *origin = NULL;
949		dsl_dir_t *dd;
950		dd = clone->ds_dir;
951
952		error = dsl_dataset_hold_obj(dd->dd_pool,
953		    dd->dd_phys->dd_origin_obj, FTAG, &origin);
954		if (error != 0) {
955			dsl_dataset_rele(clone, FTAG);
956			dsl_pool_rele(dp, FTAG);
957			return (error);
958		}
959
960		error = zfs_secpolicy_write_perms_ds(zc->zc_name, clone,
961		    ZFS_DELEG_PERM_MOUNT, cr);
962
963		dsl_dataset_name(origin, parentname);
964		if (error == 0) {
965			error = zfs_secpolicy_write_perms_ds(parentname, origin,
966			    ZFS_DELEG_PERM_PROMOTE, cr);
967		}
968		dsl_dataset_rele(clone, FTAG);
969		dsl_dataset_rele(origin, FTAG);
970	}
971	dsl_pool_rele(dp, FTAG);
972	return (error);
973}
974
975/* ARGSUSED */
976static int
977zfs_secpolicy_recv(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
978{
979	int error;
980
981	if ((error = zfs_secpolicy_write_perms(zc->zc_name,
982	    ZFS_DELEG_PERM_RECEIVE, cr)) != 0)
983		return (error);
984
985	if ((error = zfs_secpolicy_write_perms(zc->zc_name,
986	    ZFS_DELEG_PERM_MOUNT, cr)) != 0)
987		return (error);
988
989	return (zfs_secpolicy_write_perms(zc->zc_name,
990	    ZFS_DELEG_PERM_CREATE, cr));
991}
992
993int
994zfs_secpolicy_snapshot_perms(const char *name, cred_t *cr)
995{
996	return (zfs_secpolicy_write_perms(name,
997	    ZFS_DELEG_PERM_SNAPSHOT, cr));
998}
999
1000/*
1001 * Check for permission to create each snapshot in the nvlist.
1002 */
1003/* ARGSUSED */
1004static int
1005zfs_secpolicy_snapshot(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1006{
1007	nvlist_t *snaps;
1008	int error;
1009	nvpair_t *pair;
1010
1011	if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
1012		return (SET_ERROR(EINVAL));
1013	for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
1014	    pair = nvlist_next_nvpair(snaps, pair)) {
1015		char *name = nvpair_name(pair);
1016		char *atp = strchr(name, '@');
1017
1018		if (atp == NULL) {
1019			error = SET_ERROR(EINVAL);
1020			break;
1021		}
1022		*atp = '\0';
1023		error = zfs_secpolicy_snapshot_perms(name, cr);
1024		*atp = '@';
1025		if (error != 0)
1026			break;
1027	}
1028	return (error);
1029}
1030
1031/*
1032 * Check for permission to create each snapshot in the nvlist.
1033 */
1034/* ARGSUSED */
1035static int
1036zfs_secpolicy_bookmark(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1037{
1038	int error = 0;
1039
1040	for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL);
1041	    pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) {
1042		char *name = nvpair_name(pair);
1043		char *hashp = strchr(name, '#');
1044
1045		if (hashp == NULL) {
1046			error = SET_ERROR(EINVAL);
1047			break;
1048		}
1049		*hashp = '\0';
1050		error = zfs_secpolicy_write_perms(name,
1051		    ZFS_DELEG_PERM_BOOKMARK, cr);
1052		*hashp = '#';
1053		if (error != 0)
1054			break;
1055	}
1056	return (error);
1057}
1058
1059/* ARGSUSED */
1060static int
1061zfs_secpolicy_destroy_bookmarks(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1062{
1063	nvpair_t *pair, *nextpair;
1064	int error = 0;
1065
1066	for (pair = nvlist_next_nvpair(innvl, NULL); pair != NULL;
1067	    pair = nextpair) {
1068		char *name = nvpair_name(pair);
1069		char *hashp = strchr(name, '#');
1070		nextpair = nvlist_next_nvpair(innvl, pair);
1071
1072		if (hashp == NULL) {
1073			error = SET_ERROR(EINVAL);
1074			break;
1075		}
1076
1077		*hashp = '\0';
1078		error = zfs_secpolicy_write_perms(name,
1079		    ZFS_DELEG_PERM_DESTROY, cr);
1080		*hashp = '#';
1081		if (error == ENOENT) {
1082			/*
1083			 * Ignore any filesystems that don't exist (we consider
1084			 * their bookmarks "already destroyed").  Remove
1085			 * the name from the nvl here in case the filesystem
1086			 * is created between now and when we try to destroy
1087			 * the bookmark (in which case we don't want to
1088			 * destroy it since we haven't checked for permission).
1089			 */
1090			fnvlist_remove_nvpair(innvl, pair);
1091			error = 0;
1092		}
1093		if (error != 0)
1094			break;
1095	}
1096
1097	return (error);
1098}
1099
1100/* ARGSUSED */
1101static int
1102zfs_secpolicy_log_history(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1103{
1104	/*
1105	 * Even root must have a proper TSD so that we know what pool
1106	 * to log to.
1107	 */
1108	if (tsd_get(zfs_allow_log_key) == NULL)
1109		return (SET_ERROR(EPERM));
1110	return (0);
1111}
1112
1113static int
1114zfs_secpolicy_create_clone(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1115{
1116	char	parentname[MAXNAMELEN];
1117	int	error;
1118	char	*origin;
1119
1120	if ((error = zfs_get_parent(zc->zc_name, parentname,
1121	    sizeof (parentname))) != 0)
1122		return (error);
1123
1124	if (nvlist_lookup_string(innvl, "origin", &origin) == 0 &&
1125	    (error = zfs_secpolicy_write_perms(origin,
1126	    ZFS_DELEG_PERM_CLONE, cr)) != 0)
1127		return (error);
1128
1129	if ((error = zfs_secpolicy_write_perms(parentname,
1130	    ZFS_DELEG_PERM_CREATE, cr)) != 0)
1131		return (error);
1132
1133	return (zfs_secpolicy_write_perms(parentname,
1134	    ZFS_DELEG_PERM_MOUNT, cr));
1135}
1136
1137/*
1138 * Policy for pool operations - create/destroy pools, add vdevs, etc.  Requires
1139 * SYS_CONFIG privilege, which is not available in a local zone.
1140 */
1141/* ARGSUSED */
1142static int
1143zfs_secpolicy_config(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1144{
1145	if (secpolicy_sys_config(cr, B_FALSE) != 0)
1146		return (SET_ERROR(EPERM));
1147
1148	return (0);
1149}
1150
1151/*
1152 * Policy for object to name lookups.
1153 */
1154/* ARGSUSED */
1155static int
1156zfs_secpolicy_diff(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1157{
1158	int error;
1159
1160	if ((error = secpolicy_sys_config(cr, B_FALSE)) == 0)
1161		return (0);
1162
1163	error = zfs_secpolicy_write_perms(zc->zc_name, ZFS_DELEG_PERM_DIFF, cr);
1164	return (error);
1165}
1166
1167/*
1168 * Policy for fault injection.  Requires all privileges.
1169 */
1170/* ARGSUSED */
1171static int
1172zfs_secpolicy_inject(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1173{
1174	return (secpolicy_zinject(cr));
1175}
1176
1177/* ARGSUSED */
1178static int
1179zfs_secpolicy_inherit_prop(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1180{
1181	zfs_prop_t prop = zfs_name_to_prop(zc->zc_value);
1182
1183	if (prop == ZPROP_INVAL) {
1184		if (!zfs_prop_user(zc->zc_value))
1185			return (SET_ERROR(EINVAL));
1186		return (zfs_secpolicy_write_perms(zc->zc_name,
1187		    ZFS_DELEG_PERM_USERPROP, cr));
1188	} else {
1189		return (zfs_secpolicy_setprop(zc->zc_name, prop,
1190		    NULL, cr));
1191	}
1192}
1193
1194static int
1195zfs_secpolicy_userspace_one(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1196{
1197	int err = zfs_secpolicy_read(zc, innvl, cr);
1198	if (err)
1199		return (err);
1200
1201	if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
1202		return (SET_ERROR(EINVAL));
1203
1204	if (zc->zc_value[0] == 0) {
1205		/*
1206		 * They are asking about a posix uid/gid.  If it's
1207		 * themself, allow it.
1208		 */
1209		if (zc->zc_objset_type == ZFS_PROP_USERUSED ||
1210		    zc->zc_objset_type == ZFS_PROP_USERQUOTA) {
1211			if (zc->zc_guid == crgetuid(cr))
1212				return (0);
1213		} else {
1214			if (groupmember(zc->zc_guid, cr))
1215				return (0);
1216		}
1217	}
1218
1219	return (zfs_secpolicy_write_perms(zc->zc_name,
1220	    userquota_perms[zc->zc_objset_type], cr));
1221}
1222
1223static int
1224zfs_secpolicy_userspace_many(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1225{
1226	int err = zfs_secpolicy_read(zc, innvl, cr);
1227	if (err)
1228		return (err);
1229
1230	if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
1231		return (SET_ERROR(EINVAL));
1232
1233	return (zfs_secpolicy_write_perms(zc->zc_name,
1234	    userquota_perms[zc->zc_objset_type], cr));
1235}
1236
1237/* ARGSUSED */
1238static int
1239zfs_secpolicy_userspace_upgrade(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1240{
1241	return (zfs_secpolicy_setprop(zc->zc_name, ZFS_PROP_VERSION,
1242	    NULL, cr));
1243}
1244
1245/* ARGSUSED */
1246static int
1247zfs_secpolicy_hold(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1248{
1249	nvpair_t *pair;
1250	nvlist_t *holds;
1251	int error;
1252
1253	error = nvlist_lookup_nvlist(innvl, "holds", &holds);
1254	if (error != 0)
1255		return (SET_ERROR(EINVAL));
1256
1257	for (pair = nvlist_next_nvpair(holds, NULL); pair != NULL;
1258	    pair = nvlist_next_nvpair(holds, pair)) {
1259		char fsname[MAXNAMELEN];
1260		error = dmu_fsname(nvpair_name(pair), fsname);
1261		if (error != 0)
1262			return (error);
1263		error = zfs_secpolicy_write_perms(fsname,
1264		    ZFS_DELEG_PERM_HOLD, cr);
1265		if (error != 0)
1266			return (error);
1267	}
1268	return (0);
1269}
1270
1271/* ARGSUSED */
1272static int
1273zfs_secpolicy_release(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1274{
1275	nvpair_t *pair;
1276	int error;
1277
1278	for (pair = nvlist_next_nvpair(innvl, NULL); pair != NULL;
1279	    pair = nvlist_next_nvpair(innvl, pair)) {
1280		char fsname[MAXNAMELEN];
1281		error = dmu_fsname(nvpair_name(pair), fsname);
1282		if (error != 0)
1283			return (error);
1284		error = zfs_secpolicy_write_perms(fsname,
1285		    ZFS_DELEG_PERM_RELEASE, cr);
1286		if (error != 0)
1287			return (error);
1288	}
1289	return (0);
1290}
1291
1292/*
1293 * Policy for allowing temporary snapshots to be taken or released
1294 */
1295static int
1296zfs_secpolicy_tmp_snapshot(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1297{
1298	/*
1299	 * A temporary snapshot is the same as a snapshot,
1300	 * hold, destroy and release all rolled into one.
1301	 * Delegated diff alone is sufficient that we allow this.
1302	 */
1303	int error;
1304
1305	if ((error = zfs_secpolicy_write_perms(zc->zc_name,
1306	    ZFS_DELEG_PERM_DIFF, cr)) == 0)
1307		return (0);
1308
1309	error = zfs_secpolicy_snapshot_perms(zc->zc_name, cr);
1310	if (error == 0)
1311		error = zfs_secpolicy_hold(zc, innvl, cr);
1312	if (error == 0)
1313		error = zfs_secpolicy_release(zc, innvl, cr);
1314	if (error == 0)
1315		error = zfs_secpolicy_destroy(zc, innvl, cr);
1316	return (error);
1317}
1318
1319/*
1320 * Returns the nvlist as specified by the user in the zfs_cmd_t.
1321 */
1322static int
1323get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp)
1324{
1325	char *packed;
1326	int error;
1327	nvlist_t *list = NULL;
1328
1329	/*
1330	 * Read in and unpack the user-supplied nvlist.
1331	 */
1332	if (size == 0)
1333		return (SET_ERROR(EINVAL));
1334
1335	packed = kmem_alloc(size, KM_SLEEP);
1336
1337	if ((error = ddi_copyin((void *)(uintptr_t)nvl, packed, size,
1338	    iflag)) != 0) {
1339		kmem_free(packed, size);
1340		return (error);
1341	}
1342
1343	if ((error = nvlist_unpack(packed, size, &list, 0)) != 0) {
1344		kmem_free(packed, size);
1345		return (error);
1346	}
1347
1348	kmem_free(packed, size);
1349
1350	*nvp = list;
1351	return (0);
1352}
1353
1354/*
1355 * Reduce the size of this nvlist until it can be serialized in 'max' bytes.
1356 * Entries will be removed from the end of the nvlist, and one int32 entry
1357 * named "N_MORE_ERRORS" will be added indicating how many entries were
1358 * removed.
1359 */
1360static int
1361nvlist_smush(nvlist_t *errors, size_t max)
1362{
1363	size_t size;
1364
1365	size = fnvlist_size(errors);
1366
1367	if (size > max) {
1368		nvpair_t *more_errors;
1369		int n = 0;
1370
1371		if (max < 1024)
1372			return (SET_ERROR(ENOMEM));
1373
1374		fnvlist_add_int32(errors, ZPROP_N_MORE_ERRORS, 0);
1375		more_errors = nvlist_prev_nvpair(errors, NULL);
1376
1377		do {
1378			nvpair_t *pair = nvlist_prev_nvpair(errors,
1379			    more_errors);
1380			fnvlist_remove_nvpair(errors, pair);
1381			n++;
1382			size = fnvlist_size(errors);
1383		} while (size > max);
1384
1385		fnvlist_remove_nvpair(errors, more_errors);
1386		fnvlist_add_int32(errors, ZPROP_N_MORE_ERRORS, n);
1387		ASSERT3U(fnvlist_size(errors), <=, max);
1388	}
1389
1390	return (0);
1391}
1392
1393static int
1394put_nvlist(zfs_cmd_t *zc, nvlist_t *nvl)
1395{
1396	char *packed = NULL;
1397	int error = 0;
1398	size_t size;
1399
1400	size = fnvlist_size(nvl);
1401
1402	if (size > zc->zc_nvlist_dst_size) {
1403		/*
1404		 * Solaris returns ENOMEM here, because even if an error is
1405		 * returned from an ioctl(2), new zc_nvlist_dst_size will be
1406		 * passed to the userland. This is not the case for FreeBSD.
1407		 * We need to return 0, so the kernel will copy the
1408		 * zc_nvlist_dst_size back and the userland can discover that a
1409		 * bigger buffer is needed.
1410		 */
1411		error = 0;
1412	} else {
1413		packed = fnvlist_pack(nvl, &size);
1414		if (ddi_copyout(packed, (void *)(uintptr_t)zc->zc_nvlist_dst,
1415		    size, zc->zc_iflags) != 0)
1416			error = SET_ERROR(EFAULT);
1417		fnvlist_pack_free(packed, size);
1418	}
1419
1420	zc->zc_nvlist_dst_size = size;
1421	zc->zc_nvlist_dst_filled = B_TRUE;
1422	return (error);
1423}
1424
1425static int
1426getzfsvfs(const char *dsname, zfsvfs_t **zfvp)
1427{
1428	objset_t *os;
1429	int error;
1430
1431	error = dmu_objset_hold(dsname, FTAG, &os);
1432	if (error != 0)
1433		return (error);
1434	if (dmu_objset_type(os) != DMU_OST_ZFS) {
1435		dmu_objset_rele(os, FTAG);
1436		return (SET_ERROR(EINVAL));
1437	}
1438
1439	mutex_enter(&os->os_user_ptr_lock);
1440	*zfvp = dmu_objset_get_user(os);
1441	if (*zfvp) {
1442		VFS_HOLD((*zfvp)->z_vfs);
1443	} else {
1444		error = SET_ERROR(ESRCH);
1445	}
1446	mutex_exit(&os->os_user_ptr_lock);
1447	dmu_objset_rele(os, FTAG);
1448	return (error);
1449}
1450
1451/*
1452 * Find a zfsvfs_t for a mounted filesystem, or create our own, in which
1453 * case its z_vfs will be NULL, and it will be opened as the owner.
1454 * If 'writer' is set, the z_teardown_lock will be held for RW_WRITER,
1455 * which prevents all vnode ops from running.
1456 */
1457static int
1458zfsvfs_hold(const char *name, void *tag, zfsvfs_t **zfvp, boolean_t writer)
1459{
1460	int error = 0;
1461
1462	if (getzfsvfs(name, zfvp) != 0)
1463		error = zfsvfs_create(name, zfvp);
1464	if (error == 0) {
1465		rrw_enter(&(*zfvp)->z_teardown_lock, (writer) ? RW_WRITER :
1466		    RW_READER, tag);
1467		if ((*zfvp)->z_unmounted) {
1468			/*
1469			 * XXX we could probably try again, since the unmounting
1470			 * thread should be just about to disassociate the
1471			 * objset from the zfsvfs.
1472			 */
1473			rrw_exit(&(*zfvp)->z_teardown_lock, tag);
1474			return (SET_ERROR(EBUSY));
1475		}
1476	}
1477	return (error);
1478}
1479
1480static void
1481zfsvfs_rele(zfsvfs_t *zfsvfs, void *tag)
1482{
1483	rrw_exit(&zfsvfs->z_teardown_lock, tag);
1484
1485	if (zfsvfs->z_vfs) {
1486		VFS_RELE(zfsvfs->z_vfs);
1487	} else {
1488		dmu_objset_disown(zfsvfs->z_os, zfsvfs);
1489		zfsvfs_free(zfsvfs);
1490	}
1491}
1492
1493static int
1494zfs_ioc_pool_create(zfs_cmd_t *zc)
1495{
1496	int error;
1497	nvlist_t *config, *props = NULL;
1498	nvlist_t *rootprops = NULL;
1499	nvlist_t *zplprops = NULL;
1500
1501	if (error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1502	    zc->zc_iflags, &config))
1503		return (error);
1504
1505	if (zc->zc_nvlist_src_size != 0 && (error =
1506	    get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1507	    zc->zc_iflags, &props))) {
1508		nvlist_free(config);
1509		return (error);
1510	}
1511
1512	if (props) {
1513		nvlist_t *nvl = NULL;
1514		uint64_t version = SPA_VERSION;
1515
1516		(void) nvlist_lookup_uint64(props,
1517		    zpool_prop_to_name(ZPOOL_PROP_VERSION), &version);
1518		if (!SPA_VERSION_IS_SUPPORTED(version)) {
1519			error = SET_ERROR(EINVAL);
1520			goto pool_props_bad;
1521		}
1522		(void) nvlist_lookup_nvlist(props, ZPOOL_ROOTFS_PROPS, &nvl);
1523		if (nvl) {
1524			error = nvlist_dup(nvl, &rootprops, KM_SLEEP);
1525			if (error != 0) {
1526				nvlist_free(config);
1527				nvlist_free(props);
1528				return (error);
1529			}
1530			(void) nvlist_remove_all(props, ZPOOL_ROOTFS_PROPS);
1531		}
1532		VERIFY(nvlist_alloc(&zplprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
1533		error = zfs_fill_zplprops_root(version, rootprops,
1534		    zplprops, NULL);
1535		if (error != 0)
1536			goto pool_props_bad;
1537	}
1538
1539	error = spa_create(zc->zc_name, config, props, zplprops);
1540
1541	/*
1542	 * Set the remaining root properties
1543	 */
1544	if (!error && (error = zfs_set_prop_nvlist(zc->zc_name,
1545	    ZPROP_SRC_LOCAL, rootprops, NULL)) != 0)
1546		(void) spa_destroy(zc->zc_name);
1547
1548pool_props_bad:
1549	nvlist_free(rootprops);
1550	nvlist_free(zplprops);
1551	nvlist_free(config);
1552	nvlist_free(props);
1553
1554	return (error);
1555}
1556
1557static int
1558zfs_ioc_pool_destroy(zfs_cmd_t *zc)
1559{
1560	int error;
1561	zfs_log_history(zc);
1562	error = spa_destroy(zc->zc_name);
1563	if (error == 0)
1564		zvol_remove_minors(zc->zc_name);
1565	return (error);
1566}
1567
1568static int
1569zfs_ioc_pool_import(zfs_cmd_t *zc)
1570{
1571	nvlist_t *config, *props = NULL;
1572	uint64_t guid;
1573	int error;
1574
1575	if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1576	    zc->zc_iflags, &config)) != 0)
1577		return (error);
1578
1579	if (zc->zc_nvlist_src_size != 0 && (error =
1580	    get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1581	    zc->zc_iflags, &props))) {
1582		nvlist_free(config);
1583		return (error);
1584	}
1585
1586	if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, &guid) != 0 ||
1587	    guid != zc->zc_guid)
1588		error = SET_ERROR(EINVAL);
1589	else
1590		error = spa_import(zc->zc_name, config, props, zc->zc_cookie);
1591
1592	if (zc->zc_nvlist_dst != 0) {
1593		int err;
1594
1595		if ((err = put_nvlist(zc, config)) != 0)
1596			error = err;
1597	}
1598
1599	nvlist_free(config);
1600
1601	if (props)
1602		nvlist_free(props);
1603
1604	return (error);
1605}
1606
1607static int
1608zfs_ioc_pool_export(zfs_cmd_t *zc)
1609{
1610	int error;
1611	boolean_t force = (boolean_t)zc->zc_cookie;
1612	boolean_t hardforce = (boolean_t)zc->zc_guid;
1613
1614	zfs_log_history(zc);
1615	error = spa_export(zc->zc_name, NULL, force, hardforce);
1616	if (error == 0)
1617		zvol_remove_minors(zc->zc_name);
1618	return (error);
1619}
1620
1621static int
1622zfs_ioc_pool_configs(zfs_cmd_t *zc)
1623{
1624	nvlist_t *configs;
1625	int error;
1626
1627	if ((configs = spa_all_configs(&zc->zc_cookie)) == NULL)
1628		return (SET_ERROR(EEXIST));
1629
1630	error = put_nvlist(zc, configs);
1631
1632	nvlist_free(configs);
1633
1634	return (error);
1635}
1636
1637/*
1638 * inputs:
1639 * zc_name		name of the pool
1640 *
1641 * outputs:
1642 * zc_cookie		real errno
1643 * zc_nvlist_dst	config nvlist
1644 * zc_nvlist_dst_size	size of config nvlist
1645 */
1646static int
1647zfs_ioc_pool_stats(zfs_cmd_t *zc)
1648{
1649	nvlist_t *config;
1650	int error;
1651	int ret = 0;
1652
1653	error = spa_get_stats(zc->zc_name, &config, zc->zc_value,
1654	    sizeof (zc->zc_value));
1655
1656	if (config != NULL) {
1657		ret = put_nvlist(zc, config);
1658		nvlist_free(config);
1659
1660		/*
1661		 * The config may be present even if 'error' is non-zero.
1662		 * In this case we return success, and preserve the real errno
1663		 * in 'zc_cookie'.
1664		 */
1665		zc->zc_cookie = error;
1666	} else {
1667		ret = error;
1668	}
1669
1670	return (ret);
1671}
1672
1673/*
1674 * Try to import the given pool, returning pool stats as appropriate so that
1675 * user land knows which devices are available and overall pool health.
1676 */
1677static int
1678zfs_ioc_pool_tryimport(zfs_cmd_t *zc)
1679{
1680	nvlist_t *tryconfig, *config;
1681	int error;
1682
1683	if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1684	    zc->zc_iflags, &tryconfig)) != 0)
1685		return (error);
1686
1687	config = spa_tryimport(tryconfig);
1688
1689	nvlist_free(tryconfig);
1690
1691	if (config == NULL)
1692		return (SET_ERROR(EINVAL));
1693
1694	error = put_nvlist(zc, config);
1695	nvlist_free(config);
1696
1697	return (error);
1698}
1699
1700/*
1701 * inputs:
1702 * zc_name              name of the pool
1703 * zc_cookie            scan func (pool_scan_func_t)
1704 */
1705static int
1706zfs_ioc_pool_scan(zfs_cmd_t *zc)
1707{
1708	spa_t *spa;
1709	int error;
1710
1711	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1712		return (error);
1713
1714	if (zc->zc_cookie == POOL_SCAN_NONE)
1715		error = spa_scan_stop(spa);
1716	else
1717		error = spa_scan(spa, zc->zc_cookie);
1718
1719	spa_close(spa, FTAG);
1720
1721	return (error);
1722}
1723
1724static int
1725zfs_ioc_pool_freeze(zfs_cmd_t *zc)
1726{
1727	spa_t *spa;
1728	int error;
1729
1730	error = spa_open(zc->zc_name, &spa, FTAG);
1731	if (error == 0) {
1732		spa_freeze(spa);
1733		spa_close(spa, FTAG);
1734	}
1735	return (error);
1736}
1737
1738static int
1739zfs_ioc_pool_upgrade(zfs_cmd_t *zc)
1740{
1741	spa_t *spa;
1742	int error;
1743
1744	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1745		return (error);
1746
1747	if (zc->zc_cookie < spa_version(spa) ||
1748	    !SPA_VERSION_IS_SUPPORTED(zc->zc_cookie)) {
1749		spa_close(spa, FTAG);
1750		return (SET_ERROR(EINVAL));
1751	}
1752
1753	spa_upgrade(spa, zc->zc_cookie);
1754	spa_close(spa, FTAG);
1755
1756	return (error);
1757}
1758
1759static int
1760zfs_ioc_pool_get_history(zfs_cmd_t *zc)
1761{
1762	spa_t *spa;
1763	char *hist_buf;
1764	uint64_t size;
1765	int error;
1766
1767	if ((size = zc->zc_history_len) == 0)
1768		return (SET_ERROR(EINVAL));
1769
1770	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1771		return (error);
1772
1773	if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) {
1774		spa_close(spa, FTAG);
1775		return (SET_ERROR(ENOTSUP));
1776	}
1777
1778	hist_buf = kmem_alloc(size, KM_SLEEP);
1779	if ((error = spa_history_get(spa, &zc->zc_history_offset,
1780	    &zc->zc_history_len, hist_buf)) == 0) {
1781		error = ddi_copyout(hist_buf,
1782		    (void *)(uintptr_t)zc->zc_history,
1783		    zc->zc_history_len, zc->zc_iflags);
1784	}
1785
1786	spa_close(spa, FTAG);
1787	kmem_free(hist_buf, size);
1788	return (error);
1789}
1790
1791static int
1792zfs_ioc_pool_reguid(zfs_cmd_t *zc)
1793{
1794	spa_t *spa;
1795	int error;
1796
1797	error = spa_open(zc->zc_name, &spa, FTAG);
1798	if (error == 0) {
1799		error = spa_change_guid(spa);
1800		spa_close(spa, FTAG);
1801	}
1802	return (error);
1803}
1804
1805static int
1806zfs_ioc_dsobj_to_dsname(zfs_cmd_t *zc)
1807{
1808	return (dsl_dsobj_to_dsname(zc->zc_name, zc->zc_obj, zc->zc_value));
1809}
1810
1811/*
1812 * inputs:
1813 * zc_name		name of filesystem
1814 * zc_obj		object to find
1815 *
1816 * outputs:
1817 * zc_value		name of object
1818 */
1819static int
1820zfs_ioc_obj_to_path(zfs_cmd_t *zc)
1821{
1822	objset_t *os;
1823	int error;
1824
1825	/* XXX reading from objset not owned */
1826	if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os)) != 0)
1827		return (error);
1828	if (dmu_objset_type(os) != DMU_OST_ZFS) {
1829		dmu_objset_rele(os, FTAG);
1830		return (SET_ERROR(EINVAL));
1831	}
1832	error = zfs_obj_to_path(os, zc->zc_obj, zc->zc_value,
1833	    sizeof (zc->zc_value));
1834	dmu_objset_rele(os, FTAG);
1835
1836	return (error);
1837}
1838
1839/*
1840 * inputs:
1841 * zc_name		name of filesystem
1842 * zc_obj		object to find
1843 *
1844 * outputs:
1845 * zc_stat		stats on object
1846 * zc_value		path to object
1847 */
1848static int
1849zfs_ioc_obj_to_stats(zfs_cmd_t *zc)
1850{
1851	objset_t *os;
1852	int error;
1853
1854	/* XXX reading from objset not owned */
1855	if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os)) != 0)
1856		return (error);
1857	if (dmu_objset_type(os) != DMU_OST_ZFS) {
1858		dmu_objset_rele(os, FTAG);
1859		return (SET_ERROR(EINVAL));
1860	}
1861	error = zfs_obj_to_stats(os, zc->zc_obj, &zc->zc_stat, zc->zc_value,
1862	    sizeof (zc->zc_value));
1863	dmu_objset_rele(os, FTAG);
1864
1865	return (error);
1866}
1867
1868static int
1869zfs_ioc_vdev_add(zfs_cmd_t *zc)
1870{
1871	spa_t *spa;
1872	int error;
1873	nvlist_t *config, **l2cache, **spares;
1874	uint_t nl2cache = 0, nspares = 0;
1875
1876	error = spa_open(zc->zc_name, &spa, FTAG);
1877	if (error != 0)
1878		return (error);
1879
1880	error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1881	    zc->zc_iflags, &config);
1882	(void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_L2CACHE,
1883	    &l2cache, &nl2cache);
1884
1885	(void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_SPARES,
1886	    &spares, &nspares);
1887
1888#ifdef illumos
1889	/*
1890	 * A root pool with concatenated devices is not supported.
1891	 * Thus, can not add a device to a root pool.
1892	 *
1893	 * Intent log device can not be added to a rootpool because
1894	 * during mountroot, zil is replayed, a seperated log device
1895	 * can not be accessed during the mountroot time.
1896	 *
1897	 * l2cache and spare devices are ok to be added to a rootpool.
1898	 */
1899	if (spa_bootfs(spa) != 0 && nl2cache == 0 && nspares == 0) {
1900		nvlist_free(config);
1901		spa_close(spa, FTAG);
1902		return (SET_ERROR(EDOM));
1903	}
1904#endif /* illumos */
1905
1906	if (error == 0) {
1907		error = spa_vdev_add(spa, config);
1908		nvlist_free(config);
1909	}
1910	spa_close(spa, FTAG);
1911	return (error);
1912}
1913
1914/*
1915 * inputs:
1916 * zc_name		name of the pool
1917 * zc_nvlist_conf	nvlist of devices to remove
1918 * zc_cookie		to stop the remove?
1919 */
1920static int
1921zfs_ioc_vdev_remove(zfs_cmd_t *zc)
1922{
1923	spa_t *spa;
1924	int error;
1925
1926	error = spa_open(zc->zc_name, &spa, FTAG);
1927	if (error != 0)
1928		return (error);
1929	error = spa_vdev_remove(spa, zc->zc_guid, B_FALSE);
1930	spa_close(spa, FTAG);
1931	return (error);
1932}
1933
1934static int
1935zfs_ioc_vdev_set_state(zfs_cmd_t *zc)
1936{
1937	spa_t *spa;
1938	int error;
1939	vdev_state_t newstate = VDEV_STATE_UNKNOWN;
1940
1941	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1942		return (error);
1943	switch (zc->zc_cookie) {
1944	case VDEV_STATE_ONLINE:
1945		error = vdev_online(spa, zc->zc_guid, zc->zc_obj, &newstate);
1946		break;
1947
1948	case VDEV_STATE_OFFLINE:
1949		error = vdev_offline(spa, zc->zc_guid, zc->zc_obj);
1950		break;
1951
1952	case VDEV_STATE_FAULTED:
1953		if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED &&
1954		    zc->zc_obj != VDEV_AUX_EXTERNAL)
1955			zc->zc_obj = VDEV_AUX_ERR_EXCEEDED;
1956
1957		error = vdev_fault(spa, zc->zc_guid, zc->zc_obj);
1958		break;
1959
1960	case VDEV_STATE_DEGRADED:
1961		if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED &&
1962		    zc->zc_obj != VDEV_AUX_EXTERNAL)
1963			zc->zc_obj = VDEV_AUX_ERR_EXCEEDED;
1964
1965		error = vdev_degrade(spa, zc->zc_guid, zc->zc_obj);
1966		break;
1967
1968	default:
1969		error = SET_ERROR(EINVAL);
1970	}
1971	zc->zc_cookie = newstate;
1972	spa_close(spa, FTAG);
1973	return (error);
1974}
1975
1976static int
1977zfs_ioc_vdev_attach(zfs_cmd_t *zc)
1978{
1979	spa_t *spa;
1980	int replacing = zc->zc_cookie;
1981	nvlist_t *config;
1982	int error;
1983
1984	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1985		return (error);
1986
1987	if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1988	    zc->zc_iflags, &config)) == 0) {
1989		error = spa_vdev_attach(spa, zc->zc_guid, config, replacing);
1990		nvlist_free(config);
1991	}
1992
1993	spa_close(spa, FTAG);
1994	return (error);
1995}
1996
1997static int
1998zfs_ioc_vdev_detach(zfs_cmd_t *zc)
1999{
2000	spa_t *spa;
2001	int error;
2002
2003	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
2004		return (error);
2005
2006	error = spa_vdev_detach(spa, zc->zc_guid, 0, B_FALSE);
2007
2008	spa_close(spa, FTAG);
2009	return (error);
2010}
2011
2012static int
2013zfs_ioc_vdev_split(zfs_cmd_t *zc)
2014{
2015	spa_t *spa;
2016	nvlist_t *config, *props = NULL;
2017	int error;
2018	boolean_t exp = !!(zc->zc_cookie & ZPOOL_EXPORT_AFTER_SPLIT);
2019
2020	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
2021		return (error);
2022
2023	if (error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
2024	    zc->zc_iflags, &config)) {
2025		spa_close(spa, FTAG);
2026		return (error);
2027	}
2028
2029	if (zc->zc_nvlist_src_size != 0 && (error =
2030	    get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2031	    zc->zc_iflags, &props))) {
2032		spa_close(spa, FTAG);
2033		nvlist_free(config);
2034		return (error);
2035	}
2036
2037	error = spa_vdev_split_mirror(spa, zc->zc_string, config, props, exp);
2038
2039	spa_close(spa, FTAG);
2040
2041	nvlist_free(config);
2042	nvlist_free(props);
2043
2044	return (error);
2045}
2046
2047static int
2048zfs_ioc_vdev_setpath(zfs_cmd_t *zc)
2049{
2050	spa_t *spa;
2051	char *path = zc->zc_value;
2052	uint64_t guid = zc->zc_guid;
2053	int error;
2054
2055	error = spa_open(zc->zc_name, &spa, FTAG);
2056	if (error != 0)
2057		return (error);
2058
2059	error = spa_vdev_setpath(spa, guid, path);
2060	spa_close(spa, FTAG);
2061	return (error);
2062}
2063
2064static int
2065zfs_ioc_vdev_setfru(zfs_cmd_t *zc)
2066{
2067	spa_t *spa;
2068	char *fru = zc->zc_value;
2069	uint64_t guid = zc->zc_guid;
2070	int error;
2071
2072	error = spa_open(zc->zc_name, &spa, FTAG);
2073	if (error != 0)
2074		return (error);
2075
2076	error = spa_vdev_setfru(spa, guid, fru);
2077	spa_close(spa, FTAG);
2078	return (error);
2079}
2080
2081static int
2082zfs_ioc_objset_stats_impl(zfs_cmd_t *zc, objset_t *os)
2083{
2084	int error = 0;
2085	nvlist_t *nv;
2086
2087	dmu_objset_fast_stat(os, &zc->zc_objset_stats);
2088
2089	if (zc->zc_nvlist_dst != 0 &&
2090	    (error = dsl_prop_get_all(os, &nv)) == 0) {
2091		dmu_objset_stats(os, nv);
2092		/*
2093		 * NB: zvol_get_stats() will read the objset contents,
2094		 * which we aren't supposed to do with a
2095		 * DS_MODE_USER hold, because it could be
2096		 * inconsistent.  So this is a bit of a workaround...
2097		 * XXX reading with out owning
2098		 */
2099		if (!zc->zc_objset_stats.dds_inconsistent &&
2100		    dmu_objset_type(os) == DMU_OST_ZVOL) {
2101			error = zvol_get_stats(os, nv);
2102			if (error == EIO)
2103				return (error);
2104			VERIFY0(error);
2105		}
2106		error = put_nvlist(zc, nv);
2107		nvlist_free(nv);
2108	}
2109
2110	return (error);
2111}
2112
2113/*
2114 * inputs:
2115 * zc_name		name of filesystem
2116 * zc_nvlist_dst_size	size of buffer for property nvlist
2117 *
2118 * outputs:
2119 * zc_objset_stats	stats
2120 * zc_nvlist_dst	property nvlist
2121 * zc_nvlist_dst_size	size of property nvlist
2122 */
2123static int
2124zfs_ioc_objset_stats(zfs_cmd_t *zc)
2125{
2126	objset_t *os;
2127	int error;
2128
2129	error = dmu_objset_hold(zc->zc_name, FTAG, &os);
2130	if (error == 0) {
2131		error = zfs_ioc_objset_stats_impl(zc, os);
2132		dmu_objset_rele(os, FTAG);
2133	}
2134
2135	if (error == ENOMEM)
2136		error = 0;
2137	return (error);
2138}
2139
2140/*
2141 * inputs:
2142 * zc_name		name of filesystem
2143 * zc_nvlist_dst_size	size of buffer for property nvlist
2144 *
2145 * outputs:
2146 * zc_nvlist_dst	received property nvlist
2147 * zc_nvlist_dst_size	size of received property nvlist
2148 *
2149 * Gets received properties (distinct from local properties on or after
2150 * SPA_VERSION_RECVD_PROPS) for callers who want to differentiate received from
2151 * local property values.
2152 */
2153static int
2154zfs_ioc_objset_recvd_props(zfs_cmd_t *zc)
2155{
2156	int error = 0;
2157	nvlist_t *nv;
2158
2159	/*
2160	 * Without this check, we would return local property values if the
2161	 * caller has not already received properties on or after
2162	 * SPA_VERSION_RECVD_PROPS.
2163	 */
2164	if (!dsl_prop_get_hasrecvd(zc->zc_name))
2165		return (SET_ERROR(ENOTSUP));
2166
2167	if (zc->zc_nvlist_dst != 0 &&
2168	    (error = dsl_prop_get_received(zc->zc_name, &nv)) == 0) {
2169		error = put_nvlist(zc, nv);
2170		nvlist_free(nv);
2171	}
2172
2173	return (error);
2174}
2175
2176static int
2177nvl_add_zplprop(objset_t *os, nvlist_t *props, zfs_prop_t prop)
2178{
2179	uint64_t value;
2180	int error;
2181
2182	/*
2183	 * zfs_get_zplprop() will either find a value or give us
2184	 * the default value (if there is one).
2185	 */
2186	if ((error = zfs_get_zplprop(os, prop, &value)) != 0)
2187		return (error);
2188	VERIFY(nvlist_add_uint64(props, zfs_prop_to_name(prop), value) == 0);
2189	return (0);
2190}
2191
2192/*
2193 * inputs:
2194 * zc_name		name of filesystem
2195 * zc_nvlist_dst_size	size of buffer for zpl property nvlist
2196 *
2197 * outputs:
2198 * zc_nvlist_dst	zpl property nvlist
2199 * zc_nvlist_dst_size	size of zpl property nvlist
2200 */
2201static int
2202zfs_ioc_objset_zplprops(zfs_cmd_t *zc)
2203{
2204	objset_t *os;
2205	int err;
2206
2207	/* XXX reading without owning */
2208	if (err = dmu_objset_hold(zc->zc_name, FTAG, &os))
2209		return (err);
2210
2211	dmu_objset_fast_stat(os, &zc->zc_objset_stats);
2212
2213	/*
2214	 * NB: nvl_add_zplprop() will read the objset contents,
2215	 * which we aren't supposed to do with a DS_MODE_USER
2216	 * hold, because it could be inconsistent.
2217	 */
2218	if (zc->zc_nvlist_dst != 0 &&
2219	    !zc->zc_objset_stats.dds_inconsistent &&
2220	    dmu_objset_type(os) == DMU_OST_ZFS) {
2221		nvlist_t *nv;
2222
2223		VERIFY(nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2224		if ((err = nvl_add_zplprop(os, nv, ZFS_PROP_VERSION)) == 0 &&
2225		    (err = nvl_add_zplprop(os, nv, ZFS_PROP_NORMALIZE)) == 0 &&
2226		    (err = nvl_add_zplprop(os, nv, ZFS_PROP_UTF8ONLY)) == 0 &&
2227		    (err = nvl_add_zplprop(os, nv, ZFS_PROP_CASE)) == 0)
2228			err = put_nvlist(zc, nv);
2229		nvlist_free(nv);
2230	} else {
2231		err = SET_ERROR(ENOENT);
2232	}
2233	dmu_objset_rele(os, FTAG);
2234	return (err);
2235}
2236
2237boolean_t
2238dataset_name_hidden(const char *name)
2239{
2240	/*
2241	 * Skip over datasets that are not visible in this zone,
2242	 * internal datasets (which have a $ in their name), and
2243	 * temporary datasets (which have a % in their name).
2244	 */
2245	if (strchr(name, '$') != NULL)
2246		return (B_TRUE);
2247	if (strchr(name, '%') != NULL)
2248		return (B_TRUE);
2249	if (!INGLOBALZONE(curthread) && !zone_dataset_visible(name, NULL))
2250		return (B_TRUE);
2251	return (B_FALSE);
2252}
2253
2254/*
2255 * inputs:
2256 * zc_name		name of filesystem
2257 * zc_cookie		zap cursor
2258 * zc_nvlist_dst_size	size of buffer for property nvlist
2259 *
2260 * outputs:
2261 * zc_name		name of next filesystem
2262 * zc_cookie		zap cursor
2263 * zc_objset_stats	stats
2264 * zc_nvlist_dst	property nvlist
2265 * zc_nvlist_dst_size	size of property nvlist
2266 */
2267static int
2268zfs_ioc_dataset_list_next(zfs_cmd_t *zc)
2269{
2270	objset_t *os;
2271	int error;
2272	char *p;
2273	size_t orig_len = strlen(zc->zc_name);
2274
2275top:
2276	if (error = dmu_objset_hold(zc->zc_name, FTAG, &os)) {
2277		if (error == ENOENT)
2278			error = SET_ERROR(ESRCH);
2279		return (error);
2280	}
2281
2282	p = strrchr(zc->zc_name, '/');
2283	if (p == NULL || p[1] != '\0')
2284		(void) strlcat(zc->zc_name, "/", sizeof (zc->zc_name));
2285	p = zc->zc_name + strlen(zc->zc_name);
2286
2287	do {
2288		error = dmu_dir_list_next(os,
2289		    sizeof (zc->zc_name) - (p - zc->zc_name), p,
2290		    NULL, &zc->zc_cookie);
2291		if (error == ENOENT)
2292			error = SET_ERROR(ESRCH);
2293	} while (error == 0 && dataset_name_hidden(zc->zc_name));
2294	dmu_objset_rele(os, FTAG);
2295
2296	/*
2297	 * If it's an internal dataset (ie. with a '$' in its name),
2298	 * don't try to get stats for it, otherwise we'll return ENOENT.
2299	 */
2300	if (error == 0 && strchr(zc->zc_name, '$') == NULL) {
2301		error = zfs_ioc_objset_stats(zc); /* fill in the stats */
2302		if (error == ENOENT) {
2303			/* We lost a race with destroy, get the next one. */
2304			zc->zc_name[orig_len] = '\0';
2305			goto top;
2306		}
2307	}
2308	return (error);
2309}
2310
2311/*
2312 * inputs:
2313 * zc_name		name of filesystem
2314 * zc_cookie		zap cursor
2315 * zc_nvlist_dst_size	size of buffer for property nvlist
2316 * zc_simple		when set, only name is requested
2317 *
2318 * outputs:
2319 * zc_name		name of next snapshot
2320 * zc_objset_stats	stats
2321 * zc_nvlist_dst	property nvlist
2322 * zc_nvlist_dst_size	size of property nvlist
2323 */
2324static int
2325zfs_ioc_snapshot_list_next(zfs_cmd_t *zc)
2326{
2327	objset_t *os;
2328	int error;
2329
2330	error = dmu_objset_hold(zc->zc_name, FTAG, &os);
2331	if (error != 0) {
2332		return (error == ENOENT ? ESRCH : error);
2333	}
2334
2335	/*
2336	 * A dataset name of maximum length cannot have any snapshots,
2337	 * so exit immediately.
2338	 */
2339	if (strlcat(zc->zc_name, "@", sizeof (zc->zc_name)) >= MAXNAMELEN) {
2340		dmu_objset_rele(os, FTAG);
2341		return (SET_ERROR(ESRCH));
2342	}
2343
2344	error = dmu_snapshot_list_next(os,
2345	    sizeof (zc->zc_name) - strlen(zc->zc_name),
2346	    zc->zc_name + strlen(zc->zc_name), &zc->zc_obj, &zc->zc_cookie,
2347	    NULL);
2348
2349	if (error == 0 && !zc->zc_simple) {
2350		dsl_dataset_t *ds;
2351		dsl_pool_t *dp = os->os_dsl_dataset->ds_dir->dd_pool;
2352
2353		error = dsl_dataset_hold_obj(dp, zc->zc_obj, FTAG, &ds);
2354		if (error == 0) {
2355			objset_t *ossnap;
2356
2357			error = dmu_objset_from_ds(ds, &ossnap);
2358			if (error == 0)
2359				error = zfs_ioc_objset_stats_impl(zc, ossnap);
2360			dsl_dataset_rele(ds, FTAG);
2361		}
2362	} else if (error == ENOENT) {
2363		error = SET_ERROR(ESRCH);
2364	}
2365
2366	dmu_objset_rele(os, FTAG);
2367	/* if we failed, undo the @ that we tacked on to zc_name */
2368	if (error != 0)
2369		*strchr(zc->zc_name, '@') = '\0';
2370	return (error);
2371}
2372
2373static int
2374zfs_prop_set_userquota(const char *dsname, nvpair_t *pair)
2375{
2376	const char *propname = nvpair_name(pair);
2377	uint64_t *valary;
2378	unsigned int vallen;
2379	const char *domain;
2380	char *dash;
2381	zfs_userquota_prop_t type;
2382	uint64_t rid;
2383	uint64_t quota;
2384	zfsvfs_t *zfsvfs;
2385	int err;
2386
2387	if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2388		nvlist_t *attrs;
2389		VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
2390		if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2391		    &pair) != 0)
2392			return (SET_ERROR(EINVAL));
2393	}
2394
2395	/*
2396	 * A correctly constructed propname is encoded as
2397	 * userquota@<rid>-<domain>.
2398	 */
2399	if ((dash = strchr(propname, '-')) == NULL ||
2400	    nvpair_value_uint64_array(pair, &valary, &vallen) != 0 ||
2401	    vallen != 3)
2402		return (SET_ERROR(EINVAL));
2403
2404	domain = dash + 1;
2405	type = valary[0];
2406	rid = valary[1];
2407	quota = valary[2];
2408
2409	err = zfsvfs_hold(dsname, FTAG, &zfsvfs, B_FALSE);
2410	if (err == 0) {
2411		err = zfs_set_userquota(zfsvfs, type, domain, rid, quota);
2412		zfsvfs_rele(zfsvfs, FTAG);
2413	}
2414
2415	return (err);
2416}
2417
2418/*
2419 * If the named property is one that has a special function to set its value,
2420 * return 0 on success and a positive error code on failure; otherwise if it is
2421 * not one of the special properties handled by this function, return -1.
2422 *
2423 * XXX: It would be better for callers of the property interface if we handled
2424 * these special cases in dsl_prop.c (in the dsl layer).
2425 */
2426static int
2427zfs_prop_set_special(const char *dsname, zprop_source_t source,
2428    nvpair_t *pair)
2429{
2430	const char *propname = nvpair_name(pair);
2431	zfs_prop_t prop = zfs_name_to_prop(propname);
2432	uint64_t intval;
2433	int err;
2434
2435	if (prop == ZPROP_INVAL) {
2436		if (zfs_prop_userquota(propname))
2437			return (zfs_prop_set_userquota(dsname, pair));
2438		return (-1);
2439	}
2440
2441	if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2442		nvlist_t *attrs;
2443		VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
2444		VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2445		    &pair) == 0);
2446	}
2447
2448	if (zfs_prop_get_type(prop) == PROP_TYPE_STRING)
2449		return (-1);
2450
2451	VERIFY(0 == nvpair_value_uint64(pair, &intval));
2452
2453	switch (prop) {
2454	case ZFS_PROP_QUOTA:
2455		err = dsl_dir_set_quota(dsname, source, intval);
2456		break;
2457	case ZFS_PROP_REFQUOTA:
2458		err = dsl_dataset_set_refquota(dsname, source, intval);
2459		break;
2460	case ZFS_PROP_RESERVATION:
2461		err = dsl_dir_set_reservation(dsname, source, intval);
2462		break;
2463	case ZFS_PROP_REFRESERVATION:
2464		err = dsl_dataset_set_refreservation(dsname, source, intval);
2465		break;
2466	case ZFS_PROP_VOLSIZE:
2467		err = zvol_set_volsize(dsname, ddi_driver_major(zfs_dip),
2468		    intval);
2469		break;
2470	case ZFS_PROP_VERSION:
2471	{
2472		zfsvfs_t *zfsvfs;
2473
2474		if ((err = zfsvfs_hold(dsname, FTAG, &zfsvfs, B_TRUE)) != 0)
2475			break;
2476
2477		err = zfs_set_version(zfsvfs, intval);
2478		zfsvfs_rele(zfsvfs, FTAG);
2479
2480		if (err == 0 && intval >= ZPL_VERSION_USERSPACE) {
2481			zfs_cmd_t *zc;
2482
2483			zc = kmem_zalloc(sizeof (zfs_cmd_t), KM_SLEEP);
2484			(void) strcpy(zc->zc_name, dsname);
2485			(void) zfs_ioc_userspace_upgrade(zc);
2486			kmem_free(zc, sizeof (zfs_cmd_t));
2487		}
2488		break;
2489	}
2490	case ZFS_PROP_COMPRESSION:
2491	{
2492		if (intval == ZIO_COMPRESS_LZ4) {
2493			spa_t *spa;
2494
2495			if ((err = spa_open(dsname, &spa, FTAG)) != 0)
2496				return (err);
2497
2498			/*
2499			 * Setting the LZ4 compression algorithm activates
2500			 * the feature.
2501			 */
2502			if (!spa_feature_is_active(spa,
2503			    SPA_FEATURE_LZ4_COMPRESS)) {
2504				if ((err = zfs_prop_activate_feature(spa,
2505				    SPA_FEATURE_LZ4_COMPRESS)) != 0) {
2506					spa_close(spa, FTAG);
2507					return (err);
2508				}
2509			}
2510
2511			spa_close(spa, FTAG);
2512		}
2513		/*
2514		 * We still want the default set action to be performed in the
2515		 * caller, we only performed zfeature settings here.
2516		 */
2517		err = -1;
2518		break;
2519	}
2520
2521	default:
2522		err = -1;
2523	}
2524
2525	return (err);
2526}
2527
2528/*
2529 * This function is best effort. If it fails to set any of the given properties,
2530 * it continues to set as many as it can and returns the last error
2531 * encountered. If the caller provides a non-NULL errlist, it will be filled in
2532 * with the list of names of all the properties that failed along with the
2533 * corresponding error numbers.
2534 *
2535 * If every property is set successfully, zero is returned and errlist is not
2536 * modified.
2537 */
2538int
2539zfs_set_prop_nvlist(const char *dsname, zprop_source_t source, nvlist_t *nvl,
2540    nvlist_t *errlist)
2541{
2542	nvpair_t *pair;
2543	nvpair_t *propval;
2544	int rv = 0;
2545	uint64_t intval;
2546	char *strval;
2547	nvlist_t *genericnvl = fnvlist_alloc();
2548	nvlist_t *retrynvl = fnvlist_alloc();
2549
2550retry:
2551	pair = NULL;
2552	while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) {
2553		const char *propname = nvpair_name(pair);
2554		zfs_prop_t prop = zfs_name_to_prop(propname);
2555		int err = 0;
2556
2557		/* decode the property value */
2558		propval = pair;
2559		if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2560			nvlist_t *attrs;
2561			attrs = fnvpair_value_nvlist(pair);
2562			if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2563			    &propval) != 0)
2564				err = SET_ERROR(EINVAL);
2565		}
2566
2567		/* Validate value type */
2568		if (err == 0 && prop == ZPROP_INVAL) {
2569			if (zfs_prop_user(propname)) {
2570				if (nvpair_type(propval) != DATA_TYPE_STRING)
2571					err = SET_ERROR(EINVAL);
2572			} else if (zfs_prop_userquota(propname)) {
2573				if (nvpair_type(propval) !=
2574				    DATA_TYPE_UINT64_ARRAY)
2575					err = SET_ERROR(EINVAL);
2576			} else {
2577				err = SET_ERROR(EINVAL);
2578			}
2579		} else if (err == 0) {
2580			if (nvpair_type(propval) == DATA_TYPE_STRING) {
2581				if (zfs_prop_get_type(prop) != PROP_TYPE_STRING)
2582					err = SET_ERROR(EINVAL);
2583			} else if (nvpair_type(propval) == DATA_TYPE_UINT64) {
2584				const char *unused;
2585
2586				intval = fnvpair_value_uint64(propval);
2587
2588				switch (zfs_prop_get_type(prop)) {
2589				case PROP_TYPE_NUMBER:
2590					break;
2591				case PROP_TYPE_STRING:
2592					err = SET_ERROR(EINVAL);
2593					break;
2594				case PROP_TYPE_INDEX:
2595					if (zfs_prop_index_to_string(prop,
2596					    intval, &unused) != 0)
2597						err = SET_ERROR(EINVAL);
2598					break;
2599				default:
2600					cmn_err(CE_PANIC,
2601					    "unknown property type");
2602				}
2603			} else {
2604				err = SET_ERROR(EINVAL);
2605			}
2606		}
2607
2608		/* Validate permissions */
2609		if (err == 0)
2610			err = zfs_check_settable(dsname, pair, CRED());
2611
2612		if (err == 0) {
2613			err = zfs_prop_set_special(dsname, source, pair);
2614			if (err == -1) {
2615				/*
2616				 * For better performance we build up a list of
2617				 * properties to set in a single transaction.
2618				 */
2619				err = nvlist_add_nvpair(genericnvl, pair);
2620			} else if (err != 0 && nvl != retrynvl) {
2621				/*
2622				 * This may be a spurious error caused by
2623				 * receiving quota and reservation out of order.
2624				 * Try again in a second pass.
2625				 */
2626				err = nvlist_add_nvpair(retrynvl, pair);
2627			}
2628		}
2629
2630		if (err != 0) {
2631			if (errlist != NULL)
2632				fnvlist_add_int32(errlist, propname, err);
2633			rv = err;
2634		}
2635	}
2636
2637	if (nvl != retrynvl && !nvlist_empty(retrynvl)) {
2638		nvl = retrynvl;
2639		goto retry;
2640	}
2641
2642	if (!nvlist_empty(genericnvl) &&
2643	    dsl_props_set(dsname, source, genericnvl) != 0) {
2644		/*
2645		 * If this fails, we still want to set as many properties as we
2646		 * can, so try setting them individually.
2647		 */
2648		pair = NULL;
2649		while ((pair = nvlist_next_nvpair(genericnvl, pair)) != NULL) {
2650			const char *propname = nvpair_name(pair);
2651			int err = 0;
2652
2653			propval = pair;
2654			if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2655				nvlist_t *attrs;
2656				attrs = fnvpair_value_nvlist(pair);
2657				propval = fnvlist_lookup_nvpair(attrs,
2658				    ZPROP_VALUE);
2659			}
2660
2661			if (nvpair_type(propval) == DATA_TYPE_STRING) {
2662				strval = fnvpair_value_string(propval);
2663				err = dsl_prop_set_string(dsname, propname,
2664				    source, strval);
2665			} else {
2666				intval = fnvpair_value_uint64(propval);
2667				err = dsl_prop_set_int(dsname, propname, source,
2668				    intval);
2669			}
2670
2671			if (err != 0) {
2672				if (errlist != NULL) {
2673					fnvlist_add_int32(errlist, propname,
2674					    err);
2675				}
2676				rv = err;
2677			}
2678		}
2679	}
2680	nvlist_free(genericnvl);
2681	nvlist_free(retrynvl);
2682
2683	return (rv);
2684}
2685
2686/*
2687 * Check that all the properties are valid user properties.
2688 */
2689static int
2690zfs_check_userprops(const char *fsname, nvlist_t *nvl)
2691{
2692	nvpair_t *pair = NULL;
2693	int error = 0;
2694
2695	while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) {
2696		const char *propname = nvpair_name(pair);
2697
2698		if (!zfs_prop_user(propname) ||
2699		    nvpair_type(pair) != DATA_TYPE_STRING)
2700			return (SET_ERROR(EINVAL));
2701
2702		if (error = zfs_secpolicy_write_perms(fsname,
2703		    ZFS_DELEG_PERM_USERPROP, CRED()))
2704			return (error);
2705
2706		if (strlen(propname) >= ZAP_MAXNAMELEN)
2707			return (SET_ERROR(ENAMETOOLONG));
2708
2709		if (strlen(fnvpair_value_string(pair)) >= ZAP_MAXVALUELEN)
2710			return (E2BIG);
2711	}
2712	return (0);
2713}
2714
2715static void
2716props_skip(nvlist_t *props, nvlist_t *skipped, nvlist_t **newprops)
2717{
2718	nvpair_t *pair;
2719
2720	VERIFY(nvlist_alloc(newprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2721
2722	pair = NULL;
2723	while ((pair = nvlist_next_nvpair(props, pair)) != NULL) {
2724		if (nvlist_exists(skipped, nvpair_name(pair)))
2725			continue;
2726
2727		VERIFY(nvlist_add_nvpair(*newprops, pair) == 0);
2728	}
2729}
2730
2731static int
2732clear_received_props(const char *dsname, nvlist_t *props,
2733    nvlist_t *skipped)
2734{
2735	int err = 0;
2736	nvlist_t *cleared_props = NULL;
2737	props_skip(props, skipped, &cleared_props);
2738	if (!nvlist_empty(cleared_props)) {
2739		/*
2740		 * Acts on local properties until the dataset has received
2741		 * properties at least once on or after SPA_VERSION_RECVD_PROPS.
2742		 */
2743		zprop_source_t flags = (ZPROP_SRC_NONE |
2744		    (dsl_prop_get_hasrecvd(dsname) ? ZPROP_SRC_RECEIVED : 0));
2745		err = zfs_set_prop_nvlist(dsname, flags, cleared_props, NULL);
2746	}
2747	nvlist_free(cleared_props);
2748	return (err);
2749}
2750
2751/*
2752 * inputs:
2753 * zc_name		name of filesystem
2754 * zc_value		name of property to set
2755 * zc_nvlist_src{_size}	nvlist of properties to apply
2756 * zc_cookie		received properties flag
2757 *
2758 * outputs:
2759 * zc_nvlist_dst{_size} error for each unapplied received property
2760 */
2761static int
2762zfs_ioc_set_prop(zfs_cmd_t *zc)
2763{
2764	nvlist_t *nvl;
2765	boolean_t received = zc->zc_cookie;
2766	zprop_source_t source = (received ? ZPROP_SRC_RECEIVED :
2767	    ZPROP_SRC_LOCAL);
2768	nvlist_t *errors;
2769	int error;
2770
2771	if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2772	    zc->zc_iflags, &nvl)) != 0)
2773		return (error);
2774
2775	if (received) {
2776		nvlist_t *origprops;
2777
2778		if (dsl_prop_get_received(zc->zc_name, &origprops) == 0) {
2779			(void) clear_received_props(zc->zc_name,
2780			    origprops, nvl);
2781			nvlist_free(origprops);
2782		}
2783
2784		error = dsl_prop_set_hasrecvd(zc->zc_name);
2785	}
2786
2787	errors = fnvlist_alloc();
2788	if (error == 0)
2789		error = zfs_set_prop_nvlist(zc->zc_name, source, nvl, errors);
2790
2791	if (zc->zc_nvlist_dst != 0 && errors != NULL) {
2792		(void) put_nvlist(zc, errors);
2793	}
2794
2795	nvlist_free(errors);
2796	nvlist_free(nvl);
2797	return (error);
2798}
2799
2800/*
2801 * inputs:
2802 * zc_name		name of filesystem
2803 * zc_value		name of property to inherit
2804 * zc_cookie		revert to received value if TRUE
2805 *
2806 * outputs:		none
2807 */
2808static int
2809zfs_ioc_inherit_prop(zfs_cmd_t *zc)
2810{
2811	const char *propname = zc->zc_value;
2812	zfs_prop_t prop = zfs_name_to_prop(propname);
2813	boolean_t received = zc->zc_cookie;
2814	zprop_source_t source = (received
2815	    ? ZPROP_SRC_NONE		/* revert to received value, if any */
2816	    : ZPROP_SRC_INHERITED);	/* explicitly inherit */
2817
2818	if (received) {
2819		nvlist_t *dummy;
2820		nvpair_t *pair;
2821		zprop_type_t type;
2822		int err;
2823
2824		/*
2825		 * zfs_prop_set_special() expects properties in the form of an
2826		 * nvpair with type info.
2827		 */
2828		if (prop == ZPROP_INVAL) {
2829			if (!zfs_prop_user(propname))
2830				return (SET_ERROR(EINVAL));
2831
2832			type = PROP_TYPE_STRING;
2833		} else if (prop == ZFS_PROP_VOLSIZE ||
2834		    prop == ZFS_PROP_VERSION) {
2835			return (SET_ERROR(EINVAL));
2836		} else {
2837			type = zfs_prop_get_type(prop);
2838		}
2839
2840		VERIFY(nvlist_alloc(&dummy, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2841
2842		switch (type) {
2843		case PROP_TYPE_STRING:
2844			VERIFY(0 == nvlist_add_string(dummy, propname, ""));
2845			break;
2846		case PROP_TYPE_NUMBER:
2847		case PROP_TYPE_INDEX:
2848			VERIFY(0 == nvlist_add_uint64(dummy, propname, 0));
2849			break;
2850		default:
2851			nvlist_free(dummy);
2852			return (SET_ERROR(EINVAL));
2853		}
2854
2855		pair = nvlist_next_nvpair(dummy, NULL);
2856		err = zfs_prop_set_special(zc->zc_name, source, pair);
2857		nvlist_free(dummy);
2858		if (err != -1)
2859			return (err); /* special property already handled */
2860	} else {
2861		/*
2862		 * Only check this in the non-received case. We want to allow
2863		 * 'inherit -S' to revert non-inheritable properties like quota
2864		 * and reservation to the received or default values even though
2865		 * they are not considered inheritable.
2866		 */
2867		if (prop != ZPROP_INVAL && !zfs_prop_inheritable(prop))
2868			return (SET_ERROR(EINVAL));
2869	}
2870
2871	/* property name has been validated by zfs_secpolicy_inherit_prop() */
2872	return (dsl_prop_inherit(zc->zc_name, zc->zc_value, source));
2873}
2874
2875static int
2876zfs_ioc_pool_set_props(zfs_cmd_t *zc)
2877{
2878	nvlist_t *props;
2879	spa_t *spa;
2880	int error;
2881	nvpair_t *pair;
2882
2883	if (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2884	    zc->zc_iflags, &props))
2885		return (error);
2886
2887	/*
2888	 * If the only property is the configfile, then just do a spa_lookup()
2889	 * to handle the faulted case.
2890	 */
2891	pair = nvlist_next_nvpair(props, NULL);
2892	if (pair != NULL && strcmp(nvpair_name(pair),
2893	    zpool_prop_to_name(ZPOOL_PROP_CACHEFILE)) == 0 &&
2894	    nvlist_next_nvpair(props, pair) == NULL) {
2895		mutex_enter(&spa_namespace_lock);
2896		if ((spa = spa_lookup(zc->zc_name)) != NULL) {
2897			spa_configfile_set(spa, props, B_FALSE);
2898			spa_config_sync(spa, B_FALSE, B_TRUE);
2899		}
2900		mutex_exit(&spa_namespace_lock);
2901		if (spa != NULL) {
2902			nvlist_free(props);
2903			return (0);
2904		}
2905	}
2906
2907	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) {
2908		nvlist_free(props);
2909		return (error);
2910	}
2911
2912	error = spa_prop_set(spa, props);
2913
2914	nvlist_free(props);
2915	spa_close(spa, FTAG);
2916
2917	return (error);
2918}
2919
2920static int
2921zfs_ioc_pool_get_props(zfs_cmd_t *zc)
2922{
2923	spa_t *spa;
2924	int error;
2925	nvlist_t *nvp = NULL;
2926
2927	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) {
2928		/*
2929		 * If the pool is faulted, there may be properties we can still
2930		 * get (such as altroot and cachefile), so attempt to get them
2931		 * anyway.
2932		 */
2933		mutex_enter(&spa_namespace_lock);
2934		if ((spa = spa_lookup(zc->zc_name)) != NULL)
2935			error = spa_prop_get(spa, &nvp);
2936		mutex_exit(&spa_namespace_lock);
2937	} else {
2938		error = spa_prop_get(spa, &nvp);
2939		spa_close(spa, FTAG);
2940	}
2941
2942	if (error == 0 && zc->zc_nvlist_dst != 0)
2943		error = put_nvlist(zc, nvp);
2944	else
2945		error = SET_ERROR(EFAULT);
2946
2947	nvlist_free(nvp);
2948	return (error);
2949}
2950
2951/*
2952 * inputs:
2953 * zc_name		name of filesystem
2954 * zc_nvlist_src{_size}	nvlist of delegated permissions
2955 * zc_perm_action	allow/unallow flag
2956 *
2957 * outputs:		none
2958 */
2959static int
2960zfs_ioc_set_fsacl(zfs_cmd_t *zc)
2961{
2962	int error;
2963	nvlist_t *fsaclnv = NULL;
2964
2965	if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2966	    zc->zc_iflags, &fsaclnv)) != 0)
2967		return (error);
2968
2969	/*
2970	 * Verify nvlist is constructed correctly
2971	 */
2972	if ((error = zfs_deleg_verify_nvlist(fsaclnv)) != 0) {
2973		nvlist_free(fsaclnv);
2974		return (SET_ERROR(EINVAL));
2975	}
2976
2977	/*
2978	 * If we don't have PRIV_SYS_MOUNT, then validate
2979	 * that user is allowed to hand out each permission in
2980	 * the nvlist(s)
2981	 */
2982
2983	error = secpolicy_zfs(CRED());
2984	if (error != 0) {
2985		if (zc->zc_perm_action == B_FALSE) {
2986			error = dsl_deleg_can_allow(zc->zc_name,
2987			    fsaclnv, CRED());
2988		} else {
2989			error = dsl_deleg_can_unallow(zc->zc_name,
2990			    fsaclnv, CRED());
2991		}
2992	}
2993
2994	if (error == 0)
2995		error = dsl_deleg_set(zc->zc_name, fsaclnv, zc->zc_perm_action);
2996
2997	nvlist_free(fsaclnv);
2998	return (error);
2999}
3000
3001/*
3002 * inputs:
3003 * zc_name		name of filesystem
3004 *
3005 * outputs:
3006 * zc_nvlist_src{_size}	nvlist of delegated permissions
3007 */
3008static int
3009zfs_ioc_get_fsacl(zfs_cmd_t *zc)
3010{
3011	nvlist_t *nvp;
3012	int error;
3013
3014	if ((error = dsl_deleg_get(zc->zc_name, &nvp)) == 0) {
3015		error = put_nvlist(zc, nvp);
3016		nvlist_free(nvp);
3017	}
3018
3019	return (error);
3020}
3021
3022/*
3023 * Search the vfs list for a specified resource.  Returns a pointer to it
3024 * or NULL if no suitable entry is found. The caller of this routine
3025 * is responsible for releasing the returned vfs pointer.
3026 */
3027static vfs_t *
3028zfs_get_vfs(const char *resource)
3029{
3030	vfs_t *vfsp;
3031
3032	mtx_lock(&mountlist_mtx);
3033	TAILQ_FOREACH(vfsp, &mountlist, mnt_list) {
3034		if (strcmp(refstr_value(vfsp->vfs_resource), resource) == 0) {
3035			VFS_HOLD(vfsp);
3036			break;
3037		}
3038	}
3039	mtx_unlock(&mountlist_mtx);
3040	return (vfsp);
3041}
3042
3043/* ARGSUSED */
3044static void
3045zfs_create_cb(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx)
3046{
3047	zfs_creat_t *zct = arg;
3048
3049	zfs_create_fs(os, cr, zct->zct_zplprops, tx);
3050}
3051
3052#define	ZFS_PROP_UNDEFINED	((uint64_t)-1)
3053
3054/*
3055 * inputs:
3056 * os			parent objset pointer (NULL if root fs)
3057 * fuids_ok		fuids allowed in this version of the spa?
3058 * sa_ok		SAs allowed in this version of the spa?
3059 * createprops		list of properties requested by creator
3060 *
3061 * outputs:
3062 * zplprops	values for the zplprops we attach to the master node object
3063 * is_ci	true if requested file system will be purely case-insensitive
3064 *
3065 * Determine the settings for utf8only, normalization and
3066 * casesensitivity.  Specific values may have been requested by the
3067 * creator and/or we can inherit values from the parent dataset.  If
3068 * the file system is of too early a vintage, a creator can not
3069 * request settings for these properties, even if the requested
3070 * setting is the default value.  We don't actually want to create dsl
3071 * properties for these, so remove them from the source nvlist after
3072 * processing.
3073 */
3074static int
3075zfs_fill_zplprops_impl(objset_t *os, uint64_t zplver,
3076    boolean_t fuids_ok, boolean_t sa_ok, nvlist_t *createprops,
3077    nvlist_t *zplprops, boolean_t *is_ci)
3078{
3079	uint64_t sense = ZFS_PROP_UNDEFINED;
3080	uint64_t norm = ZFS_PROP_UNDEFINED;
3081	uint64_t u8 = ZFS_PROP_UNDEFINED;
3082
3083	ASSERT(zplprops != NULL);
3084
3085	/*
3086	 * Pull out creator prop choices, if any.
3087	 */
3088	if (createprops) {
3089		(void) nvlist_lookup_uint64(createprops,
3090		    zfs_prop_to_name(ZFS_PROP_VERSION), &zplver);
3091		(void) nvlist_lookup_uint64(createprops,
3092		    zfs_prop_to_name(ZFS_PROP_NORMALIZE), &norm);
3093		(void) nvlist_remove_all(createprops,
3094		    zfs_prop_to_name(ZFS_PROP_NORMALIZE));
3095		(void) nvlist_lookup_uint64(createprops,
3096		    zfs_prop_to_name(ZFS_PROP_UTF8ONLY), &u8);
3097		(void) nvlist_remove_all(createprops,
3098		    zfs_prop_to_name(ZFS_PROP_UTF8ONLY));
3099		(void) nvlist_lookup_uint64(createprops,
3100		    zfs_prop_to_name(ZFS_PROP_CASE), &sense);
3101		(void) nvlist_remove_all(createprops,
3102		    zfs_prop_to_name(ZFS_PROP_CASE));
3103	}
3104
3105	/*
3106	 * If the zpl version requested is whacky or the file system
3107	 * or pool is version is too "young" to support normalization
3108	 * and the creator tried to set a value for one of the props,
3109	 * error out.
3110	 */
3111	if ((zplver < ZPL_VERSION_INITIAL || zplver > ZPL_VERSION) ||
3112	    (zplver >= ZPL_VERSION_FUID && !fuids_ok) ||
3113	    (zplver >= ZPL_VERSION_SA && !sa_ok) ||
3114	    (zplver < ZPL_VERSION_NORMALIZATION &&
3115	    (norm != ZFS_PROP_UNDEFINED || u8 != ZFS_PROP_UNDEFINED ||
3116	    sense != ZFS_PROP_UNDEFINED)))
3117		return (SET_ERROR(ENOTSUP));
3118
3119	/*
3120	 * Put the version in the zplprops
3121	 */
3122	VERIFY(nvlist_add_uint64(zplprops,
3123	    zfs_prop_to_name(ZFS_PROP_VERSION), zplver) == 0);
3124
3125	if (norm == ZFS_PROP_UNDEFINED)
3126		VERIFY(zfs_get_zplprop(os, ZFS_PROP_NORMALIZE, &norm) == 0);
3127	VERIFY(nvlist_add_uint64(zplprops,
3128	    zfs_prop_to_name(ZFS_PROP_NORMALIZE), norm) == 0);
3129
3130	/*
3131	 * If we're normalizing, names must always be valid UTF-8 strings.
3132	 */
3133	if (norm)
3134		u8 = 1;
3135	if (u8 == ZFS_PROP_UNDEFINED)
3136		VERIFY(zfs_get_zplprop(os, ZFS_PROP_UTF8ONLY, &u8) == 0);
3137	VERIFY(nvlist_add_uint64(zplprops,
3138	    zfs_prop_to_name(ZFS_PROP_UTF8ONLY), u8) == 0);
3139
3140	if (sense == ZFS_PROP_UNDEFINED)
3141		VERIFY(zfs_get_zplprop(os, ZFS_PROP_CASE, &sense) == 0);
3142	VERIFY(nvlist_add_uint64(zplprops,
3143	    zfs_prop_to_name(ZFS_PROP_CASE), sense) == 0);
3144
3145	if (is_ci)
3146		*is_ci = (sense == ZFS_CASE_INSENSITIVE);
3147
3148	return (0);
3149}
3150
3151static int
3152zfs_fill_zplprops(const char *dataset, nvlist_t *createprops,
3153    nvlist_t *zplprops, boolean_t *is_ci)
3154{
3155	boolean_t fuids_ok, sa_ok;
3156	uint64_t zplver = ZPL_VERSION;
3157	objset_t *os = NULL;
3158	char parentname[MAXNAMELEN];
3159	char *cp;
3160	spa_t *spa;
3161	uint64_t spa_vers;
3162	int error;
3163
3164	(void) strlcpy(parentname, dataset, sizeof (parentname));
3165	cp = strrchr(parentname, '/');
3166	ASSERT(cp != NULL);
3167	cp[0] = '\0';
3168
3169	if ((error = spa_open(dataset, &spa, FTAG)) != 0)
3170		return (error);
3171
3172	spa_vers = spa_version(spa);
3173	spa_close(spa, FTAG);
3174
3175	zplver = zfs_zpl_version_map(spa_vers);
3176	fuids_ok = (zplver >= ZPL_VERSION_FUID);
3177	sa_ok = (zplver >= ZPL_VERSION_SA);
3178
3179	/*
3180	 * Open parent object set so we can inherit zplprop values.
3181	 */
3182	if ((error = dmu_objset_hold(parentname, FTAG, &os)) != 0)
3183		return (error);
3184
3185	error = zfs_fill_zplprops_impl(os, zplver, fuids_ok, sa_ok, createprops,
3186	    zplprops, is_ci);
3187	dmu_objset_rele(os, FTAG);
3188	return (error);
3189}
3190
3191static int
3192zfs_fill_zplprops_root(uint64_t spa_vers, nvlist_t *createprops,
3193    nvlist_t *zplprops, boolean_t *is_ci)
3194{
3195	boolean_t fuids_ok;
3196	boolean_t sa_ok;
3197	uint64_t zplver = ZPL_VERSION;
3198	int error;
3199
3200	zplver = zfs_zpl_version_map(spa_vers);
3201	fuids_ok = (zplver >= ZPL_VERSION_FUID);
3202	sa_ok = (zplver >= ZPL_VERSION_SA);
3203
3204	error = zfs_fill_zplprops_impl(NULL, zplver, fuids_ok, sa_ok,
3205	    createprops, zplprops, is_ci);
3206	return (error);
3207}
3208
3209/*
3210 * innvl: {
3211 *     "type" -> dmu_objset_type_t (int32)
3212 *     (optional) "props" -> { prop -> value }
3213 * }
3214 *
3215 * outnvl: propname -> error code (int32)
3216 */
3217static int
3218zfs_ioc_create(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3219{
3220	int error = 0;
3221	zfs_creat_t zct = { 0 };
3222	nvlist_t *nvprops = NULL;
3223	void (*cbfunc)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx);
3224	int32_t type32;
3225	dmu_objset_type_t type;
3226	boolean_t is_insensitive = B_FALSE;
3227
3228	if (nvlist_lookup_int32(innvl, "type", &type32) != 0)
3229		return (SET_ERROR(EINVAL));
3230	type = type32;
3231	(void) nvlist_lookup_nvlist(innvl, "props", &nvprops);
3232
3233	switch (type) {
3234	case DMU_OST_ZFS:
3235		cbfunc = zfs_create_cb;
3236		break;
3237
3238	case DMU_OST_ZVOL:
3239		cbfunc = zvol_create_cb;
3240		break;
3241
3242	default:
3243		cbfunc = NULL;
3244		break;
3245	}
3246	if (strchr(fsname, '@') ||
3247	    strchr(fsname, '%'))
3248		return (SET_ERROR(EINVAL));
3249
3250	zct.zct_props = nvprops;
3251
3252	if (cbfunc == NULL)
3253		return (SET_ERROR(EINVAL));
3254
3255	if (type == DMU_OST_ZVOL) {
3256		uint64_t volsize, volblocksize;
3257
3258		if (nvprops == NULL)
3259			return (SET_ERROR(EINVAL));
3260		if (nvlist_lookup_uint64(nvprops,
3261		    zfs_prop_to_name(ZFS_PROP_VOLSIZE), &volsize) != 0)
3262			return (SET_ERROR(EINVAL));
3263
3264		if ((error = nvlist_lookup_uint64(nvprops,
3265		    zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE),
3266		    &volblocksize)) != 0 && error != ENOENT)
3267			return (SET_ERROR(EINVAL));
3268
3269		if (error != 0)
3270			volblocksize = zfs_prop_default_numeric(
3271			    ZFS_PROP_VOLBLOCKSIZE);
3272
3273		if ((error = zvol_check_volblocksize(
3274		    volblocksize)) != 0 ||
3275		    (error = zvol_check_volsize(volsize,
3276		    volblocksize)) != 0)
3277			return (error);
3278	} else if (type == DMU_OST_ZFS) {
3279		int error;
3280
3281		/*
3282		 * We have to have normalization and
3283		 * case-folding flags correct when we do the
3284		 * file system creation, so go figure them out
3285		 * now.
3286		 */
3287		VERIFY(nvlist_alloc(&zct.zct_zplprops,
3288		    NV_UNIQUE_NAME, KM_SLEEP) == 0);
3289		error = zfs_fill_zplprops(fsname, nvprops,
3290		    zct.zct_zplprops, &is_insensitive);
3291		if (error != 0) {
3292			nvlist_free(zct.zct_zplprops);
3293			return (error);
3294		}
3295	}
3296
3297	error = dmu_objset_create(fsname, type,
3298	    is_insensitive ? DS_FLAG_CI_DATASET : 0, cbfunc, &zct);
3299	nvlist_free(zct.zct_zplprops);
3300
3301	/*
3302	 * It would be nice to do this atomically.
3303	 */
3304	if (error == 0) {
3305		error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL,
3306		    nvprops, outnvl);
3307		if (error != 0)
3308			(void) dsl_destroy_head(fsname);
3309	}
3310#ifdef __FreeBSD__
3311	if (error == 0 && type == DMU_OST_ZVOL)
3312		zvol_create_minors(fsname);
3313#endif
3314	return (error);
3315}
3316
3317/*
3318 * innvl: {
3319 *     "origin" -> name of origin snapshot
3320 *     (optional) "props" -> { prop -> value }
3321 * }
3322 *
3323 * outnvl: propname -> error code (int32)
3324 */
3325static int
3326zfs_ioc_clone(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3327{
3328	int error = 0;
3329	nvlist_t *nvprops = NULL;
3330	char *origin_name;
3331
3332	if (nvlist_lookup_string(innvl, "origin", &origin_name) != 0)
3333		return (SET_ERROR(EINVAL));
3334	(void) nvlist_lookup_nvlist(innvl, "props", &nvprops);
3335
3336	if (strchr(fsname, '@') ||
3337	    strchr(fsname, '%'))
3338		return (SET_ERROR(EINVAL));
3339
3340	if (dataset_namecheck(origin_name, NULL, NULL) != 0)
3341		return (SET_ERROR(EINVAL));
3342	error = dmu_objset_clone(fsname, origin_name);
3343	if (error != 0)
3344		return (error);
3345
3346	/*
3347	 * It would be nice to do this atomically.
3348	 */
3349	if (error == 0) {
3350		error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL,
3351		    nvprops, outnvl);
3352		if (error != 0)
3353			(void) dsl_destroy_head(fsname);
3354	}
3355#ifdef __FreeBSD__
3356	if (error == 0)
3357		zvol_create_minors(fsname);
3358#endif
3359	return (error);
3360}
3361
3362/*
3363 * innvl: {
3364 *     "snaps" -> { snapshot1, snapshot2 }
3365 *     (optional) "props" -> { prop -> value (string) }
3366 * }
3367 *
3368 * outnvl: snapshot -> error code (int32)
3369 */
3370static int
3371zfs_ioc_snapshot(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3372{
3373	nvlist_t *snaps;
3374	nvlist_t *props = NULL;
3375	int error, poollen;
3376	nvpair_t *pair;
3377
3378	(void) nvlist_lookup_nvlist(innvl, "props", &props);
3379	if ((error = zfs_check_userprops(poolname, props)) != 0)
3380		return (error);
3381
3382	if (!nvlist_empty(props) &&
3383	    zfs_earlier_version(poolname, SPA_VERSION_SNAP_PROPS))
3384		return (SET_ERROR(ENOTSUP));
3385
3386	if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
3387		return (SET_ERROR(EINVAL));
3388	poollen = strlen(poolname);
3389	for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
3390	    pair = nvlist_next_nvpair(snaps, pair)) {
3391		const char *name = nvpair_name(pair);
3392		const char *cp = strchr(name, '@');
3393
3394		/*
3395		 * The snap name must contain an @, and the part after it must
3396		 * contain only valid characters.
3397		 */
3398		if (cp == NULL ||
3399		    zfs_component_namecheck(cp + 1, NULL, NULL) != 0)
3400			return (SET_ERROR(EINVAL));
3401
3402		/*
3403		 * The snap must be in the specified pool.
3404		 */
3405		if (strncmp(name, poolname, poollen) != 0 ||
3406		    (name[poollen] != '/' && name[poollen] != '@'))
3407			return (SET_ERROR(EXDEV));
3408
3409		/* This must be the only snap of this fs. */
3410		for (nvpair_t *pair2 = nvlist_next_nvpair(snaps, pair);
3411		    pair2 != NULL; pair2 = nvlist_next_nvpair(snaps, pair2)) {
3412			if (strncmp(name, nvpair_name(pair2), cp - name + 1)
3413			    == 0) {
3414				return (SET_ERROR(EXDEV));
3415			}
3416		}
3417	}
3418
3419	error = dsl_dataset_snapshot(snaps, props, outnvl);
3420	return (error);
3421}
3422
3423/*
3424 * innvl: "message" -> string
3425 */
3426/* ARGSUSED */
3427static int
3428zfs_ioc_log_history(const char *unused, nvlist_t *innvl, nvlist_t *outnvl)
3429{
3430	char *message;
3431	spa_t *spa;
3432	int error;
3433	char *poolname;
3434
3435	/*
3436	 * The poolname in the ioctl is not set, we get it from the TSD,
3437	 * which was set at the end of the last successful ioctl that allows
3438	 * logging.  The secpolicy func already checked that it is set.
3439	 * Only one log ioctl is allowed after each successful ioctl, so
3440	 * we clear the TSD here.
3441	 */
3442	poolname = tsd_get(zfs_allow_log_key);
3443	(void) tsd_set(zfs_allow_log_key, NULL);
3444	error = spa_open(poolname, &spa, FTAG);
3445	strfree(poolname);
3446	if (error != 0)
3447		return (error);
3448
3449	if (nvlist_lookup_string(innvl, "message", &message) != 0)  {
3450		spa_close(spa, FTAG);
3451		return (SET_ERROR(EINVAL));
3452	}
3453
3454	if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) {
3455		spa_close(spa, FTAG);
3456		return (SET_ERROR(ENOTSUP));
3457	}
3458
3459	error = spa_history_log(spa, message);
3460	spa_close(spa, FTAG);
3461	return (error);
3462}
3463
3464/*
3465 * The dp_config_rwlock must not be held when calling this, because the
3466 * unmount may need to write out data.
3467 *
3468 * This function is best-effort.  Callers must deal gracefully if it
3469 * remains mounted (or is remounted after this call).
3470 *
3471 * Returns 0 if the argument is not a snapshot, or it is not currently a
3472 * filesystem, or we were able to unmount it.  Returns error code otherwise.
3473 */
3474int
3475zfs_unmount_snap(const char *snapname)
3476{
3477	vfs_t *vfsp;
3478	zfsvfs_t *zfsvfs;
3479	int err;
3480
3481	if (strchr(snapname, '@') == NULL)
3482		return (0);
3483
3484	vfsp = zfs_get_vfs(snapname);
3485	if (vfsp == NULL)
3486		return (0);
3487
3488	zfsvfs = vfsp->vfs_data;
3489	ASSERT(!dsl_pool_config_held(dmu_objset_pool(zfsvfs->z_os)));
3490
3491	err = vn_vfswlock(vfsp->vfs_vnodecovered);
3492	VFS_RELE(vfsp);
3493	if (err != 0)
3494		return (SET_ERROR(err));
3495
3496	/*
3497	 * Always force the unmount for snapshots.
3498	 */
3499
3500#ifdef illumos
3501	(void) dounmount(vfsp, MS_FORCE, kcred);
3502#else
3503	mtx_lock(&Giant);	/* dounmount() */
3504	(void) dounmount(vfsp, MS_FORCE, curthread);
3505	mtx_unlock(&Giant);	/* dounmount() */
3506#endif
3507	return (0);
3508}
3509
3510/* ARGSUSED */
3511static int
3512zfs_unmount_snap_cb(const char *snapname, void *arg)
3513{
3514	return (zfs_unmount_snap(snapname));
3515}
3516
3517/*
3518 * When a clone is destroyed, its origin may also need to be destroyed,
3519 * in which case it must be unmounted.  This routine will do that unmount
3520 * if necessary.
3521 */
3522void
3523zfs_destroy_unmount_origin(const char *fsname)
3524{
3525	int error;
3526	objset_t *os;
3527	dsl_dataset_t *ds;
3528
3529	error = dmu_objset_hold(fsname, FTAG, &os);
3530	if (error != 0)
3531		return;
3532	ds = dmu_objset_ds(os);
3533	if (dsl_dir_is_clone(ds->ds_dir) && DS_IS_DEFER_DESTROY(ds->ds_prev)) {
3534		char originname[MAXNAMELEN];
3535		dsl_dataset_name(ds->ds_prev, originname);
3536		dmu_objset_rele(os, FTAG);
3537		(void) zfs_unmount_snap(originname);
3538	} else {
3539		dmu_objset_rele(os, FTAG);
3540	}
3541}
3542
3543/*
3544 * innvl: {
3545 *     "snaps" -> { snapshot1, snapshot2 }
3546 *     (optional boolean) "defer"
3547 * }
3548 *
3549 * outnvl: snapshot -> error code (int32)
3550 *
3551 */
3552/* ARGSUSED */
3553static int
3554zfs_ioc_destroy_snaps(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3555{
3556	nvlist_t *snaps;
3557	nvpair_t *pair;
3558	boolean_t defer;
3559
3560	if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
3561		return (SET_ERROR(EINVAL));
3562	defer = nvlist_exists(innvl, "defer");
3563
3564	for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
3565	    pair = nvlist_next_nvpair(snaps, pair)) {
3566		(void) zfs_unmount_snap(nvpair_name(pair));
3567	}
3568
3569	return (dsl_destroy_snapshots_nvl(snaps, defer, outnvl));
3570}
3571
3572/*
3573 * Create bookmarks.  Bookmark names are of the form <fs>#<bmark>.
3574 * All bookmarks must be in the same pool.
3575 *
3576 * innvl: {
3577 *     bookmark1 -> snapshot1, bookmark2 -> snapshot2
3578 * }
3579 *
3580 * outnvl: bookmark -> error code (int32)
3581 *
3582 */
3583/* ARGSUSED */
3584static int
3585zfs_ioc_bookmark(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3586{
3587	for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL);
3588	    pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) {
3589		char *snap_name;
3590
3591		/*
3592		 * Verify the snapshot argument.
3593		 */
3594		if (nvpair_value_string(pair, &snap_name) != 0)
3595			return (SET_ERROR(EINVAL));
3596
3597
3598		/* Verify that the keys (bookmarks) are unique */
3599		for (nvpair_t *pair2 = nvlist_next_nvpair(innvl, pair);
3600		    pair2 != NULL; pair2 = nvlist_next_nvpair(innvl, pair2)) {
3601			if (strcmp(nvpair_name(pair), nvpair_name(pair2)) == 0)
3602				return (SET_ERROR(EINVAL));
3603		}
3604	}
3605
3606	return (dsl_bookmark_create(innvl, outnvl));
3607}
3608
3609/*
3610 * innvl: {
3611 *     property 1, property 2, ...
3612 * }
3613 *
3614 * outnvl: {
3615 *     bookmark name 1 -> { property 1, property 2, ... },
3616 *     bookmark name 2 -> { property 1, property 2, ... }
3617 * }
3618 *
3619 */
3620static int
3621zfs_ioc_get_bookmarks(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3622{
3623	return (dsl_get_bookmarks(fsname, innvl, outnvl));
3624}
3625
3626/*
3627 * innvl: {
3628 *     bookmark name 1, bookmark name 2
3629 * }
3630 *
3631 * outnvl: bookmark -> error code (int32)
3632 *
3633 */
3634static int
3635zfs_ioc_destroy_bookmarks(const char *poolname, nvlist_t *innvl,
3636    nvlist_t *outnvl)
3637{
3638	int error, poollen;
3639
3640	poollen = strlen(poolname);
3641	for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL);
3642	    pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) {
3643		const char *name = nvpair_name(pair);
3644		const char *cp = strchr(name, '#');
3645
3646		/*
3647		 * The bookmark name must contain an #, and the part after it
3648		 * must contain only valid characters.
3649		 */
3650		if (cp == NULL ||
3651		    zfs_component_namecheck(cp + 1, NULL, NULL) != 0)
3652			return (SET_ERROR(EINVAL));
3653
3654		/*
3655		 * The bookmark must be in the specified pool.
3656		 */
3657		if (strncmp(name, poolname, poollen) != 0 ||
3658		    (name[poollen] != '/' && name[poollen] != '#'))
3659			return (SET_ERROR(EXDEV));
3660		(void) zvol_remove_minor(name);
3661	}
3662
3663	error = dsl_bookmark_destroy(innvl, outnvl);
3664	return (error);
3665}
3666
3667/*
3668 * inputs:
3669 * zc_name		name of dataset to destroy
3670 * zc_objset_type	type of objset
3671 * zc_defer_destroy	mark for deferred destroy
3672 *
3673 * outputs:		none
3674 */
3675static int
3676zfs_ioc_destroy(zfs_cmd_t *zc)
3677{
3678	int err;
3679
3680	if (zc->zc_objset_type == DMU_OST_ZFS) {
3681		err = zfs_unmount_snap(zc->zc_name);
3682		if (err != 0)
3683			return (err);
3684	}
3685
3686	if (strchr(zc->zc_name, '@'))
3687		err = dsl_destroy_snapshot(zc->zc_name, zc->zc_defer_destroy);
3688	else
3689		err = dsl_destroy_head(zc->zc_name);
3690	if (zc->zc_objset_type == DMU_OST_ZVOL && err == 0)
3691		(void) zvol_remove_minor(zc->zc_name);
3692	return (err);
3693}
3694
3695/*
3696 * fsname is name of dataset to rollback (to most recent snapshot)
3697 *
3698 * innvl is not used.
3699 *
3700 * outnvl: "target" -> name of most recent snapshot
3701 * }
3702 */
3703/* ARGSUSED */
3704static int
3705zfs_ioc_rollback(const char *fsname, nvlist_t *args, nvlist_t *outnvl)
3706{
3707	zfsvfs_t *zfsvfs;
3708	int error;
3709
3710	if (getzfsvfs(fsname, &zfsvfs) == 0) {
3711		error = zfs_suspend_fs(zfsvfs);
3712		if (error == 0) {
3713			int resume_err;
3714
3715			error = dsl_dataset_rollback(fsname, zfsvfs, outnvl);
3716			resume_err = zfs_resume_fs(zfsvfs, fsname);
3717			error = error ? error : resume_err;
3718		}
3719		VFS_RELE(zfsvfs->z_vfs);
3720	} else {
3721		error = dsl_dataset_rollback(fsname, NULL, outnvl);
3722	}
3723	return (error);
3724}
3725
3726static int
3727recursive_unmount(const char *fsname, void *arg)
3728{
3729	const char *snapname = arg;
3730	char fullname[MAXNAMELEN];
3731
3732	(void) snprintf(fullname, sizeof (fullname), "%s@%s", fsname, snapname);
3733	return (zfs_unmount_snap(fullname));
3734}
3735
3736/*
3737 * inputs:
3738 * zc_name	old name of dataset
3739 * zc_value	new name of dataset
3740 * zc_cookie	recursive flag (only valid for snapshots)
3741 *
3742 * outputs:	none
3743 */
3744static int
3745zfs_ioc_rename(zfs_cmd_t *zc)
3746{
3747	boolean_t recursive = zc->zc_cookie & 1;
3748#ifdef __FreeBSD__
3749	boolean_t allow_mounted = zc->zc_cookie & 2;
3750#endif
3751	char *at;
3752
3753	zc->zc_value[sizeof (zc->zc_value) - 1] = '\0';
3754	if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
3755	    strchr(zc->zc_value, '%'))
3756		return (SET_ERROR(EINVAL));
3757
3758	at = strchr(zc->zc_name, '@');
3759	if (at != NULL) {
3760		/* snaps must be in same fs */
3761		int error;
3762
3763		if (strncmp(zc->zc_name, zc->zc_value, at - zc->zc_name + 1))
3764			return (SET_ERROR(EXDEV));
3765		*at = '\0';
3766#ifdef illumos
3767		if (zc->zc_objset_type == DMU_OST_ZFS) {
3768#else
3769		if (zc->zc_objset_type == DMU_OST_ZFS && allow_mounted) {
3770#endif
3771			error = dmu_objset_find(zc->zc_name,
3772			    recursive_unmount, at + 1,
3773			    recursive ? DS_FIND_CHILDREN : 0);
3774			if (error != 0) {
3775				*at = '@';
3776				return (error);
3777			}
3778		}
3779		error = dsl_dataset_rename_snapshot(zc->zc_name,
3780		    at + 1, strchr(zc->zc_value, '@') + 1, recursive);
3781		*at = '@';
3782
3783		return (error);
3784	} else {
3785#ifdef illumos
3786		if (zc->zc_objset_type == DMU_OST_ZVOL)
3787			(void) zvol_remove_minor(zc->zc_name);
3788#endif
3789		return (dsl_dir_rename(zc->zc_name, zc->zc_value));
3790	}
3791}
3792
3793static int
3794zfs_check_settable(const char *dsname, nvpair_t *pair, cred_t *cr)
3795{
3796	const char *propname = nvpair_name(pair);
3797	boolean_t issnap = (strchr(dsname, '@') != NULL);
3798	zfs_prop_t prop = zfs_name_to_prop(propname);
3799	uint64_t intval;
3800	int err;
3801
3802	if (prop == ZPROP_INVAL) {
3803		if (zfs_prop_user(propname)) {
3804			if (err = zfs_secpolicy_write_perms(dsname,
3805			    ZFS_DELEG_PERM_USERPROP, cr))
3806				return (err);
3807			return (0);
3808		}
3809
3810		if (!issnap && zfs_prop_userquota(propname)) {
3811			const char *perm = NULL;
3812			const char *uq_prefix =
3813			    zfs_userquota_prop_prefixes[ZFS_PROP_USERQUOTA];
3814			const char *gq_prefix =
3815			    zfs_userquota_prop_prefixes[ZFS_PROP_GROUPQUOTA];
3816
3817			if (strncmp(propname, uq_prefix,
3818			    strlen(uq_prefix)) == 0) {
3819				perm = ZFS_DELEG_PERM_USERQUOTA;
3820			} else if (strncmp(propname, gq_prefix,
3821			    strlen(gq_prefix)) == 0) {
3822				perm = ZFS_DELEG_PERM_GROUPQUOTA;
3823			} else {
3824				/* USERUSED and GROUPUSED are read-only */
3825				return (SET_ERROR(EINVAL));
3826			}
3827
3828			if (err = zfs_secpolicy_write_perms(dsname, perm, cr))
3829				return (err);
3830			return (0);
3831		}
3832
3833		return (SET_ERROR(EINVAL));
3834	}
3835
3836	if (issnap)
3837		return (SET_ERROR(EINVAL));
3838
3839	if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
3840		/*
3841		 * dsl_prop_get_all_impl() returns properties in this
3842		 * format.
3843		 */
3844		nvlist_t *attrs;
3845		VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
3846		VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
3847		    &pair) == 0);
3848	}
3849
3850	/*
3851	 * Check that this value is valid for this pool version
3852	 */
3853	switch (prop) {
3854	case ZFS_PROP_COMPRESSION:
3855		/*
3856		 * If the user specified gzip compression, make sure
3857		 * the SPA supports it. We ignore any errors here since
3858		 * we'll catch them later.
3859		 */
3860		if (nvpair_type(pair) == DATA_TYPE_UINT64 &&
3861		    nvpair_value_uint64(pair, &intval) == 0) {
3862			if (intval >= ZIO_COMPRESS_GZIP_1 &&
3863			    intval <= ZIO_COMPRESS_GZIP_9 &&
3864			    zfs_earlier_version(dsname,
3865			    SPA_VERSION_GZIP_COMPRESSION)) {
3866				return (SET_ERROR(ENOTSUP));
3867			}
3868
3869			if (intval == ZIO_COMPRESS_ZLE &&
3870			    zfs_earlier_version(dsname,
3871			    SPA_VERSION_ZLE_COMPRESSION))
3872				return (SET_ERROR(ENOTSUP));
3873
3874			if (intval == ZIO_COMPRESS_LZ4) {
3875				spa_t *spa;
3876
3877				if ((err = spa_open(dsname, &spa, FTAG)) != 0)
3878					return (err);
3879
3880				if (!spa_feature_is_enabled(spa,
3881				    SPA_FEATURE_LZ4_COMPRESS)) {
3882					spa_close(spa, FTAG);
3883					return (SET_ERROR(ENOTSUP));
3884				}
3885				spa_close(spa, FTAG);
3886			}
3887
3888			/*
3889			 * If this is a bootable dataset then
3890			 * verify that the compression algorithm
3891			 * is supported for booting. We must return
3892			 * something other than ENOTSUP since it
3893			 * implies a downrev pool version.
3894			 */
3895			if (zfs_is_bootfs(dsname) &&
3896			    !BOOTFS_COMPRESS_VALID(intval)) {
3897				return (SET_ERROR(ERANGE));
3898			}
3899		}
3900		break;
3901
3902	case ZFS_PROP_COPIES:
3903		if (zfs_earlier_version(dsname, SPA_VERSION_DITTO_BLOCKS))
3904			return (SET_ERROR(ENOTSUP));
3905		break;
3906
3907	case ZFS_PROP_DEDUP:
3908		if (zfs_earlier_version(dsname, SPA_VERSION_DEDUP))
3909			return (SET_ERROR(ENOTSUP));
3910		break;
3911
3912	case ZFS_PROP_SHARESMB:
3913		if (zpl_earlier_version(dsname, ZPL_VERSION_FUID))
3914			return (SET_ERROR(ENOTSUP));
3915		break;
3916
3917	case ZFS_PROP_ACLINHERIT:
3918		if (nvpair_type(pair) == DATA_TYPE_UINT64 &&
3919		    nvpair_value_uint64(pair, &intval) == 0) {
3920			if (intval == ZFS_ACL_PASSTHROUGH_X &&
3921			    zfs_earlier_version(dsname,
3922			    SPA_VERSION_PASSTHROUGH_X))
3923				return (SET_ERROR(ENOTSUP));
3924		}
3925		break;
3926	}
3927
3928	return (zfs_secpolicy_setprop(dsname, prop, pair, CRED()));
3929}
3930
3931/*
3932 * Checks for a race condition to make sure we don't increment a feature flag
3933 * multiple times.
3934 */
3935static int
3936zfs_prop_activate_feature_check(void *arg, dmu_tx_t *tx)
3937{
3938	spa_t *spa = dmu_tx_pool(tx)->dp_spa;
3939	spa_feature_t *featurep = arg;
3940
3941	if (!spa_feature_is_active(spa, *featurep))
3942		return (0);
3943	else
3944		return (SET_ERROR(EBUSY));
3945}
3946
3947/*
3948 * The callback invoked on feature activation in the sync task caused by
3949 * zfs_prop_activate_feature.
3950 */
3951static void
3952zfs_prop_activate_feature_sync(void *arg, dmu_tx_t *tx)
3953{
3954	spa_t *spa = dmu_tx_pool(tx)->dp_spa;
3955	spa_feature_t *featurep = arg;
3956
3957	spa_feature_incr(spa, *featurep, tx);
3958}
3959
3960/*
3961 * Activates a feature on a pool in response to a property setting. This
3962 * creates a new sync task which modifies the pool to reflect the feature
3963 * as being active.
3964 */
3965static int
3966zfs_prop_activate_feature(spa_t *spa, spa_feature_t feature)
3967{
3968	int err;
3969
3970	/* EBUSY here indicates that the feature is already active */
3971	err = dsl_sync_task(spa_name(spa),
3972	    zfs_prop_activate_feature_check, zfs_prop_activate_feature_sync,
3973	    &feature, 2);
3974
3975	if (err != 0 && err != EBUSY)
3976		return (err);
3977	else
3978		return (0);
3979}
3980
3981/*
3982 * Removes properties from the given props list that fail permission checks
3983 * needed to clear them and to restore them in case of a receive error. For each
3984 * property, make sure we have both set and inherit permissions.
3985 *
3986 * Returns the first error encountered if any permission checks fail. If the
3987 * caller provides a non-NULL errlist, it also gives the complete list of names
3988 * of all the properties that failed a permission check along with the
3989 * corresponding error numbers. The caller is responsible for freeing the
3990 * returned errlist.
3991 *
3992 * If every property checks out successfully, zero is returned and the list
3993 * pointed at by errlist is NULL.
3994 */
3995static int
3996zfs_check_clearable(char *dataset, nvlist_t *props, nvlist_t **errlist)
3997{
3998	zfs_cmd_t *zc;
3999	nvpair_t *pair, *next_pair;
4000	nvlist_t *errors;
4001	int err, rv = 0;
4002
4003	if (props == NULL)
4004		return (0);
4005
4006	VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0);
4007
4008	zc = kmem_alloc(sizeof (zfs_cmd_t), KM_SLEEP);
4009	(void) strcpy(zc->zc_name, dataset);
4010	pair = nvlist_next_nvpair(props, NULL);
4011	while (pair != NULL) {
4012		next_pair = nvlist_next_nvpair(props, pair);
4013
4014		(void) strcpy(zc->zc_value, nvpair_name(pair));
4015		if ((err = zfs_check_settable(dataset, pair, CRED())) != 0 ||
4016		    (err = zfs_secpolicy_inherit_prop(zc, NULL, CRED())) != 0) {
4017			VERIFY(nvlist_remove_nvpair(props, pair) == 0);
4018			VERIFY(nvlist_add_int32(errors,
4019			    zc->zc_value, err) == 0);
4020		}
4021		pair = next_pair;
4022	}
4023	kmem_free(zc, sizeof (zfs_cmd_t));
4024
4025	if ((pair = nvlist_next_nvpair(errors, NULL)) == NULL) {
4026		nvlist_free(errors);
4027		errors = NULL;
4028	} else {
4029		VERIFY(nvpair_value_int32(pair, &rv) == 0);
4030	}
4031
4032	if (errlist == NULL)
4033		nvlist_free(errors);
4034	else
4035		*errlist = errors;
4036
4037	return (rv);
4038}
4039
4040static boolean_t
4041propval_equals(nvpair_t *p1, nvpair_t *p2)
4042{
4043	if (nvpair_type(p1) == DATA_TYPE_NVLIST) {
4044		/* dsl_prop_get_all_impl() format */
4045		nvlist_t *attrs;
4046		VERIFY(nvpair_value_nvlist(p1, &attrs) == 0);
4047		VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
4048		    &p1) == 0);
4049	}
4050
4051	if (nvpair_type(p2) == DATA_TYPE_NVLIST) {
4052		nvlist_t *attrs;
4053		VERIFY(nvpair_value_nvlist(p2, &attrs) == 0);
4054		VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
4055		    &p2) == 0);
4056	}
4057
4058	if (nvpair_type(p1) != nvpair_type(p2))
4059		return (B_FALSE);
4060
4061	if (nvpair_type(p1) == DATA_TYPE_STRING) {
4062		char *valstr1, *valstr2;
4063
4064		VERIFY(nvpair_value_string(p1, (char **)&valstr1) == 0);
4065		VERIFY(nvpair_value_string(p2, (char **)&valstr2) == 0);
4066		return (strcmp(valstr1, valstr2) == 0);
4067	} else {
4068		uint64_t intval1, intval2;
4069
4070		VERIFY(nvpair_value_uint64(p1, &intval1) == 0);
4071		VERIFY(nvpair_value_uint64(p2, &intval2) == 0);
4072		return (intval1 == intval2);
4073	}
4074}
4075
4076/*
4077 * Remove properties from props if they are not going to change (as determined
4078 * by comparison with origprops). Remove them from origprops as well, since we
4079 * do not need to clear or restore properties that won't change.
4080 */
4081static void
4082props_reduce(nvlist_t *props, nvlist_t *origprops)
4083{
4084	nvpair_t *pair, *next_pair;
4085
4086	if (origprops == NULL)
4087		return; /* all props need to be received */
4088
4089	pair = nvlist_next_nvpair(props, NULL);
4090	while (pair != NULL) {
4091		const char *propname = nvpair_name(pair);
4092		nvpair_t *match;
4093
4094		next_pair = nvlist_next_nvpair(props, pair);
4095
4096		if ((nvlist_lookup_nvpair(origprops, propname,
4097		    &match) != 0) || !propval_equals(pair, match))
4098			goto next; /* need to set received value */
4099
4100		/* don't clear the existing received value */
4101		(void) nvlist_remove_nvpair(origprops, match);
4102		/* don't bother receiving the property */
4103		(void) nvlist_remove_nvpair(props, pair);
4104next:
4105		pair = next_pair;
4106	}
4107}
4108
4109#ifdef	DEBUG
4110static boolean_t zfs_ioc_recv_inject_err;
4111#endif
4112
4113/*
4114 * inputs:
4115 * zc_name		name of containing filesystem
4116 * zc_nvlist_src{_size}	nvlist of properties to apply
4117 * zc_value		name of snapshot to create
4118 * zc_string		name of clone origin (if DRR_FLAG_CLONE)
4119 * zc_cookie		file descriptor to recv from
4120 * zc_begin_record	the BEGIN record of the stream (not byteswapped)
4121 * zc_guid		force flag
4122 * zc_cleanup_fd	cleanup-on-exit file descriptor
4123 * zc_action_handle	handle for this guid/ds mapping (or zero on first call)
4124 *
4125 * outputs:
4126 * zc_cookie		number of bytes read
4127 * zc_nvlist_dst{_size} error for each unapplied received property
4128 * zc_obj		zprop_errflags_t
4129 * zc_action_handle	handle for this guid/ds mapping
4130 */
4131static int
4132zfs_ioc_recv(zfs_cmd_t *zc)
4133{
4134	file_t *fp;
4135	dmu_recv_cookie_t drc;
4136	boolean_t force = (boolean_t)zc->zc_guid;
4137	int fd;
4138	int error = 0;
4139	int props_error = 0;
4140	nvlist_t *errors;
4141	offset_t off;
4142	nvlist_t *props = NULL; /* sent properties */
4143	nvlist_t *origprops = NULL; /* existing properties */
4144	char *origin = NULL;
4145	char *tosnap;
4146	char tofs[ZFS_MAXNAMELEN];
4147	cap_rights_t rights;
4148	boolean_t first_recvd_props = B_FALSE;
4149
4150	if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
4151	    strchr(zc->zc_value, '@') == NULL ||
4152	    strchr(zc->zc_value, '%'))
4153		return (SET_ERROR(EINVAL));
4154
4155	(void) strcpy(tofs, zc->zc_value);
4156	tosnap = strchr(tofs, '@');
4157	*tosnap++ = '\0';
4158
4159	if (zc->zc_nvlist_src != 0 &&
4160	    (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
4161	    zc->zc_iflags, &props)) != 0)
4162		return (error);
4163
4164	fd = zc->zc_cookie;
4165	fp = getf(fd, cap_rights_init(&rights, CAP_PREAD));
4166	if (fp == NULL) {
4167		nvlist_free(props);
4168		return (SET_ERROR(EBADF));
4169	}
4170
4171	VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0);
4172
4173	if (zc->zc_string[0])
4174		origin = zc->zc_string;
4175
4176	error = dmu_recv_begin(tofs, tosnap,
4177	    &zc->zc_begin_record, force, origin, &drc);
4178	if (error != 0)
4179		goto out;
4180
4181	/*
4182	 * Set properties before we receive the stream so that they are applied
4183	 * to the new data. Note that we must call dmu_recv_stream() if
4184	 * dmu_recv_begin() succeeds.
4185	 */
4186	if (props != NULL && !drc.drc_newfs) {
4187		if (spa_version(dsl_dataset_get_spa(drc.drc_ds)) >=
4188		    SPA_VERSION_RECVD_PROPS &&
4189		    !dsl_prop_get_hasrecvd(tofs))
4190			first_recvd_props = B_TRUE;
4191
4192		/*
4193		 * If new received properties are supplied, they are to
4194		 * completely replace the existing received properties, so stash
4195		 * away the existing ones.
4196		 */
4197		if (dsl_prop_get_received(tofs, &origprops) == 0) {
4198			nvlist_t *errlist = NULL;
4199			/*
4200			 * Don't bother writing a property if its value won't
4201			 * change (and avoid the unnecessary security checks).
4202			 *
4203			 * The first receive after SPA_VERSION_RECVD_PROPS is a
4204			 * special case where we blow away all local properties
4205			 * regardless.
4206			 */
4207			if (!first_recvd_props)
4208				props_reduce(props, origprops);
4209			if (zfs_check_clearable(tofs, origprops, &errlist) != 0)
4210				(void) nvlist_merge(errors, errlist, 0);
4211			nvlist_free(errlist);
4212
4213			if (clear_received_props(tofs, origprops,
4214			    first_recvd_props ? NULL : props) != 0)
4215				zc->zc_obj |= ZPROP_ERR_NOCLEAR;
4216		} else {
4217			zc->zc_obj |= ZPROP_ERR_NOCLEAR;
4218		}
4219	}
4220
4221	if (props != NULL) {
4222		props_error = dsl_prop_set_hasrecvd(tofs);
4223
4224		if (props_error == 0) {
4225			(void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_RECEIVED,
4226			    props, errors);
4227		}
4228	}
4229
4230	if (zc->zc_nvlist_dst_size != 0 &&
4231	    (nvlist_smush(errors, zc->zc_nvlist_dst_size) != 0 ||
4232	    put_nvlist(zc, errors) != 0)) {
4233		/*
4234		 * Caller made zc->zc_nvlist_dst less than the minimum expected
4235		 * size or supplied an invalid address.
4236		 */
4237		props_error = SET_ERROR(EINVAL);
4238	}
4239
4240	off = fp->f_offset;
4241	error = dmu_recv_stream(&drc, fp, &off, zc->zc_cleanup_fd,
4242	    &zc->zc_action_handle);
4243
4244	if (error == 0) {
4245		zfsvfs_t *zfsvfs = NULL;
4246
4247		if (getzfsvfs(tofs, &zfsvfs) == 0) {
4248			/* online recv */
4249			int end_err;
4250
4251			error = zfs_suspend_fs(zfsvfs);
4252			/*
4253			 * If the suspend fails, then the recv_end will
4254			 * likely also fail, and clean up after itself.
4255			 */
4256			end_err = dmu_recv_end(&drc, zfsvfs);
4257			if (error == 0)
4258				error = zfs_resume_fs(zfsvfs, tofs);
4259			error = error ? error : end_err;
4260			VFS_RELE(zfsvfs->z_vfs);
4261		} else {
4262			error = dmu_recv_end(&drc, NULL);
4263		}
4264	}
4265
4266	zc->zc_cookie = off - fp->f_offset;
4267	if (off >= 0 && off <= MAXOFFSET_T)
4268		fp->f_offset = off;
4269
4270#ifdef	DEBUG
4271	if (zfs_ioc_recv_inject_err) {
4272		zfs_ioc_recv_inject_err = B_FALSE;
4273		error = 1;
4274	}
4275#endif
4276
4277#ifdef __FreeBSD__
4278	if (error == 0)
4279		zvol_create_minors(tofs);
4280#endif
4281
4282	/*
4283	 * On error, restore the original props.
4284	 */
4285	if (error != 0 && props != NULL && !drc.drc_newfs) {
4286		if (clear_received_props(tofs, props, NULL) != 0) {
4287			/*
4288			 * We failed to clear the received properties.
4289			 * Since we may have left a $recvd value on the
4290			 * system, we can't clear the $hasrecvd flag.
4291			 */
4292			zc->zc_obj |= ZPROP_ERR_NORESTORE;
4293		} else if (first_recvd_props) {
4294			dsl_prop_unset_hasrecvd(tofs);
4295		}
4296
4297		if (origprops == NULL && !drc.drc_newfs) {
4298			/* We failed to stash the original properties. */
4299			zc->zc_obj |= ZPROP_ERR_NORESTORE;
4300		}
4301
4302		/*
4303		 * dsl_props_set() will not convert RECEIVED to LOCAL on or
4304		 * after SPA_VERSION_RECVD_PROPS, so we need to specify LOCAL
4305		 * explictly if we're restoring local properties cleared in the
4306		 * first new-style receive.
4307		 */
4308		if (origprops != NULL &&
4309		    zfs_set_prop_nvlist(tofs, (first_recvd_props ?
4310		    ZPROP_SRC_LOCAL : ZPROP_SRC_RECEIVED),
4311		    origprops, NULL) != 0) {
4312			/*
4313			 * We stashed the original properties but failed to
4314			 * restore them.
4315			 */
4316			zc->zc_obj |= ZPROP_ERR_NORESTORE;
4317		}
4318	}
4319out:
4320	nvlist_free(props);
4321	nvlist_free(origprops);
4322	nvlist_free(errors);
4323	releasef(fd);
4324
4325	if (error == 0)
4326		error = props_error;
4327
4328	return (error);
4329}
4330
4331/*
4332 * inputs:
4333 * zc_name	name of snapshot to send
4334 * zc_cookie	file descriptor to send stream to
4335 * zc_obj	fromorigin flag (mutually exclusive with zc_fromobj)
4336 * zc_sendobj	objsetid of snapshot to send
4337 * zc_fromobj	objsetid of incremental fromsnap (may be zero)
4338 * zc_guid	if set, estimate size of stream only.  zc_cookie is ignored.
4339 *		output size in zc_objset_type.
4340 *
4341 * outputs:
4342 * zc_objset_type	estimated size, if zc_guid is set
4343 */
4344static int
4345zfs_ioc_send(zfs_cmd_t *zc)
4346{
4347	int error;
4348	offset_t off;
4349	boolean_t estimate = (zc->zc_guid != 0);
4350
4351	if (zc->zc_obj != 0) {
4352		dsl_pool_t *dp;
4353		dsl_dataset_t *tosnap;
4354
4355		error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
4356		if (error != 0)
4357			return (error);
4358
4359		error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &tosnap);
4360		if (error != 0) {
4361			dsl_pool_rele(dp, FTAG);
4362			return (error);
4363		}
4364
4365		if (dsl_dir_is_clone(tosnap->ds_dir))
4366			zc->zc_fromobj = tosnap->ds_dir->dd_phys->dd_origin_obj;
4367		dsl_dataset_rele(tosnap, FTAG);
4368		dsl_pool_rele(dp, FTAG);
4369	}
4370
4371	if (estimate) {
4372		dsl_pool_t *dp;
4373		dsl_dataset_t *tosnap;
4374		dsl_dataset_t *fromsnap = NULL;
4375
4376		error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
4377		if (error != 0)
4378			return (error);
4379
4380		error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &tosnap);
4381		if (error != 0) {
4382			dsl_pool_rele(dp, FTAG);
4383			return (error);
4384		}
4385
4386		if (zc->zc_fromobj != 0) {
4387			error = dsl_dataset_hold_obj(dp, zc->zc_fromobj,
4388			    FTAG, &fromsnap);
4389			if (error != 0) {
4390				dsl_dataset_rele(tosnap, FTAG);
4391				dsl_pool_rele(dp, FTAG);
4392				return (error);
4393			}
4394		}
4395
4396		error = dmu_send_estimate(tosnap, fromsnap,
4397		    &zc->zc_objset_type);
4398
4399		if (fromsnap != NULL)
4400			dsl_dataset_rele(fromsnap, FTAG);
4401		dsl_dataset_rele(tosnap, FTAG);
4402		dsl_pool_rele(dp, FTAG);
4403	} else {
4404		file_t *fp;
4405		cap_rights_t rights;
4406
4407		fp = getf(zc->zc_cookie,
4408		    cap_rights_init(&rights, CAP_WRITE));
4409		if (fp == NULL)
4410			return (SET_ERROR(EBADF));
4411
4412		off = fp->f_offset;
4413		error = dmu_send_obj(zc->zc_name, zc->zc_sendobj,
4414#ifdef illumos
4415		    zc->zc_fromobj, zc->zc_cookie, fp->f_vnode, &off);
4416#else
4417		    zc->zc_fromobj, zc->zc_cookie, fp, &off);
4418#endif
4419
4420		if (off >= 0 && off <= MAXOFFSET_T)
4421			fp->f_offset = off;
4422		releasef(zc->zc_cookie);
4423	}
4424	return (error);
4425}
4426
4427/*
4428 * inputs:
4429 * zc_name	name of snapshot on which to report progress
4430 * zc_cookie	file descriptor of send stream
4431 *
4432 * outputs:
4433 * zc_cookie	number of bytes written in send stream thus far
4434 */
4435static int
4436zfs_ioc_send_progress(zfs_cmd_t *zc)
4437{
4438	dsl_pool_t *dp;
4439	dsl_dataset_t *ds;
4440	dmu_sendarg_t *dsp = NULL;
4441	int error;
4442
4443	error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
4444	if (error != 0)
4445		return (error);
4446
4447	error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &ds);
4448	if (error != 0) {
4449		dsl_pool_rele(dp, FTAG);
4450		return (error);
4451	}
4452
4453	mutex_enter(&ds->ds_sendstream_lock);
4454
4455	/*
4456	 * Iterate over all the send streams currently active on this dataset.
4457	 * If there's one which matches the specified file descriptor _and_ the
4458	 * stream was started by the current process, return the progress of
4459	 * that stream.
4460	 */
4461	for (dsp = list_head(&ds->ds_sendstreams); dsp != NULL;
4462	    dsp = list_next(&ds->ds_sendstreams, dsp)) {
4463		if (dsp->dsa_outfd == zc->zc_cookie &&
4464		    dsp->dsa_proc == curproc)
4465			break;
4466	}
4467
4468	if (dsp != NULL)
4469		zc->zc_cookie = *(dsp->dsa_off);
4470	else
4471		error = SET_ERROR(ENOENT);
4472
4473	mutex_exit(&ds->ds_sendstream_lock);
4474	dsl_dataset_rele(ds, FTAG);
4475	dsl_pool_rele(dp, FTAG);
4476	return (error);
4477}
4478
4479static int
4480zfs_ioc_inject_fault(zfs_cmd_t *zc)
4481{
4482	int id, error;
4483
4484	error = zio_inject_fault(zc->zc_name, (int)zc->zc_guid, &id,
4485	    &zc->zc_inject_record);
4486
4487	if (error == 0)
4488		zc->zc_guid = (uint64_t)id;
4489
4490	return (error);
4491}
4492
4493static int
4494zfs_ioc_clear_fault(zfs_cmd_t *zc)
4495{
4496	return (zio_clear_fault((int)zc->zc_guid));
4497}
4498
4499static int
4500zfs_ioc_inject_list_next(zfs_cmd_t *zc)
4501{
4502	int id = (int)zc->zc_guid;
4503	int error;
4504
4505	error = zio_inject_list_next(&id, zc->zc_name, sizeof (zc->zc_name),
4506	    &zc->zc_inject_record);
4507
4508	zc->zc_guid = id;
4509
4510	return (error);
4511}
4512
4513static int
4514zfs_ioc_error_log(zfs_cmd_t *zc)
4515{
4516	spa_t *spa;
4517	int error;
4518	size_t count = (size_t)zc->zc_nvlist_dst_size;
4519
4520	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
4521		return (error);
4522
4523	error = spa_get_errlog(spa, (void *)(uintptr_t)zc->zc_nvlist_dst,
4524	    &count);
4525	if (error == 0)
4526		zc->zc_nvlist_dst_size = count;
4527	else
4528		zc->zc_nvlist_dst_size = spa_get_errlog_size(spa);
4529
4530	spa_close(spa, FTAG);
4531
4532	return (error);
4533}
4534
4535static int
4536zfs_ioc_clear(zfs_cmd_t *zc)
4537{
4538	spa_t *spa;
4539	vdev_t *vd;
4540	int error;
4541
4542	/*
4543	 * On zpool clear we also fix up missing slogs
4544	 */
4545	mutex_enter(&spa_namespace_lock);
4546	spa = spa_lookup(zc->zc_name);
4547	if (spa == NULL) {
4548		mutex_exit(&spa_namespace_lock);
4549		return (SET_ERROR(EIO));
4550	}
4551	if (spa_get_log_state(spa) == SPA_LOG_MISSING) {
4552		/* we need to let spa_open/spa_load clear the chains */
4553		spa_set_log_state(spa, SPA_LOG_CLEAR);
4554	}
4555	spa->spa_last_open_failed = 0;
4556	mutex_exit(&spa_namespace_lock);
4557
4558	if (zc->zc_cookie & ZPOOL_NO_REWIND) {
4559		error = spa_open(zc->zc_name, &spa, FTAG);
4560	} else {
4561		nvlist_t *policy;
4562		nvlist_t *config = NULL;
4563
4564		if (zc->zc_nvlist_src == 0)
4565			return (SET_ERROR(EINVAL));
4566
4567		if ((error = get_nvlist(zc->zc_nvlist_src,
4568		    zc->zc_nvlist_src_size, zc->zc_iflags, &policy)) == 0) {
4569			error = spa_open_rewind(zc->zc_name, &spa, FTAG,
4570			    policy, &config);
4571			if (config != NULL) {
4572				int err;
4573
4574				if ((err = put_nvlist(zc, config)) != 0)
4575					error = err;
4576				nvlist_free(config);
4577			}
4578			nvlist_free(policy);
4579		}
4580	}
4581
4582	if (error != 0)
4583		return (error);
4584
4585	spa_vdev_state_enter(spa, SCL_NONE);
4586
4587	if (zc->zc_guid == 0) {
4588		vd = NULL;
4589	} else {
4590		vd = spa_lookup_by_guid(spa, zc->zc_guid, B_TRUE);
4591		if (vd == NULL) {
4592			(void) spa_vdev_state_exit(spa, NULL, ENODEV);
4593			spa_close(spa, FTAG);
4594			return (SET_ERROR(ENODEV));
4595		}
4596	}
4597
4598	vdev_clear(spa, vd);
4599
4600	(void) spa_vdev_state_exit(spa, NULL, 0);
4601
4602	/*
4603	 * Resume any suspended I/Os.
4604	 */
4605	if (zio_resume(spa) != 0)
4606		error = SET_ERROR(EIO);
4607
4608	spa_close(spa, FTAG);
4609
4610	return (error);
4611}
4612
4613static int
4614zfs_ioc_pool_reopen(zfs_cmd_t *zc)
4615{
4616	spa_t *spa;
4617	int error;
4618
4619	error = spa_open(zc->zc_name, &spa, FTAG);
4620	if (error != 0)
4621		return (error);
4622
4623	spa_vdev_state_enter(spa, SCL_NONE);
4624
4625	/*
4626	 * If a resilver is already in progress then set the
4627	 * spa_scrub_reopen flag to B_TRUE so that we don't restart
4628	 * the scan as a side effect of the reopen. Otherwise, let
4629	 * vdev_open() decided if a resilver is required.
4630	 */
4631	spa->spa_scrub_reopen = dsl_scan_resilvering(spa->spa_dsl_pool);
4632	vdev_reopen(spa->spa_root_vdev);
4633	spa->spa_scrub_reopen = B_FALSE;
4634
4635	(void) spa_vdev_state_exit(spa, NULL, 0);
4636	spa_close(spa, FTAG);
4637	return (0);
4638}
4639/*
4640 * inputs:
4641 * zc_name	name of filesystem
4642 * zc_value	name of origin snapshot
4643 *
4644 * outputs:
4645 * zc_string	name of conflicting snapshot, if there is one
4646 */
4647static int
4648zfs_ioc_promote(zfs_cmd_t *zc)
4649{
4650	char *cp;
4651
4652	/*
4653	 * We don't need to unmount *all* the origin fs's snapshots, but
4654	 * it's easier.
4655	 */
4656	cp = strchr(zc->zc_value, '@');
4657	if (cp)
4658		*cp = '\0';
4659	(void) dmu_objset_find(zc->zc_value,
4660	    zfs_unmount_snap_cb, NULL, DS_FIND_SNAPSHOTS);
4661	return (dsl_dataset_promote(zc->zc_name, zc->zc_string));
4662}
4663
4664/*
4665 * Retrieve a single {user|group}{used|quota}@... property.
4666 *
4667 * inputs:
4668 * zc_name	name of filesystem
4669 * zc_objset_type zfs_userquota_prop_t
4670 * zc_value	domain name (eg. "S-1-234-567-89")
4671 * zc_guid	RID/UID/GID
4672 *
4673 * outputs:
4674 * zc_cookie	property value
4675 */
4676static int
4677zfs_ioc_userspace_one(zfs_cmd_t *zc)
4678{
4679	zfsvfs_t *zfsvfs;
4680	int error;
4681
4682	if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
4683		return (SET_ERROR(EINVAL));
4684
4685	error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE);
4686	if (error != 0)
4687		return (error);
4688
4689	error = zfs_userspace_one(zfsvfs,
4690	    zc->zc_objset_type, zc->zc_value, zc->zc_guid, &zc->zc_cookie);
4691	zfsvfs_rele(zfsvfs, FTAG);
4692
4693	return (error);
4694}
4695
4696/*
4697 * inputs:
4698 * zc_name		name of filesystem
4699 * zc_cookie		zap cursor
4700 * zc_objset_type	zfs_userquota_prop_t
4701 * zc_nvlist_dst[_size] buffer to fill (not really an nvlist)
4702 *
4703 * outputs:
4704 * zc_nvlist_dst[_size]	data buffer (array of zfs_useracct_t)
4705 * zc_cookie	zap cursor
4706 */
4707static int
4708zfs_ioc_userspace_many(zfs_cmd_t *zc)
4709{
4710	zfsvfs_t *zfsvfs;
4711	int bufsize = zc->zc_nvlist_dst_size;
4712
4713	if (bufsize <= 0)
4714		return (SET_ERROR(ENOMEM));
4715
4716	int error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE);
4717	if (error != 0)
4718		return (error);
4719
4720	void *buf = kmem_alloc(bufsize, KM_SLEEP);
4721
4722	error = zfs_userspace_many(zfsvfs, zc->zc_objset_type, &zc->zc_cookie,
4723	    buf, &zc->zc_nvlist_dst_size);
4724
4725	if (error == 0) {
4726		error = ddi_copyout(buf,
4727		    (void *)(uintptr_t)zc->zc_nvlist_dst,
4728		    zc->zc_nvlist_dst_size, zc->zc_iflags);
4729	}
4730	kmem_free(buf, bufsize);
4731	zfsvfs_rele(zfsvfs, FTAG);
4732
4733	return (error);
4734}
4735
4736/*
4737 * inputs:
4738 * zc_name		name of filesystem
4739 *
4740 * outputs:
4741 * none
4742 */
4743static int
4744zfs_ioc_userspace_upgrade(zfs_cmd_t *zc)
4745{
4746	objset_t *os;
4747	int error = 0;
4748	zfsvfs_t *zfsvfs;
4749
4750	if (getzfsvfs(zc->zc_name, &zfsvfs) == 0) {
4751		if (!dmu_objset_userused_enabled(zfsvfs->z_os)) {
4752			/*
4753			 * If userused is not enabled, it may be because the
4754			 * objset needs to be closed & reopened (to grow the
4755			 * objset_phys_t).  Suspend/resume the fs will do that.
4756			 */
4757			error = zfs_suspend_fs(zfsvfs);
4758			if (error == 0) {
4759				dmu_objset_refresh_ownership(zfsvfs->z_os,
4760				    zfsvfs);
4761				error = zfs_resume_fs(zfsvfs, zc->zc_name);
4762			}
4763		}
4764		if (error == 0)
4765			error = dmu_objset_userspace_upgrade(zfsvfs->z_os);
4766		VFS_RELE(zfsvfs->z_vfs);
4767	} else {
4768		/* XXX kind of reading contents without owning */
4769		error = dmu_objset_hold(zc->zc_name, FTAG, &os);
4770		if (error != 0)
4771			return (error);
4772
4773		error = dmu_objset_userspace_upgrade(os);
4774		dmu_objset_rele(os, FTAG);
4775	}
4776
4777	return (error);
4778}
4779
4780#ifdef sun
4781/*
4782 * We don't want to have a hard dependency
4783 * against some special symbols in sharefs
4784 * nfs, and smbsrv.  Determine them if needed when
4785 * the first file system is shared.
4786 * Neither sharefs, nfs or smbsrv are unloadable modules.
4787 */
4788int (*znfsexport_fs)(void *arg);
4789int (*zshare_fs)(enum sharefs_sys_op, share_t *, uint32_t);
4790int (*zsmbexport_fs)(void *arg, boolean_t add_share);
4791
4792int zfs_nfsshare_inited;
4793int zfs_smbshare_inited;
4794
4795ddi_modhandle_t nfs_mod;
4796ddi_modhandle_t sharefs_mod;
4797ddi_modhandle_t smbsrv_mod;
4798#endif	/* sun */
4799kmutex_t zfs_share_lock;
4800
4801#ifdef sun
4802static int
4803zfs_init_sharefs()
4804{
4805	int error;
4806
4807	ASSERT(MUTEX_HELD(&zfs_share_lock));
4808	/* Both NFS and SMB shares also require sharetab support. */
4809	if (sharefs_mod == NULL && ((sharefs_mod =
4810	    ddi_modopen("fs/sharefs",
4811	    KRTLD_MODE_FIRST, &error)) == NULL)) {
4812		return (SET_ERROR(ENOSYS));
4813	}
4814	if (zshare_fs == NULL && ((zshare_fs =
4815	    (int (*)(enum sharefs_sys_op, share_t *, uint32_t))
4816	    ddi_modsym(sharefs_mod, "sharefs_impl", &error)) == NULL)) {
4817		return (SET_ERROR(ENOSYS));
4818	}
4819	return (0);
4820}
4821#endif	/* sun */
4822
4823static int
4824zfs_ioc_share(zfs_cmd_t *zc)
4825{
4826#ifdef sun
4827	int error;
4828	int opcode;
4829
4830	switch (zc->zc_share.z_sharetype) {
4831	case ZFS_SHARE_NFS:
4832	case ZFS_UNSHARE_NFS:
4833		if (zfs_nfsshare_inited == 0) {
4834			mutex_enter(&zfs_share_lock);
4835			if (nfs_mod == NULL && ((nfs_mod = ddi_modopen("fs/nfs",
4836			    KRTLD_MODE_FIRST, &error)) == NULL)) {
4837				mutex_exit(&zfs_share_lock);
4838				return (SET_ERROR(ENOSYS));
4839			}
4840			if (znfsexport_fs == NULL &&
4841			    ((znfsexport_fs = (int (*)(void *))
4842			    ddi_modsym(nfs_mod,
4843			    "nfs_export", &error)) == NULL)) {
4844				mutex_exit(&zfs_share_lock);
4845				return (SET_ERROR(ENOSYS));
4846			}
4847			error = zfs_init_sharefs();
4848			if (error != 0) {
4849				mutex_exit(&zfs_share_lock);
4850				return (SET_ERROR(ENOSYS));
4851			}
4852			zfs_nfsshare_inited = 1;
4853			mutex_exit(&zfs_share_lock);
4854		}
4855		break;
4856	case ZFS_SHARE_SMB:
4857	case ZFS_UNSHARE_SMB:
4858		if (zfs_smbshare_inited == 0) {
4859			mutex_enter(&zfs_share_lock);
4860			if (smbsrv_mod == NULL && ((smbsrv_mod =
4861			    ddi_modopen("drv/smbsrv",
4862			    KRTLD_MODE_FIRST, &error)) == NULL)) {
4863				mutex_exit(&zfs_share_lock);
4864				return (SET_ERROR(ENOSYS));
4865			}
4866			if (zsmbexport_fs == NULL && ((zsmbexport_fs =
4867			    (int (*)(void *, boolean_t))ddi_modsym(smbsrv_mod,
4868			    "smb_server_share", &error)) == NULL)) {
4869				mutex_exit(&zfs_share_lock);
4870				return (SET_ERROR(ENOSYS));
4871			}
4872			error = zfs_init_sharefs();
4873			if (error != 0) {
4874				mutex_exit(&zfs_share_lock);
4875				return (SET_ERROR(ENOSYS));
4876			}
4877			zfs_smbshare_inited = 1;
4878			mutex_exit(&zfs_share_lock);
4879		}
4880		break;
4881	default:
4882		return (SET_ERROR(EINVAL));
4883	}
4884
4885	switch (zc->zc_share.z_sharetype) {
4886	case ZFS_SHARE_NFS:
4887	case ZFS_UNSHARE_NFS:
4888		if (error =
4889		    znfsexport_fs((void *)
4890		    (uintptr_t)zc->zc_share.z_exportdata))
4891			return (error);
4892		break;
4893	case ZFS_SHARE_SMB:
4894	case ZFS_UNSHARE_SMB:
4895		if (error = zsmbexport_fs((void *)
4896		    (uintptr_t)zc->zc_share.z_exportdata,
4897		    zc->zc_share.z_sharetype == ZFS_SHARE_SMB ?
4898		    B_TRUE: B_FALSE)) {
4899			return (error);
4900		}
4901		break;
4902	}
4903
4904	opcode = (zc->zc_share.z_sharetype == ZFS_SHARE_NFS ||
4905	    zc->zc_share.z_sharetype == ZFS_SHARE_SMB) ?
4906	    SHAREFS_ADD : SHAREFS_REMOVE;
4907
4908	/*
4909	 * Add or remove share from sharetab
4910	 */
4911	error = zshare_fs(opcode,
4912	    (void *)(uintptr_t)zc->zc_share.z_sharedata,
4913	    zc->zc_share.z_sharemax);
4914
4915	return (error);
4916
4917#else	/* !sun */
4918	return (ENOSYS);
4919#endif	/* !sun */
4920}
4921
4922ace_t full_access[] = {
4923	{(uid_t)-1, ACE_ALL_PERMS, ACE_EVERYONE, 0}
4924};
4925
4926/*
4927 * inputs:
4928 * zc_name		name of containing filesystem
4929 * zc_obj		object # beyond which we want next in-use object #
4930 *
4931 * outputs:
4932 * zc_obj		next in-use object #
4933 */
4934static int
4935zfs_ioc_next_obj(zfs_cmd_t *zc)
4936{
4937	objset_t *os = NULL;
4938	int error;
4939
4940	error = dmu_objset_hold(zc->zc_name, FTAG, &os);
4941	if (error != 0)
4942		return (error);
4943
4944	error = dmu_object_next(os, &zc->zc_obj, B_FALSE,
4945	    os->os_dsl_dataset->ds_phys->ds_prev_snap_txg);
4946
4947	dmu_objset_rele(os, FTAG);
4948	return (error);
4949}
4950
4951/*
4952 * inputs:
4953 * zc_name		name of filesystem
4954 * zc_value		prefix name for snapshot
4955 * zc_cleanup_fd	cleanup-on-exit file descriptor for calling process
4956 *
4957 * outputs:
4958 * zc_value		short name of new snapshot
4959 */
4960static int
4961zfs_ioc_tmp_snapshot(zfs_cmd_t *zc)
4962{
4963	char *snap_name;
4964	char *hold_name;
4965	int error;
4966	minor_t minor;
4967
4968	error = zfs_onexit_fd_hold(zc->zc_cleanup_fd, &minor);
4969	if (error != 0)
4970		return (error);
4971
4972	snap_name = kmem_asprintf("%s-%016llx", zc->zc_value,
4973	    (u_longlong_t)ddi_get_lbolt64());
4974	hold_name = kmem_asprintf("%%%s", zc->zc_value);
4975
4976	error = dsl_dataset_snapshot_tmp(zc->zc_name, snap_name, minor,
4977	    hold_name);
4978	if (error == 0)
4979		(void) strcpy(zc->zc_value, snap_name);
4980	strfree(snap_name);
4981	strfree(hold_name);
4982	zfs_onexit_fd_rele(zc->zc_cleanup_fd);
4983	return (error);
4984}
4985
4986/*
4987 * inputs:
4988 * zc_name		name of "to" snapshot
4989 * zc_value		name of "from" snapshot
4990 * zc_cookie		file descriptor to write diff data on
4991 *
4992 * outputs:
4993 * dmu_diff_record_t's to the file descriptor
4994 */
4995static int
4996zfs_ioc_diff(zfs_cmd_t *zc)
4997{
4998	file_t *fp;
4999	cap_rights_t rights;
5000	offset_t off;
5001	int error;
5002
5003	fp = getf(zc->zc_cookie, cap_rights_init(&rights, CAP_WRITE));
5004	if (fp == NULL)
5005		return (SET_ERROR(EBADF));
5006
5007	off = fp->f_offset;
5008
5009#ifdef illumos
5010	error = dmu_diff(zc->zc_name, zc->zc_value, fp->f_vnode, &off);
5011#else
5012	error = dmu_diff(zc->zc_name, zc->zc_value, fp, &off);
5013#endif
5014
5015	if (off >= 0 && off <= MAXOFFSET_T)
5016		fp->f_offset = off;
5017	releasef(zc->zc_cookie);
5018
5019	return (error);
5020}
5021
5022#ifdef sun
5023/*
5024 * Remove all ACL files in shares dir
5025 */
5026static int
5027zfs_smb_acl_purge(znode_t *dzp)
5028{
5029	zap_cursor_t	zc;
5030	zap_attribute_t	zap;
5031	zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
5032	int error;
5033
5034	for (zap_cursor_init(&zc, zfsvfs->z_os, dzp->z_id);
5035	    (error = zap_cursor_retrieve(&zc, &zap)) == 0;
5036	    zap_cursor_advance(&zc)) {
5037		if ((error = VOP_REMOVE(ZTOV(dzp), zap.za_name, kcred,
5038		    NULL, 0)) != 0)
5039			break;
5040	}
5041	zap_cursor_fini(&zc);
5042	return (error);
5043}
5044#endif	/* sun */
5045
5046static int
5047zfs_ioc_smb_acl(zfs_cmd_t *zc)
5048{
5049#ifdef sun
5050	vnode_t *vp;
5051	znode_t *dzp;
5052	vnode_t *resourcevp = NULL;
5053	znode_t *sharedir;
5054	zfsvfs_t *zfsvfs;
5055	nvlist_t *nvlist;
5056	char *src, *target;
5057	vattr_t vattr;
5058	vsecattr_t vsec;
5059	int error = 0;
5060
5061	if ((error = lookupname(zc->zc_value, UIO_SYSSPACE,
5062	    NO_FOLLOW, NULL, &vp)) != 0)
5063		return (error);
5064
5065	/* Now make sure mntpnt and dataset are ZFS */
5066
5067	if (strcmp(vp->v_vfsp->mnt_stat.f_fstypename, "zfs") != 0 ||
5068	    (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource),
5069	    zc->zc_name) != 0)) {
5070		VN_RELE(vp);
5071		return (SET_ERROR(EINVAL));
5072	}
5073
5074	dzp = VTOZ(vp);
5075	zfsvfs = dzp->z_zfsvfs;
5076	ZFS_ENTER(zfsvfs);
5077
5078	/*
5079	 * Create share dir if its missing.
5080	 */
5081	mutex_enter(&zfsvfs->z_lock);
5082	if (zfsvfs->z_shares_dir == 0) {
5083		dmu_tx_t *tx;
5084
5085		tx = dmu_tx_create(zfsvfs->z_os);
5086		dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, TRUE,
5087		    ZFS_SHARES_DIR);
5088		dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL);
5089		error = dmu_tx_assign(tx, TXG_WAIT);
5090		if (error != 0) {
5091			dmu_tx_abort(tx);
5092		} else {
5093			error = zfs_create_share_dir(zfsvfs, tx);
5094			dmu_tx_commit(tx);
5095		}
5096		if (error != 0) {
5097			mutex_exit(&zfsvfs->z_lock);
5098			VN_RELE(vp);
5099			ZFS_EXIT(zfsvfs);
5100			return (error);
5101		}
5102	}
5103	mutex_exit(&zfsvfs->z_lock);
5104
5105	ASSERT(zfsvfs->z_shares_dir);
5106	if ((error = zfs_zget(zfsvfs, zfsvfs->z_shares_dir, &sharedir)) != 0) {
5107		VN_RELE(vp);
5108		ZFS_EXIT(zfsvfs);
5109		return (error);
5110	}
5111
5112	switch (zc->zc_cookie) {
5113	case ZFS_SMB_ACL_ADD:
5114		vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE;
5115		vattr.va_type = VREG;
5116		vattr.va_mode = S_IFREG|0777;
5117		vattr.va_uid = 0;
5118		vattr.va_gid = 0;
5119
5120		vsec.vsa_mask = VSA_ACE;
5121		vsec.vsa_aclentp = &full_access;
5122		vsec.vsa_aclentsz = sizeof (full_access);
5123		vsec.vsa_aclcnt = 1;
5124
5125		error = VOP_CREATE(ZTOV(sharedir), zc->zc_string,
5126		    &vattr, EXCL, 0, &resourcevp, kcred, 0, NULL, &vsec);
5127		if (resourcevp)
5128			VN_RELE(resourcevp);
5129		break;
5130
5131	case ZFS_SMB_ACL_REMOVE:
5132		error = VOP_REMOVE(ZTOV(sharedir), zc->zc_string, kcred,
5133		    NULL, 0);
5134		break;
5135
5136	case ZFS_SMB_ACL_RENAME:
5137		if ((error = get_nvlist(zc->zc_nvlist_src,
5138		    zc->zc_nvlist_src_size, zc->zc_iflags, &nvlist)) != 0) {
5139			VN_RELE(vp);
5140			ZFS_EXIT(zfsvfs);
5141			return (error);
5142		}
5143		if (nvlist_lookup_string(nvlist, ZFS_SMB_ACL_SRC, &src) ||
5144		    nvlist_lookup_string(nvlist, ZFS_SMB_ACL_TARGET,
5145		    &target)) {
5146			VN_RELE(vp);
5147			VN_RELE(ZTOV(sharedir));
5148			ZFS_EXIT(zfsvfs);
5149			nvlist_free(nvlist);
5150			return (error);
5151		}
5152		error = VOP_RENAME(ZTOV(sharedir), src, ZTOV(sharedir), target,
5153		    kcred, NULL, 0);
5154		nvlist_free(nvlist);
5155		break;
5156
5157	case ZFS_SMB_ACL_PURGE:
5158		error = zfs_smb_acl_purge(sharedir);
5159		break;
5160
5161	default:
5162		error = SET_ERROR(EINVAL);
5163		break;
5164	}
5165
5166	VN_RELE(vp);
5167	VN_RELE(ZTOV(sharedir));
5168
5169	ZFS_EXIT(zfsvfs);
5170
5171	return (error);
5172#else	/* !sun */
5173	return (EOPNOTSUPP);
5174#endif	/* !sun */
5175}
5176
5177/*
5178 * innvl: {
5179 *     "holds" -> { snapname -> holdname (string), ... }
5180 *     (optional) "cleanup_fd" -> fd (int32)
5181 * }
5182 *
5183 * outnvl: {
5184 *     snapname -> error value (int32)
5185 *     ...
5186 * }
5187 */
5188/* ARGSUSED */
5189static int
5190zfs_ioc_hold(const char *pool, nvlist_t *args, nvlist_t *errlist)
5191{
5192	nvlist_t *holds;
5193	int cleanup_fd = -1;
5194	int error;
5195	minor_t minor = 0;
5196
5197	error = nvlist_lookup_nvlist(args, "holds", &holds);
5198	if (error != 0)
5199		return (SET_ERROR(EINVAL));
5200
5201	if (nvlist_lookup_int32(args, "cleanup_fd", &cleanup_fd) == 0) {
5202		error = zfs_onexit_fd_hold(cleanup_fd, &minor);
5203		if (error != 0)
5204			return (error);
5205	}
5206
5207	error = dsl_dataset_user_hold(holds, minor, errlist);
5208	if (minor != 0)
5209		zfs_onexit_fd_rele(cleanup_fd);
5210	return (error);
5211}
5212
5213/*
5214 * innvl is not used.
5215 *
5216 * outnvl: {
5217 *    holdname -> time added (uint64 seconds since epoch)
5218 *    ...
5219 * }
5220 */
5221/* ARGSUSED */
5222static int
5223zfs_ioc_get_holds(const char *snapname, nvlist_t *args, nvlist_t *outnvl)
5224{
5225	return (dsl_dataset_get_holds(snapname, outnvl));
5226}
5227
5228/*
5229 * innvl: {
5230 *     snapname -> { holdname, ... }
5231 *     ...
5232 * }
5233 *
5234 * outnvl: {
5235 *     snapname -> error value (int32)
5236 *     ...
5237 * }
5238 */
5239/* ARGSUSED */
5240static int
5241zfs_ioc_release(const char *pool, nvlist_t *holds, nvlist_t *errlist)
5242{
5243	return (dsl_dataset_user_release(holds, errlist));
5244}
5245
5246/*
5247 * inputs:
5248 * zc_name		name of new filesystem or snapshot
5249 * zc_value		full name of old snapshot
5250 *
5251 * outputs:
5252 * zc_cookie		space in bytes
5253 * zc_objset_type	compressed space in bytes
5254 * zc_perm_action	uncompressed space in bytes
5255 */
5256static int
5257zfs_ioc_space_written(zfs_cmd_t *zc)
5258{
5259	int error;
5260	dsl_pool_t *dp;
5261	dsl_dataset_t *new, *old;
5262
5263	error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
5264	if (error != 0)
5265		return (error);
5266	error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &new);
5267	if (error != 0) {
5268		dsl_pool_rele(dp, FTAG);
5269		return (error);
5270	}
5271	error = dsl_dataset_hold(dp, zc->zc_value, FTAG, &old);
5272	if (error != 0) {
5273		dsl_dataset_rele(new, FTAG);
5274		dsl_pool_rele(dp, FTAG);
5275		return (error);
5276	}
5277
5278	error = dsl_dataset_space_written(old, new, &zc->zc_cookie,
5279	    &zc->zc_objset_type, &zc->zc_perm_action);
5280	dsl_dataset_rele(old, FTAG);
5281	dsl_dataset_rele(new, FTAG);
5282	dsl_pool_rele(dp, FTAG);
5283	return (error);
5284}
5285
5286/*
5287 * innvl: {
5288 *     "firstsnap" -> snapshot name
5289 * }
5290 *
5291 * outnvl: {
5292 *     "used" -> space in bytes
5293 *     "compressed" -> compressed space in bytes
5294 *     "uncompressed" -> uncompressed space in bytes
5295 * }
5296 */
5297static int
5298zfs_ioc_space_snaps(const char *lastsnap, nvlist_t *innvl, nvlist_t *outnvl)
5299{
5300	int error;
5301	dsl_pool_t *dp;
5302	dsl_dataset_t *new, *old;
5303	char *firstsnap;
5304	uint64_t used, comp, uncomp;
5305
5306	if (nvlist_lookup_string(innvl, "firstsnap", &firstsnap) != 0)
5307		return (SET_ERROR(EINVAL));
5308
5309	error = dsl_pool_hold(lastsnap, FTAG, &dp);
5310	if (error != 0)
5311		return (error);
5312
5313	error = dsl_dataset_hold(dp, lastsnap, FTAG, &new);
5314	if (error != 0) {
5315		dsl_pool_rele(dp, FTAG);
5316		return (error);
5317	}
5318	error = dsl_dataset_hold(dp, firstsnap, FTAG, &old);
5319	if (error != 0) {
5320		dsl_dataset_rele(new, FTAG);
5321		dsl_pool_rele(dp, FTAG);
5322		return (error);
5323	}
5324
5325	error = dsl_dataset_space_wouldfree(old, new, &used, &comp, &uncomp);
5326	dsl_dataset_rele(old, FTAG);
5327	dsl_dataset_rele(new, FTAG);
5328	dsl_pool_rele(dp, FTAG);
5329	fnvlist_add_uint64(outnvl, "used", used);
5330	fnvlist_add_uint64(outnvl, "compressed", comp);
5331	fnvlist_add_uint64(outnvl, "uncompressed", uncomp);
5332	return (error);
5333}
5334
5335static int
5336zfs_ioc_jail(zfs_cmd_t *zc)
5337{
5338
5339	return (zone_dataset_attach(curthread->td_ucred, zc->zc_name,
5340	    (int)zc->zc_jailid));
5341}
5342
5343static int
5344zfs_ioc_unjail(zfs_cmd_t *zc)
5345{
5346
5347	return (zone_dataset_detach(curthread->td_ucred, zc->zc_name,
5348	    (int)zc->zc_jailid));
5349}
5350
5351/*
5352 * innvl: {
5353 *     "fd" -> file descriptor to write stream to (int32)
5354 *     (optional) "fromsnap" -> full snap name to send an incremental from
5355 * }
5356 *
5357 * outnvl is unused
5358 */
5359/* ARGSUSED */
5360static int
5361zfs_ioc_send_new(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl)
5362{
5363	cap_rights_t rights;
5364	int error;
5365	offset_t off;
5366	char *fromname = NULL;
5367	int fd;
5368
5369	error = nvlist_lookup_int32(innvl, "fd", &fd);
5370	if (error != 0)
5371		return (SET_ERROR(EINVAL));
5372
5373	(void) nvlist_lookup_string(innvl, "fromsnap", &fromname);
5374
5375	file_t *fp = getf(fd, cap_rights_init(&rights, CAP_READ));
5376	if (fp == NULL)
5377		return (SET_ERROR(EBADF));
5378
5379	off = fp->f_offset;
5380#ifdef illumos
5381	error = dmu_send(snapname, fromname, fd, fp->f_vnode, &off);
5382#else
5383	error = dmu_send(snapname, fromname, fd, fp, &off);
5384#endif
5385
5386#ifdef illumos
5387	if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
5388		fp->f_offset = off;
5389#else
5390	fp->f_offset = off;
5391#endif
5392
5393	releasef(fd);
5394	return (error);
5395}
5396
5397/*
5398 * Determine approximately how large a zfs send stream will be -- the number
5399 * of bytes that will be written to the fd supplied to zfs_ioc_send_new().
5400 *
5401 * innvl: {
5402 *     (optional) "fromsnap" -> full snap name to send an incremental from
5403 * }
5404 *
5405 * outnvl: {
5406 *     "space" -> bytes of space (uint64)
5407 * }
5408 */
5409static int
5410zfs_ioc_send_space(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl)
5411{
5412	dsl_pool_t *dp;
5413	dsl_dataset_t *fromsnap = NULL;
5414	dsl_dataset_t *tosnap;
5415	int error;
5416	char *fromname;
5417	uint64_t space;
5418
5419	error = dsl_pool_hold(snapname, FTAG, &dp);
5420	if (error != 0)
5421		return (error);
5422
5423	error = dsl_dataset_hold(dp, snapname, FTAG, &tosnap);
5424	if (error != 0) {
5425		dsl_pool_rele(dp, FTAG);
5426		return (error);
5427	}
5428
5429	error = nvlist_lookup_string(innvl, "fromsnap", &fromname);
5430	if (error == 0) {
5431		error = dsl_dataset_hold(dp, fromname, FTAG, &fromsnap);
5432		if (error != 0) {
5433			dsl_dataset_rele(tosnap, FTAG);
5434			dsl_pool_rele(dp, FTAG);
5435			return (error);
5436		}
5437	}
5438
5439	error = dmu_send_estimate(tosnap, fromsnap, &space);
5440	fnvlist_add_uint64(outnvl, "space", space);
5441
5442	if (fromsnap != NULL)
5443		dsl_dataset_rele(fromsnap, FTAG);
5444	dsl_dataset_rele(tosnap, FTAG);
5445	dsl_pool_rele(dp, FTAG);
5446	return (error);
5447}
5448
5449
5450static zfs_ioc_vec_t zfs_ioc_vec[ZFS_IOC_LAST - ZFS_IOC_FIRST];
5451
5452static void
5453zfs_ioctl_register_legacy(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5454    zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck,
5455    boolean_t log_history, zfs_ioc_poolcheck_t pool_check)
5456{
5457	zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST];
5458
5459	ASSERT3U(ioc, >=, ZFS_IOC_FIRST);
5460	ASSERT3U(ioc, <, ZFS_IOC_LAST);
5461	ASSERT3P(vec->zvec_legacy_func, ==, NULL);
5462	ASSERT3P(vec->zvec_func, ==, NULL);
5463
5464	vec->zvec_legacy_func = func;
5465	vec->zvec_secpolicy = secpolicy;
5466	vec->zvec_namecheck = namecheck;
5467	vec->zvec_allow_log = log_history;
5468	vec->zvec_pool_check = pool_check;
5469}
5470
5471/*
5472 * See the block comment at the beginning of this file for details on
5473 * each argument to this function.
5474 */
5475static void
5476zfs_ioctl_register(const char *name, zfs_ioc_t ioc, zfs_ioc_func_t *func,
5477    zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck,
5478    zfs_ioc_poolcheck_t pool_check, boolean_t smush_outnvlist,
5479    boolean_t allow_log)
5480{
5481	zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST];
5482
5483	ASSERT3U(ioc, >=, ZFS_IOC_FIRST);
5484	ASSERT3U(ioc, <, ZFS_IOC_LAST);
5485	ASSERT3P(vec->zvec_legacy_func, ==, NULL);
5486	ASSERT3P(vec->zvec_func, ==, NULL);
5487
5488	/* if we are logging, the name must be valid */
5489	ASSERT(!allow_log || namecheck != NO_NAME);
5490
5491	vec->zvec_name = name;
5492	vec->zvec_func = func;
5493	vec->zvec_secpolicy = secpolicy;
5494	vec->zvec_namecheck = namecheck;
5495	vec->zvec_pool_check = pool_check;
5496	vec->zvec_smush_outnvlist = smush_outnvlist;
5497	vec->zvec_allow_log = allow_log;
5498}
5499
5500static void
5501zfs_ioctl_register_pool(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5502    zfs_secpolicy_func_t *secpolicy, boolean_t log_history,
5503    zfs_ioc_poolcheck_t pool_check)
5504{
5505	zfs_ioctl_register_legacy(ioc, func, secpolicy,
5506	    POOL_NAME, log_history, pool_check);
5507}
5508
5509static void
5510zfs_ioctl_register_dataset_nolog(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5511    zfs_secpolicy_func_t *secpolicy, zfs_ioc_poolcheck_t pool_check)
5512{
5513	zfs_ioctl_register_legacy(ioc, func, secpolicy,
5514	    DATASET_NAME, B_FALSE, pool_check);
5515}
5516
5517static void
5518zfs_ioctl_register_pool_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func)
5519{
5520	zfs_ioctl_register_legacy(ioc, func, zfs_secpolicy_config,
5521	    POOL_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5522}
5523
5524static void
5525zfs_ioctl_register_pool_meta(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5526    zfs_secpolicy_func_t *secpolicy)
5527{
5528	zfs_ioctl_register_legacy(ioc, func, secpolicy,
5529	    NO_NAME, B_FALSE, POOL_CHECK_NONE);
5530}
5531
5532static void
5533zfs_ioctl_register_dataset_read_secpolicy(zfs_ioc_t ioc,
5534    zfs_ioc_legacy_func_t *func, zfs_secpolicy_func_t *secpolicy)
5535{
5536	zfs_ioctl_register_legacy(ioc, func, secpolicy,
5537	    DATASET_NAME, B_FALSE, POOL_CHECK_SUSPENDED);
5538}
5539
5540static void
5541zfs_ioctl_register_dataset_read(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func)
5542{
5543	zfs_ioctl_register_dataset_read_secpolicy(ioc, func,
5544	    zfs_secpolicy_read);
5545}
5546
5547static void
5548zfs_ioctl_register_dataset_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5549	zfs_secpolicy_func_t *secpolicy)
5550{
5551	zfs_ioctl_register_legacy(ioc, func, secpolicy,
5552	    DATASET_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5553}
5554
5555static void
5556zfs_ioctl_init(void)
5557{
5558	zfs_ioctl_register("snapshot", ZFS_IOC_SNAPSHOT,
5559	    zfs_ioc_snapshot, zfs_secpolicy_snapshot, POOL_NAME,
5560	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5561
5562	zfs_ioctl_register("log_history", ZFS_IOC_LOG_HISTORY,
5563	    zfs_ioc_log_history, zfs_secpolicy_log_history, NO_NAME,
5564	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_FALSE);
5565
5566	zfs_ioctl_register("space_snaps", ZFS_IOC_SPACE_SNAPS,
5567	    zfs_ioc_space_snaps, zfs_secpolicy_read, DATASET_NAME,
5568	    POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5569
5570	zfs_ioctl_register("send", ZFS_IOC_SEND_NEW,
5571	    zfs_ioc_send_new, zfs_secpolicy_send_new, DATASET_NAME,
5572	    POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5573
5574	zfs_ioctl_register("send_space", ZFS_IOC_SEND_SPACE,
5575	    zfs_ioc_send_space, zfs_secpolicy_read, DATASET_NAME,
5576	    POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5577
5578	zfs_ioctl_register("create", ZFS_IOC_CREATE,
5579	    zfs_ioc_create, zfs_secpolicy_create_clone, DATASET_NAME,
5580	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5581
5582	zfs_ioctl_register("clone", ZFS_IOC_CLONE,
5583	    zfs_ioc_clone, zfs_secpolicy_create_clone, DATASET_NAME,
5584	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5585
5586	zfs_ioctl_register("destroy_snaps", ZFS_IOC_DESTROY_SNAPS,
5587	    zfs_ioc_destroy_snaps, zfs_secpolicy_destroy_snaps, POOL_NAME,
5588	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5589
5590	zfs_ioctl_register("hold", ZFS_IOC_HOLD,
5591	    zfs_ioc_hold, zfs_secpolicy_hold, POOL_NAME,
5592	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5593	zfs_ioctl_register("release", ZFS_IOC_RELEASE,
5594	    zfs_ioc_release, zfs_secpolicy_release, POOL_NAME,
5595	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5596
5597	zfs_ioctl_register("get_holds", ZFS_IOC_GET_HOLDS,
5598	    zfs_ioc_get_holds, zfs_secpolicy_read, DATASET_NAME,
5599	    POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5600
5601	zfs_ioctl_register("rollback", ZFS_IOC_ROLLBACK,
5602	    zfs_ioc_rollback, zfs_secpolicy_rollback, DATASET_NAME,
5603	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_TRUE);
5604
5605	zfs_ioctl_register("bookmark", ZFS_IOC_BOOKMARK,
5606	    zfs_ioc_bookmark, zfs_secpolicy_bookmark, POOL_NAME,
5607	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5608
5609	zfs_ioctl_register("get_bookmarks", ZFS_IOC_GET_BOOKMARKS,
5610	    zfs_ioc_get_bookmarks, zfs_secpolicy_read, DATASET_NAME,
5611	    POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5612
5613	zfs_ioctl_register("destroy_bookmarks", ZFS_IOC_DESTROY_BOOKMARKS,
5614	    zfs_ioc_destroy_bookmarks, zfs_secpolicy_destroy_bookmarks,
5615	    POOL_NAME,
5616	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5617
5618	/* IOCTLS that use the legacy function signature */
5619
5620	zfs_ioctl_register_legacy(ZFS_IOC_POOL_FREEZE, zfs_ioc_pool_freeze,
5621	    zfs_secpolicy_config, NO_NAME, B_FALSE, POOL_CHECK_READONLY);
5622
5623	zfs_ioctl_register_pool(ZFS_IOC_POOL_CREATE, zfs_ioc_pool_create,
5624	    zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE);
5625	zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SCAN,
5626	    zfs_ioc_pool_scan);
5627	zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_UPGRADE,
5628	    zfs_ioc_pool_upgrade);
5629	zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ADD,
5630	    zfs_ioc_vdev_add);
5631	zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_REMOVE,
5632	    zfs_ioc_vdev_remove);
5633	zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SET_STATE,
5634	    zfs_ioc_vdev_set_state);
5635	zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ATTACH,
5636	    zfs_ioc_vdev_attach);
5637	zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_DETACH,
5638	    zfs_ioc_vdev_detach);
5639	zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETPATH,
5640	    zfs_ioc_vdev_setpath);
5641	zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETFRU,
5642	    zfs_ioc_vdev_setfru);
5643	zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SET_PROPS,
5644	    zfs_ioc_pool_set_props);
5645	zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SPLIT,
5646	    zfs_ioc_vdev_split);
5647	zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_REGUID,
5648	    zfs_ioc_pool_reguid);
5649
5650	zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_CONFIGS,
5651	    zfs_ioc_pool_configs, zfs_secpolicy_none);
5652	zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_TRYIMPORT,
5653	    zfs_ioc_pool_tryimport, zfs_secpolicy_config);
5654	zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_FAULT,
5655	    zfs_ioc_inject_fault, zfs_secpolicy_inject);
5656	zfs_ioctl_register_pool_meta(ZFS_IOC_CLEAR_FAULT,
5657	    zfs_ioc_clear_fault, zfs_secpolicy_inject);
5658	zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_LIST_NEXT,
5659	    zfs_ioc_inject_list_next, zfs_secpolicy_inject);
5660
5661	/*
5662	 * pool destroy, and export don't log the history as part of
5663	 * zfsdev_ioctl, but rather zfs_ioc_pool_export
5664	 * does the logging of those commands.
5665	 */
5666	zfs_ioctl_register_pool(ZFS_IOC_POOL_DESTROY, zfs_ioc_pool_destroy,
5667	    zfs_secpolicy_config, B_FALSE, POOL_CHECK_NONE);
5668	zfs_ioctl_register_pool(ZFS_IOC_POOL_EXPORT, zfs_ioc_pool_export,
5669	    zfs_secpolicy_config, B_FALSE, POOL_CHECK_NONE);
5670
5671	zfs_ioctl_register_pool(ZFS_IOC_POOL_STATS, zfs_ioc_pool_stats,
5672	    zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE);
5673	zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_PROPS, zfs_ioc_pool_get_props,
5674	    zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE);
5675
5676	zfs_ioctl_register_pool(ZFS_IOC_ERROR_LOG, zfs_ioc_error_log,
5677	    zfs_secpolicy_inject, B_FALSE, POOL_CHECK_NONE);
5678	zfs_ioctl_register_pool(ZFS_IOC_DSOBJ_TO_DSNAME,
5679	    zfs_ioc_dsobj_to_dsname,
5680	    zfs_secpolicy_diff, B_FALSE, POOL_CHECK_NONE);
5681	zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_HISTORY,
5682	    zfs_ioc_pool_get_history,
5683	    zfs_secpolicy_config, B_FALSE, POOL_CHECK_SUSPENDED);
5684
5685	zfs_ioctl_register_pool(ZFS_IOC_POOL_IMPORT, zfs_ioc_pool_import,
5686	    zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE);
5687
5688	zfs_ioctl_register_pool(ZFS_IOC_CLEAR, zfs_ioc_clear,
5689	    zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE);
5690	zfs_ioctl_register_pool(ZFS_IOC_POOL_REOPEN, zfs_ioc_pool_reopen,
5691	    zfs_secpolicy_config, B_TRUE, POOL_CHECK_SUSPENDED);
5692
5693	zfs_ioctl_register_dataset_read(ZFS_IOC_SPACE_WRITTEN,
5694	    zfs_ioc_space_written);
5695	zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_RECVD_PROPS,
5696	    zfs_ioc_objset_recvd_props);
5697	zfs_ioctl_register_dataset_read(ZFS_IOC_NEXT_OBJ,
5698	    zfs_ioc_next_obj);
5699	zfs_ioctl_register_dataset_read(ZFS_IOC_GET_FSACL,
5700	    zfs_ioc_get_fsacl);
5701	zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_STATS,
5702	    zfs_ioc_objset_stats);
5703	zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_ZPLPROPS,
5704	    zfs_ioc_objset_zplprops);
5705	zfs_ioctl_register_dataset_read(ZFS_IOC_DATASET_LIST_NEXT,
5706	    zfs_ioc_dataset_list_next);
5707	zfs_ioctl_register_dataset_read(ZFS_IOC_SNAPSHOT_LIST_NEXT,
5708	    zfs_ioc_snapshot_list_next);
5709	zfs_ioctl_register_dataset_read(ZFS_IOC_SEND_PROGRESS,
5710	    zfs_ioc_send_progress);
5711
5712	zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_DIFF,
5713	    zfs_ioc_diff, zfs_secpolicy_diff);
5714	zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_STATS,
5715	    zfs_ioc_obj_to_stats, zfs_secpolicy_diff);
5716	zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_PATH,
5717	    zfs_ioc_obj_to_path, zfs_secpolicy_diff);
5718	zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_ONE,
5719	    zfs_ioc_userspace_one, zfs_secpolicy_userspace_one);
5720	zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_MANY,
5721	    zfs_ioc_userspace_many, zfs_secpolicy_userspace_many);
5722	zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_SEND,
5723	    zfs_ioc_send, zfs_secpolicy_send);
5724
5725	zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_PROP, zfs_ioc_set_prop,
5726	    zfs_secpolicy_none);
5727	zfs_ioctl_register_dataset_modify(ZFS_IOC_DESTROY, zfs_ioc_destroy,
5728	    zfs_secpolicy_destroy);
5729	zfs_ioctl_register_dataset_modify(ZFS_IOC_RENAME, zfs_ioc_rename,
5730	    zfs_secpolicy_rename);
5731	zfs_ioctl_register_dataset_modify(ZFS_IOC_RECV, zfs_ioc_recv,
5732	    zfs_secpolicy_recv);
5733	zfs_ioctl_register_dataset_modify(ZFS_IOC_PROMOTE, zfs_ioc_promote,
5734	    zfs_secpolicy_promote);
5735	zfs_ioctl_register_dataset_modify(ZFS_IOC_INHERIT_PROP,
5736	    zfs_ioc_inherit_prop, zfs_secpolicy_inherit_prop);
5737	zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_FSACL, zfs_ioc_set_fsacl,
5738	    zfs_secpolicy_set_fsacl);
5739
5740	zfs_ioctl_register_dataset_nolog(ZFS_IOC_SHARE, zfs_ioc_share,
5741	    zfs_secpolicy_share, POOL_CHECK_NONE);
5742	zfs_ioctl_register_dataset_nolog(ZFS_IOC_SMB_ACL, zfs_ioc_smb_acl,
5743	    zfs_secpolicy_smb_acl, POOL_CHECK_NONE);
5744	zfs_ioctl_register_dataset_nolog(ZFS_IOC_USERSPACE_UPGRADE,
5745	    zfs_ioc_userspace_upgrade, zfs_secpolicy_userspace_upgrade,
5746	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5747	zfs_ioctl_register_dataset_nolog(ZFS_IOC_TMP_SNAPSHOT,
5748	    zfs_ioc_tmp_snapshot, zfs_secpolicy_tmp_snapshot,
5749	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5750
5751#ifdef __FreeBSD__
5752	zfs_ioctl_register_dataset_nolog(ZFS_IOC_JAIL, zfs_ioc_jail,
5753	    zfs_secpolicy_config, POOL_CHECK_NONE);
5754	zfs_ioctl_register_dataset_nolog(ZFS_IOC_UNJAIL, zfs_ioc_unjail,
5755	    zfs_secpolicy_config, POOL_CHECK_NONE);
5756#endif
5757}
5758
5759int
5760pool_status_check(const char *name, zfs_ioc_namecheck_t type,
5761    zfs_ioc_poolcheck_t check)
5762{
5763	spa_t *spa;
5764	int error;
5765
5766	ASSERT(type == POOL_NAME || type == DATASET_NAME);
5767
5768	if (check & POOL_CHECK_NONE)
5769		return (0);
5770
5771	error = spa_open(name, &spa, FTAG);
5772	if (error == 0) {
5773		if ((check & POOL_CHECK_SUSPENDED) && spa_suspended(spa))
5774			error = SET_ERROR(EAGAIN);
5775		else if ((check & POOL_CHECK_READONLY) && !spa_writeable(spa))
5776			error = SET_ERROR(EROFS);
5777		spa_close(spa, FTAG);
5778	}
5779	return (error);
5780}
5781
5782/*
5783 * Find a free minor number.
5784 */
5785minor_t
5786zfsdev_minor_alloc(void)
5787{
5788	static minor_t last_minor;
5789	minor_t m;
5790
5791	ASSERT(MUTEX_HELD(&spa_namespace_lock));
5792
5793	for (m = last_minor + 1; m != last_minor; m++) {
5794		if (m > ZFSDEV_MAX_MINOR)
5795			m = 1;
5796		if (ddi_get_soft_state(zfsdev_state, m) == NULL) {
5797			last_minor = m;
5798			return (m);
5799		}
5800	}
5801
5802	return (0);
5803}
5804
5805static int
5806zfs_ctldev_init(struct cdev *devp)
5807{
5808	minor_t minor;
5809	zfs_soft_state_t *zs;
5810
5811	ASSERT(MUTEX_HELD(&spa_namespace_lock));
5812
5813	minor = zfsdev_minor_alloc();
5814	if (minor == 0)
5815		return (SET_ERROR(ENXIO));
5816
5817	if (ddi_soft_state_zalloc(zfsdev_state, minor) != DDI_SUCCESS)
5818		return (SET_ERROR(EAGAIN));
5819
5820	devfs_set_cdevpriv((void *)(uintptr_t)minor, zfsdev_close);
5821
5822	zs = ddi_get_soft_state(zfsdev_state, minor);
5823	zs->zss_type = ZSST_CTLDEV;
5824	zfs_onexit_init((zfs_onexit_t **)&zs->zss_data);
5825
5826	return (0);
5827}
5828
5829static void
5830zfs_ctldev_destroy(zfs_onexit_t *zo, minor_t minor)
5831{
5832	ASSERT(MUTEX_HELD(&spa_namespace_lock));
5833
5834	zfs_onexit_destroy(zo);
5835	ddi_soft_state_free(zfsdev_state, minor);
5836}
5837
5838void *
5839zfsdev_get_soft_state(minor_t minor, enum zfs_soft_state_type which)
5840{
5841	zfs_soft_state_t *zp;
5842
5843	zp = ddi_get_soft_state(zfsdev_state, minor);
5844	if (zp == NULL || zp->zss_type != which)
5845		return (NULL);
5846
5847	return (zp->zss_data);
5848}
5849
5850static int
5851zfsdev_open(struct cdev *devp, int flag, int mode, struct thread *td)
5852{
5853	int error = 0;
5854
5855#ifdef sun
5856	if (getminor(*devp) != 0)
5857		return (zvol_open(devp, flag, otyp, cr));
5858#endif
5859
5860	/* This is the control device. Allocate a new minor if requested. */
5861	if (flag & FEXCL) {
5862		mutex_enter(&spa_namespace_lock);
5863		error = zfs_ctldev_init(devp);
5864		mutex_exit(&spa_namespace_lock);
5865	}
5866
5867	return (error);
5868}
5869
5870static void
5871zfsdev_close(void *data)
5872{
5873	zfs_onexit_t *zo;
5874	minor_t minor = (minor_t)(uintptr_t)data;
5875
5876	if (minor == 0)
5877		return;
5878
5879	mutex_enter(&spa_namespace_lock);
5880	zo = zfsdev_get_soft_state(minor, ZSST_CTLDEV);
5881	if (zo == NULL) {
5882		mutex_exit(&spa_namespace_lock);
5883		return;
5884	}
5885	zfs_ctldev_destroy(zo, minor);
5886	mutex_exit(&spa_namespace_lock);
5887}
5888
5889static int
5890zfsdev_ioctl(struct cdev *dev, u_long zcmd, caddr_t arg, int flag,
5891    struct thread *td)
5892{
5893	zfs_cmd_t *zc;
5894	uint_t vecnum;
5895	int error, rc, len;
5896#ifdef illumos
5897	minor_t minor = getminor(dev);
5898#else
5899	zfs_iocparm_t *zc_iocparm;
5900	int cflag, cmd, oldvecnum;
5901	boolean_t newioc, compat;
5902	cred_t *cr = td->td_ucred;
5903#endif
5904	const zfs_ioc_vec_t *vec;
5905	char *saved_poolname = NULL;
5906	nvlist_t *innvl = NULL;
5907
5908	cflag = ZFS_CMD_COMPAT_NONE;
5909	compat = B_FALSE;
5910	newioc = B_TRUE;
5911
5912	len = IOCPARM_LEN(zcmd);
5913	cmd = zcmd & 0xff;
5914
5915	/*
5916	 * Check if we are talking to supported older binaries
5917	 * and translate zfs_cmd if necessary
5918	 */
5919	if (len != sizeof(zfs_iocparm_t)) {
5920		newioc = B_FALSE;
5921		if (len == sizeof(zfs_cmd_t)) {
5922			cflag = ZFS_CMD_COMPAT_LZC;
5923			vecnum = cmd;
5924		} else if (len == sizeof(zfs_cmd_deadman_t)) {
5925			cflag = ZFS_CMD_COMPAT_DEADMAN;
5926			compat = B_TRUE;
5927			vecnum = cmd;
5928		} else if (len == sizeof(zfs_cmd_v28_t)) {
5929			cflag = ZFS_CMD_COMPAT_V28;
5930			compat = B_TRUE;
5931			vecnum = cmd;
5932		} else if (len == sizeof(zfs_cmd_v15_t)) {
5933			cflag = ZFS_CMD_COMPAT_V15;
5934			compat = B_TRUE;
5935			vecnum = zfs_ioctl_v15_to_v28[cmd];
5936		} else
5937			return (EINVAL);
5938	} else
5939		vecnum = cmd;
5940
5941#ifdef illumos
5942	vecnum = cmd - ZFS_IOC_FIRST;
5943	ASSERT3U(getmajor(dev), ==, ddi_driver_major(zfs_dip));
5944#endif
5945
5946	if (compat) {
5947		if (vecnum == ZFS_IOC_COMPAT_PASS)
5948			return (0);
5949		else if (vecnum == ZFS_IOC_COMPAT_FAIL)
5950			return (ENOTSUP);
5951	}
5952
5953	/*
5954	 * Check if we have sufficient kernel memory allocated
5955	 * for the zfs_cmd_t request.  Bail out if not so we
5956	 * will not access undefined memory region.
5957	 */
5958	if (vecnum >= sizeof (zfs_ioc_vec) / sizeof (zfs_ioc_vec[0]))
5959		return (SET_ERROR(EINVAL));
5960	vec = &zfs_ioc_vec[vecnum];
5961
5962#ifdef illumos
5963	zc = kmem_zalloc(sizeof(zfs_cmd_t), KM_SLEEP);
5964	bzero(zc, sizeof(zfs_cmd_t));
5965
5966	error = ddi_copyin((void *)arg, zc, sizeof (zfs_cmd_t), flag);
5967	if (error != 0) {
5968		error = SET_ERROR(EFAULT);
5969		goto out;
5970	}
5971#else	/* !illumos */
5972	/*
5973	 * We don't alloc/free zc only if talking to library ioctl version 2
5974	 */
5975	if (cflag != ZFS_CMD_COMPAT_LZC) {
5976		zc = kmem_zalloc(sizeof(zfs_cmd_t), KM_SLEEP);
5977		bzero(zc, sizeof(zfs_cmd_t));
5978	} else {
5979		zc = (void *)arg;
5980		error = 0;
5981	}
5982
5983	if (newioc) {
5984		zc_iocparm = (void *)arg;
5985		if (zc_iocparm->zfs_cmd_size != sizeof(zfs_cmd_t)) {
5986			error = SET_ERROR(EFAULT);
5987			goto out;
5988		}
5989		error = ddi_copyin((void *)(uintptr_t)zc_iocparm->zfs_cmd, zc,
5990		    sizeof(zfs_cmd_t), flag);
5991		if (error != 0) {
5992			error = SET_ERROR(EFAULT);
5993			goto out;
5994		}
5995	}
5996
5997	if (compat) {
5998		zfs_cmd_compat_get(zc, arg, cflag);
5999		oldvecnum = vecnum;
6000		error = zfs_ioctl_compat_pre(zc, &vecnum, cflag);
6001		if (error != 0)
6002			goto out;
6003		if (oldvecnum != vecnum)
6004			vec = &zfs_ioc_vec[vecnum];
6005	}
6006#endif	/* !illumos */
6007
6008	zc->zc_iflags = flag & FKIOCTL;
6009	if (zc->zc_nvlist_src_size != 0) {
6010		error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
6011		    zc->zc_iflags, &innvl);
6012		if (error != 0)
6013			goto out;
6014	}
6015
6016	/* rewrite innvl for backwards compatibility */
6017	if (compat)
6018		innvl = zfs_ioctl_compat_innvl(zc, innvl, vecnum, cflag);
6019
6020	/*
6021	 * Ensure that all pool/dataset names are valid before we pass down to
6022	 * the lower layers.
6023	 */
6024	zc->zc_name[sizeof (zc->zc_name) - 1] = '\0';
6025	switch (vec->zvec_namecheck) {
6026	case POOL_NAME:
6027		if (pool_namecheck(zc->zc_name, NULL, NULL) != 0)
6028			error = SET_ERROR(EINVAL);
6029		else
6030			error = pool_status_check(zc->zc_name,
6031			    vec->zvec_namecheck, vec->zvec_pool_check);
6032		break;
6033
6034	case DATASET_NAME:
6035		if (dataset_namecheck(zc->zc_name, NULL, NULL) != 0)
6036			error = SET_ERROR(EINVAL);
6037		else
6038			error = pool_status_check(zc->zc_name,
6039			    vec->zvec_namecheck, vec->zvec_pool_check);
6040		break;
6041
6042	case NO_NAME:
6043		break;
6044	}
6045
6046	if (error == 0 && !(flag & FKIOCTL))
6047		error = vec->zvec_secpolicy(zc, innvl, cr);
6048
6049	if (error != 0)
6050		goto out;
6051
6052	/* legacy ioctls can modify zc_name */
6053	len = strcspn(zc->zc_name, "/@#") + 1;
6054	saved_poolname = kmem_alloc(len, KM_SLEEP);
6055	(void) strlcpy(saved_poolname, zc->zc_name, len);
6056
6057	if (vec->zvec_func != NULL) {
6058		nvlist_t *outnvl;
6059		int puterror = 0;
6060		spa_t *spa;
6061		nvlist_t *lognv = NULL;
6062
6063		ASSERT(vec->zvec_legacy_func == NULL);
6064
6065		/*
6066		 * Add the innvl to the lognv before calling the func,
6067		 * in case the func changes the innvl.
6068		 */
6069		if (vec->zvec_allow_log) {
6070			lognv = fnvlist_alloc();
6071			fnvlist_add_string(lognv, ZPOOL_HIST_IOCTL,
6072			    vec->zvec_name);
6073			if (!nvlist_empty(innvl)) {
6074				fnvlist_add_nvlist(lognv, ZPOOL_HIST_INPUT_NVL,
6075				    innvl);
6076			}
6077		}
6078
6079		outnvl = fnvlist_alloc();
6080		error = vec->zvec_func(zc->zc_name, innvl, outnvl);
6081
6082		if (error == 0 && vec->zvec_allow_log &&
6083		    spa_open(zc->zc_name, &spa, FTAG) == 0) {
6084			if (!nvlist_empty(outnvl)) {
6085				fnvlist_add_nvlist(lognv, ZPOOL_HIST_OUTPUT_NVL,
6086				    outnvl);
6087			}
6088			(void) spa_history_log_nvl(spa, lognv);
6089			spa_close(spa, FTAG);
6090		}
6091		fnvlist_free(lognv);
6092
6093		/* rewrite outnvl for backwards compatibility */
6094		if (cflag != ZFS_CMD_COMPAT_NONE && cflag != ZFS_CMD_COMPAT_LZC)
6095			outnvl = zfs_ioctl_compat_outnvl(zc, outnvl, vecnum,
6096			    cflag);
6097
6098		if (!nvlist_empty(outnvl) || zc->zc_nvlist_dst_size != 0) {
6099			int smusherror = 0;
6100			if (vec->zvec_smush_outnvlist) {
6101				smusherror = nvlist_smush(outnvl,
6102				    zc->zc_nvlist_dst_size);
6103			}
6104			if (smusherror == 0)
6105				puterror = put_nvlist(zc, outnvl);
6106		}
6107
6108		if (puterror != 0)
6109			error = puterror;
6110
6111		nvlist_free(outnvl);
6112	} else {
6113		error = vec->zvec_legacy_func(zc);
6114	}
6115
6116out:
6117	nvlist_free(innvl);
6118
6119	if (compat) {
6120		zfs_ioctl_compat_post(zc, cmd, cflag);
6121		zfs_cmd_compat_put(zc, arg, vecnum, cflag);
6122	}
6123
6124#ifdef illumos
6125	rc = ddi_copyout(zc, (void *)arg, sizeof (zfs_cmd_t), flag);
6126	if (error == 0 && rc != 0)
6127		error = SET_ERROR(EFAULT);
6128#else
6129	if (newioc) {
6130		rc = ddi_copyout(zc, (void *)(uintptr_t)zc_iocparm->zfs_cmd,
6131		    sizeof (zfs_cmd_t), flag);
6132		if (error == 0 && rc != 0)
6133			error = SET_ERROR(EFAULT);
6134	}
6135#endif
6136	if (error == 0 && vec->zvec_allow_log) {
6137		char *s = tsd_get(zfs_allow_log_key);
6138		if (s != NULL)
6139			strfree(s);
6140		(void) tsd_set(zfs_allow_log_key, saved_poolname);
6141	} else {
6142		if (saved_poolname != NULL)
6143			strfree(saved_poolname);
6144	}
6145
6146#ifdef illumos
6147	kmem_free(zc, sizeof (zfs_cmd_t));
6148#else
6149	/*
6150	 * We don't alloc/free zc only if talking to library ioctl version 2
6151	 */
6152	if (cflag != ZFS_CMD_COMPAT_LZC)
6153		kmem_free(zc, sizeof (zfs_cmd_t));
6154#endif
6155	return (error);
6156}
6157
6158#ifdef sun
6159static int
6160zfs_attach(dev_info_t *dip, ddi_attach_cmd_t cmd)
6161{
6162	if (cmd != DDI_ATTACH)
6163		return (DDI_FAILURE);
6164
6165	if (ddi_create_minor_node(dip, "zfs", S_IFCHR, 0,
6166	    DDI_PSEUDO, 0) == DDI_FAILURE)
6167		return (DDI_FAILURE);
6168
6169	zfs_dip = dip;
6170
6171	ddi_report_dev(dip);
6172
6173	return (DDI_SUCCESS);
6174}
6175
6176static int
6177zfs_detach(dev_info_t *dip, ddi_detach_cmd_t cmd)
6178{
6179	if (spa_busy() || zfs_busy() || zvol_busy())
6180		return (DDI_FAILURE);
6181
6182	if (cmd != DDI_DETACH)
6183		return (DDI_FAILURE);
6184
6185	zfs_dip = NULL;
6186
6187	ddi_prop_remove_all(dip);
6188	ddi_remove_minor_node(dip, NULL);
6189
6190	return (DDI_SUCCESS);
6191}
6192
6193/*ARGSUSED*/
6194static int
6195zfs_info(dev_info_t *dip, ddi_info_cmd_t infocmd, void *arg, void **result)
6196{
6197	switch (infocmd) {
6198	case DDI_INFO_DEVT2DEVINFO:
6199		*result = zfs_dip;
6200		return (DDI_SUCCESS);
6201
6202	case DDI_INFO_DEVT2INSTANCE:
6203		*result = (void *)0;
6204		return (DDI_SUCCESS);
6205	}
6206
6207	return (DDI_FAILURE);
6208}
6209#endif	/* sun */
6210
6211/*
6212 * OK, so this is a little weird.
6213 *
6214 * /dev/zfs is the control node, i.e. minor 0.
6215 * /dev/zvol/[r]dsk/pool/dataset are the zvols, minor > 0.
6216 *
6217 * /dev/zfs has basically nothing to do except serve up ioctls,
6218 * so most of the standard driver entry points are in zvol.c.
6219 */
6220#ifdef sun
6221static struct cb_ops zfs_cb_ops = {
6222	zfsdev_open,	/* open */
6223	zfsdev_close,	/* close */
6224	zvol_strategy,	/* strategy */
6225	nodev,		/* print */
6226	zvol_dump,	/* dump */
6227	zvol_read,	/* read */
6228	zvol_write,	/* write */
6229	zfsdev_ioctl,	/* ioctl */
6230	nodev,		/* devmap */
6231	nodev,		/* mmap */
6232	nodev,		/* segmap */
6233	nochpoll,	/* poll */
6234	ddi_prop_op,	/* prop_op */
6235	NULL,		/* streamtab */
6236	D_NEW | D_MP | D_64BIT,		/* Driver compatibility flag */
6237	CB_REV,		/* version */
6238	nodev,		/* async read */
6239	nodev,		/* async write */
6240};
6241
6242static struct dev_ops zfs_dev_ops = {
6243	DEVO_REV,	/* version */
6244	0,		/* refcnt */
6245	zfs_info,	/* info */
6246	nulldev,	/* identify */
6247	nulldev,	/* probe */
6248	zfs_attach,	/* attach */
6249	zfs_detach,	/* detach */
6250	nodev,		/* reset */
6251	&zfs_cb_ops,	/* driver operations */
6252	NULL,		/* no bus operations */
6253	NULL,		/* power */
6254	ddi_quiesce_not_needed,	/* quiesce */
6255};
6256
6257static struct modldrv zfs_modldrv = {
6258	&mod_driverops,
6259	"ZFS storage pool",
6260	&zfs_dev_ops
6261};
6262
6263static struct modlinkage modlinkage = {
6264	MODREV_1,
6265	(void *)&zfs_modlfs,
6266	(void *)&zfs_modldrv,
6267	NULL
6268};
6269#endif	/* sun */
6270
6271static struct cdevsw zfs_cdevsw = {
6272	.d_version =	D_VERSION,
6273	.d_open =	zfsdev_open,
6274	.d_ioctl =	zfsdev_ioctl,
6275	.d_name =	ZFS_DEV_NAME
6276};
6277
6278static void
6279zfs_allow_log_destroy(void *arg)
6280{
6281	char *poolname = arg;
6282	strfree(poolname);
6283}
6284
6285static void
6286zfsdev_init(void)
6287{
6288	zfsdev = make_dev(&zfs_cdevsw, 0x0, UID_ROOT, GID_OPERATOR, 0666,
6289	    ZFS_DEV_NAME);
6290}
6291
6292static void
6293zfsdev_fini(void)
6294{
6295	if (zfsdev != NULL)
6296		destroy_dev(zfsdev);
6297}
6298
6299static struct root_hold_token *zfs_root_token;
6300struct proc *zfsproc;
6301
6302#ifdef sun
6303int
6304_init(void)
6305{
6306	int error;
6307
6308	spa_init(FREAD | FWRITE);
6309	zfs_init();
6310	zvol_init();
6311	zfs_ioctl_init();
6312
6313	if ((error = mod_install(&modlinkage)) != 0) {
6314		zvol_fini();
6315		zfs_fini();
6316		spa_fini();
6317		return (error);
6318	}
6319
6320	tsd_create(&zfs_fsyncer_key, NULL);
6321	tsd_create(&rrw_tsd_key, rrw_tsd_destroy);
6322	tsd_create(&zfs_allow_log_key, zfs_allow_log_destroy);
6323
6324	error = ldi_ident_from_mod(&modlinkage, &zfs_li);
6325	ASSERT(error == 0);
6326	mutex_init(&zfs_share_lock, NULL, MUTEX_DEFAULT, NULL);
6327
6328	return (0);
6329}
6330
6331int
6332_fini(void)
6333{
6334	int error;
6335
6336	if (spa_busy() || zfs_busy() || zvol_busy() || zio_injection_enabled)
6337		return (SET_ERROR(EBUSY));
6338
6339	if ((error = mod_remove(&modlinkage)) != 0)
6340		return (error);
6341
6342	zvol_fini();
6343	zfs_fini();
6344	spa_fini();
6345	if (zfs_nfsshare_inited)
6346		(void) ddi_modclose(nfs_mod);
6347	if (zfs_smbshare_inited)
6348		(void) ddi_modclose(smbsrv_mod);
6349	if (zfs_nfsshare_inited || zfs_smbshare_inited)
6350		(void) ddi_modclose(sharefs_mod);
6351
6352	tsd_destroy(&zfs_fsyncer_key);
6353	ldi_ident_release(zfs_li);
6354	zfs_li = NULL;
6355	mutex_destroy(&zfs_share_lock);
6356
6357	return (error);
6358}
6359
6360int
6361_info(struct modinfo *modinfop)
6362{
6363	return (mod_info(&modlinkage, modinfop));
6364}
6365#endif	/* sun */
6366
6367static int zfs__init(void);
6368static int zfs__fini(void);
6369static void zfs_shutdown(void *, int);
6370
6371static eventhandler_tag zfs_shutdown_event_tag;
6372
6373int
6374zfs__init(void)
6375{
6376
6377	zfs_root_token = root_mount_hold("ZFS");
6378
6379	mutex_init(&zfs_share_lock, NULL, MUTEX_DEFAULT, NULL);
6380
6381	spa_init(FREAD | FWRITE);
6382	zfs_init();
6383	zvol_init();
6384	zfs_ioctl_init();
6385
6386	tsd_create(&zfs_fsyncer_key, NULL);
6387	tsd_create(&rrw_tsd_key, rrw_tsd_destroy);
6388	tsd_create(&zfs_allow_log_key, zfs_allow_log_destroy);
6389
6390	printf("ZFS storage pool version: features support (" SPA_VERSION_STRING ")\n");
6391	root_mount_rel(zfs_root_token);
6392
6393	zfsdev_init();
6394
6395	return (0);
6396}
6397
6398int
6399zfs__fini(void)
6400{
6401	if (spa_busy() || zfs_busy() || zvol_busy() ||
6402	    zio_injection_enabled) {
6403		return (EBUSY);
6404	}
6405
6406	zfsdev_fini();
6407	zvol_fini();
6408	zfs_fini();
6409	spa_fini();
6410
6411	tsd_destroy(&zfs_fsyncer_key);
6412	tsd_destroy(&rrw_tsd_key);
6413	tsd_destroy(&zfs_allow_log_key);
6414
6415	mutex_destroy(&zfs_share_lock);
6416
6417	return (0);
6418}
6419
6420static void
6421zfs_shutdown(void *arg __unused, int howto __unused)
6422{
6423
6424	/*
6425	 * ZFS fini routines can not properly work in a panic-ed system.
6426	 */
6427	if (panicstr == NULL)
6428		(void)zfs__fini();
6429}
6430
6431
6432static int
6433zfs_modevent(module_t mod, int type, void *unused __unused)
6434{
6435	int err;
6436
6437	switch (type) {
6438	case MOD_LOAD:
6439		err = zfs__init();
6440		if (err == 0)
6441			zfs_shutdown_event_tag = EVENTHANDLER_REGISTER(
6442			    shutdown_post_sync, zfs_shutdown, NULL,
6443			    SHUTDOWN_PRI_FIRST);
6444		return (err);
6445	case MOD_UNLOAD:
6446		err = zfs__fini();
6447		if (err == 0 && zfs_shutdown_event_tag != NULL)
6448			EVENTHANDLER_DEREGISTER(shutdown_post_sync,
6449			    zfs_shutdown_event_tag);
6450		return (err);
6451	case MOD_SHUTDOWN:
6452		return (0);
6453	default:
6454		break;
6455	}
6456	return (EOPNOTSUPP);
6457}
6458
6459static moduledata_t zfs_mod = {
6460	"zfsctrl",
6461	zfs_modevent,
6462	0
6463};
6464DECLARE_MODULE(zfsctrl, zfs_mod, SI_SUB_VFS, SI_ORDER_ANY);
6465MODULE_VERSION(zfsctrl, 1);
6466MODULE_DEPEND(zfsctrl, opensolaris, 1, 1, 1);
6467MODULE_DEPEND(zfsctrl, krpc, 1, 1, 1);
6468MODULE_DEPEND(zfsctrl, acl_nfs4, 1, 1, 1);
6469