xref: /titanic_51/usr/src/uts/common/fs/zfs/zfs_ioctl.c (revision 14ea4bb737263733ad80a36b4f73f681c30a6b45)
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  * Copyright 2006 Sun Microsystems, Inc.  All rights reserved.
23  * Use is subject to license terms.
24  */
25 
26 #pragma ident	"%Z%%M%	%I%	%E% SMI"
27 
28 #include <sys/types.h>
29 #include <sys/param.h>
30 #include <sys/errno.h>
31 #include <sys/uio.h>
32 #include <sys/buf.h>
33 #include <sys/modctl.h>
34 #include <sys/open.h>
35 #include <sys/file.h>
36 #include <sys/kmem.h>
37 #include <sys/conf.h>
38 #include <sys/cmn_err.h>
39 #include <sys/stat.h>
40 #include <sys/zfs_ioctl.h>
41 #include <sys/zap.h>
42 #include <sys/spa.h>
43 #include <sys/vdev.h>
44 #include <sys/dmu.h>
45 #include <sys/dsl_dir.h>
46 #include <sys/dsl_dataset.h>
47 #include <sys/dsl_prop.h>
48 #include <sys/ddi.h>
49 #include <sys/sunddi.h>
50 #include <sys/sunldi.h>
51 #include <sys/policy.h>
52 #include <sys/zone.h>
53 #include <sys/nvpair.h>
54 #include <sys/pathname.h>
55 #include <sys/mount.h>
56 #include <sys/sdt.h>
57 #include <sys/fs/zfs.h>
58 #include <sys/zfs_ctldir.h>
59 #include <sys/zvol.h>
60 
61 #include "zfs_namecheck.h"
62 #include "zfs_prop.h"
63 
64 extern struct modlfs zfs_modlfs;
65 
66 extern void zfs_init(void);
67 extern void zfs_fini(void);
68 
69 ldi_ident_t zfs_li = NULL;
70 dev_info_t *zfs_dip;
71 
72 typedef int zfs_ioc_func_t(zfs_cmd_t *);
73 typedef int zfs_secpolicy_func_t(const char *, cred_t *);
74 
75 typedef struct zfs_ioc_vec {
76 	zfs_ioc_func_t		*zvec_func;
77 	zfs_secpolicy_func_t	*zvec_secpolicy;
78 	enum {
79 		no_name,
80 		pool_name,
81 		dataset_name
82 	}			zvec_namecheck;
83 } zfs_ioc_vec_t;
84 
85 /* _NOTE(PRINTFLIKE(4)) - this is printf-like, but lint is too whiney */
86 void
87 __dprintf(const char *file, const char *func, int line, const char *fmt, ...)
88 {
89 	const char *newfile;
90 	char buf[256];
91 	va_list adx;
92 
93 	/*
94 	 * Get rid of annoying "../common/" prefix to filename.
95 	 */
96 	newfile = strrchr(file, '/');
97 	if (newfile != NULL) {
98 		newfile = newfile + 1; /* Get rid of leading / */
99 	} else {
100 		newfile = file;
101 	}
102 
103 	va_start(adx, fmt);
104 	(void) vsnprintf(buf, sizeof (buf), fmt, adx);
105 	va_end(adx);
106 
107 	/*
108 	 * To get this data, use the zfs-dprintf probe as so:
109 	 * dtrace -q -n 'zfs-dprintf \
110 	 *	/stringof(arg0) == "dbuf.c"/ \
111 	 *	{printf("%s: %s", stringof(arg1), stringof(arg3))}'
112 	 * arg0 = file name
113 	 * arg1 = function name
114 	 * arg2 = line number
115 	 * arg3 = message
116 	 */
117 	DTRACE_PROBE4(zfs__dprintf,
118 	    char *, newfile, char *, func, int, line, char *, buf);
119 }
120 
121 /*
122  * Policy for top-level read operations (list pools).  Requires no privileges,
123  * and can be used in the local zone, as there is no associated dataset.
124  */
125 /* ARGSUSED */
126 static int
127 zfs_secpolicy_none(const char *unused1, cred_t *cr)
128 {
129 	return (0);
130 }
131 
132 /*
133  * Policy for dataset read operations (list children, get statistics).  Requires
134  * no privileges, but must be visible in the local zone.
135  */
136 /* ARGSUSED */
137 static int
138 zfs_secpolicy_read(const char *dataset, cred_t *cr)
139 {
140 	if (INGLOBALZONE(curproc) ||
141 	    zone_dataset_visible(dataset, NULL))
142 		return (0);
143 
144 	return (ENOENT);
145 }
146 
147 static int
148 zfs_dozonecheck(const char *dataset, cred_t *cr)
149 {
150 	uint64_t zoned;
151 	int writable = 1;
152 
153 	/*
154 	 * The dataset must be visible by this zone -- check this first
155 	 * so they don't see EPERM on something they shouldn't know about.
156 	 */
157 	if (!INGLOBALZONE(curproc) &&
158 	    !zone_dataset_visible(dataset, &writable))
159 		return (ENOENT);
160 
161 	if (dsl_prop_get_integer(dataset, "zoned", &zoned, NULL))
162 		return (ENOENT);
163 
164 	if (INGLOBALZONE(curproc)) {
165 		/*
166 		 * If the fs is zoned, only root can access it from the
167 		 * global zone.
168 		 */
169 		if (secpolicy_zfs(cr) && zoned)
170 			return (EPERM);
171 	} else {
172 		/*
173 		 * If we are in a local zone, the 'zoned' property must be set.
174 		 */
175 		if (!zoned)
176 			return (EPERM);
177 
178 		/* must be writable by this zone */
179 		if (!writable)
180 			return (EPERM);
181 	}
182 	return (0);
183 }
184 
185 /*
186  * Policy for dataset write operations (create children, set properties, etc).
187  * Requires SYS_MOUNT privilege, and must be writable in the local zone.
188  */
189 int
190 zfs_secpolicy_write(const char *dataset, cred_t *cr)
191 {
192 	int error;
193 
194 	if (error = zfs_dozonecheck(dataset, cr))
195 		return (error);
196 
197 	return (secpolicy_zfs(cr));
198 }
199 
200 /*
201  * Policy for operations that want to write a dataset's parent:
202  * create, destroy, snapshot, clone, restore.
203  */
204 static int
205 zfs_secpolicy_parent(const char *dataset, cred_t *cr)
206 {
207 	char parentname[MAXNAMELEN];
208 	char *cp;
209 
210 	/*
211 	 * Remove the @bla or /bla from the end of the name to get the parent.
212 	 */
213 	(void) strncpy(parentname, dataset, sizeof (parentname));
214 	cp = strrchr(parentname, '@');
215 	if (cp != NULL) {
216 		cp[0] = '\0';
217 	} else {
218 		cp = strrchr(parentname, '/');
219 		if (cp == NULL)
220 			return (ENOENT);
221 		cp[0] = '\0';
222 
223 	}
224 
225 	return (zfs_secpolicy_write(parentname, cr));
226 }
227 
228 /*
229  * Policy for pool operations - create/destroy pools, add vdevs, etc.  Requires
230  * SYS_CONFIG privilege, which is not available in a local zone.
231  */
232 /* ARGSUSED */
233 static int
234 zfs_secpolicy_config(const char *unused, cred_t *cr)
235 {
236 	if (secpolicy_sys_config(cr, B_FALSE) != 0)
237 		return (EPERM);
238 
239 	return (0);
240 }
241 
242 /*
243  * Policy for fault injection.  Requires all privileges.
244  */
245 /* ARGSUSED */
246 static int
247 zfs_secpolicy_inject(const char *unused, cred_t *cr)
248 {
249 	return (secpolicy_zinject(cr));
250 }
251 
252 /*
253  * Returns the nvlist as specified by the user in the zfs_cmd_t.
254  */
255 static int
256 get_nvlist(zfs_cmd_t *zc, nvlist_t **nvp)
257 {
258 	char *packed;
259 	size_t size;
260 	int error;
261 	nvlist_t *config = NULL;
262 
263 	/*
264 	 * Read in and unpack the user-supplied nvlist.
265 	 */
266 	if ((size = zc->zc_nvlist_src_size) == 0)
267 		return (EINVAL);
268 
269 	packed = kmem_alloc(size, KM_SLEEP);
270 
271 	if ((error = xcopyin((void *)(uintptr_t)zc->zc_nvlist_src, packed,
272 	    size)) != 0) {
273 		kmem_free(packed, size);
274 		return (error);
275 	}
276 
277 	if ((error = nvlist_unpack(packed, size, &config, 0)) != 0) {
278 		kmem_free(packed, size);
279 		return (error);
280 	}
281 
282 	kmem_free(packed, size);
283 
284 	*nvp = config;
285 	return (0);
286 }
287 
288 static int
289 put_nvlist(zfs_cmd_t *zc, nvlist_t *nvl)
290 {
291 	char *packed = NULL;
292 	size_t size;
293 	int error;
294 
295 	VERIFY(nvlist_size(nvl, &size, NV_ENCODE_NATIVE) == 0);
296 
297 	if (size > zc->zc_nvlist_dst_size) {
298 		error = ENOMEM;
299 	} else {
300 		VERIFY(nvlist_pack(nvl, &packed, &size, NV_ENCODE_NATIVE,
301 		    KM_SLEEP) == 0);
302 		error = xcopyout(packed, (void *)(uintptr_t)zc->zc_nvlist_dst,
303 		    size);
304 		kmem_free(packed, size);
305 	}
306 
307 	zc->zc_nvlist_dst_size = size;
308 	return (error);
309 }
310 
311 static int
312 zfs_ioc_pool_create(zfs_cmd_t *zc)
313 {
314 	int error;
315 	nvlist_t *config;
316 
317 	if ((error = get_nvlist(zc, &config)) != 0)
318 		return (error);
319 
320 	error = spa_create(zc->zc_name, config, zc->zc_value[0] == '\0' ?
321 	    NULL : zc->zc_value);
322 
323 	nvlist_free(config);
324 
325 	return (error);
326 }
327 
328 static int
329 zfs_ioc_pool_destroy(zfs_cmd_t *zc)
330 {
331 	return (spa_destroy(zc->zc_name));
332 }
333 
334 static int
335 zfs_ioc_pool_import(zfs_cmd_t *zc)
336 {
337 	int error;
338 	nvlist_t *config;
339 	uint64_t guid;
340 
341 	if ((error = get_nvlist(zc, &config)) != 0)
342 		return (error);
343 
344 	if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, &guid) != 0 ||
345 	    guid != zc->zc_guid)
346 		error = EINVAL;
347 	else
348 		error = spa_import(zc->zc_name, config,
349 		    zc->zc_value[0] == '\0' ? NULL : zc->zc_value);
350 
351 	nvlist_free(config);
352 
353 	return (error);
354 }
355 
356 static int
357 zfs_ioc_pool_export(zfs_cmd_t *zc)
358 {
359 	return (spa_export(zc->zc_name, NULL));
360 }
361 
362 static int
363 zfs_ioc_pool_configs(zfs_cmd_t *zc)
364 {
365 	nvlist_t *configs;
366 	int error;
367 
368 	if ((configs = spa_all_configs(&zc->zc_cookie)) == NULL)
369 		return (EEXIST);
370 
371 	error = put_nvlist(zc, configs);
372 
373 	nvlist_free(configs);
374 
375 	return (error);
376 }
377 
378 static int
379 zfs_ioc_pool_stats(zfs_cmd_t *zc)
380 {
381 	nvlist_t *config;
382 	int error;
383 	int ret = 0;
384 
385 	error = spa_get_stats(zc->zc_name, &config, zc->zc_value,
386 	    sizeof (zc->zc_value));
387 
388 	if (config != NULL) {
389 		ret = put_nvlist(zc, config);
390 		nvlist_free(config);
391 
392 		/*
393 		 * The config may be present even if 'error' is non-zero.
394 		 * In this case we return success, and preserve the real errno
395 		 * in 'zc_cookie'.
396 		 */
397 		zc->zc_cookie = error;
398 	} else {
399 		ret = error;
400 	}
401 
402 	return (ret);
403 }
404 
405 /*
406  * Try to import the given pool, returning pool stats as appropriate so that
407  * user land knows which devices are available and overall pool health.
408  */
409 static int
410 zfs_ioc_pool_tryimport(zfs_cmd_t *zc)
411 {
412 	nvlist_t *tryconfig, *config;
413 	int error;
414 
415 	if ((error = get_nvlist(zc, &tryconfig)) != 0)
416 		return (error);
417 
418 	config = spa_tryimport(tryconfig);
419 
420 	nvlist_free(tryconfig);
421 
422 	if (config == NULL)
423 		return (EINVAL);
424 
425 	error = put_nvlist(zc, config);
426 	nvlist_free(config);
427 
428 	return (error);
429 }
430 
431 static int
432 zfs_ioc_pool_scrub(zfs_cmd_t *zc)
433 {
434 	spa_t *spa;
435 	int error;
436 
437 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
438 		return (error);
439 
440 	error = spa_scrub(spa, zc->zc_cookie, B_FALSE);
441 
442 	spa_close(spa, FTAG);
443 
444 	return (error);
445 }
446 
447 static int
448 zfs_ioc_pool_freeze(zfs_cmd_t *zc)
449 {
450 	spa_t *spa;
451 	int error;
452 
453 	error = spa_open(zc->zc_name, &spa, FTAG);
454 	if (error == 0) {
455 		spa_freeze(spa);
456 		spa_close(spa, FTAG);
457 	}
458 	return (error);
459 }
460 
461 static int
462 zfs_ioc_pool_upgrade(zfs_cmd_t *zc)
463 {
464 	spa_t *spa;
465 	int error;
466 
467 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
468 		return (error);
469 
470 	spa_upgrade(spa);
471 
472 	spa_close(spa, FTAG);
473 
474 	return (error);
475 }
476 
477 static int
478 zfs_ioc_pool_get_history(zfs_cmd_t *zc)
479 {
480 	spa_t *spa;
481 	char *hist_buf;
482 	uint64_t size;
483 	int error;
484 
485 	if ((size = zc->zc_history_len) == 0)
486 		return (EINVAL);
487 
488 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
489 		return (error);
490 
491 	hist_buf = kmem_alloc(size, KM_SLEEP);
492 	if ((error = spa_history_get(spa, &zc->zc_history_offset,
493 	    &zc->zc_history_len, hist_buf)) == 0) {
494 		error = xcopyout(hist_buf, (char *)(uintptr_t)zc->zc_history,
495 		    zc->zc_history_len);
496 	}
497 
498 	spa_close(spa, FTAG);
499 	kmem_free(hist_buf, size);
500 	return (error);
501 }
502 
503 static int
504 zfs_ioc_pool_log_history(zfs_cmd_t *zc)
505 {
506 	spa_t *spa;
507 	char *history_str = NULL;
508 	size_t size;
509 	int error;
510 
511 	size = zc->zc_history_len;
512 	if (size == 0 || size > HIS_MAX_RECORD_LEN)
513 		return (EINVAL);
514 
515 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
516 		return (error);
517 
518 	/* add one for the NULL delimiter */
519 	size++;
520 	history_str = kmem_alloc(size, KM_SLEEP);
521 	if ((error = xcopyin((void *)(uintptr_t)zc->zc_history, history_str,
522 	    size)) != 0) {
523 		spa_close(spa, FTAG);
524 		kmem_free(history_str, size);
525 		return (error);
526 	}
527 	history_str[size - 1] = '\0';
528 
529 	error = spa_history_log(spa, history_str, zc->zc_history_offset);
530 
531 	spa_close(spa, FTAG);
532 	kmem_free(history_str, size);
533 
534 	return (error);
535 }
536 
537 static int
538 zfs_ioc_vdev_add(zfs_cmd_t *zc)
539 {
540 	spa_t *spa;
541 	int error;
542 	nvlist_t *config;
543 
544 	error = spa_open(zc->zc_name, &spa, FTAG);
545 	if (error != 0)
546 		return (error);
547 
548 	if ((error = get_nvlist(zc, &config)) == 0) {
549 		error = spa_vdev_add(spa, config);
550 		nvlist_free(config);
551 	}
552 
553 	spa_close(spa, FTAG);
554 	return (error);
555 }
556 
557 static int
558 zfs_ioc_vdev_remove(zfs_cmd_t *zc)
559 {
560 	spa_t *spa;
561 	int error;
562 
563 	error = spa_open(zc->zc_name, &spa, FTAG);
564 	if (error != 0)
565 		return (error);
566 	error = spa_vdev_remove(spa, zc->zc_guid, B_FALSE);
567 	spa_close(spa, FTAG);
568 	return (error);
569 }
570 
571 static int
572 zfs_ioc_vdev_online(zfs_cmd_t *zc)
573 {
574 	spa_t *spa;
575 	int error;
576 
577 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
578 		return (error);
579 	error = vdev_online(spa, zc->zc_guid);
580 	spa_close(spa, FTAG);
581 	return (error);
582 }
583 
584 static int
585 zfs_ioc_vdev_offline(zfs_cmd_t *zc)
586 {
587 	spa_t *spa;
588 	int istmp = zc->zc_cookie;
589 	int error;
590 
591 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
592 		return (error);
593 	error = vdev_offline(spa, zc->zc_guid, istmp);
594 	spa_close(spa, FTAG);
595 	return (error);
596 }
597 
598 static int
599 zfs_ioc_vdev_attach(zfs_cmd_t *zc)
600 {
601 	spa_t *spa;
602 	int replacing = zc->zc_cookie;
603 	nvlist_t *config;
604 	int error;
605 
606 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
607 		return (error);
608 
609 	if ((error = get_nvlist(zc, &config)) == 0) {
610 		error = spa_vdev_attach(spa, zc->zc_guid, config, replacing);
611 		nvlist_free(config);
612 	}
613 
614 	spa_close(spa, FTAG);
615 	return (error);
616 }
617 
618 static int
619 zfs_ioc_vdev_detach(zfs_cmd_t *zc)
620 {
621 	spa_t *spa;
622 	int error;
623 
624 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
625 		return (error);
626 
627 	error = spa_vdev_detach(spa, zc->zc_guid, B_FALSE);
628 
629 	spa_close(spa, FTAG);
630 	return (error);
631 }
632 
633 static int
634 zfs_ioc_vdev_setpath(zfs_cmd_t *zc)
635 {
636 	spa_t *spa;
637 	char *path = zc->zc_value;
638 	uint64_t guid = zc->zc_guid;
639 	int error;
640 
641 	error = spa_open(zc->zc_name, &spa, FTAG);
642 	if (error != 0)
643 		return (error);
644 
645 	error = spa_vdev_setpath(spa, guid, path);
646 	spa_close(spa, FTAG);
647 	return (error);
648 }
649 
650 static int
651 zfs_ioc_objset_stats(zfs_cmd_t *zc)
652 {
653 	objset_t *os = NULL;
654 	int error;
655 	nvlist_t *nv;
656 
657 retry:
658 	error = dmu_objset_open(zc->zc_name, DMU_OST_ANY,
659 	    DS_MODE_STANDARD | DS_MODE_READONLY, &os);
660 	if (error != 0) {
661 		/*
662 		 * This is ugly: dmu_objset_open() can return EBUSY if
663 		 * the objset is held exclusively. Fortunately this hold is
664 		 * only for a short while, so we retry here.
665 		 * This avoids user code having to handle EBUSY,
666 		 * for example for a "zfs list".
667 		 */
668 		if (error == EBUSY) {
669 			delay(1);
670 			goto retry;
671 		}
672 		return (error);
673 	}
674 
675 	dmu_objset_fast_stat(os, &zc->zc_objset_stats);
676 
677 	if (zc->zc_nvlist_dst != 0 &&
678 	    (error = dsl_prop_get_all(os, &nv)) == 0) {
679 		dmu_objset_stats(os, nv);
680 		/*
681 		 * NB: zvol_get_stats() will read the objset contents,
682 		 * which we aren't supposed to do with a
683 		 * DS_MODE_STANDARD open, because it could be
684 		 * inconsistent.  So this is a bit of a workaround...
685 		 */
686 		if (!zc->zc_objset_stats.dds_inconsistent &&
687 		    dmu_objset_type(os) == DMU_OST_ZVOL)
688 			VERIFY(zvol_get_stats(os, nv) == 0);
689 		error = put_nvlist(zc, nv);
690 		nvlist_free(nv);
691 	}
692 
693 	spa_altroot(dmu_objset_spa(os), zc->zc_value, sizeof (zc->zc_value));
694 
695 	dmu_objset_close(os);
696 	return (error);
697 }
698 
699 static int
700 zfs_ioc_dataset_list_next(zfs_cmd_t *zc)
701 {
702 	objset_t *os;
703 	int error;
704 	char *p;
705 
706 retry:
707 	error = dmu_objset_open(zc->zc_name, DMU_OST_ANY,
708 	    DS_MODE_STANDARD | DS_MODE_READONLY, &os);
709 	if (error != 0) {
710 		/*
711 		 * This is ugly: dmu_objset_open() can return EBUSY if
712 		 * the objset is held exclusively. Fortunately this hold is
713 		 * only for a short while, so we retry here.
714 		 * This avoids user code having to handle EBUSY,
715 		 * for example for a "zfs list".
716 		 */
717 		if (error == EBUSY) {
718 			delay(1);
719 			goto retry;
720 		}
721 		if (error == ENOENT)
722 			error = ESRCH;
723 		return (error);
724 	}
725 
726 	p = strrchr(zc->zc_name, '/');
727 	if (p == NULL || p[1] != '\0')
728 		(void) strlcat(zc->zc_name, "/", sizeof (zc->zc_name));
729 	p = zc->zc_name + strlen(zc->zc_name);
730 
731 	do {
732 		error = dmu_dir_list_next(os,
733 		    sizeof (zc->zc_name) - (p - zc->zc_name), p,
734 		    NULL, &zc->zc_cookie);
735 		if (error == ENOENT)
736 			error = ESRCH;
737 	} while (error == 0 && !INGLOBALZONE(curproc) &&
738 	    !zone_dataset_visible(zc->zc_name, NULL));
739 
740 	/*
741 	 * If it's a hidden dataset (ie. with a '$' in its name), don't
742 	 * try to get stats for it.  Userland will skip over it.
743 	 */
744 	if (error == 0 && strchr(zc->zc_name, '$') == NULL)
745 		error = zfs_ioc_objset_stats(zc); /* fill in the stats */
746 
747 	dmu_objset_close(os);
748 	return (error);
749 }
750 
751 static int
752 zfs_ioc_snapshot_list_next(zfs_cmd_t *zc)
753 {
754 	objset_t *os;
755 	int error;
756 
757 retry:
758 	error = dmu_objset_open(zc->zc_name, DMU_OST_ANY,
759 	    DS_MODE_STANDARD | DS_MODE_READONLY, &os);
760 	if (error != 0) {
761 		/*
762 		 * This is ugly: dmu_objset_open() can return EBUSY if
763 		 * the objset is held exclusively. Fortunately this hold is
764 		 * only for a short while, so we retry here.
765 		 * This avoids user code having to handle EBUSY,
766 		 * for example for a "zfs list".
767 		 */
768 		if (error == EBUSY) {
769 			delay(1);
770 			goto retry;
771 		}
772 		if (error == ENOENT)
773 			error = ESRCH;
774 		return (error);
775 	}
776 
777 	/*
778 	 * A dataset name of maximum length cannot have any snapshots,
779 	 * so exit immediately.
780 	 */
781 	if (strlcat(zc->zc_name, "@", sizeof (zc->zc_name)) >= MAXNAMELEN) {
782 		dmu_objset_close(os);
783 		return (ESRCH);
784 	}
785 
786 	error = dmu_snapshot_list_next(os,
787 	    sizeof (zc->zc_name) - strlen(zc->zc_name),
788 	    zc->zc_name + strlen(zc->zc_name), NULL, &zc->zc_cookie);
789 	if (error == ENOENT)
790 		error = ESRCH;
791 
792 	if (error == 0)
793 		error = zfs_ioc_objset_stats(zc); /* fill in the stats */
794 
795 	dmu_objset_close(os);
796 	return (error);
797 }
798 
799 static int
800 zfs_set_prop_nvlist(const char *name, dev_t dev, cred_t *cr, nvlist_t *nvl)
801 {
802 	nvpair_t *elem;
803 	int error;
804 	const char *propname;
805 	zfs_prop_t prop;
806 	uint64_t intval;
807 	char *strval;
808 
809 	elem = NULL;
810 	while ((elem = nvlist_next_nvpair(nvl, elem)) != NULL) {
811 		propname = nvpair_name(elem);
812 
813 		if ((prop = zfs_name_to_prop(propname)) ==
814 		    ZFS_PROP_INVAL) {
815 			/*
816 			 * If this is a user-defined property, it must be a
817 			 * string, and there is no further validation to do.
818 			 */
819 			if (!zfs_prop_user(propname) ||
820 			    nvpair_type(elem) != DATA_TYPE_STRING)
821 				return (EINVAL);
822 
823 			VERIFY(nvpair_value_string(elem, &strval) == 0);
824 			error = dsl_prop_set(name, propname, 1,
825 			    strlen(strval) + 1, strval);
826 			if (error == 0)
827 				continue;
828 			else
829 				break;
830 		}
831 
832 		/*
833 		 * Check permissions for special properties.
834 		 */
835 		switch (prop) {
836 		case ZFS_PROP_ZONED:
837 			/*
838 			 * Disallow setting of 'zoned' from within a local zone.
839 			 */
840 			if (!INGLOBALZONE(curproc))
841 				return (EPERM);
842 			break;
843 
844 		case ZFS_PROP_QUOTA:
845 			if (error = zfs_dozonecheck(name, cr))
846 				return (error);
847 
848 			if (!INGLOBALZONE(curproc)) {
849 				uint64_t zoned;
850 				char setpoint[MAXNAMELEN];
851 				int dslen;
852 				/*
853 				 * Unprivileged users are allowed to modify the
854 				 * quota on things *under* (ie. contained by)
855 				 * the thing they own.
856 				 */
857 				if (dsl_prop_get_integer(name, "zoned", &zoned,
858 				    setpoint))
859 					return (EPERM);
860 				if (!zoned) /* this shouldn't happen */
861 					return (EPERM);
862 				dslen = strlen(name);
863 				if (dslen <= strlen(setpoint))
864 					return (EPERM);
865 			}
866 		}
867 
868 		switch (prop) {
869 		case ZFS_PROP_QUOTA:
870 			if ((error = nvpair_value_uint64(elem, &intval)) != 0 ||
871 			    (error = dsl_dir_set_quota(name,
872 			    intval)) != 0)
873 				return (error);
874 			break;
875 
876 		case ZFS_PROP_RESERVATION:
877 			if ((error = nvpair_value_uint64(elem, &intval)) != 0 ||
878 			    (error = dsl_dir_set_reservation(name,
879 			    intval)) != 0)
880 				return (error);
881 			break;
882 
883 		case ZFS_PROP_VOLSIZE:
884 			if ((error = nvpair_value_uint64(elem, &intval)) != 0 ||
885 			    (error = zvol_set_volsize(name, dev,
886 			    intval)) != 0)
887 				return (error);
888 			break;
889 
890 		case ZFS_PROP_VOLBLOCKSIZE:
891 			if ((error = nvpair_value_uint64(elem, &intval)) != 0 ||
892 			    (error = zvol_set_volblocksize(name,
893 			    intval)) != 0)
894 				return (error);
895 			break;
896 
897 		default:
898 			if (nvpair_type(elem) == DATA_TYPE_STRING) {
899 				if (zfs_prop_get_type(prop) !=
900 				    prop_type_string)
901 					return (EINVAL);
902 				VERIFY(nvpair_value_string(elem, &strval) == 0);
903 				if ((error = dsl_prop_set(name,
904 				    nvpair_name(elem), 1, strlen(strval) + 1,
905 				    strval)) != 0)
906 					return (error);
907 			} else if (nvpair_type(elem) == DATA_TYPE_UINT64) {
908 				const char *unused;
909 
910 				VERIFY(nvpair_value_uint64(elem, &intval) == 0);
911 
912 				switch (zfs_prop_get_type(prop)) {
913 				case prop_type_number:
914 					break;
915 				case prop_type_boolean:
916 					if (intval > 1)
917 						return (EINVAL);
918 					break;
919 				case prop_type_string:
920 					return (EINVAL);
921 				case prop_type_index:
922 					if (zfs_prop_index_to_string(prop,
923 					    intval, &unused) != 0)
924 						return (EINVAL);
925 					break;
926 				default:
927 					cmn_err(CE_PANIC, "unknown property "
928 					    "type");
929 					break;
930 				}
931 
932 				if ((error = dsl_prop_set(name, propname,
933 				    8, 1, &intval)) != 0)
934 					return (error);
935 			} else {
936 				return (EINVAL);
937 			}
938 			break;
939 		}
940 	}
941 
942 	return (0);
943 }
944 
945 static int
946 zfs_ioc_set_prop(zfs_cmd_t *zc)
947 {
948 	nvlist_t *nvl;
949 	int error;
950 	zfs_prop_t prop;
951 
952 	/*
953 	 * If zc_value is set, then this is an attempt to inherit a value.
954 	 * Otherwise, zc_nvlist refers to a list of properties to set.
955 	 */
956 	if (zc->zc_value[0] != '\0') {
957 		if (!zfs_prop_user(zc->zc_value) &&
958 		    ((prop = zfs_name_to_prop(zc->zc_value)) ==
959 		    ZFS_PROP_INVAL ||
960 		    !zfs_prop_inheritable(prop)))
961 			return (EINVAL);
962 
963 		return (dsl_prop_set(zc->zc_name, zc->zc_value, 0, 0, NULL));
964 	}
965 
966 	if ((error = get_nvlist(zc, &nvl)) != 0)
967 		return (error);
968 
969 	error = zfs_set_prop_nvlist(zc->zc_name, zc->zc_dev,
970 	    (cred_t *)(uintptr_t)zc->zc_cred, nvl);
971 	nvlist_free(nvl);
972 	return (error);
973 }
974 
975 static int
976 zfs_ioc_create_minor(zfs_cmd_t *zc)
977 {
978 	return (zvol_create_minor(zc->zc_name, zc->zc_dev));
979 }
980 
981 static int
982 zfs_ioc_remove_minor(zfs_cmd_t *zc)
983 {
984 	return (zvol_remove_minor(zc->zc_name));
985 }
986 
987 /*
988  * Search the vfs list for a specified resource.  Returns a pointer to it
989  * or NULL if no suitable entry is found. The caller of this routine
990  * is responsible for releasing the returned vfs pointer.
991  */
992 static vfs_t *
993 zfs_get_vfs(const char *resource)
994 {
995 	struct vfs *vfsp;
996 	struct vfs *vfs_found = NULL;
997 
998 	vfs_list_read_lock();
999 	vfsp = rootvfs;
1000 	do {
1001 		if (strcmp(refstr_value(vfsp->vfs_resource), resource) == 0) {
1002 			VFS_HOLD(vfsp);
1003 			vfs_found = vfsp;
1004 			break;
1005 		}
1006 		vfsp = vfsp->vfs_next;
1007 	} while (vfsp != rootvfs);
1008 	vfs_list_unlock();
1009 	return (vfs_found);
1010 }
1011 
1012 static void
1013 zfs_create_cb(objset_t *os, void *arg, dmu_tx_t *tx)
1014 {
1015 	zfs_create_data_t *zc = arg;
1016 	zfs_create_fs(os, (cred_t *)(uintptr_t)zc->zc_cred, tx);
1017 }
1018 
1019 static int
1020 zfs_ioc_create(zfs_cmd_t *zc)
1021 {
1022 	objset_t *clone;
1023 	int error = 0;
1024 	zfs_create_data_t cbdata = { 0 };
1025 	void (*cbfunc)(objset_t *os, void *arg, dmu_tx_t *tx);
1026 	dmu_objset_type_t type = zc->zc_objset_type;
1027 
1028 	switch (type) {
1029 
1030 	case DMU_OST_ZFS:
1031 		cbfunc = zfs_create_cb;
1032 		break;
1033 
1034 	case DMU_OST_ZVOL:
1035 		cbfunc = zvol_create_cb;
1036 		break;
1037 
1038 	default:
1039 		cbfunc = NULL;
1040 	}
1041 	if (strchr(zc->zc_name, '@'))
1042 		return (EINVAL);
1043 
1044 	if (zc->zc_nvlist_src != NULL &&
1045 	    (error = get_nvlist(zc, &cbdata.zc_props)) != 0)
1046 		return (error);
1047 
1048 	cbdata.zc_cred = (cred_t *)(uintptr_t)zc->zc_cred;
1049 	cbdata.zc_dev = (dev_t)zc->zc_dev;
1050 
1051 	if (zc->zc_value[0] != '\0') {
1052 		/*
1053 		 * We're creating a clone of an existing snapshot.
1054 		 */
1055 		zc->zc_value[sizeof (zc->zc_value) - 1] = '\0';
1056 		if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0) {
1057 			nvlist_free(cbdata.zc_props);
1058 			return (EINVAL);
1059 		}
1060 
1061 		error = dmu_objset_open(zc->zc_value, type,
1062 		    DS_MODE_STANDARD | DS_MODE_READONLY, &clone);
1063 		if (error) {
1064 			nvlist_free(cbdata.zc_props);
1065 			return (error);
1066 		}
1067 		error = dmu_objset_create(zc->zc_name, type, clone, NULL, NULL);
1068 		dmu_objset_close(clone);
1069 	} else {
1070 		if (cbfunc == NULL) {
1071 			nvlist_free(cbdata.zc_props);
1072 			return (EINVAL);
1073 		}
1074 
1075 		if (type == DMU_OST_ZVOL) {
1076 			uint64_t volsize, volblocksize;
1077 
1078 			if (cbdata.zc_props == NULL ||
1079 			    nvlist_lookup_uint64(cbdata.zc_props,
1080 			    zfs_prop_to_name(ZFS_PROP_VOLSIZE),
1081 			    &volsize) != 0) {
1082 				nvlist_free(cbdata.zc_props);
1083 				return (EINVAL);
1084 			}
1085 
1086 			if ((error = nvlist_lookup_uint64(cbdata.zc_props,
1087 			    zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE),
1088 			    &volblocksize)) != 0 && error != ENOENT) {
1089 				nvlist_free(cbdata.zc_props);
1090 				return (EINVAL);
1091 			}
1092 
1093 			if (error != 0)
1094 				volblocksize = zfs_prop_default_numeric(
1095 				    ZFS_PROP_VOLBLOCKSIZE);
1096 
1097 			if ((error = zvol_check_volblocksize(
1098 			    volblocksize)) != 0 ||
1099 			    (error = zvol_check_volsize(volsize,
1100 			    volblocksize)) != 0) {
1101 				nvlist_free(cbdata.zc_props);
1102 				return (error);
1103 			}
1104 		}
1105 
1106 		error = dmu_objset_create(zc->zc_name, type, NULL, cbfunc,
1107 		    &cbdata);
1108 	}
1109 
1110 	/*
1111 	 * It would be nice to do this atomically.
1112 	 */
1113 	if (error == 0) {
1114 		if ((error = zfs_set_prop_nvlist(zc->zc_name,
1115 		    zc->zc_dev, (cred_t *)(uintptr_t)zc->zc_cred,
1116 		    cbdata.zc_props)) != 0)
1117 			(void) dmu_objset_destroy(zc->zc_name);
1118 	}
1119 
1120 	nvlist_free(cbdata.zc_props);
1121 	return (error);
1122 }
1123 
1124 static int
1125 zfs_ioc_snapshot(zfs_cmd_t *zc)
1126 {
1127 	if (snapshot_namecheck(zc->zc_value, NULL, NULL) != 0)
1128 		return (EINVAL);
1129 	return (dmu_objset_snapshot(zc->zc_name,
1130 	    zc->zc_value, zc->zc_cookie));
1131 }
1132 
1133 static int
1134 zfs_unmount_snap(char *name, void *arg)
1135 {
1136 	char *snapname = arg;
1137 	char *cp;
1138 	vfs_t *vfsp = NULL;
1139 
1140 	/*
1141 	 * Snapshots (which are under .zfs control) must be unmounted
1142 	 * before they can be destroyed.
1143 	 */
1144 
1145 	if (snapname) {
1146 		(void) strcat(name, "@");
1147 		(void) strcat(name, snapname);
1148 		vfsp = zfs_get_vfs(name);
1149 		cp = strchr(name, '@');
1150 		*cp = '\0';
1151 	} else if (strchr(name, '@')) {
1152 		vfsp = zfs_get_vfs(name);
1153 	}
1154 
1155 	if (vfsp) {
1156 		/*
1157 		 * Always force the unmount for snapshots.
1158 		 */
1159 		int flag = MS_FORCE;
1160 		int err;
1161 
1162 		if ((err = vn_vfswlock(vfsp->vfs_vnodecovered)) != 0) {
1163 			VFS_RELE(vfsp);
1164 			return (err);
1165 		}
1166 		VFS_RELE(vfsp);
1167 		if ((err = dounmount(vfsp, flag, kcred)) != 0)
1168 			return (err);
1169 	}
1170 	return (0);
1171 }
1172 
1173 static int
1174 zfs_ioc_destroy_snaps(zfs_cmd_t *zc)
1175 {
1176 	int err;
1177 
1178 	if (snapshot_namecheck(zc->zc_value, NULL, NULL) != 0)
1179 		return (EINVAL);
1180 	err = dmu_objset_find(zc->zc_name,
1181 	    zfs_unmount_snap, zc->zc_value, DS_FIND_CHILDREN);
1182 	if (err)
1183 		return (err);
1184 	return (dmu_snapshots_destroy(zc->zc_name, zc->zc_value));
1185 }
1186 
1187 static int
1188 zfs_ioc_destroy(zfs_cmd_t *zc)
1189 {
1190 	if (strchr(zc->zc_name, '@') && zc->zc_objset_type == DMU_OST_ZFS) {
1191 		int err = zfs_unmount_snap(zc->zc_name, NULL);
1192 		if (err)
1193 			return (err);
1194 	}
1195 
1196 	return (dmu_objset_destroy(zc->zc_name));
1197 }
1198 
1199 static int
1200 zfs_ioc_rollback(zfs_cmd_t *zc)
1201 {
1202 	return (dmu_objset_rollback(zc->zc_name));
1203 }
1204 
1205 static int
1206 zfs_ioc_rename(zfs_cmd_t *zc)
1207 {
1208 	zc->zc_value[sizeof (zc->zc_value) - 1] = '\0';
1209 	if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0)
1210 		return (EINVAL);
1211 
1212 	if (strchr(zc->zc_name, '@') != NULL &&
1213 	    zc->zc_objset_type == DMU_OST_ZFS) {
1214 		int err = zfs_unmount_snap(zc->zc_name, NULL);
1215 		if (err)
1216 			return (err);
1217 	}
1218 
1219 	return (dmu_objset_rename(zc->zc_name, zc->zc_value));
1220 }
1221 
1222 static int
1223 zfs_ioc_recvbackup(zfs_cmd_t *zc)
1224 {
1225 	file_t *fp;
1226 	int error, fd;
1227 	offset_t new_off;
1228 
1229 	if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
1230 	    strchr(zc->zc_value, '@') == NULL)
1231 		return (EINVAL);
1232 
1233 	fd = zc->zc_cookie;
1234 	fp = getf(fd);
1235 	if (fp == NULL)
1236 		return (EBADF);
1237 	error = dmu_recvbackup(zc->zc_value, &zc->zc_begin_record,
1238 	    &zc->zc_cookie, (boolean_t)zc->zc_guid, fp->f_vnode,
1239 	    fp->f_offset);
1240 
1241 	new_off = fp->f_offset + zc->zc_cookie;
1242 	if (VOP_SEEK(fp->f_vnode, fp->f_offset, &new_off) == 0)
1243 		fp->f_offset = new_off;
1244 
1245 	releasef(fd);
1246 	return (error);
1247 }
1248 
1249 static int
1250 zfs_ioc_sendbackup(zfs_cmd_t *zc)
1251 {
1252 	objset_t *fromsnap = NULL;
1253 	objset_t *tosnap;
1254 	file_t *fp;
1255 	int error;
1256 
1257 	error = dmu_objset_open(zc->zc_name, DMU_OST_ANY,
1258 	    DS_MODE_STANDARD | DS_MODE_READONLY, &tosnap);
1259 	if (error)
1260 		return (error);
1261 
1262 	if (zc->zc_value[0] != '\0') {
1263 		char buf[MAXPATHLEN];
1264 		char *cp;
1265 
1266 		(void) strncpy(buf, zc->zc_name, sizeof (buf));
1267 		cp = strchr(buf, '@');
1268 		if (cp)
1269 			*(cp+1) = 0;
1270 		(void) strncat(buf, zc->zc_value, sizeof (buf));
1271 		error = dmu_objset_open(buf, DMU_OST_ANY,
1272 		    DS_MODE_STANDARD | DS_MODE_READONLY, &fromsnap);
1273 		if (error) {
1274 			dmu_objset_close(tosnap);
1275 			return (error);
1276 		}
1277 	}
1278 
1279 	fp = getf(zc->zc_cookie);
1280 	if (fp == NULL) {
1281 		dmu_objset_close(tosnap);
1282 		if (fromsnap)
1283 			dmu_objset_close(fromsnap);
1284 		return (EBADF);
1285 	}
1286 
1287 	error = dmu_sendbackup(tosnap, fromsnap, fp->f_vnode);
1288 
1289 	releasef(zc->zc_cookie);
1290 	if (fromsnap)
1291 		dmu_objset_close(fromsnap);
1292 	dmu_objset_close(tosnap);
1293 	return (error);
1294 }
1295 
1296 static int
1297 zfs_ioc_inject_fault(zfs_cmd_t *zc)
1298 {
1299 	int id, error;
1300 
1301 	error = zio_inject_fault(zc->zc_name, (int)zc->zc_guid, &id,
1302 	    &zc->zc_inject_record);
1303 
1304 	if (error == 0)
1305 		zc->zc_guid = (uint64_t)id;
1306 
1307 	return (error);
1308 }
1309 
1310 static int
1311 zfs_ioc_clear_fault(zfs_cmd_t *zc)
1312 {
1313 	return (zio_clear_fault((int)zc->zc_guid));
1314 }
1315 
1316 static int
1317 zfs_ioc_inject_list_next(zfs_cmd_t *zc)
1318 {
1319 	int id = (int)zc->zc_guid;
1320 	int error;
1321 
1322 	error = zio_inject_list_next(&id, zc->zc_name, sizeof (zc->zc_name),
1323 	    &zc->zc_inject_record);
1324 
1325 	zc->zc_guid = id;
1326 
1327 	return (error);
1328 }
1329 
1330 static int
1331 zfs_ioc_error_log(zfs_cmd_t *zc)
1332 {
1333 	spa_t *spa;
1334 	int error;
1335 	size_t count = (size_t)zc->zc_nvlist_dst_size;
1336 
1337 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1338 		return (error);
1339 
1340 	error = spa_get_errlog(spa, (void *)(uintptr_t)zc->zc_nvlist_dst,
1341 	    &count);
1342 	if (error == 0)
1343 		zc->zc_nvlist_dst_size = count;
1344 	else
1345 		zc->zc_nvlist_dst_size = spa_get_errlog_size(spa);
1346 
1347 	spa_close(spa, FTAG);
1348 
1349 	return (error);
1350 }
1351 
1352 static int
1353 zfs_ioc_clear(zfs_cmd_t *zc)
1354 {
1355 	spa_t *spa;
1356 	vdev_t *vd;
1357 	int error;
1358 
1359 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1360 		return (error);
1361 
1362 	spa_config_enter(spa, RW_WRITER, FTAG);
1363 
1364 	if (zc->zc_guid == 0) {
1365 		vd = NULL;
1366 	} else if ((vd = spa_lookup_by_guid(spa, zc->zc_guid)) == NULL) {
1367 		spa_config_exit(spa, FTAG);
1368 		spa_close(spa, FTAG);
1369 		return (ENODEV);
1370 	}
1371 
1372 	vdev_clear(spa, vd);
1373 
1374 	spa_config_exit(spa, FTAG);
1375 
1376 	spa_close(spa, FTAG);
1377 
1378 	return (0);
1379 }
1380 
1381 static int
1382 zfs_ioc_bookmark_name(zfs_cmd_t *zc)
1383 {
1384 	spa_t *spa;
1385 	int error;
1386 	nvlist_t *nvl;
1387 
1388 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1389 		return (error);
1390 
1391 	VERIFY(nvlist_alloc(&nvl, NV_UNIQUE_NAME, KM_SLEEP) == 0);
1392 
1393 	error = spa_bookmark_name(spa, &zc->zc_bookmark, nvl);
1394 	if (error == 0)
1395 		error = put_nvlist(zc, nvl);
1396 	nvlist_free(nvl);
1397 
1398 	spa_close(spa, FTAG);
1399 
1400 	return (error);
1401 }
1402 
1403 static int
1404 zfs_ioc_promote(zfs_cmd_t *zc)
1405 {
1406 	char *cp;
1407 
1408 	/*
1409 	 * We don't need to unmount *all* the origin fs's snapshots, but
1410 	 * it's easier.
1411 	 */
1412 	cp = strchr(zc->zc_value, '@');
1413 	if (cp)
1414 		*cp = '\0';
1415 	(void) dmu_objset_find(zc->zc_value,
1416 	    zfs_unmount_snap, NULL, DS_FIND_SNAPSHOTS);
1417 	return (dsl_dataset_promote(zc->zc_name));
1418 }
1419 
1420 static zfs_ioc_vec_t zfs_ioc_vec[] = {
1421 	{ zfs_ioc_pool_create,		zfs_secpolicy_config,	pool_name },
1422 	{ zfs_ioc_pool_destroy,		zfs_secpolicy_config,	pool_name },
1423 	{ zfs_ioc_pool_import,		zfs_secpolicy_config,	pool_name },
1424 	{ zfs_ioc_pool_export,		zfs_secpolicy_config,	pool_name },
1425 	{ zfs_ioc_pool_configs,		zfs_secpolicy_none,	no_name },
1426 	{ zfs_ioc_pool_stats,		zfs_secpolicy_read,	pool_name },
1427 	{ zfs_ioc_pool_tryimport,	zfs_secpolicy_config,	no_name },
1428 	{ zfs_ioc_pool_scrub,		zfs_secpolicy_config,	pool_name },
1429 	{ zfs_ioc_pool_freeze,		zfs_secpolicy_config,	no_name },
1430 	{ zfs_ioc_pool_upgrade,		zfs_secpolicy_config,	pool_name },
1431 	{ zfs_ioc_pool_get_history,	zfs_secpolicy_config,	pool_name },
1432 	{ zfs_ioc_pool_log_history,	zfs_secpolicy_config,	pool_name },
1433 	{ zfs_ioc_vdev_add,		zfs_secpolicy_config,	pool_name },
1434 	{ zfs_ioc_vdev_remove,		zfs_secpolicy_config,	pool_name },
1435 	{ zfs_ioc_vdev_online,		zfs_secpolicy_config,	pool_name },
1436 	{ zfs_ioc_vdev_offline,		zfs_secpolicy_config,	pool_name },
1437 	{ zfs_ioc_vdev_attach,		zfs_secpolicy_config,	pool_name },
1438 	{ zfs_ioc_vdev_detach,		zfs_secpolicy_config,	pool_name },
1439 	{ zfs_ioc_vdev_setpath,		zfs_secpolicy_config,	pool_name },
1440 	{ zfs_ioc_objset_stats,		zfs_secpolicy_read,	dataset_name },
1441 	{ zfs_ioc_dataset_list_next,	zfs_secpolicy_read,	dataset_name },
1442 	{ zfs_ioc_snapshot_list_next,	zfs_secpolicy_read,	dataset_name },
1443 	{ zfs_ioc_set_prop,		zfs_secpolicy_write,	dataset_name },
1444 	{ zfs_ioc_create_minor,		zfs_secpolicy_config,	dataset_name },
1445 	{ zfs_ioc_remove_minor,		zfs_secpolicy_config,	dataset_name },
1446 	{ zfs_ioc_create,		zfs_secpolicy_parent,	dataset_name },
1447 	{ zfs_ioc_destroy,		zfs_secpolicy_parent,	dataset_name },
1448 	{ zfs_ioc_rollback,		zfs_secpolicy_write,	dataset_name },
1449 	{ zfs_ioc_rename,		zfs_secpolicy_write,	dataset_name },
1450 	{ zfs_ioc_recvbackup,		zfs_secpolicy_write,	dataset_name },
1451 	{ zfs_ioc_sendbackup,		zfs_secpolicy_write,	dataset_name },
1452 	{ zfs_ioc_inject_fault,		zfs_secpolicy_inject,	no_name },
1453 	{ zfs_ioc_clear_fault,		zfs_secpolicy_inject,	no_name },
1454 	{ zfs_ioc_inject_list_next,	zfs_secpolicy_inject,	no_name },
1455 	{ zfs_ioc_error_log,		zfs_secpolicy_inject,	pool_name },
1456 	{ zfs_ioc_clear,		zfs_secpolicy_config,	pool_name },
1457 	{ zfs_ioc_bookmark_name,	zfs_secpolicy_inject,	pool_name },
1458 	{ zfs_ioc_promote,		zfs_secpolicy_write,	dataset_name },
1459 	{ zfs_ioc_destroy_snaps,	zfs_secpolicy_write,	dataset_name },
1460 	{ zfs_ioc_snapshot,		zfs_secpolicy_write,	dataset_name }
1461 };
1462 
1463 static int
1464 zfsdev_ioctl(dev_t dev, int cmd, intptr_t arg, int flag, cred_t *cr, int *rvalp)
1465 {
1466 	zfs_cmd_t *zc;
1467 	uint_t vec;
1468 	int error, rc;
1469 
1470 	if (getminor(dev) != 0)
1471 		return (zvol_ioctl(dev, cmd, arg, flag, cr, rvalp));
1472 
1473 	vec = cmd - ZFS_IOC;
1474 
1475 	if (vec >= sizeof (zfs_ioc_vec) / sizeof (zfs_ioc_vec[0]))
1476 		return (EINVAL);
1477 
1478 	zc = kmem_zalloc(sizeof (zfs_cmd_t), KM_SLEEP);
1479 
1480 	error = xcopyin((void *)arg, zc, sizeof (zfs_cmd_t));
1481 
1482 	if (error == 0) {
1483 		zc->zc_cred = (uintptr_t)cr;
1484 		zc->zc_dev = dev;
1485 		error = zfs_ioc_vec[vec].zvec_secpolicy(zc->zc_name, cr);
1486 	}
1487 
1488 	/*
1489 	 * Ensure that all pool/dataset names are valid before we pass down to
1490 	 * the lower layers.
1491 	 */
1492 	if (error == 0) {
1493 		zc->zc_name[sizeof (zc->zc_name) - 1] = '\0';
1494 		switch (zfs_ioc_vec[vec].zvec_namecheck) {
1495 		case pool_name:
1496 			if (pool_namecheck(zc->zc_name, NULL, NULL) != 0)
1497 				error = EINVAL;
1498 			break;
1499 
1500 		case dataset_name:
1501 			if (dataset_namecheck(zc->zc_name, NULL, NULL) != 0)
1502 				error = EINVAL;
1503 			break;
1504 
1505 		case no_name:
1506 			break;
1507 		}
1508 	}
1509 
1510 	if (error == 0)
1511 		error = zfs_ioc_vec[vec].zvec_func(zc);
1512 
1513 	rc = xcopyout(zc, (void *)arg, sizeof (zfs_cmd_t));
1514 	if (error == 0)
1515 		error = rc;
1516 
1517 	kmem_free(zc, sizeof (zfs_cmd_t));
1518 	return (error);
1519 }
1520 
1521 static int
1522 zfs_attach(dev_info_t *dip, ddi_attach_cmd_t cmd)
1523 {
1524 	if (cmd != DDI_ATTACH)
1525 		return (DDI_FAILURE);
1526 
1527 	if (ddi_create_minor_node(dip, "zfs", S_IFCHR, 0,
1528 	    DDI_PSEUDO, 0) == DDI_FAILURE)
1529 		return (DDI_FAILURE);
1530 
1531 	zfs_dip = dip;
1532 
1533 	ddi_report_dev(dip);
1534 
1535 	return (DDI_SUCCESS);
1536 }
1537 
1538 static int
1539 zfs_detach(dev_info_t *dip, ddi_detach_cmd_t cmd)
1540 {
1541 	if (spa_busy() || zfs_busy() || zvol_busy())
1542 		return (DDI_FAILURE);
1543 
1544 	if (cmd != DDI_DETACH)
1545 		return (DDI_FAILURE);
1546 
1547 	zfs_dip = NULL;
1548 
1549 	ddi_prop_remove_all(dip);
1550 	ddi_remove_minor_node(dip, NULL);
1551 
1552 	return (DDI_SUCCESS);
1553 }
1554 
1555 /*ARGSUSED*/
1556 static int
1557 zfs_info(dev_info_t *dip, ddi_info_cmd_t infocmd, void *arg, void **result)
1558 {
1559 	switch (infocmd) {
1560 	case DDI_INFO_DEVT2DEVINFO:
1561 		*result = zfs_dip;
1562 		return (DDI_SUCCESS);
1563 
1564 	case DDI_INFO_DEVT2INSTANCE:
1565 		*result = (void *)0;
1566 		return (DDI_SUCCESS);
1567 	}
1568 
1569 	return (DDI_FAILURE);
1570 }
1571 
1572 /*
1573  * OK, so this is a little weird.
1574  *
1575  * /dev/zfs is the control node, i.e. minor 0.
1576  * /dev/zvol/[r]dsk/pool/dataset are the zvols, minor > 0.
1577  *
1578  * /dev/zfs has basically nothing to do except serve up ioctls,
1579  * so most of the standard driver entry points are in zvol.c.
1580  */
1581 static struct cb_ops zfs_cb_ops = {
1582 	zvol_open,	/* open */
1583 	zvol_close,	/* close */
1584 	zvol_strategy,	/* strategy */
1585 	nodev,		/* print */
1586 	nodev,		/* dump */
1587 	zvol_read,	/* read */
1588 	zvol_write,	/* write */
1589 	zfsdev_ioctl,	/* ioctl */
1590 	nodev,		/* devmap */
1591 	nodev,		/* mmap */
1592 	nodev,		/* segmap */
1593 	nochpoll,	/* poll */
1594 	ddi_prop_op,	/* prop_op */
1595 	NULL,		/* streamtab */
1596 	D_NEW | D_MP | D_64BIT,		/* Driver compatibility flag */
1597 	CB_REV,		/* version */
1598 	zvol_aread,	/* async read */
1599 	zvol_awrite,	/* async write */
1600 };
1601 
1602 static struct dev_ops zfs_dev_ops = {
1603 	DEVO_REV,	/* version */
1604 	0,		/* refcnt */
1605 	zfs_info,	/* info */
1606 	nulldev,	/* identify */
1607 	nulldev,	/* probe */
1608 	zfs_attach,	/* attach */
1609 	zfs_detach,	/* detach */
1610 	nodev,		/* reset */
1611 	&zfs_cb_ops,	/* driver operations */
1612 	NULL		/* no bus operations */
1613 };
1614 
1615 static struct modldrv zfs_modldrv = {
1616 	&mod_driverops, "ZFS storage pool version " ZFS_VERSION_STRING,
1617 	    &zfs_dev_ops
1618 };
1619 
1620 static struct modlinkage modlinkage = {
1621 	MODREV_1,
1622 	(void *)&zfs_modlfs,
1623 	(void *)&zfs_modldrv,
1624 	NULL
1625 };
1626 
1627 int
1628 _init(void)
1629 {
1630 	int error;
1631 
1632 	spa_init(FREAD | FWRITE);
1633 	zfs_init();
1634 	zvol_init();
1635 
1636 	if ((error = mod_install(&modlinkage)) != 0) {
1637 		zvol_fini();
1638 		zfs_fini();
1639 		spa_fini();
1640 		return (error);
1641 	}
1642 
1643 	error = ldi_ident_from_mod(&modlinkage, &zfs_li);
1644 	ASSERT(error == 0);
1645 
1646 	return (0);
1647 }
1648 
1649 int
1650 _fini(void)
1651 {
1652 	int error;
1653 
1654 	if (spa_busy() || zfs_busy() || zvol_busy() || zio_injection_enabled)
1655 		return (EBUSY);
1656 
1657 	if ((error = mod_remove(&modlinkage)) != 0)
1658 		return (error);
1659 
1660 	zvol_fini();
1661 	zfs_fini();
1662 	spa_fini();
1663 
1664 	ldi_ident_release(zfs_li);
1665 	zfs_li = NULL;
1666 
1667 	return (error);
1668 }
1669 
1670 int
1671 _info(struct modinfo *modinfop)
1672 {
1673 	return (mod_info(&modlinkage, modinfop));
1674 }
1675