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. All rights reserved.
25 * Portions Copyright 2011 Martin Matuska
26 * Copyright 2015, OmniTI Computer Consulting, Inc. All rights reserved.
27 * Copyright 2015 Nexenta Systems, Inc. All rights reserved.
28 * Copyright (c) 2014, 2016 Joyent, Inc. All rights reserved.
29 * Copyright (c) 2011, 2017 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 * Copyright (c) 2014 Integros [integros.com]
33 * Copyright 2016 Toomas Soome <tsoome@me.com>
34 * Copyright 2017 RackTop Systems.
35 * Copyright (c) 2017 Datto Inc.
36 */
37
38 /*
39 * ZFS ioctls.
40 *
41 * This file handles the ioctls to /dev/zfs, used for configuring ZFS storage
42 * pools and filesystems, e.g. with /sbin/zfs and /sbin/zpool.
43 *
44 * There are two ways that we handle ioctls: the legacy way where almost
45 * all of the logic is in the ioctl callback, and the new way where most
46 * of the marshalling is handled in the common entry point, zfsdev_ioctl().
47 *
48 * Non-legacy ioctls should be registered by calling
49 * zfs_ioctl_register() from zfs_ioctl_init(). The ioctl is invoked
50 * from userland by lzc_ioctl().
51 *
52 * The registration arguments are as follows:
53 *
54 * const char *name
55 * The name of the ioctl. This is used for history logging. If the
56 * ioctl returns successfully (the callback returns 0), and allow_log
57 * is true, then a history log entry will be recorded with the input &
58 * output nvlists. The log entry can be printed with "zpool history -i".
59 *
60 * zfs_ioc_t ioc
61 * The ioctl request number, which userland will pass to ioctl(2).
62 * The ioctl numbers can change from release to release, because
63 * the caller (libzfs) must be matched to the kernel.
64 *
65 * zfs_secpolicy_func_t *secpolicy
66 * This function will be called before the zfs_ioc_func_t, to
67 * determine if this operation is permitted. It should return EPERM
68 * on failure, and 0 on success. Checks include determining if the
69 * dataset is visible in this zone, and if the user has either all
70 * zfs privileges in the zone (SYS_MOUNT), or has been granted permission
71 * to do this operation on this dataset with "zfs allow".
72 *
73 * zfs_ioc_namecheck_t namecheck
74 * This specifies what to expect in the zfs_cmd_t:zc_name -- a pool
75 * name, a dataset name, or nothing. If the name is not well-formed,
76 * the ioctl will fail and the callback will not be called.
77 * Therefore, the callback can assume that the name is well-formed
78 * (e.g. is null-terminated, doesn't have more than one '@' character,
79 * doesn't have invalid characters).
80 *
81 * zfs_ioc_poolcheck_t pool_check
82 * This specifies requirements on the pool state. If the pool does
83 * not meet them (is suspended or is readonly), the ioctl will fail
84 * and the callback will not be called. If any checks are specified
85 * (i.e. it is not POOL_CHECK_NONE), namecheck must not be NO_NAME.
86 * Multiple checks can be or-ed together (e.g. POOL_CHECK_SUSPENDED |
87 * POOL_CHECK_READONLY).
88 *
89 * boolean_t smush_outnvlist
90 * If smush_outnvlist is true, then the output is presumed to be a
91 * list of errors, and it will be "smushed" down to fit into the
92 * caller's buffer, by removing some entries and replacing them with a
93 * single "N_MORE_ERRORS" entry indicating how many were removed. See
94 * nvlist_smush() for details. If smush_outnvlist is false, and the
95 * outnvlist does not fit into the userland-provided buffer, then the
96 * ioctl will fail with ENOMEM.
97 *
98 * zfs_ioc_func_t *func
99 * The callback function that will perform the operation.
100 *
101 * The callback should return 0 on success, or an error number on
102 * failure. If the function fails, the userland ioctl will return -1,
103 * and errno will be set to the callback's return value. The callback
104 * will be called with the following arguments:
105 *
106 * const char *name
107 * The name of the pool or dataset to operate on, from
108 * zfs_cmd_t:zc_name. The 'namecheck' argument specifies the
109 * expected type (pool, dataset, or none).
110 *
111 * nvlist_t *innvl
112 * The input nvlist, deserialized from zfs_cmd_t:zc_nvlist_src. Or
113 * NULL if no input nvlist was provided. Changes to this nvlist are
114 * ignored. If the input nvlist could not be deserialized, the
115 * ioctl will fail and the callback will not be called.
116 *
117 * nvlist_t *outnvl
118 * The output nvlist, initially empty. The callback can fill it in,
119 * and it will be returned to userland by serializing it into
120 * zfs_cmd_t:zc_nvlist_dst. If it is non-empty, and serialization
121 * fails (e.g. because the caller didn't supply a large enough
122 * buffer), then the overall ioctl will fail. See the
123 * 'smush_nvlist' argument above for additional behaviors.
124 *
125 * There are two typical uses of the output nvlist:
126 * - To return state, e.g. property values. In this case,
127 * smush_outnvlist should be false. If the buffer was not large
128 * enough, the caller will reallocate a larger buffer and try
129 * the ioctl again.
130 *
131 * - To return multiple errors from an ioctl which makes on-disk
132 * changes. In this case, smush_outnvlist should be true.
133 * Ioctls which make on-disk modifications should generally not
134 * use the outnvl if they succeed, because the caller can not
135 * distinguish between the operation failing, and
136 * deserialization failing.
137 */
138
139 #include <sys/types.h>
140 #include <sys/param.h>
141 #include <sys/errno.h>
142 #include <sys/uio.h>
143 #include <sys/buf.h>
144 #include <sys/modctl.h>
145 #include <sys/open.h>
146 #include <sys/file.h>
147 #include <sys/kmem.h>
148 #include <sys/conf.h>
149 #include <sys/cmn_err.h>
150 #include <sys/stat.h>
151 #include <sys/zfs_ioctl.h>
152 #include <sys/zfs_vfsops.h>
153 #include <sys/zfs_znode.h>
154 #include <sys/zap.h>
155 #include <sys/spa.h>
156 #include <sys/spa_impl.h>
157 #include <sys/vdev.h>
158 #include <sys/priv_impl.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/ddi.h>
168 #include <sys/sunddi.h>
169 #include <sys/sunldi.h>
170 #include <sys/policy.h>
171 #include <sys/zone.h>
172 #include <sys/nvpair.h>
173 #include <sys/pathname.h>
174 #include <sys/mount.h>
175 #include <sys/sdt.h>
176 #include <sys/fs/zfs.h>
177 #include <sys/zfs_ctldir.h>
178 #include <sys/zfs_dir.h>
179 #include <sys/zfs_onexit.h>
180 #include <sys/zvol.h>
181 #include <sys/dsl_scan.h>
182 #include <sharefs/share.h>
183 #include <sys/dmu_objset.h>
184 #include <sys/dmu_send.h>
185 #include <sys/dsl_destroy.h>
186 #include <sys/dsl_bookmark.h>
187 #include <sys/dsl_userhold.h>
188 #include <sys/zfeature.h>
189 #include <sys/zcp.h>
190 #include <sys/zio_checksum.h>
191 #include <sys/vdev_removal.h>
192 #include <sys/vdev_impl.h>
193 #include <sys/vdev_initialize.h>
194
195 #include "zfs_namecheck.h"
196 #include "zfs_prop.h"
197 #include "zfs_deleg.h"
198 #include "zfs_comutil.h"
199
200 #include "lua.h"
201 #include "lauxlib.h"
202
203 extern struct modlfs zfs_modlfs;
204
205 extern void zfs_init(void);
206 extern void zfs_fini(void);
207
208 ldi_ident_t zfs_li = NULL;
209 dev_info_t *zfs_dip;
210
211 uint_t zfs_fsyncer_key;
212 extern uint_t rrw_tsd_key;
213 static uint_t zfs_allow_log_key;
214
215 typedef int zfs_ioc_legacy_func_t(zfs_cmd_t *);
216 typedef int zfs_ioc_func_t(const char *, nvlist_t *, nvlist_t *);
217 typedef int zfs_secpolicy_func_t(zfs_cmd_t *, nvlist_t *, cred_t *);
218
219 typedef enum {
220 NO_NAME,
221 POOL_NAME,
222 DATASET_NAME
223 } zfs_ioc_namecheck_t;
224
225 typedef enum {
226 POOL_CHECK_NONE = 1 << 0,
227 POOL_CHECK_SUSPENDED = 1 << 1,
228 POOL_CHECK_READONLY = 1 << 2,
229 } zfs_ioc_poolcheck_t;
230
231 typedef struct zfs_ioc_vec {
232 zfs_ioc_legacy_func_t *zvec_legacy_func;
233 zfs_ioc_func_t *zvec_func;
234 zfs_secpolicy_func_t *zvec_secpolicy;
235 zfs_ioc_namecheck_t zvec_namecheck;
236 boolean_t zvec_allow_log;
237 zfs_ioc_poolcheck_t zvec_pool_check;
238 boolean_t zvec_smush_outnvlist;
239 const char *zvec_name;
240 } zfs_ioc_vec_t;
241
242 /* This array is indexed by zfs_userquota_prop_t */
243 static const char *userquota_perms[] = {
244 ZFS_DELEG_PERM_USERUSED,
245 ZFS_DELEG_PERM_USERQUOTA,
246 ZFS_DELEG_PERM_GROUPUSED,
247 ZFS_DELEG_PERM_GROUPQUOTA,
248 };
249
250 static int zfs_ioc_userspace_upgrade(zfs_cmd_t *zc);
251 static int zfs_check_settable(const char *name, nvpair_t *property,
252 cred_t *cr);
253 static int zfs_check_clearable(char *dataset, nvlist_t *props,
254 nvlist_t **errors);
255 static int zfs_fill_zplprops_root(uint64_t, nvlist_t *, nvlist_t *,
256 boolean_t *);
257 int zfs_set_prop_nvlist(const char *, zprop_source_t, nvlist_t *, nvlist_t *);
258 static int get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp);
259
260 static int zfs_prop_activate_feature(spa_t *spa, spa_feature_t feature);
261
262 /* _NOTE(PRINTFLIKE(4)) - this is printf-like, but lint is too whiney */
263 void
264 __dprintf(const char *file, const char *func, int line, const char *fmt, ...)
265 {
266 const char *newfile;
267 char buf[512];
268 va_list adx;
269
270 /*
271 * Get rid of annoying "../common/" prefix to filename.
272 */
273 newfile = strrchr(file, '/');
274 if (newfile != NULL) {
275 newfile = newfile + 1; /* Get rid of leading / */
276 } else {
277 newfile = file;
278 }
279
280 va_start(adx, fmt);
281 (void) vsnprintf(buf, sizeof (buf), fmt, adx);
282 va_end(adx);
283
284 /*
285 * To get this data, use the zfs-dprintf probe as so:
286 * dtrace -q -n 'zfs-dprintf \
287 * /stringof(arg0) == "dbuf.c"/ \
288 * {printf("%s: %s", stringof(arg1), stringof(arg3))}'
289 * arg0 = file name
290 * arg1 = function name
291 * arg2 = line number
292 * arg3 = message
293 */
294 DTRACE_PROBE4(zfs__dprintf,
295 char *, newfile, char *, func, int, line, char *, buf);
296 }
297
298 static void
299 history_str_free(char *buf)
300 {
301 kmem_free(buf, HIS_MAX_RECORD_LEN);
302 }
303
304 static char *
305 history_str_get(zfs_cmd_t *zc)
306 {
307 char *buf;
308
309 if (zc->zc_history == NULL)
310 return (NULL);
311
312 buf = kmem_alloc(HIS_MAX_RECORD_LEN, KM_SLEEP);
313 if (copyinstr((void *)(uintptr_t)zc->zc_history,
314 buf, HIS_MAX_RECORD_LEN, NULL) != 0) {
315 history_str_free(buf);
316 return (NULL);
317 }
318
319 buf[HIS_MAX_RECORD_LEN -1] = '\0';
320
321 return (buf);
322 }
323
324 /*
325 * Check to see if the named dataset is currently defined as bootable
326 */
327 static boolean_t
328 zfs_is_bootfs(const char *name)
329 {
330 objset_t *os;
331
332 if (dmu_objset_hold(name, FTAG, &os) == 0) {
333 boolean_t ret;
334 ret = (dmu_objset_id(os) == spa_bootfs(dmu_objset_spa(os)));
335 dmu_objset_rele(os, FTAG);
336 return (ret);
337 }
338 return (B_FALSE);
339 }
340
341 /*
342 * Return non-zero if the spa version is less than requested version.
343 */
344 static int
345 zfs_earlier_version(const char *name, int version)
346 {
347 spa_t *spa;
348
349 if (spa_open(name, &spa, FTAG) == 0) {
350 if (spa_version(spa) < version) {
351 spa_close(spa, FTAG);
352 return (1);
353 }
354 spa_close(spa, FTAG);
355 }
356 return (0);
357 }
358
359 /*
360 * Return TRUE if the ZPL version is less than requested version.
361 */
362 static boolean_t
363 zpl_earlier_version(const char *name, int version)
364 {
365 objset_t *os;
366 boolean_t rc = B_TRUE;
367
368 if (dmu_objset_hold(name, FTAG, &os) == 0) {
369 uint64_t zplversion;
370
371 if (dmu_objset_type(os) != DMU_OST_ZFS) {
372 dmu_objset_rele(os, FTAG);
373 return (B_TRUE);
374 }
375 /* XXX reading from non-owned objset */
376 if (zfs_get_zplprop(os, ZFS_PROP_VERSION, &zplversion) == 0)
377 rc = zplversion < version;
378 dmu_objset_rele(os, FTAG);
379 }
380 return (rc);
381 }
382
383 static void
384 zfs_log_history(zfs_cmd_t *zc)
385 {
386 spa_t *spa;
387 char *buf;
388
389 if ((buf = history_str_get(zc)) == NULL)
390 return;
391
392 if (spa_open(zc->zc_name, &spa, FTAG) == 0) {
393 if (spa_version(spa) >= SPA_VERSION_ZPOOL_HISTORY)
394 (void) spa_history_log(spa, buf);
395 spa_close(spa, FTAG);
396 }
397 history_str_free(buf);
398 }
399
400 /*
401 * Policy for top-level read operations (list pools). Requires no privileges,
402 * and can be used in the local zone, as there is no associated dataset.
403 */
404 /* ARGSUSED */
405 static int
406 zfs_secpolicy_none(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
407 {
408 return (0);
409 }
410
411 /*
412 * Policy for dataset read operations (list children, get statistics). Requires
413 * no privileges, but must be visible in the local zone.
414 */
415 /* ARGSUSED */
416 static int
417 zfs_secpolicy_read(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
418 {
419 if (INGLOBALZONE(curproc) ||
420 zone_dataset_visible(zc->zc_name, NULL))
421 return (0);
422
423 return (SET_ERROR(ENOENT));
424 }
425
426 static int
427 zfs_dozonecheck_impl(const char *dataset, uint64_t zoned, cred_t *cr)
428 {
429 int writable = 1;
430
431 /*
432 * The dataset must be visible by this zone -- check this first
433 * so they don't see EPERM on something they shouldn't know about.
434 */
435 if (!INGLOBALZONE(curproc) &&
436 !zone_dataset_visible(dataset, &writable))
437 return (SET_ERROR(ENOENT));
438
439 if (INGLOBALZONE(curproc)) {
440 /*
441 * If the fs is zoned, only root can access it from the
442 * global zone.
443 */
444 if (secpolicy_zfs(cr) && zoned)
445 return (SET_ERROR(EPERM));
446 } else {
447 /*
448 * If we are in a local zone, the 'zoned' property must be set.
449 */
450 if (!zoned)
451 return (SET_ERROR(EPERM));
452
453 /* must be writable by this zone */
454 if (!writable)
455 return (SET_ERROR(EPERM));
456 }
457 return (0);
458 }
459
460 static int
461 zfs_dozonecheck(const char *dataset, cred_t *cr)
462 {
463 uint64_t zoned;
464
465 if (dsl_prop_get_integer(dataset, "zoned", &zoned, NULL))
466 return (SET_ERROR(ENOENT));
467
468 return (zfs_dozonecheck_impl(dataset, zoned, cr));
469 }
470
471 static int
472 zfs_dozonecheck_ds(const char *dataset, dsl_dataset_t *ds, cred_t *cr)
473 {
474 uint64_t zoned;
475
476 if (dsl_prop_get_int_ds(ds, "zoned", &zoned))
477 return (SET_ERROR(ENOENT));
478
479 return (zfs_dozonecheck_impl(dataset, zoned, cr));
480 }
481
482 static int
483 zfs_secpolicy_write_perms_ds(const char *name, dsl_dataset_t *ds,
484 const char *perm, cred_t *cr)
485 {
486 int error;
487
488 error = zfs_dozonecheck_ds(name, ds, cr);
489 if (error == 0) {
490 error = secpolicy_zfs(cr);
491 if (error != 0)
492 error = dsl_deleg_access_impl(ds, perm, cr);
493 }
494 return (error);
495 }
496
497 static int
498 zfs_secpolicy_write_perms(const char *name, const char *perm, cred_t *cr)
499 {
500 int error;
501 dsl_dataset_t *ds;
502 dsl_pool_t *dp;
503
504 /*
505 * First do a quick check for root in the global zone, which
506 * is allowed to do all write_perms. This ensures that zfs_ioc_*
507 * will get to handle nonexistent datasets.
508 */
509 if (INGLOBALZONE(curproc) && secpolicy_zfs(cr) == 0)
510 return (0);
511
512 error = dsl_pool_hold(name, FTAG, &dp);
513 if (error != 0)
514 return (error);
515
516 error = dsl_dataset_hold(dp, name, FTAG, &ds);
517 if (error != 0) {
518 dsl_pool_rele(dp, FTAG);
519 return (error);
520 }
521
522 error = zfs_secpolicy_write_perms_ds(name, ds, perm, cr);
523
524 dsl_dataset_rele(ds, FTAG);
525 dsl_pool_rele(dp, FTAG);
526 return (error);
527 }
528
529 /*
530 * Policy for setting the security label property.
531 *
532 * Returns 0 for success, non-zero for access and other errors.
533 */
534 static int
535 zfs_set_slabel_policy(const char *name, char *strval, cred_t *cr)
536 {
537 char ds_hexsl[MAXNAMELEN];
538 bslabel_t ds_sl, new_sl;
539 boolean_t new_default = FALSE;
540 uint64_t zoned;
541 int needed_priv = -1;
542 int error;
543
544 /* First get the existing dataset label. */
545 error = dsl_prop_get(name, zfs_prop_to_name(ZFS_PROP_MLSLABEL),
546 1, sizeof (ds_hexsl), &ds_hexsl, NULL);
547 if (error != 0)
548 return (SET_ERROR(EPERM));
549
550 if (strcasecmp(strval, ZFS_MLSLABEL_DEFAULT) == 0)
551 new_default = TRUE;
552
553 /* The label must be translatable */
554 if (!new_default && (hexstr_to_label(strval, &new_sl) != 0))
555 return (SET_ERROR(EINVAL));
556
557 /*
558 * In a non-global zone, disallow attempts to set a label that
559 * doesn't match that of the zone; otherwise no other checks
560 * are needed.
561 */
562 if (!INGLOBALZONE(curproc)) {
563 if (new_default || !blequal(&new_sl, CR_SL(CRED())))
564 return (SET_ERROR(EPERM));
565 return (0);
566 }
567
568 /*
569 * For global-zone datasets (i.e., those whose zoned property is
570 * "off", verify that the specified new label is valid for the
571 * global zone.
572 */
573 if (dsl_prop_get_integer(name,
574 zfs_prop_to_name(ZFS_PROP_ZONED), &zoned, NULL))
575 return (SET_ERROR(EPERM));
576 if (!zoned) {
577 if (zfs_check_global_label(name, strval) != 0)
578 return (SET_ERROR(EPERM));
579 }
580
581 /*
582 * If the existing dataset label is nondefault, check if the
583 * dataset is mounted (label cannot be changed while mounted).
584 * Get the zfsvfs; if there isn't one, then the dataset isn't
585 * mounted (or isn't a dataset, doesn't exist, ...).
586 */
587 if (strcasecmp(ds_hexsl, ZFS_MLSLABEL_DEFAULT) != 0) {
588 objset_t *os;
589 static char *setsl_tag = "setsl_tag";
590
591 /*
592 * Try to own the dataset; abort if there is any error,
593 * (e.g., already mounted, in use, or other error).
594 */
595 error = dmu_objset_own(name, DMU_OST_ZFS, B_TRUE,
596 setsl_tag, &os);
597 if (error != 0)
598 return (SET_ERROR(EPERM));
599
600 dmu_objset_disown(os, setsl_tag);
601
602 if (new_default) {
603 needed_priv = PRIV_FILE_DOWNGRADE_SL;
604 goto out_check;
605 }
606
607 if (hexstr_to_label(strval, &new_sl) != 0)
608 return (SET_ERROR(EPERM));
609
610 if (blstrictdom(&ds_sl, &new_sl))
611 needed_priv = PRIV_FILE_DOWNGRADE_SL;
612 else if (blstrictdom(&new_sl, &ds_sl))
613 needed_priv = PRIV_FILE_UPGRADE_SL;
614 } else {
615 /* dataset currently has a default label */
616 if (!new_default)
617 needed_priv = PRIV_FILE_UPGRADE_SL;
618 }
619
620 out_check:
621 if (needed_priv != -1)
622 return (PRIV_POLICY(cr, needed_priv, B_FALSE, EPERM, NULL));
623 return (0);
624 }
625
626 static int
627 zfs_secpolicy_setprop(const char *dsname, zfs_prop_t prop, nvpair_t *propval,
628 cred_t *cr)
629 {
630 char *strval;
631
632 /*
633 * Check permissions for special properties.
634 */
635 switch (prop) {
636 case ZFS_PROP_ZONED:
637 /*
638 * Disallow setting of 'zoned' from within a local zone.
639 */
640 if (!INGLOBALZONE(curproc))
641 return (SET_ERROR(EPERM));
642 break;
643
644 case ZFS_PROP_QUOTA:
645 case ZFS_PROP_FILESYSTEM_LIMIT:
646 case ZFS_PROP_SNAPSHOT_LIMIT:
647 if (!INGLOBALZONE(curproc)) {
648 uint64_t zoned;
649 char setpoint[ZFS_MAX_DATASET_NAME_LEN];
650 /*
651 * Unprivileged users are allowed to modify the
652 * limit on things *under* (ie. contained by)
653 * the thing they own.
654 */
655 if (dsl_prop_get_integer(dsname, "zoned", &zoned,
656 setpoint))
657 return (SET_ERROR(EPERM));
658 if (!zoned || strlen(dsname) <= strlen(setpoint))
659 return (SET_ERROR(EPERM));
660 }
661 break;
662
663 case ZFS_PROP_MLSLABEL:
664 if (!is_system_labeled())
665 return (SET_ERROR(EPERM));
666
667 if (nvpair_value_string(propval, &strval) == 0) {
668 int err;
669
670 err = zfs_set_slabel_policy(dsname, strval, CRED());
671 if (err != 0)
672 return (err);
673 }
674 break;
675 }
676
677 return (zfs_secpolicy_write_perms(dsname, zfs_prop_to_name(prop), cr));
678 }
679
680 /* ARGSUSED */
681 static int
682 zfs_secpolicy_set_fsacl(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
683 {
684 int error;
685
686 error = zfs_dozonecheck(zc->zc_name, cr);
687 if (error != 0)
688 return (error);
689
690 /*
691 * permission to set permissions will be evaluated later in
692 * dsl_deleg_can_allow()
693 */
694 return (0);
695 }
696
697 /* ARGSUSED */
698 static int
699 zfs_secpolicy_rollback(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
700 {
701 return (zfs_secpolicy_write_perms(zc->zc_name,
702 ZFS_DELEG_PERM_ROLLBACK, cr));
703 }
704
705 /* ARGSUSED */
706 static int
707 zfs_secpolicy_send(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
708 {
709 dsl_pool_t *dp;
710 dsl_dataset_t *ds;
711 char *cp;
712 int error;
713
714 /*
715 * Generate the current snapshot name from the given objsetid, then
716 * use that name for the secpolicy/zone checks.
717 */
718 cp = strchr(zc->zc_name, '@');
719 if (cp == NULL)
720 return (SET_ERROR(EINVAL));
721 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
722 if (error != 0)
723 return (error);
724
725 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &ds);
726 if (error != 0) {
727 dsl_pool_rele(dp, FTAG);
728 return (error);
729 }
730
731 dsl_dataset_name(ds, zc->zc_name);
732
733 error = zfs_secpolicy_write_perms_ds(zc->zc_name, ds,
734 ZFS_DELEG_PERM_SEND, cr);
735 dsl_dataset_rele(ds, FTAG);
736 dsl_pool_rele(dp, FTAG);
737
738 return (error);
739 }
740
741 /* ARGSUSED */
742 static int
743 zfs_secpolicy_send_new(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
744 {
745 return (zfs_secpolicy_write_perms(zc->zc_name,
746 ZFS_DELEG_PERM_SEND, cr));
747 }
748
749 /* ARGSUSED */
750 static int
751 zfs_secpolicy_deleg_share(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
752 {
753 vnode_t *vp;
754 int error;
755
756 if ((error = lookupname(zc->zc_value, UIO_SYSSPACE,
757 NO_FOLLOW, NULL, &vp)) != 0)
758 return (error);
759
760 /* Now make sure mntpnt and dataset are ZFS */
761
762 if (vp->v_vfsp->vfs_fstype != zfsfstype ||
763 (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource),
764 zc->zc_name) != 0)) {
765 VN_RELE(vp);
766 return (SET_ERROR(EPERM));
767 }
768
769 VN_RELE(vp);
770 return (dsl_deleg_access(zc->zc_name,
771 ZFS_DELEG_PERM_SHARE, cr));
772 }
773
774 int
775 zfs_secpolicy_share(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
776 {
777 if (!INGLOBALZONE(curproc))
778 return (SET_ERROR(EPERM));
779
780 if (secpolicy_nfs(cr) == 0) {
781 return (0);
782 } else {
783 return (zfs_secpolicy_deleg_share(zc, innvl, cr));
784 }
785 }
786
787 int
788 zfs_secpolicy_smb_acl(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
789 {
790 if (!INGLOBALZONE(curproc))
791 return (SET_ERROR(EPERM));
792
793 if (secpolicy_smb(cr) == 0) {
794 return (0);
795 } else {
796 return (zfs_secpolicy_deleg_share(zc, innvl, cr));
797 }
798 }
799
800 static int
801 zfs_get_parent(const char *datasetname, char *parent, int parentsize)
802 {
803 char *cp;
804
805 /*
806 * Remove the @bla or /bla from the end of the name to get the parent.
807 */
808 (void) strncpy(parent, datasetname, parentsize);
809 cp = strrchr(parent, '@');
810 if (cp != NULL) {
811 cp[0] = '\0';
812 } else {
813 cp = strrchr(parent, '/');
814 if (cp == NULL)
815 return (SET_ERROR(ENOENT));
816 cp[0] = '\0';
817 }
818
819 return (0);
820 }
821
822 int
823 zfs_secpolicy_destroy_perms(const char *name, cred_t *cr)
824 {
825 int error;
826
827 if ((error = zfs_secpolicy_write_perms(name,
828 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
829 return (error);
830
831 return (zfs_secpolicy_write_perms(name, ZFS_DELEG_PERM_DESTROY, cr));
832 }
833
834 /* ARGSUSED */
835 static int
836 zfs_secpolicy_destroy(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
837 {
838 return (zfs_secpolicy_destroy_perms(zc->zc_name, cr));
839 }
840
841 /*
842 * Destroying snapshots with delegated permissions requires
843 * descendant mount and destroy permissions.
844 */
845 /* ARGSUSED */
846 static int
847 zfs_secpolicy_destroy_snaps(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
848 {
849 nvlist_t *snaps;
850 nvpair_t *pair, *nextpair;
851 int error = 0;
852
853 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
854 return (SET_ERROR(EINVAL));
855 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
856 pair = nextpair) {
857 nextpair = nvlist_next_nvpair(snaps, pair);
858 error = zfs_secpolicy_destroy_perms(nvpair_name(pair), cr);
859 if (error == ENOENT) {
860 /*
861 * Ignore any snapshots that don't exist (we consider
862 * them "already destroyed"). Remove the name from the
863 * nvl here in case the snapshot is created between
864 * now and when we try to destroy it (in which case
865 * we don't want to destroy it since we haven't
866 * checked for permission).
867 */
868 fnvlist_remove_nvpair(snaps, pair);
869 error = 0;
870 }
871 if (error != 0)
872 break;
873 }
874
875 return (error);
876 }
877
878 int
879 zfs_secpolicy_rename_perms(const char *from, const char *to, cred_t *cr)
880 {
881 char parentname[ZFS_MAX_DATASET_NAME_LEN];
882 int error;
883
884 if ((error = zfs_secpolicy_write_perms(from,
885 ZFS_DELEG_PERM_RENAME, cr)) != 0)
886 return (error);
887
888 if ((error = zfs_secpolicy_write_perms(from,
889 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
890 return (error);
891
892 if ((error = zfs_get_parent(to, parentname,
893 sizeof (parentname))) != 0)
894 return (error);
895
896 if ((error = zfs_secpolicy_write_perms(parentname,
897 ZFS_DELEG_PERM_CREATE, cr)) != 0)
898 return (error);
899
900 if ((error = zfs_secpolicy_write_perms(parentname,
901 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
902 return (error);
903
904 return (error);
905 }
906
907 /* ARGSUSED */
908 static int
909 zfs_secpolicy_rename(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
910 {
911 return (zfs_secpolicy_rename_perms(zc->zc_name, zc->zc_value, cr));
912 }
913
914 /* ARGSUSED */
915 static int
916 zfs_secpolicy_promote(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
917 {
918 dsl_pool_t *dp;
919 dsl_dataset_t *clone;
920 int error;
921
922 error = zfs_secpolicy_write_perms(zc->zc_name,
923 ZFS_DELEG_PERM_PROMOTE, cr);
924 if (error != 0)
925 return (error);
926
927 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
928 if (error != 0)
929 return (error);
930
931 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &clone);
932
933 if (error == 0) {
934 char parentname[ZFS_MAX_DATASET_NAME_LEN];
935 dsl_dataset_t *origin = NULL;
936 dsl_dir_t *dd;
937 dd = clone->ds_dir;
938
939 error = dsl_dataset_hold_obj(dd->dd_pool,
940 dsl_dir_phys(dd)->dd_origin_obj, FTAG, &origin);
941 if (error != 0) {
942 dsl_dataset_rele(clone, FTAG);
943 dsl_pool_rele(dp, FTAG);
944 return (error);
945 }
946
947 error = zfs_secpolicy_write_perms_ds(zc->zc_name, clone,
948 ZFS_DELEG_PERM_MOUNT, cr);
949
950 dsl_dataset_name(origin, parentname);
951 if (error == 0) {
952 error = zfs_secpolicy_write_perms_ds(parentname, origin,
953 ZFS_DELEG_PERM_PROMOTE, cr);
954 }
955 dsl_dataset_rele(clone, FTAG);
956 dsl_dataset_rele(origin, FTAG);
957 }
958 dsl_pool_rele(dp, FTAG);
959 return (error);
960 }
961
962 /* ARGSUSED */
963 static int
964 zfs_secpolicy_recv(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
965 {
966 int error;
967
968 if ((error = zfs_secpolicy_write_perms(zc->zc_name,
969 ZFS_DELEG_PERM_RECEIVE, cr)) != 0)
970 return (error);
971
972 if ((error = zfs_secpolicy_write_perms(zc->zc_name,
973 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
974 return (error);
975
976 return (zfs_secpolicy_write_perms(zc->zc_name,
977 ZFS_DELEG_PERM_CREATE, cr));
978 }
979
980 int
981 zfs_secpolicy_snapshot_perms(const char *name, cred_t *cr)
982 {
983 return (zfs_secpolicy_write_perms(name,
984 ZFS_DELEG_PERM_SNAPSHOT, cr));
985 }
986
987 /*
988 * Check for permission to create each snapshot in the nvlist.
989 */
990 /* ARGSUSED */
991 static int
992 zfs_secpolicy_snapshot(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
993 {
994 nvlist_t *snaps;
995 int error = 0;
996 nvpair_t *pair;
997
998 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
999 return (SET_ERROR(EINVAL));
1000 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
1001 pair = nvlist_next_nvpair(snaps, pair)) {
1002 char *name = nvpair_name(pair);
1003 char *atp = strchr(name, '@');
1004
1005 if (atp == NULL) {
1006 error = SET_ERROR(EINVAL);
1007 break;
1008 }
1009 *atp = '\0';
1010 error = zfs_secpolicy_snapshot_perms(name, cr);
1011 *atp = '@';
1012 if (error != 0)
1013 break;
1014 }
1015 return (error);
1016 }
1017
1018 /*
1019 * Check for permission to create each snapshot in the nvlist.
1020 */
1021 /* ARGSUSED */
1022 static int
1023 zfs_secpolicy_bookmark(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1024 {
1025 int error = 0;
1026
1027 for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL);
1028 pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) {
1029 char *name = nvpair_name(pair);
1030 char *hashp = strchr(name, '#');
1031
1032 if (hashp == NULL) {
1033 error = SET_ERROR(EINVAL);
1034 break;
1035 }
1036 *hashp = '\0';
1037 error = zfs_secpolicy_write_perms(name,
1038 ZFS_DELEG_PERM_BOOKMARK, cr);
1039 *hashp = '#';
1040 if (error != 0)
1041 break;
1042 }
1043 return (error);
1044 }
1045
1046 /* ARGSUSED */
1047 static int
1048 zfs_secpolicy_remap(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1049 {
1050 return (zfs_secpolicy_write_perms(zc->zc_name,
1051 ZFS_DELEG_PERM_REMAP, cr));
1052 }
1053
1054 /* ARGSUSED */
1055 static int
1056 zfs_secpolicy_destroy_bookmarks(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1057 {
1058 nvpair_t *pair, *nextpair;
1059 int error = 0;
1060
1061 for (pair = nvlist_next_nvpair(innvl, NULL); pair != NULL;
1062 pair = nextpair) {
1063 char *name = nvpair_name(pair);
1064 char *hashp = strchr(name, '#');
1065 nextpair = nvlist_next_nvpair(innvl, pair);
1066
1067 if (hashp == NULL) {
1068 error = SET_ERROR(EINVAL);
1069 break;
1070 }
1071
1072 *hashp = '\0';
1073 error = zfs_secpolicy_write_perms(name,
1074 ZFS_DELEG_PERM_DESTROY, cr);
1075 *hashp = '#';
1076 if (error == ENOENT) {
1077 /*
1078 * Ignore any filesystems that don't exist (we consider
1079 * their bookmarks "already destroyed"). Remove
1080 * the name from the nvl here in case the filesystem
1081 * is created between now and when we try to destroy
1082 * the bookmark (in which case we don't want to
1083 * destroy it since we haven't checked for permission).
1084 */
1085 fnvlist_remove_nvpair(innvl, pair);
1086 error = 0;
1087 }
1088 if (error != 0)
1089 break;
1090 }
1091
1092 return (error);
1093 }
1094
1095 /* ARGSUSED */
1096 static int
1097 zfs_secpolicy_log_history(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1098 {
1099 /*
1100 * Even root must have a proper TSD so that we know what pool
1101 * to log to.
1102 */
1103 if (tsd_get(zfs_allow_log_key) == NULL)
1104 return (SET_ERROR(EPERM));
1105 return (0);
1106 }
1107
1108 static int
1109 zfs_secpolicy_create_clone(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1110 {
1111 char parentname[ZFS_MAX_DATASET_NAME_LEN];
1112 int error;
1113 char *origin;
1114
1115 if ((error = zfs_get_parent(zc->zc_name, parentname,
1116 sizeof (parentname))) != 0)
1117 return (error);
1118
1119 if (nvlist_lookup_string(innvl, "origin", &origin) == 0 &&
1120 (error = zfs_secpolicy_write_perms(origin,
1121 ZFS_DELEG_PERM_CLONE, cr)) != 0)
1122 return (error);
1123
1124 if ((error = zfs_secpolicy_write_perms(parentname,
1125 ZFS_DELEG_PERM_CREATE, cr)) != 0)
1126 return (error);
1127
1128 return (zfs_secpolicy_write_perms(parentname,
1129 ZFS_DELEG_PERM_MOUNT, cr));
1130 }
1131
1132 /*
1133 * Policy for pool operations - create/destroy pools, add vdevs, etc. Requires
1134 * SYS_CONFIG privilege, which is not available in a local zone.
1135 */
1136 /* ARGSUSED */
1137 static int
1138 zfs_secpolicy_config(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1139 {
1140 if (secpolicy_sys_config(cr, B_FALSE) != 0)
1141 return (SET_ERROR(EPERM));
1142
1143 return (0);
1144 }
1145
1146 /*
1147 * Policy for object to name lookups.
1148 */
1149 /* ARGSUSED */
1150 static int
1151 zfs_secpolicy_diff(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1152 {
1153 int error;
1154
1155 if ((error = secpolicy_sys_config(cr, B_FALSE)) == 0)
1156 return (0);
1157
1158 error = zfs_secpolicy_write_perms(zc->zc_name, ZFS_DELEG_PERM_DIFF, cr);
1159 return (error);
1160 }
1161
1162 /*
1163 * Policy for fault injection. Requires all privileges.
1164 */
1165 /* ARGSUSED */
1166 static int
1167 zfs_secpolicy_inject(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1168 {
1169 return (secpolicy_zinject(cr));
1170 }
1171
1172 /* ARGSUSED */
1173 static int
1174 zfs_secpolicy_inherit_prop(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1175 {
1176 zfs_prop_t prop = zfs_name_to_prop(zc->zc_value);
1177
1178 if (prop == ZPROP_INVAL) {
1179 if (!zfs_prop_user(zc->zc_value))
1180 return (SET_ERROR(EINVAL));
1181 return (zfs_secpolicy_write_perms(zc->zc_name,
1182 ZFS_DELEG_PERM_USERPROP, cr));
1183 } else {
1184 return (zfs_secpolicy_setprop(zc->zc_name, prop,
1185 NULL, cr));
1186 }
1187 }
1188
1189 static int
1190 zfs_secpolicy_userspace_one(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1191 {
1192 int err = zfs_secpolicy_read(zc, innvl, cr);
1193 if (err)
1194 return (err);
1195
1196 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
1197 return (SET_ERROR(EINVAL));
1198
1199 if (zc->zc_value[0] == 0) {
1200 /*
1201 * They are asking about a posix uid/gid. If it's
1202 * themself, allow it.
1203 */
1204 if (zc->zc_objset_type == ZFS_PROP_USERUSED ||
1205 zc->zc_objset_type == ZFS_PROP_USERQUOTA) {
1206 if (zc->zc_guid == crgetuid(cr))
1207 return (0);
1208 } else {
1209 if (groupmember(zc->zc_guid, cr))
1210 return (0);
1211 }
1212 }
1213
1214 return (zfs_secpolicy_write_perms(zc->zc_name,
1215 userquota_perms[zc->zc_objset_type], cr));
1216 }
1217
1218 static int
1219 zfs_secpolicy_userspace_many(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1220 {
1221 int err = zfs_secpolicy_read(zc, innvl, cr);
1222 if (err)
1223 return (err);
1224
1225 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
1226 return (SET_ERROR(EINVAL));
1227
1228 return (zfs_secpolicy_write_perms(zc->zc_name,
1229 userquota_perms[zc->zc_objset_type], cr));
1230 }
1231
1232 /* ARGSUSED */
1233 static int
1234 zfs_secpolicy_userspace_upgrade(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1235 {
1236 return (zfs_secpolicy_setprop(zc->zc_name, ZFS_PROP_VERSION,
1237 NULL, cr));
1238 }
1239
1240 /* ARGSUSED */
1241 static int
1242 zfs_secpolicy_hold(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1243 {
1244 nvpair_t *pair;
1245 nvlist_t *holds;
1246 int error;
1247
1248 error = nvlist_lookup_nvlist(innvl, "holds", &holds);
1249 if (error != 0)
1250 return (SET_ERROR(EINVAL));
1251
1252 for (pair = nvlist_next_nvpair(holds, NULL); pair != NULL;
1253 pair = nvlist_next_nvpair(holds, pair)) {
1254 char fsname[ZFS_MAX_DATASET_NAME_LEN];
1255 error = dmu_fsname(nvpair_name(pair), fsname);
1256 if (error != 0)
1257 return (error);
1258 error = zfs_secpolicy_write_perms(fsname,
1259 ZFS_DELEG_PERM_HOLD, cr);
1260 if (error != 0)
1261 return (error);
1262 }
1263 return (0);
1264 }
1265
1266 /* ARGSUSED */
1267 static int
1268 zfs_secpolicy_release(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1269 {
1270 nvpair_t *pair;
1271 int error;
1272
1273 for (pair = nvlist_next_nvpair(innvl, NULL); pair != NULL;
1274 pair = nvlist_next_nvpair(innvl, pair)) {
1275 char fsname[ZFS_MAX_DATASET_NAME_LEN];
1276 error = dmu_fsname(nvpair_name(pair), fsname);
1277 if (error != 0)
1278 return (error);
1279 error = zfs_secpolicy_write_perms(fsname,
1280 ZFS_DELEG_PERM_RELEASE, cr);
1281 if (error != 0)
1282 return (error);
1283 }
1284 return (0);
1285 }
1286
1287 /*
1288 * Policy for allowing temporary snapshots to be taken or released
1289 */
1290 static int
1291 zfs_secpolicy_tmp_snapshot(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1292 {
1293 /*
1294 * A temporary snapshot is the same as a snapshot,
1295 * hold, destroy and release all rolled into one.
1296 * Delegated diff alone is sufficient that we allow this.
1297 */
1298 int error;
1299
1300 if ((error = zfs_secpolicy_write_perms(zc->zc_name,
1301 ZFS_DELEG_PERM_DIFF, cr)) == 0)
1302 return (0);
1303
1304 error = zfs_secpolicy_snapshot_perms(zc->zc_name, cr);
1305 if (error == 0)
1306 error = zfs_secpolicy_hold(zc, innvl, cr);
1307 if (error == 0)
1308 error = zfs_secpolicy_release(zc, innvl, cr);
1309 if (error == 0)
1310 error = zfs_secpolicy_destroy(zc, innvl, cr);
1311 return (error);
1312 }
1313
1314 /*
1315 * Returns the nvlist as specified by the user in the zfs_cmd_t.
1316 */
1317 static int
1318 get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp)
1319 {
1320 char *packed;
1321 int error;
1322 nvlist_t *list = NULL;
1323
1324 /*
1325 * Read in and unpack the user-supplied nvlist.
1326 */
1327 if (size == 0)
1328 return (SET_ERROR(EINVAL));
1329
1330 packed = kmem_alloc(size, KM_SLEEP);
1331
1332 if ((error = ddi_copyin((void *)(uintptr_t)nvl, packed, size,
1333 iflag)) != 0) {
1334 kmem_free(packed, size);
1335 return (SET_ERROR(EFAULT));
1336 }
1337
1338 if ((error = nvlist_unpack(packed, size, &list, 0)) != 0) {
1339 kmem_free(packed, size);
1340 return (error);
1341 }
1342
1343 kmem_free(packed, size);
1344
1345 *nvp = list;
1346 return (0);
1347 }
1348
1349 /*
1350 * Reduce the size of this nvlist until it can be serialized in 'max' bytes.
1351 * Entries will be removed from the end of the nvlist, and one int32 entry
1352 * named "N_MORE_ERRORS" will be added indicating how many entries were
1353 * removed.
1354 */
1355 static int
1356 nvlist_smush(nvlist_t *errors, size_t max)
1357 {
1358 size_t size;
1359
1360 size = fnvlist_size(errors);
1361
1362 if (size > max) {
1363 nvpair_t *more_errors;
1364 int n = 0;
1365
1366 if (max < 1024)
1367 return (SET_ERROR(ENOMEM));
1368
1369 fnvlist_add_int32(errors, ZPROP_N_MORE_ERRORS, 0);
1370 more_errors = nvlist_prev_nvpair(errors, NULL);
1371
1372 do {
1373 nvpair_t *pair = nvlist_prev_nvpair(errors,
1374 more_errors);
1375 fnvlist_remove_nvpair(errors, pair);
1376 n++;
1377 size = fnvlist_size(errors);
1378 } while (size > max);
1379
1380 fnvlist_remove_nvpair(errors, more_errors);
1381 fnvlist_add_int32(errors, ZPROP_N_MORE_ERRORS, n);
1382 ASSERT3U(fnvlist_size(errors), <=, max);
1383 }
1384
1385 return (0);
1386 }
1387
1388 static int
1389 put_nvlist(zfs_cmd_t *zc, nvlist_t *nvl)
1390 {
1391 char *packed = NULL;
1392 int error = 0;
1393 size_t size;
1394
1395 size = fnvlist_size(nvl);
1396
1397 if (size > zc->zc_nvlist_dst_size) {
1398 error = SET_ERROR(ENOMEM);
1399 } else {
1400 packed = fnvlist_pack(nvl, &size);
1401 if (ddi_copyout(packed, (void *)(uintptr_t)zc->zc_nvlist_dst,
1402 size, zc->zc_iflags) != 0)
1403 error = SET_ERROR(EFAULT);
1404 fnvlist_pack_free(packed, size);
1405 }
1406
1407 zc->zc_nvlist_dst_size = size;
1408 zc->zc_nvlist_dst_filled = B_TRUE;
1409 return (error);
1410 }
1411
1412 int
1413 getzfsvfs_impl(objset_t *os, zfsvfs_t **zfvp)
1414 {
1415 int error = 0;
1416 if (dmu_objset_type(os) != DMU_OST_ZFS) {
1417 return (SET_ERROR(EINVAL));
1418 }
1419
1420 mutex_enter(&os->os_user_ptr_lock);
1421 *zfvp = dmu_objset_get_user(os);
1422 if (*zfvp) {
1423 VFS_HOLD((*zfvp)->z_vfs);
1424 } else {
1425 error = SET_ERROR(ESRCH);
1426 }
1427 mutex_exit(&os->os_user_ptr_lock);
1428 return (error);
1429 }
1430
1431 int
1432 getzfsvfs(const char *dsname, zfsvfs_t **zfvp)
1433 {
1434 objset_t *os;
1435 int error;
1436
1437 error = dmu_objset_hold(dsname, FTAG, &os);
1438 if (error != 0)
1439 return (error);
1440
1441 error = getzfsvfs_impl(os, zfvp);
1442 dmu_objset_rele(os, FTAG);
1443 return (error);
1444 }
1445
1446 /*
1447 * Find a zfsvfs_t for a mounted filesystem, or create our own, in which
1448 * case its z_vfs will be NULL, and it will be opened as the owner.
1449 * If 'writer' is set, the z_teardown_lock will be held for RW_WRITER,
1450 * which prevents all vnode ops from running.
1451 */
1452 static int
1453 zfsvfs_hold(const char *name, void *tag, zfsvfs_t **zfvp, boolean_t writer)
1454 {
1455 int error = 0;
1456
1457 if (getzfsvfs(name, zfvp) != 0)
1458 error = zfsvfs_create(name, zfvp);
1459 if (error == 0) {
1460 rrm_enter(&(*zfvp)->z_teardown_lock, (writer) ? RW_WRITER :
1461 RW_READER, tag);
1462 if ((*zfvp)->z_unmounted) {
1463 /*
1464 * XXX we could probably try again, since the unmounting
1465 * thread should be just about to disassociate the
1466 * objset from the zfsvfs.
1467 */
1468 rrm_exit(&(*zfvp)->z_teardown_lock, tag);
1469 return (SET_ERROR(EBUSY));
1470 }
1471 }
1472 return (error);
1473 }
1474
1475 static void
1476 zfsvfs_rele(zfsvfs_t *zfsvfs, void *tag)
1477 {
1478 rrm_exit(&zfsvfs->z_teardown_lock, tag);
1479
1480 if (zfsvfs->z_vfs) {
1481 VFS_RELE(zfsvfs->z_vfs);
1482 } else {
1483 dmu_objset_disown(zfsvfs->z_os, zfsvfs);
1484 zfsvfs_free(zfsvfs);
1485 }
1486 }
1487
1488 static int
1489 zfs_ioc_pool_create(zfs_cmd_t *zc)
1490 {
1491 int error;
1492 nvlist_t *config, *props = NULL;
1493 nvlist_t *rootprops = NULL;
1494 nvlist_t *zplprops = NULL;
1495 char *spa_name = zc->zc_name;
1496
1497 if (error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1498 zc->zc_iflags, &config))
1499 return (error);
1500
1501 if (zc->zc_nvlist_src_size != 0 && (error =
1502 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1503 zc->zc_iflags, &props))) {
1504 nvlist_free(config);
1505 return (error);
1506 }
1507
1508 if (props) {
1509 nvlist_t *nvl = NULL;
1510 uint64_t version = SPA_VERSION;
1511 char *tname;
1512
1513 (void) nvlist_lookup_uint64(props,
1514 zpool_prop_to_name(ZPOOL_PROP_VERSION), &version);
1515 if (!SPA_VERSION_IS_SUPPORTED(version)) {
1516 error = SET_ERROR(EINVAL);
1517 goto pool_props_bad;
1518 }
1519 (void) nvlist_lookup_nvlist(props, ZPOOL_ROOTFS_PROPS, &nvl);
1520 if (nvl) {
1521 error = nvlist_dup(nvl, &rootprops, KM_SLEEP);
1522 if (error != 0) {
1523 nvlist_free(config);
1524 nvlist_free(props);
1525 return (error);
1526 }
1527 (void) nvlist_remove_all(props, ZPOOL_ROOTFS_PROPS);
1528 }
1529 VERIFY(nvlist_alloc(&zplprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
1530 error = zfs_fill_zplprops_root(version, rootprops,
1531 zplprops, NULL);
1532 if (error != 0)
1533 goto pool_props_bad;
1534
1535 if (nvlist_lookup_string(props,
1536 zpool_prop_to_name(ZPOOL_PROP_TNAME), &tname) == 0)
1537 spa_name = tname;
1538 }
1539
1540 error = spa_create(zc->zc_name, config, props, zplprops);
1541
1542 /*
1543 * Set the remaining root properties
1544 */
1545 if (!error && (error = zfs_set_prop_nvlist(spa_name,
1546 ZPROP_SRC_LOCAL, rootprops, NULL)) != 0)
1547 (void) spa_destroy(spa_name);
1548
1549 pool_props_bad:
1550 nvlist_free(rootprops);
1551 nvlist_free(zplprops);
1552 nvlist_free(config);
1553 nvlist_free(props);
1554
1555 return (error);
1556 }
1557
1558 static int
1559 zfs_ioc_pool_destroy(zfs_cmd_t *zc)
1560 {
1561 int error;
1562 zfs_log_history(zc);
1563 error = spa_destroy(zc->zc_name);
1564 if (error == 0)
1565 zvol_remove_minors(zc->zc_name);
1566 return (error);
1567 }
1568
1569 static int
1570 zfs_ioc_pool_import(zfs_cmd_t *zc)
1571 {
1572 nvlist_t *config, *props = NULL;
1573 uint64_t guid;
1574 int error;
1575
1576 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1577 zc->zc_iflags, &config)) != 0)
1578 return (error);
1579
1580 if (zc->zc_nvlist_src_size != 0 && (error =
1581 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1582 zc->zc_iflags, &props))) {
1583 nvlist_free(config);
1584 return (error);
1585 }
1586
1587 if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, &guid) != 0 ||
1588 guid != zc->zc_guid)
1589 error = SET_ERROR(EINVAL);
1590 else
1591 error = spa_import(zc->zc_name, config, props, zc->zc_cookie);
1592
1593 if (zc->zc_nvlist_dst != 0) {
1594 int err;
1595
1596 if ((err = put_nvlist(zc, config)) != 0)
1597 error = err;
1598 }
1599
1600 nvlist_free(config);
1601
1602 nvlist_free(props);
1603
1604 return (error);
1605 }
1606
1607 static int
1608 zfs_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
1621 static int
1622 zfs_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 */
1646 static int
1647 zfs_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 */
1677 static int
1678 zfs_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 * zc_flags scrub pause/resume flag (pool_scrub_cmd_t)
1705 */
1706 static int
1707 zfs_ioc_pool_scan(zfs_cmd_t *zc)
1708 {
1709 spa_t *spa;
1710 int error;
1711
1712 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1713 return (error);
1714
1715 if (zc->zc_flags >= POOL_SCRUB_FLAGS_END)
1716 return (SET_ERROR(EINVAL));
1717
1718 if (zc->zc_flags == POOL_SCRUB_PAUSE)
1719 error = spa_scrub_pause_resume(spa, POOL_SCRUB_PAUSE);
1720 else if (zc->zc_cookie == POOL_SCAN_NONE)
1721 error = spa_scan_stop(spa);
1722 else
1723 error = spa_scan(spa, zc->zc_cookie);
1724
1725 spa_close(spa, FTAG);
1726
1727 return (error);
1728 }
1729
1730 static int
1731 zfs_ioc_pool_freeze(zfs_cmd_t *zc)
1732 {
1733 spa_t *spa;
1734 int error;
1735
1736 error = spa_open(zc->zc_name, &spa, FTAG);
1737 if (error == 0) {
1738 spa_freeze(spa);
1739 spa_close(spa, FTAG);
1740 }
1741 return (error);
1742 }
1743
1744 static int
1745 zfs_ioc_pool_upgrade(zfs_cmd_t *zc)
1746 {
1747 spa_t *spa;
1748 int error;
1749
1750 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1751 return (error);
1752
1753 if (zc->zc_cookie < spa_version(spa) ||
1754 !SPA_VERSION_IS_SUPPORTED(zc->zc_cookie)) {
1755 spa_close(spa, FTAG);
1756 return (SET_ERROR(EINVAL));
1757 }
1758
1759 spa_upgrade(spa, zc->zc_cookie);
1760 spa_close(spa, FTAG);
1761
1762 return (error);
1763 }
1764
1765 static int
1766 zfs_ioc_pool_get_history(zfs_cmd_t *zc)
1767 {
1768 spa_t *spa;
1769 char *hist_buf;
1770 uint64_t size;
1771 int error;
1772
1773 if ((size = zc->zc_history_len) == 0)
1774 return (SET_ERROR(EINVAL));
1775
1776 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1777 return (error);
1778
1779 if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) {
1780 spa_close(spa, FTAG);
1781 return (SET_ERROR(ENOTSUP));
1782 }
1783
1784 hist_buf = kmem_alloc(size, KM_SLEEP);
1785 if ((error = spa_history_get(spa, &zc->zc_history_offset,
1786 &zc->zc_history_len, hist_buf)) == 0) {
1787 error = ddi_copyout(hist_buf,
1788 (void *)(uintptr_t)zc->zc_history,
1789 zc->zc_history_len, zc->zc_iflags);
1790 }
1791
1792 spa_close(spa, FTAG);
1793 kmem_free(hist_buf, size);
1794 return (error);
1795 }
1796
1797 static int
1798 zfs_ioc_pool_reguid(zfs_cmd_t *zc)
1799 {
1800 spa_t *spa;
1801 int error;
1802
1803 error = spa_open(zc->zc_name, &spa, FTAG);
1804 if (error == 0) {
1805 error = spa_change_guid(spa);
1806 spa_close(spa, FTAG);
1807 }
1808 return (error);
1809 }
1810
1811 static int
1812 zfs_ioc_dsobj_to_dsname(zfs_cmd_t *zc)
1813 {
1814 return (dsl_dsobj_to_dsname(zc->zc_name, zc->zc_obj, zc->zc_value));
1815 }
1816
1817 /*
1818 * inputs:
1819 * zc_name name of filesystem
1820 * zc_obj object to find
1821 *
1822 * outputs:
1823 * zc_value name of object
1824 */
1825 static int
1826 zfs_ioc_obj_to_path(zfs_cmd_t *zc)
1827 {
1828 objset_t *os;
1829 int error;
1830
1831 /* XXX reading from objset not owned */
1832 if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os)) != 0)
1833 return (error);
1834 if (dmu_objset_type(os) != DMU_OST_ZFS) {
1835 dmu_objset_rele(os, FTAG);
1836 return (SET_ERROR(EINVAL));
1837 }
1838 error = zfs_obj_to_path(os, zc->zc_obj, zc->zc_value,
1839 sizeof (zc->zc_value));
1840 dmu_objset_rele(os, FTAG);
1841
1842 return (error);
1843 }
1844
1845 /*
1846 * inputs:
1847 * zc_name name of filesystem
1848 * zc_obj object to find
1849 *
1850 * outputs:
1851 * zc_stat stats on object
1852 * zc_value path to object
1853 */
1854 static int
1855 zfs_ioc_obj_to_stats(zfs_cmd_t *zc)
1856 {
1857 objset_t *os;
1858 int error;
1859
1860 /* XXX reading from objset not owned */
1861 if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os)) != 0)
1862 return (error);
1863 if (dmu_objset_type(os) != DMU_OST_ZFS) {
1864 dmu_objset_rele(os, FTAG);
1865 return (SET_ERROR(EINVAL));
1866 }
1867 error = zfs_obj_to_stats(os, zc->zc_obj, &zc->zc_stat, zc->zc_value,
1868 sizeof (zc->zc_value));
1869 dmu_objset_rele(os, FTAG);
1870
1871 return (error);
1872 }
1873
1874 static int
1875 zfs_ioc_vdev_add(zfs_cmd_t *zc)
1876 {
1877 spa_t *spa;
1878 int error;
1879 nvlist_t *config, **l2cache, **spares;
1880 uint_t nl2cache = 0, nspares = 0;
1881
1882 error = spa_open(zc->zc_name, &spa, FTAG);
1883 if (error != 0)
1884 return (error);
1885
1886 error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1887 zc->zc_iflags, &config);
1888 (void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_L2CACHE,
1889 &l2cache, &nl2cache);
1890
1891 (void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_SPARES,
1892 &spares, &nspares);
1893
1894 /*
1895 * A root pool with concatenated devices is not supported.
1896 * Thus, can not add a device to a root pool.
1897 *
1898 * Intent log device can not be added to a rootpool because
1899 * during mountroot, zil is replayed, a seperated log device
1900 * can not be accessed during the mountroot time.
1901 *
1902 * l2cache and spare devices are ok to be added to a rootpool.
1903 */
1904 if (spa_bootfs(spa) != 0 && nl2cache == 0 && nspares == 0) {
1905 nvlist_free(config);
1906 spa_close(spa, FTAG);
1907 return (SET_ERROR(EDOM));
1908 }
1909
1910 if (error == 0) {
1911 error = spa_vdev_add(spa, config);
1912 nvlist_free(config);
1913 }
1914 spa_close(spa, FTAG);
1915 return (error);
1916 }
1917
1918 /*
1919 * inputs:
1920 * zc_name name of the pool
1921 * zc_guid guid of vdev to remove
1922 * zc_cookie cancel removal
1923 */
1924 static int
1925 zfs_ioc_vdev_remove(zfs_cmd_t *zc)
1926 {
1927 spa_t *spa;
1928 int error;
1929
1930 error = spa_open(zc->zc_name, &spa, FTAG);
1931 if (error != 0)
1932 return (error);
1933 if (zc->zc_cookie != 0) {
1934 error = spa_vdev_remove_cancel(spa);
1935 } else {
1936 error = spa_vdev_remove(spa, zc->zc_guid, B_FALSE);
1937 }
1938 spa_close(spa, FTAG);
1939 return (error);
1940 }
1941
1942 static int
1943 zfs_ioc_vdev_set_state(zfs_cmd_t *zc)
1944 {
1945 spa_t *spa;
1946 int error;
1947 vdev_state_t newstate = VDEV_STATE_UNKNOWN;
1948
1949 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1950 return (error);
1951 switch (zc->zc_cookie) {
1952 case VDEV_STATE_ONLINE:
1953 error = vdev_online(spa, zc->zc_guid, zc->zc_obj, &newstate);
1954 break;
1955
1956 case VDEV_STATE_OFFLINE:
1957 error = vdev_offline(spa, zc->zc_guid, zc->zc_obj);
1958 break;
1959
1960 case VDEV_STATE_FAULTED:
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_fault(spa, zc->zc_guid, zc->zc_obj);
1966 break;
1967
1968 case VDEV_STATE_DEGRADED:
1969 if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED &&
1970 zc->zc_obj != VDEV_AUX_EXTERNAL)
1971 zc->zc_obj = VDEV_AUX_ERR_EXCEEDED;
1972
1973 error = vdev_degrade(spa, zc->zc_guid, zc->zc_obj);
1974 break;
1975
1976 default:
1977 error = SET_ERROR(EINVAL);
1978 }
1979 zc->zc_cookie = newstate;
1980 spa_close(spa, FTAG);
1981 return (error);
1982 }
1983
1984 static int
1985 zfs_ioc_vdev_attach(zfs_cmd_t *zc)
1986 {
1987 spa_t *spa;
1988 int replacing = zc->zc_cookie;
1989 nvlist_t *config;
1990 int error;
1991
1992 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1993 return (error);
1994
1995 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1996 zc->zc_iflags, &config)) == 0) {
1997 error = spa_vdev_attach(spa, zc->zc_guid, config, replacing);
1998 nvlist_free(config);
1999 }
2000
2001 spa_close(spa, FTAG);
2002 return (error);
2003 }
2004
2005 static int
2006 zfs_ioc_vdev_detach(zfs_cmd_t *zc)
2007 {
2008 spa_t *spa;
2009 int error;
2010
2011 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
2012 return (error);
2013
2014 error = spa_vdev_detach(spa, zc->zc_guid, 0, B_FALSE);
2015
2016 spa_close(spa, FTAG);
2017 return (error);
2018 }
2019
2020 static int
2021 zfs_ioc_vdev_split(zfs_cmd_t *zc)
2022 {
2023 spa_t *spa;
2024 nvlist_t *config, *props = NULL;
2025 int error;
2026 boolean_t exp = !!(zc->zc_cookie & ZPOOL_EXPORT_AFTER_SPLIT);
2027
2028 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
2029 return (error);
2030
2031 if (error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
2032 zc->zc_iflags, &config)) {
2033 spa_close(spa, FTAG);
2034 return (error);
2035 }
2036
2037 if (zc->zc_nvlist_src_size != 0 && (error =
2038 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2039 zc->zc_iflags, &props))) {
2040 spa_close(spa, FTAG);
2041 nvlist_free(config);
2042 return (error);
2043 }
2044
2045 error = spa_vdev_split_mirror(spa, zc->zc_string, config, props, exp);
2046
2047 spa_close(spa, FTAG);
2048
2049 nvlist_free(config);
2050 nvlist_free(props);
2051
2052 return (error);
2053 }
2054
2055 static int
2056 zfs_ioc_vdev_setpath(zfs_cmd_t *zc)
2057 {
2058 spa_t *spa;
2059 char *path = zc->zc_value;
2060 uint64_t guid = zc->zc_guid;
2061 int error;
2062
2063 error = spa_open(zc->zc_name, &spa, FTAG);
2064 if (error != 0)
2065 return (error);
2066
2067 error = spa_vdev_setpath(spa, guid, path);
2068 spa_close(spa, FTAG);
2069 return (error);
2070 }
2071
2072 static int
2073 zfs_ioc_vdev_setfru(zfs_cmd_t *zc)
2074 {
2075 spa_t *spa;
2076 char *fru = zc->zc_value;
2077 uint64_t guid = zc->zc_guid;
2078 int error;
2079
2080 error = spa_open(zc->zc_name, &spa, FTAG);
2081 if (error != 0)
2082 return (error);
2083
2084 error = spa_vdev_setfru(spa, guid, fru);
2085 spa_close(spa, FTAG);
2086 return (error);
2087 }
2088
2089 static int
2090 zfs_ioc_objset_stats_impl(zfs_cmd_t *zc, objset_t *os)
2091 {
2092 int error = 0;
2093 nvlist_t *nv;
2094
2095 dmu_objset_fast_stat(os, &zc->zc_objset_stats);
2096
2097 if (zc->zc_nvlist_dst != 0 &&
2098 (error = dsl_prop_get_all(os, &nv)) == 0) {
2099 dmu_objset_stats(os, nv);
2100 /*
2101 * NB: zvol_get_stats() will read the objset contents,
2102 * which we aren't supposed to do with a
2103 * DS_MODE_USER hold, because it could be
2104 * inconsistent. So this is a bit of a workaround...
2105 * XXX reading with out owning
2106 */
2107 if (!zc->zc_objset_stats.dds_inconsistent &&
2108 dmu_objset_type(os) == DMU_OST_ZVOL) {
2109 error = zvol_get_stats(os, nv);
2110 if (error == EIO)
2111 return (error);
2112 VERIFY0(error);
2113 }
2114 error = put_nvlist(zc, nv);
2115 nvlist_free(nv);
2116 }
2117
2118 return (error);
2119 }
2120
2121 /*
2122 * inputs:
2123 * zc_name name of filesystem
2124 * zc_nvlist_dst_size size of buffer for property nvlist
2125 *
2126 * outputs:
2127 * zc_objset_stats stats
2128 * zc_nvlist_dst property nvlist
2129 * zc_nvlist_dst_size size of property nvlist
2130 */
2131 static int
2132 zfs_ioc_objset_stats(zfs_cmd_t *zc)
2133 {
2134 objset_t *os;
2135 int error;
2136
2137 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
2138 if (error == 0) {
2139 error = zfs_ioc_objset_stats_impl(zc, os);
2140 dmu_objset_rele(os, FTAG);
2141 }
2142
2143 return (error);
2144 }
2145
2146 /*
2147 * inputs:
2148 * zc_name name of filesystem
2149 * zc_nvlist_dst_size size of buffer for property nvlist
2150 *
2151 * outputs:
2152 * zc_nvlist_dst received property nvlist
2153 * zc_nvlist_dst_size size of received property nvlist
2154 *
2155 * Gets received properties (distinct from local properties on or after
2156 * SPA_VERSION_RECVD_PROPS) for callers who want to differentiate received from
2157 * local property values.
2158 */
2159 static int
2160 zfs_ioc_objset_recvd_props(zfs_cmd_t *zc)
2161 {
2162 int error = 0;
2163 nvlist_t *nv;
2164
2165 /*
2166 * Without this check, we would return local property values if the
2167 * caller has not already received properties on or after
2168 * SPA_VERSION_RECVD_PROPS.
2169 */
2170 if (!dsl_prop_get_hasrecvd(zc->zc_name))
2171 return (SET_ERROR(ENOTSUP));
2172
2173 if (zc->zc_nvlist_dst != 0 &&
2174 (error = dsl_prop_get_received(zc->zc_name, &nv)) == 0) {
2175 error = put_nvlist(zc, nv);
2176 nvlist_free(nv);
2177 }
2178
2179 return (error);
2180 }
2181
2182 static int
2183 nvl_add_zplprop(objset_t *os, nvlist_t *props, zfs_prop_t prop)
2184 {
2185 uint64_t value;
2186 int error;
2187
2188 /*
2189 * zfs_get_zplprop() will either find a value or give us
2190 * the default value (if there is one).
2191 */
2192 if ((error = zfs_get_zplprop(os, prop, &value)) != 0)
2193 return (error);
2194 VERIFY(nvlist_add_uint64(props, zfs_prop_to_name(prop), value) == 0);
2195 return (0);
2196 }
2197
2198 /*
2199 * inputs:
2200 * zc_name name of filesystem
2201 * zc_nvlist_dst_size size of buffer for zpl property nvlist
2202 *
2203 * outputs:
2204 * zc_nvlist_dst zpl property nvlist
2205 * zc_nvlist_dst_size size of zpl property nvlist
2206 */
2207 static int
2208 zfs_ioc_objset_zplprops(zfs_cmd_t *zc)
2209 {
2210 objset_t *os;
2211 int err;
2212
2213 /* XXX reading without owning */
2214 if (err = dmu_objset_hold(zc->zc_name, FTAG, &os))
2215 return (err);
2216
2217 dmu_objset_fast_stat(os, &zc->zc_objset_stats);
2218
2219 /*
2220 * NB: nvl_add_zplprop() will read the objset contents,
2221 * which we aren't supposed to do with a DS_MODE_USER
2222 * hold, because it could be inconsistent.
2223 */
2224 if (zc->zc_nvlist_dst != NULL &&
2225 !zc->zc_objset_stats.dds_inconsistent &&
2226 dmu_objset_type(os) == DMU_OST_ZFS) {
2227 nvlist_t *nv;
2228
2229 VERIFY(nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2230 if ((err = nvl_add_zplprop(os, nv, ZFS_PROP_VERSION)) == 0 &&
2231 (err = nvl_add_zplprop(os, nv, ZFS_PROP_NORMALIZE)) == 0 &&
2232 (err = nvl_add_zplprop(os, nv, ZFS_PROP_UTF8ONLY)) == 0 &&
2233 (err = nvl_add_zplprop(os, nv, ZFS_PROP_CASE)) == 0)
2234 err = put_nvlist(zc, nv);
2235 nvlist_free(nv);
2236 } else {
2237 err = SET_ERROR(ENOENT);
2238 }
2239 dmu_objset_rele(os, FTAG);
2240 return (err);
2241 }
2242
2243 static boolean_t
2244 dataset_name_hidden(const char *name)
2245 {
2246 /*
2247 * Skip over datasets that are not visible in this zone,
2248 * internal datasets (which have a $ in their name), and
2249 * temporary datasets (which have a % in their name).
2250 */
2251 if (strchr(name, '$') != NULL)
2252 return (B_TRUE);
2253 if (strchr(name, '%') != NULL)
2254 return (B_TRUE);
2255 if (!INGLOBALZONE(curproc) && !zone_dataset_visible(name, NULL))
2256 return (B_TRUE);
2257 return (B_FALSE);
2258 }
2259
2260 /*
2261 * inputs:
2262 * zc_name name of filesystem
2263 * zc_cookie zap cursor
2264 * zc_nvlist_dst_size size of buffer for property nvlist
2265 *
2266 * outputs:
2267 * zc_name name of next filesystem
2268 * zc_cookie zap cursor
2269 * zc_objset_stats stats
2270 * zc_nvlist_dst property nvlist
2271 * zc_nvlist_dst_size size of property nvlist
2272 */
2273 static int
2274 zfs_ioc_dataset_list_next(zfs_cmd_t *zc)
2275 {
2276 objset_t *os;
2277 int error;
2278 char *p;
2279 size_t orig_len = strlen(zc->zc_name);
2280
2281 top:
2282 if (error = dmu_objset_hold(zc->zc_name, FTAG, &os)) {
2283 if (error == ENOENT)
2284 error = SET_ERROR(ESRCH);
2285 return (error);
2286 }
2287
2288 p = strrchr(zc->zc_name, '/');
2289 if (p == NULL || p[1] != '\0')
2290 (void) strlcat(zc->zc_name, "/", sizeof (zc->zc_name));
2291 p = zc->zc_name + strlen(zc->zc_name);
2292
2293 do {
2294 error = dmu_dir_list_next(os,
2295 sizeof (zc->zc_name) - (p - zc->zc_name), p,
2296 NULL, &zc->zc_cookie);
2297 if (error == ENOENT)
2298 error = SET_ERROR(ESRCH);
2299 } while (error == 0 && dataset_name_hidden(zc->zc_name));
2300 dmu_objset_rele(os, FTAG);
2301
2302 /*
2303 * If it's an internal dataset (ie. with a '$' in its name),
2304 * don't try to get stats for it, otherwise we'll return ENOENT.
2305 */
2306 if (error == 0 && strchr(zc->zc_name, '$') == NULL) {
2307 error = zfs_ioc_objset_stats(zc); /* fill in the stats */
2308 if (error == ENOENT) {
2309 /* We lost a race with destroy, get the next one. */
2310 zc->zc_name[orig_len] = '\0';
2311 goto top;
2312 }
2313 }
2314 return (error);
2315 }
2316
2317 /*
2318 * inputs:
2319 * zc_name name of filesystem
2320 * zc_cookie zap cursor
2321 * zc_nvlist_dst_size size of buffer for property nvlist
2322 * zc_simple when set, only name is requested
2323 *
2324 * outputs:
2325 * zc_name name of next snapshot
2326 * zc_objset_stats stats
2327 * zc_nvlist_dst property nvlist
2328 * zc_nvlist_dst_size size of property nvlist
2329 */
2330 static int
2331 zfs_ioc_snapshot_list_next(zfs_cmd_t *zc)
2332 {
2333 objset_t *os;
2334 int error;
2335
2336 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
2337 if (error != 0) {
2338 return (error == ENOENT ? ESRCH : error);
2339 }
2340
2341 /*
2342 * A dataset name of maximum length cannot have any snapshots,
2343 * so exit immediately.
2344 */
2345 if (strlcat(zc->zc_name, "@", sizeof (zc->zc_name)) >=
2346 ZFS_MAX_DATASET_NAME_LEN) {
2347 dmu_objset_rele(os, FTAG);
2348 return (SET_ERROR(ESRCH));
2349 }
2350
2351 error = dmu_snapshot_list_next(os,
2352 sizeof (zc->zc_name) - strlen(zc->zc_name),
2353 zc->zc_name + strlen(zc->zc_name), &zc->zc_obj, &zc->zc_cookie,
2354 NULL);
2355
2356 if (error == 0 && !zc->zc_simple) {
2357 dsl_dataset_t *ds;
2358 dsl_pool_t *dp = os->os_dsl_dataset->ds_dir->dd_pool;
2359
2360 error = dsl_dataset_hold_obj(dp, zc->zc_obj, FTAG, &ds);
2361 if (error == 0) {
2362 objset_t *ossnap;
2363
2364 error = dmu_objset_from_ds(ds, &ossnap);
2365 if (error == 0)
2366 error = zfs_ioc_objset_stats_impl(zc, ossnap);
2367 dsl_dataset_rele(ds, FTAG);
2368 }
2369 } else if (error == ENOENT) {
2370 error = SET_ERROR(ESRCH);
2371 }
2372
2373 dmu_objset_rele(os, FTAG);
2374 /* if we failed, undo the @ that we tacked on to zc_name */
2375 if (error != 0)
2376 *strchr(zc->zc_name, '@') = '\0';
2377 return (error);
2378 }
2379
2380 static int
2381 zfs_prop_set_userquota(const char *dsname, nvpair_t *pair)
2382 {
2383 const char *propname = nvpair_name(pair);
2384 uint64_t *valary;
2385 unsigned int vallen;
2386 const char *domain;
2387 char *dash;
2388 zfs_userquota_prop_t type;
2389 uint64_t rid;
2390 uint64_t quota;
2391 zfsvfs_t *zfsvfs;
2392 int err;
2393
2394 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2395 nvlist_t *attrs;
2396 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
2397 if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2398 &pair) != 0)
2399 return (SET_ERROR(EINVAL));
2400 }
2401
2402 /*
2403 * A correctly constructed propname is encoded as
2404 * userquota@<rid>-<domain>.
2405 */
2406 if ((dash = strchr(propname, '-')) == NULL ||
2407 nvpair_value_uint64_array(pair, &valary, &vallen) != 0 ||
2408 vallen != 3)
2409 return (SET_ERROR(EINVAL));
2410
2411 domain = dash + 1;
2412 type = valary[0];
2413 rid = valary[1];
2414 quota = valary[2];
2415
2416 err = zfsvfs_hold(dsname, FTAG, &zfsvfs, B_FALSE);
2417 if (err == 0) {
2418 err = zfs_set_userquota(zfsvfs, type, domain, rid, quota);
2419 zfsvfs_rele(zfsvfs, FTAG);
2420 }
2421
2422 return (err);
2423 }
2424
2425 /*
2426 * If the named property is one that has a special function to set its value,
2427 * return 0 on success and a positive error code on failure; otherwise if it is
2428 * not one of the special properties handled by this function, return -1.
2429 *
2430 * XXX: It would be better for callers of the property interface if we handled
2431 * these special cases in dsl_prop.c (in the dsl layer).
2432 */
2433 static int
2434 zfs_prop_set_special(const char *dsname, zprop_source_t source,
2435 nvpair_t *pair)
2436 {
2437 const char *propname = nvpair_name(pair);
2438 zfs_prop_t prop = zfs_name_to_prop(propname);
2439 uint64_t intval;
2440 int err = -1;
2441
2442 if (prop == ZPROP_INVAL) {
2443 if (zfs_prop_userquota(propname))
2444 return (zfs_prop_set_userquota(dsname, pair));
2445 return (-1);
2446 }
2447
2448 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2449 nvlist_t *attrs;
2450 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
2451 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2452 &pair) == 0);
2453 }
2454
2455 if (zfs_prop_get_type(prop) == PROP_TYPE_STRING)
2456 return (-1);
2457
2458 VERIFY(0 == nvpair_value_uint64(pair, &intval));
2459
2460 switch (prop) {
2461 case ZFS_PROP_QUOTA:
2462 err = dsl_dir_set_quota(dsname, source, intval);
2463 break;
2464 case ZFS_PROP_REFQUOTA:
2465 err = dsl_dataset_set_refquota(dsname, source, intval);
2466 break;
2467 case ZFS_PROP_FILESYSTEM_LIMIT:
2468 case ZFS_PROP_SNAPSHOT_LIMIT:
2469 if (intval == UINT64_MAX) {
2470 /* clearing the limit, just do it */
2471 err = 0;
2472 } else {
2473 err = dsl_dir_activate_fs_ss_limit(dsname);
2474 }
2475 /*
2476 * Set err to -1 to force the zfs_set_prop_nvlist code down the
2477 * default path to set the value in the nvlist.
2478 */
2479 if (err == 0)
2480 err = -1;
2481 break;
2482 case ZFS_PROP_RESERVATION:
2483 err = dsl_dir_set_reservation(dsname, source, intval);
2484 break;
2485 case ZFS_PROP_REFRESERVATION:
2486 err = dsl_dataset_set_refreservation(dsname, source, intval);
2487 break;
2488 case ZFS_PROP_VOLSIZE:
2489 err = zvol_set_volsize(dsname, intval);
2490 break;
2491 case ZFS_PROP_VERSION:
2492 {
2493 zfsvfs_t *zfsvfs;
2494
2495 if ((err = zfsvfs_hold(dsname, FTAG, &zfsvfs, B_TRUE)) != 0)
2496 break;
2497
2498 err = zfs_set_version(zfsvfs, intval);
2499 zfsvfs_rele(zfsvfs, FTAG);
2500
2501 if (err == 0 && intval >= ZPL_VERSION_USERSPACE) {
2502 zfs_cmd_t *zc;
2503
2504 zc = kmem_zalloc(sizeof (zfs_cmd_t), KM_SLEEP);
2505 (void) strcpy(zc->zc_name, dsname);
2506 (void) zfs_ioc_userspace_upgrade(zc);
2507 kmem_free(zc, sizeof (zfs_cmd_t));
2508 }
2509 break;
2510 }
2511 default:
2512 err = -1;
2513 }
2514
2515 return (err);
2516 }
2517
2518 /*
2519 * This function is best effort. If it fails to set any of the given properties,
2520 * it continues to set as many as it can and returns the last error
2521 * encountered. If the caller provides a non-NULL errlist, it will be filled in
2522 * with the list of names of all the properties that failed along with the
2523 * corresponding error numbers.
2524 *
2525 * If every property is set successfully, zero is returned and errlist is not
2526 * modified.
2527 */
2528 int
2529 zfs_set_prop_nvlist(const char *dsname, zprop_source_t source, nvlist_t *nvl,
2530 nvlist_t *errlist)
2531 {
2532 nvpair_t *pair;
2533 nvpair_t *propval;
2534 int rv = 0;
2535 uint64_t intval;
2536 char *strval;
2537 nvlist_t *genericnvl = fnvlist_alloc();
2538 nvlist_t *retrynvl = fnvlist_alloc();
2539
2540 retry:
2541 pair = NULL;
2542 while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) {
2543 const char *propname = nvpair_name(pair);
2544 zfs_prop_t prop = zfs_name_to_prop(propname);
2545 int err = 0;
2546
2547 /* decode the property value */
2548 propval = pair;
2549 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2550 nvlist_t *attrs;
2551 attrs = fnvpair_value_nvlist(pair);
2552 if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2553 &propval) != 0)
2554 err = SET_ERROR(EINVAL);
2555 }
2556
2557 /* Validate value type */
2558 if (err == 0 && prop == ZPROP_INVAL) {
2559 if (zfs_prop_user(propname)) {
2560 if (nvpair_type(propval) != DATA_TYPE_STRING)
2561 err = SET_ERROR(EINVAL);
2562 } else if (zfs_prop_userquota(propname)) {
2563 if (nvpair_type(propval) !=
2564 DATA_TYPE_UINT64_ARRAY)
2565 err = SET_ERROR(EINVAL);
2566 } else {
2567 err = SET_ERROR(EINVAL);
2568 }
2569 } else if (err == 0) {
2570 if (nvpair_type(propval) == DATA_TYPE_STRING) {
2571 if (zfs_prop_get_type(prop) != PROP_TYPE_STRING)
2572 err = SET_ERROR(EINVAL);
2573 } else if (nvpair_type(propval) == DATA_TYPE_UINT64) {
2574 const char *unused;
2575
2576 intval = fnvpair_value_uint64(propval);
2577
2578 switch (zfs_prop_get_type(prop)) {
2579 case PROP_TYPE_NUMBER:
2580 break;
2581 case PROP_TYPE_STRING:
2582 err = SET_ERROR(EINVAL);
2583 break;
2584 case PROP_TYPE_INDEX:
2585 if (zfs_prop_index_to_string(prop,
2586 intval, &unused) != 0)
2587 err = SET_ERROR(EINVAL);
2588 break;
2589 default:
2590 cmn_err(CE_PANIC,
2591 "unknown property type");
2592 }
2593 } else {
2594 err = SET_ERROR(EINVAL);
2595 }
2596 }
2597
2598 /* Validate permissions */
2599 if (err == 0)
2600 err = zfs_check_settable(dsname, pair, CRED());
2601
2602 if (err == 0) {
2603 err = zfs_prop_set_special(dsname, source, pair);
2604 if (err == -1) {
2605 /*
2606 * For better performance we build up a list of
2607 * properties to set in a single transaction.
2608 */
2609 err = nvlist_add_nvpair(genericnvl, pair);
2610 } else if (err != 0 && nvl != retrynvl) {
2611 /*
2612 * This may be a spurious error caused by
2613 * receiving quota and reservation out of order.
2614 * Try again in a second pass.
2615 */
2616 err = nvlist_add_nvpair(retrynvl, pair);
2617 }
2618 }
2619
2620 if (err != 0) {
2621 if (errlist != NULL)
2622 fnvlist_add_int32(errlist, propname, err);
2623 rv = err;
2624 }
2625 }
2626
2627 if (nvl != retrynvl && !nvlist_empty(retrynvl)) {
2628 nvl = retrynvl;
2629 goto retry;
2630 }
2631
2632 if (!nvlist_empty(genericnvl) &&
2633 dsl_props_set(dsname, source, genericnvl) != 0) {
2634 /*
2635 * If this fails, we still want to set as many properties as we
2636 * can, so try setting them individually.
2637 */
2638 pair = NULL;
2639 while ((pair = nvlist_next_nvpair(genericnvl, pair)) != NULL) {
2640 const char *propname = nvpair_name(pair);
2641 int err = 0;
2642
2643 propval = pair;
2644 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2645 nvlist_t *attrs;
2646 attrs = fnvpair_value_nvlist(pair);
2647 propval = fnvlist_lookup_nvpair(attrs,
2648 ZPROP_VALUE);
2649 }
2650
2651 if (nvpair_type(propval) == DATA_TYPE_STRING) {
2652 strval = fnvpair_value_string(propval);
2653 err = dsl_prop_set_string(dsname, propname,
2654 source, strval);
2655 } else {
2656 intval = fnvpair_value_uint64(propval);
2657 err = dsl_prop_set_int(dsname, propname, source,
2658 intval);
2659 }
2660
2661 if (err != 0) {
2662 if (errlist != NULL) {
2663 fnvlist_add_int32(errlist, propname,
2664 err);
2665 }
2666 rv = err;
2667 }
2668 }
2669 }
2670 nvlist_free(genericnvl);
2671 nvlist_free(retrynvl);
2672
2673 return (rv);
2674 }
2675
2676 /*
2677 * Check that all the properties are valid user properties.
2678 */
2679 static int
2680 zfs_check_userprops(const char *fsname, nvlist_t *nvl)
2681 {
2682 nvpair_t *pair = NULL;
2683 int error = 0;
2684
2685 while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) {
2686 const char *propname = nvpair_name(pair);
2687
2688 if (!zfs_prop_user(propname) ||
2689 nvpair_type(pair) != DATA_TYPE_STRING)
2690 return (SET_ERROR(EINVAL));
2691
2692 if (error = zfs_secpolicy_write_perms(fsname,
2693 ZFS_DELEG_PERM_USERPROP, CRED()))
2694 return (error);
2695
2696 if (strlen(propname) >= ZAP_MAXNAMELEN)
2697 return (SET_ERROR(ENAMETOOLONG));
2698
2699 if (strlen(fnvpair_value_string(pair)) >= ZAP_MAXVALUELEN)
2700 return (E2BIG);
2701 }
2702 return (0);
2703 }
2704
2705 static void
2706 props_skip(nvlist_t *props, nvlist_t *skipped, nvlist_t **newprops)
2707 {
2708 nvpair_t *pair;
2709
2710 VERIFY(nvlist_alloc(newprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2711
2712 pair = NULL;
2713 while ((pair = nvlist_next_nvpair(props, pair)) != NULL) {
2714 if (nvlist_exists(skipped, nvpair_name(pair)))
2715 continue;
2716
2717 VERIFY(nvlist_add_nvpair(*newprops, pair) == 0);
2718 }
2719 }
2720
2721 static int
2722 clear_received_props(const char *dsname, nvlist_t *props,
2723 nvlist_t *skipped)
2724 {
2725 int err = 0;
2726 nvlist_t *cleared_props = NULL;
2727 props_skip(props, skipped, &cleared_props);
2728 if (!nvlist_empty(cleared_props)) {
2729 /*
2730 * Acts on local properties until the dataset has received
2731 * properties at least once on or after SPA_VERSION_RECVD_PROPS.
2732 */
2733 zprop_source_t flags = (ZPROP_SRC_NONE |
2734 (dsl_prop_get_hasrecvd(dsname) ? ZPROP_SRC_RECEIVED : 0));
2735 err = zfs_set_prop_nvlist(dsname, flags, cleared_props, NULL);
2736 }
2737 nvlist_free(cleared_props);
2738 return (err);
2739 }
2740
2741 /*
2742 * inputs:
2743 * zc_name name of filesystem
2744 * zc_value name of property to set
2745 * zc_nvlist_src{_size} nvlist of properties to apply
2746 * zc_cookie received properties flag
2747 *
2748 * outputs:
2749 * zc_nvlist_dst{_size} error for each unapplied received property
2750 */
2751 static int
2752 zfs_ioc_set_prop(zfs_cmd_t *zc)
2753 {
2754 nvlist_t *nvl;
2755 boolean_t received = zc->zc_cookie;
2756 zprop_source_t source = (received ? ZPROP_SRC_RECEIVED :
2757 ZPROP_SRC_LOCAL);
2758 nvlist_t *errors;
2759 int error;
2760
2761 if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2762 zc->zc_iflags, &nvl)) != 0)
2763 return (error);
2764
2765 if (received) {
2766 nvlist_t *origprops;
2767
2768 if (dsl_prop_get_received(zc->zc_name, &origprops) == 0) {
2769 (void) clear_received_props(zc->zc_name,
2770 origprops, nvl);
2771 nvlist_free(origprops);
2772 }
2773
2774 error = dsl_prop_set_hasrecvd(zc->zc_name);
2775 }
2776
2777 errors = fnvlist_alloc();
2778 if (error == 0)
2779 error = zfs_set_prop_nvlist(zc->zc_name, source, nvl, errors);
2780
2781 if (zc->zc_nvlist_dst != NULL && errors != NULL) {
2782 (void) put_nvlist(zc, errors);
2783 }
2784
2785 nvlist_free(errors);
2786 nvlist_free(nvl);
2787 return (error);
2788 }
2789
2790 /*
2791 * inputs:
2792 * zc_name name of filesystem
2793 * zc_value name of property to inherit
2794 * zc_cookie revert to received value if TRUE
2795 *
2796 * outputs: none
2797 */
2798 static int
2799 zfs_ioc_inherit_prop(zfs_cmd_t *zc)
2800 {
2801 const char *propname = zc->zc_value;
2802 zfs_prop_t prop = zfs_name_to_prop(propname);
2803 boolean_t received = zc->zc_cookie;
2804 zprop_source_t source = (received
2805 ? ZPROP_SRC_NONE /* revert to received value, if any */
2806 : ZPROP_SRC_INHERITED); /* explicitly inherit */
2807
2808 if (received) {
2809 nvlist_t *dummy;
2810 nvpair_t *pair;
2811 zprop_type_t type;
2812 int err;
2813
2814 /*
2815 * zfs_prop_set_special() expects properties in the form of an
2816 * nvpair with type info.
2817 */
2818 if (prop == ZPROP_INVAL) {
2819 if (!zfs_prop_user(propname))
2820 return (SET_ERROR(EINVAL));
2821
2822 type = PROP_TYPE_STRING;
2823 } else if (prop == ZFS_PROP_VOLSIZE ||
2824 prop == ZFS_PROP_VERSION) {
2825 return (SET_ERROR(EINVAL));
2826 } else {
2827 type = zfs_prop_get_type(prop);
2828 }
2829
2830 VERIFY(nvlist_alloc(&dummy, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2831
2832 switch (type) {
2833 case PROP_TYPE_STRING:
2834 VERIFY(0 == nvlist_add_string(dummy, propname, ""));
2835 break;
2836 case PROP_TYPE_NUMBER:
2837 case PROP_TYPE_INDEX:
2838 VERIFY(0 == nvlist_add_uint64(dummy, propname, 0));
2839 break;
2840 default:
2841 nvlist_free(dummy);
2842 return (SET_ERROR(EINVAL));
2843 }
2844
2845 pair = nvlist_next_nvpair(dummy, NULL);
2846 err = zfs_prop_set_special(zc->zc_name, source, pair);
2847 nvlist_free(dummy);
2848 if (err != -1)
2849 return (err); /* special property already handled */
2850 } else {
2851 /*
2852 * Only check this in the non-received case. We want to allow
2853 * 'inherit -S' to revert non-inheritable properties like quota
2854 * and reservation to the received or default values even though
2855 * they are not considered inheritable.
2856 */
2857 if (prop != ZPROP_INVAL && !zfs_prop_inheritable(prop))
2858 return (SET_ERROR(EINVAL));
2859 }
2860
2861 /* property name has been validated by zfs_secpolicy_inherit_prop() */
2862 return (dsl_prop_inherit(zc->zc_name, zc->zc_value, source));
2863 }
2864
2865 static int
2866 zfs_ioc_pool_set_props(zfs_cmd_t *zc)
2867 {
2868 nvlist_t *props;
2869 spa_t *spa;
2870 int error;
2871 nvpair_t *pair;
2872
2873 if (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2874 zc->zc_iflags, &props))
2875 return (error);
2876
2877 /*
2878 * If the only property is the configfile, then just do a spa_lookup()
2879 * to handle the faulted case.
2880 */
2881 pair = nvlist_next_nvpair(props, NULL);
2882 if (pair != NULL && strcmp(nvpair_name(pair),
2883 zpool_prop_to_name(ZPOOL_PROP_CACHEFILE)) == 0 &&
2884 nvlist_next_nvpair(props, pair) == NULL) {
2885 mutex_enter(&spa_namespace_lock);
2886 if ((spa = spa_lookup(zc->zc_name)) != NULL) {
2887 spa_configfile_set(spa, props, B_FALSE);
2888 spa_write_cachefile(spa, B_FALSE, B_TRUE);
2889 }
2890 mutex_exit(&spa_namespace_lock);
2891 if (spa != NULL) {
2892 nvlist_free(props);
2893 return (0);
2894 }
2895 }
2896
2897 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) {
2898 nvlist_free(props);
2899 return (error);
2900 }
2901
2902 error = spa_prop_set(spa, props);
2903
2904 nvlist_free(props);
2905 spa_close(spa, FTAG);
2906
2907 return (error);
2908 }
2909
2910 static int
2911 zfs_ioc_pool_get_props(zfs_cmd_t *zc)
2912 {
2913 spa_t *spa;
2914 int error;
2915 nvlist_t *nvp = NULL;
2916
2917 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) {
2918 /*
2919 * If the pool is faulted, there may be properties we can still
2920 * get (such as altroot and cachefile), so attempt to get them
2921 * anyway.
2922 */
2923 mutex_enter(&spa_namespace_lock);
2924 if ((spa = spa_lookup(zc->zc_name)) != NULL)
2925 error = spa_prop_get(spa, &nvp);
2926 mutex_exit(&spa_namespace_lock);
2927 } else {
2928 error = spa_prop_get(spa, &nvp);
2929 spa_close(spa, FTAG);
2930 }
2931
2932 if (error == 0 && zc->zc_nvlist_dst != NULL)
2933 error = put_nvlist(zc, nvp);
2934 else
2935 error = SET_ERROR(EFAULT);
2936
2937 nvlist_free(nvp);
2938 return (error);
2939 }
2940
2941 /*
2942 * inputs:
2943 * zc_name name of filesystem
2944 * zc_nvlist_src{_size} nvlist of delegated permissions
2945 * zc_perm_action allow/unallow flag
2946 *
2947 * outputs: none
2948 */
2949 static int
2950 zfs_ioc_set_fsacl(zfs_cmd_t *zc)
2951 {
2952 int error;
2953 nvlist_t *fsaclnv = NULL;
2954
2955 if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2956 zc->zc_iflags, &fsaclnv)) != 0)
2957 return (error);
2958
2959 /*
2960 * Verify nvlist is constructed correctly
2961 */
2962 if ((error = zfs_deleg_verify_nvlist(fsaclnv)) != 0) {
2963 nvlist_free(fsaclnv);
2964 return (SET_ERROR(EINVAL));
2965 }
2966
2967 /*
2968 * If we don't have PRIV_SYS_MOUNT, then validate
2969 * that user is allowed to hand out each permission in
2970 * the nvlist(s)
2971 */
2972
2973 error = secpolicy_zfs(CRED());
2974 if (error != 0) {
2975 if (zc->zc_perm_action == B_FALSE) {
2976 error = dsl_deleg_can_allow(zc->zc_name,
2977 fsaclnv, CRED());
2978 } else {
2979 error = dsl_deleg_can_unallow(zc->zc_name,
2980 fsaclnv, CRED());
2981 }
2982 }
2983
2984 if (error == 0)
2985 error = dsl_deleg_set(zc->zc_name, fsaclnv, zc->zc_perm_action);
2986
2987 nvlist_free(fsaclnv);
2988 return (error);
2989 }
2990
2991 /*
2992 * inputs:
2993 * zc_name name of filesystem
2994 *
2995 * outputs:
2996 * zc_nvlist_src{_size} nvlist of delegated permissions
2997 */
2998 static int
2999 zfs_ioc_get_fsacl(zfs_cmd_t *zc)
3000 {
3001 nvlist_t *nvp;
3002 int error;
3003
3004 if ((error = dsl_deleg_get(zc->zc_name, &nvp)) == 0) {
3005 error = put_nvlist(zc, nvp);
3006 nvlist_free(nvp);
3007 }
3008
3009 return (error);
3010 }
3011
3012 /* ARGSUSED */
3013 static void
3014 zfs_create_cb(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx)
3015 {
3016 zfs_creat_t *zct = arg;
3017
3018 zfs_create_fs(os, cr, zct->zct_zplprops, tx);
3019 }
3020
3021 #define ZFS_PROP_UNDEFINED ((uint64_t)-1)
3022
3023 /*
3024 * inputs:
3025 * os parent objset pointer (NULL if root fs)
3026 * fuids_ok fuids allowed in this version of the spa?
3027 * sa_ok SAs allowed in this version of the spa?
3028 * createprops list of properties requested by creator
3029 *
3030 * outputs:
3031 * zplprops values for the zplprops we attach to the master node object
3032 * is_ci true if requested file system will be purely case-insensitive
3033 *
3034 * Determine the settings for utf8only, normalization and
3035 * casesensitivity. Specific values may have been requested by the
3036 * creator and/or we can inherit values from the parent dataset. If
3037 * the file system is of too early a vintage, a creator can not
3038 * request settings for these properties, even if the requested
3039 * setting is the default value. We don't actually want to create dsl
3040 * properties for these, so remove them from the source nvlist after
3041 * processing.
3042 */
3043 static int
3044 zfs_fill_zplprops_impl(objset_t *os, uint64_t zplver,
3045 boolean_t fuids_ok, boolean_t sa_ok, nvlist_t *createprops,
3046 nvlist_t *zplprops, boolean_t *is_ci)
3047 {
3048 uint64_t sense = ZFS_PROP_UNDEFINED;
3049 uint64_t norm = ZFS_PROP_UNDEFINED;
3050 uint64_t u8 = ZFS_PROP_UNDEFINED;
3051
3052 ASSERT(zplprops != NULL);
3053
3054 if (os != NULL && os->os_phys->os_type != DMU_OST_ZFS)
3055 return (SET_ERROR(EINVAL));
3056
3057 /*
3058 * Pull out creator prop choices, if any.
3059 */
3060 if (createprops) {
3061 (void) nvlist_lookup_uint64(createprops,
3062 zfs_prop_to_name(ZFS_PROP_VERSION), &zplver);
3063 (void) nvlist_lookup_uint64(createprops,
3064 zfs_prop_to_name(ZFS_PROP_NORMALIZE), &norm);
3065 (void) nvlist_remove_all(createprops,
3066 zfs_prop_to_name(ZFS_PROP_NORMALIZE));
3067 (void) nvlist_lookup_uint64(createprops,
3068 zfs_prop_to_name(ZFS_PROP_UTF8ONLY), &u8);
3069 (void) nvlist_remove_all(createprops,
3070 zfs_prop_to_name(ZFS_PROP_UTF8ONLY));
3071 (void) nvlist_lookup_uint64(createprops,
3072 zfs_prop_to_name(ZFS_PROP_CASE), &sense);
3073 (void) nvlist_remove_all(createprops,
3074 zfs_prop_to_name(ZFS_PROP_CASE));
3075 }
3076
3077 /*
3078 * If the zpl version requested is whacky or the file system
3079 * or pool is version is too "young" to support normalization
3080 * and the creator tried to set a value for one of the props,
3081 * error out.
3082 */
3083 if ((zplver < ZPL_VERSION_INITIAL || zplver > ZPL_VERSION) ||
3084 (zplver >= ZPL_VERSION_FUID && !fuids_ok) ||
3085 (zplver >= ZPL_VERSION_SA && !sa_ok) ||
3086 (zplver < ZPL_VERSION_NORMALIZATION &&
3087 (norm != ZFS_PROP_UNDEFINED || u8 != ZFS_PROP_UNDEFINED ||
3088 sense != ZFS_PROP_UNDEFINED)))
3089 return (SET_ERROR(ENOTSUP));
3090
3091 /*
3092 * Put the version in the zplprops
3093 */
3094 VERIFY(nvlist_add_uint64(zplprops,
3095 zfs_prop_to_name(ZFS_PROP_VERSION), zplver) == 0);
3096
3097 if (norm == ZFS_PROP_UNDEFINED)
3098 VERIFY(zfs_get_zplprop(os, ZFS_PROP_NORMALIZE, &norm) == 0);
3099 VERIFY(nvlist_add_uint64(zplprops,
3100 zfs_prop_to_name(ZFS_PROP_NORMALIZE), norm) == 0);
3101
3102 /*
3103 * If we're normalizing, names must always be valid UTF-8 strings.
3104 */
3105 if (norm)
3106 u8 = 1;
3107 if (u8 == ZFS_PROP_UNDEFINED)
3108 VERIFY(zfs_get_zplprop(os, ZFS_PROP_UTF8ONLY, &u8) == 0);
3109 VERIFY(nvlist_add_uint64(zplprops,
3110 zfs_prop_to_name(ZFS_PROP_UTF8ONLY), u8) == 0);
3111
3112 if (sense == ZFS_PROP_UNDEFINED)
3113 VERIFY(zfs_get_zplprop(os, ZFS_PROP_CASE, &sense) == 0);
3114 VERIFY(nvlist_add_uint64(zplprops,
3115 zfs_prop_to_name(ZFS_PROP_CASE), sense) == 0);
3116
3117 if (is_ci)
3118 *is_ci = (sense == ZFS_CASE_INSENSITIVE);
3119
3120 return (0);
3121 }
3122
3123 static int
3124 zfs_fill_zplprops(const char *dataset, nvlist_t *createprops,
3125 nvlist_t *zplprops, boolean_t *is_ci)
3126 {
3127 boolean_t fuids_ok, sa_ok;
3128 uint64_t zplver = ZPL_VERSION;
3129 objset_t *os = NULL;
3130 char parentname[ZFS_MAX_DATASET_NAME_LEN];
3131 char *cp;
3132 spa_t *spa;
3133 uint64_t spa_vers;
3134 int error;
3135
3136 (void) strlcpy(parentname, dataset, sizeof (parentname));
3137 cp = strrchr(parentname, '/');
3138 ASSERT(cp != NULL);
3139 cp[0] = '\0';
3140
3141 if ((error = spa_open(dataset, &spa, FTAG)) != 0)
3142 return (error);
3143
3144 spa_vers = spa_version(spa);
3145 spa_close(spa, FTAG);
3146
3147 zplver = zfs_zpl_version_map(spa_vers);
3148 fuids_ok = (zplver >= ZPL_VERSION_FUID);
3149 sa_ok = (zplver >= ZPL_VERSION_SA);
3150
3151 /*
3152 * Open parent object set so we can inherit zplprop values.
3153 */
3154 if ((error = dmu_objset_hold(parentname, FTAG, &os)) != 0)
3155 return (error);
3156
3157 error = zfs_fill_zplprops_impl(os, zplver, fuids_ok, sa_ok, createprops,
3158 zplprops, is_ci);
3159 dmu_objset_rele(os, FTAG);
3160 return (error);
3161 }
3162
3163 static int
3164 zfs_fill_zplprops_root(uint64_t spa_vers, nvlist_t *createprops,
3165 nvlist_t *zplprops, boolean_t *is_ci)
3166 {
3167 boolean_t fuids_ok;
3168 boolean_t sa_ok;
3169 uint64_t zplver = ZPL_VERSION;
3170 int error;
3171
3172 zplver = zfs_zpl_version_map(spa_vers);
3173 fuids_ok = (zplver >= ZPL_VERSION_FUID);
3174 sa_ok = (zplver >= ZPL_VERSION_SA);
3175
3176 error = zfs_fill_zplprops_impl(NULL, zplver, fuids_ok, sa_ok,
3177 createprops, zplprops, is_ci);
3178 return (error);
3179 }
3180
3181 /*
3182 * innvl: {
3183 * "type" -> dmu_objset_type_t (int32)
3184 * (optional) "props" -> { prop -> value }
3185 * }
3186 *
3187 * outnvl: propname -> error code (int32)
3188 */
3189 static int
3190 zfs_ioc_create(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3191 {
3192 int error = 0;
3193 zfs_creat_t zct = { 0 };
3194 nvlist_t *nvprops = NULL;
3195 void (*cbfunc)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx);
3196 int32_t type32;
3197 dmu_objset_type_t type;
3198 boolean_t is_insensitive = B_FALSE;
3199
3200 if (nvlist_lookup_int32(innvl, "type", &type32) != 0)
3201 return (SET_ERROR(EINVAL));
3202 type = type32;
3203 (void) nvlist_lookup_nvlist(innvl, "props", &nvprops);
3204
3205 switch (type) {
3206 case DMU_OST_ZFS:
3207 cbfunc = zfs_create_cb;
3208 break;
3209
3210 case DMU_OST_ZVOL:
3211 cbfunc = zvol_create_cb;
3212 break;
3213
3214 default:
3215 cbfunc = NULL;
3216 break;
3217 }
3218 if (strchr(fsname, '@') ||
3219 strchr(fsname, '%'))
3220 return (SET_ERROR(EINVAL));
3221
3222 zct.zct_props = nvprops;
3223
3224 if (cbfunc == NULL)
3225 return (SET_ERROR(EINVAL));
3226
3227 if (type == DMU_OST_ZVOL) {
3228 uint64_t volsize, volblocksize;
3229
3230 if (nvprops == NULL)
3231 return (SET_ERROR(EINVAL));
3232 if (nvlist_lookup_uint64(nvprops,
3233 zfs_prop_to_name(ZFS_PROP_VOLSIZE), &volsize) != 0)
3234 return (SET_ERROR(EINVAL));
3235
3236 if ((error = nvlist_lookup_uint64(nvprops,
3237 zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE),
3238 &volblocksize)) != 0 && error != ENOENT)
3239 return (SET_ERROR(EINVAL));
3240
3241 if (error != 0)
3242 volblocksize = zfs_prop_default_numeric(
3243 ZFS_PROP_VOLBLOCKSIZE);
3244
3245 if ((error = zvol_check_volblocksize(
3246 volblocksize)) != 0 ||
3247 (error = zvol_check_volsize(volsize,
3248 volblocksize)) != 0)
3249 return (error);
3250 } else if (type == DMU_OST_ZFS) {
3251 int error;
3252
3253 /*
3254 * We have to have normalization and
3255 * case-folding flags correct when we do the
3256 * file system creation, so go figure them out
3257 * now.
3258 */
3259 VERIFY(nvlist_alloc(&zct.zct_zplprops,
3260 NV_UNIQUE_NAME, KM_SLEEP) == 0);
3261 error = zfs_fill_zplprops(fsname, nvprops,
3262 zct.zct_zplprops, &is_insensitive);
3263 if (error != 0) {
3264 nvlist_free(zct.zct_zplprops);
3265 return (error);
3266 }
3267 }
3268
3269 error = dmu_objset_create(fsname, type,
3270 is_insensitive ? DS_FLAG_CI_DATASET : 0, cbfunc, &zct);
3271 nvlist_free(zct.zct_zplprops);
3272
3273 /*
3274 * It would be nice to do this atomically.
3275 */
3276 if (error == 0) {
3277 error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL,
3278 nvprops, outnvl);
3279 if (error != 0)
3280 (void) dsl_destroy_head(fsname);
3281 }
3282 return (error);
3283 }
3284
3285 /*
3286 * innvl: {
3287 * "origin" -> name of origin snapshot
3288 * (optional) "props" -> { prop -> value }
3289 * }
3290 *
3291 * outnvl: propname -> error code (int32)
3292 */
3293 static int
3294 zfs_ioc_clone(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3295 {
3296 int error = 0;
3297 nvlist_t *nvprops = NULL;
3298 char *origin_name;
3299
3300 if (nvlist_lookup_string(innvl, "origin", &origin_name) != 0)
3301 return (SET_ERROR(EINVAL));
3302 (void) nvlist_lookup_nvlist(innvl, "props", &nvprops);
3303
3304 if (strchr(fsname, '@') ||
3305 strchr(fsname, '%'))
3306 return (SET_ERROR(EINVAL));
3307
3308 if (dataset_namecheck(origin_name, NULL, NULL) != 0)
3309 return (SET_ERROR(EINVAL));
3310 error = dmu_objset_clone(fsname, origin_name);
3311 if (error != 0)
3312 return (error);
3313
3314 /*
3315 * It would be nice to do this atomically.
3316 */
3317 if (error == 0) {
3318 error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL,
3319 nvprops, outnvl);
3320 if (error != 0)
3321 (void) dsl_destroy_head(fsname);
3322 }
3323 return (error);
3324 }
3325
3326 /* ARGSUSED */
3327 static int
3328 zfs_ioc_remap(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3329 {
3330 if (strchr(fsname, '@') ||
3331 strchr(fsname, '%'))
3332 return (SET_ERROR(EINVAL));
3333
3334 return (dmu_objset_remap_indirects(fsname));
3335 }
3336
3337 /*
3338 * innvl: {
3339 * "snaps" -> { snapshot1, snapshot2 }
3340 * (optional) "props" -> { prop -> value (string) }
3341 * }
3342 *
3343 * outnvl: snapshot -> error code (int32)
3344 */
3345 static int
3346 zfs_ioc_snapshot(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3347 {
3348 nvlist_t *snaps;
3349 nvlist_t *props = NULL;
3350 int error, poollen;
3351 nvpair_t *pair;
3352
3353 (void) nvlist_lookup_nvlist(innvl, "props", &props);
3354 if ((error = zfs_check_userprops(poolname, props)) != 0)
3355 return (error);
3356
3357 if (!nvlist_empty(props) &&
3358 zfs_earlier_version(poolname, SPA_VERSION_SNAP_PROPS))
3359 return (SET_ERROR(ENOTSUP));
3360
3361 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
3362 return (SET_ERROR(EINVAL));
3363 poollen = strlen(poolname);
3364 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
3365 pair = nvlist_next_nvpair(snaps, pair)) {
3366 const char *name = nvpair_name(pair);
3367 const char *cp = strchr(name, '@');
3368
3369 /*
3370 * The snap name must contain an @, and the part after it must
3371 * contain only valid characters.
3372 */
3373 if (cp == NULL ||
3374 zfs_component_namecheck(cp + 1, NULL, NULL) != 0)
3375 return (SET_ERROR(EINVAL));
3376
3377 /*
3378 * The snap must be in the specified pool.
3379 */
3380 if (strncmp(name, poolname, poollen) != 0 ||
3381 (name[poollen] != '/' && name[poollen] != '@'))
3382 return (SET_ERROR(EXDEV));
3383
3384 /* This must be the only snap of this fs. */
3385 for (nvpair_t *pair2 = nvlist_next_nvpair(snaps, pair);
3386 pair2 != NULL; pair2 = nvlist_next_nvpair(snaps, pair2)) {
3387 if (strncmp(name, nvpair_name(pair2), cp - name + 1)
3388 == 0) {
3389 return (SET_ERROR(EXDEV));
3390 }
3391 }
3392 }
3393
3394 error = dsl_dataset_snapshot(snaps, props, outnvl);
3395 return (error);
3396 }
3397
3398 /*
3399 * innvl: "message" -> string
3400 */
3401 /* ARGSUSED */
3402 static int
3403 zfs_ioc_log_history(const char *unused, nvlist_t *innvl, nvlist_t *outnvl)
3404 {
3405 char *message;
3406 spa_t *spa;
3407 int error;
3408 char *poolname;
3409
3410 /*
3411 * The poolname in the ioctl is not set, we get it from the TSD,
3412 * which was set at the end of the last successful ioctl that allows
3413 * logging. The secpolicy func already checked that it is set.
3414 * Only one log ioctl is allowed after each successful ioctl, so
3415 * we clear the TSD here.
3416 */
3417 poolname = tsd_get(zfs_allow_log_key);
3418 (void) tsd_set(zfs_allow_log_key, NULL);
3419 error = spa_open(poolname, &spa, FTAG);
3420 strfree(poolname);
3421 if (error != 0)
3422 return (error);
3423
3424 if (nvlist_lookup_string(innvl, "message", &message) != 0) {
3425 spa_close(spa, FTAG);
3426 return (SET_ERROR(EINVAL));
3427 }
3428
3429 if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) {
3430 spa_close(spa, FTAG);
3431 return (SET_ERROR(ENOTSUP));
3432 }
3433
3434 error = spa_history_log(spa, message);
3435 spa_close(spa, FTAG);
3436 return (error);
3437 }
3438
3439 /*
3440 * The dp_config_rwlock must not be held when calling this, because the
3441 * unmount may need to write out data.
3442 *
3443 * This function is best-effort. Callers must deal gracefully if it
3444 * remains mounted (or is remounted after this call).
3445 *
3446 * Returns 0 if the argument is not a snapshot, or it is not currently a
3447 * filesystem, or we were able to unmount it. Returns error code otherwise.
3448 */
3449 void
3450 zfs_unmount_snap(const char *snapname)
3451 {
3452 vfs_t *vfsp = NULL;
3453 zfsvfs_t *zfsvfs = NULL;
3454
3455 if (strchr(snapname, '@') == NULL)
3456 return;
3457
3458 int err = getzfsvfs(snapname, &zfsvfs);
3459 if (err != 0) {
3460 ASSERT3P(zfsvfs, ==, NULL);
3461 return;
3462 }
3463 vfsp = zfsvfs->z_vfs;
3464
3465 ASSERT(!dsl_pool_config_held(dmu_objset_pool(zfsvfs->z_os)));
3466
3467 err = vn_vfswlock(vfsp->vfs_vnodecovered);
3468 VFS_RELE(vfsp);
3469 if (err != 0)
3470 return;
3471
3472 /*
3473 * Always force the unmount for snapshots.
3474 */
3475 (void) dounmount(vfsp, MS_FORCE, kcred);
3476 }
3477
3478 /* ARGSUSED */
3479 static int
3480 zfs_unmount_snap_cb(const char *snapname, void *arg)
3481 {
3482 zfs_unmount_snap(snapname);
3483 return (0);
3484 }
3485
3486 /*
3487 * When a clone is destroyed, its origin may also need to be destroyed,
3488 * in which case it must be unmounted. This routine will do that unmount
3489 * if necessary.
3490 */
3491 void
3492 zfs_destroy_unmount_origin(const char *fsname)
3493 {
3494 int error;
3495 objset_t *os;
3496 dsl_dataset_t *ds;
3497
3498 error = dmu_objset_hold(fsname, FTAG, &os);
3499 if (error != 0)
3500 return;
3501 ds = dmu_objset_ds(os);
3502 if (dsl_dir_is_clone(ds->ds_dir) && DS_IS_DEFER_DESTROY(ds->ds_prev)) {
3503 char originname[ZFS_MAX_DATASET_NAME_LEN];
3504 dsl_dataset_name(ds->ds_prev, originname);
3505 dmu_objset_rele(os, FTAG);
3506 zfs_unmount_snap(originname);
3507 } else {
3508 dmu_objset_rele(os, FTAG);
3509 }
3510 }
3511
3512 /*
3513 * innvl: {
3514 * "snaps" -> { snapshot1, snapshot2 }
3515 * (optional boolean) "defer"
3516 * }
3517 *
3518 * outnvl: snapshot -> error code (int32)
3519 *
3520 */
3521 /* ARGSUSED */
3522 static int
3523 zfs_ioc_destroy_snaps(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3524 {
3525 nvlist_t *snaps;
3526 nvpair_t *pair;
3527 boolean_t defer;
3528
3529 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
3530 return (SET_ERROR(EINVAL));
3531 defer = nvlist_exists(innvl, "defer");
3532
3533 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
3534 pair = nvlist_next_nvpair(snaps, pair)) {
3535 zfs_unmount_snap(nvpair_name(pair));
3536 }
3537
3538 return (dsl_destroy_snapshots_nvl(snaps, defer, outnvl));
3539 }
3540
3541 /*
3542 * Create bookmarks. Bookmark names are of the form <fs>#<bmark>.
3543 * All bookmarks must be in the same pool.
3544 *
3545 * innvl: {
3546 * bookmark1 -> snapshot1, bookmark2 -> snapshot2
3547 * }
3548 *
3549 * outnvl: bookmark -> error code (int32)
3550 *
3551 */
3552 /* ARGSUSED */
3553 static int
3554 zfs_ioc_bookmark(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3555 {
3556 for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL);
3557 pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) {
3558 char *snap_name;
3559
3560 /*
3561 * Verify the snapshot argument.
3562 */
3563 if (nvpair_value_string(pair, &snap_name) != 0)
3564 return (SET_ERROR(EINVAL));
3565
3566
3567 /* Verify that the keys (bookmarks) are unique */
3568 for (nvpair_t *pair2 = nvlist_next_nvpair(innvl, pair);
3569 pair2 != NULL; pair2 = nvlist_next_nvpair(innvl, pair2)) {
3570 if (strcmp(nvpair_name(pair), nvpair_name(pair2)) == 0)
3571 return (SET_ERROR(EINVAL));
3572 }
3573 }
3574
3575 return (dsl_bookmark_create(innvl, outnvl));
3576 }
3577
3578 /*
3579 * innvl: {
3580 * property 1, property 2, ...
3581 * }
3582 *
3583 * outnvl: {
3584 * bookmark name 1 -> { property 1, property 2, ... },
3585 * bookmark name 2 -> { property 1, property 2, ... }
3586 * }
3587 *
3588 */
3589 static int
3590 zfs_ioc_get_bookmarks(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3591 {
3592 return (dsl_get_bookmarks(fsname, innvl, outnvl));
3593 }
3594
3595 /*
3596 * innvl: {
3597 * bookmark name 1, bookmark name 2
3598 * }
3599 *
3600 * outnvl: bookmark -> error code (int32)
3601 *
3602 */
3603 static int
3604 zfs_ioc_destroy_bookmarks(const char *poolname, nvlist_t *innvl,
3605 nvlist_t *outnvl)
3606 {
3607 int error, poollen;
3608
3609 poollen = strlen(poolname);
3610 for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL);
3611 pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) {
3612 const char *name = nvpair_name(pair);
3613 const char *cp = strchr(name, '#');
3614
3615 /*
3616 * The bookmark name must contain an #, and the part after it
3617 * must contain only valid characters.
3618 */
3619 if (cp == NULL ||
3620 zfs_component_namecheck(cp + 1, NULL, NULL) != 0)
3621 return (SET_ERROR(EINVAL));
3622
3623 /*
3624 * The bookmark must be in the specified pool.
3625 */
3626 if (strncmp(name, poolname, poollen) != 0 ||
3627 (name[poollen] != '/' && name[poollen] != '#'))
3628 return (SET_ERROR(EXDEV));
3629 }
3630
3631 error = dsl_bookmark_destroy(innvl, outnvl);
3632 return (error);
3633 }
3634
3635 static int
3636 zfs_ioc_channel_program(const char *poolname, nvlist_t *innvl,
3637 nvlist_t *outnvl)
3638 {
3639 char *program;
3640 uint64_t instrlimit, memlimit;
3641 boolean_t sync_flag;
3642 nvpair_t *nvarg = NULL;
3643
3644 if (0 != nvlist_lookup_string(innvl, ZCP_ARG_PROGRAM, &program)) {
3645 return (EINVAL);
3646 }
3647 if (0 != nvlist_lookup_boolean_value(innvl, ZCP_ARG_SYNC, &sync_flag)) {
3648 sync_flag = B_TRUE;
3649 }
3650 if (0 != nvlist_lookup_uint64(innvl, ZCP_ARG_INSTRLIMIT, &instrlimit)) {
3651 instrlimit = ZCP_DEFAULT_INSTRLIMIT;
3652 }
3653 if (0 != nvlist_lookup_uint64(innvl, ZCP_ARG_MEMLIMIT, &memlimit)) {
3654 memlimit = ZCP_DEFAULT_MEMLIMIT;
3655 }
3656 if (0 != nvlist_lookup_nvpair(innvl, ZCP_ARG_ARGLIST, &nvarg)) {
3657 return (EINVAL);
3658 }
3659
3660 if (instrlimit == 0 || instrlimit > zfs_lua_max_instrlimit)
3661 return (EINVAL);
3662 if (memlimit == 0 || memlimit > zfs_lua_max_memlimit)
3663 return (EINVAL);
3664
3665 return (zcp_eval(poolname, program, sync_flag, instrlimit, memlimit,
3666 nvarg, outnvl));
3667 }
3668
3669 /*
3670 * innvl: unused
3671 * outnvl: empty
3672 */
3673 /* ARGSUSED */
3674 static int
3675 zfs_ioc_pool_checkpoint(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3676 {
3677 return (spa_checkpoint(poolname));
3678 }
3679
3680 /*
3681 * innvl: unused
3682 * outnvl: empty
3683 */
3684 /* ARGSUSED */
3685 static int
3686 zfs_ioc_pool_discard_checkpoint(const char *poolname, nvlist_t *innvl,
3687 nvlist_t *outnvl)
3688 {
3689 return (spa_checkpoint_discard(poolname));
3690 }
3691
3692 /*
3693 * inputs:
3694 * zc_name name of dataset to destroy
3695 * zc_defer_destroy mark for deferred destroy
3696 *
3697 * outputs: none
3698 */
3699 static int
3700 zfs_ioc_destroy(zfs_cmd_t *zc)
3701 {
3702 objset_t *os;
3703 dmu_objset_type_t ost;
3704 int err;
3705
3706 err = dmu_objset_hold(zc->zc_name, FTAG, &os);
3707 if (err != 0)
3708 return (err);
3709 ost = dmu_objset_type(os);
3710 dmu_objset_rele(os, FTAG);
3711
3712 if (ost == DMU_OST_ZFS)
3713 zfs_unmount_snap(zc->zc_name);
3714
3715 if (strchr(zc->zc_name, '@'))
3716 err = dsl_destroy_snapshot(zc->zc_name, zc->zc_defer_destroy);
3717 else
3718 err = dsl_destroy_head(zc->zc_name);
3719 if (ost == DMU_OST_ZVOL && err == 0)
3720 (void) zvol_remove_minor(zc->zc_name);
3721 return (err);
3722 }
3723
3724 /*
3725 * innvl: {
3726 * vdevs: {
3727 * guid 1, guid 2, ...
3728 * },
3729 * func: POOL_INITIALIZE_{CANCEL|DO|SUSPEND}
3730 * }
3731 *
3732 * outnvl: {
3733 * [func: EINVAL (if provided command type didn't make sense)],
3734 * [vdevs: {
3735 * guid1: errno, (see function body for possible errnos)
3736 * ...
3737 * }]
3738 * }
3739 *
3740 */
3741 static int
3742 zfs_ioc_pool_initialize(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3743 {
3744 spa_t *spa;
3745 int error;
3746
3747 error = spa_open(poolname, &spa, FTAG);
3748 if (error != 0)
3749 return (error);
3750
3751 uint64_t cmd_type;
3752 if (nvlist_lookup_uint64(innvl, ZPOOL_INITIALIZE_COMMAND,
3753 &cmd_type) != 0) {
3754 spa_close(spa, FTAG);
3755 return (SET_ERROR(EINVAL));
3756 }
3757 if (!(cmd_type == POOL_INITIALIZE_CANCEL ||
3758 cmd_type == POOL_INITIALIZE_DO ||
3759 cmd_type == POOL_INITIALIZE_SUSPEND)) {
3760 spa_close(spa, FTAG);
3761 return (SET_ERROR(EINVAL));
3762 }
3763
3764 nvlist_t *vdev_guids;
3765 if (nvlist_lookup_nvlist(innvl, ZPOOL_INITIALIZE_VDEVS,
3766 &vdev_guids) != 0) {
3767 spa_close(spa, FTAG);
3768 return (SET_ERROR(EINVAL));
3769 }
3770
3771 nvlist_t *vdev_errlist = fnvlist_alloc();
3772 int total_errors = 0;
3773
3774 for (nvpair_t *pair = nvlist_next_nvpair(vdev_guids, NULL);
3775 pair != NULL; pair = nvlist_next_nvpair(vdev_guids, pair)) {
3776 uint64_t vdev_guid = fnvpair_value_uint64(pair);
3777
3778 error = spa_vdev_initialize(spa, vdev_guid, cmd_type);
3779 if (error != 0) {
3780 char guid_as_str[MAXNAMELEN];
3781
3782 (void) snprintf(guid_as_str, sizeof (guid_as_str),
3783 "%llu", (unsigned long long)vdev_guid);
3784 fnvlist_add_int64(vdev_errlist, guid_as_str, error);
3785 total_errors++;
3786 }
3787 }
3788 if (fnvlist_size(vdev_errlist) > 0) {
3789 fnvlist_add_nvlist(outnvl, ZPOOL_INITIALIZE_VDEVS,
3790 vdev_errlist);
3791 }
3792 fnvlist_free(vdev_errlist);
3793
3794 spa_close(spa, FTAG);
3795 return (total_errors > 0 ? EINVAL : 0);
3796 }
3797
3798 /*
3799 * fsname is name of dataset to rollback (to most recent snapshot)
3800 *
3801 * innvl may contain name of expected target snapshot
3802 *
3803 * outnvl: "target" -> name of most recent snapshot
3804 * }
3805 */
3806 /* ARGSUSED */
3807 static int
3808 zfs_ioc_rollback(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3809 {
3810 zfsvfs_t *zfsvfs;
3811 char *target = NULL;
3812 int error;
3813
3814 (void) nvlist_lookup_string(innvl, "target", &target);
3815 if (target != NULL) {
3816 const char *cp = strchr(target, '@');
3817
3818 /*
3819 * The snap name must contain an @, and the part after it must
3820 * contain only valid characters.
3821 */
3822 if (cp == NULL ||
3823 zfs_component_namecheck(cp + 1, NULL, NULL) != 0)
3824 return (SET_ERROR(EINVAL));
3825 }
3826
3827 if (getzfsvfs(fsname, &zfsvfs) == 0) {
3828 dsl_dataset_t *ds;
3829
3830 ds = dmu_objset_ds(zfsvfs->z_os);
3831 error = zfs_suspend_fs(zfsvfs);
3832 if (error == 0) {
3833 int resume_err;
3834
3835 error = dsl_dataset_rollback(fsname, target, zfsvfs,
3836 outnvl);
3837 resume_err = zfs_resume_fs(zfsvfs, ds);
3838 error = error ? error : resume_err;
3839 }
3840 VFS_RELE(zfsvfs->z_vfs);
3841 } else {
3842 error = dsl_dataset_rollback(fsname, target, NULL, outnvl);
3843 }
3844 return (error);
3845 }
3846
3847 static int
3848 recursive_unmount(const char *fsname, void *arg)
3849 {
3850 const char *snapname = arg;
3851 char fullname[ZFS_MAX_DATASET_NAME_LEN];
3852
3853 (void) snprintf(fullname, sizeof (fullname), "%s@%s", fsname, snapname);
3854 zfs_unmount_snap(fullname);
3855
3856 return (0);
3857 }
3858
3859 /*
3860 * inputs:
3861 * zc_name old name of dataset
3862 * zc_value new name of dataset
3863 * zc_cookie recursive flag (only valid for snapshots)
3864 *
3865 * outputs: none
3866 */
3867 static int
3868 zfs_ioc_rename(zfs_cmd_t *zc)
3869 {
3870 objset_t *os;
3871 dmu_objset_type_t ost;
3872 boolean_t recursive = zc->zc_cookie & 1;
3873 char *at;
3874 int err;
3875
3876 /* "zfs rename" from and to ...%recv datasets should both fail */
3877 zc->zc_name[sizeof (zc->zc_name) - 1] = '\0';
3878 zc->zc_value[sizeof (zc->zc_value) - 1] = '\0';
3879 if (dataset_namecheck(zc->zc_name, NULL, NULL) != 0 ||
3880 dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
3881 strchr(zc->zc_name, '%') || strchr(zc->zc_value, '%'))
3882 return (SET_ERROR(EINVAL));
3883
3884 err = dmu_objset_hold(zc->zc_name, FTAG, &os);
3885 if (err != 0)
3886 return (err);
3887 ost = dmu_objset_type(os);
3888 dmu_objset_rele(os, FTAG);
3889
3890 at = strchr(zc->zc_name, '@');
3891 if (at != NULL) {
3892 /* snaps must be in same fs */
3893 int error;
3894
3895 if (strncmp(zc->zc_name, zc->zc_value, at - zc->zc_name + 1))
3896 return (SET_ERROR(EXDEV));
3897 *at = '\0';
3898 if (ost == DMU_OST_ZFS) {
3899 error = dmu_objset_find(zc->zc_name,
3900 recursive_unmount, at + 1,
3901 recursive ? DS_FIND_CHILDREN : 0);
3902 if (error != 0) {
3903 *at = '@';
3904 return (error);
3905 }
3906 }
3907 error = dsl_dataset_rename_snapshot(zc->zc_name,
3908 at + 1, strchr(zc->zc_value, '@') + 1, recursive);
3909 *at = '@';
3910
3911 return (error);
3912 } else {
3913 if (ost == DMU_OST_ZVOL)
3914 (void) zvol_remove_minor(zc->zc_name);
3915 return (dsl_dir_rename(zc->zc_name, zc->zc_value));
3916 }
3917 }
3918
3919 static int
3920 zfs_check_settable(const char *dsname, nvpair_t *pair, cred_t *cr)
3921 {
3922 const char *propname = nvpair_name(pair);
3923 boolean_t issnap = (strchr(dsname, '@') != NULL);
3924 zfs_prop_t prop = zfs_name_to_prop(propname);
3925 uint64_t intval;
3926 int err;
3927
3928 if (prop == ZPROP_INVAL) {
3929 if (zfs_prop_user(propname)) {
3930 if (err = zfs_secpolicy_write_perms(dsname,
3931 ZFS_DELEG_PERM_USERPROP, cr))
3932 return (err);
3933 return (0);
3934 }
3935
3936 if (!issnap && zfs_prop_userquota(propname)) {
3937 const char *perm = NULL;
3938 const char *uq_prefix =
3939 zfs_userquota_prop_prefixes[ZFS_PROP_USERQUOTA];
3940 const char *gq_prefix =
3941 zfs_userquota_prop_prefixes[ZFS_PROP_GROUPQUOTA];
3942
3943 if (strncmp(propname, uq_prefix,
3944 strlen(uq_prefix)) == 0) {
3945 perm = ZFS_DELEG_PERM_USERQUOTA;
3946 } else if (strncmp(propname, gq_prefix,
3947 strlen(gq_prefix)) == 0) {
3948 perm = ZFS_DELEG_PERM_GROUPQUOTA;
3949 } else {
3950 /* USERUSED and GROUPUSED are read-only */
3951 return (SET_ERROR(EINVAL));
3952 }
3953
3954 if (err = zfs_secpolicy_write_perms(dsname, perm, cr))
3955 return (err);
3956 return (0);
3957 }
3958
3959 return (SET_ERROR(EINVAL));
3960 }
3961
3962 if (issnap)
3963 return (SET_ERROR(EINVAL));
3964
3965 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
3966 /*
3967 * dsl_prop_get_all_impl() returns properties in this
3968 * format.
3969 */
3970 nvlist_t *attrs;
3971 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
3972 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
3973 &pair) == 0);
3974 }
3975
3976 /*
3977 * Check that this value is valid for this pool version
3978 */
3979 switch (prop) {
3980 case ZFS_PROP_COMPRESSION:
3981 /*
3982 * If the user specified gzip compression, make sure
3983 * the SPA supports it. We ignore any errors here since
3984 * we'll catch them later.
3985 */
3986 if (nvpair_value_uint64(pair, &intval) == 0) {
3987 if (intval >= ZIO_COMPRESS_GZIP_1 &&
3988 intval <= ZIO_COMPRESS_GZIP_9 &&
3989 zfs_earlier_version(dsname,
3990 SPA_VERSION_GZIP_COMPRESSION)) {
3991 return (SET_ERROR(ENOTSUP));
3992 }
3993
3994 if (intval == ZIO_COMPRESS_ZLE &&
3995 zfs_earlier_version(dsname,
3996 SPA_VERSION_ZLE_COMPRESSION))
3997 return (SET_ERROR(ENOTSUP));
3998
3999 if (intval == ZIO_COMPRESS_LZ4) {
4000 spa_t *spa;
4001
4002 if ((err = spa_open(dsname, &spa, FTAG)) != 0)
4003 return (err);
4004
4005 if (!spa_feature_is_enabled(spa,
4006 SPA_FEATURE_LZ4_COMPRESS)) {
4007 spa_close(spa, FTAG);
4008 return (SET_ERROR(ENOTSUP));
4009 }
4010 spa_close(spa, FTAG);
4011 }
4012
4013 /*
4014 * If this is a bootable dataset then
4015 * verify that the compression algorithm
4016 * is supported for booting. We must return
4017 * something other than ENOTSUP since it
4018 * implies a downrev pool version.
4019 */
4020 if (zfs_is_bootfs(dsname) &&
4021 !BOOTFS_COMPRESS_VALID(intval)) {
4022 return (SET_ERROR(ERANGE));
4023 }
4024 }
4025 break;
4026
4027 case ZFS_PROP_COPIES:
4028 if (zfs_earlier_version(dsname, SPA_VERSION_DITTO_BLOCKS))
4029 return (SET_ERROR(ENOTSUP));
4030 break;
4031
4032 case ZFS_PROP_RECORDSIZE:
4033 /* Record sizes above 128k need the feature to be enabled */
4034 if (nvpair_value_uint64(pair, &intval) == 0 &&
4035 intval > SPA_OLD_MAXBLOCKSIZE) {
4036 spa_t *spa;
4037
4038 /*
4039 * We don't allow setting the property above 1MB,
4040 * unless the tunable has been changed.
4041 */
4042 if (intval > zfs_max_recordsize ||
4043 intval > SPA_MAXBLOCKSIZE)
4044 return (SET_ERROR(ERANGE));
4045
4046 if ((err = spa_open(dsname, &spa, FTAG)) != 0)
4047 return (err);
4048
4049 if (!spa_feature_is_enabled(spa,
4050 SPA_FEATURE_LARGE_BLOCKS)) {
4051 spa_close(spa, FTAG);
4052 return (SET_ERROR(ENOTSUP));
4053 }
4054 spa_close(spa, FTAG);
4055 }
4056 break;
4057
4058 case ZFS_PROP_SHARESMB:
4059 if (zpl_earlier_version(dsname, ZPL_VERSION_FUID))
4060 return (SET_ERROR(ENOTSUP));
4061 break;
4062
4063 case ZFS_PROP_ACLINHERIT:
4064 if (nvpair_type(pair) == DATA_TYPE_UINT64 &&
4065 nvpair_value_uint64(pair, &intval) == 0) {
4066 if (intval == ZFS_ACL_PASSTHROUGH_X &&
4067 zfs_earlier_version(dsname,
4068 SPA_VERSION_PASSTHROUGH_X))
4069 return (SET_ERROR(ENOTSUP));
4070 }
4071 break;
4072
4073 case ZFS_PROP_CHECKSUM:
4074 case ZFS_PROP_DEDUP:
4075 {
4076 spa_feature_t feature;
4077 spa_t *spa;
4078
4079 /* dedup feature version checks */
4080 if (prop == ZFS_PROP_DEDUP &&
4081 zfs_earlier_version(dsname, SPA_VERSION_DEDUP))
4082 return (SET_ERROR(ENOTSUP));
4083
4084 if (nvpair_value_uint64(pair, &intval) != 0)
4085 return (SET_ERROR(EINVAL));
4086
4087 /* check prop value is enabled in features */
4088 feature = zio_checksum_to_feature(intval & ZIO_CHECKSUM_MASK);
4089 if (feature == SPA_FEATURE_NONE)
4090 break;
4091
4092 if ((err = spa_open(dsname, &spa, FTAG)) != 0)
4093 return (err);
4094
4095 if (!spa_feature_is_enabled(spa, feature)) {
4096 spa_close(spa, FTAG);
4097 return (SET_ERROR(ENOTSUP));
4098 }
4099 spa_close(spa, FTAG);
4100 break;
4101 }
4102 }
4103
4104 return (zfs_secpolicy_setprop(dsname, prop, pair, CRED()));
4105 }
4106
4107 /*
4108 * Checks for a race condition to make sure we don't increment a feature flag
4109 * multiple times.
4110 */
4111 static int
4112 zfs_prop_activate_feature_check(void *arg, dmu_tx_t *tx)
4113 {
4114 spa_t *spa = dmu_tx_pool(tx)->dp_spa;
4115 spa_feature_t *featurep = arg;
4116
4117 if (!spa_feature_is_active(spa, *featurep))
4118 return (0);
4119 else
4120 return (SET_ERROR(EBUSY));
4121 }
4122
4123 /*
4124 * The callback invoked on feature activation in the sync task caused by
4125 * zfs_prop_activate_feature.
4126 */
4127 static void
4128 zfs_prop_activate_feature_sync(void *arg, dmu_tx_t *tx)
4129 {
4130 spa_t *spa = dmu_tx_pool(tx)->dp_spa;
4131 spa_feature_t *featurep = arg;
4132
4133 spa_feature_incr(spa, *featurep, tx);
4134 }
4135
4136 /*
4137 * Activates a feature on a pool in response to a property setting. This
4138 * creates a new sync task which modifies the pool to reflect the feature
4139 * as being active.
4140 */
4141 static int
4142 zfs_prop_activate_feature(spa_t *spa, spa_feature_t feature)
4143 {
4144 int err;
4145
4146 /* EBUSY here indicates that the feature is already active */
4147 err = dsl_sync_task(spa_name(spa),
4148 zfs_prop_activate_feature_check, zfs_prop_activate_feature_sync,
4149 &feature, 2, ZFS_SPACE_CHECK_RESERVED);
4150
4151 if (err != 0 && err != EBUSY)
4152 return (err);
4153 else
4154 return (0);
4155 }
4156
4157 /*
4158 * Removes properties from the given props list that fail permission checks
4159 * needed to clear them and to restore them in case of a receive error. For each
4160 * property, make sure we have both set and inherit permissions.
4161 *
4162 * Returns the first error encountered if any permission checks fail. If the
4163 * caller provides a non-NULL errlist, it also gives the complete list of names
4164 * of all the properties that failed a permission check along with the
4165 * corresponding error numbers. The caller is responsible for freeing the
4166 * returned errlist.
4167 *
4168 * If every property checks out successfully, zero is returned and the list
4169 * pointed at by errlist is NULL.
4170 */
4171 static int
4172 zfs_check_clearable(char *dataset, nvlist_t *props, nvlist_t **errlist)
4173 {
4174 zfs_cmd_t *zc;
4175 nvpair_t *pair, *next_pair;
4176 nvlist_t *errors;
4177 int err, rv = 0;
4178
4179 if (props == NULL)
4180 return (0);
4181
4182 VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0);
4183
4184 zc = kmem_alloc(sizeof (zfs_cmd_t), KM_SLEEP);
4185 (void) strcpy(zc->zc_name, dataset);
4186 pair = nvlist_next_nvpair(props, NULL);
4187 while (pair != NULL) {
4188 next_pair = nvlist_next_nvpair(props, pair);
4189
4190 (void) strcpy(zc->zc_value, nvpair_name(pair));
4191 if ((err = zfs_check_settable(dataset, pair, CRED())) != 0 ||
4192 (err = zfs_secpolicy_inherit_prop(zc, NULL, CRED())) != 0) {
4193 VERIFY(nvlist_remove_nvpair(props, pair) == 0);
4194 VERIFY(nvlist_add_int32(errors,
4195 zc->zc_value, err) == 0);
4196 }
4197 pair = next_pair;
4198 }
4199 kmem_free(zc, sizeof (zfs_cmd_t));
4200
4201 if ((pair = nvlist_next_nvpair(errors, NULL)) == NULL) {
4202 nvlist_free(errors);
4203 errors = NULL;
4204 } else {
4205 VERIFY(nvpair_value_int32(pair, &rv) == 0);
4206 }
4207
4208 if (errlist == NULL)
4209 nvlist_free(errors);
4210 else
4211 *errlist = errors;
4212
4213 return (rv);
4214 }
4215
4216 static boolean_t
4217 propval_equals(nvpair_t *p1, nvpair_t *p2)
4218 {
4219 if (nvpair_type(p1) == DATA_TYPE_NVLIST) {
4220 /* dsl_prop_get_all_impl() format */
4221 nvlist_t *attrs;
4222 VERIFY(nvpair_value_nvlist(p1, &attrs) == 0);
4223 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
4224 &p1) == 0);
4225 }
4226
4227 if (nvpair_type(p2) == DATA_TYPE_NVLIST) {
4228 nvlist_t *attrs;
4229 VERIFY(nvpair_value_nvlist(p2, &attrs) == 0);
4230 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
4231 &p2) == 0);
4232 }
4233
4234 if (nvpair_type(p1) != nvpair_type(p2))
4235 return (B_FALSE);
4236
4237 if (nvpair_type(p1) == DATA_TYPE_STRING) {
4238 char *valstr1, *valstr2;
4239
4240 VERIFY(nvpair_value_string(p1, (char **)&valstr1) == 0);
4241 VERIFY(nvpair_value_string(p2, (char **)&valstr2) == 0);
4242 return (strcmp(valstr1, valstr2) == 0);
4243 } else {
4244 uint64_t intval1, intval2;
4245
4246 VERIFY(nvpair_value_uint64(p1, &intval1) == 0);
4247 VERIFY(nvpair_value_uint64(p2, &intval2) == 0);
4248 return (intval1 == intval2);
4249 }
4250 }
4251
4252 /*
4253 * Remove properties from props if they are not going to change (as determined
4254 * by comparison with origprops). Remove them from origprops as well, since we
4255 * do not need to clear or restore properties that won't change.
4256 */
4257 static void
4258 props_reduce(nvlist_t *props, nvlist_t *origprops)
4259 {
4260 nvpair_t *pair, *next_pair;
4261
4262 if (origprops == NULL)
4263 return; /* all props need to be received */
4264
4265 pair = nvlist_next_nvpair(props, NULL);
4266 while (pair != NULL) {
4267 const char *propname = nvpair_name(pair);
4268 nvpair_t *match;
4269
4270 next_pair = nvlist_next_nvpair(props, pair);
4271
4272 if ((nvlist_lookup_nvpair(origprops, propname,
4273 &match) != 0) || !propval_equals(pair, match))
4274 goto next; /* need to set received value */
4275
4276 /* don't clear the existing received value */
4277 (void) nvlist_remove_nvpair(origprops, match);
4278 /* don't bother receiving the property */
4279 (void) nvlist_remove_nvpair(props, pair);
4280 next:
4281 pair = next_pair;
4282 }
4283 }
4284
4285 /*
4286 * Extract properties that cannot be set PRIOR to the receipt of a dataset.
4287 * For example, refquota cannot be set until after the receipt of a dataset,
4288 * because in replication streams, an older/earlier snapshot may exceed the
4289 * refquota. We want to receive the older/earlier snapshot, but setting
4290 * refquota pre-receipt will set the dsl's ACTUAL quota, which will prevent
4291 * the older/earlier snapshot from being received (with EDQUOT).
4292 *
4293 * The ZFS test "zfs_receive_011_pos" demonstrates such a scenario.
4294 *
4295 * libzfs will need to be judicious handling errors encountered by props
4296 * extracted by this function.
4297 */
4298 static nvlist_t *
4299 extract_delay_props(nvlist_t *props)
4300 {
4301 nvlist_t *delayprops;
4302 nvpair_t *nvp, *tmp;
4303 static const zfs_prop_t delayable[] = { ZFS_PROP_REFQUOTA, 0 };
4304 int i;
4305
4306 VERIFY(nvlist_alloc(&delayprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
4307
4308 for (nvp = nvlist_next_nvpair(props, NULL); nvp != NULL;
4309 nvp = nvlist_next_nvpair(props, nvp)) {
4310 /*
4311 * strcmp() is safe because zfs_prop_to_name() always returns
4312 * a bounded string.
4313 */
4314 for (i = 0; delayable[i] != 0; i++) {
4315 if (strcmp(zfs_prop_to_name(delayable[i]),
4316 nvpair_name(nvp)) == 0) {
4317 break;
4318 }
4319 }
4320 if (delayable[i] != 0) {
4321 tmp = nvlist_prev_nvpair(props, nvp);
4322 VERIFY(nvlist_add_nvpair(delayprops, nvp) == 0);
4323 VERIFY(nvlist_remove_nvpair(props, nvp) == 0);
4324 nvp = tmp;
4325 }
4326 }
4327
4328 if (nvlist_empty(delayprops)) {
4329 nvlist_free(delayprops);
4330 delayprops = NULL;
4331 }
4332 return (delayprops);
4333 }
4334
4335 #ifdef DEBUG
4336 static boolean_t zfs_ioc_recv_inject_err;
4337 #endif
4338
4339 /*
4340 * inputs:
4341 * zc_name name of containing filesystem
4342 * zc_nvlist_src{_size} nvlist of properties to apply
4343 * zc_value name of snapshot to create
4344 * zc_string name of clone origin (if DRR_FLAG_CLONE)
4345 * zc_cookie file descriptor to recv from
4346 * zc_begin_record the BEGIN record of the stream (not byteswapped)
4347 * zc_guid force flag
4348 * zc_cleanup_fd cleanup-on-exit file descriptor
4349 * zc_action_handle handle for this guid/ds mapping (or zero on first call)
4350 * zc_resumable if data is incomplete assume sender will resume
4351 *
4352 * outputs:
4353 * zc_cookie number of bytes read
4354 * zc_nvlist_dst{_size} error for each unapplied received property
4355 * zc_obj zprop_errflags_t
4356 * zc_action_handle handle for this guid/ds mapping
4357 */
4358 static int
4359 zfs_ioc_recv(zfs_cmd_t *zc)
4360 {
4361 file_t *fp;
4362 dmu_recv_cookie_t drc;
4363 boolean_t force = (boolean_t)zc->zc_guid;
4364 int fd;
4365 int error = 0;
4366 int props_error = 0;
4367 nvlist_t *errors;
4368 offset_t off;
4369 nvlist_t *props = NULL; /* sent properties */
4370 nvlist_t *origprops = NULL; /* existing properties */
4371 nvlist_t *delayprops = NULL; /* sent properties applied post-receive */
4372 char *origin = NULL;
4373 char *tosnap;
4374 char tofs[ZFS_MAX_DATASET_NAME_LEN];
4375 boolean_t first_recvd_props = B_FALSE;
4376
4377 if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
4378 strchr(zc->zc_value, '@') == NULL ||
4379 strchr(zc->zc_value, '%'))
4380 return (SET_ERROR(EINVAL));
4381
4382 (void) strcpy(tofs, zc->zc_value);
4383 tosnap = strchr(tofs, '@');
4384 *tosnap++ = '\0';
4385
4386 if (zc->zc_nvlist_src != NULL &&
4387 (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
4388 zc->zc_iflags, &props)) != 0)
4389 return (error);
4390
4391 fd = zc->zc_cookie;
4392 fp = getf(fd);
4393 if (fp == NULL) {
4394 nvlist_free(props);
4395 return (SET_ERROR(EBADF));
4396 }
4397
4398 errors = fnvlist_alloc();
4399
4400 if (zc->zc_string[0])
4401 origin = zc->zc_string;
4402
4403 error = dmu_recv_begin(tofs, tosnap,
4404 &zc->zc_begin_record, force, zc->zc_resumable, origin, &drc);
4405 if (error != 0)
4406 goto out;
4407
4408 /*
4409 * Set properties before we receive the stream so that they are applied
4410 * to the new data. Note that we must call dmu_recv_stream() if
4411 * dmu_recv_begin() succeeds.
4412 */
4413 if (props != NULL && !drc.drc_newfs) {
4414 if (spa_version(dsl_dataset_get_spa(drc.drc_ds)) >=
4415 SPA_VERSION_RECVD_PROPS &&
4416 !dsl_prop_get_hasrecvd(tofs))
4417 first_recvd_props = B_TRUE;
4418
4419 /*
4420 * If new received properties are supplied, they are to
4421 * completely replace the existing received properties, so stash
4422 * away the existing ones.
4423 */
4424 if (dsl_prop_get_received(tofs, &origprops) == 0) {
4425 nvlist_t *errlist = NULL;
4426 /*
4427 * Don't bother writing a property if its value won't
4428 * change (and avoid the unnecessary security checks).
4429 *
4430 * The first receive after SPA_VERSION_RECVD_PROPS is a
4431 * special case where we blow away all local properties
4432 * regardless.
4433 */
4434 if (!first_recvd_props)
4435 props_reduce(props, origprops);
4436 if (zfs_check_clearable(tofs, origprops, &errlist) != 0)
4437 (void) nvlist_merge(errors, errlist, 0);
4438 nvlist_free(errlist);
4439
4440 if (clear_received_props(tofs, origprops,
4441 first_recvd_props ? NULL : props) != 0)
4442 zc->zc_obj |= ZPROP_ERR_NOCLEAR;
4443 } else {
4444 zc->zc_obj |= ZPROP_ERR_NOCLEAR;
4445 }
4446 }
4447
4448 if (props != NULL) {
4449 props_error = dsl_prop_set_hasrecvd(tofs);
4450
4451 if (props_error == 0) {
4452 delayprops = extract_delay_props(props);
4453 (void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_RECEIVED,
4454 props, errors);
4455 }
4456 }
4457
4458 off = fp->f_offset;
4459 error = dmu_recv_stream(&drc, fp->f_vnode, &off, zc->zc_cleanup_fd,
4460 &zc->zc_action_handle);
4461
4462 if (error == 0) {
4463 zfsvfs_t *zfsvfs = NULL;
4464
4465 if (getzfsvfs(tofs, &zfsvfs) == 0) {
4466 /* online recv */
4467 dsl_dataset_t *ds;
4468 int end_err;
4469
4470 ds = dmu_objset_ds(zfsvfs->z_os);
4471 error = zfs_suspend_fs(zfsvfs);
4472 /*
4473 * If the suspend fails, then the recv_end will
4474 * likely also fail, and clean up after itself.
4475 */
4476 end_err = dmu_recv_end(&drc, zfsvfs);
4477 if (error == 0)
4478 error = zfs_resume_fs(zfsvfs, ds);
4479 error = error ? error : end_err;
4480 VFS_RELE(zfsvfs->z_vfs);
4481 } else {
4482 error = dmu_recv_end(&drc, NULL);
4483 }
4484
4485 /* Set delayed properties now, after we're done receiving. */
4486 if (delayprops != NULL && error == 0) {
4487 (void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_RECEIVED,
4488 delayprops, errors);
4489 }
4490 }
4491
4492 if (delayprops != NULL) {
4493 /*
4494 * Merge delayed props back in with initial props, in case
4495 * we're DEBUG and zfs_ioc_recv_inject_err is set (which means
4496 * we have to make sure clear_received_props() includes
4497 * the delayed properties).
4498 *
4499 * Since zfs_ioc_recv_inject_err is only in DEBUG kernels,
4500 * using ASSERT() will be just like a VERIFY.
4501 */
4502 ASSERT(nvlist_merge(props, delayprops, 0) == 0);
4503 nvlist_free(delayprops);
4504 }
4505
4506 /*
4507 * Now that all props, initial and delayed, are set, report the prop
4508 * errors to the caller.
4509 */
4510 if (zc->zc_nvlist_dst_size != 0 &&
4511 (nvlist_smush(errors, zc->zc_nvlist_dst_size) != 0 ||
4512 put_nvlist(zc, errors) != 0)) {
4513 /*
4514 * Caller made zc->zc_nvlist_dst less than the minimum expected
4515 * size or supplied an invalid address.
4516 */
4517 props_error = SET_ERROR(EINVAL);
4518 }
4519
4520 zc->zc_cookie = off - fp->f_offset;
4521 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
4522 fp->f_offset = off;
4523
4524 #ifdef DEBUG
4525 if (zfs_ioc_recv_inject_err) {
4526 zfs_ioc_recv_inject_err = B_FALSE;
4527 error = 1;
4528 }
4529 #endif
4530 /*
4531 * On error, restore the original props.
4532 */
4533 if (error != 0 && props != NULL && !drc.drc_newfs) {
4534 if (clear_received_props(tofs, props, NULL) != 0) {
4535 /*
4536 * We failed to clear the received properties.
4537 * Since we may have left a $recvd value on the
4538 * system, we can't clear the $hasrecvd flag.
4539 */
4540 zc->zc_obj |= ZPROP_ERR_NORESTORE;
4541 } else if (first_recvd_props) {
4542 dsl_prop_unset_hasrecvd(tofs);
4543 }
4544
4545 if (origprops == NULL && !drc.drc_newfs) {
4546 /* We failed to stash the original properties. */
4547 zc->zc_obj |= ZPROP_ERR_NORESTORE;
4548 }
4549
4550 /*
4551 * dsl_props_set() will not convert RECEIVED to LOCAL on or
4552 * after SPA_VERSION_RECVD_PROPS, so we need to specify LOCAL
4553 * explictly if we're restoring local properties cleared in the
4554 * first new-style receive.
4555 */
4556 if (origprops != NULL &&
4557 zfs_set_prop_nvlist(tofs, (first_recvd_props ?
4558 ZPROP_SRC_LOCAL : ZPROP_SRC_RECEIVED),
4559 origprops, NULL) != 0) {
4560 /*
4561 * We stashed the original properties but failed to
4562 * restore them.
4563 */
4564 zc->zc_obj |= ZPROP_ERR_NORESTORE;
4565 }
4566 }
4567 out:
4568 nvlist_free(props);
4569 nvlist_free(origprops);
4570 nvlist_free(errors);
4571 releasef(fd);
4572
4573 if (error == 0)
4574 error = props_error;
4575
4576 return (error);
4577 }
4578
4579 /*
4580 * inputs:
4581 * zc_name name of snapshot to send
4582 * zc_cookie file descriptor to send stream to
4583 * zc_obj fromorigin flag (mutually exclusive with zc_fromobj)
4584 * zc_sendobj objsetid of snapshot to send
4585 * zc_fromobj objsetid of incremental fromsnap (may be zero)
4586 * zc_guid if set, estimate size of stream only. zc_cookie is ignored.
4587 * output size in zc_objset_type.
4588 * zc_flags lzc_send_flags
4589 *
4590 * outputs:
4591 * zc_objset_type estimated size, if zc_guid is set
4592 */
4593 static int
4594 zfs_ioc_send(zfs_cmd_t *zc)
4595 {
4596 int error;
4597 offset_t off;
4598 boolean_t estimate = (zc->zc_guid != 0);
4599 boolean_t embedok = (zc->zc_flags & 0x1);
4600 boolean_t large_block_ok = (zc->zc_flags & 0x2);
4601 boolean_t compressok = (zc->zc_flags & 0x4);
4602
4603 if (zc->zc_obj != 0) {
4604 dsl_pool_t *dp;
4605 dsl_dataset_t *tosnap;
4606
4607 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
4608 if (error != 0)
4609 return (error);
4610
4611 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &tosnap);
4612 if (error != 0) {
4613 dsl_pool_rele(dp, FTAG);
4614 return (error);
4615 }
4616
4617 if (dsl_dir_is_clone(tosnap->ds_dir))
4618 zc->zc_fromobj =
4619 dsl_dir_phys(tosnap->ds_dir)->dd_origin_obj;
4620 dsl_dataset_rele(tosnap, FTAG);
4621 dsl_pool_rele(dp, FTAG);
4622 }
4623
4624 if (estimate) {
4625 dsl_pool_t *dp;
4626 dsl_dataset_t *tosnap;
4627 dsl_dataset_t *fromsnap = NULL;
4628
4629 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
4630 if (error != 0)
4631 return (error);
4632
4633 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &tosnap);
4634 if (error != 0) {
4635 dsl_pool_rele(dp, FTAG);
4636 return (error);
4637 }
4638
4639 if (zc->zc_fromobj != 0) {
4640 error = dsl_dataset_hold_obj(dp, zc->zc_fromobj,
4641 FTAG, &fromsnap);
4642 if (error != 0) {
4643 dsl_dataset_rele(tosnap, FTAG);
4644 dsl_pool_rele(dp, FTAG);
4645 return (error);
4646 }
4647 }
4648
4649 error = dmu_send_estimate(tosnap, fromsnap, compressok,
4650 &zc->zc_objset_type);
4651
4652 if (fromsnap != NULL)
4653 dsl_dataset_rele(fromsnap, FTAG);
4654 dsl_dataset_rele(tosnap, FTAG);
4655 dsl_pool_rele(dp, FTAG);
4656 } else {
4657 file_t *fp = getf(zc->zc_cookie);
4658 if (fp == NULL)
4659 return (SET_ERROR(EBADF));
4660
4661 off = fp->f_offset;
4662 error = dmu_send_obj(zc->zc_name, zc->zc_sendobj,
4663 zc->zc_fromobj, embedok, large_block_ok, compressok,
4664 zc->zc_cookie, fp->f_vnode, &off);
4665
4666 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
4667 fp->f_offset = off;
4668 releasef(zc->zc_cookie);
4669 }
4670 return (error);
4671 }
4672
4673 /*
4674 * inputs:
4675 * zc_name name of snapshot on which to report progress
4676 * zc_cookie file descriptor of send stream
4677 *
4678 * outputs:
4679 * zc_cookie number of bytes written in send stream thus far
4680 */
4681 static int
4682 zfs_ioc_send_progress(zfs_cmd_t *zc)
4683 {
4684 dsl_pool_t *dp;
4685 dsl_dataset_t *ds;
4686 dmu_sendarg_t *dsp = NULL;
4687 int error;
4688
4689 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
4690 if (error != 0)
4691 return (error);
4692
4693 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &ds);
4694 if (error != 0) {
4695 dsl_pool_rele(dp, FTAG);
4696 return (error);
4697 }
4698
4699 mutex_enter(&ds->ds_sendstream_lock);
4700
4701 /*
4702 * Iterate over all the send streams currently active on this dataset.
4703 * If there's one which matches the specified file descriptor _and_ the
4704 * stream was started by the current process, return the progress of
4705 * that stream.
4706 */
4707 for (dsp = list_head(&ds->ds_sendstreams); dsp != NULL;
4708 dsp = list_next(&ds->ds_sendstreams, dsp)) {
4709 if (dsp->dsa_outfd == zc->zc_cookie &&
4710 dsp->dsa_proc == curproc)
4711 break;
4712 }
4713
4714 if (dsp != NULL)
4715 zc->zc_cookie = *(dsp->dsa_off);
4716 else
4717 error = SET_ERROR(ENOENT);
4718
4719 mutex_exit(&ds->ds_sendstream_lock);
4720 dsl_dataset_rele(ds, FTAG);
4721 dsl_pool_rele(dp, FTAG);
4722 return (error);
4723 }
4724
4725 static int
4726 zfs_ioc_inject_fault(zfs_cmd_t *zc)
4727 {
4728 int id, error;
4729
4730 error = zio_inject_fault(zc->zc_name, (int)zc->zc_guid, &id,
4731 &zc->zc_inject_record);
4732
4733 if (error == 0)
4734 zc->zc_guid = (uint64_t)id;
4735
4736 return (error);
4737 }
4738
4739 static int
4740 zfs_ioc_clear_fault(zfs_cmd_t *zc)
4741 {
4742 return (zio_clear_fault((int)zc->zc_guid));
4743 }
4744
4745 static int
4746 zfs_ioc_inject_list_next(zfs_cmd_t *zc)
4747 {
4748 int id = (int)zc->zc_guid;
4749 int error;
4750
4751 error = zio_inject_list_next(&id, zc->zc_name, sizeof (zc->zc_name),
4752 &zc->zc_inject_record);
4753
4754 zc->zc_guid = id;
4755
4756 return (error);
4757 }
4758
4759 static int
4760 zfs_ioc_error_log(zfs_cmd_t *zc)
4761 {
4762 spa_t *spa;
4763 int error;
4764 size_t count = (size_t)zc->zc_nvlist_dst_size;
4765
4766 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
4767 return (error);
4768
4769 error = spa_get_errlog(spa, (void *)(uintptr_t)zc->zc_nvlist_dst,
4770 &count);
4771 if (error == 0)
4772 zc->zc_nvlist_dst_size = count;
4773 else
4774 zc->zc_nvlist_dst_size = spa_get_errlog_size(spa);
4775
4776 spa_close(spa, FTAG);
4777
4778 return (error);
4779 }
4780
4781 static int
4782 zfs_ioc_clear(zfs_cmd_t *zc)
4783 {
4784 spa_t *spa;
4785 vdev_t *vd;
4786 int error;
4787
4788 /*
4789 * On zpool clear we also fix up missing slogs
4790 */
4791 mutex_enter(&spa_namespace_lock);
4792 spa = spa_lookup(zc->zc_name);
4793 if (spa == NULL) {
4794 mutex_exit(&spa_namespace_lock);
4795 return (SET_ERROR(EIO));
4796 }
4797 if (spa_get_log_state(spa) == SPA_LOG_MISSING) {
4798 /* we need to let spa_open/spa_load clear the chains */
4799 spa_set_log_state(spa, SPA_LOG_CLEAR);
4800 }
4801 spa->spa_last_open_failed = 0;
4802 mutex_exit(&spa_namespace_lock);
4803
4804 if (zc->zc_cookie & ZPOOL_NO_REWIND) {
4805 error = spa_open(zc->zc_name, &spa, FTAG);
4806 } else {
4807 nvlist_t *policy;
4808 nvlist_t *config = NULL;
4809
4810 if (zc->zc_nvlist_src == NULL)
4811 return (SET_ERROR(EINVAL));
4812
4813 if ((error = get_nvlist(zc->zc_nvlist_src,
4814 zc->zc_nvlist_src_size, zc->zc_iflags, &policy)) == 0) {
4815 error = spa_open_rewind(zc->zc_name, &spa, FTAG,
4816 policy, &config);
4817 if (config != NULL) {
4818 int err;
4819
4820 if ((err = put_nvlist(zc, config)) != 0)
4821 error = err;
4822 nvlist_free(config);
4823 }
4824 nvlist_free(policy);
4825 }
4826 }
4827
4828 if (error != 0)
4829 return (error);
4830
4831 spa_vdev_state_enter(spa, SCL_NONE);
4832
4833 if (zc->zc_guid == 0) {
4834 vd = NULL;
4835 } else {
4836 vd = spa_lookup_by_guid(spa, zc->zc_guid, B_TRUE);
4837 if (vd == NULL) {
4838 (void) spa_vdev_state_exit(spa, NULL, ENODEV);
4839 spa_close(spa, FTAG);
4840 return (SET_ERROR(ENODEV));
4841 }
4842 }
4843
4844 vdev_clear(spa, vd);
4845
4846 (void) spa_vdev_state_exit(spa, NULL, 0);
4847
4848 /*
4849 * Resume any suspended I/Os.
4850 */
4851 if (zio_resume(spa) != 0)
4852 error = SET_ERROR(EIO);
4853
4854 spa_close(spa, FTAG);
4855
4856 return (error);
4857 }
4858
4859 static int
4860 zfs_ioc_pool_reopen(zfs_cmd_t *zc)
4861 {
4862 spa_t *spa;
4863 int error;
4864
4865 error = spa_open(zc->zc_name, &spa, FTAG);
4866 if (error != 0)
4867 return (error);
4868
4869 spa_vdev_state_enter(spa, SCL_NONE);
4870
4871 /*
4872 * If a resilver is already in progress then set the
4873 * spa_scrub_reopen flag to B_TRUE so that we don't restart
4874 * the scan as a side effect of the reopen. Otherwise, let
4875 * vdev_open() decided if a resilver is required.
4876 */
4877 spa->spa_scrub_reopen = dsl_scan_resilvering(spa->spa_dsl_pool);
4878 vdev_reopen(spa->spa_root_vdev);
4879 spa->spa_scrub_reopen = B_FALSE;
4880
4881 (void) spa_vdev_state_exit(spa, NULL, 0);
4882 spa_close(spa, FTAG);
4883 return (0);
4884 }
4885 /*
4886 * inputs:
4887 * zc_name name of filesystem
4888 *
4889 * outputs:
4890 * zc_string name of conflicting snapshot, if there is one
4891 */
4892 static int
4893 zfs_ioc_promote(zfs_cmd_t *zc)
4894 {
4895 dsl_pool_t *dp;
4896 dsl_dataset_t *ds, *ods;
4897 char origin[ZFS_MAX_DATASET_NAME_LEN];
4898 char *cp;
4899 int error;
4900
4901 zc->zc_name[sizeof (zc->zc_name) - 1] = '\0';
4902 if (dataset_namecheck(zc->zc_name, NULL, NULL) != 0 ||
4903 strchr(zc->zc_name, '%'))
4904 return (SET_ERROR(EINVAL));
4905
4906 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
4907 if (error != 0)
4908 return (error);
4909
4910 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &ds);
4911 if (error != 0) {
4912 dsl_pool_rele(dp, FTAG);
4913 return (error);
4914 }
4915
4916 if (!dsl_dir_is_clone(ds->ds_dir)) {
4917 dsl_dataset_rele(ds, FTAG);
4918 dsl_pool_rele(dp, FTAG);
4919 return (SET_ERROR(EINVAL));
4920 }
4921
4922 error = dsl_dataset_hold_obj(dp,
4923 dsl_dir_phys(ds->ds_dir)->dd_origin_obj, FTAG, &ods);
4924 if (error != 0) {
4925 dsl_dataset_rele(ds, FTAG);
4926 dsl_pool_rele(dp, FTAG);
4927 return (error);
4928 }
4929
4930 dsl_dataset_name(ods, origin);
4931 dsl_dataset_rele(ods, FTAG);
4932 dsl_dataset_rele(ds, FTAG);
4933 dsl_pool_rele(dp, FTAG);
4934
4935 /*
4936 * We don't need to unmount *all* the origin fs's snapshots, but
4937 * it's easier.
4938 */
4939 cp = strchr(origin, '@');
4940 if (cp)
4941 *cp = '\0';
4942 (void) dmu_objset_find(origin,
4943 zfs_unmount_snap_cb, NULL, DS_FIND_SNAPSHOTS);
4944 return (dsl_dataset_promote(zc->zc_name, zc->zc_string));
4945 }
4946
4947 /*
4948 * Retrieve a single {user|group}{used|quota}@... property.
4949 *
4950 * inputs:
4951 * zc_name name of filesystem
4952 * zc_objset_type zfs_userquota_prop_t
4953 * zc_value domain name (eg. "S-1-234-567-89")
4954 * zc_guid RID/UID/GID
4955 *
4956 * outputs:
4957 * zc_cookie property value
4958 */
4959 static int
4960 zfs_ioc_userspace_one(zfs_cmd_t *zc)
4961 {
4962 zfsvfs_t *zfsvfs;
4963 int error;
4964
4965 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
4966 return (SET_ERROR(EINVAL));
4967
4968 error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE);
4969 if (error != 0)
4970 return (error);
4971
4972 error = zfs_userspace_one(zfsvfs,
4973 zc->zc_objset_type, zc->zc_value, zc->zc_guid, &zc->zc_cookie);
4974 zfsvfs_rele(zfsvfs, FTAG);
4975
4976 return (error);
4977 }
4978
4979 /*
4980 * inputs:
4981 * zc_name name of filesystem
4982 * zc_cookie zap cursor
4983 * zc_objset_type zfs_userquota_prop_t
4984 * zc_nvlist_dst[_size] buffer to fill (not really an nvlist)
4985 *
4986 * outputs:
4987 * zc_nvlist_dst[_size] data buffer (array of zfs_useracct_t)
4988 * zc_cookie zap cursor
4989 */
4990 static int
4991 zfs_ioc_userspace_many(zfs_cmd_t *zc)
4992 {
4993 zfsvfs_t *zfsvfs;
4994 int bufsize = zc->zc_nvlist_dst_size;
4995
4996 if (bufsize <= 0)
4997 return (SET_ERROR(ENOMEM));
4998
4999 int error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE);
5000 if (error != 0)
5001 return (error);
5002
5003 void *buf = kmem_alloc(bufsize, KM_SLEEP);
5004
5005 error = zfs_userspace_many(zfsvfs, zc->zc_objset_type, &zc->zc_cookie,
5006 buf, &zc->zc_nvlist_dst_size);
5007
5008 if (error == 0) {
5009 error = xcopyout(buf,
5010 (void *)(uintptr_t)zc->zc_nvlist_dst,
5011 zc->zc_nvlist_dst_size);
5012 }
5013 kmem_free(buf, bufsize);
5014 zfsvfs_rele(zfsvfs, FTAG);
5015
5016 return (error);
5017 }
5018
5019 /*
5020 * inputs:
5021 * zc_name name of filesystem
5022 *
5023 * outputs:
5024 * none
5025 */
5026 static int
5027 zfs_ioc_userspace_upgrade(zfs_cmd_t *zc)
5028 {
5029 objset_t *os;
5030 int error = 0;
5031 zfsvfs_t *zfsvfs;
5032
5033 if (getzfsvfs(zc->zc_name, &zfsvfs) == 0) {
5034 if (!dmu_objset_userused_enabled(zfsvfs->z_os)) {
5035 /*
5036 * If userused is not enabled, it may be because the
5037 * objset needs to be closed & reopened (to grow the
5038 * objset_phys_t). Suspend/resume the fs will do that.
5039 */
5040 dsl_dataset_t *ds, *newds;
5041
5042 ds = dmu_objset_ds(zfsvfs->z_os);
5043 error = zfs_suspend_fs(zfsvfs);
5044 if (error == 0) {
5045 dmu_objset_refresh_ownership(ds, &newds,
5046 zfsvfs);
5047 error = zfs_resume_fs(zfsvfs, newds);
5048 }
5049 }
5050 if (error == 0)
5051 error = dmu_objset_userspace_upgrade(zfsvfs->z_os);
5052 VFS_RELE(zfsvfs->z_vfs);
5053 } else {
5054 /* XXX kind of reading contents without owning */
5055 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
5056 if (error != 0)
5057 return (error);
5058
5059 error = dmu_objset_userspace_upgrade(os);
5060 dmu_objset_rele(os, FTAG);
5061 }
5062
5063 return (error);
5064 }
5065
5066 /*
5067 * We don't want to have a hard dependency
5068 * against some special symbols in sharefs
5069 * nfs, and smbsrv. Determine them if needed when
5070 * the first file system is shared.
5071 * Neither sharefs, nfs or smbsrv are unloadable modules.
5072 */
5073 int (*znfsexport_fs)(void *arg);
5074 int (*zshare_fs)(enum sharefs_sys_op, share_t *, uint32_t);
5075 int (*zsmbexport_fs)(void *arg, boolean_t add_share);
5076
5077 int zfs_nfsshare_inited;
5078 int zfs_smbshare_inited;
5079
5080 ddi_modhandle_t nfs_mod;
5081 ddi_modhandle_t sharefs_mod;
5082 ddi_modhandle_t smbsrv_mod;
5083 kmutex_t zfs_share_lock;
5084
5085 static int
5086 zfs_init_sharefs()
5087 {
5088 int error;
5089
5090 ASSERT(MUTEX_HELD(&zfs_share_lock));
5091 /* Both NFS and SMB shares also require sharetab support. */
5092 if (sharefs_mod == NULL && ((sharefs_mod =
5093 ddi_modopen("fs/sharefs",
5094 KRTLD_MODE_FIRST, &error)) == NULL)) {
5095 return (SET_ERROR(ENOSYS));
5096 }
5097 if (zshare_fs == NULL && ((zshare_fs =
5098 (int (*)(enum sharefs_sys_op, share_t *, uint32_t))
5099 ddi_modsym(sharefs_mod, "sharefs_impl", &error)) == NULL)) {
5100 return (SET_ERROR(ENOSYS));
5101 }
5102 return (0);
5103 }
5104
5105 static int
5106 zfs_ioc_share(zfs_cmd_t *zc)
5107 {
5108 int error;
5109 int opcode;
5110
5111 switch (zc->zc_share.z_sharetype) {
5112 case ZFS_SHARE_NFS:
5113 case ZFS_UNSHARE_NFS:
5114 if (zfs_nfsshare_inited == 0) {
5115 mutex_enter(&zfs_share_lock);
5116 if (nfs_mod == NULL && ((nfs_mod = ddi_modopen("fs/nfs",
5117 KRTLD_MODE_FIRST, &error)) == NULL)) {
5118 mutex_exit(&zfs_share_lock);
5119 return (SET_ERROR(ENOSYS));
5120 }
5121 if (znfsexport_fs == NULL &&
5122 ((znfsexport_fs = (int (*)(void *))
5123 ddi_modsym(nfs_mod,
5124 "nfs_export", &error)) == NULL)) {
5125 mutex_exit(&zfs_share_lock);
5126 return (SET_ERROR(ENOSYS));
5127 }
5128 error = zfs_init_sharefs();
5129 if (error != 0) {
5130 mutex_exit(&zfs_share_lock);
5131 return (SET_ERROR(ENOSYS));
5132 }
5133 zfs_nfsshare_inited = 1;
5134 mutex_exit(&zfs_share_lock);
5135 }
5136 break;
5137 case ZFS_SHARE_SMB:
5138 case ZFS_UNSHARE_SMB:
5139 if (zfs_smbshare_inited == 0) {
5140 mutex_enter(&zfs_share_lock);
5141 if (smbsrv_mod == NULL && ((smbsrv_mod =
5142 ddi_modopen("drv/smbsrv",
5143 KRTLD_MODE_FIRST, &error)) == NULL)) {
5144 mutex_exit(&zfs_share_lock);
5145 return (SET_ERROR(ENOSYS));
5146 }
5147 if (zsmbexport_fs == NULL && ((zsmbexport_fs =
5148 (int (*)(void *, boolean_t))ddi_modsym(smbsrv_mod,
5149 "smb_server_share", &error)) == NULL)) {
5150 mutex_exit(&zfs_share_lock);
5151 return (SET_ERROR(ENOSYS));
5152 }
5153 error = zfs_init_sharefs();
5154 if (error != 0) {
5155 mutex_exit(&zfs_share_lock);
5156 return (SET_ERROR(ENOSYS));
5157 }
5158 zfs_smbshare_inited = 1;
5159 mutex_exit(&zfs_share_lock);
5160 }
5161 break;
5162 default:
5163 return (SET_ERROR(EINVAL));
5164 }
5165
5166 switch (zc->zc_share.z_sharetype) {
5167 case ZFS_SHARE_NFS:
5168 case ZFS_UNSHARE_NFS:
5169 if (error =
5170 znfsexport_fs((void *)
5171 (uintptr_t)zc->zc_share.z_exportdata))
5172 return (error);
5173 break;
5174 case ZFS_SHARE_SMB:
5175 case ZFS_UNSHARE_SMB:
5176 if (error = zsmbexport_fs((void *)
5177 (uintptr_t)zc->zc_share.z_exportdata,
5178 zc->zc_share.z_sharetype == ZFS_SHARE_SMB ?
5179 B_TRUE: B_FALSE)) {
5180 return (error);
5181 }
5182 break;
5183 }
5184
5185 opcode = (zc->zc_share.z_sharetype == ZFS_SHARE_NFS ||
5186 zc->zc_share.z_sharetype == ZFS_SHARE_SMB) ?
5187 SHAREFS_ADD : SHAREFS_REMOVE;
5188
5189 /*
5190 * Add or remove share from sharetab
5191 */
5192 error = zshare_fs(opcode,
5193 (void *)(uintptr_t)zc->zc_share.z_sharedata,
5194 zc->zc_share.z_sharemax);
5195
5196 return (error);
5197
5198 }
5199
5200 ace_t full_access[] = {
5201 {(uid_t)-1, ACE_ALL_PERMS, ACE_EVERYONE, 0}
5202 };
5203
5204 /*
5205 * inputs:
5206 * zc_name name of containing filesystem
5207 * zc_obj object # beyond which we want next in-use object #
5208 *
5209 * outputs:
5210 * zc_obj next in-use object #
5211 */
5212 static int
5213 zfs_ioc_next_obj(zfs_cmd_t *zc)
5214 {
5215 objset_t *os = NULL;
5216 int error;
5217
5218 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
5219 if (error != 0)
5220 return (error);
5221
5222 error = dmu_object_next(os, &zc->zc_obj, B_FALSE,
5223 dsl_dataset_phys(os->os_dsl_dataset)->ds_prev_snap_txg);
5224
5225 dmu_objset_rele(os, FTAG);
5226 return (error);
5227 }
5228
5229 /*
5230 * inputs:
5231 * zc_name name of filesystem
5232 * zc_value prefix name for snapshot
5233 * zc_cleanup_fd cleanup-on-exit file descriptor for calling process
5234 *
5235 * outputs:
5236 * zc_value short name of new snapshot
5237 */
5238 static int
5239 zfs_ioc_tmp_snapshot(zfs_cmd_t *zc)
5240 {
5241 char *snap_name;
5242 char *hold_name;
5243 int error;
5244 minor_t minor;
5245
5246 error = zfs_onexit_fd_hold(zc->zc_cleanup_fd, &minor);
5247 if (error != 0)
5248 return (error);
5249
5250 snap_name = kmem_asprintf("%s-%016llx", zc->zc_value,
5251 (u_longlong_t)ddi_get_lbolt64());
5252 hold_name = kmem_asprintf("%%%s", zc->zc_value);
5253
5254 error = dsl_dataset_snapshot_tmp(zc->zc_name, snap_name, minor,
5255 hold_name);
5256 if (error == 0)
5257 (void) strcpy(zc->zc_value, snap_name);
5258 strfree(snap_name);
5259 strfree(hold_name);
5260 zfs_onexit_fd_rele(zc->zc_cleanup_fd);
5261 return (error);
5262 }
5263
5264 /*
5265 * inputs:
5266 * zc_name name of "to" snapshot
5267 * zc_value name of "from" snapshot
5268 * zc_cookie file descriptor to write diff data on
5269 *
5270 * outputs:
5271 * dmu_diff_record_t's to the file descriptor
5272 */
5273 static int
5274 zfs_ioc_diff(zfs_cmd_t *zc)
5275 {
5276 file_t *fp;
5277 offset_t off;
5278 int error;
5279
5280 fp = getf(zc->zc_cookie);
5281 if (fp == NULL)
5282 return (SET_ERROR(EBADF));
5283
5284 off = fp->f_offset;
5285
5286 error = dmu_diff(zc->zc_name, zc->zc_value, fp->f_vnode, &off);
5287
5288 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
5289 fp->f_offset = off;
5290 releasef(zc->zc_cookie);
5291
5292 return (error);
5293 }
5294
5295 /*
5296 * Remove all ACL files in shares dir
5297 */
5298 static int
5299 zfs_smb_acl_purge(znode_t *dzp)
5300 {
5301 zap_cursor_t zc;
5302 zap_attribute_t zap;
5303 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
5304 int error;
5305
5306 for (zap_cursor_init(&zc, zfsvfs->z_os, dzp->z_id);
5307 (error = zap_cursor_retrieve(&zc, &zap)) == 0;
5308 zap_cursor_advance(&zc)) {
5309 if ((error = VOP_REMOVE(ZTOV(dzp), zap.za_name, kcred,
5310 NULL, 0)) != 0)
5311 break;
5312 }
5313 zap_cursor_fini(&zc);
5314 return (error);
5315 }
5316
5317 static int
5318 zfs_ioc_smb_acl(zfs_cmd_t *zc)
5319 {
5320 vnode_t *vp;
5321 znode_t *dzp;
5322 vnode_t *resourcevp = NULL;
5323 znode_t *sharedir;
5324 zfsvfs_t *zfsvfs;
5325 nvlist_t *nvlist;
5326 char *src, *target;
5327 vattr_t vattr;
5328 vsecattr_t vsec;
5329 int error = 0;
5330
5331 if ((error = lookupname(zc->zc_value, UIO_SYSSPACE,
5332 NO_FOLLOW, NULL, &vp)) != 0)
5333 return (error);
5334
5335 /* Now make sure mntpnt and dataset are ZFS */
5336
5337 if (vp->v_vfsp->vfs_fstype != zfsfstype ||
5338 (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource),
5339 zc->zc_name) != 0)) {
5340 VN_RELE(vp);
5341 return (SET_ERROR(EINVAL));
5342 }
5343
5344 dzp = VTOZ(vp);
5345 zfsvfs = dzp->z_zfsvfs;
5346 ZFS_ENTER(zfsvfs);
5347
5348 /*
5349 * Create share dir if its missing.
5350 */
5351 mutex_enter(&zfsvfs->z_lock);
5352 if (zfsvfs->z_shares_dir == 0) {
5353 dmu_tx_t *tx;
5354
5355 tx = dmu_tx_create(zfsvfs->z_os);
5356 dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, TRUE,
5357 ZFS_SHARES_DIR);
5358 dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL);
5359 error = dmu_tx_assign(tx, TXG_WAIT);
5360 if (error != 0) {
5361 dmu_tx_abort(tx);
5362 } else {
5363 error = zfs_create_share_dir(zfsvfs, tx);
5364 dmu_tx_commit(tx);
5365 }
5366 if (error != 0) {
5367 mutex_exit(&zfsvfs->z_lock);
5368 VN_RELE(vp);
5369 ZFS_EXIT(zfsvfs);
5370 return (error);
5371 }
5372 }
5373 mutex_exit(&zfsvfs->z_lock);
5374
5375 ASSERT(zfsvfs->z_shares_dir);
5376 if ((error = zfs_zget(zfsvfs, zfsvfs->z_shares_dir, &sharedir)) != 0) {
5377 VN_RELE(vp);
5378 ZFS_EXIT(zfsvfs);
5379 return (error);
5380 }
5381
5382 switch (zc->zc_cookie) {
5383 case ZFS_SMB_ACL_ADD:
5384 vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE;
5385 vattr.va_type = VREG;
5386 vattr.va_mode = S_IFREG|0777;
5387 vattr.va_uid = 0;
5388 vattr.va_gid = 0;
5389
5390 vsec.vsa_mask = VSA_ACE;
5391 vsec.vsa_aclentp = &full_access;
5392 vsec.vsa_aclentsz = sizeof (full_access);
5393 vsec.vsa_aclcnt = 1;
5394
5395 error = VOP_CREATE(ZTOV(sharedir), zc->zc_string,
5396 &vattr, EXCL, 0, &resourcevp, kcred, 0, NULL, &vsec);
5397 if (resourcevp)
5398 VN_RELE(resourcevp);
5399 break;
5400
5401 case ZFS_SMB_ACL_REMOVE:
5402 error = VOP_REMOVE(ZTOV(sharedir), zc->zc_string, kcred,
5403 NULL, 0);
5404 break;
5405
5406 case ZFS_SMB_ACL_RENAME:
5407 if ((error = get_nvlist(zc->zc_nvlist_src,
5408 zc->zc_nvlist_src_size, zc->zc_iflags, &nvlist)) != 0) {
5409 VN_RELE(vp);
5410 VN_RELE(ZTOV(sharedir));
5411 ZFS_EXIT(zfsvfs);
5412 return (error);
5413 }
5414 if (nvlist_lookup_string(nvlist, ZFS_SMB_ACL_SRC, &src) ||
5415 nvlist_lookup_string(nvlist, ZFS_SMB_ACL_TARGET,
5416 &target)) {
5417 VN_RELE(vp);
5418 VN_RELE(ZTOV(sharedir));
5419 ZFS_EXIT(zfsvfs);
5420 nvlist_free(nvlist);
5421 return (error);
5422 }
5423 error = VOP_RENAME(ZTOV(sharedir), src, ZTOV(sharedir), target,
5424 kcred, NULL, 0);
5425 nvlist_free(nvlist);
5426 break;
5427
5428 case ZFS_SMB_ACL_PURGE:
5429 error = zfs_smb_acl_purge(sharedir);
5430 break;
5431
5432 default:
5433 error = SET_ERROR(EINVAL);
5434 break;
5435 }
5436
5437 VN_RELE(vp);
5438 VN_RELE(ZTOV(sharedir));
5439
5440 ZFS_EXIT(zfsvfs);
5441
5442 return (error);
5443 }
5444
5445 /*
5446 * innvl: {
5447 * "holds" -> { snapname -> holdname (string), ... }
5448 * (optional) "cleanup_fd" -> fd (int32)
5449 * }
5450 *
5451 * outnvl: {
5452 * snapname -> error value (int32)
5453 * ...
5454 * }
5455 */
5456 /* ARGSUSED */
5457 static int
5458 zfs_ioc_hold(const char *pool, nvlist_t *args, nvlist_t *errlist)
5459 {
5460 nvpair_t *pair;
5461 nvlist_t *holds;
5462 int cleanup_fd = -1;
5463 int error;
5464 minor_t minor = 0;
5465
5466 error = nvlist_lookup_nvlist(args, "holds", &holds);
5467 if (error != 0)
5468 return (SET_ERROR(EINVAL));
5469
5470 /* make sure the user didn't pass us any invalid (empty) tags */
5471 for (pair = nvlist_next_nvpair(holds, NULL); pair != NULL;
5472 pair = nvlist_next_nvpair(holds, pair)) {
5473 char *htag;
5474
5475 error = nvpair_value_string(pair, &htag);
5476 if (error != 0)
5477 return (SET_ERROR(error));
5478
5479 if (strlen(htag) == 0)
5480 return (SET_ERROR(EINVAL));
5481 }
5482
5483 if (nvlist_lookup_int32(args, "cleanup_fd", &cleanup_fd) == 0) {
5484 error = zfs_onexit_fd_hold(cleanup_fd, &minor);
5485 if (error != 0)
5486 return (error);
5487 }
5488
5489 error = dsl_dataset_user_hold(holds, minor, errlist);
5490 if (minor != 0)
5491 zfs_onexit_fd_rele(cleanup_fd);
5492 return (error);
5493 }
5494
5495 /*
5496 * innvl is not used.
5497 *
5498 * outnvl: {
5499 * holdname -> time added (uint64 seconds since epoch)
5500 * ...
5501 * }
5502 */
5503 /* ARGSUSED */
5504 static int
5505 zfs_ioc_get_holds(const char *snapname, nvlist_t *args, nvlist_t *outnvl)
5506 {
5507 return (dsl_dataset_get_holds(snapname, outnvl));
5508 }
5509
5510 /*
5511 * innvl: {
5512 * snapname -> { holdname, ... }
5513 * ...
5514 * }
5515 *
5516 * outnvl: {
5517 * snapname -> error value (int32)
5518 * ...
5519 * }
5520 */
5521 /* ARGSUSED */
5522 static int
5523 zfs_ioc_release(const char *pool, nvlist_t *holds, nvlist_t *errlist)
5524 {
5525 return (dsl_dataset_user_release(holds, errlist));
5526 }
5527
5528 /*
5529 * inputs:
5530 * zc_name name of new filesystem or snapshot
5531 * zc_value full name of old snapshot
5532 *
5533 * outputs:
5534 * zc_cookie space in bytes
5535 * zc_objset_type compressed space in bytes
5536 * zc_perm_action uncompressed space in bytes
5537 */
5538 static int
5539 zfs_ioc_space_written(zfs_cmd_t *zc)
5540 {
5541 int error;
5542 dsl_pool_t *dp;
5543 dsl_dataset_t *new, *old;
5544
5545 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
5546 if (error != 0)
5547 return (error);
5548 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &new);
5549 if (error != 0) {
5550 dsl_pool_rele(dp, FTAG);
5551 return (error);
5552 }
5553 error = dsl_dataset_hold(dp, zc->zc_value, FTAG, &old);
5554 if (error != 0) {
5555 dsl_dataset_rele(new, FTAG);
5556 dsl_pool_rele(dp, FTAG);
5557 return (error);
5558 }
5559
5560 error = dsl_dataset_space_written(old, new, &zc->zc_cookie,
5561 &zc->zc_objset_type, &zc->zc_perm_action);
5562 dsl_dataset_rele(old, FTAG);
5563 dsl_dataset_rele(new, FTAG);
5564 dsl_pool_rele(dp, FTAG);
5565 return (error);
5566 }
5567
5568 /*
5569 * innvl: {
5570 * "firstsnap" -> snapshot name
5571 * }
5572 *
5573 * outnvl: {
5574 * "used" -> space in bytes
5575 * "compressed" -> compressed space in bytes
5576 * "uncompressed" -> uncompressed space in bytes
5577 * }
5578 */
5579 static int
5580 zfs_ioc_space_snaps(const char *lastsnap, nvlist_t *innvl, nvlist_t *outnvl)
5581 {
5582 int error;
5583 dsl_pool_t *dp;
5584 dsl_dataset_t *new, *old;
5585 char *firstsnap;
5586 uint64_t used, comp, uncomp;
5587
5588 if (nvlist_lookup_string(innvl, "firstsnap", &firstsnap) != 0)
5589 return (SET_ERROR(EINVAL));
5590
5591 error = dsl_pool_hold(lastsnap, FTAG, &dp);
5592 if (error != 0)
5593 return (error);
5594
5595 error = dsl_dataset_hold(dp, lastsnap, FTAG, &new);
5596 if (error == 0 && !new->ds_is_snapshot) {
5597 dsl_dataset_rele(new, FTAG);
5598 error = SET_ERROR(EINVAL);
5599 }
5600 if (error != 0) {
5601 dsl_pool_rele(dp, FTAG);
5602 return (error);
5603 }
5604 error = dsl_dataset_hold(dp, firstsnap, FTAG, &old);
5605 if (error == 0 && !old->ds_is_snapshot) {
5606 dsl_dataset_rele(old, FTAG);
5607 error = SET_ERROR(EINVAL);
5608 }
5609 if (error != 0) {
5610 dsl_dataset_rele(new, FTAG);
5611 dsl_pool_rele(dp, FTAG);
5612 return (error);
5613 }
5614
5615 error = dsl_dataset_space_wouldfree(old, new, &used, &comp, &uncomp);
5616 dsl_dataset_rele(old, FTAG);
5617 dsl_dataset_rele(new, FTAG);
5618 dsl_pool_rele(dp, FTAG);
5619 fnvlist_add_uint64(outnvl, "used", used);
5620 fnvlist_add_uint64(outnvl, "compressed", comp);
5621 fnvlist_add_uint64(outnvl, "uncompressed", uncomp);
5622 return (error);
5623 }
5624
5625 /*
5626 * innvl: {
5627 * "fd" -> file descriptor to write stream to (int32)
5628 * (optional) "fromsnap" -> full snap name to send an incremental from
5629 * (optional) "largeblockok" -> (value ignored)
5630 * indicates that blocks > 128KB are permitted
5631 * (optional) "embedok" -> (value ignored)
5632 * presence indicates DRR_WRITE_EMBEDDED records are permitted
5633 * (optional) "compressok" -> (value ignored)
5634 * presence indicates compressed DRR_WRITE records are permitted
5635 * (optional) "resume_object" and "resume_offset" -> (uint64)
5636 * if present, resume send stream from specified object and offset.
5637 * }
5638 *
5639 * outnvl is unused
5640 */
5641 /* ARGSUSED */
5642 static int
5643 zfs_ioc_send_new(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl)
5644 {
5645 int error;
5646 offset_t off;
5647 char *fromname = NULL;
5648 int fd;
5649 boolean_t largeblockok;
5650 boolean_t embedok;
5651 boolean_t compressok;
5652 uint64_t resumeobj = 0;
5653 uint64_t resumeoff = 0;
5654
5655 error = nvlist_lookup_int32(innvl, "fd", &fd);
5656 if (error != 0)
5657 return (SET_ERROR(EINVAL));
5658
5659 (void) nvlist_lookup_string(innvl, "fromsnap", &fromname);
5660
5661 largeblockok = nvlist_exists(innvl, "largeblockok");
5662 embedok = nvlist_exists(innvl, "embedok");
5663 compressok = nvlist_exists(innvl, "compressok");
5664
5665 (void) nvlist_lookup_uint64(innvl, "resume_object", &resumeobj);
5666 (void) nvlist_lookup_uint64(innvl, "resume_offset", &resumeoff);
5667
5668 file_t *fp = getf(fd);
5669 if (fp == NULL)
5670 return (SET_ERROR(EBADF));
5671
5672 off = fp->f_offset;
5673 error = dmu_send(snapname, fromname, embedok, largeblockok, compressok,
5674 fd, resumeobj, resumeoff, fp->f_vnode, &off);
5675
5676 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
5677 fp->f_offset = off;
5678 releasef(fd);
5679 return (error);
5680 }
5681
5682 /*
5683 * Determine approximately how large a zfs send stream will be -- the number
5684 * of bytes that will be written to the fd supplied to zfs_ioc_send_new().
5685 *
5686 * innvl: {
5687 * (optional) "from" -> full snap or bookmark name to send an incremental
5688 * from
5689 * (optional) "largeblockok" -> (value ignored)
5690 * indicates that blocks > 128KB are permitted
5691 * (optional) "embedok" -> (value ignored)
5692 * presence indicates DRR_WRITE_EMBEDDED records are permitted
5693 * (optional) "compressok" -> (value ignored)
5694 * presence indicates compressed DRR_WRITE records are permitted
5695 * }
5696 *
5697 * outnvl: {
5698 * "space" -> bytes of space (uint64)
5699 * }
5700 */
5701 static int
5702 zfs_ioc_send_space(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl)
5703 {
5704 dsl_pool_t *dp;
5705 dsl_dataset_t *tosnap;
5706 int error;
5707 char *fromname;
5708 boolean_t compressok;
5709 uint64_t space;
5710
5711 error = dsl_pool_hold(snapname, FTAG, &dp);
5712 if (error != 0)
5713 return (error);
5714
5715 error = dsl_dataset_hold(dp, snapname, FTAG, &tosnap);
5716 if (error != 0) {
5717 dsl_pool_rele(dp, FTAG);
5718 return (error);
5719 }
5720
5721 compressok = nvlist_exists(innvl, "compressok");
5722
5723 error = nvlist_lookup_string(innvl, "from", &fromname);
5724 if (error == 0) {
5725 if (strchr(fromname, '@') != NULL) {
5726 /*
5727 * If from is a snapshot, hold it and use the more
5728 * efficient dmu_send_estimate to estimate send space
5729 * size using deadlists.
5730 */
5731 dsl_dataset_t *fromsnap;
5732 error = dsl_dataset_hold(dp, fromname, FTAG, &fromsnap);
5733 if (error != 0)
5734 goto out;
5735 error = dmu_send_estimate(tosnap, fromsnap, compressok,
5736 &space);
5737 dsl_dataset_rele(fromsnap, FTAG);
5738 } else if (strchr(fromname, '#') != NULL) {
5739 /*
5740 * If from is a bookmark, fetch the creation TXG of the
5741 * snapshot it was created from and use that to find
5742 * blocks that were born after it.
5743 */
5744 zfs_bookmark_phys_t frombm;
5745
5746 error = dsl_bookmark_lookup(dp, fromname, tosnap,
5747 &frombm);
5748 if (error != 0)
5749 goto out;
5750 error = dmu_send_estimate_from_txg(tosnap,
5751 frombm.zbm_creation_txg, compressok, &space);
5752 } else {
5753 /*
5754 * from is not properly formatted as a snapshot or
5755 * bookmark
5756 */
5757 error = SET_ERROR(EINVAL);
5758 goto out;
5759 }
5760 } else {
5761 /*
5762 * If estimating the size of a full send, use dmu_send_estimate.
5763 */
5764 error = dmu_send_estimate(tosnap, NULL, compressok, &space);
5765 }
5766
5767 fnvlist_add_uint64(outnvl, "space", space);
5768
5769 out:
5770 dsl_dataset_rele(tosnap, FTAG);
5771 dsl_pool_rele(dp, FTAG);
5772 return (error);
5773 }
5774
5775 static zfs_ioc_vec_t zfs_ioc_vec[ZFS_IOC_LAST - ZFS_IOC_FIRST];
5776
5777 static void
5778 zfs_ioctl_register_legacy(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5779 zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck,
5780 boolean_t log_history, zfs_ioc_poolcheck_t pool_check)
5781 {
5782 zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST];
5783
5784 ASSERT3U(ioc, >=, ZFS_IOC_FIRST);
5785 ASSERT3U(ioc, <, ZFS_IOC_LAST);
5786 ASSERT3P(vec->zvec_legacy_func, ==, NULL);
5787 ASSERT3P(vec->zvec_func, ==, NULL);
5788
5789 vec->zvec_legacy_func = func;
5790 vec->zvec_secpolicy = secpolicy;
5791 vec->zvec_namecheck = namecheck;
5792 vec->zvec_allow_log = log_history;
5793 vec->zvec_pool_check = pool_check;
5794 }
5795
5796 /*
5797 * See the block comment at the beginning of this file for details on
5798 * each argument to this function.
5799 */
5800 static void
5801 zfs_ioctl_register(const char *name, zfs_ioc_t ioc, zfs_ioc_func_t *func,
5802 zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck,
5803 zfs_ioc_poolcheck_t pool_check, boolean_t smush_outnvlist,
5804 boolean_t allow_log)
5805 {
5806 zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST];
5807
5808 ASSERT3U(ioc, >=, ZFS_IOC_FIRST);
5809 ASSERT3U(ioc, <, ZFS_IOC_LAST);
5810 ASSERT3P(vec->zvec_legacy_func, ==, NULL);
5811 ASSERT3P(vec->zvec_func, ==, NULL);
5812
5813 /* if we are logging, the name must be valid */
5814 ASSERT(!allow_log || namecheck != NO_NAME);
5815
5816 vec->zvec_name = name;
5817 vec->zvec_func = func;
5818 vec->zvec_secpolicy = secpolicy;
5819 vec->zvec_namecheck = namecheck;
5820 vec->zvec_pool_check = pool_check;
5821 vec->zvec_smush_outnvlist = smush_outnvlist;
5822 vec->zvec_allow_log = allow_log;
5823 }
5824
5825 static void
5826 zfs_ioctl_register_pool(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5827 zfs_secpolicy_func_t *secpolicy, boolean_t log_history,
5828 zfs_ioc_poolcheck_t pool_check)
5829 {
5830 zfs_ioctl_register_legacy(ioc, func, secpolicy,
5831 POOL_NAME, log_history, pool_check);
5832 }
5833
5834 static void
5835 zfs_ioctl_register_dataset_nolog(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5836 zfs_secpolicy_func_t *secpolicy, zfs_ioc_poolcheck_t pool_check)
5837 {
5838 zfs_ioctl_register_legacy(ioc, func, secpolicy,
5839 DATASET_NAME, B_FALSE, pool_check);
5840 }
5841
5842 static void
5843 zfs_ioctl_register_pool_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func)
5844 {
5845 zfs_ioctl_register_legacy(ioc, func, zfs_secpolicy_config,
5846 POOL_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5847 }
5848
5849 static void
5850 zfs_ioctl_register_pool_meta(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5851 zfs_secpolicy_func_t *secpolicy)
5852 {
5853 zfs_ioctl_register_legacy(ioc, func, secpolicy,
5854 NO_NAME, B_FALSE, POOL_CHECK_NONE);
5855 }
5856
5857 static void
5858 zfs_ioctl_register_dataset_read_secpolicy(zfs_ioc_t ioc,
5859 zfs_ioc_legacy_func_t *func, zfs_secpolicy_func_t *secpolicy)
5860 {
5861 zfs_ioctl_register_legacy(ioc, func, secpolicy,
5862 DATASET_NAME, B_FALSE, POOL_CHECK_SUSPENDED);
5863 }
5864
5865 static void
5866 zfs_ioctl_register_dataset_read(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func)
5867 {
5868 zfs_ioctl_register_dataset_read_secpolicy(ioc, func,
5869 zfs_secpolicy_read);
5870 }
5871
5872 static void
5873 zfs_ioctl_register_dataset_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5874 zfs_secpolicy_func_t *secpolicy)
5875 {
5876 zfs_ioctl_register_legacy(ioc, func, secpolicy,
5877 DATASET_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5878 }
5879
5880 static void
5881 zfs_ioctl_init(void)
5882 {
5883 zfs_ioctl_register("snapshot", ZFS_IOC_SNAPSHOT,
5884 zfs_ioc_snapshot, zfs_secpolicy_snapshot, POOL_NAME,
5885 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5886
5887 zfs_ioctl_register("log_history", ZFS_IOC_LOG_HISTORY,
5888 zfs_ioc_log_history, zfs_secpolicy_log_history, NO_NAME,
5889 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_FALSE);
5890
5891 zfs_ioctl_register("space_snaps", ZFS_IOC_SPACE_SNAPS,
5892 zfs_ioc_space_snaps, zfs_secpolicy_read, DATASET_NAME,
5893 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5894
5895 zfs_ioctl_register("send", ZFS_IOC_SEND_NEW,
5896 zfs_ioc_send_new, zfs_secpolicy_send_new, DATASET_NAME,
5897 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5898
5899 zfs_ioctl_register("send_space", ZFS_IOC_SEND_SPACE,
5900 zfs_ioc_send_space, zfs_secpolicy_read, DATASET_NAME,
5901 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5902
5903 zfs_ioctl_register("create", ZFS_IOC_CREATE,
5904 zfs_ioc_create, zfs_secpolicy_create_clone, DATASET_NAME,
5905 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5906
5907 zfs_ioctl_register("clone", ZFS_IOC_CLONE,
5908 zfs_ioc_clone, zfs_secpolicy_create_clone, DATASET_NAME,
5909 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5910
5911 zfs_ioctl_register("remap", ZFS_IOC_REMAP,
5912 zfs_ioc_remap, zfs_secpolicy_remap, DATASET_NAME,
5913 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_TRUE);
5914
5915 zfs_ioctl_register("destroy_snaps", ZFS_IOC_DESTROY_SNAPS,
5916 zfs_ioc_destroy_snaps, zfs_secpolicy_destroy_snaps, POOL_NAME,
5917 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5918
5919 zfs_ioctl_register("hold", ZFS_IOC_HOLD,
5920 zfs_ioc_hold, zfs_secpolicy_hold, POOL_NAME,
5921 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5922 zfs_ioctl_register("release", ZFS_IOC_RELEASE,
5923 zfs_ioc_release, zfs_secpolicy_release, POOL_NAME,
5924 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5925
5926 zfs_ioctl_register("get_holds", ZFS_IOC_GET_HOLDS,
5927 zfs_ioc_get_holds, zfs_secpolicy_read, DATASET_NAME,
5928 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5929
5930 zfs_ioctl_register("rollback", ZFS_IOC_ROLLBACK,
5931 zfs_ioc_rollback, zfs_secpolicy_rollback, DATASET_NAME,
5932 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_TRUE);
5933
5934 zfs_ioctl_register("bookmark", ZFS_IOC_BOOKMARK,
5935 zfs_ioc_bookmark, zfs_secpolicy_bookmark, POOL_NAME,
5936 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5937
5938 zfs_ioctl_register("get_bookmarks", ZFS_IOC_GET_BOOKMARKS,
5939 zfs_ioc_get_bookmarks, zfs_secpolicy_read, DATASET_NAME,
5940 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5941
5942 zfs_ioctl_register("destroy_bookmarks", ZFS_IOC_DESTROY_BOOKMARKS,
5943 zfs_ioc_destroy_bookmarks, zfs_secpolicy_destroy_bookmarks,
5944 POOL_NAME,
5945 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5946
5947 zfs_ioctl_register("channel_program", ZFS_IOC_CHANNEL_PROGRAM,
5948 zfs_ioc_channel_program, zfs_secpolicy_config,
5949 POOL_NAME, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE,
5950 B_TRUE);
5951
5952 zfs_ioctl_register("zpool_checkpoint", ZFS_IOC_POOL_CHECKPOINT,
5953 zfs_ioc_pool_checkpoint, zfs_secpolicy_config, POOL_NAME,
5954 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5955
5956 zfs_ioctl_register("zpool_discard_checkpoint",
5957 ZFS_IOC_POOL_DISCARD_CHECKPOINT, zfs_ioc_pool_discard_checkpoint,
5958 zfs_secpolicy_config, POOL_NAME,
5959 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5960
5961 zfs_ioctl_register("initialize", ZFS_IOC_POOL_INITIALIZE,
5962 zfs_ioc_pool_initialize, zfs_secpolicy_config, POOL_NAME,
5963 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5964
5965 /* IOCTLS that use the legacy function signature */
5966
5967 zfs_ioctl_register_legacy(ZFS_IOC_POOL_FREEZE, zfs_ioc_pool_freeze,
5968 zfs_secpolicy_config, NO_NAME, B_FALSE, POOL_CHECK_READONLY);
5969
5970 zfs_ioctl_register_pool(ZFS_IOC_POOL_CREATE, zfs_ioc_pool_create,
5971 zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE);
5972 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SCAN,
5973 zfs_ioc_pool_scan);
5974 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_UPGRADE,
5975 zfs_ioc_pool_upgrade);
5976 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ADD,
5977 zfs_ioc_vdev_add);
5978 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_REMOVE,
5979 zfs_ioc_vdev_remove);
5980 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SET_STATE,
5981 zfs_ioc_vdev_set_state);
5982 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ATTACH,
5983 zfs_ioc_vdev_attach);
5984 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_DETACH,
5985 zfs_ioc_vdev_detach);
5986 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETPATH,
5987 zfs_ioc_vdev_setpath);
5988 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETFRU,
5989 zfs_ioc_vdev_setfru);
5990 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SET_PROPS,
5991 zfs_ioc_pool_set_props);
5992 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SPLIT,
5993 zfs_ioc_vdev_split);
5994 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_REGUID,
5995 zfs_ioc_pool_reguid);
5996
5997 zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_CONFIGS,
5998 zfs_ioc_pool_configs, zfs_secpolicy_none);
5999 zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_TRYIMPORT,
6000 zfs_ioc_pool_tryimport, zfs_secpolicy_config);
6001 zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_FAULT,
6002 zfs_ioc_inject_fault, zfs_secpolicy_inject);
6003 zfs_ioctl_register_pool_meta(ZFS_IOC_CLEAR_FAULT,
6004 zfs_ioc_clear_fault, zfs_secpolicy_inject);
6005 zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_LIST_NEXT,
6006 zfs_ioc_inject_list_next, zfs_secpolicy_inject);
6007
6008 /*
6009 * pool destroy, and export don't log the history as part of
6010 * zfsdev_ioctl, but rather zfs_ioc_pool_export
6011 * does the logging of those commands.
6012 */
6013 zfs_ioctl_register_pool(ZFS_IOC_POOL_DESTROY, zfs_ioc_pool_destroy,
6014 zfs_secpolicy_config, B_FALSE, POOL_CHECK_NONE);
6015 zfs_ioctl_register_pool(ZFS_IOC_POOL_EXPORT, zfs_ioc_pool_export,
6016 zfs_secpolicy_config, B_FALSE, POOL_CHECK_NONE);
6017
6018 zfs_ioctl_register_pool(ZFS_IOC_POOL_STATS, zfs_ioc_pool_stats,
6019 zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE);
6020 zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_PROPS, zfs_ioc_pool_get_props,
6021 zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE);
6022
6023 zfs_ioctl_register_pool(ZFS_IOC_ERROR_LOG, zfs_ioc_error_log,
6024 zfs_secpolicy_inject, B_FALSE, POOL_CHECK_SUSPENDED);
6025 zfs_ioctl_register_pool(ZFS_IOC_DSOBJ_TO_DSNAME,
6026 zfs_ioc_dsobj_to_dsname,
6027 zfs_secpolicy_diff, B_FALSE, POOL_CHECK_SUSPENDED);
6028 zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_HISTORY,
6029 zfs_ioc_pool_get_history,
6030 zfs_secpolicy_config, B_FALSE, POOL_CHECK_SUSPENDED);
6031
6032 zfs_ioctl_register_pool(ZFS_IOC_POOL_IMPORT, zfs_ioc_pool_import,
6033 zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE);
6034
6035 zfs_ioctl_register_pool(ZFS_IOC_CLEAR, zfs_ioc_clear,
6036 zfs_secpolicy_config, B_TRUE, POOL_CHECK_READONLY);
6037 zfs_ioctl_register_pool(ZFS_IOC_POOL_REOPEN, zfs_ioc_pool_reopen,
6038 zfs_secpolicy_config, B_TRUE, POOL_CHECK_SUSPENDED);
6039
6040 zfs_ioctl_register_dataset_read(ZFS_IOC_SPACE_WRITTEN,
6041 zfs_ioc_space_written);
6042 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_RECVD_PROPS,
6043 zfs_ioc_objset_recvd_props);
6044 zfs_ioctl_register_dataset_read(ZFS_IOC_NEXT_OBJ,
6045 zfs_ioc_next_obj);
6046 zfs_ioctl_register_dataset_read(ZFS_IOC_GET_FSACL,
6047 zfs_ioc_get_fsacl);
6048 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_STATS,
6049 zfs_ioc_objset_stats);
6050 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_ZPLPROPS,
6051 zfs_ioc_objset_zplprops);
6052 zfs_ioctl_register_dataset_read(ZFS_IOC_DATASET_LIST_NEXT,
6053 zfs_ioc_dataset_list_next);
6054 zfs_ioctl_register_dataset_read(ZFS_IOC_SNAPSHOT_LIST_NEXT,
6055 zfs_ioc_snapshot_list_next);
6056 zfs_ioctl_register_dataset_read(ZFS_IOC_SEND_PROGRESS,
6057 zfs_ioc_send_progress);
6058
6059 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_DIFF,
6060 zfs_ioc_diff, zfs_secpolicy_diff);
6061 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_STATS,
6062 zfs_ioc_obj_to_stats, zfs_secpolicy_diff);
6063 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_PATH,
6064 zfs_ioc_obj_to_path, zfs_secpolicy_diff);
6065 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_ONE,
6066 zfs_ioc_userspace_one, zfs_secpolicy_userspace_one);
6067 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_MANY,
6068 zfs_ioc_userspace_many, zfs_secpolicy_userspace_many);
6069 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_SEND,
6070 zfs_ioc_send, zfs_secpolicy_send);
6071
6072 zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_PROP, zfs_ioc_set_prop,
6073 zfs_secpolicy_none);
6074 zfs_ioctl_register_dataset_modify(ZFS_IOC_DESTROY, zfs_ioc_destroy,
6075 zfs_secpolicy_destroy);
6076 zfs_ioctl_register_dataset_modify(ZFS_IOC_RENAME, zfs_ioc_rename,
6077 zfs_secpolicy_rename);
6078 zfs_ioctl_register_dataset_modify(ZFS_IOC_RECV, zfs_ioc_recv,
6079 zfs_secpolicy_recv);
6080 zfs_ioctl_register_dataset_modify(ZFS_IOC_PROMOTE, zfs_ioc_promote,
6081 zfs_secpolicy_promote);
6082 zfs_ioctl_register_dataset_modify(ZFS_IOC_INHERIT_PROP,
6083 zfs_ioc_inherit_prop, zfs_secpolicy_inherit_prop);
6084 zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_FSACL, zfs_ioc_set_fsacl,
6085 zfs_secpolicy_set_fsacl);
6086
6087 zfs_ioctl_register_dataset_nolog(ZFS_IOC_SHARE, zfs_ioc_share,
6088 zfs_secpolicy_share, POOL_CHECK_NONE);
6089 zfs_ioctl_register_dataset_nolog(ZFS_IOC_SMB_ACL, zfs_ioc_smb_acl,
6090 zfs_secpolicy_smb_acl, POOL_CHECK_NONE);
6091 zfs_ioctl_register_dataset_nolog(ZFS_IOC_USERSPACE_UPGRADE,
6092 zfs_ioc_userspace_upgrade, zfs_secpolicy_userspace_upgrade,
6093 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
6094 zfs_ioctl_register_dataset_nolog(ZFS_IOC_TMP_SNAPSHOT,
6095 zfs_ioc_tmp_snapshot, zfs_secpolicy_tmp_snapshot,
6096 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
6097 }
6098
6099 int
6100 pool_status_check(const char *name, zfs_ioc_namecheck_t type,
6101 zfs_ioc_poolcheck_t check)
6102 {
6103 spa_t *spa;
6104 int error;
6105
6106 ASSERT(type == POOL_NAME || type == DATASET_NAME);
6107
6108 if (check & POOL_CHECK_NONE)
6109 return (0);
6110
6111 error = spa_open(name, &spa, FTAG);
6112 if (error == 0) {
6113 if ((check & POOL_CHECK_SUSPENDED) && spa_suspended(spa))
6114 error = SET_ERROR(EAGAIN);
6115 else if ((check & POOL_CHECK_READONLY) && !spa_writeable(spa))
6116 error = SET_ERROR(EROFS);
6117 spa_close(spa, FTAG);
6118 }
6119 return (error);
6120 }
6121
6122 /*
6123 * Find a free minor number.
6124 */
6125 minor_t
6126 zfsdev_minor_alloc(void)
6127 {
6128 static minor_t last_minor;
6129 minor_t m;
6130
6131 ASSERT(MUTEX_HELD(&zfsdev_state_lock));
6132
6133 for (m = last_minor + 1; m != last_minor; m++) {
6134 if (m > ZFSDEV_MAX_MINOR)
6135 m = 1;
6136 if (ddi_get_soft_state(zfsdev_state, m) == NULL) {
6137 last_minor = m;
6138 return (m);
6139 }
6140 }
6141
6142 return (0);
6143 }
6144
6145 static int
6146 zfs_ctldev_init(dev_t *devp)
6147 {
6148 minor_t minor;
6149 zfs_soft_state_t *zs;
6150
6151 ASSERT(MUTEX_HELD(&zfsdev_state_lock));
6152 ASSERT(getminor(*devp) == 0);
6153
6154 minor = zfsdev_minor_alloc();
6155 if (minor == 0)
6156 return (SET_ERROR(ENXIO));
6157
6158 if (ddi_soft_state_zalloc(zfsdev_state, minor) != DDI_SUCCESS)
6159 return (SET_ERROR(EAGAIN));
6160
6161 *devp = makedevice(getemajor(*devp), minor);
6162
6163 zs = ddi_get_soft_state(zfsdev_state, minor);
6164 zs->zss_type = ZSST_CTLDEV;
6165 zfs_onexit_init((zfs_onexit_t **)&zs->zss_data);
6166
6167 return (0);
6168 }
6169
6170 static void
6171 zfs_ctldev_destroy(zfs_onexit_t *zo, minor_t minor)
6172 {
6173 ASSERT(MUTEX_HELD(&zfsdev_state_lock));
6174
6175 zfs_onexit_destroy(zo);
6176 ddi_soft_state_free(zfsdev_state, minor);
6177 }
6178
6179 void *
6180 zfsdev_get_soft_state(minor_t minor, enum zfs_soft_state_type which)
6181 {
6182 zfs_soft_state_t *zp;
6183
6184 zp = ddi_get_soft_state(zfsdev_state, minor);
6185 if (zp == NULL || zp->zss_type != which)
6186 return (NULL);
6187
6188 return (zp->zss_data);
6189 }
6190
6191 static int
6192 zfsdev_open(dev_t *devp, int flag, int otyp, cred_t *cr)
6193 {
6194 int error = 0;
6195
6196 if (getminor(*devp) != 0)
6197 return (zvol_open(devp, flag, otyp, cr));
6198
6199 /* This is the control device. Allocate a new minor if requested. */
6200 if (flag & FEXCL) {
6201 mutex_enter(&zfsdev_state_lock);
6202 error = zfs_ctldev_init(devp);
6203 mutex_exit(&zfsdev_state_lock);
6204 }
6205
6206 return (error);
6207 }
6208
6209 static int
6210 zfsdev_close(dev_t dev, int flag, int otyp, cred_t *cr)
6211 {
6212 zfs_onexit_t *zo;
6213 minor_t minor = getminor(dev);
6214
6215 if (minor == 0)
6216 return (0);
6217
6218 mutex_enter(&zfsdev_state_lock);
6219 zo = zfsdev_get_soft_state(minor, ZSST_CTLDEV);
6220 if (zo == NULL) {
6221 mutex_exit(&zfsdev_state_lock);
6222 return (zvol_close(dev, flag, otyp, cr));
6223 }
6224 zfs_ctldev_destroy(zo, minor);
6225 mutex_exit(&zfsdev_state_lock);
6226
6227 return (0);
6228 }
6229
6230 static int
6231 zfsdev_ioctl(dev_t dev, int cmd, intptr_t arg, int flag, cred_t *cr, int *rvalp)
6232 {
6233 zfs_cmd_t *zc;
6234 uint_t vecnum;
6235 int error, rc, len;
6236 minor_t minor = getminor(dev);
6237 const zfs_ioc_vec_t *vec;
6238 char *saved_poolname = NULL;
6239 nvlist_t *innvl = NULL;
6240
6241 if (minor != 0 &&
6242 zfsdev_get_soft_state(minor, ZSST_CTLDEV) == NULL)
6243 return (zvol_ioctl(dev, cmd, arg, flag, cr, rvalp));
6244
6245 vecnum = cmd - ZFS_IOC_FIRST;
6246 ASSERT3U(getmajor(dev), ==, ddi_driver_major(zfs_dip));
6247
6248 if (vecnum >= sizeof (zfs_ioc_vec) / sizeof (zfs_ioc_vec[0]))
6249 return (SET_ERROR(EINVAL));
6250 vec = &zfs_ioc_vec[vecnum];
6251
6252 zc = kmem_zalloc(sizeof (zfs_cmd_t), KM_SLEEP);
6253
6254 error = ddi_copyin((void *)arg, zc, sizeof (zfs_cmd_t), flag);
6255 if (error != 0) {
6256 error = SET_ERROR(EFAULT);
6257 goto out;
6258 }
6259
6260 zc->zc_iflags = flag & FKIOCTL;
6261 if (zc->zc_nvlist_src_size != 0) {
6262 error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
6263 zc->zc_iflags, &innvl);
6264 if (error != 0)
6265 goto out;
6266 }
6267
6268 /*
6269 * Ensure that all pool/dataset names are valid before we pass down to
6270 * the lower layers.
6271 */
6272 zc->zc_name[sizeof (zc->zc_name) - 1] = '\0';
6273 switch (vec->zvec_namecheck) {
6274 case POOL_NAME:
6275 if (pool_namecheck(zc->zc_name, NULL, NULL) != 0)
6276 error = SET_ERROR(EINVAL);
6277 else
6278 error = pool_status_check(zc->zc_name,
6279 vec->zvec_namecheck, vec->zvec_pool_check);
6280 break;
6281
6282 case DATASET_NAME:
6283 if (dataset_namecheck(zc->zc_name, NULL, NULL) != 0)
6284 error = SET_ERROR(EINVAL);
6285 else
6286 error = pool_status_check(zc->zc_name,
6287 vec->zvec_namecheck, vec->zvec_pool_check);
6288 break;
6289
6290 case NO_NAME:
6291 break;
6292 }
6293
6294
6295 if (error == 0)
6296 error = vec->zvec_secpolicy(zc, innvl, cr);
6297
6298 if (error != 0)
6299 goto out;
6300
6301 /* legacy ioctls can modify zc_name */
6302 len = strcspn(zc->zc_name, "/@#") + 1;
6303 saved_poolname = kmem_alloc(len, KM_SLEEP);
6304 (void) strlcpy(saved_poolname, zc->zc_name, len);
6305
6306 if (vec->zvec_func != NULL) {
6307 nvlist_t *outnvl;
6308 int puterror = 0;
6309 spa_t *spa;
6310 nvlist_t *lognv = NULL;
6311
6312 ASSERT(vec->zvec_legacy_func == NULL);
6313
6314 /*
6315 * Add the innvl to the lognv before calling the func,
6316 * in case the func changes the innvl.
6317 */
6318 if (vec->zvec_allow_log) {
6319 lognv = fnvlist_alloc();
6320 fnvlist_add_string(lognv, ZPOOL_HIST_IOCTL,
6321 vec->zvec_name);
6322 if (!nvlist_empty(innvl)) {
6323 fnvlist_add_nvlist(lognv, ZPOOL_HIST_INPUT_NVL,
6324 innvl);
6325 }
6326 }
6327
6328 outnvl = fnvlist_alloc();
6329 error = vec->zvec_func(zc->zc_name, innvl, outnvl);
6330
6331 /*
6332 * Some commands can partially execute, modfiy state, and still
6333 * return an error. In these cases, attempt to record what
6334 * was modified.
6335 */
6336 if ((error == 0 ||
6337 (cmd == ZFS_IOC_CHANNEL_PROGRAM && error != EINVAL)) &&
6338 vec->zvec_allow_log &&
6339 spa_open(zc->zc_name, &spa, FTAG) == 0) {
6340 if (!nvlist_empty(outnvl)) {
6341 fnvlist_add_nvlist(lognv, ZPOOL_HIST_OUTPUT_NVL,
6342 outnvl);
6343 }
6344 if (error != 0) {
6345 fnvlist_add_int64(lognv, ZPOOL_HIST_ERRNO,
6346 error);
6347 }
6348 (void) spa_history_log_nvl(spa, lognv);
6349 spa_close(spa, FTAG);
6350 }
6351 fnvlist_free(lognv);
6352
6353 if (!nvlist_empty(outnvl) || zc->zc_nvlist_dst_size != 0) {
6354 int smusherror = 0;
6355 if (vec->zvec_smush_outnvlist) {
6356 smusherror = nvlist_smush(outnvl,
6357 zc->zc_nvlist_dst_size);
6358 }
6359 if (smusherror == 0)
6360 puterror = put_nvlist(zc, outnvl);
6361 }
6362
6363 if (puterror != 0)
6364 error = puterror;
6365
6366 nvlist_free(outnvl);
6367 } else {
6368 error = vec->zvec_legacy_func(zc);
6369 }
6370
6371 out:
6372 nvlist_free(innvl);
6373 rc = ddi_copyout(zc, (void *)arg, sizeof (zfs_cmd_t), flag);
6374 if (error == 0 && rc != 0)
6375 error = SET_ERROR(EFAULT);
6376 if (error == 0 && vec->zvec_allow_log) {
6377 char *s = tsd_get(zfs_allow_log_key);
6378 if (s != NULL)
6379 strfree(s);
6380 (void) tsd_set(zfs_allow_log_key, saved_poolname);
6381 } else {
6382 if (saved_poolname != NULL)
6383 strfree(saved_poolname);
6384 }
6385
6386 kmem_free(zc, sizeof (zfs_cmd_t));
6387 return (error);
6388 }
6389
6390 static int
6391 zfs_attach(dev_info_t *dip, ddi_attach_cmd_t cmd)
6392 {
6393 if (cmd != DDI_ATTACH)
6394 return (DDI_FAILURE);
6395
6396 if (ddi_create_minor_node(dip, "zfs", S_IFCHR, 0,
6397 DDI_PSEUDO, 0) == DDI_FAILURE)
6398 return (DDI_FAILURE);
6399
6400 zfs_dip = dip;
6401
6402 ddi_report_dev(dip);
6403
6404 return (DDI_SUCCESS);
6405 }
6406
6407 static int
6408 zfs_detach(dev_info_t *dip, ddi_detach_cmd_t cmd)
6409 {
6410 if (spa_busy() || zfs_busy() || zvol_busy())
6411 return (DDI_FAILURE);
6412
6413 if (cmd != DDI_DETACH)
6414 return (DDI_FAILURE);
6415
6416 zfs_dip = NULL;
6417
6418 ddi_prop_remove_all(dip);
6419 ddi_remove_minor_node(dip, NULL);
6420
6421 return (DDI_SUCCESS);
6422 }
6423
6424 /*ARGSUSED*/
6425 static int
6426 zfs_info(dev_info_t *dip, ddi_info_cmd_t infocmd, void *arg, void **result)
6427 {
6428 switch (infocmd) {
6429 case DDI_INFO_DEVT2DEVINFO:
6430 *result = zfs_dip;
6431 return (DDI_SUCCESS);
6432
6433 case DDI_INFO_DEVT2INSTANCE:
6434 *result = (void *)0;
6435 return (DDI_SUCCESS);
6436 }
6437
6438 return (DDI_FAILURE);
6439 }
6440
6441 /*
6442 * OK, so this is a little weird.
6443 *
6444 * /dev/zfs is the control node, i.e. minor 0.
6445 * /dev/zvol/[r]dsk/pool/dataset are the zvols, minor > 0.
6446 *
6447 * /dev/zfs has basically nothing to do except serve up ioctls,
6448 * so most of the standard driver entry points are in zvol.c.
6449 */
6450 static struct cb_ops zfs_cb_ops = {
6451 zfsdev_open, /* open */
6452 zfsdev_close, /* close */
6453 zvol_strategy, /* strategy */
6454 nodev, /* print */
6455 zvol_dump, /* dump */
6456 zvol_read, /* read */
6457 zvol_write, /* write */
6458 zfsdev_ioctl, /* ioctl */
6459 nodev, /* devmap */
6460 nodev, /* mmap */
6461 nodev, /* segmap */
6462 nochpoll, /* poll */
6463 ddi_prop_op, /* prop_op */
6464 NULL, /* streamtab */
6465 D_NEW | D_MP | D_64BIT, /* Driver compatibility flag */
6466 CB_REV, /* version */
6467 nodev, /* async read */
6468 nodev, /* async write */
6469 };
6470
6471 static struct dev_ops zfs_dev_ops = {
6472 DEVO_REV, /* version */
6473 0, /* refcnt */
6474 zfs_info, /* info */
6475 nulldev, /* identify */
6476 nulldev, /* probe */
6477 zfs_attach, /* attach */
6478 zfs_detach, /* detach */
6479 nodev, /* reset */
6480 &zfs_cb_ops, /* driver operations */
6481 NULL, /* no bus operations */
6482 NULL, /* power */
6483 ddi_quiesce_not_needed, /* quiesce */
6484 };
6485
6486 static struct modldrv zfs_modldrv = {
6487 &mod_driverops,
6488 "ZFS storage pool",
6489 &zfs_dev_ops
6490 };
6491
6492 static struct modlinkage modlinkage = {
6493 MODREV_1,
6494 (void *)&zfs_modlfs,
6495 (void *)&zfs_modldrv,
6496 NULL
6497 };
6498
6499 static void
6500 zfs_allow_log_destroy(void *arg)
6501 {
6502 char *poolname = arg;
6503 strfree(poolname);
6504 }
6505
6506 int
6507 _init(void)
6508 {
6509 int error;
6510
6511 spa_init(FREAD | FWRITE);
6512 zfs_init();
6513 zvol_init();
6514 zfs_ioctl_init();
6515
6516 if ((error = mod_install(&modlinkage)) != 0) {
6517 zvol_fini();
6518 zfs_fini();
6519 spa_fini();
6520 return (error);
6521 }
6522
6523 tsd_create(&zfs_fsyncer_key, NULL);
6524 tsd_create(&rrw_tsd_key, rrw_tsd_destroy);
6525 tsd_create(&zfs_allow_log_key, zfs_allow_log_destroy);
6526
6527 error = ldi_ident_from_mod(&modlinkage, &zfs_li);
6528 ASSERT(error == 0);
6529 mutex_init(&zfs_share_lock, NULL, MUTEX_DEFAULT, NULL);
6530
6531 return (0);
6532 }
6533
6534 int
6535 _fini(void)
6536 {
6537 int error;
6538
6539 if (spa_busy() || zfs_busy() || zvol_busy() || zio_injection_enabled)
6540 return (SET_ERROR(EBUSY));
6541
6542 if ((error = mod_remove(&modlinkage)) != 0)
6543 return (error);
6544
6545 zvol_fini();
6546 zfs_fini();
6547 spa_fini();
6548 if (zfs_nfsshare_inited)
6549 (void) ddi_modclose(nfs_mod);
6550 if (zfs_smbshare_inited)
6551 (void) ddi_modclose(smbsrv_mod);
6552 if (zfs_nfsshare_inited || zfs_smbshare_inited)
6553 (void) ddi_modclose(sharefs_mod);
6554
6555 tsd_destroy(&zfs_fsyncer_key);
6556 ldi_ident_release(zfs_li);
6557 zfs_li = NULL;
6558 mutex_destroy(&zfs_share_lock);
6559
6560 return (error);
6561 }
6562
6563 int
6564 _info(struct modinfo *modinfop)
6565 {
6566 return (mod_info(&modlinkage, modinfop));
6567 }