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 * Portions Copyright 2011 Martin Matuska
25 * Copyright 2011 Nexenta Systems, Inc. All rights reserved.
26 * Copyright (c) 2012, Joyent, Inc. All rights reserved.
27 * Copyright (c) 2012 by Delphix. All rights reserved.
28 * Copyright (c) 2013 by Saso Kiselkov. All rights reserved.
29 */
30
31 /*
32 * ZFS ioctls.
33 *
34 * This file handles the ioctls to /dev/zfs, used for configuring ZFS storage
35 * pools and filesystems, e.g. with /sbin/zfs and /sbin/zpool.
36 *
37 * There are two ways that we handle ioctls: the legacy way where almost
38 * all of the logic is in the ioctl callback, and the new way where most
39 * of the marshalling is handled in the common entry point, zfsdev_ioctl().
40 *
41 * Non-legacy ioctls should be registered by calling
42 * zfs_ioctl_register() from zfs_ioctl_init(). The ioctl is invoked
43 * from userland by lzc_ioctl().
44 *
45 * The registration arguments are as follows:
46 *
47 * const char *name
48 * The name of the ioctl. This is used for history logging. If the
49 * ioctl returns successfully (the callback returns 0), and allow_log
50 * is true, then a history log entry will be recorded with the input &
51 * output nvlists. The log entry can be printed with "zpool history -i".
52 *
53 * zfs_ioc_t ioc
54 * The ioctl request number, which userland will pass to ioctl(2).
55 * The ioctl numbers can change from release to release, because
56 * the caller (libzfs) must be matched to the kernel.
57 *
58 * zfs_secpolicy_func_t *secpolicy
59 * This function will be called before the zfs_ioc_func_t, to
60 * determine if this operation is permitted. It should return EPERM
61 * on failure, and 0 on success. Checks include determining if the
62 * dataset is visible in this zone, and if the user has either all
63 * zfs privileges in the zone (SYS_MOUNT), or has been granted permission
64 * to do this operation on this dataset with "zfs allow".
65 *
66 * zfs_ioc_namecheck_t namecheck
67 * This specifies what to expect in the zfs_cmd_t:zc_name -- a pool
68 * name, a dataset name, or nothing. If the name is not well-formed,
69 * the ioctl will fail and the callback will not be called.
70 * Therefore, the callback can assume that the name is well-formed
71 * (e.g. is null-terminated, doesn't have more than one '@' character,
72 * doesn't have invalid characters).
73 *
74 * zfs_ioc_poolcheck_t pool_check
75 * This specifies requirements on the pool state. If the pool does
76 * not meet them (is suspended or is readonly), the ioctl will fail
77 * and the callback will not be called. If any checks are specified
78 * (i.e. it is not POOL_CHECK_NONE), namecheck must not be NO_NAME.
79 * Multiple checks can be or-ed together (e.g. POOL_CHECK_SUSPENDED |
80 * POOL_CHECK_READONLY).
81 *
82 * boolean_t smush_outnvlist
83 * If smush_outnvlist is true, then the output is presumed to be a
84 * list of errors, and it will be "smushed" down to fit into the
85 * caller's buffer, by removing some entries and replacing them with a
86 * single "N_MORE_ERRORS" entry indicating how many were removed. See
87 * nvlist_smush() for details. If smush_outnvlist is false, and the
88 * outnvlist does not fit into the userland-provided buffer, then the
89 * ioctl will fail with ENOMEM.
90 *
91 * zfs_ioc_func_t *func
92 * The callback function that will perform the operation.
93 *
94 * The callback should return 0 on success, or an error number on
95 * failure. If the function fails, the userland ioctl will return -1,
96 * and errno will be set to the callback's return value. The callback
97 * will be called with the following arguments:
98 *
99 * const char *name
100 * The name of the pool or dataset to operate on, from
101 * zfs_cmd_t:zc_name. The 'namecheck' argument specifies the
102 * expected type (pool, dataset, or none).
103 *
104 * nvlist_t *innvl
105 * The input nvlist, deserialized from zfs_cmd_t:zc_nvlist_src. Or
106 * NULL if no input nvlist was provided. Changes to this nvlist are
107 * ignored. If the input nvlist could not be deserialized, the
108 * ioctl will fail and the callback will not be called.
109 *
110 * nvlist_t *outnvl
111 * The output nvlist, initially empty. The callback can fill it in,
112 * and it will be returned to userland by serializing it into
113 * zfs_cmd_t:zc_nvlist_dst. If it is non-empty, and serialization
114 * fails (e.g. because the caller didn't supply a large enough
115 * buffer), then the overall ioctl will fail. See the
116 * 'smush_nvlist' argument above for additional behaviors.
117 *
118 * There are two typical uses of the output nvlist:
119 * - To return state, e.g. property values. In this case,
120 * smush_outnvlist should be false. If the buffer was not large
121 * enough, the caller will reallocate a larger buffer and try
122 * the ioctl again.
123 *
124 * - To return multiple errors from an ioctl which makes on-disk
125 * changes. In this case, smush_outnvlist should be true.
126 * Ioctls which make on-disk modifications should generally not
127 * use the outnvl if they succeed, because the caller can not
128 * distinguish between the operation failing, and
129 * deserialization failing.
130 */
131
132 #include <sys/types.h>
133 #include <sys/param.h>
134 #include <sys/errno.h>
135 #include <sys/uio.h>
136 #include <sys/buf.h>
137 #include <sys/modctl.h>
138 #include <sys/open.h>
139 #include <sys/file.h>
140 #include <sys/kmem.h>
141 #include <sys/conf.h>
142 #include <sys/cmn_err.h>
143 #include <sys/stat.h>
144 #include <sys/zfs_ioctl.h>
145 #include <sys/zfs_vfsops.h>
146 #include <sys/zfs_znode.h>
147 #include <sys/zap.h>
148 #include <sys/spa.h>
149 #include <sys/spa_impl.h>
150 #include <sys/vdev.h>
151 #include <sys/priv_impl.h>
152 #include <sys/dmu.h>
153 #include <sys/dsl_dir.h>
154 #include <sys/dsl_dataset.h>
155 #include <sys/dsl_prop.h>
156 #include <sys/dsl_deleg.h>
157 #include <sys/dmu_objset.h>
158 #include <sys/dmu_impl.h>
159 #include <sys/ddi.h>
160 #include <sys/sunddi.h>
161 #include <sys/sunldi.h>
162 #include <sys/policy.h>
163 #include <sys/zone.h>
164 #include <sys/nvpair.h>
165 #include <sys/pathname.h>
166 #include <sys/mount.h>
167 #include <sys/sdt.h>
168 #include <sys/fs/zfs.h>
169 #include <sys/zfs_ctldir.h>
170 #include <sys/zfs_dir.h>
171 #include <sys/zfs_onexit.h>
172 #include <sys/zvol.h>
173 #include <sys/dsl_scan.h>
174 #include <sharefs/share.h>
175 #include <sys/dmu_objset.h>
176 #include <sys/zfeature.h>
177
178 #include "zfs_namecheck.h"
179 #include "zfs_prop.h"
180 #include "zfs_deleg.h"
181 #include "zfs_comutil.h"
182
183 extern struct modlfs zfs_modlfs;
184
185 extern void zfs_init(void);
186 extern void zfs_fini(void);
187
188 ldi_ident_t zfs_li = NULL;
189 dev_info_t *zfs_dip;
190
191 uint_t zfs_fsyncer_key;
192 extern uint_t rrw_tsd_key;
193 static uint_t zfs_allow_log_key;
194
195 typedef int zfs_ioc_legacy_func_t(zfs_cmd_t *);
196 typedef int zfs_ioc_func_t(const char *, nvlist_t *, nvlist_t *);
197 typedef int zfs_secpolicy_func_t(zfs_cmd_t *, nvlist_t *, cred_t *);
198
199 typedef enum {
200 NO_NAME,
201 POOL_NAME,
202 DATASET_NAME
203 } zfs_ioc_namecheck_t;
204
205 typedef enum {
206 POOL_CHECK_NONE = 1 << 0,
207 POOL_CHECK_SUSPENDED = 1 << 1,
208 POOL_CHECK_READONLY = 1 << 2,
209 } zfs_ioc_poolcheck_t;
210
211 typedef struct zfs_ioc_vec {
212 zfs_ioc_legacy_func_t *zvec_legacy_func;
213 zfs_ioc_func_t *zvec_func;
214 zfs_secpolicy_func_t *zvec_secpolicy;
215 zfs_ioc_namecheck_t zvec_namecheck;
216 boolean_t zvec_allow_log;
217 zfs_ioc_poolcheck_t zvec_pool_check;
218 boolean_t zvec_smush_outnvlist;
219 const char *zvec_name;
220 } zfs_ioc_vec_t;
221
222 /* This array is indexed by zfs_userquota_prop_t */
223 static const char *userquota_perms[] = {
224 ZFS_DELEG_PERM_USERUSED,
225 ZFS_DELEG_PERM_USERQUOTA,
226 ZFS_DELEG_PERM_GROUPUSED,
227 ZFS_DELEG_PERM_GROUPQUOTA,
228 };
229
230 static int zfs_ioc_userspace_upgrade(zfs_cmd_t *zc);
231 static int zfs_check_settable(const char *name, nvpair_t *property,
232 cred_t *cr);
233 static int zfs_check_clearable(char *dataset, nvlist_t *props,
234 nvlist_t **errors);
235 static int zfs_fill_zplprops_root(uint64_t, nvlist_t *, nvlist_t *,
236 boolean_t *);
237 int zfs_set_prop_nvlist(const char *, zprop_source_t, nvlist_t *, nvlist_t *);
238 static int get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp);
239
240 static int zfs_prop_activate_feature(dsl_pool_t *dp, zfeature_info_t *feature);
241 static int zfs_prop_activate_feature_check(void *arg1, void *arg2,
242 dmu_tx_t *tx);
243 static void zfs_prop_activate_feature_sync(void *arg1, void *arg2,
244 dmu_tx_t *tx);
245
246 /* _NOTE(PRINTFLIKE(4)) - this is printf-like, but lint is too whiney */
247 void
248 __dprintf(const char *file, const char *func, int line, const char *fmt, ...)
249 {
250 const char *newfile;
251 char buf[512];
252 va_list adx;
253
254 /*
255 * Get rid of annoying "../common/" prefix to filename.
256 */
257 newfile = strrchr(file, '/');
258 if (newfile != NULL) {
259 newfile = newfile + 1; /* Get rid of leading / */
260 } else {
261 newfile = file;
262 }
263
264 va_start(adx, fmt);
265 (void) vsnprintf(buf, sizeof (buf), fmt, adx);
266 va_end(adx);
267
268 /*
269 * To get this data, use the zfs-dprintf probe as so:
270 * dtrace -q -n 'zfs-dprintf \
271 * /stringof(arg0) == "dbuf.c"/ \
272 * {printf("%s: %s", stringof(arg1), stringof(arg3))}'
273 * arg0 = file name
274 * arg1 = function name
275 * arg2 = line number
276 * arg3 = message
277 */
278 DTRACE_PROBE4(zfs__dprintf,
279 char *, newfile, char *, func, int, line, char *, buf);
280 }
281
282 static void
283 history_str_free(char *buf)
284 {
285 kmem_free(buf, HIS_MAX_RECORD_LEN);
286 }
287
288 static char *
289 history_str_get(zfs_cmd_t *zc)
290 {
291 char *buf;
292
293 if (zc->zc_history == NULL)
294 return (NULL);
295
296 buf = kmem_alloc(HIS_MAX_RECORD_LEN, KM_SLEEP);
297 if (copyinstr((void *)(uintptr_t)zc->zc_history,
298 buf, HIS_MAX_RECORD_LEN, NULL) != 0) {
299 history_str_free(buf);
300 return (NULL);
301 }
302
303 buf[HIS_MAX_RECORD_LEN -1] = '\0';
304
305 return (buf);
306 }
307
308 /*
309 * Check to see if the named dataset is currently defined as bootable
310 */
311 static boolean_t
312 zfs_is_bootfs(const char *name)
313 {
314 objset_t *os;
315
316 if (dmu_objset_hold(name, FTAG, &os) == 0) {
317 boolean_t ret;
318 ret = (dmu_objset_id(os) == spa_bootfs(dmu_objset_spa(os)));
319 dmu_objset_rele(os, FTAG);
320 return (ret);
321 }
322 return (B_FALSE);
323 }
324
325 /*
326 * zfs_earlier_version
327 *
328 * Return non-zero if the spa version is less than requested version.
329 */
330 static int
331 zfs_earlier_version(const char *name, int version)
332 {
333 spa_t *spa;
334
335 if (spa_open(name, &spa, FTAG) == 0) {
336 if (spa_version(spa) < version) {
337 spa_close(spa, FTAG);
338 return (1);
339 }
340 spa_close(spa, FTAG);
341 }
342 return (0);
343 }
344
345 /*
346 * zpl_earlier_version
347 *
348 * Return TRUE if the ZPL version is less than requested version.
349 */
350 static boolean_t
351 zpl_earlier_version(const char *name, int version)
352 {
353 objset_t *os;
354 boolean_t rc = B_TRUE;
355
356 if (dmu_objset_hold(name, FTAG, &os) == 0) {
357 uint64_t zplversion;
358
359 if (dmu_objset_type(os) != DMU_OST_ZFS) {
360 dmu_objset_rele(os, FTAG);
361 return (B_TRUE);
362 }
363 /* XXX reading from non-owned objset */
364 if (zfs_get_zplprop(os, ZFS_PROP_VERSION, &zplversion) == 0)
365 rc = zplversion < version;
366 dmu_objset_rele(os, FTAG);
367 }
368 return (rc);
369 }
370
371 static void
372 zfs_log_history(zfs_cmd_t *zc)
373 {
374 spa_t *spa;
375 char *buf;
376
377 if ((buf = history_str_get(zc)) == NULL)
378 return;
379
380 if (spa_open(zc->zc_name, &spa, FTAG) == 0) {
381 if (spa_version(spa) >= SPA_VERSION_ZPOOL_HISTORY)
382 (void) spa_history_log(spa, buf);
383 spa_close(spa, FTAG);
384 }
385 history_str_free(buf);
386 }
387
388 /*
389 * Policy for top-level read operations (list pools). Requires no privileges,
390 * and can be used in the local zone, as there is no associated dataset.
391 */
392 /* ARGSUSED */
393 static int
394 zfs_secpolicy_none(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
395 {
396 return (0);
397 }
398
399 /*
400 * Policy for dataset read operations (list children, get statistics). Requires
401 * no privileges, but must be visible in the local zone.
402 */
403 /* ARGSUSED */
404 static int
405 zfs_secpolicy_read(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
406 {
407 if (INGLOBALZONE(curproc) ||
408 zone_dataset_visible(zc->zc_name, NULL))
409 return (0);
410
411 return (ENOENT);
412 }
413
414 static int
415 zfs_dozonecheck_impl(const char *dataset, uint64_t zoned, cred_t *cr)
416 {
417 int writable = 1;
418
419 /*
420 * The dataset must be visible by this zone -- check this first
421 * so they don't see EPERM on something they shouldn't know about.
422 */
423 if (!INGLOBALZONE(curproc) &&
424 !zone_dataset_visible(dataset, &writable))
425 return (ENOENT);
426
427 if (INGLOBALZONE(curproc)) {
428 /*
429 * If the fs is zoned, only root can access it from the
430 * global zone.
431 */
432 if (secpolicy_zfs(cr) && zoned)
433 return (EPERM);
434 } else {
435 /*
436 * If we are in a local zone, the 'zoned' property must be set.
437 */
438 if (!zoned)
439 return (EPERM);
440
441 /* must be writable by this zone */
442 if (!writable)
443 return (EPERM);
444 }
445 return (0);
446 }
447
448 static int
449 zfs_dozonecheck(const char *dataset, cred_t *cr)
450 {
451 uint64_t zoned;
452
453 if (dsl_prop_get_integer(dataset, "zoned", &zoned, NULL))
454 return (ENOENT);
455
456 return (zfs_dozonecheck_impl(dataset, zoned, cr));
457 }
458
459 static int
460 zfs_dozonecheck_ds(const char *dataset, dsl_dataset_t *ds, cred_t *cr)
461 {
462 uint64_t zoned;
463
464 rw_enter(&ds->ds_dir->dd_pool->dp_config_rwlock, RW_READER);
465 if (dsl_prop_get_ds(ds, "zoned", 8, 1, &zoned, NULL)) {
466 rw_exit(&ds->ds_dir->dd_pool->dp_config_rwlock);
467 return (ENOENT);
468 }
469 rw_exit(&ds->ds_dir->dd_pool->dp_config_rwlock);
470
471 return (zfs_dozonecheck_impl(dataset, zoned, cr));
472 }
473
474 static int
475 zfs_secpolicy_write_perms(const char *name, const char *perm, cred_t *cr)
476 {
477 int error;
478 dsl_dataset_t *ds;
479
480 error = dsl_dataset_hold(name, FTAG, &ds);
481 if (error != 0)
482 return (error);
483
484 error = zfs_dozonecheck_ds(name, ds, cr);
485 if (error == 0) {
486 error = secpolicy_zfs(cr);
487 if (error)
488 error = dsl_deleg_access_impl(ds, perm, cr);
489 }
490
491 dsl_dataset_rele(ds, FTAG);
492 return (error);
493 }
494
495 static int
496 zfs_secpolicy_write_perms_ds(const char *name, dsl_dataset_t *ds,
497 const char *perm, cred_t *cr)
498 {
499 int error;
500
501 error = zfs_dozonecheck_ds(name, ds, cr);
502 if (error == 0) {
503 error = secpolicy_zfs(cr);
504 if (error)
505 error = dsl_deleg_access_impl(ds, perm, cr);
506 }
507 return (error);
508 }
509
510 /*
511 * Policy for setting the security label property.
512 *
513 * Returns 0 for success, non-zero for access and other errors.
514 */
515 static int
516 zfs_set_slabel_policy(const char *name, char *strval, cred_t *cr)
517 {
518 char ds_hexsl[MAXNAMELEN];
519 bslabel_t ds_sl, new_sl;
520 boolean_t new_default = FALSE;
521 uint64_t zoned;
522 int needed_priv = -1;
523 int error;
524
525 /* First get the existing dataset label. */
526 error = dsl_prop_get(name, zfs_prop_to_name(ZFS_PROP_MLSLABEL),
527 1, sizeof (ds_hexsl), &ds_hexsl, NULL);
528 if (error)
529 return (EPERM);
530
531 if (strcasecmp(strval, ZFS_MLSLABEL_DEFAULT) == 0)
532 new_default = TRUE;
533
534 /* The label must be translatable */
535 if (!new_default && (hexstr_to_label(strval, &new_sl) != 0))
536 return (EINVAL);
537
538 /*
539 * In a non-global zone, disallow attempts to set a label that
540 * doesn't match that of the zone; otherwise no other checks
541 * are needed.
542 */
543 if (!INGLOBALZONE(curproc)) {
544 if (new_default || !blequal(&new_sl, CR_SL(CRED())))
545 return (EPERM);
546 return (0);
547 }
548
549 /*
550 * For global-zone datasets (i.e., those whose zoned property is
551 * "off", verify that the specified new label is valid for the
552 * global zone.
553 */
554 if (dsl_prop_get_integer(name,
555 zfs_prop_to_name(ZFS_PROP_ZONED), &zoned, NULL))
556 return (EPERM);
557 if (!zoned) {
558 if (zfs_check_global_label(name, strval) != 0)
559 return (EPERM);
560 }
561
562 /*
563 * If the existing dataset label is nondefault, check if the
564 * dataset is mounted (label cannot be changed while mounted).
565 * Get the zfsvfs; if there isn't one, then the dataset isn't
566 * mounted (or isn't a dataset, doesn't exist, ...).
567 */
568 if (strcasecmp(ds_hexsl, ZFS_MLSLABEL_DEFAULT) != 0) {
569 objset_t *os;
570 static char *setsl_tag = "setsl_tag";
571
572 /*
573 * Try to own the dataset; abort if there is any error,
574 * (e.g., already mounted, in use, or other error).
575 */
576 error = dmu_objset_own(name, DMU_OST_ZFS, B_TRUE,
577 setsl_tag, &os);
578 if (error)
579 return (EPERM);
580
581 dmu_objset_disown(os, setsl_tag);
582
583 if (new_default) {
584 needed_priv = PRIV_FILE_DOWNGRADE_SL;
585 goto out_check;
586 }
587
588 if (hexstr_to_label(strval, &new_sl) != 0)
589 return (EPERM);
590
591 if (blstrictdom(&ds_sl, &new_sl))
592 needed_priv = PRIV_FILE_DOWNGRADE_SL;
593 else if (blstrictdom(&new_sl, &ds_sl))
594 needed_priv = PRIV_FILE_UPGRADE_SL;
595 } else {
596 /* dataset currently has a default label */
597 if (!new_default)
598 needed_priv = PRIV_FILE_UPGRADE_SL;
599 }
600
601 out_check:
602 if (needed_priv != -1)
603 return (PRIV_POLICY(cr, needed_priv, B_FALSE, EPERM, NULL));
604 return (0);
605 }
606
607 static int
608 zfs_secpolicy_setprop(const char *dsname, zfs_prop_t prop, nvpair_t *propval,
609 cred_t *cr)
610 {
611 char *strval;
612
613 /*
614 * Check permissions for special properties.
615 */
616 switch (prop) {
617 case ZFS_PROP_ZONED:
618 /*
619 * Disallow setting of 'zoned' from within a local zone.
620 */
621 if (!INGLOBALZONE(curproc))
622 return (EPERM);
623 break;
624
625 case ZFS_PROP_QUOTA:
626 if (!INGLOBALZONE(curproc)) {
627 uint64_t zoned;
628 char setpoint[MAXNAMELEN];
629 /*
630 * Unprivileged users are allowed to modify the
631 * quota on things *under* (ie. contained by)
632 * the thing they own.
633 */
634 if (dsl_prop_get_integer(dsname, "zoned", &zoned,
635 setpoint))
636 return (EPERM);
637 if (!zoned || strlen(dsname) <= strlen(setpoint))
638 return (EPERM);
639 }
640 break;
641
642 case ZFS_PROP_MLSLABEL:
643 if (!is_system_labeled())
644 return (EPERM);
645
646 if (nvpair_value_string(propval, &strval) == 0) {
647 int err;
648
649 err = zfs_set_slabel_policy(dsname, strval, CRED());
650 if (err != 0)
651 return (err);
652 }
653 break;
654 }
655
656 return (zfs_secpolicy_write_perms(dsname, zfs_prop_to_name(prop), cr));
657 }
658
659 /* ARGSUSED */
660 static int
661 zfs_secpolicy_set_fsacl(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
662 {
663 int error;
664
665 error = zfs_dozonecheck(zc->zc_name, cr);
666 if (error)
667 return (error);
668
669 /*
670 * permission to set permissions will be evaluated later in
671 * dsl_deleg_can_allow()
672 */
673 return (0);
674 }
675
676 /* ARGSUSED */
677 static int
678 zfs_secpolicy_rollback(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
679 {
680 return (zfs_secpolicy_write_perms(zc->zc_name,
681 ZFS_DELEG_PERM_ROLLBACK, cr));
682 }
683
684 /* ARGSUSED */
685 static int
686 zfs_secpolicy_send(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
687 {
688 spa_t *spa;
689 dsl_pool_t *dp;
690 dsl_dataset_t *ds;
691 char *cp;
692 int error;
693
694 /*
695 * Generate the current snapshot name from the given objsetid, then
696 * use that name for the secpolicy/zone checks.
697 */
698 cp = strchr(zc->zc_name, '@');
699 if (cp == NULL)
700 return (EINVAL);
701 error = spa_open(zc->zc_name, &spa, FTAG);
702 if (error)
703 return (error);
704
705 dp = spa_get_dsl(spa);
706 rw_enter(&dp->dp_config_rwlock, RW_READER);
707 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &ds);
708 rw_exit(&dp->dp_config_rwlock);
709 spa_close(spa, FTAG);
710 if (error)
711 return (error);
712
713 dsl_dataset_name(ds, zc->zc_name);
714
715 error = zfs_secpolicy_write_perms_ds(zc->zc_name, ds,
716 ZFS_DELEG_PERM_SEND, cr);
717 dsl_dataset_rele(ds, FTAG);
718
719 return (error);
720 }
721
722 /* ARGSUSED */
723 static int
724 zfs_secpolicy_send_new(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
725 {
726 return (zfs_secpolicy_write_perms(zc->zc_name,
727 ZFS_DELEG_PERM_SEND, cr));
728 }
729
730 /* ARGSUSED */
731 static int
732 zfs_secpolicy_deleg_share(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
733 {
734 vnode_t *vp;
735 int error;
736
737 if ((error = lookupname(zc->zc_value, UIO_SYSSPACE,
738 NO_FOLLOW, NULL, &vp)) != 0)
739 return (error);
740
741 /* Now make sure mntpnt and dataset are ZFS */
742
743 if (vp->v_vfsp->vfs_fstype != zfsfstype ||
744 (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource),
745 zc->zc_name) != 0)) {
746 VN_RELE(vp);
747 return (EPERM);
748 }
749
750 VN_RELE(vp);
751 return (dsl_deleg_access(zc->zc_name,
752 ZFS_DELEG_PERM_SHARE, cr));
753 }
754
755 int
756 zfs_secpolicy_share(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
757 {
758 if (!INGLOBALZONE(curproc))
759 return (EPERM);
760
761 if (secpolicy_nfs(cr) == 0) {
762 return (0);
763 } else {
764 return (zfs_secpolicy_deleg_share(zc, innvl, cr));
765 }
766 }
767
768 int
769 zfs_secpolicy_smb_acl(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
770 {
771 if (!INGLOBALZONE(curproc))
772 return (EPERM);
773
774 if (secpolicy_smb(cr) == 0) {
775 return (0);
776 } else {
777 return (zfs_secpolicy_deleg_share(zc, innvl, cr));
778 }
779 }
780
781 static int
782 zfs_get_parent(const char *datasetname, char *parent, int parentsize)
783 {
784 char *cp;
785
786 /*
787 * Remove the @bla or /bla from the end of the name to get the parent.
788 */
789 (void) strncpy(parent, datasetname, parentsize);
790 cp = strrchr(parent, '@');
791 if (cp != NULL) {
792 cp[0] = '\0';
793 } else {
794 cp = strrchr(parent, '/');
795 if (cp == NULL)
796 return (ENOENT);
797 cp[0] = '\0';
798 }
799
800 return (0);
801 }
802
803 int
804 zfs_secpolicy_destroy_perms(const char *name, cred_t *cr)
805 {
806 int error;
807
808 if ((error = zfs_secpolicy_write_perms(name,
809 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
810 return (error);
811
812 return (zfs_secpolicy_write_perms(name, ZFS_DELEG_PERM_DESTROY, cr));
813 }
814
815 /* ARGSUSED */
816 static int
817 zfs_secpolicy_destroy(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
818 {
819 return (zfs_secpolicy_destroy_perms(zc->zc_name, cr));
820 }
821
822 /*
823 * Destroying snapshots with delegated permissions requires
824 * descendant mount and destroy permissions.
825 */
826 /* ARGSUSED */
827 static int
828 zfs_secpolicy_destroy_snaps(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
829 {
830 nvlist_t *snaps;
831 nvpair_t *pair, *nextpair;
832 int error = 0;
833
834 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
835 return (EINVAL);
836 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
837 pair = nextpair) {
838 dsl_dataset_t *ds;
839
840 nextpair = nvlist_next_nvpair(snaps, pair);
841 error = dsl_dataset_hold(nvpair_name(pair), FTAG, &ds);
842 if (error == 0) {
843 dsl_dataset_rele(ds, FTAG);
844 } else if (error == ENOENT) {
845 /*
846 * Ignore any snapshots that don't exist (we consider
847 * them "already destroyed"). Remove the name from the
848 * nvl here in case the snapshot is created between
849 * now and when we try to destroy it (in which case
850 * we don't want to destroy it since we haven't
851 * checked for permission).
852 */
853 fnvlist_remove_nvpair(snaps, pair);
854 error = 0;
855 continue;
856 } else {
857 break;
858 }
859 error = zfs_secpolicy_destroy_perms(nvpair_name(pair), cr);
860 if (error != 0)
861 break;
862 }
863
864 return (error);
865 }
866
867 int
868 zfs_secpolicy_rename_perms(const char *from, const char *to, cred_t *cr)
869 {
870 char parentname[MAXNAMELEN];
871 int error;
872
873 if ((error = zfs_secpolicy_write_perms(from,
874 ZFS_DELEG_PERM_RENAME, cr)) != 0)
875 return (error);
876
877 if ((error = zfs_secpolicy_write_perms(from,
878 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
879 return (error);
880
881 if ((error = zfs_get_parent(to, parentname,
882 sizeof (parentname))) != 0)
883 return (error);
884
885 if ((error = zfs_secpolicy_write_perms(parentname,
886 ZFS_DELEG_PERM_CREATE, cr)) != 0)
887 return (error);
888
889 if ((error = zfs_secpolicy_write_perms(parentname,
890 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
891 return (error);
892
893 return (error);
894 }
895
896 /* ARGSUSED */
897 static int
898 zfs_secpolicy_rename(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
899 {
900 return (zfs_secpolicy_rename_perms(zc->zc_name, zc->zc_value, cr));
901 }
902
903 /* ARGSUSED */
904 static int
905 zfs_secpolicy_promote(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
906 {
907 char parentname[MAXNAMELEN];
908 objset_t *clone;
909 int error;
910
911 error = zfs_secpolicy_write_perms(zc->zc_name,
912 ZFS_DELEG_PERM_PROMOTE, cr);
913 if (error)
914 return (error);
915
916 error = dmu_objset_hold(zc->zc_name, FTAG, &clone);
917
918 if (error == 0) {
919 dsl_dataset_t *pclone = NULL;
920 dsl_dir_t *dd;
921 dd = clone->os_dsl_dataset->ds_dir;
922
923 rw_enter(&dd->dd_pool->dp_config_rwlock, RW_READER);
924 error = dsl_dataset_hold_obj(dd->dd_pool,
925 dd->dd_phys->dd_origin_obj, FTAG, &pclone);
926 rw_exit(&dd->dd_pool->dp_config_rwlock);
927 if (error) {
928 dmu_objset_rele(clone, FTAG);
929 return (error);
930 }
931
932 error = zfs_secpolicy_write_perms(zc->zc_name,
933 ZFS_DELEG_PERM_MOUNT, cr);
934
935 dsl_dataset_name(pclone, parentname);
936 dmu_objset_rele(clone, FTAG);
937 dsl_dataset_rele(pclone, FTAG);
938 if (error == 0)
939 error = zfs_secpolicy_write_perms(parentname,
940 ZFS_DELEG_PERM_PROMOTE, cr);
941 }
942 return (error);
943 }
944
945 /* ARGSUSED */
946 static int
947 zfs_secpolicy_recv(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
948 {
949 int error;
950
951 if ((error = zfs_secpolicy_write_perms(zc->zc_name,
952 ZFS_DELEG_PERM_RECEIVE, cr)) != 0)
953 return (error);
954
955 if ((error = zfs_secpolicy_write_perms(zc->zc_name,
956 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
957 return (error);
958
959 return (zfs_secpolicy_write_perms(zc->zc_name,
960 ZFS_DELEG_PERM_CREATE, cr));
961 }
962
963 int
964 zfs_secpolicy_snapshot_perms(const char *name, cred_t *cr)
965 {
966 return (zfs_secpolicy_write_perms(name,
967 ZFS_DELEG_PERM_SNAPSHOT, cr));
968 }
969
970 /*
971 * Check for permission to create each snapshot in the nvlist.
972 */
973 /* ARGSUSED */
974 static int
975 zfs_secpolicy_snapshot(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
976 {
977 nvlist_t *snaps;
978 int error;
979 nvpair_t *pair;
980
981 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
982 return (EINVAL);
983 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
984 pair = nvlist_next_nvpair(snaps, pair)) {
985 char *name = nvpair_name(pair);
986 char *atp = strchr(name, '@');
987
988 if (atp == NULL) {
989 error = EINVAL;
990 break;
991 }
992 *atp = '\0';
993 error = zfs_secpolicy_snapshot_perms(name, cr);
994 *atp = '@';
995 if (error != 0)
996 break;
997 }
998 return (error);
999 }
1000
1001 /* ARGSUSED */
1002 static int
1003 zfs_secpolicy_log_history(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1004 {
1005 /*
1006 * Even root must have a proper TSD so that we know what pool
1007 * to log to.
1008 */
1009 if (tsd_get(zfs_allow_log_key) == NULL)
1010 return (EPERM);
1011 return (0);
1012 }
1013
1014 static int
1015 zfs_secpolicy_create_clone(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1016 {
1017 char parentname[MAXNAMELEN];
1018 int error;
1019 char *origin;
1020
1021 if ((error = zfs_get_parent(zc->zc_name, parentname,
1022 sizeof (parentname))) != 0)
1023 return (error);
1024
1025 if (nvlist_lookup_string(innvl, "origin", &origin) == 0 &&
1026 (error = zfs_secpolicy_write_perms(origin,
1027 ZFS_DELEG_PERM_CLONE, cr)) != 0)
1028 return (error);
1029
1030 if ((error = zfs_secpolicy_write_perms(parentname,
1031 ZFS_DELEG_PERM_CREATE, cr)) != 0)
1032 return (error);
1033
1034 return (zfs_secpolicy_write_perms(parentname,
1035 ZFS_DELEG_PERM_MOUNT, cr));
1036 }
1037
1038 /*
1039 * Policy for pool operations - create/destroy pools, add vdevs, etc. Requires
1040 * SYS_CONFIG privilege, which is not available in a local zone.
1041 */
1042 /* ARGSUSED */
1043 static int
1044 zfs_secpolicy_config(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1045 {
1046 if (secpolicy_sys_config(cr, B_FALSE) != 0)
1047 return (EPERM);
1048
1049 return (0);
1050 }
1051
1052 /*
1053 * Policy for object to name lookups.
1054 */
1055 /* ARGSUSED */
1056 static int
1057 zfs_secpolicy_diff(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1058 {
1059 int error;
1060
1061 if ((error = secpolicy_sys_config(cr, B_FALSE)) == 0)
1062 return (0);
1063
1064 error = zfs_secpolicy_write_perms(zc->zc_name, ZFS_DELEG_PERM_DIFF, cr);
1065 return (error);
1066 }
1067
1068 /*
1069 * Policy for fault injection. Requires all privileges.
1070 */
1071 /* ARGSUSED */
1072 static int
1073 zfs_secpolicy_inject(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1074 {
1075 return (secpolicy_zinject(cr));
1076 }
1077
1078 /* ARGSUSED */
1079 static int
1080 zfs_secpolicy_inherit_prop(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1081 {
1082 zfs_prop_t prop = zfs_name_to_prop(zc->zc_value);
1083
1084 if (prop == ZPROP_INVAL) {
1085 if (!zfs_prop_user(zc->zc_value))
1086 return (EINVAL);
1087 return (zfs_secpolicy_write_perms(zc->zc_name,
1088 ZFS_DELEG_PERM_USERPROP, cr));
1089 } else {
1090 return (zfs_secpolicy_setprop(zc->zc_name, prop,
1091 NULL, cr));
1092 }
1093 }
1094
1095 static int
1096 zfs_secpolicy_userspace_one(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1097 {
1098 int err = zfs_secpolicy_read(zc, innvl, cr);
1099 if (err)
1100 return (err);
1101
1102 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
1103 return (EINVAL);
1104
1105 if (zc->zc_value[0] == 0) {
1106 /*
1107 * They are asking about a posix uid/gid. If it's
1108 * themself, allow it.
1109 */
1110 if (zc->zc_objset_type == ZFS_PROP_USERUSED ||
1111 zc->zc_objset_type == ZFS_PROP_USERQUOTA) {
1112 if (zc->zc_guid == crgetuid(cr))
1113 return (0);
1114 } else {
1115 if (groupmember(zc->zc_guid, cr))
1116 return (0);
1117 }
1118 }
1119
1120 return (zfs_secpolicy_write_perms(zc->zc_name,
1121 userquota_perms[zc->zc_objset_type], cr));
1122 }
1123
1124 static int
1125 zfs_secpolicy_userspace_many(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1126 {
1127 int err = zfs_secpolicy_read(zc, innvl, cr);
1128 if (err)
1129 return (err);
1130
1131 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
1132 return (EINVAL);
1133
1134 return (zfs_secpolicy_write_perms(zc->zc_name,
1135 userquota_perms[zc->zc_objset_type], cr));
1136 }
1137
1138 /* ARGSUSED */
1139 static int
1140 zfs_secpolicy_userspace_upgrade(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1141 {
1142 return (zfs_secpolicy_setprop(zc->zc_name, ZFS_PROP_VERSION,
1143 NULL, cr));
1144 }
1145
1146 /* ARGSUSED */
1147 static int
1148 zfs_secpolicy_hold(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1149 {
1150 return (zfs_secpolicy_write_perms(zc->zc_name,
1151 ZFS_DELEG_PERM_HOLD, cr));
1152 }
1153
1154 /* ARGSUSED */
1155 static int
1156 zfs_secpolicy_release(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1157 {
1158 return (zfs_secpolicy_write_perms(zc->zc_name,
1159 ZFS_DELEG_PERM_RELEASE, cr));
1160 }
1161
1162 /*
1163 * Policy for allowing temporary snapshots to be taken or released
1164 */
1165 static int
1166 zfs_secpolicy_tmp_snapshot(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1167 {
1168 /*
1169 * A temporary snapshot is the same as a snapshot,
1170 * hold, destroy and release all rolled into one.
1171 * Delegated diff alone is sufficient that we allow this.
1172 */
1173 int error;
1174
1175 if ((error = zfs_secpolicy_write_perms(zc->zc_name,
1176 ZFS_DELEG_PERM_DIFF, cr)) == 0)
1177 return (0);
1178
1179 error = zfs_secpolicy_snapshot_perms(zc->zc_name, cr);
1180 if (!error)
1181 error = zfs_secpolicy_hold(zc, innvl, cr);
1182 if (!error)
1183 error = zfs_secpolicy_release(zc, innvl, cr);
1184 if (!error)
1185 error = zfs_secpolicy_destroy(zc, innvl, cr);
1186 return (error);
1187 }
1188
1189 /*
1190 * Returns the nvlist as specified by the user in the zfs_cmd_t.
1191 */
1192 static int
1193 get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp)
1194 {
1195 char *packed;
1196 int error;
1197 nvlist_t *list = NULL;
1198
1199 /*
1200 * Read in and unpack the user-supplied nvlist.
1201 */
1202 if (size == 0)
1203 return (EINVAL);
1204
1205 packed = kmem_alloc(size, KM_SLEEP);
1206
1207 if ((error = ddi_copyin((void *)(uintptr_t)nvl, packed, size,
1208 iflag)) != 0) {
1209 kmem_free(packed, size);
1210 return (error);
1211 }
1212
1213 if ((error = nvlist_unpack(packed, size, &list, 0)) != 0) {
1214 kmem_free(packed, size);
1215 return (error);
1216 }
1217
1218 kmem_free(packed, size);
1219
1220 *nvp = list;
1221 return (0);
1222 }
1223
1224 /*
1225 * Reduce the size of this nvlist until it can be serialized in 'max' bytes.
1226 * Entries will be removed from the end of the nvlist, and one int32 entry
1227 * named "N_MORE_ERRORS" will be added indicating how many entries were
1228 * removed.
1229 */
1230 static int
1231 nvlist_smush(nvlist_t *errors, size_t max)
1232 {
1233 size_t size;
1234
1235 size = fnvlist_size(errors);
1236
1237 if (size > max) {
1238 nvpair_t *more_errors;
1239 int n = 0;
1240
1241 if (max < 1024)
1242 return (ENOMEM);
1243
1244 fnvlist_add_int32(errors, ZPROP_N_MORE_ERRORS, 0);
1245 more_errors = nvlist_prev_nvpair(errors, NULL);
1246
1247 do {
1248 nvpair_t *pair = nvlist_prev_nvpair(errors,
1249 more_errors);
1250 fnvlist_remove_nvpair(errors, pair);
1251 n++;
1252 size = fnvlist_size(errors);
1253 } while (size > max);
1254
1255 fnvlist_remove_nvpair(errors, more_errors);
1256 fnvlist_add_int32(errors, ZPROP_N_MORE_ERRORS, n);
1257 ASSERT3U(fnvlist_size(errors), <=, max);
1258 }
1259
1260 return (0);
1261 }
1262
1263 static int
1264 put_nvlist(zfs_cmd_t *zc, nvlist_t *nvl)
1265 {
1266 char *packed = NULL;
1267 int error = 0;
1268 size_t size;
1269
1270 size = fnvlist_size(nvl);
1271
1272 if (size > zc->zc_nvlist_dst_size) {
1273 error = ENOMEM;
1274 } else {
1275 packed = fnvlist_pack(nvl, &size);
1276 if (ddi_copyout(packed, (void *)(uintptr_t)zc->zc_nvlist_dst,
1277 size, zc->zc_iflags) != 0)
1278 error = EFAULT;
1279 fnvlist_pack_free(packed, size);
1280 }
1281
1282 zc->zc_nvlist_dst_size = size;
1283 zc->zc_nvlist_dst_filled = B_TRUE;
1284 return (error);
1285 }
1286
1287 static int
1288 getzfsvfs(const char *dsname, zfsvfs_t **zfvp)
1289 {
1290 objset_t *os;
1291 int error;
1292
1293 error = dmu_objset_hold(dsname, FTAG, &os);
1294 if (error)
1295 return (error);
1296 if (dmu_objset_type(os) != DMU_OST_ZFS) {
1297 dmu_objset_rele(os, FTAG);
1298 return (EINVAL);
1299 }
1300
1301 mutex_enter(&os->os_user_ptr_lock);
1302 *zfvp = dmu_objset_get_user(os);
1303 if (*zfvp) {
1304 VFS_HOLD((*zfvp)->z_vfs);
1305 } else {
1306 error = ESRCH;
1307 }
1308 mutex_exit(&os->os_user_ptr_lock);
1309 dmu_objset_rele(os, FTAG);
1310 return (error);
1311 }
1312
1313 /*
1314 * Find a zfsvfs_t for a mounted filesystem, or create our own, in which
1315 * case its z_vfs will be NULL, and it will be opened as the owner.
1316 * If 'writer' is set, the z_teardown_lock will be held for RW_WRITER,
1317 * which prevents all vnode ops from running.
1318 */
1319 static int
1320 zfsvfs_hold(const char *name, void *tag, zfsvfs_t **zfvp, boolean_t writer)
1321 {
1322 int error = 0;
1323
1324 if (getzfsvfs(name, zfvp) != 0)
1325 error = zfsvfs_create(name, zfvp);
1326 if (error == 0) {
1327 rrw_enter(&(*zfvp)->z_teardown_lock, (writer) ? RW_WRITER :
1328 RW_READER, tag);
1329 if ((*zfvp)->z_unmounted) {
1330 /*
1331 * XXX we could probably try again, since the unmounting
1332 * thread should be just about to disassociate the
1333 * objset from the zfsvfs.
1334 */
1335 rrw_exit(&(*zfvp)->z_teardown_lock, tag);
1336 return (EBUSY);
1337 }
1338 }
1339 return (error);
1340 }
1341
1342 static void
1343 zfsvfs_rele(zfsvfs_t *zfsvfs, void *tag)
1344 {
1345 rrw_exit(&zfsvfs->z_teardown_lock, tag);
1346
1347 if (zfsvfs->z_vfs) {
1348 VFS_RELE(zfsvfs->z_vfs);
1349 } else {
1350 dmu_objset_disown(zfsvfs->z_os, zfsvfs);
1351 zfsvfs_free(zfsvfs);
1352 }
1353 }
1354
1355 static int
1356 zfs_ioc_pool_create(zfs_cmd_t *zc)
1357 {
1358 int error;
1359 nvlist_t *config, *props = NULL;
1360 nvlist_t *rootprops = NULL;
1361 nvlist_t *zplprops = NULL;
1362
1363 if (error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1364 zc->zc_iflags, &config))
1365 return (error);
1366
1367 if (zc->zc_nvlist_src_size != 0 && (error =
1368 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1369 zc->zc_iflags, &props))) {
1370 nvlist_free(config);
1371 return (error);
1372 }
1373
1374 if (props) {
1375 nvlist_t *nvl = NULL;
1376 uint64_t version = SPA_VERSION;
1377
1378 (void) nvlist_lookup_uint64(props,
1379 zpool_prop_to_name(ZPOOL_PROP_VERSION), &version);
1380 if (!SPA_VERSION_IS_SUPPORTED(version)) {
1381 error = EINVAL;
1382 goto pool_props_bad;
1383 }
1384 (void) nvlist_lookup_nvlist(props, ZPOOL_ROOTFS_PROPS, &nvl);
1385 if (nvl) {
1386 error = nvlist_dup(nvl, &rootprops, KM_SLEEP);
1387 if (error != 0) {
1388 nvlist_free(config);
1389 nvlist_free(props);
1390 return (error);
1391 }
1392 (void) nvlist_remove_all(props, ZPOOL_ROOTFS_PROPS);
1393 }
1394 VERIFY(nvlist_alloc(&zplprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
1395 error = zfs_fill_zplprops_root(version, rootprops,
1396 zplprops, NULL);
1397 if (error)
1398 goto pool_props_bad;
1399 }
1400
1401 error = spa_create(zc->zc_name, config, props, zplprops);
1402
1403 /*
1404 * Set the remaining root properties
1405 */
1406 if (!error && (error = zfs_set_prop_nvlist(zc->zc_name,
1407 ZPROP_SRC_LOCAL, rootprops, NULL)) != 0)
1408 (void) spa_destroy(zc->zc_name);
1409
1410 pool_props_bad:
1411 nvlist_free(rootprops);
1412 nvlist_free(zplprops);
1413 nvlist_free(config);
1414 nvlist_free(props);
1415
1416 return (error);
1417 }
1418
1419 static int
1420 zfs_ioc_pool_destroy(zfs_cmd_t *zc)
1421 {
1422 int error;
1423 zfs_log_history(zc);
1424 error = spa_destroy(zc->zc_name);
1425 if (error == 0)
1426 zvol_remove_minors(zc->zc_name);
1427 return (error);
1428 }
1429
1430 static int
1431 zfs_ioc_pool_import(zfs_cmd_t *zc)
1432 {
1433 nvlist_t *config, *props = NULL;
1434 uint64_t guid;
1435 int error;
1436
1437 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1438 zc->zc_iflags, &config)) != 0)
1439 return (error);
1440
1441 if (zc->zc_nvlist_src_size != 0 && (error =
1442 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1443 zc->zc_iflags, &props))) {
1444 nvlist_free(config);
1445 return (error);
1446 }
1447
1448 if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, &guid) != 0 ||
1449 guid != zc->zc_guid)
1450 error = EINVAL;
1451 else
1452 error = spa_import(zc->zc_name, config, props, zc->zc_cookie);
1453
1454 if (zc->zc_nvlist_dst != 0) {
1455 int err;
1456
1457 if ((err = put_nvlist(zc, config)) != 0)
1458 error = err;
1459 }
1460
1461 nvlist_free(config);
1462
1463 if (props)
1464 nvlist_free(props);
1465
1466 return (error);
1467 }
1468
1469 static int
1470 zfs_ioc_pool_export(zfs_cmd_t *zc)
1471 {
1472 int error;
1473 boolean_t force = (boolean_t)zc->zc_cookie;
1474 boolean_t hardforce = (boolean_t)zc->zc_guid;
1475
1476 zfs_log_history(zc);
1477 error = spa_export(zc->zc_name, NULL, force, hardforce);
1478 if (error == 0)
1479 zvol_remove_minors(zc->zc_name);
1480 return (error);
1481 }
1482
1483 static int
1484 zfs_ioc_pool_configs(zfs_cmd_t *zc)
1485 {
1486 nvlist_t *configs;
1487 int error;
1488
1489 if ((configs = spa_all_configs(&zc->zc_cookie)) == NULL)
1490 return (EEXIST);
1491
1492 error = put_nvlist(zc, configs);
1493
1494 nvlist_free(configs);
1495
1496 return (error);
1497 }
1498
1499 /*
1500 * inputs:
1501 * zc_name name of the pool
1502 *
1503 * outputs:
1504 * zc_cookie real errno
1505 * zc_nvlist_dst config nvlist
1506 * zc_nvlist_dst_size size of config nvlist
1507 */
1508 static int
1509 zfs_ioc_pool_stats(zfs_cmd_t *zc)
1510 {
1511 nvlist_t *config;
1512 int error;
1513 int ret = 0;
1514
1515 error = spa_get_stats(zc->zc_name, &config, zc->zc_value,
1516 sizeof (zc->zc_value));
1517
1518 if (config != NULL) {
1519 ret = put_nvlist(zc, config);
1520 nvlist_free(config);
1521
1522 /*
1523 * The config may be present even if 'error' is non-zero.
1524 * In this case we return success, and preserve the real errno
1525 * in 'zc_cookie'.
1526 */
1527 zc->zc_cookie = error;
1528 } else {
1529 ret = error;
1530 }
1531
1532 return (ret);
1533 }
1534
1535 /*
1536 * Try to import the given pool, returning pool stats as appropriate so that
1537 * user land knows which devices are available and overall pool health.
1538 */
1539 static int
1540 zfs_ioc_pool_tryimport(zfs_cmd_t *zc)
1541 {
1542 nvlist_t *tryconfig, *config;
1543 int error;
1544
1545 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1546 zc->zc_iflags, &tryconfig)) != 0)
1547 return (error);
1548
1549 config = spa_tryimport(tryconfig);
1550
1551 nvlist_free(tryconfig);
1552
1553 if (config == NULL)
1554 return (EINVAL);
1555
1556 error = put_nvlist(zc, config);
1557 nvlist_free(config);
1558
1559 return (error);
1560 }
1561
1562 /*
1563 * inputs:
1564 * zc_name name of the pool
1565 * zc_cookie scan func (pool_scan_func_t)
1566 */
1567 static int
1568 zfs_ioc_pool_scan(zfs_cmd_t *zc)
1569 {
1570 spa_t *spa;
1571 int error;
1572
1573 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1574 return (error);
1575
1576 if (zc->zc_cookie == POOL_SCAN_NONE)
1577 error = spa_scan_stop(spa);
1578 else
1579 error = spa_scan(spa, zc->zc_cookie);
1580
1581 spa_close(spa, FTAG);
1582
1583 return (error);
1584 }
1585
1586 static int
1587 zfs_ioc_pool_freeze(zfs_cmd_t *zc)
1588 {
1589 spa_t *spa;
1590 int error;
1591
1592 error = spa_open(zc->zc_name, &spa, FTAG);
1593 if (error == 0) {
1594 spa_freeze(spa);
1595 spa_close(spa, FTAG);
1596 }
1597 return (error);
1598 }
1599
1600 static int
1601 zfs_ioc_pool_upgrade(zfs_cmd_t *zc)
1602 {
1603 spa_t *spa;
1604 int error;
1605
1606 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1607 return (error);
1608
1609 if (zc->zc_cookie < spa_version(spa) ||
1610 !SPA_VERSION_IS_SUPPORTED(zc->zc_cookie)) {
1611 spa_close(spa, FTAG);
1612 return (EINVAL);
1613 }
1614
1615 spa_upgrade(spa, zc->zc_cookie);
1616 spa_close(spa, FTAG);
1617
1618 return (error);
1619 }
1620
1621 static int
1622 zfs_ioc_pool_get_history(zfs_cmd_t *zc)
1623 {
1624 spa_t *spa;
1625 char *hist_buf;
1626 uint64_t size;
1627 int error;
1628
1629 if ((size = zc->zc_history_len) == 0)
1630 return (EINVAL);
1631
1632 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1633 return (error);
1634
1635 if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) {
1636 spa_close(spa, FTAG);
1637 return (ENOTSUP);
1638 }
1639
1640 hist_buf = kmem_alloc(size, KM_SLEEP);
1641 if ((error = spa_history_get(spa, &zc->zc_history_offset,
1642 &zc->zc_history_len, hist_buf)) == 0) {
1643 error = ddi_copyout(hist_buf,
1644 (void *)(uintptr_t)zc->zc_history,
1645 zc->zc_history_len, zc->zc_iflags);
1646 }
1647
1648 spa_close(spa, FTAG);
1649 kmem_free(hist_buf, size);
1650 return (error);
1651 }
1652
1653 static int
1654 zfs_ioc_pool_reguid(zfs_cmd_t *zc)
1655 {
1656 spa_t *spa;
1657 int error;
1658
1659 error = spa_open(zc->zc_name, &spa, FTAG);
1660 if (error == 0) {
1661 error = spa_change_guid(spa);
1662 spa_close(spa, FTAG);
1663 }
1664 return (error);
1665 }
1666
1667 static int
1668 zfs_ioc_dsobj_to_dsname(zfs_cmd_t *zc)
1669 {
1670 int error;
1671
1672 if (error = dsl_dsobj_to_dsname(zc->zc_name, zc->zc_obj, zc->zc_value))
1673 return (error);
1674
1675 return (0);
1676 }
1677
1678 /*
1679 * inputs:
1680 * zc_name name of filesystem
1681 * zc_obj object to find
1682 *
1683 * outputs:
1684 * zc_value name of object
1685 */
1686 static int
1687 zfs_ioc_obj_to_path(zfs_cmd_t *zc)
1688 {
1689 objset_t *os;
1690 int error;
1691
1692 /* XXX reading from objset not owned */
1693 if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os)) != 0)
1694 return (error);
1695 if (dmu_objset_type(os) != DMU_OST_ZFS) {
1696 dmu_objset_rele(os, FTAG);
1697 return (EINVAL);
1698 }
1699 error = zfs_obj_to_path(os, zc->zc_obj, zc->zc_value,
1700 sizeof (zc->zc_value));
1701 dmu_objset_rele(os, FTAG);
1702
1703 return (error);
1704 }
1705
1706 /*
1707 * inputs:
1708 * zc_name name of filesystem
1709 * zc_obj object to find
1710 *
1711 * outputs:
1712 * zc_stat stats on object
1713 * zc_value path to object
1714 */
1715 static int
1716 zfs_ioc_obj_to_stats(zfs_cmd_t *zc)
1717 {
1718 objset_t *os;
1719 int error;
1720
1721 /* XXX reading from objset not owned */
1722 if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os)) != 0)
1723 return (error);
1724 if (dmu_objset_type(os) != DMU_OST_ZFS) {
1725 dmu_objset_rele(os, FTAG);
1726 return (EINVAL);
1727 }
1728 error = zfs_obj_to_stats(os, zc->zc_obj, &zc->zc_stat, zc->zc_value,
1729 sizeof (zc->zc_value));
1730 dmu_objset_rele(os, FTAG);
1731
1732 return (error);
1733 }
1734
1735 static int
1736 zfs_ioc_vdev_add(zfs_cmd_t *zc)
1737 {
1738 spa_t *spa;
1739 int error;
1740 nvlist_t *config, **l2cache, **spares;
1741 uint_t nl2cache = 0, nspares = 0;
1742
1743 error = spa_open(zc->zc_name, &spa, FTAG);
1744 if (error != 0)
1745 return (error);
1746
1747 error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1748 zc->zc_iflags, &config);
1749 (void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_L2CACHE,
1750 &l2cache, &nl2cache);
1751
1752 (void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_SPARES,
1753 &spares, &nspares);
1754
1755 /*
1756 * A root pool with concatenated devices is not supported.
1757 * Thus, can not add a device to a root pool.
1758 *
1759 * Intent log device can not be added to a rootpool because
1760 * during mountroot, zil is replayed, a seperated log device
1761 * can not be accessed during the mountroot time.
1762 *
1763 * l2cache and spare devices are ok to be added to a rootpool.
1764 */
1765 if (spa_bootfs(spa) != 0 && nl2cache == 0 && nspares == 0) {
1766 nvlist_free(config);
1767 spa_close(spa, FTAG);
1768 return (EDOM);
1769 }
1770
1771 if (error == 0) {
1772 error = spa_vdev_add(spa, config);
1773 nvlist_free(config);
1774 }
1775 spa_close(spa, FTAG);
1776 return (error);
1777 }
1778
1779 /*
1780 * inputs:
1781 * zc_name name of the pool
1782 * zc_nvlist_conf nvlist of devices to remove
1783 * zc_cookie to stop the remove?
1784 */
1785 static int
1786 zfs_ioc_vdev_remove(zfs_cmd_t *zc)
1787 {
1788 spa_t *spa;
1789 int error;
1790
1791 error = spa_open(zc->zc_name, &spa, FTAG);
1792 if (error != 0)
1793 return (error);
1794 error = spa_vdev_remove(spa, zc->zc_guid, B_FALSE);
1795 spa_close(spa, FTAG);
1796 return (error);
1797 }
1798
1799 static int
1800 zfs_ioc_vdev_set_state(zfs_cmd_t *zc)
1801 {
1802 spa_t *spa;
1803 int error;
1804 vdev_state_t newstate = VDEV_STATE_UNKNOWN;
1805
1806 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1807 return (error);
1808 switch (zc->zc_cookie) {
1809 case VDEV_STATE_ONLINE:
1810 error = vdev_online(spa, zc->zc_guid, zc->zc_obj, &newstate);
1811 break;
1812
1813 case VDEV_STATE_OFFLINE:
1814 error = vdev_offline(spa, zc->zc_guid, zc->zc_obj);
1815 break;
1816
1817 case VDEV_STATE_FAULTED:
1818 if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED &&
1819 zc->zc_obj != VDEV_AUX_EXTERNAL)
1820 zc->zc_obj = VDEV_AUX_ERR_EXCEEDED;
1821
1822 error = vdev_fault(spa, zc->zc_guid, zc->zc_obj);
1823 break;
1824
1825 case VDEV_STATE_DEGRADED:
1826 if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED &&
1827 zc->zc_obj != VDEV_AUX_EXTERNAL)
1828 zc->zc_obj = VDEV_AUX_ERR_EXCEEDED;
1829
1830 error = vdev_degrade(spa, zc->zc_guid, zc->zc_obj);
1831 break;
1832
1833 default:
1834 error = EINVAL;
1835 }
1836 zc->zc_cookie = newstate;
1837 spa_close(spa, FTAG);
1838 return (error);
1839 }
1840
1841 static int
1842 zfs_ioc_vdev_attach(zfs_cmd_t *zc)
1843 {
1844 spa_t *spa;
1845 int replacing = zc->zc_cookie;
1846 nvlist_t *config;
1847 int error;
1848
1849 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1850 return (error);
1851
1852 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1853 zc->zc_iflags, &config)) == 0) {
1854 error = spa_vdev_attach(spa, zc->zc_guid, config, replacing);
1855 nvlist_free(config);
1856 }
1857
1858 spa_close(spa, FTAG);
1859 return (error);
1860 }
1861
1862 static int
1863 zfs_ioc_vdev_detach(zfs_cmd_t *zc)
1864 {
1865 spa_t *spa;
1866 int error;
1867
1868 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1869 return (error);
1870
1871 error = spa_vdev_detach(spa, zc->zc_guid, 0, B_FALSE);
1872
1873 spa_close(spa, FTAG);
1874 return (error);
1875 }
1876
1877 static int
1878 zfs_ioc_vdev_split(zfs_cmd_t *zc)
1879 {
1880 spa_t *spa;
1881 nvlist_t *config, *props = NULL;
1882 int error;
1883 boolean_t exp = !!(zc->zc_cookie & ZPOOL_EXPORT_AFTER_SPLIT);
1884
1885 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1886 return (error);
1887
1888 if (error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1889 zc->zc_iflags, &config)) {
1890 spa_close(spa, FTAG);
1891 return (error);
1892 }
1893
1894 if (zc->zc_nvlist_src_size != 0 && (error =
1895 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1896 zc->zc_iflags, &props))) {
1897 spa_close(spa, FTAG);
1898 nvlist_free(config);
1899 return (error);
1900 }
1901
1902 error = spa_vdev_split_mirror(spa, zc->zc_string, config, props, exp);
1903
1904 spa_close(spa, FTAG);
1905
1906 nvlist_free(config);
1907 nvlist_free(props);
1908
1909 return (error);
1910 }
1911
1912 static int
1913 zfs_ioc_vdev_setpath(zfs_cmd_t *zc)
1914 {
1915 spa_t *spa;
1916 char *path = zc->zc_value;
1917 uint64_t guid = zc->zc_guid;
1918 int error;
1919
1920 error = spa_open(zc->zc_name, &spa, FTAG);
1921 if (error != 0)
1922 return (error);
1923
1924 error = spa_vdev_setpath(spa, guid, path);
1925 spa_close(spa, FTAG);
1926 return (error);
1927 }
1928
1929 static int
1930 zfs_ioc_vdev_setfru(zfs_cmd_t *zc)
1931 {
1932 spa_t *spa;
1933 char *fru = zc->zc_value;
1934 uint64_t guid = zc->zc_guid;
1935 int error;
1936
1937 error = spa_open(zc->zc_name, &spa, FTAG);
1938 if (error != 0)
1939 return (error);
1940
1941 error = spa_vdev_setfru(spa, guid, fru);
1942 spa_close(spa, FTAG);
1943 return (error);
1944 }
1945
1946 static int
1947 zfs_ioc_objset_stats_impl(zfs_cmd_t *zc, objset_t *os)
1948 {
1949 int error = 0;
1950 nvlist_t *nv;
1951
1952 dmu_objset_fast_stat(os, &zc->zc_objset_stats);
1953
1954 if (zc->zc_nvlist_dst != 0 &&
1955 (error = dsl_prop_get_all(os, &nv)) == 0) {
1956 dmu_objset_stats(os, nv);
1957 /*
1958 * NB: zvol_get_stats() will read the objset contents,
1959 * which we aren't supposed to do with a
1960 * DS_MODE_USER hold, because it could be
1961 * inconsistent. So this is a bit of a workaround...
1962 * XXX reading with out owning
1963 */
1964 if (!zc->zc_objset_stats.dds_inconsistent &&
1965 dmu_objset_type(os) == DMU_OST_ZVOL) {
1966 error = zvol_get_stats(os, nv);
1967 if (error == EIO)
1968 return (error);
1969 VERIFY0(error);
1970 }
1971 error = put_nvlist(zc, nv);
1972 nvlist_free(nv);
1973 }
1974
1975 return (error);
1976 }
1977
1978 /*
1979 * inputs:
1980 * zc_name name of filesystem
1981 * zc_nvlist_dst_size size of buffer for property nvlist
1982 *
1983 * outputs:
1984 * zc_objset_stats stats
1985 * zc_nvlist_dst property nvlist
1986 * zc_nvlist_dst_size size of property nvlist
1987 */
1988 static int
1989 zfs_ioc_objset_stats(zfs_cmd_t *zc)
1990 {
1991 objset_t *os = NULL;
1992 int error;
1993
1994 if (error = dmu_objset_hold(zc->zc_name, FTAG, &os))
1995 return (error);
1996
1997 error = zfs_ioc_objset_stats_impl(zc, os);
1998
1999 dmu_objset_rele(os, FTAG);
2000
2001 return (error);
2002 }
2003
2004 /*
2005 * inputs:
2006 * zc_name name of filesystem
2007 * zc_nvlist_dst_size size of buffer for property nvlist
2008 *
2009 * outputs:
2010 * zc_nvlist_dst received property nvlist
2011 * zc_nvlist_dst_size size of received property nvlist
2012 *
2013 * Gets received properties (distinct from local properties on or after
2014 * SPA_VERSION_RECVD_PROPS) for callers who want to differentiate received from
2015 * local property values.
2016 */
2017 static int
2018 zfs_ioc_objset_recvd_props(zfs_cmd_t *zc)
2019 {
2020 objset_t *os = NULL;
2021 int error;
2022 nvlist_t *nv;
2023
2024 if (error = dmu_objset_hold(zc->zc_name, FTAG, &os))
2025 return (error);
2026
2027 /*
2028 * Without this check, we would return local property values if the
2029 * caller has not already received properties on or after
2030 * SPA_VERSION_RECVD_PROPS.
2031 */
2032 if (!dsl_prop_get_hasrecvd(os)) {
2033 dmu_objset_rele(os, FTAG);
2034 return (ENOTSUP);
2035 }
2036
2037 if (zc->zc_nvlist_dst != 0 &&
2038 (error = dsl_prop_get_received(os, &nv)) == 0) {
2039 error = put_nvlist(zc, nv);
2040 nvlist_free(nv);
2041 }
2042
2043 dmu_objset_rele(os, FTAG);
2044 return (error);
2045 }
2046
2047 static int
2048 nvl_add_zplprop(objset_t *os, nvlist_t *props, zfs_prop_t prop)
2049 {
2050 uint64_t value;
2051 int error;
2052
2053 /*
2054 * zfs_get_zplprop() will either find a value or give us
2055 * the default value (if there is one).
2056 */
2057 if ((error = zfs_get_zplprop(os, prop, &value)) != 0)
2058 return (error);
2059 VERIFY(nvlist_add_uint64(props, zfs_prop_to_name(prop), value) == 0);
2060 return (0);
2061 }
2062
2063 /*
2064 * inputs:
2065 * zc_name name of filesystem
2066 * zc_nvlist_dst_size size of buffer for zpl property nvlist
2067 *
2068 * outputs:
2069 * zc_nvlist_dst zpl property nvlist
2070 * zc_nvlist_dst_size size of zpl property nvlist
2071 */
2072 static int
2073 zfs_ioc_objset_zplprops(zfs_cmd_t *zc)
2074 {
2075 objset_t *os;
2076 int err;
2077
2078 /* XXX reading without owning */
2079 if (err = dmu_objset_hold(zc->zc_name, FTAG, &os))
2080 return (err);
2081
2082 dmu_objset_fast_stat(os, &zc->zc_objset_stats);
2083
2084 /*
2085 * NB: nvl_add_zplprop() will read the objset contents,
2086 * which we aren't supposed to do with a DS_MODE_USER
2087 * hold, because it could be inconsistent.
2088 */
2089 if (zc->zc_nvlist_dst != NULL &&
2090 !zc->zc_objset_stats.dds_inconsistent &&
2091 dmu_objset_type(os) == DMU_OST_ZFS) {
2092 nvlist_t *nv;
2093
2094 VERIFY(nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2095 if ((err = nvl_add_zplprop(os, nv, ZFS_PROP_VERSION)) == 0 &&
2096 (err = nvl_add_zplprop(os, nv, ZFS_PROP_NORMALIZE)) == 0 &&
2097 (err = nvl_add_zplprop(os, nv, ZFS_PROP_UTF8ONLY)) == 0 &&
2098 (err = nvl_add_zplprop(os, nv, ZFS_PROP_CASE)) == 0)
2099 err = put_nvlist(zc, nv);
2100 nvlist_free(nv);
2101 } else {
2102 err = ENOENT;
2103 }
2104 dmu_objset_rele(os, FTAG);
2105 return (err);
2106 }
2107
2108 static boolean_t
2109 dataset_name_hidden(const char *name)
2110 {
2111 /*
2112 * Skip over datasets that are not visible in this zone,
2113 * internal datasets (which have a $ in their name), and
2114 * temporary datasets (which have a % in their name).
2115 */
2116 if (strchr(name, '$') != NULL)
2117 return (B_TRUE);
2118 if (strchr(name, '%') != NULL)
2119 return (B_TRUE);
2120 if (!INGLOBALZONE(curproc) && !zone_dataset_visible(name, NULL))
2121 return (B_TRUE);
2122 return (B_FALSE);
2123 }
2124
2125 /*
2126 * inputs:
2127 * zc_name name of filesystem
2128 * zc_cookie zap cursor
2129 * zc_nvlist_dst_size size of buffer for property nvlist
2130 *
2131 * outputs:
2132 * zc_name name of next filesystem
2133 * zc_cookie zap cursor
2134 * zc_objset_stats stats
2135 * zc_nvlist_dst property nvlist
2136 * zc_nvlist_dst_size size of property nvlist
2137 */
2138 static int
2139 zfs_ioc_dataset_list_next(zfs_cmd_t *zc)
2140 {
2141 objset_t *os;
2142 int error;
2143 char *p;
2144 size_t orig_len = strlen(zc->zc_name);
2145
2146 top:
2147 if (error = dmu_objset_hold(zc->zc_name, FTAG, &os)) {
2148 if (error == ENOENT)
2149 error = ESRCH;
2150 return (error);
2151 }
2152
2153 p = strrchr(zc->zc_name, '/');
2154 if (p == NULL || p[1] != '\0')
2155 (void) strlcat(zc->zc_name, "/", sizeof (zc->zc_name));
2156 p = zc->zc_name + strlen(zc->zc_name);
2157
2158 /*
2159 * Pre-fetch the datasets. dmu_objset_prefetch() always returns 0
2160 * but is not declared void because its called by dmu_objset_find().
2161 */
2162 if (zc->zc_cookie == 0) {
2163 uint64_t cookie = 0;
2164 int len = sizeof (zc->zc_name) - (p - zc->zc_name);
2165
2166 while (dmu_dir_list_next(os, len, p, NULL, &cookie) == 0) {
2167 if (!dataset_name_hidden(zc->zc_name))
2168 (void) dmu_objset_prefetch(zc->zc_name, NULL);
2169 }
2170 }
2171
2172 do {
2173 error = dmu_dir_list_next(os,
2174 sizeof (zc->zc_name) - (p - zc->zc_name), p,
2175 NULL, &zc->zc_cookie);
2176 if (error == ENOENT)
2177 error = ESRCH;
2178 } while (error == 0 && dataset_name_hidden(zc->zc_name));
2179 dmu_objset_rele(os, FTAG);
2180
2181 /*
2182 * If it's an internal dataset (ie. with a '$' in its name),
2183 * don't try to get stats for it, otherwise we'll return ENOENT.
2184 */
2185 if (error == 0 && strchr(zc->zc_name, '$') == NULL) {
2186 error = zfs_ioc_objset_stats(zc); /* fill in the stats */
2187 if (error == ENOENT) {
2188 /* We lost a race with destroy, get the next one. */
2189 zc->zc_name[orig_len] = '\0';
2190 goto top;
2191 }
2192 }
2193 return (error);
2194 }
2195
2196 /*
2197 * inputs:
2198 * zc_name name of filesystem
2199 * zc_cookie zap cursor
2200 * zc_nvlist_dst_size size of buffer for property nvlist
2201 *
2202 * outputs:
2203 * zc_name name of next snapshot
2204 * zc_objset_stats stats
2205 * zc_nvlist_dst property nvlist
2206 * zc_nvlist_dst_size size of property nvlist
2207 */
2208 static int
2209 zfs_ioc_snapshot_list_next(zfs_cmd_t *zc)
2210 {
2211 objset_t *os;
2212 int error;
2213
2214 top:
2215 if (zc->zc_cookie == 0)
2216 (void) dmu_objset_find(zc->zc_name, dmu_objset_prefetch,
2217 NULL, DS_FIND_SNAPSHOTS);
2218
2219 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
2220 if (error)
2221 return (error == ENOENT ? ESRCH : error);
2222
2223 /*
2224 * A dataset name of maximum length cannot have any snapshots,
2225 * so exit immediately.
2226 */
2227 if (strlcat(zc->zc_name, "@", sizeof (zc->zc_name)) >= MAXNAMELEN) {
2228 dmu_objset_rele(os, FTAG);
2229 return (ESRCH);
2230 }
2231
2232 error = dmu_snapshot_list_next(os,
2233 sizeof (zc->zc_name) - strlen(zc->zc_name),
2234 zc->zc_name + strlen(zc->zc_name), &zc->zc_obj, &zc->zc_cookie,
2235 NULL);
2236
2237 if (error == 0) {
2238 dsl_dataset_t *ds;
2239 dsl_pool_t *dp = os->os_dsl_dataset->ds_dir->dd_pool;
2240
2241 /*
2242 * Since we probably don't have a hold on this snapshot,
2243 * it's possible that the objsetid could have been destroyed
2244 * and reused for a new objset. It's OK if this happens during
2245 * a zfs send operation, since the new createtxg will be
2246 * beyond the range we're interested in.
2247 */
2248 rw_enter(&dp->dp_config_rwlock, RW_READER);
2249 error = dsl_dataset_hold_obj(dp, zc->zc_obj, FTAG, &ds);
2250 rw_exit(&dp->dp_config_rwlock);
2251 if (error) {
2252 if (error == ENOENT) {
2253 /* Racing with destroy, get the next one. */
2254 *strchr(zc->zc_name, '@') = '\0';
2255 dmu_objset_rele(os, FTAG);
2256 goto top;
2257 }
2258 } else {
2259 objset_t *ossnap;
2260
2261 error = dmu_objset_from_ds(ds, &ossnap);
2262 if (error == 0)
2263 error = zfs_ioc_objset_stats_impl(zc, ossnap);
2264 dsl_dataset_rele(ds, FTAG);
2265 }
2266 } else if (error == ENOENT) {
2267 error = ESRCH;
2268 }
2269
2270 dmu_objset_rele(os, FTAG);
2271 /* if we failed, undo the @ that we tacked on to zc_name */
2272 if (error)
2273 *strchr(zc->zc_name, '@') = '\0';
2274 return (error);
2275 }
2276
2277 static int
2278 zfs_prop_set_userquota(const char *dsname, nvpair_t *pair)
2279 {
2280 const char *propname = nvpair_name(pair);
2281 uint64_t *valary;
2282 unsigned int vallen;
2283 const char *domain;
2284 char *dash;
2285 zfs_userquota_prop_t type;
2286 uint64_t rid;
2287 uint64_t quota;
2288 zfsvfs_t *zfsvfs;
2289 int err;
2290
2291 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2292 nvlist_t *attrs;
2293 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
2294 if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2295 &pair) != 0)
2296 return (EINVAL);
2297 }
2298
2299 /*
2300 * A correctly constructed propname is encoded as
2301 * userquota@<rid>-<domain>.
2302 */
2303 if ((dash = strchr(propname, '-')) == NULL ||
2304 nvpair_value_uint64_array(pair, &valary, &vallen) != 0 ||
2305 vallen != 3)
2306 return (EINVAL);
2307
2308 domain = dash + 1;
2309 type = valary[0];
2310 rid = valary[1];
2311 quota = valary[2];
2312
2313 err = zfsvfs_hold(dsname, FTAG, &zfsvfs, B_FALSE);
2314 if (err == 0) {
2315 err = zfs_set_userquota(zfsvfs, type, domain, rid, quota);
2316 zfsvfs_rele(zfsvfs, FTAG);
2317 }
2318
2319 return (err);
2320 }
2321
2322 /*
2323 * If the named property is one that has a special function to set its value,
2324 * return 0 on success and a positive error code on failure; otherwise if it is
2325 * not one of the special properties handled by this function, return -1.
2326 *
2327 * XXX: It would be better for callers of the property interface if we handled
2328 * these special cases in dsl_prop.c (in the dsl layer).
2329 */
2330 static int
2331 zfs_prop_set_special(const char *dsname, zprop_source_t source,
2332 nvpair_t *pair)
2333 {
2334 const char *propname = nvpair_name(pair);
2335 zfs_prop_t prop = zfs_name_to_prop(propname);
2336 uint64_t intval;
2337 int err;
2338
2339 if (prop == ZPROP_INVAL) {
2340 if (zfs_prop_userquota(propname))
2341 return (zfs_prop_set_userquota(dsname, pair));
2342 return (-1);
2343 }
2344
2345 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2346 nvlist_t *attrs;
2347 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
2348 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2349 &pair) == 0);
2350 }
2351
2352 if (zfs_prop_get_type(prop) == PROP_TYPE_STRING)
2353 return (-1);
2354
2355 VERIFY(0 == nvpair_value_uint64(pair, &intval));
2356
2357 switch (prop) {
2358 case ZFS_PROP_QUOTA:
2359 err = dsl_dir_set_quota(dsname, source, intval);
2360 break;
2361 case ZFS_PROP_REFQUOTA:
2362 err = dsl_dataset_set_quota(dsname, source, intval);
2363 break;
2364 case ZFS_PROP_RESERVATION:
2365 err = dsl_dir_set_reservation(dsname, source, intval);
2366 break;
2367 case ZFS_PROP_REFRESERVATION:
2368 err = dsl_dataset_set_reservation(dsname, source, intval);
2369 break;
2370 case ZFS_PROP_VOLSIZE:
2371 err = zvol_set_volsize(dsname, ddi_driver_major(zfs_dip),
2372 intval);
2373 break;
2374 case ZFS_PROP_VERSION:
2375 {
2376 zfsvfs_t *zfsvfs;
2377
2378 if ((err = zfsvfs_hold(dsname, FTAG, &zfsvfs, B_TRUE)) != 0)
2379 break;
2380
2381 err = zfs_set_version(zfsvfs, intval);
2382 zfsvfs_rele(zfsvfs, FTAG);
2383
2384 if (err == 0 && intval >= ZPL_VERSION_USERSPACE) {
2385 zfs_cmd_t *zc;
2386
2387 zc = kmem_zalloc(sizeof (zfs_cmd_t), KM_SLEEP);
2388 (void) strcpy(zc->zc_name, dsname);
2389 (void) zfs_ioc_userspace_upgrade(zc);
2390 kmem_free(zc, sizeof (zfs_cmd_t));
2391 }
2392 break;
2393 }
2394 case ZFS_PROP_CHECKSUM:
2395 {
2396 if (intval == ZIO_CHECKSUM_EDONR512_256) {
2397 zfeature_info_t *feature =
2398 &spa_feature_table[SPA_FEATURE_EDONR_CKSUM];
2399 spa_t *spa;
2400 dsl_pool_t *dp;
2401
2402 if ((err = spa_open(dsname, &spa, FTAG)) != 0)
2403 return (err);
2404
2405 dp = spa->spa_dsl_pool;
2406
2407 /*
2408 * Setting the Edon-R checksum algorithm activates
2409 * the feature.
2410 */
2411 if (!spa_feature_is_active(spa, feature)) {
2412 if ((err = zfs_prop_activate_feature(dp,
2413 feature)) != 0) {
2414 spa_close(spa, FTAG);
2415 return (err);
2416 }
2417 }
2418
2419 spa_close(spa, FTAG);
2420 }
2421 /*
2422 * We still want the default set action to be performed in the
2423 * caller, we only performed zfeature settings here.
2424 */
2425 err = -1;
2426 break;
2427 }
2428 case ZFS_PROP_COMPRESSION:
2429 {
2430 if (intval == ZIO_COMPRESS_LZ4) {
2431 zfeature_info_t *feature =
2432 &spa_feature_table[SPA_FEATURE_LZ4_COMPRESS];
2433 spa_t *spa;
2434 dsl_pool_t *dp;
2435
2436 if ((err = spa_open(dsname, &spa, FTAG)) != 0)
2437 return (err);
2438
2439 dp = spa->spa_dsl_pool;
2440
2441 /*
2442 * Setting the LZ4 compression algorithm activates
2443 * the feature.
2444 */
2445 if (!spa_feature_is_active(spa, feature)) {
2446 if ((err = zfs_prop_activate_feature(dp,
2447 feature)) != 0) {
2448 spa_close(spa, FTAG);
2449 return (err);
2450 }
2451 }
2452
2453 spa_close(spa, FTAG);
2454 }
2455 /*
2456 * We still want the default set action to be performed in the
2457 * caller, we only performed zfeature settings here.
2458 */
2459 err = -1;
2460 break;
2461 }
2462
2463 default:
2464 err = -1;
2465 }
2466
2467 return (err);
2468 }
2469
2470 /*
2471 * This function is best effort. If it fails to set any of the given properties,
2472 * it continues to set as many as it can and returns the last error
2473 * encountered. If the caller provides a non-NULL errlist, it will be filled in
2474 * with the list of names of all the properties that failed along with the
2475 * corresponding error numbers.
2476 *
2477 * If every property is set successfully, zero is returned and errlist is not
2478 * modified.
2479 */
2480 int
2481 zfs_set_prop_nvlist(const char *dsname, zprop_source_t source, nvlist_t *nvl,
2482 nvlist_t *errlist)
2483 {
2484 nvpair_t *pair;
2485 nvpair_t *propval;
2486 int rv = 0;
2487 uint64_t intval;
2488 char *strval;
2489 nvlist_t *genericnvl = fnvlist_alloc();
2490 nvlist_t *retrynvl = fnvlist_alloc();
2491
2492 retry:
2493 pair = NULL;
2494 while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) {
2495 const char *propname = nvpair_name(pair);
2496 zfs_prop_t prop = zfs_name_to_prop(propname);
2497 int err = 0;
2498
2499 /* decode the property value */
2500 propval = pair;
2501 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2502 nvlist_t *attrs;
2503 attrs = fnvpair_value_nvlist(pair);
2504 if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2505 &propval) != 0)
2506 err = EINVAL;
2507 }
2508
2509 /* Validate value type */
2510 if (err == 0 && prop == ZPROP_INVAL) {
2511 if (zfs_prop_user(propname)) {
2512 if (nvpair_type(propval) != DATA_TYPE_STRING)
2513 err = EINVAL;
2514 } else if (zfs_prop_userquota(propname)) {
2515 if (nvpair_type(propval) !=
2516 DATA_TYPE_UINT64_ARRAY)
2517 err = EINVAL;
2518 } else {
2519 err = EINVAL;
2520 }
2521 } else if (err == 0) {
2522 if (nvpair_type(propval) == DATA_TYPE_STRING) {
2523 if (zfs_prop_get_type(prop) != PROP_TYPE_STRING)
2524 err = EINVAL;
2525 } else if (nvpair_type(propval) == DATA_TYPE_UINT64) {
2526 const char *unused;
2527
2528 intval = fnvpair_value_uint64(propval);
2529
2530 switch (zfs_prop_get_type(prop)) {
2531 case PROP_TYPE_NUMBER:
2532 break;
2533 case PROP_TYPE_STRING:
2534 err = EINVAL;
2535 break;
2536 case PROP_TYPE_INDEX:
2537 if (zfs_prop_index_to_string(prop,
2538 intval, &unused) != 0)
2539 err = EINVAL;
2540 break;
2541 default:
2542 cmn_err(CE_PANIC,
2543 "unknown property type");
2544 }
2545 } else {
2546 err = EINVAL;
2547 }
2548 }
2549
2550 /* Validate permissions */
2551 if (err == 0)
2552 err = zfs_check_settable(dsname, pair, CRED());
2553
2554 if (err == 0) {
2555 err = zfs_prop_set_special(dsname, source, pair);
2556 if (err == -1) {
2557 /*
2558 * For better performance we build up a list of
2559 * properties to set in a single transaction.
2560 */
2561 err = nvlist_add_nvpair(genericnvl, pair);
2562 } else if (err != 0 && nvl != retrynvl) {
2563 /*
2564 * This may be a spurious error caused by
2565 * receiving quota and reservation out of order.
2566 * Try again in a second pass.
2567 */
2568 err = nvlist_add_nvpair(retrynvl, pair);
2569 }
2570 }
2571
2572 if (err != 0) {
2573 if (errlist != NULL)
2574 fnvlist_add_int32(errlist, propname, err);
2575 rv = err;
2576 }
2577 }
2578
2579 if (nvl != retrynvl && !nvlist_empty(retrynvl)) {
2580 nvl = retrynvl;
2581 goto retry;
2582 }
2583
2584 if (!nvlist_empty(genericnvl) &&
2585 dsl_props_set(dsname, source, genericnvl) != 0) {
2586 /*
2587 * If this fails, we still want to set as many properties as we
2588 * can, so try setting them individually.
2589 */
2590 pair = NULL;
2591 while ((pair = nvlist_next_nvpair(genericnvl, pair)) != NULL) {
2592 const char *propname = nvpair_name(pair);
2593 int err = 0;
2594
2595 propval = pair;
2596 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2597 nvlist_t *attrs;
2598 attrs = fnvpair_value_nvlist(pair);
2599 propval = fnvlist_lookup_nvpair(attrs,
2600 ZPROP_VALUE);
2601 }
2602
2603 if (nvpair_type(propval) == DATA_TYPE_STRING) {
2604 strval = fnvpair_value_string(propval);
2605 err = dsl_prop_set(dsname, propname, source, 1,
2606 strlen(strval) + 1, strval);
2607 } else {
2608 intval = fnvpair_value_uint64(propval);
2609 err = dsl_prop_set(dsname, propname, source, 8,
2610 1, &intval);
2611 }
2612
2613 if (err != 0) {
2614 if (errlist != NULL) {
2615 fnvlist_add_int32(errlist, propname,
2616 err);
2617 }
2618 rv = err;
2619 }
2620 }
2621 }
2622 nvlist_free(genericnvl);
2623 nvlist_free(retrynvl);
2624
2625 return (rv);
2626 }
2627
2628 /*
2629 * Check that all the properties are valid user properties.
2630 */
2631 static int
2632 zfs_check_userprops(const char *fsname, nvlist_t *nvl)
2633 {
2634 nvpair_t *pair = NULL;
2635 int error = 0;
2636
2637 while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) {
2638 const char *propname = nvpair_name(pair);
2639 char *valstr;
2640
2641 if (!zfs_prop_user(propname) ||
2642 nvpair_type(pair) != DATA_TYPE_STRING)
2643 return (EINVAL);
2644
2645 if (error = zfs_secpolicy_write_perms(fsname,
2646 ZFS_DELEG_PERM_USERPROP, CRED()))
2647 return (error);
2648
2649 if (strlen(propname) >= ZAP_MAXNAMELEN)
2650 return (ENAMETOOLONG);
2651
2652 VERIFY(nvpair_value_string(pair, &valstr) == 0);
2653 if (strlen(valstr) >= ZAP_MAXVALUELEN)
2654 return (E2BIG);
2655 }
2656 return (0);
2657 }
2658
2659 static void
2660 props_skip(nvlist_t *props, nvlist_t *skipped, nvlist_t **newprops)
2661 {
2662 nvpair_t *pair;
2663
2664 VERIFY(nvlist_alloc(newprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2665
2666 pair = NULL;
2667 while ((pair = nvlist_next_nvpair(props, pair)) != NULL) {
2668 if (nvlist_exists(skipped, nvpair_name(pair)))
2669 continue;
2670
2671 VERIFY(nvlist_add_nvpair(*newprops, pair) == 0);
2672 }
2673 }
2674
2675 static int
2676 clear_received_props(objset_t *os, const char *fs, nvlist_t *props,
2677 nvlist_t *skipped)
2678 {
2679 int err = 0;
2680 nvlist_t *cleared_props = NULL;
2681 props_skip(props, skipped, &cleared_props);
2682 if (!nvlist_empty(cleared_props)) {
2683 /*
2684 * Acts on local properties until the dataset has received
2685 * properties at least once on or after SPA_VERSION_RECVD_PROPS.
2686 */
2687 zprop_source_t flags = (ZPROP_SRC_NONE |
2688 (dsl_prop_get_hasrecvd(os) ? ZPROP_SRC_RECEIVED : 0));
2689 err = zfs_set_prop_nvlist(fs, flags, cleared_props, NULL);
2690 }
2691 nvlist_free(cleared_props);
2692 return (err);
2693 }
2694
2695 /*
2696 * inputs:
2697 * zc_name name of filesystem
2698 * zc_value name of property to set
2699 * zc_nvlist_src{_size} nvlist of properties to apply
2700 * zc_cookie received properties flag
2701 *
2702 * outputs:
2703 * zc_nvlist_dst{_size} error for each unapplied received property
2704 */
2705 static int
2706 zfs_ioc_set_prop(zfs_cmd_t *zc)
2707 {
2708 nvlist_t *nvl;
2709 boolean_t received = zc->zc_cookie;
2710 zprop_source_t source = (received ? ZPROP_SRC_RECEIVED :
2711 ZPROP_SRC_LOCAL);
2712 nvlist_t *errors;
2713 int error;
2714
2715 if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2716 zc->zc_iflags, &nvl)) != 0)
2717 return (error);
2718
2719 if (received) {
2720 nvlist_t *origprops;
2721 objset_t *os;
2722
2723 if (dmu_objset_hold(zc->zc_name, FTAG, &os) == 0) {
2724 if (dsl_prop_get_received(os, &origprops) == 0) {
2725 (void) clear_received_props(os,
2726 zc->zc_name, origprops, nvl);
2727 nvlist_free(origprops);
2728 }
2729
2730 dsl_prop_set_hasrecvd(os);
2731 dmu_objset_rele(os, FTAG);
2732 }
2733 }
2734
2735 errors = fnvlist_alloc();
2736 error = zfs_set_prop_nvlist(zc->zc_name, source, nvl, errors);
2737
2738 if (zc->zc_nvlist_dst != NULL && errors != NULL) {
2739 (void) put_nvlist(zc, errors);
2740 }
2741
2742 nvlist_free(errors);
2743 nvlist_free(nvl);
2744 return (error);
2745 }
2746
2747 /*
2748 * inputs:
2749 * zc_name name of filesystem
2750 * zc_value name of property to inherit
2751 * zc_cookie revert to received value if TRUE
2752 *
2753 * outputs: none
2754 */
2755 static int
2756 zfs_ioc_inherit_prop(zfs_cmd_t *zc)
2757 {
2758 const char *propname = zc->zc_value;
2759 zfs_prop_t prop = zfs_name_to_prop(propname);
2760 boolean_t received = zc->zc_cookie;
2761 zprop_source_t source = (received
2762 ? ZPROP_SRC_NONE /* revert to received value, if any */
2763 : ZPROP_SRC_INHERITED); /* explicitly inherit */
2764
2765 if (received) {
2766 nvlist_t *dummy;
2767 nvpair_t *pair;
2768 zprop_type_t type;
2769 int err;
2770
2771 /*
2772 * zfs_prop_set_special() expects properties in the form of an
2773 * nvpair with type info.
2774 */
2775 if (prop == ZPROP_INVAL) {
2776 if (!zfs_prop_user(propname))
2777 return (EINVAL);
2778
2779 type = PROP_TYPE_STRING;
2780 } else if (prop == ZFS_PROP_VOLSIZE ||
2781 prop == ZFS_PROP_VERSION) {
2782 return (EINVAL);
2783 } else {
2784 type = zfs_prop_get_type(prop);
2785 }
2786
2787 VERIFY(nvlist_alloc(&dummy, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2788
2789 switch (type) {
2790 case PROP_TYPE_STRING:
2791 VERIFY(0 == nvlist_add_string(dummy, propname, ""));
2792 break;
2793 case PROP_TYPE_NUMBER:
2794 case PROP_TYPE_INDEX:
2795 VERIFY(0 == nvlist_add_uint64(dummy, propname, 0));
2796 break;
2797 default:
2798 nvlist_free(dummy);
2799 return (EINVAL);
2800 }
2801
2802 pair = nvlist_next_nvpair(dummy, NULL);
2803 err = zfs_prop_set_special(zc->zc_name, source, pair);
2804 nvlist_free(dummy);
2805 if (err != -1)
2806 return (err); /* special property already handled */
2807 } else {
2808 /*
2809 * Only check this in the non-received case. We want to allow
2810 * 'inherit -S' to revert non-inheritable properties like quota
2811 * and reservation to the received or default values even though
2812 * they are not considered inheritable.
2813 */
2814 if (prop != ZPROP_INVAL && !zfs_prop_inheritable(prop))
2815 return (EINVAL);
2816 }
2817
2818 /* property name has been validated by zfs_secpolicy_inherit_prop() */
2819 return (dsl_prop_set(zc->zc_name, zc->zc_value, source, 0, 0, NULL));
2820 }
2821
2822 static int
2823 zfs_ioc_pool_set_props(zfs_cmd_t *zc)
2824 {
2825 nvlist_t *props;
2826 spa_t *spa;
2827 int error;
2828 nvpair_t *pair;
2829
2830 if (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2831 zc->zc_iflags, &props))
2832 return (error);
2833
2834 /*
2835 * If the only property is the configfile, then just do a spa_lookup()
2836 * to handle the faulted case.
2837 */
2838 pair = nvlist_next_nvpair(props, NULL);
2839 if (pair != NULL && strcmp(nvpair_name(pair),
2840 zpool_prop_to_name(ZPOOL_PROP_CACHEFILE)) == 0 &&
2841 nvlist_next_nvpair(props, pair) == NULL) {
2842 mutex_enter(&spa_namespace_lock);
2843 if ((spa = spa_lookup(zc->zc_name)) != NULL) {
2844 spa_configfile_set(spa, props, B_FALSE);
2845 spa_config_sync(spa, B_FALSE, B_TRUE);
2846 }
2847 mutex_exit(&spa_namespace_lock);
2848 if (spa != NULL) {
2849 nvlist_free(props);
2850 return (0);
2851 }
2852 }
2853
2854 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) {
2855 nvlist_free(props);
2856 return (error);
2857 }
2858
2859 error = spa_prop_set(spa, props);
2860
2861 nvlist_free(props);
2862 spa_close(spa, FTAG);
2863
2864 return (error);
2865 }
2866
2867 static int
2868 zfs_ioc_pool_get_props(zfs_cmd_t *zc)
2869 {
2870 spa_t *spa;
2871 int error;
2872 nvlist_t *nvp = NULL;
2873
2874 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) {
2875 /*
2876 * If the pool is faulted, there may be properties we can still
2877 * get (such as altroot and cachefile), so attempt to get them
2878 * anyway.
2879 */
2880 mutex_enter(&spa_namespace_lock);
2881 if ((spa = spa_lookup(zc->zc_name)) != NULL)
2882 error = spa_prop_get(spa, &nvp);
2883 mutex_exit(&spa_namespace_lock);
2884 } else {
2885 error = spa_prop_get(spa, &nvp);
2886 spa_close(spa, FTAG);
2887 }
2888
2889 if (error == 0 && zc->zc_nvlist_dst != NULL)
2890 error = put_nvlist(zc, nvp);
2891 else
2892 error = EFAULT;
2893
2894 nvlist_free(nvp);
2895 return (error);
2896 }
2897
2898 /*
2899 * inputs:
2900 * zc_name name of filesystem
2901 * zc_nvlist_src{_size} nvlist of delegated permissions
2902 * zc_perm_action allow/unallow flag
2903 *
2904 * outputs: none
2905 */
2906 static int
2907 zfs_ioc_set_fsacl(zfs_cmd_t *zc)
2908 {
2909 int error;
2910 nvlist_t *fsaclnv = NULL;
2911
2912 if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2913 zc->zc_iflags, &fsaclnv)) != 0)
2914 return (error);
2915
2916 /*
2917 * Verify nvlist is constructed correctly
2918 */
2919 if ((error = zfs_deleg_verify_nvlist(fsaclnv)) != 0) {
2920 nvlist_free(fsaclnv);
2921 return (EINVAL);
2922 }
2923
2924 /*
2925 * If we don't have PRIV_SYS_MOUNT, then validate
2926 * that user is allowed to hand out each permission in
2927 * the nvlist(s)
2928 */
2929
2930 error = secpolicy_zfs(CRED());
2931 if (error) {
2932 if (zc->zc_perm_action == B_FALSE) {
2933 error = dsl_deleg_can_allow(zc->zc_name,
2934 fsaclnv, CRED());
2935 } else {
2936 error = dsl_deleg_can_unallow(zc->zc_name,
2937 fsaclnv, CRED());
2938 }
2939 }
2940
2941 if (error == 0)
2942 error = dsl_deleg_set(zc->zc_name, fsaclnv, zc->zc_perm_action);
2943
2944 nvlist_free(fsaclnv);
2945 return (error);
2946 }
2947
2948 /*
2949 * inputs:
2950 * zc_name name of filesystem
2951 *
2952 * outputs:
2953 * zc_nvlist_src{_size} nvlist of delegated permissions
2954 */
2955 static int
2956 zfs_ioc_get_fsacl(zfs_cmd_t *zc)
2957 {
2958 nvlist_t *nvp;
2959 int error;
2960
2961 if ((error = dsl_deleg_get(zc->zc_name, &nvp)) == 0) {
2962 error = put_nvlist(zc, nvp);
2963 nvlist_free(nvp);
2964 }
2965
2966 return (error);
2967 }
2968
2969 /*
2970 * Search the vfs list for a specified resource. Returns a pointer to it
2971 * or NULL if no suitable entry is found. The caller of this routine
2972 * is responsible for releasing the returned vfs pointer.
2973 */
2974 static vfs_t *
2975 zfs_get_vfs(const char *resource)
2976 {
2977 struct vfs *vfsp;
2978 struct vfs *vfs_found = NULL;
2979
2980 vfs_list_read_lock();
2981 vfsp = rootvfs;
2982 do {
2983 if (strcmp(refstr_value(vfsp->vfs_resource), resource) == 0) {
2984 VFS_HOLD(vfsp);
2985 vfs_found = vfsp;
2986 break;
2987 }
2988 vfsp = vfsp->vfs_next;
2989 } while (vfsp != rootvfs);
2990 vfs_list_unlock();
2991 return (vfs_found);
2992 }
2993
2994 /* ARGSUSED */
2995 static void
2996 zfs_create_cb(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx)
2997 {
2998 zfs_creat_t *zct = arg;
2999
3000 zfs_create_fs(os, cr, zct->zct_zplprops, tx);
3001 }
3002
3003 #define ZFS_PROP_UNDEFINED ((uint64_t)-1)
3004
3005 /*
3006 * inputs:
3007 * createprops list of properties requested by creator
3008 * default_zplver zpl version to use if unspecified in createprops
3009 * fuids_ok fuids allowed in this version of the spa?
3010 * os parent objset pointer (NULL if root fs)
3011 *
3012 * outputs:
3013 * zplprops values for the zplprops we attach to the master node object
3014 * is_ci true if requested file system will be purely case-insensitive
3015 *
3016 * Determine the settings for utf8only, normalization and
3017 * casesensitivity. Specific values may have been requested by the
3018 * creator and/or we can inherit values from the parent dataset. If
3019 * the file system is of too early a vintage, a creator can not
3020 * request settings for these properties, even if the requested
3021 * setting is the default value. We don't actually want to create dsl
3022 * properties for these, so remove them from the source nvlist after
3023 * processing.
3024 */
3025 static int
3026 zfs_fill_zplprops_impl(objset_t *os, uint64_t zplver,
3027 boolean_t fuids_ok, boolean_t sa_ok, nvlist_t *createprops,
3028 nvlist_t *zplprops, boolean_t *is_ci)
3029 {
3030 uint64_t sense = ZFS_PROP_UNDEFINED;
3031 uint64_t norm = ZFS_PROP_UNDEFINED;
3032 uint64_t u8 = ZFS_PROP_UNDEFINED;
3033
3034 ASSERT(zplprops != NULL);
3035
3036 /*
3037 * Pull out creator prop choices, if any.
3038 */
3039 if (createprops) {
3040 (void) nvlist_lookup_uint64(createprops,
3041 zfs_prop_to_name(ZFS_PROP_VERSION), &zplver);
3042 (void) nvlist_lookup_uint64(createprops,
3043 zfs_prop_to_name(ZFS_PROP_NORMALIZE), &norm);
3044 (void) nvlist_remove_all(createprops,
3045 zfs_prop_to_name(ZFS_PROP_NORMALIZE));
3046 (void) nvlist_lookup_uint64(createprops,
3047 zfs_prop_to_name(ZFS_PROP_UTF8ONLY), &u8);
3048 (void) nvlist_remove_all(createprops,
3049 zfs_prop_to_name(ZFS_PROP_UTF8ONLY));
3050 (void) nvlist_lookup_uint64(createprops,
3051 zfs_prop_to_name(ZFS_PROP_CASE), &sense);
3052 (void) nvlist_remove_all(createprops,
3053 zfs_prop_to_name(ZFS_PROP_CASE));
3054 }
3055
3056 /*
3057 * If the zpl version requested is whacky or the file system
3058 * or pool is version is too "young" to support normalization
3059 * and the creator tried to set a value for one of the props,
3060 * error out.
3061 */
3062 if ((zplver < ZPL_VERSION_INITIAL || zplver > ZPL_VERSION) ||
3063 (zplver >= ZPL_VERSION_FUID && !fuids_ok) ||
3064 (zplver >= ZPL_VERSION_SA && !sa_ok) ||
3065 (zplver < ZPL_VERSION_NORMALIZATION &&
3066 (norm != ZFS_PROP_UNDEFINED || u8 != ZFS_PROP_UNDEFINED ||
3067 sense != ZFS_PROP_UNDEFINED)))
3068 return (ENOTSUP);
3069
3070 /*
3071 * Put the version in the zplprops
3072 */
3073 VERIFY(nvlist_add_uint64(zplprops,
3074 zfs_prop_to_name(ZFS_PROP_VERSION), zplver) == 0);
3075
3076 if (norm == ZFS_PROP_UNDEFINED)
3077 VERIFY(zfs_get_zplprop(os, ZFS_PROP_NORMALIZE, &norm) == 0);
3078 VERIFY(nvlist_add_uint64(zplprops,
3079 zfs_prop_to_name(ZFS_PROP_NORMALIZE), norm) == 0);
3080
3081 /*
3082 * If we're normalizing, names must always be valid UTF-8 strings.
3083 */
3084 if (norm)
3085 u8 = 1;
3086 if (u8 == ZFS_PROP_UNDEFINED)
3087 VERIFY(zfs_get_zplprop(os, ZFS_PROP_UTF8ONLY, &u8) == 0);
3088 VERIFY(nvlist_add_uint64(zplprops,
3089 zfs_prop_to_name(ZFS_PROP_UTF8ONLY), u8) == 0);
3090
3091 if (sense == ZFS_PROP_UNDEFINED)
3092 VERIFY(zfs_get_zplprop(os, ZFS_PROP_CASE, &sense) == 0);
3093 VERIFY(nvlist_add_uint64(zplprops,
3094 zfs_prop_to_name(ZFS_PROP_CASE), sense) == 0);
3095
3096 if (is_ci)
3097 *is_ci = (sense == ZFS_CASE_INSENSITIVE);
3098
3099 return (0);
3100 }
3101
3102 static int
3103 zfs_fill_zplprops(const char *dataset, nvlist_t *createprops,
3104 nvlist_t *zplprops, boolean_t *is_ci)
3105 {
3106 boolean_t fuids_ok, sa_ok;
3107 uint64_t zplver = ZPL_VERSION;
3108 objset_t *os = NULL;
3109 char parentname[MAXNAMELEN];
3110 char *cp;
3111 spa_t *spa;
3112 uint64_t spa_vers;
3113 int error;
3114
3115 (void) strlcpy(parentname, dataset, sizeof (parentname));
3116 cp = strrchr(parentname, '/');
3117 ASSERT(cp != NULL);
3118 cp[0] = '\0';
3119
3120 if ((error = spa_open(dataset, &spa, FTAG)) != 0)
3121 return (error);
3122
3123 spa_vers = spa_version(spa);
3124 spa_close(spa, FTAG);
3125
3126 zplver = zfs_zpl_version_map(spa_vers);
3127 fuids_ok = (zplver >= ZPL_VERSION_FUID);
3128 sa_ok = (zplver >= ZPL_VERSION_SA);
3129
3130 /*
3131 * Open parent object set so we can inherit zplprop values.
3132 */
3133 if ((error = dmu_objset_hold(parentname, FTAG, &os)) != 0)
3134 return (error);
3135
3136 error = zfs_fill_zplprops_impl(os, zplver, fuids_ok, sa_ok, createprops,
3137 zplprops, is_ci);
3138 dmu_objset_rele(os, FTAG);
3139 return (error);
3140 }
3141
3142 static int
3143 zfs_fill_zplprops_root(uint64_t spa_vers, nvlist_t *createprops,
3144 nvlist_t *zplprops, boolean_t *is_ci)
3145 {
3146 boolean_t fuids_ok;
3147 boolean_t sa_ok;
3148 uint64_t zplver = ZPL_VERSION;
3149 int error;
3150
3151 zplver = zfs_zpl_version_map(spa_vers);
3152 fuids_ok = (zplver >= ZPL_VERSION_FUID);
3153 sa_ok = (zplver >= ZPL_VERSION_SA);
3154
3155 error = zfs_fill_zplprops_impl(NULL, zplver, fuids_ok, sa_ok,
3156 createprops, zplprops, is_ci);
3157 return (error);
3158 }
3159
3160 /*
3161 * innvl: {
3162 * "type" -> dmu_objset_type_t (int32)
3163 * (optional) "props" -> { prop -> value }
3164 * }
3165 *
3166 * outnvl: propname -> error code (int32)
3167 */
3168 static int
3169 zfs_ioc_create(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3170 {
3171 int error = 0;
3172 zfs_creat_t zct = { 0 };
3173 nvlist_t *nvprops = NULL;
3174 void (*cbfunc)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx);
3175 int32_t type32;
3176 dmu_objset_type_t type;
3177 boolean_t is_insensitive = B_FALSE;
3178
3179 if (nvlist_lookup_int32(innvl, "type", &type32) != 0)
3180 return (EINVAL);
3181 type = type32;
3182 (void) nvlist_lookup_nvlist(innvl, "props", &nvprops);
3183
3184 switch (type) {
3185 case DMU_OST_ZFS:
3186 cbfunc = zfs_create_cb;
3187 break;
3188
3189 case DMU_OST_ZVOL:
3190 cbfunc = zvol_create_cb;
3191 break;
3192
3193 default:
3194 cbfunc = NULL;
3195 break;
3196 }
3197 if (strchr(fsname, '@') ||
3198 strchr(fsname, '%'))
3199 return (EINVAL);
3200
3201 zct.zct_props = nvprops;
3202
3203 if (cbfunc == NULL)
3204 return (EINVAL);
3205
3206 if (type == DMU_OST_ZVOL) {
3207 uint64_t volsize, volblocksize;
3208
3209 if (nvprops == NULL)
3210 return (EINVAL);
3211 if (nvlist_lookup_uint64(nvprops,
3212 zfs_prop_to_name(ZFS_PROP_VOLSIZE), &volsize) != 0)
3213 return (EINVAL);
3214
3215 if ((error = nvlist_lookup_uint64(nvprops,
3216 zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE),
3217 &volblocksize)) != 0 && error != ENOENT)
3218 return (EINVAL);
3219
3220 if (error != 0)
3221 volblocksize = zfs_prop_default_numeric(
3222 ZFS_PROP_VOLBLOCKSIZE);
3223
3224 if ((error = zvol_check_volblocksize(
3225 volblocksize)) != 0 ||
3226 (error = zvol_check_volsize(volsize,
3227 volblocksize)) != 0)
3228 return (error);
3229 } else if (type == DMU_OST_ZFS) {
3230 int error;
3231
3232 /*
3233 * We have to have normalization and
3234 * case-folding flags correct when we do the
3235 * file system creation, so go figure them out
3236 * now.
3237 */
3238 VERIFY(nvlist_alloc(&zct.zct_zplprops,
3239 NV_UNIQUE_NAME, KM_SLEEP) == 0);
3240 error = zfs_fill_zplprops(fsname, nvprops,
3241 zct.zct_zplprops, &is_insensitive);
3242 if (error != 0) {
3243 nvlist_free(zct.zct_zplprops);
3244 return (error);
3245 }
3246 }
3247
3248 error = dmu_objset_create(fsname, type,
3249 is_insensitive ? DS_FLAG_CI_DATASET : 0, cbfunc, &zct);
3250 nvlist_free(zct.zct_zplprops);
3251
3252 /*
3253 * It would be nice to do this atomically.
3254 */
3255 if (error == 0) {
3256 error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL,
3257 nvprops, outnvl);
3258 if (error != 0)
3259 (void) dmu_objset_destroy(fsname, B_FALSE);
3260 }
3261 return (error);
3262 }
3263
3264 /*
3265 * innvl: {
3266 * "origin" -> name of origin snapshot
3267 * (optional) "props" -> { prop -> value }
3268 * }
3269 *
3270 * outnvl: propname -> error code (int32)
3271 */
3272 static int
3273 zfs_ioc_clone(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3274 {
3275 int error = 0;
3276 nvlist_t *nvprops = NULL;
3277 char *origin_name;
3278 dsl_dataset_t *origin;
3279
3280 if (nvlist_lookup_string(innvl, "origin", &origin_name) != 0)
3281 return (EINVAL);
3282 (void) nvlist_lookup_nvlist(innvl, "props", &nvprops);
3283
3284 if (strchr(fsname, '@') ||
3285 strchr(fsname, '%'))
3286 return (EINVAL);
3287
3288 if (dataset_namecheck(origin_name, NULL, NULL) != 0)
3289 return (EINVAL);
3290
3291 error = dsl_dataset_hold(origin_name, FTAG, &origin);
3292 if (error)
3293 return (error);
3294
3295 error = dmu_objset_clone(fsname, origin, 0);
3296 dsl_dataset_rele(origin, FTAG);
3297 if (error)
3298 return (error);
3299
3300 /*
3301 * It would be nice to do this atomically.
3302 */
3303 if (error == 0) {
3304 error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL,
3305 nvprops, outnvl);
3306 if (error != 0)
3307 (void) dmu_objset_destroy(fsname, B_FALSE);
3308 }
3309 return (error);
3310 }
3311
3312 /*
3313 * innvl: {
3314 * "snaps" -> { snapshot1, snapshot2 }
3315 * (optional) "props" -> { prop -> value (string) }
3316 * }
3317 *
3318 * outnvl: snapshot -> error code (int32)
3319 *
3320 */
3321 static int
3322 zfs_ioc_snapshot(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3323 {
3324 nvlist_t *snaps;
3325 nvlist_t *props = NULL;
3326 int error, poollen;
3327 nvpair_t *pair;
3328
3329 (void) nvlist_lookup_nvlist(innvl, "props", &props);
3330 if ((error = zfs_check_userprops(poolname, props)) != 0)
3331 return (error);
3332
3333 if (!nvlist_empty(props) &&
3334 zfs_earlier_version(poolname, SPA_VERSION_SNAP_PROPS))
3335 return (ENOTSUP);
3336
3337 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
3338 return (EINVAL);
3339 poollen = strlen(poolname);
3340 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
3341 pair = nvlist_next_nvpair(snaps, pair)) {
3342 const char *name = nvpair_name(pair);
3343 const char *cp = strchr(name, '@');
3344
3345 /*
3346 * The snap name must contain an @, and the part after it must
3347 * contain only valid characters.
3348 */
3349 if (cp == NULL || snapshot_namecheck(cp + 1, NULL, NULL) != 0)
3350 return (EINVAL);
3351
3352 /*
3353 * The snap must be in the specified pool.
3354 */
3355 if (strncmp(name, poolname, poollen) != 0 ||
3356 (name[poollen] != '/' && name[poollen] != '@'))
3357 return (EXDEV);
3358
3359 /* This must be the only snap of this fs. */
3360 for (nvpair_t *pair2 = nvlist_next_nvpair(snaps, pair);
3361 pair2 != NULL; pair2 = nvlist_next_nvpair(snaps, pair2)) {
3362 if (strncmp(name, nvpair_name(pair2), cp - name + 1)
3363 == 0) {
3364 return (EXDEV);
3365 }
3366 }
3367 }
3368
3369 error = dmu_objset_snapshot(snaps, props, outnvl);
3370 return (error);
3371 }
3372
3373 /*
3374 * innvl: "message" -> string
3375 */
3376 /* ARGSUSED */
3377 static int
3378 zfs_ioc_log_history(const char *unused, nvlist_t *innvl, nvlist_t *outnvl)
3379 {
3380 char *message;
3381 spa_t *spa;
3382 int error;
3383 char *poolname;
3384
3385 /*
3386 * The poolname in the ioctl is not set, we get it from the TSD,
3387 * which was set at the end of the last successful ioctl that allows
3388 * logging. The secpolicy func already checked that it is set.
3389 * Only one log ioctl is allowed after each successful ioctl, so
3390 * we clear the TSD here.
3391 */
3392 poolname = tsd_get(zfs_allow_log_key);
3393 (void) tsd_set(zfs_allow_log_key, NULL);
3394 error = spa_open(poolname, &spa, FTAG);
3395 strfree(poolname);
3396 if (error != 0)
3397 return (error);
3398
3399 if (nvlist_lookup_string(innvl, "message", &message) != 0) {
3400 spa_close(spa, FTAG);
3401 return (EINVAL);
3402 }
3403
3404 if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) {
3405 spa_close(spa, FTAG);
3406 return (ENOTSUP);
3407 }
3408
3409 error = spa_history_log(spa, message);
3410 spa_close(spa, FTAG);
3411 return (error);
3412 }
3413
3414 /* ARGSUSED */
3415 int
3416 zfs_unmount_snap(const char *name, void *arg)
3417 {
3418 vfs_t *vfsp;
3419 int err;
3420
3421 if (strchr(name, '@') == NULL)
3422 return (0);
3423
3424 vfsp = zfs_get_vfs(name);
3425 if (vfsp == NULL)
3426 return (0);
3427
3428 if ((err = vn_vfswlock(vfsp->vfs_vnodecovered)) != 0) {
3429 VFS_RELE(vfsp);
3430 return (err);
3431 }
3432 VFS_RELE(vfsp);
3433
3434 /*
3435 * Always force the unmount for snapshots.
3436 */
3437 return (dounmount(vfsp, MS_FORCE, kcred));
3438 }
3439
3440 /*
3441 * innvl: {
3442 * "snaps" -> { snapshot1, snapshot2 }
3443 * (optional boolean) "defer"
3444 * }
3445 *
3446 * outnvl: snapshot -> error code (int32)
3447 *
3448 */
3449 static int
3450 zfs_ioc_destroy_snaps(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3451 {
3452 int poollen;
3453 nvlist_t *snaps;
3454 nvpair_t *pair;
3455 boolean_t defer;
3456
3457 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
3458 return (EINVAL);
3459 defer = nvlist_exists(innvl, "defer");
3460
3461 poollen = strlen(poolname);
3462 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
3463 pair = nvlist_next_nvpair(snaps, pair)) {
3464 const char *name = nvpair_name(pair);
3465
3466 /*
3467 * The snap must be in the specified pool.
3468 */
3469 if (strncmp(name, poolname, poollen) != 0 ||
3470 (name[poollen] != '/' && name[poollen] != '@'))
3471 return (EXDEV);
3472
3473 /*
3474 * Ignore failures to unmount; dmu_snapshots_destroy_nvl()
3475 * will deal with this gracefully (by filling in outnvl).
3476 */
3477 (void) zfs_unmount_snap(name, NULL);
3478 }
3479
3480 return (dmu_snapshots_destroy_nvl(snaps, defer, outnvl));
3481 }
3482
3483 /*
3484 * inputs:
3485 * zc_name name of dataset to destroy
3486 * zc_objset_type type of objset
3487 * zc_defer_destroy mark for deferred destroy
3488 *
3489 * outputs: none
3490 */
3491 static int
3492 zfs_ioc_destroy(zfs_cmd_t *zc)
3493 {
3494 int err;
3495 if (strchr(zc->zc_name, '@') && zc->zc_objset_type == DMU_OST_ZFS) {
3496 err = zfs_unmount_snap(zc->zc_name, NULL);
3497 if (err)
3498 return (err);
3499 }
3500
3501 err = dmu_objset_destroy(zc->zc_name, zc->zc_defer_destroy);
3502 if (zc->zc_objset_type == DMU_OST_ZVOL && err == 0)
3503 (void) zvol_remove_minor(zc->zc_name);
3504 return (err);
3505 }
3506
3507 /*
3508 * inputs:
3509 * zc_name name of dataset to rollback (to most recent snapshot)
3510 *
3511 * outputs: none
3512 */
3513 static int
3514 zfs_ioc_rollback(zfs_cmd_t *zc)
3515 {
3516 dsl_dataset_t *ds, *clone;
3517 int error;
3518 zfsvfs_t *zfsvfs;
3519 char *clone_name;
3520
3521 error = dsl_dataset_hold(zc->zc_name, FTAG, &ds);
3522 if (error)
3523 return (error);
3524
3525 /* must not be a snapshot */
3526 if (dsl_dataset_is_snapshot(ds)) {
3527 dsl_dataset_rele(ds, FTAG);
3528 return (EINVAL);
3529 }
3530
3531 /* must have a most recent snapshot */
3532 if (ds->ds_phys->ds_prev_snap_txg < TXG_INITIAL) {
3533 dsl_dataset_rele(ds, FTAG);
3534 return (EINVAL);
3535 }
3536
3537 /*
3538 * Create clone of most recent snapshot.
3539 */
3540 clone_name = kmem_asprintf("%s/%%rollback", zc->zc_name);
3541 error = dmu_objset_clone(clone_name, ds->ds_prev, DS_FLAG_INCONSISTENT);
3542 if (error)
3543 goto out;
3544
3545 error = dsl_dataset_own(clone_name, B_TRUE, FTAG, &clone);
3546 if (error)
3547 goto out;
3548
3549 /*
3550 * Do clone swap.
3551 */
3552 if (getzfsvfs(zc->zc_name, &zfsvfs) == 0) {
3553 error = zfs_suspend_fs(zfsvfs);
3554 if (error == 0) {
3555 int resume_err;
3556
3557 if (dsl_dataset_tryown(ds, B_FALSE, FTAG)) {
3558 error = dsl_dataset_clone_swap(clone, ds,
3559 B_TRUE);
3560 dsl_dataset_disown(ds, FTAG);
3561 ds = NULL;
3562 } else {
3563 error = EBUSY;
3564 }
3565 resume_err = zfs_resume_fs(zfsvfs, zc->zc_name);
3566 error = error ? error : resume_err;
3567 }
3568 VFS_RELE(zfsvfs->z_vfs);
3569 } else {
3570 if (dsl_dataset_tryown(ds, B_FALSE, FTAG)) {
3571 error = dsl_dataset_clone_swap(clone, ds, B_TRUE);
3572 dsl_dataset_disown(ds, FTAG);
3573 ds = NULL;
3574 } else {
3575 error = EBUSY;
3576 }
3577 }
3578
3579 /*
3580 * Destroy clone (which also closes it).
3581 */
3582 (void) dsl_dataset_destroy(clone, FTAG, B_FALSE);
3583
3584 out:
3585 strfree(clone_name);
3586 if (ds)
3587 dsl_dataset_rele(ds, FTAG);
3588 return (error);
3589 }
3590
3591 /*
3592 * inputs:
3593 * zc_name old name of dataset
3594 * zc_value new name of dataset
3595 * zc_cookie recursive flag (only valid for snapshots)
3596 *
3597 * outputs: none
3598 */
3599 static int
3600 zfs_ioc_rename(zfs_cmd_t *zc)
3601 {
3602 boolean_t recursive = zc->zc_cookie & 1;
3603
3604 zc->zc_value[sizeof (zc->zc_value) - 1] = '\0';
3605 if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
3606 strchr(zc->zc_value, '%'))
3607 return (EINVAL);
3608
3609 /*
3610 * Unmount snapshot unless we're doing a recursive rename,
3611 * in which case the dataset code figures out which snapshots
3612 * to unmount.
3613 */
3614 if (!recursive && strchr(zc->zc_name, '@') != NULL &&
3615 zc->zc_objset_type == DMU_OST_ZFS) {
3616 int err = zfs_unmount_snap(zc->zc_name, NULL);
3617 if (err)
3618 return (err);
3619 }
3620 if (zc->zc_objset_type == DMU_OST_ZVOL)
3621 (void) zvol_remove_minor(zc->zc_name);
3622 return (dmu_objset_rename(zc->zc_name, zc->zc_value, recursive));
3623 }
3624
3625 static int
3626 zfs_check_settable(const char *dsname, nvpair_t *pair, cred_t *cr)
3627 {
3628 const char *propname = nvpair_name(pair);
3629 boolean_t issnap = (strchr(dsname, '@') != NULL);
3630 zfs_prop_t prop = zfs_name_to_prop(propname);
3631 uint64_t intval;
3632 int err;
3633
3634 if (prop == ZPROP_INVAL) {
3635 if (zfs_prop_user(propname)) {
3636 if (err = zfs_secpolicy_write_perms(dsname,
3637 ZFS_DELEG_PERM_USERPROP, cr))
3638 return (err);
3639 return (0);
3640 }
3641
3642 if (!issnap && zfs_prop_userquota(propname)) {
3643 const char *perm = NULL;
3644 const char *uq_prefix =
3645 zfs_userquota_prop_prefixes[ZFS_PROP_USERQUOTA];
3646 const char *gq_prefix =
3647 zfs_userquota_prop_prefixes[ZFS_PROP_GROUPQUOTA];
3648
3649 if (strncmp(propname, uq_prefix,
3650 strlen(uq_prefix)) == 0) {
3651 perm = ZFS_DELEG_PERM_USERQUOTA;
3652 } else if (strncmp(propname, gq_prefix,
3653 strlen(gq_prefix)) == 0) {
3654 perm = ZFS_DELEG_PERM_GROUPQUOTA;
3655 } else {
3656 /* USERUSED and GROUPUSED are read-only */
3657 return (EINVAL);
3658 }
3659
3660 if (err = zfs_secpolicy_write_perms(dsname, perm, cr))
3661 return (err);
3662 return (0);
3663 }
3664
3665 return (EINVAL);
3666 }
3667
3668 if (issnap)
3669 return (EINVAL);
3670
3671 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
3672 /*
3673 * dsl_prop_get_all_impl() returns properties in this
3674 * format.
3675 */
3676 nvlist_t *attrs;
3677 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
3678 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
3679 &pair) == 0);
3680 }
3681
3682 /*
3683 * Check that this value is valid for this pool version
3684 */
3685 switch (prop) {
3686 case ZFS_PROP_CHECKSUM:
3687 if (nvpair_type(pair) == DATA_TYPE_UINT64 &&
3688 nvpair_value_uint64(pair, &intval) == 0) {
3689 if (intval == ZIO_CHECKSUM_EDONR512_256) {
3690 zfeature_info_t *feature =
3691 &spa_feature_table[SPA_FEATURE_EDONR_CKSUM];
3692 spa_t *spa;
3693
3694 if ((err = spa_open(dsname, &spa, FTAG)) != 0)
3695 return (err);
3696
3697 if (!spa_feature_is_enabled(spa, feature)) {
3698 spa_close(spa, FTAG);
3699 return (ENOTSUP);
3700 }
3701 spa_close(spa, FTAG);
3702 }
3703 }
3704 break;
3705 case ZFS_PROP_COMPRESSION:
3706 /*
3707 * If the user specified gzip compression, make sure
3708 * the SPA supports it. We ignore any errors here since
3709 * we'll catch them later.
3710 */
3711 if (nvpair_type(pair) == DATA_TYPE_UINT64 &&
3712 nvpair_value_uint64(pair, &intval) == 0) {
3713 if (intval >= ZIO_COMPRESS_GZIP_1 &&
3714 intval <= ZIO_COMPRESS_GZIP_9 &&
3715 zfs_earlier_version(dsname,
3716 SPA_VERSION_GZIP_COMPRESSION)) {
3717 return (ENOTSUP);
3718 }
3719
3720 if (intval == ZIO_COMPRESS_ZLE &&
3721 zfs_earlier_version(dsname,
3722 SPA_VERSION_ZLE_COMPRESSION))
3723 return (ENOTSUP);
3724
3725 if (intval == ZIO_COMPRESS_LZ4) {
3726 zfeature_info_t *feature =
3727 &spa_feature_table[
3728 SPA_FEATURE_LZ4_COMPRESS];
3729 spa_t *spa;
3730
3731 if ((err = spa_open(dsname, &spa, FTAG)) != 0)
3732 return (err);
3733
3734 if (!spa_feature_is_enabled(spa, feature)) {
3735 spa_close(spa, FTAG);
3736 return (ENOTSUP);
3737 }
3738 spa_close(spa, FTAG);
3739 }
3740
3741 /*
3742 * If this is a bootable dataset then
3743 * verify that the compression algorithm
3744 * is supported for booting. We must return
3745 * something other than ENOTSUP since it
3746 * implies a downrev pool version.
3747 */
3748 if (zfs_is_bootfs(dsname) &&
3749 !BOOTFS_COMPRESS_VALID(intval)) {
3750 return (ERANGE);
3751 }
3752 }
3753 break;
3754
3755 case ZFS_PROP_COPIES:
3756 if (zfs_earlier_version(dsname, SPA_VERSION_DITTO_BLOCKS))
3757 return (ENOTSUP);
3758 break;
3759
3760 case ZFS_PROP_DEDUP:
3761 if (zfs_earlier_version(dsname, SPA_VERSION_DEDUP))
3762 return (ENOTSUP);
3763 break;
3764
3765 case ZFS_PROP_SHARESMB:
3766 if (zpl_earlier_version(dsname, ZPL_VERSION_FUID))
3767 return (ENOTSUP);
3768 break;
3769
3770 case ZFS_PROP_ACLINHERIT:
3771 if (nvpair_type(pair) == DATA_TYPE_UINT64 &&
3772 nvpair_value_uint64(pair, &intval) == 0) {
3773 if (intval == ZFS_ACL_PASSTHROUGH_X &&
3774 zfs_earlier_version(dsname,
3775 SPA_VERSION_PASSTHROUGH_X))
3776 return (ENOTSUP);
3777 }
3778 break;
3779 }
3780
3781 return (zfs_secpolicy_setprop(dsname, prop, pair, CRED()));
3782 }
3783
3784 /*
3785 * Activates a feature on a pool in response to a property setting. This
3786 * creates a new sync task which modifies the pool to reflect the feature
3787 * as being active.
3788 */
3789 static int
3790 zfs_prop_activate_feature(dsl_pool_t *dp, zfeature_info_t *feature)
3791 {
3792 int err;
3793
3794 /* EBUSY here indicates that the feature is already active */
3795 err = dsl_sync_task_do(dp, zfs_prop_activate_feature_check,
3796 zfs_prop_activate_feature_sync, dp->dp_spa, feature, 2);
3797
3798 if (err != 0 && err != EBUSY)
3799 return (err);
3800 else
3801 return (0);
3802 }
3803
3804 /*
3805 * Checks for a race condition to make sure we don't increment a feature flag
3806 * multiple times.
3807 */
3808 /*ARGSUSED*/
3809 static int
3810 zfs_prop_activate_feature_check(void *arg1, void *arg2, dmu_tx_t *tx)
3811 {
3812 spa_t *spa = arg1;
3813 zfeature_info_t *feature = arg2;
3814
3815 if (!spa_feature_is_active(spa, feature))
3816 return (0);
3817 else
3818 return (EBUSY);
3819 }
3820
3821 /*
3822 * The callback invoked on feature activation in the sync task caused by
3823 * zfs_prop_activate_feature.
3824 */
3825 static void
3826 zfs_prop_activate_feature_sync(void *arg1, void *arg2, dmu_tx_t *tx)
3827 {
3828 spa_t *spa = arg1;
3829 zfeature_info_t *feature = arg2;
3830
3831 spa_feature_incr(spa, feature, tx);
3832 }
3833
3834 /*
3835 * Removes properties from the given props list that fail permission checks
3836 * needed to clear them and to restore them in case of a receive error. For each
3837 * property, make sure we have both set and inherit permissions.
3838 *
3839 * Returns the first error encountered if any permission checks fail. If the
3840 * caller provides a non-NULL errlist, it also gives the complete list of names
3841 * of all the properties that failed a permission check along with the
3842 * corresponding error numbers. The caller is responsible for freeing the
3843 * returned errlist.
3844 *
3845 * If every property checks out successfully, zero is returned and the list
3846 * pointed at by errlist is NULL.
3847 */
3848 static int
3849 zfs_check_clearable(char *dataset, nvlist_t *props, nvlist_t **errlist)
3850 {
3851 zfs_cmd_t *zc;
3852 nvpair_t *pair, *next_pair;
3853 nvlist_t *errors;
3854 int err, rv = 0;
3855
3856 if (props == NULL)
3857 return (0);
3858
3859 VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0);
3860
3861 zc = kmem_alloc(sizeof (zfs_cmd_t), KM_SLEEP);
3862 (void) strcpy(zc->zc_name, dataset);
3863 pair = nvlist_next_nvpair(props, NULL);
3864 while (pair != NULL) {
3865 next_pair = nvlist_next_nvpair(props, pair);
3866
3867 (void) strcpy(zc->zc_value, nvpair_name(pair));
3868 if ((err = zfs_check_settable(dataset, pair, CRED())) != 0 ||
3869 (err = zfs_secpolicy_inherit_prop(zc, NULL, CRED())) != 0) {
3870 VERIFY(nvlist_remove_nvpair(props, pair) == 0);
3871 VERIFY(nvlist_add_int32(errors,
3872 zc->zc_value, err) == 0);
3873 }
3874 pair = next_pair;
3875 }
3876 kmem_free(zc, sizeof (zfs_cmd_t));
3877
3878 if ((pair = nvlist_next_nvpair(errors, NULL)) == NULL) {
3879 nvlist_free(errors);
3880 errors = NULL;
3881 } else {
3882 VERIFY(nvpair_value_int32(pair, &rv) == 0);
3883 }
3884
3885 if (errlist == NULL)
3886 nvlist_free(errors);
3887 else
3888 *errlist = errors;
3889
3890 return (rv);
3891 }
3892
3893 static boolean_t
3894 propval_equals(nvpair_t *p1, nvpair_t *p2)
3895 {
3896 if (nvpair_type(p1) == DATA_TYPE_NVLIST) {
3897 /* dsl_prop_get_all_impl() format */
3898 nvlist_t *attrs;
3899 VERIFY(nvpair_value_nvlist(p1, &attrs) == 0);
3900 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
3901 &p1) == 0);
3902 }
3903
3904 if (nvpair_type(p2) == DATA_TYPE_NVLIST) {
3905 nvlist_t *attrs;
3906 VERIFY(nvpair_value_nvlist(p2, &attrs) == 0);
3907 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
3908 &p2) == 0);
3909 }
3910
3911 if (nvpair_type(p1) != nvpair_type(p2))
3912 return (B_FALSE);
3913
3914 if (nvpair_type(p1) == DATA_TYPE_STRING) {
3915 char *valstr1, *valstr2;
3916
3917 VERIFY(nvpair_value_string(p1, (char **)&valstr1) == 0);
3918 VERIFY(nvpair_value_string(p2, (char **)&valstr2) == 0);
3919 return (strcmp(valstr1, valstr2) == 0);
3920 } else {
3921 uint64_t intval1, intval2;
3922
3923 VERIFY(nvpair_value_uint64(p1, &intval1) == 0);
3924 VERIFY(nvpair_value_uint64(p2, &intval2) == 0);
3925 return (intval1 == intval2);
3926 }
3927 }
3928
3929 /*
3930 * Remove properties from props if they are not going to change (as determined
3931 * by comparison with origprops). Remove them from origprops as well, since we
3932 * do not need to clear or restore properties that won't change.
3933 */
3934 static void
3935 props_reduce(nvlist_t *props, nvlist_t *origprops)
3936 {
3937 nvpair_t *pair, *next_pair;
3938
3939 if (origprops == NULL)
3940 return; /* all props need to be received */
3941
3942 pair = nvlist_next_nvpair(props, NULL);
3943 while (pair != NULL) {
3944 const char *propname = nvpair_name(pair);
3945 nvpair_t *match;
3946
3947 next_pair = nvlist_next_nvpair(props, pair);
3948
3949 if ((nvlist_lookup_nvpair(origprops, propname,
3950 &match) != 0) || !propval_equals(pair, match))
3951 goto next; /* need to set received value */
3952
3953 /* don't clear the existing received value */
3954 (void) nvlist_remove_nvpair(origprops, match);
3955 /* don't bother receiving the property */
3956 (void) nvlist_remove_nvpair(props, pair);
3957 next:
3958 pair = next_pair;
3959 }
3960 }
3961
3962 #ifdef DEBUG
3963 static boolean_t zfs_ioc_recv_inject_err;
3964 #endif
3965
3966 /*
3967 * inputs:
3968 * zc_name name of containing filesystem
3969 * zc_nvlist_src{_size} nvlist of properties to apply
3970 * zc_value name of snapshot to create
3971 * zc_string name of clone origin (if DRR_FLAG_CLONE)
3972 * zc_cookie file descriptor to recv from
3973 * zc_begin_record the BEGIN record of the stream (not byteswapped)
3974 * zc_guid force flag
3975 * zc_cleanup_fd cleanup-on-exit file descriptor
3976 * zc_action_handle handle for this guid/ds mapping (or zero on first call)
3977 *
3978 * outputs:
3979 * zc_cookie number of bytes read
3980 * zc_nvlist_dst{_size} error for each unapplied received property
3981 * zc_obj zprop_errflags_t
3982 * zc_action_handle handle for this guid/ds mapping
3983 */
3984 static int
3985 zfs_ioc_recv(zfs_cmd_t *zc)
3986 {
3987 file_t *fp;
3988 objset_t *os;
3989 dmu_recv_cookie_t drc;
3990 boolean_t force = (boolean_t)zc->zc_guid;
3991 int fd;
3992 int error = 0;
3993 int props_error = 0;
3994 nvlist_t *errors;
3995 offset_t off;
3996 nvlist_t *props = NULL; /* sent properties */
3997 nvlist_t *origprops = NULL; /* existing properties */
3998 objset_t *origin = NULL;
3999 char *tosnap;
4000 char tofs[ZFS_MAXNAMELEN];
4001 boolean_t first_recvd_props = B_FALSE;
4002
4003 if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
4004 strchr(zc->zc_value, '@') == NULL ||
4005 strchr(zc->zc_value, '%'))
4006 return (EINVAL);
4007
4008 (void) strcpy(tofs, zc->zc_value);
4009 tosnap = strchr(tofs, '@');
4010 *tosnap++ = '\0';
4011
4012 if (zc->zc_nvlist_src != NULL &&
4013 (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
4014 zc->zc_iflags, &props)) != 0)
4015 return (error);
4016
4017 fd = zc->zc_cookie;
4018 fp = getf(fd);
4019 if (fp == NULL) {
4020 nvlist_free(props);
4021 return (EBADF);
4022 }
4023
4024 VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0);
4025
4026 if (props && dmu_objset_hold(tofs, FTAG, &os) == 0) {
4027 if ((spa_version(os->os_spa) >= SPA_VERSION_RECVD_PROPS) &&
4028 !dsl_prop_get_hasrecvd(os)) {
4029 first_recvd_props = B_TRUE;
4030 }
4031
4032 /*
4033 * If new received properties are supplied, they are to
4034 * completely replace the existing received properties, so stash
4035 * away the existing ones.
4036 */
4037 if (dsl_prop_get_received(os, &origprops) == 0) {
4038 nvlist_t *errlist = NULL;
4039 /*
4040 * Don't bother writing a property if its value won't
4041 * change (and avoid the unnecessary security checks).
4042 *
4043 * The first receive after SPA_VERSION_RECVD_PROPS is a
4044 * special case where we blow away all local properties
4045 * regardless.
4046 */
4047 if (!first_recvd_props)
4048 props_reduce(props, origprops);
4049 if (zfs_check_clearable(tofs, origprops,
4050 &errlist) != 0)
4051 (void) nvlist_merge(errors, errlist, 0);
4052 nvlist_free(errlist);
4053 }
4054
4055 dmu_objset_rele(os, FTAG);
4056 }
4057
4058 if (zc->zc_string[0]) {
4059 error = dmu_objset_hold(zc->zc_string, FTAG, &origin);
4060 if (error)
4061 goto out;
4062 }
4063
4064 error = dmu_recv_begin(tofs, tosnap, zc->zc_top_ds,
4065 &zc->zc_begin_record, force, origin, &drc);
4066 if (origin)
4067 dmu_objset_rele(origin, FTAG);
4068 if (error)
4069 goto out;
4070
4071 /*
4072 * Set properties before we receive the stream so that they are applied
4073 * to the new data. Note that we must call dmu_recv_stream() if
4074 * dmu_recv_begin() succeeds.
4075 */
4076 if (props) {
4077 if (dmu_objset_from_ds(drc.drc_logical_ds, &os) == 0) {
4078 if (drc.drc_newfs) {
4079 if (spa_version(os->os_spa) >=
4080 SPA_VERSION_RECVD_PROPS)
4081 first_recvd_props = B_TRUE;
4082 } else if (origprops != NULL) {
4083 if (clear_received_props(os, tofs, origprops,
4084 first_recvd_props ? NULL : props) != 0)
4085 zc->zc_obj |= ZPROP_ERR_NOCLEAR;
4086 } else {
4087 zc->zc_obj |= ZPROP_ERR_NOCLEAR;
4088 }
4089 dsl_prop_set_hasrecvd(os);
4090 } else if (!drc.drc_newfs) {
4091 zc->zc_obj |= ZPROP_ERR_NOCLEAR;
4092 }
4093
4094 (void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_RECEIVED,
4095 props, errors);
4096 }
4097
4098 if (zc->zc_nvlist_dst_size != 0 &&
4099 (nvlist_smush(errors, zc->zc_nvlist_dst_size) != 0 ||
4100 put_nvlist(zc, errors) != 0)) {
4101 /*
4102 * Caller made zc->zc_nvlist_dst less than the minimum expected
4103 * size or supplied an invalid address.
4104 */
4105 props_error = EINVAL;
4106 }
4107
4108 off = fp->f_offset;
4109 error = dmu_recv_stream(&drc, fp->f_vnode, &off, zc->zc_cleanup_fd,
4110 &zc->zc_action_handle);
4111
4112 if (error == 0) {
4113 zfsvfs_t *zfsvfs = NULL;
4114
4115 if (getzfsvfs(tofs, &zfsvfs) == 0) {
4116 /* online recv */
4117 int end_err;
4118
4119 error = zfs_suspend_fs(zfsvfs);
4120 /*
4121 * If the suspend fails, then the recv_end will
4122 * likely also fail, and clean up after itself.
4123 */
4124 end_err = dmu_recv_end(&drc);
4125 if (error == 0)
4126 error = zfs_resume_fs(zfsvfs, tofs);
4127 error = error ? error : end_err;
4128 VFS_RELE(zfsvfs->z_vfs);
4129 } else {
4130 error = dmu_recv_end(&drc);
4131 }
4132 }
4133
4134 zc->zc_cookie = off - fp->f_offset;
4135 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
4136 fp->f_offset = off;
4137
4138 #ifdef DEBUG
4139 if (zfs_ioc_recv_inject_err) {
4140 zfs_ioc_recv_inject_err = B_FALSE;
4141 error = 1;
4142 }
4143 #endif
4144 /*
4145 * On error, restore the original props.
4146 */
4147 if (error && props) {
4148 if (dmu_objset_hold(tofs, FTAG, &os) == 0) {
4149 if (clear_received_props(os, tofs, props, NULL) != 0) {
4150 /*
4151 * We failed to clear the received properties.
4152 * Since we may have left a $recvd value on the
4153 * system, we can't clear the $hasrecvd flag.
4154 */
4155 zc->zc_obj |= ZPROP_ERR_NORESTORE;
4156 } else if (first_recvd_props) {
4157 dsl_prop_unset_hasrecvd(os);
4158 }
4159 dmu_objset_rele(os, FTAG);
4160 } else if (!drc.drc_newfs) {
4161 /* We failed to clear the received properties. */
4162 zc->zc_obj |= ZPROP_ERR_NORESTORE;
4163 }
4164
4165 if (origprops == NULL && !drc.drc_newfs) {
4166 /* We failed to stash the original properties. */
4167 zc->zc_obj |= ZPROP_ERR_NORESTORE;
4168 }
4169
4170 /*
4171 * dsl_props_set() will not convert RECEIVED to LOCAL on or
4172 * after SPA_VERSION_RECVD_PROPS, so we need to specify LOCAL
4173 * explictly if we're restoring local properties cleared in the
4174 * first new-style receive.
4175 */
4176 if (origprops != NULL &&
4177 zfs_set_prop_nvlist(tofs, (first_recvd_props ?
4178 ZPROP_SRC_LOCAL : ZPROP_SRC_RECEIVED),
4179 origprops, NULL) != 0) {
4180 /*
4181 * We stashed the original properties but failed to
4182 * restore them.
4183 */
4184 zc->zc_obj |= ZPROP_ERR_NORESTORE;
4185 }
4186 }
4187 out:
4188 nvlist_free(props);
4189 nvlist_free(origprops);
4190 nvlist_free(errors);
4191 releasef(fd);
4192
4193 if (error == 0)
4194 error = props_error;
4195
4196 return (error);
4197 }
4198
4199 /*
4200 * inputs:
4201 * zc_name name of snapshot to send
4202 * zc_cookie file descriptor to send stream to
4203 * zc_obj fromorigin flag (mutually exclusive with zc_fromobj)
4204 * zc_sendobj objsetid of snapshot to send
4205 * zc_fromobj objsetid of incremental fromsnap (may be zero)
4206 * zc_guid if set, estimate size of stream only. zc_cookie is ignored.
4207 * output size in zc_objset_type.
4208 *
4209 * outputs: none
4210 */
4211 static int
4212 zfs_ioc_send(zfs_cmd_t *zc)
4213 {
4214 objset_t *fromsnap = NULL;
4215 objset_t *tosnap;
4216 int error;
4217 offset_t off;
4218 dsl_dataset_t *ds;
4219 dsl_dataset_t *dsfrom = NULL;
4220 spa_t *spa;
4221 dsl_pool_t *dp;
4222 boolean_t estimate = (zc->zc_guid != 0);
4223
4224 error = spa_open(zc->zc_name, &spa, FTAG);
4225 if (error)
4226 return (error);
4227
4228 dp = spa_get_dsl(spa);
4229 rw_enter(&dp->dp_config_rwlock, RW_READER);
4230 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &ds);
4231 rw_exit(&dp->dp_config_rwlock);
4232 spa_close(spa, FTAG);
4233 if (error)
4234 return (error);
4235
4236 error = dmu_objset_from_ds(ds, &tosnap);
4237 if (error) {
4238 dsl_dataset_rele(ds, FTAG);
4239 return (error);
4240 }
4241
4242 if (zc->zc_fromobj != 0) {
4243 rw_enter(&dp->dp_config_rwlock, RW_READER);
4244 error = dsl_dataset_hold_obj(dp, zc->zc_fromobj, FTAG, &dsfrom);
4245 rw_exit(&dp->dp_config_rwlock);
4246 if (error) {
4247 dsl_dataset_rele(ds, FTAG);
4248 return (error);
4249 }
4250 error = dmu_objset_from_ds(dsfrom, &fromsnap);
4251 if (error) {
4252 dsl_dataset_rele(dsfrom, FTAG);
4253 dsl_dataset_rele(ds, FTAG);
4254 return (error);
4255 }
4256 }
4257
4258 if (zc->zc_obj) {
4259 dsl_pool_t *dp = ds->ds_dir->dd_pool;
4260
4261 if (fromsnap != NULL) {
4262 dsl_dataset_rele(dsfrom, FTAG);
4263 dsl_dataset_rele(ds, FTAG);
4264 return (EINVAL);
4265 }
4266
4267 if (dsl_dir_is_clone(ds->ds_dir)) {
4268 rw_enter(&dp->dp_config_rwlock, RW_READER);
4269 error = dsl_dataset_hold_obj(dp,
4270 ds->ds_dir->dd_phys->dd_origin_obj, FTAG, &dsfrom);
4271 rw_exit(&dp->dp_config_rwlock);
4272 if (error) {
4273 dsl_dataset_rele(ds, FTAG);
4274 return (error);
4275 }
4276 error = dmu_objset_from_ds(dsfrom, &fromsnap);
4277 if (error) {
4278 dsl_dataset_rele(dsfrom, FTAG);
4279 dsl_dataset_rele(ds, FTAG);
4280 return (error);
4281 }
4282 }
4283 }
4284
4285 if (estimate) {
4286 error = dmu_send_estimate(tosnap, fromsnap,
4287 &zc->zc_objset_type);
4288 } else {
4289 file_t *fp = getf(zc->zc_cookie);
4290 if (fp == NULL) {
4291 dsl_dataset_rele(ds, FTAG);
4292 if (dsfrom)
4293 dsl_dataset_rele(dsfrom, FTAG);
4294 return (EBADF);
4295 }
4296
4297 off = fp->f_offset;
4298 error = dmu_send(tosnap, fromsnap,
4299 zc->zc_cookie, fp->f_vnode, &off);
4300
4301 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
4302 fp->f_offset = off;
4303 releasef(zc->zc_cookie);
4304 }
4305 if (dsfrom)
4306 dsl_dataset_rele(dsfrom, FTAG);
4307 dsl_dataset_rele(ds, FTAG);
4308 return (error);
4309 }
4310
4311 /*
4312 * inputs:
4313 * zc_name name of snapshot on which to report progress
4314 * zc_cookie file descriptor of send stream
4315 *
4316 * outputs:
4317 * zc_cookie number of bytes written in send stream thus far
4318 */
4319 static int
4320 zfs_ioc_send_progress(zfs_cmd_t *zc)
4321 {
4322 dsl_dataset_t *ds;
4323 dmu_sendarg_t *dsp = NULL;
4324 int error;
4325
4326 if ((error = dsl_dataset_hold(zc->zc_name, FTAG, &ds)) != 0)
4327 return (error);
4328
4329 mutex_enter(&ds->ds_sendstream_lock);
4330
4331 /*
4332 * Iterate over all the send streams currently active on this dataset.
4333 * If there's one which matches the specified file descriptor _and_ the
4334 * stream was started by the current process, return the progress of
4335 * that stream.
4336 */
4337 for (dsp = list_head(&ds->ds_sendstreams); dsp != NULL;
4338 dsp = list_next(&ds->ds_sendstreams, dsp)) {
4339 if (dsp->dsa_outfd == zc->zc_cookie &&
4340 dsp->dsa_proc == curproc)
4341 break;
4342 }
4343
4344 if (dsp != NULL)
4345 zc->zc_cookie = *(dsp->dsa_off);
4346 else
4347 error = ENOENT;
4348
4349 mutex_exit(&ds->ds_sendstream_lock);
4350 dsl_dataset_rele(ds, FTAG);
4351 return (error);
4352 }
4353
4354 static int
4355 zfs_ioc_inject_fault(zfs_cmd_t *zc)
4356 {
4357 int id, error;
4358
4359 error = zio_inject_fault(zc->zc_name, (int)zc->zc_guid, &id,
4360 &zc->zc_inject_record);
4361
4362 if (error == 0)
4363 zc->zc_guid = (uint64_t)id;
4364
4365 return (error);
4366 }
4367
4368 static int
4369 zfs_ioc_clear_fault(zfs_cmd_t *zc)
4370 {
4371 return (zio_clear_fault((int)zc->zc_guid));
4372 }
4373
4374 static int
4375 zfs_ioc_inject_list_next(zfs_cmd_t *zc)
4376 {
4377 int id = (int)zc->zc_guid;
4378 int error;
4379
4380 error = zio_inject_list_next(&id, zc->zc_name, sizeof (zc->zc_name),
4381 &zc->zc_inject_record);
4382
4383 zc->zc_guid = id;
4384
4385 return (error);
4386 }
4387
4388 static int
4389 zfs_ioc_error_log(zfs_cmd_t *zc)
4390 {
4391 spa_t *spa;
4392 int error;
4393 size_t count = (size_t)zc->zc_nvlist_dst_size;
4394
4395 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
4396 return (error);
4397
4398 error = spa_get_errlog(spa, (void *)(uintptr_t)zc->zc_nvlist_dst,
4399 &count);
4400 if (error == 0)
4401 zc->zc_nvlist_dst_size = count;
4402 else
4403 zc->zc_nvlist_dst_size = spa_get_errlog_size(spa);
4404
4405 spa_close(spa, FTAG);
4406
4407 return (error);
4408 }
4409
4410 static int
4411 zfs_ioc_clear(zfs_cmd_t *zc)
4412 {
4413 spa_t *spa;
4414 vdev_t *vd;
4415 int error;
4416
4417 /*
4418 * On zpool clear we also fix up missing slogs
4419 */
4420 mutex_enter(&spa_namespace_lock);
4421 spa = spa_lookup(zc->zc_name);
4422 if (spa == NULL) {
4423 mutex_exit(&spa_namespace_lock);
4424 return (EIO);
4425 }
4426 if (spa_get_log_state(spa) == SPA_LOG_MISSING) {
4427 /* we need to let spa_open/spa_load clear the chains */
4428 spa_set_log_state(spa, SPA_LOG_CLEAR);
4429 }
4430 spa->spa_last_open_failed = 0;
4431 mutex_exit(&spa_namespace_lock);
4432
4433 if (zc->zc_cookie & ZPOOL_NO_REWIND) {
4434 error = spa_open(zc->zc_name, &spa, FTAG);
4435 } else {
4436 nvlist_t *policy;
4437 nvlist_t *config = NULL;
4438
4439 if (zc->zc_nvlist_src == NULL)
4440 return (EINVAL);
4441
4442 if ((error = get_nvlist(zc->zc_nvlist_src,
4443 zc->zc_nvlist_src_size, zc->zc_iflags, &policy)) == 0) {
4444 error = spa_open_rewind(zc->zc_name, &spa, FTAG,
4445 policy, &config);
4446 if (config != NULL) {
4447 int err;
4448
4449 if ((err = put_nvlist(zc, config)) != 0)
4450 error = err;
4451 nvlist_free(config);
4452 }
4453 nvlist_free(policy);
4454 }
4455 }
4456
4457 if (error)
4458 return (error);
4459
4460 spa_vdev_state_enter(spa, SCL_NONE);
4461
4462 if (zc->zc_guid == 0) {
4463 vd = NULL;
4464 } else {
4465 vd = spa_lookup_by_guid(spa, zc->zc_guid, B_TRUE);
4466 if (vd == NULL) {
4467 (void) spa_vdev_state_exit(spa, NULL, ENODEV);
4468 spa_close(spa, FTAG);
4469 return (ENODEV);
4470 }
4471 }
4472
4473 vdev_clear(spa, vd);
4474
4475 (void) spa_vdev_state_exit(spa, NULL, 0);
4476
4477 /*
4478 * Resume any suspended I/Os.
4479 */
4480 if (zio_resume(spa) != 0)
4481 error = EIO;
4482
4483 spa_close(spa, FTAG);
4484
4485 return (error);
4486 }
4487
4488 static int
4489 zfs_ioc_pool_reopen(zfs_cmd_t *zc)
4490 {
4491 spa_t *spa;
4492 int error;
4493
4494 error = spa_open(zc->zc_name, &spa, FTAG);
4495 if (error)
4496 return (error);
4497
4498 spa_vdev_state_enter(spa, SCL_NONE);
4499
4500 /*
4501 * If a resilver is already in progress then set the
4502 * spa_scrub_reopen flag to B_TRUE so that we don't restart
4503 * the scan as a side effect of the reopen. Otherwise, let
4504 * vdev_open() decided if a resilver is required.
4505 */
4506 spa->spa_scrub_reopen = dsl_scan_resilvering(spa->spa_dsl_pool);
4507 vdev_reopen(spa->spa_root_vdev);
4508 spa->spa_scrub_reopen = B_FALSE;
4509
4510 (void) spa_vdev_state_exit(spa, NULL, 0);
4511 spa_close(spa, FTAG);
4512 return (0);
4513 }
4514 /*
4515 * inputs:
4516 * zc_name name of filesystem
4517 * zc_value name of origin snapshot
4518 *
4519 * outputs:
4520 * zc_string name of conflicting snapshot, if there is one
4521 */
4522 static int
4523 zfs_ioc_promote(zfs_cmd_t *zc)
4524 {
4525 char *cp;
4526
4527 /*
4528 * We don't need to unmount *all* the origin fs's snapshots, but
4529 * it's easier.
4530 */
4531 cp = strchr(zc->zc_value, '@');
4532 if (cp)
4533 *cp = '\0';
4534 (void) dmu_objset_find(zc->zc_value,
4535 zfs_unmount_snap, NULL, DS_FIND_SNAPSHOTS);
4536 return (dsl_dataset_promote(zc->zc_name, zc->zc_string));
4537 }
4538
4539 /*
4540 * Retrieve a single {user|group}{used|quota}@... property.
4541 *
4542 * inputs:
4543 * zc_name name of filesystem
4544 * zc_objset_type zfs_userquota_prop_t
4545 * zc_value domain name (eg. "S-1-234-567-89")
4546 * zc_guid RID/UID/GID
4547 *
4548 * outputs:
4549 * zc_cookie property value
4550 */
4551 static int
4552 zfs_ioc_userspace_one(zfs_cmd_t *zc)
4553 {
4554 zfsvfs_t *zfsvfs;
4555 int error;
4556
4557 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
4558 return (EINVAL);
4559
4560 error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE);
4561 if (error)
4562 return (error);
4563
4564 error = zfs_userspace_one(zfsvfs,
4565 zc->zc_objset_type, zc->zc_value, zc->zc_guid, &zc->zc_cookie);
4566 zfsvfs_rele(zfsvfs, FTAG);
4567
4568 return (error);
4569 }
4570
4571 /*
4572 * inputs:
4573 * zc_name name of filesystem
4574 * zc_cookie zap cursor
4575 * zc_objset_type zfs_userquota_prop_t
4576 * zc_nvlist_dst[_size] buffer to fill (not really an nvlist)
4577 *
4578 * outputs:
4579 * zc_nvlist_dst[_size] data buffer (array of zfs_useracct_t)
4580 * zc_cookie zap cursor
4581 */
4582 static int
4583 zfs_ioc_userspace_many(zfs_cmd_t *zc)
4584 {
4585 zfsvfs_t *zfsvfs;
4586 int bufsize = zc->zc_nvlist_dst_size;
4587
4588 if (bufsize <= 0)
4589 return (ENOMEM);
4590
4591 int error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE);
4592 if (error)
4593 return (error);
4594
4595 void *buf = kmem_alloc(bufsize, KM_SLEEP);
4596
4597 error = zfs_userspace_many(zfsvfs, zc->zc_objset_type, &zc->zc_cookie,
4598 buf, &zc->zc_nvlist_dst_size);
4599
4600 if (error == 0) {
4601 error = xcopyout(buf,
4602 (void *)(uintptr_t)zc->zc_nvlist_dst,
4603 zc->zc_nvlist_dst_size);
4604 }
4605 kmem_free(buf, bufsize);
4606 zfsvfs_rele(zfsvfs, FTAG);
4607
4608 return (error);
4609 }
4610
4611 /*
4612 * inputs:
4613 * zc_name name of filesystem
4614 *
4615 * outputs:
4616 * none
4617 */
4618 static int
4619 zfs_ioc_userspace_upgrade(zfs_cmd_t *zc)
4620 {
4621 objset_t *os;
4622 int error = 0;
4623 zfsvfs_t *zfsvfs;
4624
4625 if (getzfsvfs(zc->zc_name, &zfsvfs) == 0) {
4626 if (!dmu_objset_userused_enabled(zfsvfs->z_os)) {
4627 /*
4628 * If userused is not enabled, it may be because the
4629 * objset needs to be closed & reopened (to grow the
4630 * objset_phys_t). Suspend/resume the fs will do that.
4631 */
4632 error = zfs_suspend_fs(zfsvfs);
4633 if (error == 0)
4634 error = zfs_resume_fs(zfsvfs, zc->zc_name);
4635 }
4636 if (error == 0)
4637 error = dmu_objset_userspace_upgrade(zfsvfs->z_os);
4638 VFS_RELE(zfsvfs->z_vfs);
4639 } else {
4640 /* XXX kind of reading contents without owning */
4641 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
4642 if (error)
4643 return (error);
4644
4645 error = dmu_objset_userspace_upgrade(os);
4646 dmu_objset_rele(os, FTAG);
4647 }
4648
4649 return (error);
4650 }
4651
4652 /*
4653 * We don't want to have a hard dependency
4654 * against some special symbols in sharefs
4655 * nfs, and smbsrv. Determine them if needed when
4656 * the first file system is shared.
4657 * Neither sharefs, nfs or smbsrv are unloadable modules.
4658 */
4659 int (*znfsexport_fs)(void *arg);
4660 int (*zshare_fs)(enum sharefs_sys_op, share_t *, uint32_t);
4661 int (*zsmbexport_fs)(void *arg, boolean_t add_share);
4662
4663 int zfs_nfsshare_inited;
4664 int zfs_smbshare_inited;
4665
4666 ddi_modhandle_t nfs_mod;
4667 ddi_modhandle_t sharefs_mod;
4668 ddi_modhandle_t smbsrv_mod;
4669 kmutex_t zfs_share_lock;
4670
4671 static int
4672 zfs_init_sharefs()
4673 {
4674 int error;
4675
4676 ASSERT(MUTEX_HELD(&zfs_share_lock));
4677 /* Both NFS and SMB shares also require sharetab support. */
4678 if (sharefs_mod == NULL && ((sharefs_mod =
4679 ddi_modopen("fs/sharefs",
4680 KRTLD_MODE_FIRST, &error)) == NULL)) {
4681 return (ENOSYS);
4682 }
4683 if (zshare_fs == NULL && ((zshare_fs =
4684 (int (*)(enum sharefs_sys_op, share_t *, uint32_t))
4685 ddi_modsym(sharefs_mod, "sharefs_impl", &error)) == NULL)) {
4686 return (ENOSYS);
4687 }
4688 return (0);
4689 }
4690
4691 static int
4692 zfs_ioc_share(zfs_cmd_t *zc)
4693 {
4694 int error;
4695 int opcode;
4696
4697 switch (zc->zc_share.z_sharetype) {
4698 case ZFS_SHARE_NFS:
4699 case ZFS_UNSHARE_NFS:
4700 if (zfs_nfsshare_inited == 0) {
4701 mutex_enter(&zfs_share_lock);
4702 if (nfs_mod == NULL && ((nfs_mod = ddi_modopen("fs/nfs",
4703 KRTLD_MODE_FIRST, &error)) == NULL)) {
4704 mutex_exit(&zfs_share_lock);
4705 return (ENOSYS);
4706 }
4707 if (znfsexport_fs == NULL &&
4708 ((znfsexport_fs = (int (*)(void *))
4709 ddi_modsym(nfs_mod,
4710 "nfs_export", &error)) == NULL)) {
4711 mutex_exit(&zfs_share_lock);
4712 return (ENOSYS);
4713 }
4714 error = zfs_init_sharefs();
4715 if (error) {
4716 mutex_exit(&zfs_share_lock);
4717 return (ENOSYS);
4718 }
4719 zfs_nfsshare_inited = 1;
4720 mutex_exit(&zfs_share_lock);
4721 }
4722 break;
4723 case ZFS_SHARE_SMB:
4724 case ZFS_UNSHARE_SMB:
4725 if (zfs_smbshare_inited == 0) {
4726 mutex_enter(&zfs_share_lock);
4727 if (smbsrv_mod == NULL && ((smbsrv_mod =
4728 ddi_modopen("drv/smbsrv",
4729 KRTLD_MODE_FIRST, &error)) == NULL)) {
4730 mutex_exit(&zfs_share_lock);
4731 return (ENOSYS);
4732 }
4733 if (zsmbexport_fs == NULL && ((zsmbexport_fs =
4734 (int (*)(void *, boolean_t))ddi_modsym(smbsrv_mod,
4735 "smb_server_share", &error)) == NULL)) {
4736 mutex_exit(&zfs_share_lock);
4737 return (ENOSYS);
4738 }
4739 error = zfs_init_sharefs();
4740 if (error) {
4741 mutex_exit(&zfs_share_lock);
4742 return (ENOSYS);
4743 }
4744 zfs_smbshare_inited = 1;
4745 mutex_exit(&zfs_share_lock);
4746 }
4747 break;
4748 default:
4749 return (EINVAL);
4750 }
4751
4752 switch (zc->zc_share.z_sharetype) {
4753 case ZFS_SHARE_NFS:
4754 case ZFS_UNSHARE_NFS:
4755 if (error =
4756 znfsexport_fs((void *)
4757 (uintptr_t)zc->zc_share.z_exportdata))
4758 return (error);
4759 break;
4760 case ZFS_SHARE_SMB:
4761 case ZFS_UNSHARE_SMB:
4762 if (error = zsmbexport_fs((void *)
4763 (uintptr_t)zc->zc_share.z_exportdata,
4764 zc->zc_share.z_sharetype == ZFS_SHARE_SMB ?
4765 B_TRUE: B_FALSE)) {
4766 return (error);
4767 }
4768 break;
4769 }
4770
4771 opcode = (zc->zc_share.z_sharetype == ZFS_SHARE_NFS ||
4772 zc->zc_share.z_sharetype == ZFS_SHARE_SMB) ?
4773 SHAREFS_ADD : SHAREFS_REMOVE;
4774
4775 /*
4776 * Add or remove share from sharetab
4777 */
4778 error = zshare_fs(opcode,
4779 (void *)(uintptr_t)zc->zc_share.z_sharedata,
4780 zc->zc_share.z_sharemax);
4781
4782 return (error);
4783
4784 }
4785
4786 ace_t full_access[] = {
4787 {(uid_t)-1, ACE_ALL_PERMS, ACE_EVERYONE, 0}
4788 };
4789
4790 /*
4791 * inputs:
4792 * zc_name name of containing filesystem
4793 * zc_obj object # beyond which we want next in-use object #
4794 *
4795 * outputs:
4796 * zc_obj next in-use object #
4797 */
4798 static int
4799 zfs_ioc_next_obj(zfs_cmd_t *zc)
4800 {
4801 objset_t *os = NULL;
4802 int error;
4803
4804 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
4805 if (error)
4806 return (error);
4807
4808 error = dmu_object_next(os, &zc->zc_obj, B_FALSE,
4809 os->os_dsl_dataset->ds_phys->ds_prev_snap_txg);
4810
4811 dmu_objset_rele(os, FTAG);
4812 return (error);
4813 }
4814
4815 /*
4816 * inputs:
4817 * zc_name name of filesystem
4818 * zc_value prefix name for snapshot
4819 * zc_cleanup_fd cleanup-on-exit file descriptor for calling process
4820 *
4821 * outputs:
4822 * zc_value short name of new snapshot
4823 */
4824 static int
4825 zfs_ioc_tmp_snapshot(zfs_cmd_t *zc)
4826 {
4827 char *snap_name;
4828 int error;
4829
4830 snap_name = kmem_asprintf("%s@%s-%016llx", zc->zc_name, zc->zc_value,
4831 (u_longlong_t)ddi_get_lbolt64());
4832
4833 if (strlen(snap_name) >= MAXPATHLEN) {
4834 strfree(snap_name);
4835 return (E2BIG);
4836 }
4837
4838 error = dmu_objset_snapshot_tmp(snap_name, "%temp", zc->zc_cleanup_fd);
4839 if (error != 0) {
4840 strfree(snap_name);
4841 return (error);
4842 }
4843
4844 (void) strcpy(zc->zc_value, strchr(snap_name, '@') + 1);
4845 strfree(snap_name);
4846 return (0);
4847 }
4848
4849 /*
4850 * inputs:
4851 * zc_name name of "to" snapshot
4852 * zc_value name of "from" snapshot
4853 * zc_cookie file descriptor to write diff data on
4854 *
4855 * outputs:
4856 * dmu_diff_record_t's to the file descriptor
4857 */
4858 static int
4859 zfs_ioc_diff(zfs_cmd_t *zc)
4860 {
4861 objset_t *fromsnap;
4862 objset_t *tosnap;
4863 file_t *fp;
4864 offset_t off;
4865 int error;
4866
4867 error = dmu_objset_hold(zc->zc_name, FTAG, &tosnap);
4868 if (error)
4869 return (error);
4870
4871 error = dmu_objset_hold(zc->zc_value, FTAG, &fromsnap);
4872 if (error) {
4873 dmu_objset_rele(tosnap, FTAG);
4874 return (error);
4875 }
4876
4877 fp = getf(zc->zc_cookie);
4878 if (fp == NULL) {
4879 dmu_objset_rele(fromsnap, FTAG);
4880 dmu_objset_rele(tosnap, FTAG);
4881 return (EBADF);
4882 }
4883
4884 off = fp->f_offset;
4885
4886 error = dmu_diff(tosnap, fromsnap, fp->f_vnode, &off);
4887
4888 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
4889 fp->f_offset = off;
4890 releasef(zc->zc_cookie);
4891
4892 dmu_objset_rele(fromsnap, FTAG);
4893 dmu_objset_rele(tosnap, FTAG);
4894 return (error);
4895 }
4896
4897 /*
4898 * Remove all ACL files in shares dir
4899 */
4900 static int
4901 zfs_smb_acl_purge(znode_t *dzp)
4902 {
4903 zap_cursor_t zc;
4904 zap_attribute_t zap;
4905 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
4906 int error;
4907
4908 for (zap_cursor_init(&zc, zfsvfs->z_os, dzp->z_id);
4909 (error = zap_cursor_retrieve(&zc, &zap)) == 0;
4910 zap_cursor_advance(&zc)) {
4911 if ((error = VOP_REMOVE(ZTOV(dzp), zap.za_name, kcred,
4912 NULL, 0)) != 0)
4913 break;
4914 }
4915 zap_cursor_fini(&zc);
4916 return (error);
4917 }
4918
4919 static int
4920 zfs_ioc_smb_acl(zfs_cmd_t *zc)
4921 {
4922 vnode_t *vp;
4923 znode_t *dzp;
4924 vnode_t *resourcevp = NULL;
4925 znode_t *sharedir;
4926 zfsvfs_t *zfsvfs;
4927 nvlist_t *nvlist;
4928 char *src, *target;
4929 vattr_t vattr;
4930 vsecattr_t vsec;
4931 int error = 0;
4932
4933 if ((error = lookupname(zc->zc_value, UIO_SYSSPACE,
4934 NO_FOLLOW, NULL, &vp)) != 0)
4935 return (error);
4936
4937 /* Now make sure mntpnt and dataset are ZFS */
4938
4939 if (vp->v_vfsp->vfs_fstype != zfsfstype ||
4940 (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource),
4941 zc->zc_name) != 0)) {
4942 VN_RELE(vp);
4943 return (EINVAL);
4944 }
4945
4946 dzp = VTOZ(vp);
4947 zfsvfs = dzp->z_zfsvfs;
4948 ZFS_ENTER(zfsvfs);
4949
4950 /*
4951 * Create share dir if its missing.
4952 */
4953 mutex_enter(&zfsvfs->z_lock);
4954 if (zfsvfs->z_shares_dir == 0) {
4955 dmu_tx_t *tx;
4956
4957 tx = dmu_tx_create(zfsvfs->z_os);
4958 dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, TRUE,
4959 ZFS_SHARES_DIR);
4960 dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL);
4961 error = dmu_tx_assign(tx, TXG_WAIT);
4962 if (error) {
4963 dmu_tx_abort(tx);
4964 } else {
4965 error = zfs_create_share_dir(zfsvfs, tx);
4966 dmu_tx_commit(tx);
4967 }
4968 if (error) {
4969 mutex_exit(&zfsvfs->z_lock);
4970 VN_RELE(vp);
4971 ZFS_EXIT(zfsvfs);
4972 return (error);
4973 }
4974 }
4975 mutex_exit(&zfsvfs->z_lock);
4976
4977 ASSERT(zfsvfs->z_shares_dir);
4978 if ((error = zfs_zget(zfsvfs, zfsvfs->z_shares_dir, &sharedir)) != 0) {
4979 VN_RELE(vp);
4980 ZFS_EXIT(zfsvfs);
4981 return (error);
4982 }
4983
4984 switch (zc->zc_cookie) {
4985 case ZFS_SMB_ACL_ADD:
4986 vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE;
4987 vattr.va_type = VREG;
4988 vattr.va_mode = S_IFREG|0777;
4989 vattr.va_uid = 0;
4990 vattr.va_gid = 0;
4991
4992 vsec.vsa_mask = VSA_ACE;
4993 vsec.vsa_aclentp = &full_access;
4994 vsec.vsa_aclentsz = sizeof (full_access);
4995 vsec.vsa_aclcnt = 1;
4996
4997 error = VOP_CREATE(ZTOV(sharedir), zc->zc_string,
4998 &vattr, EXCL, 0, &resourcevp, kcred, 0, NULL, &vsec);
4999 if (resourcevp)
5000 VN_RELE(resourcevp);
5001 break;
5002
5003 case ZFS_SMB_ACL_REMOVE:
5004 error = VOP_REMOVE(ZTOV(sharedir), zc->zc_string, kcred,
5005 NULL, 0);
5006 break;
5007
5008 case ZFS_SMB_ACL_RENAME:
5009 if ((error = get_nvlist(zc->zc_nvlist_src,
5010 zc->zc_nvlist_src_size, zc->zc_iflags, &nvlist)) != 0) {
5011 VN_RELE(vp);
5012 ZFS_EXIT(zfsvfs);
5013 return (error);
5014 }
5015 if (nvlist_lookup_string(nvlist, ZFS_SMB_ACL_SRC, &src) ||
5016 nvlist_lookup_string(nvlist, ZFS_SMB_ACL_TARGET,
5017 &target)) {
5018 VN_RELE(vp);
5019 VN_RELE(ZTOV(sharedir));
5020 ZFS_EXIT(zfsvfs);
5021 nvlist_free(nvlist);
5022 return (error);
5023 }
5024 error = VOP_RENAME(ZTOV(sharedir), src, ZTOV(sharedir), target,
5025 kcred, NULL, 0);
5026 nvlist_free(nvlist);
5027 break;
5028
5029 case ZFS_SMB_ACL_PURGE:
5030 error = zfs_smb_acl_purge(sharedir);
5031 break;
5032
5033 default:
5034 error = EINVAL;
5035 break;
5036 }
5037
5038 VN_RELE(vp);
5039 VN_RELE(ZTOV(sharedir));
5040
5041 ZFS_EXIT(zfsvfs);
5042
5043 return (error);
5044 }
5045
5046 /*
5047 * inputs:
5048 * zc_name name of filesystem
5049 * zc_value short name of snap
5050 * zc_string user-supplied tag for this hold
5051 * zc_cookie recursive flag
5052 * zc_temphold set if hold is temporary
5053 * zc_cleanup_fd cleanup-on-exit file descriptor for calling process
5054 * zc_sendobj if non-zero, the objid for zc_name@zc_value
5055 * zc_createtxg if zc_sendobj is non-zero, snap must have zc_createtxg
5056 *
5057 * outputs: none
5058 */
5059 static int
5060 zfs_ioc_hold(zfs_cmd_t *zc)
5061 {
5062 boolean_t recursive = zc->zc_cookie;
5063 spa_t *spa;
5064 dsl_pool_t *dp;
5065 dsl_dataset_t *ds;
5066 int error;
5067 minor_t minor = 0;
5068
5069 if (snapshot_namecheck(zc->zc_value, NULL, NULL) != 0)
5070 return (EINVAL);
5071
5072 if (zc->zc_sendobj == 0) {
5073 return (dsl_dataset_user_hold(zc->zc_name, zc->zc_value,
5074 zc->zc_string, recursive, zc->zc_temphold,
5075 zc->zc_cleanup_fd));
5076 }
5077
5078 if (recursive)
5079 return (EINVAL);
5080
5081 error = spa_open(zc->zc_name, &spa, FTAG);
5082 if (error)
5083 return (error);
5084
5085 dp = spa_get_dsl(spa);
5086 rw_enter(&dp->dp_config_rwlock, RW_READER);
5087 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &ds);
5088 rw_exit(&dp->dp_config_rwlock);
5089 spa_close(spa, FTAG);
5090 if (error)
5091 return (error);
5092
5093 /*
5094 * Until we have a hold on this snapshot, it's possible that
5095 * zc_sendobj could've been destroyed and reused as part
5096 * of a later txg. Make sure we're looking at the right object.
5097 */
5098 if (zc->zc_createtxg != ds->ds_phys->ds_creation_txg) {
5099 dsl_dataset_rele(ds, FTAG);
5100 return (ENOENT);
5101 }
5102
5103 if (zc->zc_cleanup_fd != -1 && zc->zc_temphold) {
5104 error = zfs_onexit_fd_hold(zc->zc_cleanup_fd, &minor);
5105 if (error) {
5106 dsl_dataset_rele(ds, FTAG);
5107 return (error);
5108 }
5109 }
5110
5111 error = dsl_dataset_user_hold_for_send(ds, zc->zc_string,
5112 zc->zc_temphold);
5113 if (minor != 0) {
5114 if (error == 0) {
5115 dsl_register_onexit_hold_cleanup(ds, zc->zc_string,
5116 minor);
5117 }
5118 zfs_onexit_fd_rele(zc->zc_cleanup_fd);
5119 }
5120 dsl_dataset_rele(ds, FTAG);
5121
5122 return (error);
5123 }
5124
5125 /*
5126 * inputs:
5127 * zc_name name of dataset from which we're releasing a user hold
5128 * zc_value short name of snap
5129 * zc_string user-supplied tag for this hold
5130 * zc_cookie recursive flag
5131 *
5132 * outputs: none
5133 */
5134 static int
5135 zfs_ioc_release(zfs_cmd_t *zc)
5136 {
5137 boolean_t recursive = zc->zc_cookie;
5138
5139 if (snapshot_namecheck(zc->zc_value, NULL, NULL) != 0)
5140 return (EINVAL);
5141
5142 return (dsl_dataset_user_release(zc->zc_name, zc->zc_value,
5143 zc->zc_string, recursive));
5144 }
5145
5146 /*
5147 * inputs:
5148 * zc_name name of filesystem
5149 *
5150 * outputs:
5151 * zc_nvlist_src{_size} nvlist of snapshot holds
5152 */
5153 static int
5154 zfs_ioc_get_holds(zfs_cmd_t *zc)
5155 {
5156 nvlist_t *nvp;
5157 int error;
5158
5159 if ((error = dsl_dataset_get_holds(zc->zc_name, &nvp)) == 0) {
5160 error = put_nvlist(zc, nvp);
5161 nvlist_free(nvp);
5162 }
5163
5164 return (error);
5165 }
5166
5167 /*
5168 * inputs:
5169 * zc_name name of new filesystem or snapshot
5170 * zc_value full name of old snapshot
5171 *
5172 * outputs:
5173 * zc_cookie space in bytes
5174 * zc_objset_type compressed space in bytes
5175 * zc_perm_action uncompressed space in bytes
5176 */
5177 static int
5178 zfs_ioc_space_written(zfs_cmd_t *zc)
5179 {
5180 int error;
5181 dsl_dataset_t *new, *old;
5182
5183 error = dsl_dataset_hold(zc->zc_name, FTAG, &new);
5184 if (error != 0)
5185 return (error);
5186 error = dsl_dataset_hold(zc->zc_value, FTAG, &old);
5187 if (error != 0) {
5188 dsl_dataset_rele(new, FTAG);
5189 return (error);
5190 }
5191
5192 error = dsl_dataset_space_written(old, new, &zc->zc_cookie,
5193 &zc->zc_objset_type, &zc->zc_perm_action);
5194 dsl_dataset_rele(old, FTAG);
5195 dsl_dataset_rele(new, FTAG);
5196 return (error);
5197 }
5198 /*
5199 * innvl: {
5200 * "firstsnap" -> snapshot name
5201 * }
5202 *
5203 * outnvl: {
5204 * "used" -> space in bytes
5205 * "compressed" -> compressed space in bytes
5206 * "uncompressed" -> uncompressed space in bytes
5207 * }
5208 */
5209 static int
5210 zfs_ioc_space_snaps(const char *lastsnap, nvlist_t *innvl, nvlist_t *outnvl)
5211 {
5212 int error;
5213 dsl_dataset_t *new, *old;
5214 char *firstsnap;
5215 uint64_t used, comp, uncomp;
5216
5217 if (nvlist_lookup_string(innvl, "firstsnap", &firstsnap) != 0)
5218 return (EINVAL);
5219
5220 error = dsl_dataset_hold(lastsnap, FTAG, &new);
5221 if (error != 0)
5222 return (error);
5223 error = dsl_dataset_hold(firstsnap, FTAG, &old);
5224 if (error != 0) {
5225 dsl_dataset_rele(new, FTAG);
5226 return (error);
5227 }
5228
5229 error = dsl_dataset_space_wouldfree(old, new, &used, &comp, &uncomp);
5230 dsl_dataset_rele(old, FTAG);
5231 dsl_dataset_rele(new, FTAG);
5232 fnvlist_add_uint64(outnvl, "used", used);
5233 fnvlist_add_uint64(outnvl, "compressed", comp);
5234 fnvlist_add_uint64(outnvl, "uncompressed", uncomp);
5235 return (error);
5236 }
5237
5238 /*
5239 * innvl: {
5240 * "fd" -> file descriptor to write stream to (int32)
5241 * (optional) "fromsnap" -> full snap name to send an incremental from
5242 * }
5243 *
5244 * outnvl is unused
5245 */
5246 /* ARGSUSED */
5247 static int
5248 zfs_ioc_send_new(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl)
5249 {
5250 objset_t *fromsnap = NULL;
5251 objset_t *tosnap;
5252 int error;
5253 offset_t off;
5254 char *fromname;
5255 int fd;
5256
5257 error = nvlist_lookup_int32(innvl, "fd", &fd);
5258 if (error != 0)
5259 return (EINVAL);
5260
5261 error = dmu_objset_hold(snapname, FTAG, &tosnap);
5262 if (error)
5263 return (error);
5264
5265 error = nvlist_lookup_string(innvl, "fromsnap", &fromname);
5266 if (error == 0) {
5267 error = dmu_objset_hold(fromname, FTAG, &fromsnap);
5268 if (error) {
5269 dmu_objset_rele(tosnap, FTAG);
5270 return (error);
5271 }
5272 }
5273
5274 file_t *fp = getf(fd);
5275 if (fp == NULL) {
5276 dmu_objset_rele(tosnap, FTAG);
5277 if (fromsnap != NULL)
5278 dmu_objset_rele(fromsnap, FTAG);
5279 return (EBADF);
5280 }
5281
5282 off = fp->f_offset;
5283 error = dmu_send(tosnap, fromsnap, fd, fp->f_vnode, &off);
5284
5285 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
5286 fp->f_offset = off;
5287 releasef(fd);
5288 if (fromsnap != NULL)
5289 dmu_objset_rele(fromsnap, FTAG);
5290 dmu_objset_rele(tosnap, FTAG);
5291 return (error);
5292 }
5293
5294 /*
5295 * Determine approximately how large a zfs send stream will be -- the number
5296 * of bytes that will be written to the fd supplied to zfs_ioc_send_new().
5297 *
5298 * innvl: {
5299 * (optional) "fromsnap" -> full snap name to send an incremental from
5300 * }
5301 *
5302 * outnvl: {
5303 * "space" -> bytes of space (uint64)
5304 * }
5305 */
5306 static int
5307 zfs_ioc_send_space(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl)
5308 {
5309 objset_t *fromsnap = NULL;
5310 objset_t *tosnap;
5311 int error;
5312 char *fromname;
5313 uint64_t space;
5314
5315 error = dmu_objset_hold(snapname, FTAG, &tosnap);
5316 if (error)
5317 return (error);
5318
5319 error = nvlist_lookup_string(innvl, "fromsnap", &fromname);
5320 if (error == 0) {
5321 error = dmu_objset_hold(fromname, FTAG, &fromsnap);
5322 if (error) {
5323 dmu_objset_rele(tosnap, FTAG);
5324 return (error);
5325 }
5326 }
5327
5328 error = dmu_send_estimate(tosnap, fromsnap, &space);
5329 fnvlist_add_uint64(outnvl, "space", space);
5330
5331 if (fromsnap != NULL)
5332 dmu_objset_rele(fromsnap, FTAG);
5333 dmu_objset_rele(tosnap, FTAG);
5334 return (error);
5335 }
5336
5337
5338 static zfs_ioc_vec_t zfs_ioc_vec[ZFS_IOC_LAST - ZFS_IOC_FIRST];
5339
5340 static void
5341 zfs_ioctl_register_legacy(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5342 zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck,
5343 boolean_t log_history, zfs_ioc_poolcheck_t pool_check)
5344 {
5345 zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST];
5346
5347 ASSERT3U(ioc, >=, ZFS_IOC_FIRST);
5348 ASSERT3U(ioc, <, ZFS_IOC_LAST);
5349 ASSERT3P(vec->zvec_legacy_func, ==, NULL);
5350 ASSERT3P(vec->zvec_func, ==, NULL);
5351
5352 vec->zvec_legacy_func = func;
5353 vec->zvec_secpolicy = secpolicy;
5354 vec->zvec_namecheck = namecheck;
5355 vec->zvec_allow_log = log_history;
5356 vec->zvec_pool_check = pool_check;
5357 }
5358
5359 /*
5360 * See the block comment at the beginning of this file for details on
5361 * each argument to this function.
5362 */
5363 static void
5364 zfs_ioctl_register(const char *name, zfs_ioc_t ioc, zfs_ioc_func_t *func,
5365 zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck,
5366 zfs_ioc_poolcheck_t pool_check, boolean_t smush_outnvlist,
5367 boolean_t allow_log)
5368 {
5369 zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST];
5370
5371 ASSERT3U(ioc, >=, ZFS_IOC_FIRST);
5372 ASSERT3U(ioc, <, ZFS_IOC_LAST);
5373 ASSERT3P(vec->zvec_legacy_func, ==, NULL);
5374 ASSERT3P(vec->zvec_func, ==, NULL);
5375
5376 /* if we are logging, the name must be valid */
5377 ASSERT(!allow_log || namecheck != NO_NAME);
5378
5379 vec->zvec_name = name;
5380 vec->zvec_func = func;
5381 vec->zvec_secpolicy = secpolicy;
5382 vec->zvec_namecheck = namecheck;
5383 vec->zvec_pool_check = pool_check;
5384 vec->zvec_smush_outnvlist = smush_outnvlist;
5385 vec->zvec_allow_log = allow_log;
5386 }
5387
5388 static void
5389 zfs_ioctl_register_pool(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5390 zfs_secpolicy_func_t *secpolicy, boolean_t log_history,
5391 zfs_ioc_poolcheck_t pool_check)
5392 {
5393 zfs_ioctl_register_legacy(ioc, func, secpolicy,
5394 POOL_NAME, log_history, pool_check);
5395 }
5396
5397 static void
5398 zfs_ioctl_register_dataset_nolog(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5399 zfs_secpolicy_func_t *secpolicy, zfs_ioc_poolcheck_t pool_check)
5400 {
5401 zfs_ioctl_register_legacy(ioc, func, secpolicy,
5402 DATASET_NAME, B_FALSE, pool_check);
5403 }
5404
5405 static void
5406 zfs_ioctl_register_pool_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func)
5407 {
5408 zfs_ioctl_register_legacy(ioc, func, zfs_secpolicy_config,
5409 POOL_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5410 }
5411
5412 static void
5413 zfs_ioctl_register_pool_meta(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5414 zfs_secpolicy_func_t *secpolicy)
5415 {
5416 zfs_ioctl_register_legacy(ioc, func, secpolicy,
5417 NO_NAME, B_FALSE, POOL_CHECK_NONE);
5418 }
5419
5420 static void
5421 zfs_ioctl_register_dataset_read_secpolicy(zfs_ioc_t ioc,
5422 zfs_ioc_legacy_func_t *func, zfs_secpolicy_func_t *secpolicy)
5423 {
5424 zfs_ioctl_register_legacy(ioc, func, secpolicy,
5425 DATASET_NAME, B_FALSE, POOL_CHECK_SUSPENDED);
5426 }
5427
5428 static void
5429 zfs_ioctl_register_dataset_read(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func)
5430 {
5431 zfs_ioctl_register_dataset_read_secpolicy(ioc, func,
5432 zfs_secpolicy_read);
5433 }
5434
5435 static void
5436 zfs_ioctl_register_dataset_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5437 zfs_secpolicy_func_t *secpolicy)
5438 {
5439 zfs_ioctl_register_legacy(ioc, func, secpolicy,
5440 DATASET_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5441 }
5442
5443 static void
5444 zfs_ioctl_init(void)
5445 {
5446 zfs_ioctl_register("snapshot", ZFS_IOC_SNAPSHOT,
5447 zfs_ioc_snapshot, zfs_secpolicy_snapshot, POOL_NAME,
5448 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5449
5450 zfs_ioctl_register("log_history", ZFS_IOC_LOG_HISTORY,
5451 zfs_ioc_log_history, zfs_secpolicy_log_history, NO_NAME,
5452 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_FALSE);
5453
5454 zfs_ioctl_register("space_snaps", ZFS_IOC_SPACE_SNAPS,
5455 zfs_ioc_space_snaps, zfs_secpolicy_read, DATASET_NAME,
5456 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5457
5458 zfs_ioctl_register("send", ZFS_IOC_SEND_NEW,
5459 zfs_ioc_send_new, zfs_secpolicy_send_new, DATASET_NAME,
5460 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5461
5462 zfs_ioctl_register("send_space", ZFS_IOC_SEND_SPACE,
5463 zfs_ioc_send_space, zfs_secpolicy_read, DATASET_NAME,
5464 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5465
5466 zfs_ioctl_register("create", ZFS_IOC_CREATE,
5467 zfs_ioc_create, zfs_secpolicy_create_clone, DATASET_NAME,
5468 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5469
5470 zfs_ioctl_register("clone", ZFS_IOC_CLONE,
5471 zfs_ioc_clone, zfs_secpolicy_create_clone, DATASET_NAME,
5472 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5473
5474 zfs_ioctl_register("destroy_snaps", ZFS_IOC_DESTROY_SNAPS,
5475 zfs_ioc_destroy_snaps, zfs_secpolicy_destroy_snaps, POOL_NAME,
5476 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5477
5478 /* IOCTLS that use the legacy function signature */
5479
5480 zfs_ioctl_register_legacy(ZFS_IOC_POOL_FREEZE, zfs_ioc_pool_freeze,
5481 zfs_secpolicy_config, NO_NAME, B_FALSE, POOL_CHECK_READONLY);
5482
5483 zfs_ioctl_register_pool(ZFS_IOC_POOL_CREATE, zfs_ioc_pool_create,
5484 zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE);
5485 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SCAN,
5486 zfs_ioc_pool_scan);
5487 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_UPGRADE,
5488 zfs_ioc_pool_upgrade);
5489 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ADD,
5490 zfs_ioc_vdev_add);
5491 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_REMOVE,
5492 zfs_ioc_vdev_remove);
5493 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SET_STATE,
5494 zfs_ioc_vdev_set_state);
5495 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ATTACH,
5496 zfs_ioc_vdev_attach);
5497 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_DETACH,
5498 zfs_ioc_vdev_detach);
5499 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETPATH,
5500 zfs_ioc_vdev_setpath);
5501 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETFRU,
5502 zfs_ioc_vdev_setfru);
5503 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SET_PROPS,
5504 zfs_ioc_pool_set_props);
5505 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SPLIT,
5506 zfs_ioc_vdev_split);
5507 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_REGUID,
5508 zfs_ioc_pool_reguid);
5509
5510 zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_CONFIGS,
5511 zfs_ioc_pool_configs, zfs_secpolicy_none);
5512 zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_TRYIMPORT,
5513 zfs_ioc_pool_tryimport, zfs_secpolicy_config);
5514 zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_FAULT,
5515 zfs_ioc_inject_fault, zfs_secpolicy_inject);
5516 zfs_ioctl_register_pool_meta(ZFS_IOC_CLEAR_FAULT,
5517 zfs_ioc_clear_fault, zfs_secpolicy_inject);
5518 zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_LIST_NEXT,
5519 zfs_ioc_inject_list_next, zfs_secpolicy_inject);
5520
5521 /*
5522 * pool destroy, and export don't log the history as part of
5523 * zfsdev_ioctl, but rather zfs_ioc_pool_export
5524 * does the logging of those commands.
5525 */
5526 zfs_ioctl_register_pool(ZFS_IOC_POOL_DESTROY, zfs_ioc_pool_destroy,
5527 zfs_secpolicy_config, B_FALSE, POOL_CHECK_NONE);
5528 zfs_ioctl_register_pool(ZFS_IOC_POOL_EXPORT, zfs_ioc_pool_export,
5529 zfs_secpolicy_config, B_FALSE, POOL_CHECK_NONE);
5530
5531 zfs_ioctl_register_pool(ZFS_IOC_POOL_STATS, zfs_ioc_pool_stats,
5532 zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE);
5533 zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_PROPS, zfs_ioc_pool_get_props,
5534 zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE);
5535
5536 zfs_ioctl_register_pool(ZFS_IOC_ERROR_LOG, zfs_ioc_error_log,
5537 zfs_secpolicy_inject, B_FALSE, POOL_CHECK_SUSPENDED);
5538 zfs_ioctl_register_pool(ZFS_IOC_DSOBJ_TO_DSNAME,
5539 zfs_ioc_dsobj_to_dsname,
5540 zfs_secpolicy_diff, B_FALSE, POOL_CHECK_SUSPENDED);
5541 zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_HISTORY,
5542 zfs_ioc_pool_get_history,
5543 zfs_secpolicy_config, B_FALSE, POOL_CHECK_SUSPENDED);
5544
5545 zfs_ioctl_register_pool(ZFS_IOC_POOL_IMPORT, zfs_ioc_pool_import,
5546 zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE);
5547
5548 zfs_ioctl_register_pool(ZFS_IOC_CLEAR, zfs_ioc_clear,
5549 zfs_secpolicy_config, B_TRUE, POOL_CHECK_SUSPENDED);
5550 zfs_ioctl_register_pool(ZFS_IOC_POOL_REOPEN, zfs_ioc_pool_reopen,
5551 zfs_secpolicy_config, B_TRUE, POOL_CHECK_SUSPENDED);
5552
5553 zfs_ioctl_register_dataset_read(ZFS_IOC_SPACE_WRITTEN,
5554 zfs_ioc_space_written);
5555 zfs_ioctl_register_dataset_read(ZFS_IOC_GET_HOLDS,
5556 zfs_ioc_get_holds);
5557 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_RECVD_PROPS,
5558 zfs_ioc_objset_recvd_props);
5559 zfs_ioctl_register_dataset_read(ZFS_IOC_NEXT_OBJ,
5560 zfs_ioc_next_obj);
5561 zfs_ioctl_register_dataset_read(ZFS_IOC_GET_FSACL,
5562 zfs_ioc_get_fsacl);
5563 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_STATS,
5564 zfs_ioc_objset_stats);
5565 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_ZPLPROPS,
5566 zfs_ioc_objset_zplprops);
5567 zfs_ioctl_register_dataset_read(ZFS_IOC_DATASET_LIST_NEXT,
5568 zfs_ioc_dataset_list_next);
5569 zfs_ioctl_register_dataset_read(ZFS_IOC_SNAPSHOT_LIST_NEXT,
5570 zfs_ioc_snapshot_list_next);
5571 zfs_ioctl_register_dataset_read(ZFS_IOC_SEND_PROGRESS,
5572 zfs_ioc_send_progress);
5573
5574 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_DIFF,
5575 zfs_ioc_diff, zfs_secpolicy_diff);
5576 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_STATS,
5577 zfs_ioc_obj_to_stats, zfs_secpolicy_diff);
5578 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_PATH,
5579 zfs_ioc_obj_to_path, zfs_secpolicy_diff);
5580 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_ONE,
5581 zfs_ioc_userspace_one, zfs_secpolicy_userspace_one);
5582 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_MANY,
5583 zfs_ioc_userspace_many, zfs_secpolicy_userspace_many);
5584 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_SEND,
5585 zfs_ioc_send, zfs_secpolicy_send);
5586
5587 zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_PROP, zfs_ioc_set_prop,
5588 zfs_secpolicy_none);
5589 zfs_ioctl_register_dataset_modify(ZFS_IOC_DESTROY, zfs_ioc_destroy,
5590 zfs_secpolicy_destroy);
5591 zfs_ioctl_register_dataset_modify(ZFS_IOC_ROLLBACK, zfs_ioc_rollback,
5592 zfs_secpolicy_rollback);
5593 zfs_ioctl_register_dataset_modify(ZFS_IOC_RENAME, zfs_ioc_rename,
5594 zfs_secpolicy_rename);
5595 zfs_ioctl_register_dataset_modify(ZFS_IOC_RECV, zfs_ioc_recv,
5596 zfs_secpolicy_recv);
5597 zfs_ioctl_register_dataset_modify(ZFS_IOC_PROMOTE, zfs_ioc_promote,
5598 zfs_secpolicy_promote);
5599 zfs_ioctl_register_dataset_modify(ZFS_IOC_HOLD, zfs_ioc_hold,
5600 zfs_secpolicy_hold);
5601 zfs_ioctl_register_dataset_modify(ZFS_IOC_RELEASE, zfs_ioc_release,
5602 zfs_secpolicy_release);
5603 zfs_ioctl_register_dataset_modify(ZFS_IOC_INHERIT_PROP,
5604 zfs_ioc_inherit_prop, zfs_secpolicy_inherit_prop);
5605 zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_FSACL, zfs_ioc_set_fsacl,
5606 zfs_secpolicy_set_fsacl);
5607
5608 zfs_ioctl_register_dataset_nolog(ZFS_IOC_SHARE, zfs_ioc_share,
5609 zfs_secpolicy_share, POOL_CHECK_NONE);
5610 zfs_ioctl_register_dataset_nolog(ZFS_IOC_SMB_ACL, zfs_ioc_smb_acl,
5611 zfs_secpolicy_smb_acl, POOL_CHECK_NONE);
5612 zfs_ioctl_register_dataset_nolog(ZFS_IOC_USERSPACE_UPGRADE,
5613 zfs_ioc_userspace_upgrade, zfs_secpolicy_userspace_upgrade,
5614 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5615 zfs_ioctl_register_dataset_nolog(ZFS_IOC_TMP_SNAPSHOT,
5616 zfs_ioc_tmp_snapshot, zfs_secpolicy_tmp_snapshot,
5617 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5618 }
5619
5620 int
5621 pool_status_check(const char *name, zfs_ioc_namecheck_t type,
5622 zfs_ioc_poolcheck_t check)
5623 {
5624 spa_t *spa;
5625 int error;
5626
5627 ASSERT(type == POOL_NAME || type == DATASET_NAME);
5628
5629 if (check & POOL_CHECK_NONE)
5630 return (0);
5631
5632 error = spa_open(name, &spa, FTAG);
5633 if (error == 0) {
5634 if ((check & POOL_CHECK_SUSPENDED) && spa_suspended(spa))
5635 error = EAGAIN;
5636 else if ((check & POOL_CHECK_READONLY) && !spa_writeable(spa))
5637 error = EROFS;
5638 spa_close(spa, FTAG);
5639 }
5640 return (error);
5641 }
5642
5643 /*
5644 * Find a free minor number.
5645 */
5646 minor_t
5647 zfsdev_minor_alloc(void)
5648 {
5649 static minor_t last_minor;
5650 minor_t m;
5651
5652 ASSERT(MUTEX_HELD(&zfsdev_state_lock));
5653
5654 for (m = last_minor + 1; m != last_minor; m++) {
5655 if (m > ZFSDEV_MAX_MINOR)
5656 m = 1;
5657 if (ddi_get_soft_state(zfsdev_state, m) == NULL) {
5658 last_minor = m;
5659 return (m);
5660 }
5661 }
5662
5663 return (0);
5664 }
5665
5666 static int
5667 zfs_ctldev_init(dev_t *devp)
5668 {
5669 minor_t minor;
5670 zfs_soft_state_t *zs;
5671
5672 ASSERT(MUTEX_HELD(&zfsdev_state_lock));
5673 ASSERT(getminor(*devp) == 0);
5674
5675 minor = zfsdev_minor_alloc();
5676 if (minor == 0)
5677 return (ENXIO);
5678
5679 if (ddi_soft_state_zalloc(zfsdev_state, minor) != DDI_SUCCESS)
5680 return (EAGAIN);
5681
5682 *devp = makedevice(getemajor(*devp), minor);
5683
5684 zs = ddi_get_soft_state(zfsdev_state, minor);
5685 zs->zss_type = ZSST_CTLDEV;
5686 zfs_onexit_init((zfs_onexit_t **)&zs->zss_data);
5687
5688 return (0);
5689 }
5690
5691 static void
5692 zfs_ctldev_destroy(zfs_onexit_t *zo, minor_t minor)
5693 {
5694 ASSERT(MUTEX_HELD(&zfsdev_state_lock));
5695
5696 zfs_onexit_destroy(zo);
5697 ddi_soft_state_free(zfsdev_state, minor);
5698 }
5699
5700 void *
5701 zfsdev_get_soft_state(minor_t minor, enum zfs_soft_state_type which)
5702 {
5703 zfs_soft_state_t *zp;
5704
5705 zp = ddi_get_soft_state(zfsdev_state, minor);
5706 if (zp == NULL || zp->zss_type != which)
5707 return (NULL);
5708
5709 return (zp->zss_data);
5710 }
5711
5712 static int
5713 zfsdev_open(dev_t *devp, int flag, int otyp, cred_t *cr)
5714 {
5715 int error = 0;
5716
5717 if (getminor(*devp) != 0)
5718 return (zvol_open(devp, flag, otyp, cr));
5719
5720 /* This is the control device. Allocate a new minor if requested. */
5721 if (flag & FEXCL) {
5722 mutex_enter(&zfsdev_state_lock);
5723 error = zfs_ctldev_init(devp);
5724 mutex_exit(&zfsdev_state_lock);
5725 }
5726
5727 return (error);
5728 }
5729
5730 static int
5731 zfsdev_close(dev_t dev, int flag, int otyp, cred_t *cr)
5732 {
5733 zfs_onexit_t *zo;
5734 minor_t minor = getminor(dev);
5735
5736 if (minor == 0)
5737 return (0);
5738
5739 mutex_enter(&zfsdev_state_lock);
5740 zo = zfsdev_get_soft_state(minor, ZSST_CTLDEV);
5741 if (zo == NULL) {
5742 mutex_exit(&zfsdev_state_lock);
5743 return (zvol_close(dev, flag, otyp, cr));
5744 }
5745 zfs_ctldev_destroy(zo, minor);
5746 mutex_exit(&zfsdev_state_lock);
5747
5748 return (0);
5749 }
5750
5751 static int
5752 zfsdev_ioctl(dev_t dev, int cmd, intptr_t arg, int flag, cred_t *cr, int *rvalp)
5753 {
5754 zfs_cmd_t *zc;
5755 uint_t vecnum;
5756 int error, rc, len;
5757 minor_t minor = getminor(dev);
5758 const zfs_ioc_vec_t *vec;
5759 char *saved_poolname = NULL;
5760 nvlist_t *innvl = NULL;
5761
5762 if (minor != 0 &&
5763 zfsdev_get_soft_state(minor, ZSST_CTLDEV) == NULL)
5764 return (zvol_ioctl(dev, cmd, arg, flag, cr, rvalp));
5765
5766 vecnum = cmd - ZFS_IOC_FIRST;
5767 ASSERT3U(getmajor(dev), ==, ddi_driver_major(zfs_dip));
5768
5769 if (vecnum >= sizeof (zfs_ioc_vec) / sizeof (zfs_ioc_vec[0]))
5770 return (EINVAL);
5771 vec = &zfs_ioc_vec[vecnum];
5772
5773 zc = kmem_zalloc(sizeof (zfs_cmd_t), KM_SLEEP);
5774
5775 error = ddi_copyin((void *)arg, zc, sizeof (zfs_cmd_t), flag);
5776 if (error != 0) {
5777 error = EFAULT;
5778 goto out;
5779 }
5780
5781 zc->zc_iflags = flag & FKIOCTL;
5782 if (zc->zc_nvlist_src_size != 0) {
5783 error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
5784 zc->zc_iflags, &innvl);
5785 if (error != 0)
5786 goto out;
5787 }
5788
5789 /*
5790 * Ensure that all pool/dataset names are valid before we pass down to
5791 * the lower layers.
5792 */
5793 zc->zc_name[sizeof (zc->zc_name) - 1] = '\0';
5794 switch (vec->zvec_namecheck) {
5795 case POOL_NAME:
5796 if (pool_namecheck(zc->zc_name, NULL, NULL) != 0)
5797 error = EINVAL;
5798 else
5799 error = pool_status_check(zc->zc_name,
5800 vec->zvec_namecheck, vec->zvec_pool_check);
5801 break;
5802
5803 case DATASET_NAME:
5804 if (dataset_namecheck(zc->zc_name, NULL, NULL) != 0)
5805 error = EINVAL;
5806 else
5807 error = pool_status_check(zc->zc_name,
5808 vec->zvec_namecheck, vec->zvec_pool_check);
5809 break;
5810
5811 case NO_NAME:
5812 break;
5813 }
5814
5815
5816 if (error == 0 && !(flag & FKIOCTL))
5817 error = vec->zvec_secpolicy(zc, innvl, cr);
5818
5819 if (error != 0)
5820 goto out;
5821
5822 /* legacy ioctls can modify zc_name */
5823 len = strcspn(zc->zc_name, "/@") + 1;
5824 saved_poolname = kmem_alloc(len, KM_SLEEP);
5825 (void) strlcpy(saved_poolname, zc->zc_name, len);
5826
5827 if (vec->zvec_func != NULL) {
5828 nvlist_t *outnvl;
5829 int puterror = 0;
5830 spa_t *spa;
5831 nvlist_t *lognv = NULL;
5832
5833 ASSERT(vec->zvec_legacy_func == NULL);
5834
5835 /*
5836 * Add the innvl to the lognv before calling the func,
5837 * in case the func changes the innvl.
5838 */
5839 if (vec->zvec_allow_log) {
5840 lognv = fnvlist_alloc();
5841 fnvlist_add_string(lognv, ZPOOL_HIST_IOCTL,
5842 vec->zvec_name);
5843 if (!nvlist_empty(innvl)) {
5844 fnvlist_add_nvlist(lognv, ZPOOL_HIST_INPUT_NVL,
5845 innvl);
5846 }
5847 }
5848
5849 outnvl = fnvlist_alloc();
5850 error = vec->zvec_func(zc->zc_name, innvl, outnvl);
5851
5852 if (error == 0 && vec->zvec_allow_log &&
5853 spa_open(zc->zc_name, &spa, FTAG) == 0) {
5854 if (!nvlist_empty(outnvl)) {
5855 fnvlist_add_nvlist(lognv, ZPOOL_HIST_OUTPUT_NVL,
5856 outnvl);
5857 }
5858 (void) spa_history_log_nvl(spa, lognv);
5859 spa_close(spa, FTAG);
5860 }
5861 fnvlist_free(lognv);
5862
5863 if (!nvlist_empty(outnvl) || zc->zc_nvlist_dst_size != 0) {
5864 int smusherror = 0;
5865 if (vec->zvec_smush_outnvlist) {
5866 smusherror = nvlist_smush(outnvl,
5867 zc->zc_nvlist_dst_size);
5868 }
5869 if (smusherror == 0)
5870 puterror = put_nvlist(zc, outnvl);
5871 }
5872
5873 if (puterror != 0)
5874 error = puterror;
5875
5876 nvlist_free(outnvl);
5877 } else {
5878 error = vec->zvec_legacy_func(zc);
5879 }
5880
5881 out:
5882 nvlist_free(innvl);
5883 rc = ddi_copyout(zc, (void *)arg, sizeof (zfs_cmd_t), flag);
5884 if (error == 0 && rc != 0)
5885 error = EFAULT;
5886 if (error == 0 && vec->zvec_allow_log) {
5887 char *s = tsd_get(zfs_allow_log_key);
5888 if (s != NULL)
5889 strfree(s);
5890 (void) tsd_set(zfs_allow_log_key, saved_poolname);
5891 } else {
5892 if (saved_poolname != NULL)
5893 strfree(saved_poolname);
5894 }
5895
5896 kmem_free(zc, sizeof (zfs_cmd_t));
5897 return (error);
5898 }
5899
5900 static int
5901 zfs_attach(dev_info_t *dip, ddi_attach_cmd_t cmd)
5902 {
5903 if (cmd != DDI_ATTACH)
5904 return (DDI_FAILURE);
5905
5906 if (ddi_create_minor_node(dip, "zfs", S_IFCHR, 0,
5907 DDI_PSEUDO, 0) == DDI_FAILURE)
5908 return (DDI_FAILURE);
5909
5910 zfs_dip = dip;
5911
5912 ddi_report_dev(dip);
5913
5914 return (DDI_SUCCESS);
5915 }
5916
5917 static int
5918 zfs_detach(dev_info_t *dip, ddi_detach_cmd_t cmd)
5919 {
5920 if (spa_busy() || zfs_busy() || zvol_busy())
5921 return (DDI_FAILURE);
5922
5923 if (cmd != DDI_DETACH)
5924 return (DDI_FAILURE);
5925
5926 zfs_dip = NULL;
5927
5928 ddi_prop_remove_all(dip);
5929 ddi_remove_minor_node(dip, NULL);
5930
5931 return (DDI_SUCCESS);
5932 }
5933
5934 /*ARGSUSED*/
5935 static int
5936 zfs_info(dev_info_t *dip, ddi_info_cmd_t infocmd, void *arg, void **result)
5937 {
5938 switch (infocmd) {
5939 case DDI_INFO_DEVT2DEVINFO:
5940 *result = zfs_dip;
5941 return (DDI_SUCCESS);
5942
5943 case DDI_INFO_DEVT2INSTANCE:
5944 *result = (void *)0;
5945 return (DDI_SUCCESS);
5946 }
5947
5948 return (DDI_FAILURE);
5949 }
5950
5951 /*
5952 * OK, so this is a little weird.
5953 *
5954 * /dev/zfs is the control node, i.e. minor 0.
5955 * /dev/zvol/[r]dsk/pool/dataset are the zvols, minor > 0.
5956 *
5957 * /dev/zfs has basically nothing to do except serve up ioctls,
5958 * so most of the standard driver entry points are in zvol.c.
5959 */
5960 static struct cb_ops zfs_cb_ops = {
5961 zfsdev_open, /* open */
5962 zfsdev_close, /* close */
5963 zvol_strategy, /* strategy */
5964 nodev, /* print */
5965 zvol_dump, /* dump */
5966 zvol_read, /* read */
5967 zvol_write, /* write */
5968 zfsdev_ioctl, /* ioctl */
5969 nodev, /* devmap */
5970 nodev, /* mmap */
5971 nodev, /* segmap */
5972 nochpoll, /* poll */
5973 ddi_prop_op, /* prop_op */
5974 NULL, /* streamtab */
5975 D_NEW | D_MP | D_64BIT, /* Driver compatibility flag */
5976 CB_REV, /* version */
5977 nodev, /* async read */
5978 nodev, /* async write */
5979 };
5980
5981 static struct dev_ops zfs_dev_ops = {
5982 DEVO_REV, /* version */
5983 0, /* refcnt */
5984 zfs_info, /* info */
5985 nulldev, /* identify */
5986 nulldev, /* probe */
5987 zfs_attach, /* attach */
5988 zfs_detach, /* detach */
5989 nodev, /* reset */
5990 &zfs_cb_ops, /* driver operations */
5991 NULL, /* no bus operations */
5992 NULL, /* power */
5993 ddi_quiesce_not_needed, /* quiesce */
5994 };
5995
5996 static struct modldrv zfs_modldrv = {
5997 &mod_driverops,
5998 "ZFS storage pool",
5999 &zfs_dev_ops
6000 };
6001
6002 static struct modlinkage modlinkage = {
6003 MODREV_1,
6004 (void *)&zfs_modlfs,
6005 (void *)&zfs_modldrv,
6006 NULL
6007 };
6008
6009 static void
6010 zfs_allow_log_destroy(void *arg)
6011 {
6012 char *poolname = arg;
6013 strfree(poolname);
6014 }
6015
6016 int
6017 _init(void)
6018 {
6019 int error;
6020
6021 spa_init(FREAD | FWRITE);
6022 zfs_init();
6023 zvol_init();
6024 zfs_ioctl_init();
6025
6026 if ((error = mod_install(&modlinkage)) != 0) {
6027 zvol_fini();
6028 zfs_fini();
6029 spa_fini();
6030 return (error);
6031 }
6032
6033 tsd_create(&zfs_fsyncer_key, NULL);
6034 tsd_create(&rrw_tsd_key, rrw_tsd_destroy);
6035 tsd_create(&zfs_allow_log_key, zfs_allow_log_destroy);
6036
6037 error = ldi_ident_from_mod(&modlinkage, &zfs_li);
6038 ASSERT(error == 0);
6039 mutex_init(&zfs_share_lock, NULL, MUTEX_DEFAULT, NULL);
6040
6041 return (0);
6042 }
6043
6044 int
6045 _fini(void)
6046 {
6047 int error;
6048
6049 if (spa_busy() || zfs_busy() || zvol_busy() || zio_injection_enabled)
6050 return (EBUSY);
6051
6052 if ((error = mod_remove(&modlinkage)) != 0)
6053 return (error);
6054
6055 zvol_fini();
6056 zfs_fini();
6057 spa_fini();
6058 if (zfs_nfsshare_inited)
6059 (void) ddi_modclose(nfs_mod);
6060 if (zfs_smbshare_inited)
6061 (void) ddi_modclose(smbsrv_mod);
6062 if (zfs_nfsshare_inited || zfs_smbshare_inited)
6063 (void) ddi_modclose(sharefs_mod);
6064
6065 tsd_destroy(&zfs_fsyncer_key);
6066 ldi_ident_release(zfs_li);
6067 zfs_li = NULL;
6068 mutex_destroy(&zfs_share_lock);
6069
6070 return (error);
6071 }
6072
6073 int
6074 _info(struct modinfo *modinfop)
6075 {
6076 return (mod_info(&modlinkage, modinfop));
6077 }