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 2015 Nexenta Systems, Inc. All rights reserved.
26 * Copyright (c) 2014, Joyent, Inc. All rights reserved.
27 * Copyright (c) 2011, 2015 by Delphix. All rights reserved.
28 * Copyright (c) 2013 by Saso Kiselkov. All rights reserved.
29 * Copyright (c) 2013 Steven Hartland. All rights reserved.
30 */
31
32 /*
33 * ZFS ioctls.
34 *
35 * This file handles the ioctls to /dev/zfs, used for configuring ZFS storage
36 * pools and filesystems, e.g. with /sbin/zfs and /sbin/zpool.
37 *
38 * There are two ways that we handle ioctls: the legacy way where almost
39 * all of the logic is in the ioctl callback, and the new way where most
40 * of the marshalling is handled in the common entry point, zfsdev_ioctl().
41 *
42 * Non-legacy ioctls should be registered by calling
43 * zfs_ioctl_register() from zfs_ioctl_init(). The ioctl is invoked
44 * from userland by lzc_ioctl().
45 *
46 * The registration arguments are as follows:
47 *
48 * const char *name
49 * The name of the ioctl. This is used for history logging. If the
50 * ioctl returns successfully (the callback returns 0), and allow_log
51 * is true, then a history log entry will be recorded with the input &
52 * output nvlists. The log entry can be printed with "zpool history -i".
53 *
54 * zfs_ioc_t ioc
55 * The ioctl request number, which userland will pass to ioctl(2).
56 * The ioctl numbers can change from release to release, because
57 * the caller (libzfs) must be matched to the kernel.
58 *
59 * zfs_secpolicy_func_t *secpolicy
60 * This function will be called before the zfs_ioc_func_t, to
61 * determine if this operation is permitted. It should return EPERM
62 * on failure, and 0 on success. Checks include determining if the
63 * dataset is visible in this zone, and if the user has either all
64 * zfs privileges in the zone (SYS_MOUNT), or has been granted permission
65 * to do this operation on this dataset with "zfs allow".
66 *
67 * zfs_ioc_namecheck_t namecheck
68 * This specifies what to expect in the zfs_cmd_t:zc_name -- a pool
69 * name, a dataset name, or nothing. If the name is not well-formed,
70 * the ioctl will fail and the callback will not be called.
71 * Therefore, the callback can assume that the name is well-formed
72 * (e.g. is null-terminated, doesn't have more than one '@' character,
73 * doesn't have invalid characters).
74 *
75 * zfs_ioc_poolcheck_t pool_check
76 * This specifies requirements on the pool state. If the pool does
77 * not meet them (is suspended or is readonly), the ioctl will fail
78 * and the callback will not be called. If any checks are specified
79 * (i.e. it is not POOL_CHECK_NONE), namecheck must not be NO_NAME.
80 * Multiple checks can be or-ed together (e.g. POOL_CHECK_SUSPENDED |
81 * POOL_CHECK_READONLY).
82 *
83 * boolean_t smush_outnvlist
84 * If smush_outnvlist is true, then the output is presumed to be a
85 * list of errors, and it will be "smushed" down to fit into the
86 * caller's buffer, by removing some entries and replacing them with a
87 * single "N_MORE_ERRORS" entry indicating how many were removed. See
88 * nvlist_smush() for details. If smush_outnvlist is false, and the
89 * outnvlist does not fit into the userland-provided buffer, then the
90 * ioctl will fail with ENOMEM.
91 *
92 * zfs_ioc_func_t *func
93 * The callback function that will perform the operation.
94 *
95 * The callback should return 0 on success, or an error number on
96 * failure. If the function fails, the userland ioctl will return -1,
97 * and errno will be set to the callback's return value. The callback
98 * will be called with the following arguments:
99 *
100 * const char *name
101 * The name of the pool or dataset to operate on, from
102 * zfs_cmd_t:zc_name. The 'namecheck' argument specifies the
103 * expected type (pool, dataset, or none).
104 *
105 * nvlist_t *innvl
106 * The input nvlist, deserialized from zfs_cmd_t:zc_nvlist_src. Or
107 * NULL if no input nvlist was provided. Changes to this nvlist are
108 * ignored. If the input nvlist could not be deserialized, the
109 * ioctl will fail and the callback will not be called.
110 *
111 * nvlist_t *outnvl
112 * The output nvlist, initially empty. The callback can fill it in,
113 * and it will be returned to userland by serializing it into
114 * zfs_cmd_t:zc_nvlist_dst. If it is non-empty, and serialization
115 * fails (e.g. because the caller didn't supply a large enough
116 * buffer), then the overall ioctl will fail. See the
117 * 'smush_nvlist' argument above for additional behaviors.
118 *
119 * There are two typical uses of the output nvlist:
120 * - To return state, e.g. property values. In this case,
121 * smush_outnvlist should be false. If the buffer was not large
122 * enough, the caller will reallocate a larger buffer and try
123 * the ioctl again.
124 *
125 * - To return multiple errors from an ioctl which makes on-disk
126 * changes. In this case, smush_outnvlist should be true.
127 * Ioctls which make on-disk modifications should generally not
128 * use the outnvl if they succeed, because the caller can not
129 * distinguish between the operation failing, and
130 * deserialization failing.
131 */
132
133 #include <sys/types.h>
134 #include <sys/param.h>
135 #include <sys/errno.h>
136 #include <sys/uio.h>
137 #include <sys/buf.h>
138 #include <sys/modctl.h>
139 #include <sys/open.h>
140 #include <sys/file.h>
141 #include <sys/kmem.h>
142 #include <sys/conf.h>
143 #include <sys/cmn_err.h>
144 #include <sys/stat.h>
145 #include <sys/zfs_ioctl.h>
146 #include <sys/zfs_vfsops.h>
147 #include <sys/zfs_znode.h>
148 #include <sys/zap.h>
149 #include <sys/spa.h>
150 #include <sys/spa_impl.h>
151 #include <sys/vdev.h>
152 #include <sys/priv_impl.h>
153 #include <sys/dmu.h>
154 #include <sys/dsl_dir.h>
155 #include <sys/dsl_dataset.h>
156 #include <sys/dsl_prop.h>
157 #include <sys/dsl_deleg.h>
158 #include <sys/dmu_objset.h>
159 #include <sys/dmu_impl.h>
160 #include <sys/dmu_tx.h>
161 #include <sys/ddi.h>
162 #include <sys/sunddi.h>
163 #include <sys/sunldi.h>
164 #include <sys/policy.h>
165 #include <sys/zone.h>
166 #include <sys/nvpair.h>
167 #include <sys/pathname.h>
168 #include <sys/mount.h>
169 #include <sys/sdt.h>
170 #include <sys/fs/zfs.h>
171 #include <sys/zfs_ctldir.h>
172 #include <sys/zfs_dir.h>
173 #include <sys/zfs_onexit.h>
174 #include <sys/zvol.h>
175 #include <sys/dsl_scan.h>
176 #include <sharefs/share.h>
177 #include <sys/dmu_objset.h>
178 #include <sys/dmu_send.h>
179 #include <sys/dsl_destroy.h>
180 #include <sys/dsl_bookmark.h>
181 #include <sys/dsl_userhold.h>
182 #include <sys/zfeature.h>
183 #include <sys/zio_checksum.h>
184
185 #include "zfs_namecheck.h"
186 #include "zfs_prop.h"
187 #include "zfs_deleg.h"
188 #include "zfs_comutil.h"
189
190 extern struct modlfs zfs_modlfs;
191
192 extern void zfs_init(void);
193 extern void zfs_fini(void);
194
195 ldi_ident_t zfs_li = NULL;
196 dev_info_t *zfs_dip;
197
198 uint_t zfs_fsyncer_key;
199 extern uint_t rrw_tsd_key;
200 static uint_t zfs_allow_log_key;
201
202 typedef int zfs_ioc_legacy_func_t(zfs_cmd_t *);
203 typedef int zfs_ioc_func_t(const char *, nvlist_t *, nvlist_t *);
204 typedef int zfs_secpolicy_func_t(zfs_cmd_t *, nvlist_t *, cred_t *);
205
206 typedef enum {
207 NO_NAME,
208 POOL_NAME,
209 DATASET_NAME
210 } zfs_ioc_namecheck_t;
211
212 typedef enum {
213 POOL_CHECK_NONE = 1 << 0,
214 POOL_CHECK_SUSPENDED = 1 << 1,
215 POOL_CHECK_READONLY = 1 << 2,
216 } zfs_ioc_poolcheck_t;
217
218 typedef struct zfs_ioc_vec {
219 zfs_ioc_legacy_func_t *zvec_legacy_func;
220 zfs_ioc_func_t *zvec_func;
221 zfs_secpolicy_func_t *zvec_secpolicy;
222 zfs_ioc_namecheck_t zvec_namecheck;
223 boolean_t zvec_allow_log;
224 zfs_ioc_poolcheck_t zvec_pool_check;
225 boolean_t zvec_smush_outnvlist;
226 const char *zvec_name;
227 } zfs_ioc_vec_t;
228
229 /* This array is indexed by zfs_userquota_prop_t */
230 static const char *userquota_perms[] = {
231 ZFS_DELEG_PERM_USERUSED,
232 ZFS_DELEG_PERM_USERQUOTA,
233 ZFS_DELEG_PERM_GROUPUSED,
234 ZFS_DELEG_PERM_GROUPQUOTA,
235 };
236
237 static int zfs_ioc_userspace_upgrade(zfs_cmd_t *zc);
238 static int zfs_check_settable(const char *name, nvpair_t *property,
239 cred_t *cr);
240 static int zfs_check_clearable(char *dataset, nvlist_t *props,
241 nvlist_t **errors);
242 static int zfs_fill_zplprops_root(uint64_t, nvlist_t *, nvlist_t *,
243 boolean_t *);
244 int zfs_set_prop_nvlist(const char *, zprop_source_t, nvlist_t *, nvlist_t *);
245 static int get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp);
246
247 static int zfs_prop_activate_feature(spa_t *spa, spa_feature_t feature);
248
249 /* _NOTE(PRINTFLIKE(4)) - this is printf-like, but lint is too whiney */
250 void
251 __dprintf(const char *file, const char *func, int line, const char *fmt, ...)
252 {
253 const char *newfile;
254 char buf[512];
255 va_list adx;
256
257 /*
258 * Get rid of annoying "../common/" prefix to filename.
259 */
260 newfile = strrchr(file, '/');
261 if (newfile != NULL) {
262 newfile = newfile + 1; /* Get rid of leading / */
263 } else {
264 newfile = file;
265 }
266
267 va_start(adx, fmt);
268 (void) vsnprintf(buf, sizeof (buf), fmt, adx);
269 va_end(adx);
270
271 /*
272 * To get this data, use the zfs-dprintf probe as so:
273 * dtrace -q -n 'zfs-dprintf \
274 * /stringof(arg0) == "dbuf.c"/ \
275 * {printf("%s: %s", stringof(arg1), stringof(arg3))}'
276 * arg0 = file name
277 * arg1 = function name
278 * arg2 = line number
279 * arg3 = message
280 */
281 DTRACE_PROBE4(zfs__dprintf,
282 char *, newfile, char *, func, int, line, char *, buf);
283 }
284
285 static void
286 history_str_free(char *buf)
287 {
288 kmem_free(buf, HIS_MAX_RECORD_LEN);
289 }
290
291 static char *
292 history_str_get(zfs_cmd_t *zc)
293 {
294 char *buf;
295
296 if (zc->zc_history == NULL)
297 return (NULL);
298
299 buf = kmem_alloc(HIS_MAX_RECORD_LEN, KM_SLEEP);
300 if (copyinstr((void *)(uintptr_t)zc->zc_history,
301 buf, HIS_MAX_RECORD_LEN, NULL) != 0) {
302 history_str_free(buf);
303 return (NULL);
304 }
305
306 buf[HIS_MAX_RECORD_LEN -1] = '\0';
307
308 return (buf);
309 }
310
311 /*
312 * Check to see if the named dataset is currently defined as bootable
313 */
314 static boolean_t
315 zfs_is_bootfs(const char *name)
316 {
317 objset_t *os;
318
319 if (dmu_objset_hold(name, FTAG, &os) == 0) {
320 boolean_t ret;
321 ret = (dmu_objset_id(os) == spa_bootfs(dmu_objset_spa(os)));
322 dmu_objset_rele(os, FTAG);
323 return (ret);
324 }
325 return (B_FALSE);
326 }
327
328 /*
329 * Return non-zero if the spa version is less than requested version.
330 */
331 static int
332 zfs_earlier_version(const char *name, int version)
333 {
334 spa_t *spa;
335
336 if (spa_open(name, &spa, FTAG) == 0) {
337 if (spa_version(spa) < version) {
338 spa_close(spa, FTAG);
339 return (1);
340 }
341 spa_close(spa, FTAG);
342 }
343 return (0);
344 }
345
346 /*
347 * Return TRUE if the ZPL version is less than requested version.
348 */
349 static boolean_t
350 zpl_earlier_version(const char *name, int version)
351 {
352 objset_t *os;
353 boolean_t rc = B_TRUE;
354
355 if (dmu_objset_hold(name, FTAG, &os) == 0) {
356 uint64_t zplversion;
357
358 if (dmu_objset_type(os) != DMU_OST_ZFS) {
359 dmu_objset_rele(os, FTAG);
360 return (B_TRUE);
361 }
362 /* XXX reading from non-owned objset */
363 if (zfs_get_zplprop(os, ZFS_PROP_VERSION, &zplversion) == 0)
364 rc = zplversion < version;
365 dmu_objset_rele(os, FTAG);
366 }
367 return (rc);
368 }
369
370 static void
371 zfs_log_history(zfs_cmd_t *zc)
372 {
373 spa_t *spa;
374 char *buf;
375
376 if ((buf = history_str_get(zc)) == NULL)
377 return;
378
379 if (spa_open(zc->zc_name, &spa, FTAG) == 0) {
380 if (spa_version(spa) >= SPA_VERSION_ZPOOL_HISTORY)
381 (void) spa_history_log(spa, buf);
382 spa_close(spa, FTAG);
383 }
384 history_str_free(buf);
385 }
386
387 /*
388 * Policy for top-level read operations (list pools). Requires no privileges,
389 * and can be used in the local zone, as there is no associated dataset.
390 */
391 /* ARGSUSED */
392 static int
393 zfs_secpolicy_none(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
394 {
395 return (0);
396 }
397
398 /*
399 * Policy for dataset read operations (list children, get statistics). Requires
400 * no privileges, but must be visible in the local zone.
401 */
402 /* ARGSUSED */
403 static int
404 zfs_secpolicy_read(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
405 {
406 if (INGLOBALZONE(curproc) ||
407 zone_dataset_visible(zc->zc_name, NULL))
408 return (0);
409
410 return (SET_ERROR(ENOENT));
411 }
412
413 static int
414 zfs_dozonecheck_impl(const char *dataset, uint64_t zoned, cred_t *cr)
415 {
416 int writable = 1;
417
418 /*
419 * The dataset must be visible by this zone -- check this first
420 * so they don't see EPERM on something they shouldn't know about.
421 */
422 if (!INGLOBALZONE(curproc) &&
423 !zone_dataset_visible(dataset, &writable))
424 return (SET_ERROR(ENOENT));
425
426 if (INGLOBALZONE(curproc)) {
427 /*
428 * If the fs is zoned, only root can access it from the
429 * global zone.
430 */
431 if (secpolicy_zfs(cr) && zoned)
432 return (SET_ERROR(EPERM));
433 } else {
434 /*
435 * If we are in a local zone, the 'zoned' property must be set.
436 */
437 if (!zoned)
438 return (SET_ERROR(EPERM));
439
440 /* must be writable by this zone */
441 if (!writable)
442 return (SET_ERROR(EPERM));
443 }
444 return (0);
445 }
446
447 static int
448 zfs_dozonecheck(const char *dataset, cred_t *cr)
449 {
450 uint64_t zoned;
451
452 if (dsl_prop_get_integer(dataset, "zoned", &zoned, NULL))
453 return (SET_ERROR(ENOENT));
454
455 return (zfs_dozonecheck_impl(dataset, zoned, cr));
456 }
457
458 static int
459 zfs_dozonecheck_ds(const char *dataset, dsl_dataset_t *ds, cred_t *cr)
460 {
461 uint64_t zoned;
462
463 if (dsl_prop_get_int_ds(ds, "zoned", &zoned))
464 return (SET_ERROR(ENOENT));
465
466 return (zfs_dozonecheck_impl(dataset, zoned, cr));
467 }
468
469 static int
470 zfs_secpolicy_write_perms_ds(const char *name, dsl_dataset_t *ds,
471 const char *perm, cred_t *cr)
472 {
473 int error;
474
475 error = zfs_dozonecheck_ds(name, ds, cr);
476 if (error == 0) {
477 error = secpolicy_zfs(cr);
478 if (error != 0)
479 error = dsl_deleg_access_impl(ds, perm, cr);
480 }
481 return (error);
482 }
483
484 static int
485 zfs_secpolicy_write_perms(const char *name, const char *perm, cred_t *cr)
486 {
487 int error;
488 dsl_dataset_t *ds;
489 dsl_pool_t *dp;
490
491 error = dsl_pool_hold(name, FTAG, &dp);
492 if (error != 0)
493 return (error);
494
495 error = dsl_dataset_hold(dp, name, FTAG, &ds);
496 if (error != 0) {
497 dsl_pool_rele(dp, FTAG);
498 return (error);
499 }
500
501 error = zfs_secpolicy_write_perms_ds(name, ds, perm, cr);
502
503 dsl_dataset_rele(ds, FTAG);
504 dsl_pool_rele(dp, FTAG);
505 return (error);
506 }
507
508 /*
509 * Policy for setting the security label property.
510 *
511 * Returns 0 for success, non-zero for access and other errors.
512 */
513 static int
514 zfs_set_slabel_policy(const char *name, char *strval, cred_t *cr)
515 {
516 char ds_hexsl[MAXNAMELEN];
517 bslabel_t ds_sl, new_sl;
518 boolean_t new_default = FALSE;
519 uint64_t zoned;
520 int needed_priv = -1;
521 int error;
522
523 /* First get the existing dataset label. */
524 error = dsl_prop_get(name, zfs_prop_to_name(ZFS_PROP_MLSLABEL),
525 1, sizeof (ds_hexsl), &ds_hexsl, NULL);
526 if (error != 0)
527 return (SET_ERROR(EPERM));
528
529 if (strcasecmp(strval, ZFS_MLSLABEL_DEFAULT) == 0)
530 new_default = TRUE;
531
532 /* The label must be translatable */
533 if (!new_default && (hexstr_to_label(strval, &new_sl) != 0))
534 return (SET_ERROR(EINVAL));
535
536 /*
537 * In a non-global zone, disallow attempts to set a label that
538 * doesn't match that of the zone; otherwise no other checks
539 * are needed.
540 */
541 if (!INGLOBALZONE(curproc)) {
542 if (new_default || !blequal(&new_sl, CR_SL(CRED())))
543 return (SET_ERROR(EPERM));
544 return (0);
545 }
546
547 /*
548 * For global-zone datasets (i.e., those whose zoned property is
549 * "off", verify that the specified new label is valid for the
550 * global zone.
551 */
552 if (dsl_prop_get_integer(name,
553 zfs_prop_to_name(ZFS_PROP_ZONED), &zoned, NULL))
554 return (SET_ERROR(EPERM));
555 if (!zoned) {
556 if (zfs_check_global_label(name, strval) != 0)
557 return (SET_ERROR(EPERM));
558 }
559
560 /*
561 * If the existing dataset label is nondefault, check if the
562 * dataset is mounted (label cannot be changed while mounted).
563 * Get the zfsvfs; if there isn't one, then the dataset isn't
564 * mounted (or isn't a dataset, doesn't exist, ...).
565 */
566 if (strcasecmp(ds_hexsl, ZFS_MLSLABEL_DEFAULT) != 0) {
567 objset_t *os;
568 static char *setsl_tag = "setsl_tag";
569
570 /*
571 * Try to own the dataset; abort if there is any error,
572 * (e.g., already mounted, in use, or other error).
573 */
574 error = dmu_objset_own(name, DMU_OST_ZFS, B_TRUE,
575 setsl_tag, &os);
576 if (error != 0)
577 return (SET_ERROR(EPERM));
578
579 dmu_objset_disown(os, setsl_tag);
580
581 if (new_default) {
582 needed_priv = PRIV_FILE_DOWNGRADE_SL;
583 goto out_check;
584 }
585
586 if (hexstr_to_label(strval, &new_sl) != 0)
587 return (SET_ERROR(EPERM));
588
589 if (blstrictdom(&ds_sl, &new_sl))
590 needed_priv = PRIV_FILE_DOWNGRADE_SL;
591 else if (blstrictdom(&new_sl, &ds_sl))
592 needed_priv = PRIV_FILE_UPGRADE_SL;
593 } else {
594 /* dataset currently has a default label */
595 if (!new_default)
596 needed_priv = PRIV_FILE_UPGRADE_SL;
597 }
598
599 out_check:
600 if (needed_priv != -1)
601 return (PRIV_POLICY(cr, needed_priv, B_FALSE, EPERM, NULL));
602 return (0);
603 }
604
605 static int
606 zfs_secpolicy_setprop(const char *dsname, zfs_prop_t prop, nvpair_t *propval,
607 cred_t *cr)
608 {
609 char *strval;
610
611 /*
612 * Check permissions for special properties.
613 */
614 switch (prop) {
615 case ZFS_PROP_ZONED:
616 /*
617 * Disallow setting of 'zoned' from within a local zone.
618 */
619 if (!INGLOBALZONE(curproc))
620 return (SET_ERROR(EPERM));
621 break;
622
623 case ZFS_PROP_QUOTA:
624 case ZFS_PROP_FILESYSTEM_LIMIT:
625 case ZFS_PROP_SNAPSHOT_LIMIT:
626 if (!INGLOBALZONE(curproc)) {
627 uint64_t zoned;
628 char setpoint[MAXNAMELEN];
629 /*
630 * Unprivileged users are allowed to modify the
631 * limit 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 (SET_ERROR(EPERM));
637 if (!zoned || strlen(dsname) <= strlen(setpoint))
638 return (SET_ERROR(EPERM));
639 }
640 break;
641
642 case ZFS_PROP_MLSLABEL:
643 if (!is_system_labeled())
644 return (SET_ERROR(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 != 0)
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 dsl_pool_t *dp;
689 dsl_dataset_t *ds;
690 char *cp;
691 int error;
692
693 /*
694 * Generate the current snapshot name from the given objsetid, then
695 * use that name for the secpolicy/zone checks.
696 */
697 cp = strchr(zc->zc_name, '@');
698 if (cp == NULL)
699 return (SET_ERROR(EINVAL));
700 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
701 if (error != 0)
702 return (error);
703
704 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &ds);
705 if (error != 0) {
706 dsl_pool_rele(dp, FTAG);
707 return (error);
708 }
709
710 dsl_dataset_name(ds, zc->zc_name);
711
712 error = zfs_secpolicy_write_perms_ds(zc->zc_name, ds,
713 ZFS_DELEG_PERM_SEND, cr);
714 dsl_dataset_rele(ds, FTAG);
715 dsl_pool_rele(dp, FTAG);
716
717 return (error);
718 }
719
720 /* ARGSUSED */
721 static int
722 zfs_secpolicy_send_new(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
723 {
724 return (zfs_secpolicy_write_perms(zc->zc_name,
725 ZFS_DELEG_PERM_SEND, cr));
726 }
727
728 /* ARGSUSED */
729 static int
730 zfs_secpolicy_deleg_share(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
731 {
732 vnode_t *vp;
733 int error;
734
735 if ((error = lookupname(zc->zc_value, UIO_SYSSPACE,
736 NO_FOLLOW, NULL, &vp)) != 0)
737 return (error);
738
739 /* Now make sure mntpnt and dataset are ZFS */
740
741 if (vp->v_vfsp->vfs_fstype != zfsfstype ||
742 (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource),
743 zc->zc_name) != 0)) {
744 VN_RELE(vp);
745 return (SET_ERROR(EPERM));
746 }
747
748 VN_RELE(vp);
749 return (dsl_deleg_access(zc->zc_name,
750 ZFS_DELEG_PERM_SHARE, cr));
751 }
752
753 int
754 zfs_secpolicy_share(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
755 {
756 if (!INGLOBALZONE(curproc))
757 return (SET_ERROR(EPERM));
758
759 if (secpolicy_nfs(cr) == 0) {
760 return (0);
761 } else {
762 return (zfs_secpolicy_deleg_share(zc, innvl, cr));
763 }
764 }
765
766 int
767 zfs_secpolicy_smb_acl(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
768 {
769 if (!INGLOBALZONE(curproc))
770 return (SET_ERROR(EPERM));
771
772 if (secpolicy_smb(cr) == 0) {
773 return (0);
774 } else {
775 return (zfs_secpolicy_deleg_share(zc, innvl, cr));
776 }
777 }
778
779 static int
780 zfs_get_parent(const char *datasetname, char *parent, int parentsize)
781 {
782 char *cp;
783
784 /*
785 * Remove the @bla or /bla from the end of the name to get the parent.
786 */
787 (void) strncpy(parent, datasetname, parentsize);
788 cp = strrchr(parent, '@');
789 if (cp != NULL) {
790 cp[0] = '\0';
791 } else {
792 cp = strrchr(parent, '/');
793 if (cp == NULL)
794 return (SET_ERROR(ENOENT));
795 cp[0] = '\0';
796 }
797
798 return (0);
799 }
800
801 int
802 zfs_secpolicy_destroy_perms(const char *name, cred_t *cr)
803 {
804 int error;
805
806 if ((error = zfs_secpolicy_write_perms(name,
807 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
808 return (error);
809
810 return (zfs_secpolicy_write_perms(name, ZFS_DELEG_PERM_DESTROY, cr));
811 }
812
813 /* ARGSUSED */
814 static int
815 zfs_secpolicy_destroy(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
816 {
817 return (zfs_secpolicy_destroy_perms(zc->zc_name, cr));
818 }
819
820 /*
821 * Destroying snapshots with delegated permissions requires
822 * descendant mount and destroy permissions.
823 */
824 /* ARGSUSED */
825 static int
826 zfs_secpolicy_destroy_snaps(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
827 {
828 nvlist_t *snaps;
829 nvpair_t *pair, *nextpair;
830 int error = 0;
831
832 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
833 return (SET_ERROR(EINVAL));
834 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
835 pair = nextpair) {
836 nextpair = nvlist_next_nvpair(snaps, pair);
837 error = zfs_secpolicy_destroy_perms(nvpair_name(pair), cr);
838 if (error == ENOENT) {
839 /*
840 * Ignore any snapshots that don't exist (we consider
841 * them "already destroyed"). Remove the name from the
842 * nvl here in case the snapshot is created between
843 * now and when we try to destroy it (in which case
844 * we don't want to destroy it since we haven't
845 * checked for permission).
846 */
847 fnvlist_remove_nvpair(snaps, pair);
848 error = 0;
849 }
850 if (error != 0)
851 break;
852 }
853
854 return (error);
855 }
856
857 int
858 zfs_secpolicy_rename_perms(const char *from, const char *to, cred_t *cr)
859 {
860 char parentname[MAXNAMELEN];
861 int error;
862
863 if ((error = zfs_secpolicy_write_perms(from,
864 ZFS_DELEG_PERM_RENAME, cr)) != 0)
865 return (error);
866
867 if ((error = zfs_secpolicy_write_perms(from,
868 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
869 return (error);
870
871 if ((error = zfs_get_parent(to, parentname,
872 sizeof (parentname))) != 0)
873 return (error);
874
875 if ((error = zfs_secpolicy_write_perms(parentname,
876 ZFS_DELEG_PERM_CREATE, cr)) != 0)
877 return (error);
878
879 if ((error = zfs_secpolicy_write_perms(parentname,
880 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
881 return (error);
882
883 return (error);
884 }
885
886 /* ARGSUSED */
887 static int
888 zfs_secpolicy_rename(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
889 {
890 return (zfs_secpolicy_rename_perms(zc->zc_name, zc->zc_value, cr));
891 }
892
893 /* ARGSUSED */
894 static int
895 zfs_secpolicy_promote(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
896 {
897 dsl_pool_t *dp;
898 dsl_dataset_t *clone;
899 int error;
900
901 error = zfs_secpolicy_write_perms(zc->zc_name,
902 ZFS_DELEG_PERM_PROMOTE, cr);
903 if (error != 0)
904 return (error);
905
906 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
907 if (error != 0)
908 return (error);
909
910 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &clone);
911
912 if (error == 0) {
913 char parentname[MAXNAMELEN];
914 dsl_dataset_t *origin = NULL;
915 dsl_dir_t *dd;
916 dd = clone->ds_dir;
917
918 error = dsl_dataset_hold_obj(dd->dd_pool,
919 dsl_dir_phys(dd)->dd_origin_obj, FTAG, &origin);
920 if (error != 0) {
921 dsl_dataset_rele(clone, FTAG);
922 dsl_pool_rele(dp, FTAG);
923 return (error);
924 }
925
926 error = zfs_secpolicy_write_perms_ds(zc->zc_name, clone,
927 ZFS_DELEG_PERM_MOUNT, cr);
928
929 dsl_dataset_name(origin, parentname);
930 if (error == 0) {
931 error = zfs_secpolicy_write_perms_ds(parentname, origin,
932 ZFS_DELEG_PERM_PROMOTE, cr);
933 }
934 dsl_dataset_rele(clone, FTAG);
935 dsl_dataset_rele(origin, FTAG);
936 }
937 dsl_pool_rele(dp, FTAG);
938 return (error);
939 }
940
941 /* ARGSUSED */
942 static int
943 zfs_secpolicy_recv(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
944 {
945 int error;
946
947 if ((error = zfs_secpolicy_write_perms(zc->zc_name,
948 ZFS_DELEG_PERM_RECEIVE, cr)) != 0)
949 return (error);
950
951 if ((error = zfs_secpolicy_write_perms(zc->zc_name,
952 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
953 return (error);
954
955 return (zfs_secpolicy_write_perms(zc->zc_name,
956 ZFS_DELEG_PERM_CREATE, cr));
957 }
958
959 int
960 zfs_secpolicy_snapshot_perms(const char *name, cred_t *cr)
961 {
962 return (zfs_secpolicy_write_perms(name,
963 ZFS_DELEG_PERM_SNAPSHOT, cr));
964 }
965
966 /*
967 * Check for permission to create each snapshot in the nvlist.
968 */
969 /* ARGSUSED */
970 static int
971 zfs_secpolicy_snapshot(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
972 {
973 nvlist_t *snaps;
974 int error = 0;
975 nvpair_t *pair;
976
977 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
978 return (SET_ERROR(EINVAL));
979 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
980 pair = nvlist_next_nvpair(snaps, pair)) {
981 char *name = nvpair_name(pair);
982 char *atp = strchr(name, '@');
983
984 if (atp == NULL) {
985 error = SET_ERROR(EINVAL);
986 break;
987 }
988 *atp = '\0';
989 error = zfs_secpolicy_snapshot_perms(name, cr);
990 *atp = '@';
991 if (error != 0)
992 break;
993 }
994 return (error);
995 }
996
997 /*
998 * Check for permission to create each snapshot in the nvlist.
999 */
1000 /* ARGSUSED */
1001 static int
1002 zfs_secpolicy_bookmark(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1003 {
1004 int error = 0;
1005
1006 for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL);
1007 pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) {
1008 char *name = nvpair_name(pair);
1009 char *hashp = strchr(name, '#');
1010
1011 if (hashp == NULL) {
1012 error = SET_ERROR(EINVAL);
1013 break;
1014 }
1015 *hashp = '\0';
1016 error = zfs_secpolicy_write_perms(name,
1017 ZFS_DELEG_PERM_BOOKMARK, cr);
1018 *hashp = '#';
1019 if (error != 0)
1020 break;
1021 }
1022 return (error);
1023 }
1024
1025 /* ARGSUSED */
1026 static int
1027 zfs_secpolicy_destroy_bookmarks(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1028 {
1029 nvpair_t *pair, *nextpair;
1030 int error = 0;
1031
1032 for (pair = nvlist_next_nvpair(innvl, NULL); pair != NULL;
1033 pair = nextpair) {
1034 char *name = nvpair_name(pair);
1035 char *hashp = strchr(name, '#');
1036 nextpair = nvlist_next_nvpair(innvl, pair);
1037
1038 if (hashp == NULL) {
1039 error = SET_ERROR(EINVAL);
1040 break;
1041 }
1042
1043 *hashp = '\0';
1044 error = zfs_secpolicy_write_perms(name,
1045 ZFS_DELEG_PERM_DESTROY, cr);
1046 *hashp = '#';
1047 if (error == ENOENT) {
1048 /*
1049 * Ignore any filesystems that don't exist (we consider
1050 * their bookmarks "already destroyed"). Remove
1051 * the name from the nvl here in case the filesystem
1052 * is created between now and when we try to destroy
1053 * the bookmark (in which case we don't want to
1054 * destroy it since we haven't checked for permission).
1055 */
1056 fnvlist_remove_nvpair(innvl, pair);
1057 error = 0;
1058 }
1059 if (error != 0)
1060 break;
1061 }
1062
1063 return (error);
1064 }
1065
1066 /* ARGSUSED */
1067 static int
1068 zfs_secpolicy_log_history(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1069 {
1070 /*
1071 * Even root must have a proper TSD so that we know what pool
1072 * to log to.
1073 */
1074 if (tsd_get(zfs_allow_log_key) == NULL)
1075 return (SET_ERROR(EPERM));
1076 return (0);
1077 }
1078
1079 static int
1080 zfs_secpolicy_create_clone(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1081 {
1082 char parentname[MAXNAMELEN];
1083 int error;
1084 char *origin;
1085
1086 if ((error = zfs_get_parent(zc->zc_name, parentname,
1087 sizeof (parentname))) != 0)
1088 return (error);
1089
1090 if (nvlist_lookup_string(innvl, "origin", &origin) == 0 &&
1091 (error = zfs_secpolicy_write_perms(origin,
1092 ZFS_DELEG_PERM_CLONE, cr)) != 0)
1093 return (error);
1094
1095 if ((error = zfs_secpolicy_write_perms(parentname,
1096 ZFS_DELEG_PERM_CREATE, cr)) != 0)
1097 return (error);
1098
1099 return (zfs_secpolicy_write_perms(parentname,
1100 ZFS_DELEG_PERM_MOUNT, cr));
1101 }
1102
1103 /*
1104 * Policy for pool operations - create/destroy pools, add vdevs, etc. Requires
1105 * SYS_CONFIG privilege, which is not available in a local zone.
1106 */
1107 /* ARGSUSED */
1108 static int
1109 zfs_secpolicy_config(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1110 {
1111 if (secpolicy_sys_config(cr, B_FALSE) != 0)
1112 return (SET_ERROR(EPERM));
1113
1114 return (0);
1115 }
1116
1117 /*
1118 * Policy for object to name lookups.
1119 */
1120 /* ARGSUSED */
1121 static int
1122 zfs_secpolicy_diff(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1123 {
1124 int error;
1125
1126 if ((error = secpolicy_sys_config(cr, B_FALSE)) == 0)
1127 return (0);
1128
1129 error = zfs_secpolicy_write_perms(zc->zc_name, ZFS_DELEG_PERM_DIFF, cr);
1130 return (error);
1131 }
1132
1133 /*
1134 * Policy for fault injection. Requires all privileges.
1135 */
1136 /* ARGSUSED */
1137 static int
1138 zfs_secpolicy_inject(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1139 {
1140 return (secpolicy_zinject(cr));
1141 }
1142
1143 /* ARGSUSED */
1144 static int
1145 zfs_secpolicy_inherit_prop(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1146 {
1147 zfs_prop_t prop = zfs_name_to_prop(zc->zc_value);
1148
1149 if (prop == ZPROP_INVAL) {
1150 if (!zfs_prop_user(zc->zc_value))
1151 return (SET_ERROR(EINVAL));
1152 return (zfs_secpolicy_write_perms(zc->zc_name,
1153 ZFS_DELEG_PERM_USERPROP, cr));
1154 } else {
1155 return (zfs_secpolicy_setprop(zc->zc_name, prop,
1156 NULL, cr));
1157 }
1158 }
1159
1160 static int
1161 zfs_secpolicy_userspace_one(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1162 {
1163 int err = zfs_secpolicy_read(zc, innvl, cr);
1164 if (err)
1165 return (err);
1166
1167 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
1168 return (SET_ERROR(EINVAL));
1169
1170 if (zc->zc_value[0] == 0) {
1171 /*
1172 * They are asking about a posix uid/gid. If it's
1173 * themself, allow it.
1174 */
1175 if (zc->zc_objset_type == ZFS_PROP_USERUSED ||
1176 zc->zc_objset_type == ZFS_PROP_USERQUOTA) {
1177 if (zc->zc_guid == crgetuid(cr))
1178 return (0);
1179 } else {
1180 if (groupmember(zc->zc_guid, cr))
1181 return (0);
1182 }
1183 }
1184
1185 return (zfs_secpolicy_write_perms(zc->zc_name,
1186 userquota_perms[zc->zc_objset_type], cr));
1187 }
1188
1189 static int
1190 zfs_secpolicy_userspace_many(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1191 {
1192 int err = zfs_secpolicy_read(zc, innvl, cr);
1193 if (err)
1194 return (err);
1195
1196 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
1197 return (SET_ERROR(EINVAL));
1198
1199 return (zfs_secpolicy_write_perms(zc->zc_name,
1200 userquota_perms[zc->zc_objset_type], cr));
1201 }
1202
1203 /* ARGSUSED */
1204 static int
1205 zfs_secpolicy_userspace_upgrade(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1206 {
1207 return (zfs_secpolicy_setprop(zc->zc_name, ZFS_PROP_VERSION,
1208 NULL, cr));
1209 }
1210
1211 /* ARGSUSED */
1212 static int
1213 zfs_secpolicy_hold(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1214 {
1215 nvpair_t *pair;
1216 nvlist_t *holds;
1217 int error;
1218
1219 error = nvlist_lookup_nvlist(innvl, "holds", &holds);
1220 if (error != 0)
1221 return (SET_ERROR(EINVAL));
1222
1223 for (pair = nvlist_next_nvpair(holds, NULL); pair != NULL;
1224 pair = nvlist_next_nvpair(holds, pair)) {
1225 char fsname[MAXNAMELEN];
1226 error = dmu_fsname(nvpair_name(pair), fsname);
1227 if (error != 0)
1228 return (error);
1229 error = zfs_secpolicy_write_perms(fsname,
1230 ZFS_DELEG_PERM_HOLD, cr);
1231 if (error != 0)
1232 return (error);
1233 }
1234 return (0);
1235 }
1236
1237 /* ARGSUSED */
1238 static int
1239 zfs_secpolicy_release(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1240 {
1241 nvpair_t *pair;
1242 int error;
1243
1244 for (pair = nvlist_next_nvpair(innvl, NULL); pair != NULL;
1245 pair = nvlist_next_nvpair(innvl, pair)) {
1246 char fsname[MAXNAMELEN];
1247 error = dmu_fsname(nvpair_name(pair), fsname);
1248 if (error != 0)
1249 return (error);
1250 error = zfs_secpolicy_write_perms(fsname,
1251 ZFS_DELEG_PERM_RELEASE, cr);
1252 if (error != 0)
1253 return (error);
1254 }
1255 return (0);
1256 }
1257
1258 /*
1259 * Policy for allowing temporary snapshots to be taken or released
1260 */
1261 static int
1262 zfs_secpolicy_tmp_snapshot(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1263 {
1264 /*
1265 * A temporary snapshot is the same as a snapshot,
1266 * hold, destroy and release all rolled into one.
1267 * Delegated diff alone is sufficient that we allow this.
1268 */
1269 int error;
1270
1271 if ((error = zfs_secpolicy_write_perms(zc->zc_name,
1272 ZFS_DELEG_PERM_DIFF, cr)) == 0)
1273 return (0);
1274
1275 error = zfs_secpolicy_snapshot_perms(zc->zc_name, cr);
1276 if (error == 0)
1277 error = zfs_secpolicy_hold(zc, innvl, cr);
1278 if (error == 0)
1279 error = zfs_secpolicy_release(zc, innvl, cr);
1280 if (error == 0)
1281 error = zfs_secpolicy_destroy(zc, innvl, cr);
1282 return (error);
1283 }
1284
1285 /*
1286 * Returns the nvlist as specified by the user in the zfs_cmd_t.
1287 */
1288 static int
1289 get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp)
1290 {
1291 char *packed;
1292 int error;
1293 nvlist_t *list = NULL;
1294
1295 /*
1296 * Read in and unpack the user-supplied nvlist.
1297 */
1298 if (size == 0)
1299 return (SET_ERROR(EINVAL));
1300
1301 packed = kmem_alloc(size, KM_SLEEP);
1302
1303 if ((error = ddi_copyin((void *)(uintptr_t)nvl, packed, size,
1304 iflag)) != 0) {
1305 kmem_free(packed, size);
1306 return (error);
1307 }
1308
1309 if ((error = nvlist_unpack(packed, size, &list, 0)) != 0) {
1310 kmem_free(packed, size);
1311 return (error);
1312 }
1313
1314 kmem_free(packed, size);
1315
1316 *nvp = list;
1317 return (0);
1318 }
1319
1320 /*
1321 * Reduce the size of this nvlist until it can be serialized in 'max' bytes.
1322 * Entries will be removed from the end of the nvlist, and one int32 entry
1323 * named "N_MORE_ERRORS" will be added indicating how many entries were
1324 * removed.
1325 */
1326 static int
1327 nvlist_smush(nvlist_t *errors, size_t max)
1328 {
1329 size_t size;
1330
1331 size = fnvlist_size(errors);
1332
1333 if (size > max) {
1334 nvpair_t *more_errors;
1335 int n = 0;
1336
1337 if (max < 1024)
1338 return (SET_ERROR(ENOMEM));
1339
1340 fnvlist_add_int32(errors, ZPROP_N_MORE_ERRORS, 0);
1341 more_errors = nvlist_prev_nvpair(errors, NULL);
1342
1343 do {
1344 nvpair_t *pair = nvlist_prev_nvpair(errors,
1345 more_errors);
1346 fnvlist_remove_nvpair(errors, pair);
1347 n++;
1348 size = fnvlist_size(errors);
1349 } while (size > max);
1350
1351 fnvlist_remove_nvpair(errors, more_errors);
1352 fnvlist_add_int32(errors, ZPROP_N_MORE_ERRORS, n);
1353 ASSERT3U(fnvlist_size(errors), <=, max);
1354 }
1355
1356 return (0);
1357 }
1358
1359 static int
1360 put_nvlist(zfs_cmd_t *zc, nvlist_t *nvl)
1361 {
1362 char *packed = NULL;
1363 int error = 0;
1364 size_t size;
1365
1366 size = fnvlist_size(nvl);
1367
1368 if (size > zc->zc_nvlist_dst_size) {
1369 error = SET_ERROR(ENOMEM);
1370 } else {
1371 packed = fnvlist_pack(nvl, &size);
1372 if (ddi_copyout(packed, (void *)(uintptr_t)zc->zc_nvlist_dst,
1373 size, zc->zc_iflags) != 0)
1374 error = SET_ERROR(EFAULT);
1375 fnvlist_pack_free(packed, size);
1376 }
1377
1378 zc->zc_nvlist_dst_size = size;
1379 zc->zc_nvlist_dst_filled = B_TRUE;
1380 return (error);
1381 }
1382
1383 static int
1384 getzfsvfs(const char *dsname, zfsvfs_t **zfvp)
1385 {
1386 objset_t *os;
1387 int error;
1388
1389 error = dmu_objset_hold(dsname, FTAG, &os);
1390 if (error != 0)
1391 return (error);
1392 if (dmu_objset_type(os) != DMU_OST_ZFS) {
1393 dmu_objset_rele(os, FTAG);
1394 return (SET_ERROR(EINVAL));
1395 }
1396
1397 mutex_enter(&os->os_user_ptr_lock);
1398 *zfvp = dmu_objset_get_user(os);
1399 if (*zfvp) {
1400 VFS_HOLD((*zfvp)->z_vfs);
1401 } else {
1402 error = SET_ERROR(ESRCH);
1403 }
1404 mutex_exit(&os->os_user_ptr_lock);
1405 dmu_objset_rele(os, FTAG);
1406 return (error);
1407 }
1408
1409 /*
1410 * Find a zfsvfs_t for a mounted filesystem, or create our own, in which
1411 * case its z_vfs will be NULL, and it will be opened as the owner.
1412 * If 'writer' is set, the z_teardown_lock will be held for RW_WRITER,
1413 * which prevents all vnode ops from running.
1414 */
1415 static int
1416 zfsvfs_hold(const char *name, void *tag, zfsvfs_t **zfvp, boolean_t writer)
1417 {
1418 int error = 0;
1419
1420 if (getzfsvfs(name, zfvp) != 0)
1421 error = zfsvfs_create(name, zfvp);
1422 if (error == 0) {
1423 rrm_enter(&(*zfvp)->z_teardown_lock, (writer) ? RW_WRITER :
1424 RW_READER, tag);
1425 if ((*zfvp)->z_unmounted) {
1426 /*
1427 * XXX we could probably try again, since the unmounting
1428 * thread should be just about to disassociate the
1429 * objset from the zfsvfs.
1430 */
1431 rrm_exit(&(*zfvp)->z_teardown_lock, tag);
1432 return (SET_ERROR(EBUSY));
1433 }
1434 }
1435 return (error);
1436 }
1437
1438 static void
1439 zfsvfs_rele(zfsvfs_t *zfsvfs, void *tag)
1440 {
1441 rrm_exit(&zfsvfs->z_teardown_lock, tag);
1442
1443 if (zfsvfs->z_vfs) {
1444 VFS_RELE(zfsvfs->z_vfs);
1445 } else {
1446 dmu_objset_disown(zfsvfs->z_os, zfsvfs);
1447 zfsvfs_free(zfsvfs);
1448 }
1449 }
1450
1451 static int
1452 zfs_ioc_pool_create(zfs_cmd_t *zc)
1453 {
1454 int error;
1455 nvlist_t *config, *props = NULL;
1456 nvlist_t *rootprops = NULL;
1457 nvlist_t *zplprops = NULL;
1458
1459 if (error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1460 zc->zc_iflags, &config))
1461 return (error);
1462
1463 if (zc->zc_nvlist_src_size != 0 && (error =
1464 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1465 zc->zc_iflags, &props))) {
1466 nvlist_free(config);
1467 return (error);
1468 }
1469
1470 if (props) {
1471 nvlist_t *nvl = NULL;
1472 uint64_t version = SPA_VERSION;
1473
1474 (void) nvlist_lookup_uint64(props,
1475 zpool_prop_to_name(ZPOOL_PROP_VERSION), &version);
1476 if (!SPA_VERSION_IS_SUPPORTED(version)) {
1477 error = SET_ERROR(EINVAL);
1478 goto pool_props_bad;
1479 }
1480 (void) nvlist_lookup_nvlist(props, ZPOOL_ROOTFS_PROPS, &nvl);
1481 if (nvl) {
1482 error = nvlist_dup(nvl, &rootprops, KM_SLEEP);
1483 if (error != 0) {
1484 nvlist_free(config);
1485 nvlist_free(props);
1486 return (error);
1487 }
1488 (void) nvlist_remove_all(props, ZPOOL_ROOTFS_PROPS);
1489 }
1490 VERIFY(nvlist_alloc(&zplprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
1491 error = zfs_fill_zplprops_root(version, rootprops,
1492 zplprops, NULL);
1493 if (error != 0)
1494 goto pool_props_bad;
1495 }
1496
1497 error = spa_create(zc->zc_name, config, props, zplprops);
1498
1499 /*
1500 * Set the remaining root properties
1501 */
1502 if (!error && (error = zfs_set_prop_nvlist(zc->zc_name,
1503 ZPROP_SRC_LOCAL, rootprops, NULL)) != 0)
1504 (void) spa_destroy(zc->zc_name);
1505
1506 pool_props_bad:
1507 nvlist_free(rootprops);
1508 nvlist_free(zplprops);
1509 nvlist_free(config);
1510 nvlist_free(props);
1511
1512 return (error);
1513 }
1514
1515 static int
1516 zfs_ioc_pool_destroy(zfs_cmd_t *zc)
1517 {
1518 int error;
1519 zfs_log_history(zc);
1520 error = spa_destroy(zc->zc_name);
1521 if (error == 0)
1522 zvol_remove_minors(zc->zc_name);
1523 return (error);
1524 }
1525
1526 static int
1527 zfs_ioc_pool_import(zfs_cmd_t *zc)
1528 {
1529 nvlist_t *config, *props = NULL;
1530 uint64_t guid;
1531 int error;
1532
1533 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1534 zc->zc_iflags, &config)) != 0)
1535 return (error);
1536
1537 if (zc->zc_nvlist_src_size != 0 && (error =
1538 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1539 zc->zc_iflags, &props))) {
1540 nvlist_free(config);
1541 return (error);
1542 }
1543
1544 if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, &guid) != 0 ||
1545 guid != zc->zc_guid)
1546 error = SET_ERROR(EINVAL);
1547 else
1548 error = spa_import(zc->zc_name, config, props, zc->zc_cookie);
1549
1550 if (zc->zc_nvlist_dst != 0) {
1551 int err;
1552
1553 if ((err = put_nvlist(zc, config)) != 0)
1554 error = err;
1555 }
1556
1557 nvlist_free(config);
1558
1559 if (props)
1560 nvlist_free(props);
1561
1562 return (error);
1563 }
1564
1565 static int
1566 zfs_ioc_pool_export(zfs_cmd_t *zc)
1567 {
1568 int error;
1569 boolean_t force = (boolean_t)zc->zc_cookie;
1570 boolean_t hardforce = (boolean_t)zc->zc_guid;
1571
1572 zfs_log_history(zc);
1573 error = spa_export(zc->zc_name, NULL, force, hardforce);
1574 if (error == 0)
1575 zvol_remove_minors(zc->zc_name);
1576 return (error);
1577 }
1578
1579 static int
1580 zfs_ioc_pool_configs(zfs_cmd_t *zc)
1581 {
1582 nvlist_t *configs;
1583 int error;
1584
1585 if ((configs = spa_all_configs(&zc->zc_cookie)) == NULL)
1586 return (SET_ERROR(EEXIST));
1587
1588 error = put_nvlist(zc, configs);
1589
1590 nvlist_free(configs);
1591
1592 return (error);
1593 }
1594
1595 /*
1596 * inputs:
1597 * zc_name name of the pool
1598 *
1599 * outputs:
1600 * zc_cookie real errno
1601 * zc_nvlist_dst config nvlist
1602 * zc_nvlist_dst_size size of config nvlist
1603 */
1604 static int
1605 zfs_ioc_pool_stats(zfs_cmd_t *zc)
1606 {
1607 nvlist_t *config;
1608 int error;
1609 int ret = 0;
1610
1611 error = spa_get_stats(zc->zc_name, &config, zc->zc_value,
1612 sizeof (zc->zc_value));
1613
1614 if (config != NULL) {
1615 ret = put_nvlist(zc, config);
1616 nvlist_free(config);
1617
1618 /*
1619 * The config may be present even if 'error' is non-zero.
1620 * In this case we return success, and preserve the real errno
1621 * in 'zc_cookie'.
1622 */
1623 zc->zc_cookie = error;
1624 } else {
1625 ret = error;
1626 }
1627
1628 return (ret);
1629 }
1630
1631 /*
1632 * Try to import the given pool, returning pool stats as appropriate so that
1633 * user land knows which devices are available and overall pool health.
1634 */
1635 static int
1636 zfs_ioc_pool_tryimport(zfs_cmd_t *zc)
1637 {
1638 nvlist_t *tryconfig, *config;
1639 int error;
1640
1641 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1642 zc->zc_iflags, &tryconfig)) != 0)
1643 return (error);
1644
1645 config = spa_tryimport(tryconfig);
1646
1647 nvlist_free(tryconfig);
1648
1649 if (config == NULL)
1650 return (SET_ERROR(EINVAL));
1651
1652 error = put_nvlist(zc, config);
1653 nvlist_free(config);
1654
1655 return (error);
1656 }
1657
1658 /*
1659 * inputs:
1660 * zc_name name of the pool
1661 * zc_cookie scan func (pool_scan_func_t)
1662 */
1663 static int
1664 zfs_ioc_pool_scan(zfs_cmd_t *zc)
1665 {
1666 spa_t *spa;
1667 int error;
1668
1669 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1670 return (error);
1671
1672 if (zc->zc_cookie == POOL_SCAN_NONE)
1673 error = spa_scan_stop(spa);
1674 else
1675 error = spa_scan(spa, zc->zc_cookie);
1676
1677 spa_close(spa, FTAG);
1678
1679 return (error);
1680 }
1681
1682 static int
1683 zfs_ioc_pool_freeze(zfs_cmd_t *zc)
1684 {
1685 spa_t *spa;
1686 int error;
1687
1688 error = spa_open(zc->zc_name, &spa, FTAG);
1689 if (error == 0) {
1690 spa_freeze(spa);
1691 spa_close(spa, FTAG);
1692 }
1693 return (error);
1694 }
1695
1696 static int
1697 zfs_ioc_pool_upgrade(zfs_cmd_t *zc)
1698 {
1699 spa_t *spa;
1700 int error;
1701
1702 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1703 return (error);
1704
1705 if (zc->zc_cookie < spa_version(spa) ||
1706 !SPA_VERSION_IS_SUPPORTED(zc->zc_cookie)) {
1707 spa_close(spa, FTAG);
1708 return (SET_ERROR(EINVAL));
1709 }
1710
1711 spa_upgrade(spa, zc->zc_cookie);
1712 spa_close(spa, FTAG);
1713
1714 return (error);
1715 }
1716
1717 static int
1718 zfs_ioc_pool_get_history(zfs_cmd_t *zc)
1719 {
1720 spa_t *spa;
1721 char *hist_buf;
1722 uint64_t size;
1723 int error;
1724
1725 if ((size = zc->zc_history_len) == 0)
1726 return (SET_ERROR(EINVAL));
1727
1728 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1729 return (error);
1730
1731 if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) {
1732 spa_close(spa, FTAG);
1733 return (SET_ERROR(ENOTSUP));
1734 }
1735
1736 hist_buf = kmem_alloc(size, KM_SLEEP);
1737 if ((error = spa_history_get(spa, &zc->zc_history_offset,
1738 &zc->zc_history_len, hist_buf)) == 0) {
1739 error = ddi_copyout(hist_buf,
1740 (void *)(uintptr_t)zc->zc_history,
1741 zc->zc_history_len, zc->zc_iflags);
1742 }
1743
1744 spa_close(spa, FTAG);
1745 kmem_free(hist_buf, size);
1746 return (error);
1747 }
1748
1749 static int
1750 zfs_ioc_pool_reguid(zfs_cmd_t *zc)
1751 {
1752 spa_t *spa;
1753 int error;
1754
1755 error = spa_open(zc->zc_name, &spa, FTAG);
1756 if (error == 0) {
1757 error = spa_change_guid(spa);
1758 spa_close(spa, FTAG);
1759 }
1760 return (error);
1761 }
1762
1763 static int
1764 zfs_ioc_dsobj_to_dsname(zfs_cmd_t *zc)
1765 {
1766 return (dsl_dsobj_to_dsname(zc->zc_name, zc->zc_obj, zc->zc_value));
1767 }
1768
1769 /*
1770 * inputs:
1771 * zc_name name of filesystem
1772 * zc_obj object to find
1773 *
1774 * outputs:
1775 * zc_value name of object
1776 */
1777 static int
1778 zfs_ioc_obj_to_path(zfs_cmd_t *zc)
1779 {
1780 objset_t *os;
1781 int error;
1782
1783 /* XXX reading from objset not owned */
1784 if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os)) != 0)
1785 return (error);
1786 if (dmu_objset_type(os) != DMU_OST_ZFS) {
1787 dmu_objset_rele(os, FTAG);
1788 return (SET_ERROR(EINVAL));
1789 }
1790 error = zfs_obj_to_path(os, zc->zc_obj, zc->zc_value,
1791 sizeof (zc->zc_value));
1792 dmu_objset_rele(os, FTAG);
1793
1794 return (error);
1795 }
1796
1797 /*
1798 * inputs:
1799 * zc_name name of filesystem
1800 * zc_obj object to find
1801 *
1802 * outputs:
1803 * zc_stat stats on object
1804 * zc_value path to object
1805 */
1806 static int
1807 zfs_ioc_obj_to_stats(zfs_cmd_t *zc)
1808 {
1809 objset_t *os;
1810 int error;
1811
1812 /* XXX reading from objset not owned */
1813 if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os)) != 0)
1814 return (error);
1815 if (dmu_objset_type(os) != DMU_OST_ZFS) {
1816 dmu_objset_rele(os, FTAG);
1817 return (SET_ERROR(EINVAL));
1818 }
1819 error = zfs_obj_to_stats(os, zc->zc_obj, &zc->zc_stat, zc->zc_value,
1820 sizeof (zc->zc_value));
1821 dmu_objset_rele(os, FTAG);
1822
1823 return (error);
1824 }
1825
1826 static int
1827 zfs_ioc_vdev_add(zfs_cmd_t *zc)
1828 {
1829 spa_t *spa;
1830 int error;
1831 nvlist_t *config, **l2cache, **spares;
1832 uint_t nl2cache = 0, nspares = 0;
1833
1834 error = spa_open(zc->zc_name, &spa, FTAG);
1835 if (error != 0)
1836 return (error);
1837
1838 error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1839 zc->zc_iflags, &config);
1840 (void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_L2CACHE,
1841 &l2cache, &nl2cache);
1842
1843 (void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_SPARES,
1844 &spares, &nspares);
1845
1846 /*
1847 * A root pool with concatenated devices is not supported.
1848 * Thus, can not add a device to a root pool.
1849 *
1850 * Intent log device can not be added to a rootpool because
1851 * during mountroot, zil is replayed, a seperated log device
1852 * can not be accessed during the mountroot time.
1853 *
1854 * l2cache and spare devices are ok to be added to a rootpool.
1855 */
1856 if (spa_bootfs(spa) != 0 && nl2cache == 0 && nspares == 0) {
1857 nvlist_free(config);
1858 spa_close(spa, FTAG);
1859 return (SET_ERROR(EDOM));
1860 }
1861
1862 if (error == 0) {
1863 error = spa_vdev_add(spa, config);
1864 nvlist_free(config);
1865 }
1866 spa_close(spa, FTAG);
1867 return (error);
1868 }
1869
1870 /*
1871 * inputs:
1872 * zc_name name of the pool
1873 * zc_nvlist_conf nvlist of devices to remove
1874 * zc_cookie to stop the remove?
1875 */
1876 static int
1877 zfs_ioc_vdev_remove(zfs_cmd_t *zc)
1878 {
1879 spa_t *spa;
1880 int error;
1881
1882 error = spa_open(zc->zc_name, &spa, FTAG);
1883 if (error != 0)
1884 return (error);
1885 error = spa_vdev_remove(spa, zc->zc_guid, B_FALSE);
1886 spa_close(spa, FTAG);
1887 return (error);
1888 }
1889
1890 static int
1891 zfs_ioc_vdev_set_state(zfs_cmd_t *zc)
1892 {
1893 spa_t *spa;
1894 int error;
1895 vdev_state_t newstate = VDEV_STATE_UNKNOWN;
1896
1897 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1898 return (error);
1899 switch (zc->zc_cookie) {
1900 case VDEV_STATE_ONLINE:
1901 error = vdev_online(spa, zc->zc_guid, zc->zc_obj, &newstate);
1902 break;
1903
1904 case VDEV_STATE_OFFLINE:
1905 error = vdev_offline(spa, zc->zc_guid, zc->zc_obj);
1906 break;
1907
1908 case VDEV_STATE_FAULTED:
1909 if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED &&
1910 zc->zc_obj != VDEV_AUX_EXTERNAL)
1911 zc->zc_obj = VDEV_AUX_ERR_EXCEEDED;
1912
1913 error = vdev_fault(spa, zc->zc_guid, zc->zc_obj);
1914 break;
1915
1916 case VDEV_STATE_DEGRADED:
1917 if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED &&
1918 zc->zc_obj != VDEV_AUX_EXTERNAL)
1919 zc->zc_obj = VDEV_AUX_ERR_EXCEEDED;
1920
1921 error = vdev_degrade(spa, zc->zc_guid, zc->zc_obj);
1922 break;
1923
1924 default:
1925 error = SET_ERROR(EINVAL);
1926 }
1927 zc->zc_cookie = newstate;
1928 spa_close(spa, FTAG);
1929 return (error);
1930 }
1931
1932 static int
1933 zfs_ioc_vdev_attach(zfs_cmd_t *zc)
1934 {
1935 spa_t *spa;
1936 int replacing = zc->zc_cookie;
1937 nvlist_t *config;
1938 int error;
1939
1940 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1941 return (error);
1942
1943 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1944 zc->zc_iflags, &config)) == 0) {
1945 error = spa_vdev_attach(spa, zc->zc_guid, config, replacing);
1946 nvlist_free(config);
1947 }
1948
1949 spa_close(spa, FTAG);
1950 return (error);
1951 }
1952
1953 static int
1954 zfs_ioc_vdev_detach(zfs_cmd_t *zc)
1955 {
1956 spa_t *spa;
1957 int error;
1958
1959 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1960 return (error);
1961
1962 error = spa_vdev_detach(spa, zc->zc_guid, 0, B_FALSE);
1963
1964 spa_close(spa, FTAG);
1965 return (error);
1966 }
1967
1968 static int
1969 zfs_ioc_vdev_split(zfs_cmd_t *zc)
1970 {
1971 spa_t *spa;
1972 nvlist_t *config, *props = NULL;
1973 int error;
1974 boolean_t exp = !!(zc->zc_cookie & ZPOOL_EXPORT_AFTER_SPLIT);
1975
1976 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1977 return (error);
1978
1979 if (error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1980 zc->zc_iflags, &config)) {
1981 spa_close(spa, FTAG);
1982 return (error);
1983 }
1984
1985 if (zc->zc_nvlist_src_size != 0 && (error =
1986 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1987 zc->zc_iflags, &props))) {
1988 spa_close(spa, FTAG);
1989 nvlist_free(config);
1990 return (error);
1991 }
1992
1993 error = spa_vdev_split_mirror(spa, zc->zc_string, config, props, exp);
1994
1995 spa_close(spa, FTAG);
1996
1997 nvlist_free(config);
1998 nvlist_free(props);
1999
2000 return (error);
2001 }
2002
2003 static int
2004 zfs_ioc_vdev_setpath(zfs_cmd_t *zc)
2005 {
2006 spa_t *spa;
2007 char *path = zc->zc_value;
2008 uint64_t guid = zc->zc_guid;
2009 int error;
2010
2011 error = spa_open(zc->zc_name, &spa, FTAG);
2012 if (error != 0)
2013 return (error);
2014
2015 error = spa_vdev_setpath(spa, guid, path);
2016 spa_close(spa, FTAG);
2017 return (error);
2018 }
2019
2020 static int
2021 zfs_ioc_vdev_setfru(zfs_cmd_t *zc)
2022 {
2023 spa_t *spa;
2024 char *fru = zc->zc_value;
2025 uint64_t guid = zc->zc_guid;
2026 int error;
2027
2028 error = spa_open(zc->zc_name, &spa, FTAG);
2029 if (error != 0)
2030 return (error);
2031
2032 error = spa_vdev_setfru(spa, guid, fru);
2033 spa_close(spa, FTAG);
2034 return (error);
2035 }
2036
2037 static int
2038 zfs_ioc_objset_stats_impl(zfs_cmd_t *zc, objset_t *os)
2039 {
2040 int error = 0;
2041 nvlist_t *nv;
2042
2043 dmu_objset_fast_stat(os, &zc->zc_objset_stats);
2044
2045 if (zc->zc_nvlist_dst != 0 &&
2046 (error = dsl_prop_get_all(os, &nv)) == 0) {
2047 dmu_objset_stats(os, nv);
2048 /*
2049 * NB: zvol_get_stats() will read the objset contents,
2050 * which we aren't supposed to do with a
2051 * DS_MODE_USER hold, because it could be
2052 * inconsistent. So this is a bit of a workaround...
2053 * XXX reading with out owning
2054 */
2055 if (!zc->zc_objset_stats.dds_inconsistent &&
2056 dmu_objset_type(os) == DMU_OST_ZVOL) {
2057 error = zvol_get_stats(os, nv);
2058 if (error == EIO)
2059 return (error);
2060 VERIFY0(error);
2061 }
2062 error = put_nvlist(zc, nv);
2063 nvlist_free(nv);
2064 }
2065
2066 return (error);
2067 }
2068
2069 /*
2070 * inputs:
2071 * zc_name name of filesystem
2072 * zc_nvlist_dst_size size of buffer for property nvlist
2073 *
2074 * outputs:
2075 * zc_objset_stats stats
2076 * zc_nvlist_dst property nvlist
2077 * zc_nvlist_dst_size size of property nvlist
2078 */
2079 static int
2080 zfs_ioc_objset_stats(zfs_cmd_t *zc)
2081 {
2082 objset_t *os;
2083 int error;
2084
2085 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
2086 if (error == 0) {
2087 error = zfs_ioc_objset_stats_impl(zc, os);
2088 dmu_objset_rele(os, FTAG);
2089 }
2090
2091 return (error);
2092 }
2093
2094 /*
2095 * inputs:
2096 * zc_name name of filesystem
2097 * zc_nvlist_dst_size size of buffer for property nvlist
2098 *
2099 * outputs:
2100 * zc_nvlist_dst received property nvlist
2101 * zc_nvlist_dst_size size of received property nvlist
2102 *
2103 * Gets received properties (distinct from local properties on or after
2104 * SPA_VERSION_RECVD_PROPS) for callers who want to differentiate received from
2105 * local property values.
2106 */
2107 static int
2108 zfs_ioc_objset_recvd_props(zfs_cmd_t *zc)
2109 {
2110 int error = 0;
2111 nvlist_t *nv;
2112
2113 /*
2114 * Without this check, we would return local property values if the
2115 * caller has not already received properties on or after
2116 * SPA_VERSION_RECVD_PROPS.
2117 */
2118 if (!dsl_prop_get_hasrecvd(zc->zc_name))
2119 return (SET_ERROR(ENOTSUP));
2120
2121 if (zc->zc_nvlist_dst != 0 &&
2122 (error = dsl_prop_get_received(zc->zc_name, &nv)) == 0) {
2123 error = put_nvlist(zc, nv);
2124 nvlist_free(nv);
2125 }
2126
2127 return (error);
2128 }
2129
2130 static int
2131 nvl_add_zplprop(objset_t *os, nvlist_t *props, zfs_prop_t prop)
2132 {
2133 uint64_t value;
2134 int error;
2135
2136 /*
2137 * zfs_get_zplprop() will either find a value or give us
2138 * the default value (if there is one).
2139 */
2140 if ((error = zfs_get_zplprop(os, prop, &value)) != 0)
2141 return (error);
2142 VERIFY(nvlist_add_uint64(props, zfs_prop_to_name(prop), value) == 0);
2143 return (0);
2144 }
2145
2146 /*
2147 * inputs:
2148 * zc_name name of filesystem
2149 * zc_nvlist_dst_size size of buffer for zpl property nvlist
2150 *
2151 * outputs:
2152 * zc_nvlist_dst zpl property nvlist
2153 * zc_nvlist_dst_size size of zpl property nvlist
2154 */
2155 static int
2156 zfs_ioc_objset_zplprops(zfs_cmd_t *zc)
2157 {
2158 objset_t *os;
2159 int err;
2160
2161 /* XXX reading without owning */
2162 if (err = dmu_objset_hold(zc->zc_name, FTAG, &os))
2163 return (err);
2164
2165 dmu_objset_fast_stat(os, &zc->zc_objset_stats);
2166
2167 /*
2168 * NB: nvl_add_zplprop() will read the objset contents,
2169 * which we aren't supposed to do with a DS_MODE_USER
2170 * hold, because it could be inconsistent.
2171 */
2172 if (zc->zc_nvlist_dst != NULL &&
2173 !zc->zc_objset_stats.dds_inconsistent &&
2174 dmu_objset_type(os) == DMU_OST_ZFS) {
2175 nvlist_t *nv;
2176
2177 VERIFY(nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2178 if ((err = nvl_add_zplprop(os, nv, ZFS_PROP_VERSION)) == 0 &&
2179 (err = nvl_add_zplprop(os, nv, ZFS_PROP_NORMALIZE)) == 0 &&
2180 (err = nvl_add_zplprop(os, nv, ZFS_PROP_UTF8ONLY)) == 0 &&
2181 (err = nvl_add_zplprop(os, nv, ZFS_PROP_CASE)) == 0)
2182 err = put_nvlist(zc, nv);
2183 nvlist_free(nv);
2184 } else {
2185 err = SET_ERROR(ENOENT);
2186 }
2187 dmu_objset_rele(os, FTAG);
2188 return (err);
2189 }
2190
2191 static boolean_t
2192 dataset_name_hidden(const char *name)
2193 {
2194 /*
2195 * Skip over datasets that are not visible in this zone,
2196 * internal datasets (which have a $ in their name), and
2197 * temporary datasets (which have a % in their name).
2198 */
2199 if (strchr(name, '$') != NULL)
2200 return (B_TRUE);
2201 if (strchr(name, '%') != NULL)
2202 return (B_TRUE);
2203 if (!INGLOBALZONE(curproc) && !zone_dataset_visible(name, NULL))
2204 return (B_TRUE);
2205 return (B_FALSE);
2206 }
2207
2208 /*
2209 * inputs:
2210 * zc_name name of filesystem
2211 * zc_cookie zap cursor
2212 * zc_nvlist_dst_size size of buffer for property nvlist
2213 *
2214 * outputs:
2215 * zc_name name of next filesystem
2216 * zc_cookie zap cursor
2217 * zc_objset_stats stats
2218 * zc_nvlist_dst property nvlist
2219 * zc_nvlist_dst_size size of property nvlist
2220 */
2221 static int
2222 zfs_ioc_dataset_list_next(zfs_cmd_t *zc)
2223 {
2224 objset_t *os;
2225 int error;
2226 char *p;
2227 size_t orig_len = strlen(zc->zc_name);
2228
2229 top:
2230 if (error = dmu_objset_hold(zc->zc_name, FTAG, &os)) {
2231 if (error == ENOENT)
2232 error = SET_ERROR(ESRCH);
2233 return (error);
2234 }
2235
2236 p = strrchr(zc->zc_name, '/');
2237 if (p == NULL || p[1] != '\0')
2238 (void) strlcat(zc->zc_name, "/", sizeof (zc->zc_name));
2239 p = zc->zc_name + strlen(zc->zc_name);
2240
2241 do {
2242 error = dmu_dir_list_next(os,
2243 sizeof (zc->zc_name) - (p - zc->zc_name), p,
2244 NULL, &zc->zc_cookie);
2245 if (error == ENOENT)
2246 error = SET_ERROR(ESRCH);
2247 } while (error == 0 && dataset_name_hidden(zc->zc_name));
2248 dmu_objset_rele(os, FTAG);
2249
2250 /*
2251 * If it's an internal dataset (ie. with a '$' in its name),
2252 * don't try to get stats for it, otherwise we'll return ENOENT.
2253 */
2254 if (error == 0 && strchr(zc->zc_name, '$') == NULL) {
2255 error = zfs_ioc_objset_stats(zc); /* fill in the stats */
2256 if (error == ENOENT) {
2257 /* We lost a race with destroy, get the next one. */
2258 zc->zc_name[orig_len] = '\0';
2259 goto top;
2260 }
2261 }
2262 return (error);
2263 }
2264
2265 /*
2266 * inputs:
2267 * zc_name name of filesystem
2268 * zc_cookie zap cursor
2269 * zc_nvlist_dst_size size of buffer for property nvlist
2270 *
2271 * outputs:
2272 * zc_name name of next snapshot
2273 * zc_objset_stats stats
2274 * zc_nvlist_dst property nvlist
2275 * zc_nvlist_dst_size size of property nvlist
2276 */
2277 static int
2278 zfs_ioc_snapshot_list_next(zfs_cmd_t *zc)
2279 {
2280 objset_t *os;
2281 int error;
2282
2283 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
2284 if (error != 0) {
2285 return (error == ENOENT ? ESRCH : error);
2286 }
2287
2288 /*
2289 * A dataset name of maximum length cannot have any snapshots,
2290 * so exit immediately.
2291 */
2292 if (strlcat(zc->zc_name, "@", sizeof (zc->zc_name)) >= MAXNAMELEN) {
2293 dmu_objset_rele(os, FTAG);
2294 return (SET_ERROR(ESRCH));
2295 }
2296
2297 error = dmu_snapshot_list_next(os,
2298 sizeof (zc->zc_name) - strlen(zc->zc_name),
2299 zc->zc_name + strlen(zc->zc_name), &zc->zc_obj, &zc->zc_cookie,
2300 NULL);
2301
2302 if (error == 0) {
2303 dsl_dataset_t *ds;
2304 dsl_pool_t *dp = os->os_dsl_dataset->ds_dir->dd_pool;
2305
2306 error = dsl_dataset_hold_obj(dp, zc->zc_obj, FTAG, &ds);
2307 if (error == 0) {
2308 objset_t *ossnap;
2309
2310 error = dmu_objset_from_ds(ds, &ossnap);
2311 if (error == 0)
2312 error = zfs_ioc_objset_stats_impl(zc, ossnap);
2313 dsl_dataset_rele(ds, FTAG);
2314 }
2315 } else if (error == ENOENT) {
2316 error = SET_ERROR(ESRCH);
2317 }
2318
2319 dmu_objset_rele(os, FTAG);
2320 /* if we failed, undo the @ that we tacked on to zc_name */
2321 if (error != 0)
2322 *strchr(zc->zc_name, '@') = '\0';
2323 return (error);
2324 }
2325
2326 static int
2327 zfs_prop_set_userquota(const char *dsname, nvpair_t *pair)
2328 {
2329 const char *propname = nvpair_name(pair);
2330 uint64_t *valary;
2331 unsigned int vallen;
2332 const char *domain;
2333 char *dash;
2334 zfs_userquota_prop_t type;
2335 uint64_t rid;
2336 uint64_t quota;
2337 zfsvfs_t *zfsvfs;
2338 int err;
2339
2340 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2341 nvlist_t *attrs;
2342 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
2343 if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2344 &pair) != 0)
2345 return (SET_ERROR(EINVAL));
2346 }
2347
2348 /*
2349 * A correctly constructed propname is encoded as
2350 * userquota@<rid>-<domain>.
2351 */
2352 if ((dash = strchr(propname, '-')) == NULL ||
2353 nvpair_value_uint64_array(pair, &valary, &vallen) != 0 ||
2354 vallen != 3)
2355 return (SET_ERROR(EINVAL));
2356
2357 domain = dash + 1;
2358 type = valary[0];
2359 rid = valary[1];
2360 quota = valary[2];
2361
2362 err = zfsvfs_hold(dsname, FTAG, &zfsvfs, B_FALSE);
2363 if (err == 0) {
2364 err = zfs_set_userquota(zfsvfs, type, domain, rid, quota);
2365 zfsvfs_rele(zfsvfs, FTAG);
2366 }
2367
2368 return (err);
2369 }
2370
2371 /*
2372 * If the named property is one that has a special function to set its value,
2373 * return 0 on success and a positive error code on failure; otherwise if it is
2374 * not one of the special properties handled by this function, return -1.
2375 *
2376 * XXX: It would be better for callers of the property interface if we handled
2377 * these special cases in dsl_prop.c (in the dsl layer).
2378 */
2379 static int
2380 zfs_prop_set_special(const char *dsname, zprop_source_t source,
2381 nvpair_t *pair)
2382 {
2383 const char *propname = nvpair_name(pair);
2384 zfs_prop_t prop = zfs_name_to_prop(propname);
2385 uint64_t intval;
2386 int err = -1;
2387
2388 if (prop == ZPROP_INVAL) {
2389 if (zfs_prop_userquota(propname))
2390 return (zfs_prop_set_userquota(dsname, pair));
2391 return (-1);
2392 }
2393
2394 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2395 nvlist_t *attrs;
2396 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
2397 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2398 &pair) == 0);
2399 }
2400
2401 if (zfs_prop_get_type(prop) == PROP_TYPE_STRING)
2402 return (-1);
2403
2404 VERIFY(0 == nvpair_value_uint64(pair, &intval));
2405
2406 switch (prop) {
2407 case ZFS_PROP_FSID_GUID:
2408 err = dsl_dataset_set_fsid_guid(dsname, source, intval);
2409 break;
2410 case ZFS_PROP_QUOTA:
2411 err = dsl_dir_set_quota(dsname, source, intval);
2412 break;
2413 case ZFS_PROP_REFQUOTA:
2414 err = dsl_dataset_set_refquota(dsname, source, intval);
2415 break;
2416 case ZFS_PROP_FILESYSTEM_LIMIT:
2417 case ZFS_PROP_SNAPSHOT_LIMIT:
2418 if (intval == UINT64_MAX) {
2419 /* clearing the limit, just do it */
2420 err = 0;
2421 } else {
2422 err = dsl_dir_activate_fs_ss_limit(dsname);
2423 }
2424 /*
2425 * Set err to -1 to force the zfs_set_prop_nvlist code down the
2426 * default path to set the value in the nvlist.
2427 */
2428 if (err == 0)
2429 err = -1;
2430 break;
2431 case ZFS_PROP_RESERVATION:
2432 err = dsl_dir_set_reservation(dsname, source, intval);
2433 break;
2434 case ZFS_PROP_REFRESERVATION:
2435 err = dsl_dataset_set_refreservation(dsname, source, intval);
2436 break;
2437 case ZFS_PROP_VOLSIZE:
2438 err = zvol_set_volsize(dsname, intval);
2439 break;
2440 case ZFS_PROP_VERSION:
2441 {
2442 zfsvfs_t *zfsvfs;
2443
2444 if ((err = zfsvfs_hold(dsname, FTAG, &zfsvfs, B_TRUE)) != 0)
2445 break;
2446
2447 err = zfs_set_version(zfsvfs, intval);
2448 zfsvfs_rele(zfsvfs, FTAG);
2449
2450 if (err == 0 && intval >= ZPL_VERSION_USERSPACE) {
2451 zfs_cmd_t *zc;
2452
2453 zc = kmem_zalloc(sizeof (zfs_cmd_t), KM_SLEEP);
2454 (void) strcpy(zc->zc_name, dsname);
2455 (void) zfs_ioc_userspace_upgrade(zc);
2456 kmem_free(zc, sizeof (zfs_cmd_t));
2457 }
2458 break;
2459 }
2460 default:
2461 err = -1;
2462 }
2463
2464 return (err);
2465 }
2466
2467 /*
2468 * This function is best effort. If it fails to set any of the given properties,
2469 * it continues to set as many as it can and returns the last error
2470 * encountered. If the caller provides a non-NULL errlist, it will be filled in
2471 * with the list of names of all the properties that failed along with the
2472 * corresponding error numbers.
2473 *
2474 * If every property is set successfully, zero is returned and errlist is not
2475 * modified.
2476 */
2477 int
2478 zfs_set_prop_nvlist(const char *dsname, zprop_source_t source, nvlist_t *nvl,
2479 nvlist_t *errlist)
2480 {
2481 nvpair_t *pair;
2482 nvpair_t *propval;
2483 int rv = 0;
2484 uint64_t intval;
2485 char *strval;
2486 nvlist_t *genericnvl = fnvlist_alloc();
2487 nvlist_t *retrynvl = fnvlist_alloc();
2488
2489 retry:
2490 pair = NULL;
2491 while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) {
2492 const char *propname = nvpair_name(pair);
2493 zfs_prop_t prop = zfs_name_to_prop(propname);
2494 int err = 0;
2495
2496 /* decode the property value */
2497 propval = pair;
2498 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2499 nvlist_t *attrs;
2500 attrs = fnvpair_value_nvlist(pair);
2501 if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2502 &propval) != 0)
2503 err = SET_ERROR(EINVAL);
2504 }
2505
2506 /* Validate value type */
2507 if (err == 0 && prop == ZPROP_INVAL) {
2508 if (zfs_prop_user(propname)) {
2509 if (nvpair_type(propval) != DATA_TYPE_STRING)
2510 err = SET_ERROR(EINVAL);
2511 } else if (zfs_prop_userquota(propname)) {
2512 if (nvpair_type(propval) !=
2513 DATA_TYPE_UINT64_ARRAY)
2514 err = SET_ERROR(EINVAL);
2515 } else {
2516 err = SET_ERROR(EINVAL);
2517 }
2518 } else if (err == 0) {
2519 if (nvpair_type(propval) == DATA_TYPE_STRING) {
2520 if (zfs_prop_get_type(prop) != PROP_TYPE_STRING)
2521 err = SET_ERROR(EINVAL);
2522 } else if (nvpair_type(propval) == DATA_TYPE_UINT64) {
2523 const char *unused;
2524
2525 intval = fnvpair_value_uint64(propval);
2526
2527 switch (zfs_prop_get_type(prop)) {
2528 case PROP_TYPE_NUMBER:
2529 break;
2530 case PROP_TYPE_STRING:
2531 err = SET_ERROR(EINVAL);
2532 break;
2533 case PROP_TYPE_INDEX:
2534 if (zfs_prop_index_to_string(prop,
2535 intval, &unused) != 0)
2536 err = SET_ERROR(EINVAL);
2537 break;
2538 default:
2539 cmn_err(CE_PANIC,
2540 "unknown property type");
2541 }
2542 } else {
2543 err = SET_ERROR(EINVAL);
2544 }
2545 }
2546
2547 /* Validate permissions */
2548 if (err == 0)
2549 err = zfs_check_settable(dsname, pair, CRED());
2550
2551 if (err == 0) {
2552 err = zfs_prop_set_special(dsname, source, pair);
2553 if (err == -1) {
2554 /*
2555 * For better performance we build up a list of
2556 * properties to set in a single transaction.
2557 */
2558 err = nvlist_add_nvpair(genericnvl, pair);
2559 } else if (err != 0 && nvl != retrynvl) {
2560 /*
2561 * This may be a spurious error caused by
2562 * receiving quota and reservation out of order.
2563 * Try again in a second pass.
2564 */
2565 err = nvlist_add_nvpair(retrynvl, pair);
2566 }
2567 }
2568
2569 if (err != 0) {
2570 if (errlist != NULL)
2571 fnvlist_add_int32(errlist, propname, err);
2572 rv = err;
2573 }
2574 }
2575
2576 if (nvl != retrynvl && !nvlist_empty(retrynvl)) {
2577 nvl = retrynvl;
2578 goto retry;
2579 }
2580
2581 if (!nvlist_empty(genericnvl) &&
2582 dsl_props_set(dsname, source, genericnvl) != 0) {
2583 /*
2584 * If this fails, we still want to set as many properties as we
2585 * can, so try setting them individually.
2586 */
2587 pair = NULL;
2588 while ((pair = nvlist_next_nvpair(genericnvl, pair)) != NULL) {
2589 const char *propname = nvpair_name(pair);
2590 int err = 0;
2591
2592 propval = pair;
2593 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2594 nvlist_t *attrs;
2595 attrs = fnvpair_value_nvlist(pair);
2596 propval = fnvlist_lookup_nvpair(attrs,
2597 ZPROP_VALUE);
2598 }
2599
2600 if (nvpair_type(propval) == DATA_TYPE_STRING) {
2601 strval = fnvpair_value_string(propval);
2602 err = dsl_prop_set_string(dsname, propname,
2603 source, strval);
2604 } else {
2605 intval = fnvpair_value_uint64(propval);
2606 err = dsl_prop_set_int(dsname, propname, source,
2607 intval);
2608 }
2609
2610 if (err != 0) {
2611 if (errlist != NULL) {
2612 fnvlist_add_int32(errlist, propname,
2613 err);
2614 }
2615 rv = err;
2616 }
2617 }
2618 }
2619 nvlist_free(genericnvl);
2620 nvlist_free(retrynvl);
2621
2622 return (rv);
2623 }
2624
2625 /*
2626 * Check that all the properties are valid user properties.
2627 */
2628 static int
2629 zfs_check_userprops(const char *fsname, nvlist_t *nvl)
2630 {
2631 nvpair_t *pair = NULL;
2632 int error = 0;
2633
2634 while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) {
2635 const char *propname = nvpair_name(pair);
2636
2637 if (!zfs_prop_user(propname) ||
2638 nvpair_type(pair) != DATA_TYPE_STRING)
2639 return (SET_ERROR(EINVAL));
2640
2641 if (error = zfs_secpolicy_write_perms(fsname,
2642 ZFS_DELEG_PERM_USERPROP, CRED()))
2643 return (error);
2644
2645 if (strlen(propname) >= ZAP_MAXNAMELEN)
2646 return (SET_ERROR(ENAMETOOLONG));
2647
2648 if (strlen(fnvpair_value_string(pair)) >= ZAP_MAXVALUELEN)
2649 return (E2BIG);
2650 }
2651 return (0);
2652 }
2653
2654 static void
2655 props_skip(nvlist_t *props, nvlist_t *skipped, nvlist_t **newprops)
2656 {
2657 nvpair_t *pair;
2658
2659 VERIFY(nvlist_alloc(newprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2660
2661 pair = NULL;
2662 while ((pair = nvlist_next_nvpair(props, pair)) != NULL) {
2663 if (nvlist_exists(skipped, nvpair_name(pair)))
2664 continue;
2665
2666 VERIFY(nvlist_add_nvpair(*newprops, pair) == 0);
2667 }
2668 }
2669
2670 static int
2671 clear_received_props(const char *dsname, nvlist_t *props,
2672 nvlist_t *skipped)
2673 {
2674 int err = 0;
2675 nvlist_t *cleared_props = NULL;
2676 props_skip(props, skipped, &cleared_props);
2677 if (!nvlist_empty(cleared_props)) {
2678 /*
2679 * Acts on local properties until the dataset has received
2680 * properties at least once on or after SPA_VERSION_RECVD_PROPS.
2681 */
2682 zprop_source_t flags = (ZPROP_SRC_NONE |
2683 (dsl_prop_get_hasrecvd(dsname) ? ZPROP_SRC_RECEIVED : 0));
2684 err = zfs_set_prop_nvlist(dsname, flags, cleared_props, NULL);
2685 }
2686 nvlist_free(cleared_props);
2687 return (err);
2688 }
2689
2690 /*
2691 * inputs:
2692 * zc_name name of filesystem
2693 * zc_value name of property to set
2694 * zc_nvlist_src{_size} nvlist of properties to apply
2695 * zc_cookie received properties flag
2696 *
2697 * outputs:
2698 * zc_nvlist_dst{_size} error for each unapplied received property
2699 */
2700 static int
2701 zfs_ioc_set_prop(zfs_cmd_t *zc)
2702 {
2703 nvlist_t *nvl;
2704 boolean_t received = zc->zc_cookie;
2705 zprop_source_t source = (received ? ZPROP_SRC_RECEIVED :
2706 ZPROP_SRC_LOCAL);
2707 nvlist_t *errors;
2708 int error;
2709
2710 if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2711 zc->zc_iflags, &nvl)) != 0)
2712 return (error);
2713
2714 if (received) {
2715 nvlist_t *origprops;
2716
2717 if (dsl_prop_get_received(zc->zc_name, &origprops) == 0) {
2718 (void) clear_received_props(zc->zc_name,
2719 origprops, nvl);
2720 nvlist_free(origprops);
2721 }
2722
2723 error = dsl_prop_set_hasrecvd(zc->zc_name);
2724 }
2725
2726 errors = fnvlist_alloc();
2727 if (error == 0)
2728 error = zfs_set_prop_nvlist(zc->zc_name, source, nvl, errors);
2729
2730 if (zc->zc_nvlist_dst != NULL && errors != NULL) {
2731 (void) put_nvlist(zc, errors);
2732 }
2733
2734 nvlist_free(errors);
2735 nvlist_free(nvl);
2736 return (error);
2737 }
2738
2739 /*
2740 * inputs:
2741 * zc_name name of filesystem
2742 * zc_value name of property to inherit
2743 * zc_cookie revert to received value if TRUE
2744 *
2745 * outputs: none
2746 */
2747 static int
2748 zfs_ioc_inherit_prop(zfs_cmd_t *zc)
2749 {
2750 const char *propname = zc->zc_value;
2751 zfs_prop_t prop = zfs_name_to_prop(propname);
2752 boolean_t received = zc->zc_cookie;
2753 zprop_source_t source = (received
2754 ? ZPROP_SRC_NONE /* revert to received value, if any */
2755 : ZPROP_SRC_INHERITED); /* explicitly inherit */
2756
2757 if (received) {
2758 nvlist_t *dummy;
2759 nvpair_t *pair;
2760 zprop_type_t type;
2761 int err;
2762
2763 /*
2764 * zfs_prop_set_special() expects properties in the form of an
2765 * nvpair with type info.
2766 */
2767 if (prop == ZPROP_INVAL) {
2768 if (!zfs_prop_user(propname))
2769 return (SET_ERROR(EINVAL));
2770
2771 type = PROP_TYPE_STRING;
2772 } else if (prop == ZFS_PROP_VOLSIZE ||
2773 prop == ZFS_PROP_VERSION) {
2774 return (SET_ERROR(EINVAL));
2775 } else {
2776 type = zfs_prop_get_type(prop);
2777 }
2778
2779 VERIFY(nvlist_alloc(&dummy, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2780
2781 switch (type) {
2782 case PROP_TYPE_STRING:
2783 VERIFY(0 == nvlist_add_string(dummy, propname, ""));
2784 break;
2785 case PROP_TYPE_NUMBER:
2786 case PROP_TYPE_INDEX:
2787 VERIFY(0 == nvlist_add_uint64(dummy, propname, 0));
2788 break;
2789 default:
2790 nvlist_free(dummy);
2791 return (SET_ERROR(EINVAL));
2792 }
2793
2794 pair = nvlist_next_nvpair(dummy, NULL);
2795 err = zfs_prop_set_special(zc->zc_name, source, pair);
2796 nvlist_free(dummy);
2797 if (err != -1)
2798 return (err); /* special property already handled */
2799 } else {
2800 /*
2801 * Only check this in the non-received case. We want to allow
2802 * 'inherit -S' to revert non-inheritable properties like quota
2803 * and reservation to the received or default values even though
2804 * they are not considered inheritable.
2805 */
2806 if (prop != ZPROP_INVAL && !zfs_prop_inheritable(prop))
2807 return (SET_ERROR(EINVAL));
2808 }
2809
2810 /* property name has been validated by zfs_secpolicy_inherit_prop() */
2811 return (dsl_prop_inherit(zc->zc_name, zc->zc_value, source));
2812 }
2813
2814 static int
2815 zfs_ioc_pool_set_props(zfs_cmd_t *zc)
2816 {
2817 nvlist_t *props;
2818 spa_t *spa;
2819 int error;
2820 nvpair_t *pair;
2821
2822 if (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2823 zc->zc_iflags, &props))
2824 return (error);
2825
2826 /*
2827 * If the only property is the configfile, then just do a spa_lookup()
2828 * to handle the faulted case.
2829 */
2830 pair = nvlist_next_nvpair(props, NULL);
2831 if (pair != NULL && strcmp(nvpair_name(pair),
2832 zpool_prop_to_name(ZPOOL_PROP_CACHEFILE)) == 0 &&
2833 nvlist_next_nvpair(props, pair) == NULL) {
2834 mutex_enter(&spa_namespace_lock);
2835 if ((spa = spa_lookup(zc->zc_name)) != NULL) {
2836 spa_configfile_set(spa, props, B_FALSE);
2837 spa_config_sync(spa, B_FALSE, B_TRUE);
2838 }
2839 mutex_exit(&spa_namespace_lock);
2840 if (spa != NULL) {
2841 nvlist_free(props);
2842 return (0);
2843 }
2844 }
2845
2846 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) {
2847 nvlist_free(props);
2848 return (error);
2849 }
2850
2851 error = spa_prop_set(spa, props);
2852
2853 nvlist_free(props);
2854 spa_close(spa, FTAG);
2855
2856 return (error);
2857 }
2858
2859 static int
2860 zfs_ioc_pool_get_props(zfs_cmd_t *zc)
2861 {
2862 spa_t *spa;
2863 int error;
2864 nvlist_t *nvp = NULL;
2865
2866 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) {
2867 /*
2868 * If the pool is faulted, there may be properties we can still
2869 * get (such as altroot and cachefile), so attempt to get them
2870 * anyway.
2871 */
2872 mutex_enter(&spa_namespace_lock);
2873 if ((spa = spa_lookup(zc->zc_name)) != NULL)
2874 error = spa_prop_get(spa, &nvp);
2875 mutex_exit(&spa_namespace_lock);
2876 } else {
2877 error = spa_prop_get(spa, &nvp);
2878 spa_close(spa, FTAG);
2879 }
2880
2881 if (error == 0 && zc->zc_nvlist_dst != NULL)
2882 error = put_nvlist(zc, nvp);
2883 else
2884 error = SET_ERROR(EFAULT);
2885
2886 nvlist_free(nvp);
2887 return (error);
2888 }
2889
2890 /*
2891 * inputs:
2892 * zc_name name of filesystem
2893 * zc_nvlist_src{_size} nvlist of delegated permissions
2894 * zc_perm_action allow/unallow flag
2895 *
2896 * outputs: none
2897 */
2898 static int
2899 zfs_ioc_set_fsacl(zfs_cmd_t *zc)
2900 {
2901 int error;
2902 nvlist_t *fsaclnv = NULL;
2903
2904 if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2905 zc->zc_iflags, &fsaclnv)) != 0)
2906 return (error);
2907
2908 /*
2909 * Verify nvlist is constructed correctly
2910 */
2911 if ((error = zfs_deleg_verify_nvlist(fsaclnv)) != 0) {
2912 nvlist_free(fsaclnv);
2913 return (SET_ERROR(EINVAL));
2914 }
2915
2916 /*
2917 * If we don't have PRIV_SYS_MOUNT, then validate
2918 * that user is allowed to hand out each permission in
2919 * the nvlist(s)
2920 */
2921
2922 error = secpolicy_zfs(CRED());
2923 if (error != 0) {
2924 if (zc->zc_perm_action == B_FALSE) {
2925 error = dsl_deleg_can_allow(zc->zc_name,
2926 fsaclnv, CRED());
2927 } else {
2928 error = dsl_deleg_can_unallow(zc->zc_name,
2929 fsaclnv, CRED());
2930 }
2931 }
2932
2933 if (error == 0)
2934 error = dsl_deleg_set(zc->zc_name, fsaclnv, zc->zc_perm_action);
2935
2936 nvlist_free(fsaclnv);
2937 return (error);
2938 }
2939
2940 /*
2941 * inputs:
2942 * zc_name name of filesystem
2943 *
2944 * outputs:
2945 * zc_nvlist_src{_size} nvlist of delegated permissions
2946 */
2947 static int
2948 zfs_ioc_get_fsacl(zfs_cmd_t *zc)
2949 {
2950 nvlist_t *nvp;
2951 int error;
2952
2953 if ((error = dsl_deleg_get(zc->zc_name, &nvp)) == 0) {
2954 error = put_nvlist(zc, nvp);
2955 nvlist_free(nvp);
2956 }
2957
2958 return (error);
2959 }
2960
2961 /*
2962 * Search the vfs list for a specified resource. Returns a pointer to it
2963 * or NULL if no suitable entry is found. The caller of this routine
2964 * is responsible for releasing the returned vfs pointer.
2965 */
2966 static vfs_t *
2967 zfs_get_vfs(const char *resource)
2968 {
2969 struct vfs *vfsp;
2970 struct vfs *vfs_found = NULL;
2971
2972 vfs_list_read_lock();
2973 vfsp = rootvfs;
2974 do {
2975 if (strcmp(refstr_value(vfsp->vfs_resource), resource) == 0) {
2976 VFS_HOLD(vfsp);
2977 vfs_found = vfsp;
2978 break;
2979 }
2980 vfsp = vfsp->vfs_next;
2981 } while (vfsp != rootvfs);
2982 vfs_list_unlock();
2983 return (vfs_found);
2984 }
2985
2986 /* ARGSUSED */
2987 static void
2988 zfs_create_cb(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx)
2989 {
2990 zfs_creat_t *zct = arg;
2991
2992 zfs_create_fs(os, cr, zct->zct_zplprops, tx);
2993 }
2994
2995 #define ZFS_PROP_UNDEFINED ((uint64_t)-1)
2996
2997 /*
2998 * inputs:
2999 * os parent objset pointer (NULL if root fs)
3000 * fuids_ok fuids allowed in this version of the spa?
3001 * sa_ok SAs allowed in this version of the spa?
3002 * createprops list of properties requested by creator
3003 *
3004 * outputs:
3005 * zplprops values for the zplprops we attach to the master node object
3006 * is_ci true if requested file system will be purely case-insensitive
3007 *
3008 * Determine the settings for utf8only, normalization and
3009 * casesensitivity. Specific values may have been requested by the
3010 * creator and/or we can inherit values from the parent dataset. If
3011 * the file system is of too early a vintage, a creator can not
3012 * request settings for these properties, even if the requested
3013 * setting is the default value. We don't actually want to create dsl
3014 * properties for these, so remove them from the source nvlist after
3015 * processing.
3016 */
3017 static int
3018 zfs_fill_zplprops_impl(objset_t *os, uint64_t zplver,
3019 boolean_t fuids_ok, boolean_t sa_ok, nvlist_t *createprops,
3020 nvlist_t *zplprops, boolean_t *is_ci)
3021 {
3022 uint64_t sense = ZFS_PROP_UNDEFINED;
3023 uint64_t norm = ZFS_PROP_UNDEFINED;
3024 uint64_t u8 = ZFS_PROP_UNDEFINED;
3025
3026 ASSERT(zplprops != NULL);
3027
3028 /*
3029 * Pull out creator prop choices, if any.
3030 */
3031 if (createprops) {
3032 (void) nvlist_lookup_uint64(createprops,
3033 zfs_prop_to_name(ZFS_PROP_VERSION), &zplver);
3034 (void) nvlist_lookup_uint64(createprops,
3035 zfs_prop_to_name(ZFS_PROP_NORMALIZE), &norm);
3036 (void) nvlist_remove_all(createprops,
3037 zfs_prop_to_name(ZFS_PROP_NORMALIZE));
3038 (void) nvlist_lookup_uint64(createprops,
3039 zfs_prop_to_name(ZFS_PROP_UTF8ONLY), &u8);
3040 (void) nvlist_remove_all(createprops,
3041 zfs_prop_to_name(ZFS_PROP_UTF8ONLY));
3042 (void) nvlist_lookup_uint64(createprops,
3043 zfs_prop_to_name(ZFS_PROP_CASE), &sense);
3044 (void) nvlist_remove_all(createprops,
3045 zfs_prop_to_name(ZFS_PROP_CASE));
3046 }
3047
3048 /*
3049 * If the zpl version requested is whacky or the file system
3050 * or pool is version is too "young" to support normalization
3051 * and the creator tried to set a value for one of the props,
3052 * error out.
3053 */
3054 if ((zplver < ZPL_VERSION_INITIAL || zplver > ZPL_VERSION) ||
3055 (zplver >= ZPL_VERSION_FUID && !fuids_ok) ||
3056 (zplver >= ZPL_VERSION_SA && !sa_ok) ||
3057 (zplver < ZPL_VERSION_NORMALIZATION &&
3058 (norm != ZFS_PROP_UNDEFINED || u8 != ZFS_PROP_UNDEFINED ||
3059 sense != ZFS_PROP_UNDEFINED)))
3060 return (SET_ERROR(ENOTSUP));
3061
3062 /*
3063 * Put the version in the zplprops
3064 */
3065 VERIFY(nvlist_add_uint64(zplprops,
3066 zfs_prop_to_name(ZFS_PROP_VERSION), zplver) == 0);
3067
3068 if (norm == ZFS_PROP_UNDEFINED)
3069 VERIFY(zfs_get_zplprop(os, ZFS_PROP_NORMALIZE, &norm) == 0);
3070 VERIFY(nvlist_add_uint64(zplprops,
3071 zfs_prop_to_name(ZFS_PROP_NORMALIZE), norm) == 0);
3072
3073 /*
3074 * If we're normalizing, names must always be valid UTF-8 strings.
3075 */
3076 if (norm)
3077 u8 = 1;
3078 if (u8 == ZFS_PROP_UNDEFINED)
3079 VERIFY(zfs_get_zplprop(os, ZFS_PROP_UTF8ONLY, &u8) == 0);
3080 VERIFY(nvlist_add_uint64(zplprops,
3081 zfs_prop_to_name(ZFS_PROP_UTF8ONLY), u8) == 0);
3082
3083 if (sense == ZFS_PROP_UNDEFINED)
3084 VERIFY(zfs_get_zplprop(os, ZFS_PROP_CASE, &sense) == 0);
3085 VERIFY(nvlist_add_uint64(zplprops,
3086 zfs_prop_to_name(ZFS_PROP_CASE), sense) == 0);
3087
3088 if (is_ci)
3089 *is_ci = (sense == ZFS_CASE_INSENSITIVE);
3090
3091 return (0);
3092 }
3093
3094 static int
3095 zfs_fill_zplprops(const char *dataset, nvlist_t *createprops,
3096 nvlist_t *zplprops, boolean_t *is_ci)
3097 {
3098 boolean_t fuids_ok, sa_ok;
3099 uint64_t zplver = ZPL_VERSION;
3100 objset_t *os = NULL;
3101 char parentname[MAXNAMELEN];
3102 char *cp;
3103 spa_t *spa;
3104 uint64_t spa_vers;
3105 int error;
3106
3107 (void) strlcpy(parentname, dataset, sizeof (parentname));
3108 cp = strrchr(parentname, '/');
3109 ASSERT(cp != NULL);
3110 cp[0] = '\0';
3111
3112 if ((error = spa_open(dataset, &spa, FTAG)) != 0)
3113 return (error);
3114
3115 spa_vers = spa_version(spa);
3116 spa_close(spa, FTAG);
3117
3118 zplver = zfs_zpl_version_map(spa_vers);
3119 fuids_ok = (zplver >= ZPL_VERSION_FUID);
3120 sa_ok = (zplver >= ZPL_VERSION_SA);
3121
3122 /*
3123 * Open parent object set so we can inherit zplprop values.
3124 */
3125 if ((error = dmu_objset_hold(parentname, FTAG, &os)) != 0)
3126 return (error);
3127
3128 error = zfs_fill_zplprops_impl(os, zplver, fuids_ok, sa_ok, createprops,
3129 zplprops, is_ci);
3130 dmu_objset_rele(os, FTAG);
3131 return (error);
3132 }
3133
3134 static int
3135 zfs_fill_zplprops_root(uint64_t spa_vers, nvlist_t *createprops,
3136 nvlist_t *zplprops, boolean_t *is_ci)
3137 {
3138 boolean_t fuids_ok;
3139 boolean_t sa_ok;
3140 uint64_t zplver = ZPL_VERSION;
3141 int error;
3142
3143 zplver = zfs_zpl_version_map(spa_vers);
3144 fuids_ok = (zplver >= ZPL_VERSION_FUID);
3145 sa_ok = (zplver >= ZPL_VERSION_SA);
3146
3147 error = zfs_fill_zplprops_impl(NULL, zplver, fuids_ok, sa_ok,
3148 createprops, zplprops, is_ci);
3149 return (error);
3150 }
3151
3152 /*
3153 * innvl: {
3154 * "type" -> dmu_objset_type_t (int32)
3155 * (optional) "props" -> { prop -> value }
3156 * }
3157 *
3158 * outnvl: propname -> error code (int32)
3159 */
3160 static int
3161 zfs_ioc_create(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3162 {
3163 int error = 0;
3164 zfs_creat_t zct = { 0 };
3165 nvlist_t *nvprops = NULL;
3166 void (*cbfunc)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx);
3167 int32_t type32;
3168 dmu_objset_type_t type;
3169 boolean_t is_insensitive = B_FALSE;
3170
3171 if (nvlist_lookup_int32(innvl, "type", &type32) != 0)
3172 return (SET_ERROR(EINVAL));
3173 type = type32;
3174 (void) nvlist_lookup_nvlist(innvl, "props", &nvprops);
3175
3176 switch (type) {
3177 case DMU_OST_ZFS:
3178 cbfunc = zfs_create_cb;
3179 break;
3180
3181 case DMU_OST_ZVOL:
3182 cbfunc = zvol_create_cb;
3183 break;
3184
3185 default:
3186 cbfunc = NULL;
3187 break;
3188 }
3189 if (strchr(fsname, '@') ||
3190 strchr(fsname, '%'))
3191 return (SET_ERROR(EINVAL));
3192
3193 zct.zct_props = nvprops;
3194
3195 if (cbfunc == NULL)
3196 return (SET_ERROR(EINVAL));
3197
3198 if (type == DMU_OST_ZVOL) {
3199 uint64_t volsize, volblocksize;
3200
3201 if (nvprops == NULL)
3202 return (SET_ERROR(EINVAL));
3203 if (nvlist_lookup_uint64(nvprops,
3204 zfs_prop_to_name(ZFS_PROP_VOLSIZE), &volsize) != 0)
3205 return (SET_ERROR(EINVAL));
3206
3207 if ((error = nvlist_lookup_uint64(nvprops,
3208 zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE),
3209 &volblocksize)) != 0 && error != ENOENT)
3210 return (SET_ERROR(EINVAL));
3211
3212 if (error != 0)
3213 volblocksize = zfs_prop_default_numeric(
3214 ZFS_PROP_VOLBLOCKSIZE);
3215
3216 if ((error = zvol_check_volblocksize(
3217 volblocksize)) != 0 ||
3218 (error = zvol_check_volsize(volsize,
3219 volblocksize)) != 0)
3220 return (error);
3221 } else if (type == DMU_OST_ZFS) {
3222 int error;
3223
3224 /*
3225 * We have to have normalization and
3226 * case-folding flags correct when we do the
3227 * file system creation, so go figure them out
3228 * now.
3229 */
3230 VERIFY(nvlist_alloc(&zct.zct_zplprops,
3231 NV_UNIQUE_NAME, KM_SLEEP) == 0);
3232 error = zfs_fill_zplprops(fsname, nvprops,
3233 zct.zct_zplprops, &is_insensitive);
3234 if (error != 0) {
3235 nvlist_free(zct.zct_zplprops);
3236 return (error);
3237 }
3238 }
3239
3240 error = dmu_objset_create(fsname, type,
3241 is_insensitive ? DS_FLAG_CI_DATASET : 0, cbfunc, &zct);
3242 nvlist_free(zct.zct_zplprops);
3243
3244 /*
3245 * It would be nice to do this atomically.
3246 */
3247 if (error == 0) {
3248 error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL,
3249 nvprops, outnvl);
3250 if (error != 0)
3251 (void) dsl_destroy_head(fsname);
3252 }
3253 return (error);
3254 }
3255
3256 /*
3257 * innvl: {
3258 * "origin" -> name of origin snapshot
3259 * (optional) "props" -> { prop -> value }
3260 * }
3261 *
3262 * outnvl: propname -> error code (int32)
3263 */
3264 static int
3265 zfs_ioc_clone(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3266 {
3267 int error = 0;
3268 nvlist_t *nvprops = NULL;
3269 char *origin_name;
3270
3271 if (nvlist_lookup_string(innvl, "origin", &origin_name) != 0)
3272 return (SET_ERROR(EINVAL));
3273 (void) nvlist_lookup_nvlist(innvl, "props", &nvprops);
3274
3275 if (strchr(fsname, '@') ||
3276 strchr(fsname, '%'))
3277 return (SET_ERROR(EINVAL));
3278
3279 if (dataset_namecheck(origin_name, NULL, NULL) != 0)
3280 return (SET_ERROR(EINVAL));
3281 error = dmu_objset_clone(fsname, origin_name);
3282 if (error != 0)
3283 return (error);
3284
3285 /*
3286 * It would be nice to do this atomically.
3287 */
3288 if (error == 0) {
3289 error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL,
3290 nvprops, outnvl);
3291 if (error != 0)
3292 (void) dsl_destroy_head(fsname);
3293 }
3294 return (error);
3295 }
3296
3297 /*
3298 * innvl: {
3299 * "snaps" -> { snapshot1, snapshot2 }
3300 * (optional) "props" -> { prop -> value (string) }
3301 * }
3302 *
3303 * outnvl: snapshot -> error code (int32)
3304 */
3305 static int
3306 zfs_ioc_snapshot(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3307 {
3308 nvlist_t *snaps;
3309 nvlist_t *props = NULL;
3310 int error, poollen;
3311 nvpair_t *pair;
3312
3313 (void) nvlist_lookup_nvlist(innvl, "props", &props);
3314 if ((error = zfs_check_userprops(poolname, props)) != 0)
3315 return (error);
3316
3317 if (!nvlist_empty(props) &&
3318 zfs_earlier_version(poolname, SPA_VERSION_SNAP_PROPS))
3319 return (SET_ERROR(ENOTSUP));
3320
3321 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
3322 return (SET_ERROR(EINVAL));
3323 poollen = strlen(poolname);
3324 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
3325 pair = nvlist_next_nvpair(snaps, pair)) {
3326 const char *name = nvpair_name(pair);
3327 const char *cp = strchr(name, '@');
3328
3329 /*
3330 * The snap name must contain an @, and the part after it must
3331 * contain only valid characters.
3332 */
3333 if (cp == NULL ||
3334 zfs_component_namecheck(cp + 1, NULL, NULL) != 0)
3335 return (SET_ERROR(EINVAL));
3336
3337 /*
3338 * The snap must be in the specified pool.
3339 */
3340 if (strncmp(name, poolname, poollen) != 0 ||
3341 (name[poollen] != '/' && name[poollen] != '@'))
3342 return (SET_ERROR(EXDEV));
3343
3344 /* This must be the only snap of this fs. */
3345 for (nvpair_t *pair2 = nvlist_next_nvpair(snaps, pair);
3346 pair2 != NULL; pair2 = nvlist_next_nvpair(snaps, pair2)) {
3347 if (strncmp(name, nvpair_name(pair2), cp - name + 1)
3348 == 0) {
3349 return (SET_ERROR(EXDEV));
3350 }
3351 }
3352 }
3353
3354 error = dsl_dataset_snapshot(snaps, props, outnvl);
3355 return (error);
3356 }
3357
3358 /*
3359 * innvl: "message" -> string
3360 */
3361 /* ARGSUSED */
3362 static int
3363 zfs_ioc_log_history(const char *unused, nvlist_t *innvl, nvlist_t *outnvl)
3364 {
3365 char *message;
3366 spa_t *spa;
3367 int error;
3368 char *poolname;
3369
3370 /*
3371 * The poolname in the ioctl is not set, we get it from the TSD,
3372 * which was set at the end of the last successful ioctl that allows
3373 * logging. The secpolicy func already checked that it is set.
3374 * Only one log ioctl is allowed after each successful ioctl, so
3375 * we clear the TSD here.
3376 */
3377 poolname = tsd_get(zfs_allow_log_key);
3378 (void) tsd_set(zfs_allow_log_key, NULL);
3379 error = spa_open(poolname, &spa, FTAG);
3380 strfree(poolname);
3381 if (error != 0)
3382 return (error);
3383
3384 if (nvlist_lookup_string(innvl, "message", &message) != 0) {
3385 spa_close(spa, FTAG);
3386 return (SET_ERROR(EINVAL));
3387 }
3388
3389 if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) {
3390 spa_close(spa, FTAG);
3391 return (SET_ERROR(ENOTSUP));
3392 }
3393
3394 error = spa_history_log(spa, message);
3395 spa_close(spa, FTAG);
3396 return (error);
3397 }
3398
3399 /*
3400 * The dp_config_rwlock must not be held when calling this, because the
3401 * unmount may need to write out data.
3402 *
3403 * This function is best-effort. Callers must deal gracefully if it
3404 * remains mounted (or is remounted after this call).
3405 *
3406 * Returns 0 if the argument is not a snapshot, or it is not currently a
3407 * filesystem, or we were able to unmount it. Returns error code otherwise.
3408 */
3409 int
3410 zfs_unmount_snap(const char *snapname)
3411 {
3412 vfs_t *vfsp;
3413 zfsvfs_t *zfsvfs;
3414 int err;
3415
3416 if (strchr(snapname, '@') == NULL)
3417 return (0);
3418
3419 vfsp = zfs_get_vfs(snapname);
3420 if (vfsp == NULL)
3421 return (0);
3422
3423 zfsvfs = vfsp->vfs_data;
3424 ASSERT(!dsl_pool_config_held(dmu_objset_pool(zfsvfs->z_os)));
3425
3426 err = vn_vfswlock(vfsp->vfs_vnodecovered);
3427 VFS_RELE(vfsp);
3428 if (err != 0)
3429 return (SET_ERROR(err));
3430
3431 /*
3432 * Always force the unmount for snapshots.
3433 */
3434 (void) dounmount(vfsp, MS_FORCE, kcred);
3435 return (0);
3436 }
3437
3438 /* ARGSUSED */
3439 static int
3440 zfs_unmount_snap_cb(const char *snapname, void *arg)
3441 {
3442 return (zfs_unmount_snap(snapname));
3443 }
3444
3445 /*
3446 * When a clone is destroyed, its origin may also need to be destroyed,
3447 * in which case it must be unmounted. This routine will do that unmount
3448 * if necessary.
3449 */
3450 void
3451 zfs_destroy_unmount_origin(const char *fsname)
3452 {
3453 int error;
3454 objset_t *os;
3455 dsl_dataset_t *ds;
3456
3457 error = dmu_objset_hold(fsname, FTAG, &os);
3458 if (error != 0)
3459 return;
3460 ds = dmu_objset_ds(os);
3461 if (dsl_dir_is_clone(ds->ds_dir) && DS_IS_DEFER_DESTROY(ds->ds_prev)) {
3462 char originname[MAXNAMELEN];
3463 dsl_dataset_name(ds->ds_prev, originname);
3464 dmu_objset_rele(os, FTAG);
3465 (void) zfs_unmount_snap(originname);
3466 } else {
3467 dmu_objset_rele(os, FTAG);
3468 }
3469 }
3470
3471 /*
3472 * innvl: {
3473 * "snaps" -> { snapshot1, snapshot2 }
3474 * (optional boolean) "defer"
3475 * }
3476 *
3477 * outnvl: snapshot -> error code (int32)
3478 *
3479 */
3480 /* ARGSUSED */
3481 static int
3482 zfs_ioc_destroy_snaps(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3483 {
3484 nvlist_t *snaps;
3485 nvpair_t *pair;
3486 boolean_t defer;
3487
3488 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
3489 return (SET_ERROR(EINVAL));
3490 defer = nvlist_exists(innvl, "defer");
3491
3492 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
3493 pair = nvlist_next_nvpair(snaps, pair)) {
3494 (void) zfs_unmount_snap(nvpair_name(pair));
3495 }
3496
3497 return (dsl_destroy_snapshots_nvl(snaps, defer, outnvl));
3498 }
3499
3500 /*
3501 * Create bookmarks. Bookmark names are of the form <fs>#<bmark>.
3502 * All bookmarks must be in the same pool.
3503 *
3504 * innvl: {
3505 * bookmark1 -> snapshot1, bookmark2 -> snapshot2
3506 * }
3507 *
3508 * outnvl: bookmark -> error code (int32)
3509 *
3510 */
3511 /* ARGSUSED */
3512 static int
3513 zfs_ioc_bookmark(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3514 {
3515 for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL);
3516 pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) {
3517 char *snap_name;
3518
3519 /*
3520 * Verify the snapshot argument.
3521 */
3522 if (nvpair_value_string(pair, &snap_name) != 0)
3523 return (SET_ERROR(EINVAL));
3524
3525
3526 /* Verify that the keys (bookmarks) are unique */
3527 for (nvpair_t *pair2 = nvlist_next_nvpair(innvl, pair);
3528 pair2 != NULL; pair2 = nvlist_next_nvpair(innvl, pair2)) {
3529 if (strcmp(nvpair_name(pair), nvpair_name(pair2)) == 0)
3530 return (SET_ERROR(EINVAL));
3531 }
3532 }
3533
3534 return (dsl_bookmark_create(innvl, outnvl));
3535 }
3536
3537 /*
3538 * innvl: {
3539 * property 1, property 2, ...
3540 * }
3541 *
3542 * outnvl: {
3543 * bookmark name 1 -> { property 1, property 2, ... },
3544 * bookmark name 2 -> { property 1, property 2, ... }
3545 * }
3546 *
3547 */
3548 static int
3549 zfs_ioc_get_bookmarks(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3550 {
3551 return (dsl_get_bookmarks(fsname, innvl, outnvl));
3552 }
3553
3554 /*
3555 * innvl: {
3556 * bookmark name 1, bookmark name 2
3557 * }
3558 *
3559 * outnvl: bookmark -> error code (int32)
3560 *
3561 */
3562 static int
3563 zfs_ioc_destroy_bookmarks(const char *poolname, nvlist_t *innvl,
3564 nvlist_t *outnvl)
3565 {
3566 int error, poollen;
3567
3568 poollen = strlen(poolname);
3569 for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL);
3570 pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) {
3571 const char *name = nvpair_name(pair);
3572 const char *cp = strchr(name, '#');
3573
3574 /*
3575 * The bookmark name must contain an #, and the part after it
3576 * must contain only valid characters.
3577 */
3578 if (cp == NULL ||
3579 zfs_component_namecheck(cp + 1, NULL, NULL) != 0)
3580 return (SET_ERROR(EINVAL));
3581
3582 /*
3583 * The bookmark must be in the specified pool.
3584 */
3585 if (strncmp(name, poolname, poollen) != 0 ||
3586 (name[poollen] != '/' && name[poollen] != '#'))
3587 return (SET_ERROR(EXDEV));
3588 }
3589
3590 error = dsl_bookmark_destroy(innvl, outnvl);
3591 return (error);
3592 }
3593
3594 /*
3595 * inputs:
3596 * zc_name name of dataset to destroy
3597 * zc_objset_type type of objset
3598 * zc_defer_destroy mark for deferred destroy
3599 *
3600 * outputs: none
3601 */
3602 static int
3603 zfs_ioc_destroy(zfs_cmd_t *zc)
3604 {
3605 int err;
3606
3607 if (zc->zc_objset_type == DMU_OST_ZFS) {
3608 err = zfs_unmount_snap(zc->zc_name);
3609 if (err != 0)
3610 return (err);
3611 }
3612
3613 if (strchr(zc->zc_name, '@'))
3614 err = dsl_destroy_snapshot(zc->zc_name, zc->zc_defer_destroy);
3615 else
3616 err = dsl_destroy_head(zc->zc_name);
3617 if (zc->zc_objset_type == DMU_OST_ZVOL && err == 0)
3618 (void) zvol_remove_minor(zc->zc_name);
3619 return (err);
3620 }
3621
3622 /*
3623 * fsname is name of dataset to rollback (to most recent snapshot)
3624 *
3625 * innvl is not used.
3626 *
3627 * outnvl: "target" -> name of most recent snapshot
3628 * }
3629 */
3630 /* ARGSUSED */
3631 static int
3632 zfs_ioc_rollback(const char *fsname, nvlist_t *args, nvlist_t *outnvl)
3633 {
3634 zfsvfs_t *zfsvfs;
3635 int error;
3636
3637 if (getzfsvfs(fsname, &zfsvfs) == 0) {
3638 error = zfs_suspend_fs(zfsvfs);
3639 if (error == 0) {
3640 int resume_err;
3641
3642 error = dsl_dataset_rollback(fsname, zfsvfs, outnvl);
3643 resume_err = zfs_resume_fs(zfsvfs, fsname);
3644 error = error ? error : resume_err;
3645 }
3646 VFS_RELE(zfsvfs->z_vfs);
3647 } else {
3648 error = dsl_dataset_rollback(fsname, NULL, outnvl);
3649 }
3650 return (error);
3651 }
3652
3653 static int
3654 recursive_unmount(const char *fsname, void *arg)
3655 {
3656 const char *snapname = arg;
3657 char fullname[MAXNAMELEN];
3658
3659 (void) snprintf(fullname, sizeof (fullname), "%s@%s", fsname, snapname);
3660 return (zfs_unmount_snap(fullname));
3661 }
3662
3663 /*
3664 * inputs:
3665 * zc_name old name of dataset
3666 * zc_value new name of dataset
3667 * zc_cookie recursive flag (only valid for snapshots)
3668 *
3669 * outputs: none
3670 */
3671 static int
3672 zfs_ioc_rename(zfs_cmd_t *zc)
3673 {
3674 boolean_t recursive = zc->zc_cookie & 1;
3675 char *at;
3676
3677 zc->zc_value[sizeof (zc->zc_value) - 1] = '\0';
3678 if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
3679 strchr(zc->zc_value, '%'))
3680 return (SET_ERROR(EINVAL));
3681
3682 at = strchr(zc->zc_name, '@');
3683 if (at != NULL) {
3684 /* snaps must be in same fs */
3685 int error;
3686
3687 if (strncmp(zc->zc_name, zc->zc_value, at - zc->zc_name + 1))
3688 return (SET_ERROR(EXDEV));
3689 *at = '\0';
3690 if (zc->zc_objset_type == DMU_OST_ZFS) {
3691 error = dmu_objset_find(zc->zc_name,
3692 recursive_unmount, at + 1,
3693 recursive ? DS_FIND_CHILDREN : 0);
3694 if (error != 0) {
3695 *at = '@';
3696 return (error);
3697 }
3698 }
3699 error = dsl_dataset_rename_snapshot(zc->zc_name,
3700 at + 1, strchr(zc->zc_value, '@') + 1, recursive);
3701 *at = '@';
3702
3703 return (error);
3704 } else {
3705 if (zc->zc_objset_type == DMU_OST_ZVOL)
3706 (void) zvol_remove_minor(zc->zc_name);
3707 return (dsl_dir_rename(zc->zc_name, zc->zc_value));
3708 }
3709 }
3710
3711 static int
3712 zfs_check_settable(const char *dsname, nvpair_t *pair, cred_t *cr)
3713 {
3714 const char *propname = nvpair_name(pair);
3715 boolean_t issnap = (strchr(dsname, '@') != NULL);
3716 zfs_prop_t prop = zfs_name_to_prop(propname);
3717 uint64_t intval;
3718 int err;
3719
3720 if (prop == ZPROP_INVAL) {
3721 if (zfs_prop_user(propname)) {
3722 if (err = zfs_secpolicy_write_perms(dsname,
3723 ZFS_DELEG_PERM_USERPROP, cr))
3724 return (err);
3725 return (0);
3726 }
3727
3728 if (!issnap && zfs_prop_userquota(propname)) {
3729 const char *perm = NULL;
3730 const char *uq_prefix =
3731 zfs_userquota_prop_prefixes[ZFS_PROP_USERQUOTA];
3732 const char *gq_prefix =
3733 zfs_userquota_prop_prefixes[ZFS_PROP_GROUPQUOTA];
3734
3735 if (strncmp(propname, uq_prefix,
3736 strlen(uq_prefix)) == 0) {
3737 perm = ZFS_DELEG_PERM_USERQUOTA;
3738 } else if (strncmp(propname, gq_prefix,
3739 strlen(gq_prefix)) == 0) {
3740 perm = ZFS_DELEG_PERM_GROUPQUOTA;
3741 } else {
3742 /* USERUSED and GROUPUSED are read-only */
3743 return (SET_ERROR(EINVAL));
3744 }
3745
3746 if (err = zfs_secpolicy_write_perms(dsname, perm, cr))
3747 return (err);
3748 return (0);
3749 }
3750
3751 return (SET_ERROR(EINVAL));
3752 }
3753
3754 if (issnap)
3755 return (SET_ERROR(EINVAL));
3756
3757 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
3758 /*
3759 * dsl_prop_get_all_impl() returns properties in this
3760 * format.
3761 */
3762 nvlist_t *attrs;
3763 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
3764 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
3765 &pair) == 0);
3766 }
3767
3768 /*
3769 * Check that this value is valid for this pool version
3770 */
3771 switch (prop) {
3772 case ZFS_PROP_COMPRESSION:
3773 /*
3774 * If the user specified gzip compression, make sure
3775 * the SPA supports it. We ignore any errors here since
3776 * we'll catch them later.
3777 */
3778 if (nvpair_value_uint64(pair, &intval) == 0) {
3779 if (intval >= ZIO_COMPRESS_GZIP_1 &&
3780 intval <= ZIO_COMPRESS_GZIP_9 &&
3781 zfs_earlier_version(dsname,
3782 SPA_VERSION_GZIP_COMPRESSION)) {
3783 return (SET_ERROR(ENOTSUP));
3784 }
3785
3786 if (intval == ZIO_COMPRESS_ZLE &&
3787 zfs_earlier_version(dsname,
3788 SPA_VERSION_ZLE_COMPRESSION))
3789 return (SET_ERROR(ENOTSUP));
3790
3791 if (intval == ZIO_COMPRESS_LZ4) {
3792 spa_t *spa;
3793
3794 if ((err = spa_open(dsname, &spa, FTAG)) != 0)
3795 return (err);
3796
3797 if (!spa_feature_is_enabled(spa,
3798 SPA_FEATURE_LZ4_COMPRESS)) {
3799 spa_close(spa, FTAG);
3800 return (SET_ERROR(ENOTSUP));
3801 }
3802 spa_close(spa, FTAG);
3803 }
3804
3805 /*
3806 * If this is a bootable dataset then
3807 * verify that the compression algorithm
3808 * is supported for booting. We must return
3809 * something other than ENOTSUP since it
3810 * implies a downrev pool version.
3811 */
3812 if (zfs_is_bootfs(dsname) &&
3813 !BOOTFS_COMPRESS_VALID(intval)) {
3814 return (SET_ERROR(ERANGE));
3815 }
3816 }
3817 break;
3818
3819 case ZFS_PROP_COPIES:
3820 if (zfs_earlier_version(dsname, SPA_VERSION_DITTO_BLOCKS))
3821 return (SET_ERROR(ENOTSUP));
3822 break;
3823
3824 case ZFS_PROP_RECORDSIZE:
3825 /* Record sizes above 128k need the feature to be enabled */
3826 if (nvpair_value_uint64(pair, &intval) == 0 &&
3827 intval > SPA_OLD_MAXBLOCKSIZE) {
3828 spa_t *spa;
3829
3830 /*
3831 * If this is a bootable dataset then
3832 * the we don't allow large (>128K) blocks,
3833 * because GRUB doesn't support them.
3834 */
3835 if (zfs_is_bootfs(dsname) &&
3836 intval > SPA_OLD_MAXBLOCKSIZE) {
3837 return (SET_ERROR(ERANGE));
3838 }
3839
3840 /*
3841 * We don't allow setting the property above 1MB,
3842 * unless the tunable has been changed.
3843 */
3844 if (intval > zfs_max_recordsize ||
3845 intval > SPA_MAXBLOCKSIZE)
3846 return (SET_ERROR(ERANGE));
3847
3848 if ((err = spa_open(dsname, &spa, FTAG)) != 0)
3849 return (err);
3850
3851 if (!spa_feature_is_enabled(spa,
3852 SPA_FEATURE_LARGE_BLOCKS)) {
3853 spa_close(spa, FTAG);
3854 return (SET_ERROR(ENOTSUP));
3855 }
3856 spa_close(spa, FTAG);
3857 }
3858 break;
3859
3860 case ZFS_PROP_SHARESMB:
3861 if (zpl_earlier_version(dsname, ZPL_VERSION_FUID))
3862 return (SET_ERROR(ENOTSUP));
3863 break;
3864
3865 case ZFS_PROP_ACLINHERIT:
3866 if (nvpair_type(pair) == DATA_TYPE_UINT64 &&
3867 nvpair_value_uint64(pair, &intval) == 0) {
3868 if (intval == ZFS_ACL_PASSTHROUGH_X &&
3869 zfs_earlier_version(dsname,
3870 SPA_VERSION_PASSTHROUGH_X))
3871 return (SET_ERROR(ENOTSUP));
3872 }
3873 break;
3874
3875 case ZFS_PROP_CHECKSUM:
3876 case ZFS_PROP_DEDUP:
3877 {
3878 spa_feature_t feature;
3879 spa_t *spa;
3880
3881 /* dedup feature version checks */
3882 if (prop == ZFS_PROP_DEDUP &&
3883 zfs_earlier_version(dsname, SPA_VERSION_DEDUP))
3884 return (SET_ERROR(ENOTSUP));
3885
3886 if (nvpair_value_uint64(pair, &intval) != 0)
3887 return (SET_ERROR(EINVAL));
3888
3889 /* check prop value is enabled in features */
3890 feature = zio_checksum_to_feature(intval);
3891 if (feature == SPA_FEATURE_NONE)
3892 break;
3893
3894 if ((err = spa_open(dsname, &spa, FTAG)) != 0)
3895 return (err);
3896 /*
3897 * Salted checksums are not supported on root pools.
3898 */
3899 if (spa_bootfs(spa) != 0 &&
3900 intval < ZIO_CHECKSUM_FUNCTIONS &&
3901 (zio_checksum_table[intval].ci_flags &
3902 ZCHECKSUM_FLAG_SALTED)) {
3903 spa_close(spa, FTAG);
3904 return (SET_ERROR(ERANGE));
3905 }
3906 if (!spa_feature_is_enabled(spa, feature)) {
3907 spa_close(spa, FTAG);
3908 return (SET_ERROR(ENOTSUP));
3909 }
3910 spa_close(spa, FTAG);
3911 break;
3912 }
3913 }
3914
3915 return (zfs_secpolicy_setprop(dsname, prop, pair, CRED()));
3916 }
3917
3918 /*
3919 * Checks for a race condition to make sure we don't increment a feature flag
3920 * multiple times.
3921 */
3922 static int
3923 zfs_prop_activate_feature_check(void *arg, dmu_tx_t *tx)
3924 {
3925 spa_t *spa = dmu_tx_pool(tx)->dp_spa;
3926 spa_feature_t *featurep = arg;
3927
3928 if (!spa_feature_is_active(spa, *featurep))
3929 return (0);
3930 else
3931 return (SET_ERROR(EBUSY));
3932 }
3933
3934 /*
3935 * The callback invoked on feature activation in the sync task caused by
3936 * zfs_prop_activate_feature.
3937 */
3938 static void
3939 zfs_prop_activate_feature_sync(void *arg, dmu_tx_t *tx)
3940 {
3941 spa_t *spa = dmu_tx_pool(tx)->dp_spa;
3942 spa_feature_t *featurep = arg;
3943
3944 spa_feature_incr(spa, *featurep, tx);
3945 }
3946
3947 /*
3948 * Activates a feature on a pool in response to a property setting. This
3949 * creates a new sync task which modifies the pool to reflect the feature
3950 * as being active.
3951 */
3952 static int
3953 zfs_prop_activate_feature(spa_t *spa, spa_feature_t feature)
3954 {
3955 int err;
3956
3957 /* EBUSY here indicates that the feature is already active */
3958 err = dsl_sync_task(spa_name(spa),
3959 zfs_prop_activate_feature_check, zfs_prop_activate_feature_sync,
3960 &feature, 2, ZFS_SPACE_CHECK_RESERVED);
3961
3962 if (err != 0 && err != EBUSY)
3963 return (err);
3964 else
3965 return (0);
3966 }
3967
3968 /*
3969 * Removes properties from the given props list that fail permission checks
3970 * needed to clear them and to restore them in case of a receive error. For each
3971 * property, make sure we have both set and inherit permissions.
3972 *
3973 * Returns the first error encountered if any permission checks fail. If the
3974 * caller provides a non-NULL errlist, it also gives the complete list of names
3975 * of all the properties that failed a permission check along with the
3976 * corresponding error numbers. The caller is responsible for freeing the
3977 * returned errlist.
3978 *
3979 * If every property checks out successfully, zero is returned and the list
3980 * pointed at by errlist is NULL.
3981 */
3982 static int
3983 zfs_check_clearable(char *dataset, nvlist_t *props, nvlist_t **errlist)
3984 {
3985 zfs_cmd_t *zc;
3986 nvpair_t *pair, *next_pair;
3987 nvlist_t *errors;
3988 int err, rv = 0;
3989
3990 if (props == NULL)
3991 return (0);
3992
3993 VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0);
3994
3995 zc = kmem_alloc(sizeof (zfs_cmd_t), KM_SLEEP);
3996 (void) strcpy(zc->zc_name, dataset);
3997 pair = nvlist_next_nvpair(props, NULL);
3998 while (pair != NULL) {
3999 next_pair = nvlist_next_nvpair(props, pair);
4000
4001 (void) strcpy(zc->zc_value, nvpair_name(pair));
4002 if ((err = zfs_check_settable(dataset, pair, CRED())) != 0 ||
4003 (err = zfs_secpolicy_inherit_prop(zc, NULL, CRED())) != 0) {
4004 VERIFY(nvlist_remove_nvpair(props, pair) == 0);
4005 VERIFY(nvlist_add_int32(errors,
4006 zc->zc_value, err) == 0);
4007 }
4008 pair = next_pair;
4009 }
4010 kmem_free(zc, sizeof (zfs_cmd_t));
4011
4012 if ((pair = nvlist_next_nvpair(errors, NULL)) == NULL) {
4013 nvlist_free(errors);
4014 errors = NULL;
4015 } else {
4016 VERIFY(nvpair_value_int32(pair, &rv) == 0);
4017 }
4018
4019 if (errlist == NULL)
4020 nvlist_free(errors);
4021 else
4022 *errlist = errors;
4023
4024 return (rv);
4025 }
4026
4027 static boolean_t
4028 propval_equals(nvpair_t *p1, nvpair_t *p2)
4029 {
4030 if (nvpair_type(p1) == DATA_TYPE_NVLIST) {
4031 /* dsl_prop_get_all_impl() format */
4032 nvlist_t *attrs;
4033 VERIFY(nvpair_value_nvlist(p1, &attrs) == 0);
4034 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
4035 &p1) == 0);
4036 }
4037
4038 if (nvpair_type(p2) == DATA_TYPE_NVLIST) {
4039 nvlist_t *attrs;
4040 VERIFY(nvpair_value_nvlist(p2, &attrs) == 0);
4041 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
4042 &p2) == 0);
4043 }
4044
4045 if (nvpair_type(p1) != nvpair_type(p2))
4046 return (B_FALSE);
4047
4048 if (nvpair_type(p1) == DATA_TYPE_STRING) {
4049 char *valstr1, *valstr2;
4050
4051 VERIFY(nvpair_value_string(p1, (char **)&valstr1) == 0);
4052 VERIFY(nvpair_value_string(p2, (char **)&valstr2) == 0);
4053 return (strcmp(valstr1, valstr2) == 0);
4054 } else {
4055 uint64_t intval1, intval2;
4056
4057 VERIFY(nvpair_value_uint64(p1, &intval1) == 0);
4058 VERIFY(nvpair_value_uint64(p2, &intval2) == 0);
4059 return (intval1 == intval2);
4060 }
4061 }
4062
4063 /*
4064 * Remove properties from props if they are not going to change (as determined
4065 * by comparison with origprops). Remove them from origprops as well, since we
4066 * do not need to clear or restore properties that won't change.
4067 */
4068 static void
4069 props_reduce(nvlist_t *props, nvlist_t *origprops)
4070 {
4071 nvpair_t *pair, *next_pair;
4072
4073 if (origprops == NULL)
4074 return; /* all props need to be received */
4075
4076 pair = nvlist_next_nvpair(props, NULL);
4077 while (pair != NULL) {
4078 const char *propname = nvpair_name(pair);
4079 nvpair_t *match;
4080
4081 next_pair = nvlist_next_nvpair(props, pair);
4082
4083 if ((nvlist_lookup_nvpair(origprops, propname,
4084 &match) != 0) || !propval_equals(pair, match))
4085 goto next; /* need to set received value */
4086
4087 /* don't clear the existing received value */
4088 (void) nvlist_remove_nvpair(origprops, match);
4089 /* don't bother receiving the property */
4090 (void) nvlist_remove_nvpair(props, pair);
4091 next:
4092 pair = next_pair;
4093 }
4094 }
4095
4096 #ifdef DEBUG
4097 static boolean_t zfs_ioc_recv_inject_err;
4098 #endif
4099
4100 /*
4101 * inputs:
4102 * zc_name name of containing filesystem
4103 * zc_nvlist_src{_size} nvlist of properties to apply
4104 * zc_value name of snapshot to create
4105 * zc_string name of clone origin (if DRR_FLAG_CLONE)
4106 * zc_cookie file descriptor to recv from
4107 * zc_begin_record the BEGIN record of the stream (not byteswapped)
4108 * zc_guid force flag
4109 * zc_cleanup_fd cleanup-on-exit file descriptor
4110 * zc_action_handle handle for this guid/ds mapping (or zero on first call)
4111 * zc_resumable if data is incomplete assume sender will resume
4112 *
4113 * outputs:
4114 * zc_cookie number of bytes read
4115 * zc_nvlist_dst{_size} error for each unapplied received property
4116 * zc_obj zprop_errflags_t
4117 * zc_action_handle handle for this guid/ds mapping
4118 */
4119 static int
4120 zfs_ioc_recv(zfs_cmd_t *zc)
4121 {
4122 file_t *fp;
4123 dmu_recv_cookie_t drc;
4124 boolean_t force = (boolean_t)zc->zc_guid;
4125 int fd;
4126 int error = 0;
4127 int props_error = 0;
4128 nvlist_t *errors;
4129 offset_t off;
4130 nvlist_t *props = NULL; /* sent properties */
4131 nvlist_t *origprops = NULL; /* existing properties */
4132 char *origin = NULL;
4133 char *tosnap;
4134 char tofs[ZFS_MAXNAMELEN];
4135 boolean_t first_recvd_props = B_FALSE;
4136
4137 if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
4138 strchr(zc->zc_value, '@') == NULL ||
4139 strchr(zc->zc_value, '%'))
4140 return (SET_ERROR(EINVAL));
4141
4142 (void) strcpy(tofs, zc->zc_value);
4143 tosnap = strchr(tofs, '@');
4144 *tosnap++ = '\0';
4145
4146 if (zc->zc_nvlist_src != NULL &&
4147 (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
4148 zc->zc_iflags, &props)) != 0)
4149 return (error);
4150
4151 fd = zc->zc_cookie;
4152 fp = getf(fd);
4153 if (fp == NULL) {
4154 nvlist_free(props);
4155 return (SET_ERROR(EBADF));
4156 }
4157
4158 errors = fnvlist_alloc();
4159
4160 if (zc->zc_string[0])
4161 origin = zc->zc_string;
4162
4163 error = dmu_recv_begin(tofs, tosnap,
4164 &zc->zc_begin_record, force, zc->zc_resumable, origin, &drc);
4165 if (error != 0)
4166 goto out;
4167
4168 /*
4169 * Set properties before we receive the stream so that they are applied
4170 * to the new data. Note that we must call dmu_recv_stream() if
4171 * dmu_recv_begin() succeeds.
4172 */
4173 if (props != NULL && !drc.drc_newfs) {
4174 if (spa_version(dsl_dataset_get_spa(drc.drc_ds)) >=
4175 SPA_VERSION_RECVD_PROPS &&
4176 !dsl_prop_get_hasrecvd(tofs))
4177 first_recvd_props = B_TRUE;
4178
4179 /*
4180 * If new received properties are supplied, they are to
4181 * completely replace the existing received properties, so stash
4182 * away the existing ones.
4183 */
4184 if (dsl_prop_get_received(tofs, &origprops) == 0) {
4185 nvlist_t *errlist = NULL;
4186 /*
4187 * Don't bother writing a property if its value won't
4188 * change (and avoid the unnecessary security checks).
4189 *
4190 * The first receive after SPA_VERSION_RECVD_PROPS is a
4191 * special case where we blow away all local properties
4192 * regardless.
4193 */
4194 if (!first_recvd_props)
4195 props_reduce(props, origprops);
4196 if (zfs_check_clearable(tofs, origprops, &errlist) != 0)
4197 (void) nvlist_merge(errors, errlist, 0);
4198 nvlist_free(errlist);
4199
4200 if (clear_received_props(tofs, origprops,
4201 first_recvd_props ? NULL : props) != 0)
4202 zc->zc_obj |= ZPROP_ERR_NOCLEAR;
4203 } else {
4204 zc->zc_obj |= ZPROP_ERR_NOCLEAR;
4205 }
4206 }
4207
4208 if (props != NULL) {
4209 props_error = dsl_prop_set_hasrecvd(tofs);
4210
4211 if (props_error == 0) {
4212 (void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_RECEIVED,
4213 props, errors);
4214 }
4215 }
4216
4217 if (zc->zc_nvlist_dst_size != 0 &&
4218 (nvlist_smush(errors, zc->zc_nvlist_dst_size) != 0 ||
4219 put_nvlist(zc, errors) != 0)) {
4220 /*
4221 * Caller made zc->zc_nvlist_dst less than the minimum expected
4222 * size or supplied an invalid address.
4223 */
4224 props_error = SET_ERROR(EINVAL);
4225 }
4226
4227 off = fp->f_offset;
4228 error = dmu_recv_stream(&drc, fp->f_vnode, &off, zc->zc_cleanup_fd,
4229 &zc->zc_action_handle);
4230
4231 if (error == 0) {
4232 zfsvfs_t *zfsvfs = NULL;
4233
4234 if (getzfsvfs(tofs, &zfsvfs) == 0) {
4235 /* online recv */
4236 int end_err;
4237
4238 error = zfs_suspend_fs(zfsvfs);
4239 /*
4240 * If the suspend fails, then the recv_end will
4241 * likely also fail, and clean up after itself.
4242 */
4243 end_err = dmu_recv_end(&drc, zfsvfs);
4244 if (error == 0)
4245 error = zfs_resume_fs(zfsvfs, tofs);
4246 error = error ? error : end_err;
4247 VFS_RELE(zfsvfs->z_vfs);
4248 } else {
4249 error = dmu_recv_end(&drc, NULL);
4250 }
4251 }
4252
4253 zc->zc_cookie = off - fp->f_offset;
4254 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
4255 fp->f_offset = off;
4256
4257 #ifdef DEBUG
4258 if (zfs_ioc_recv_inject_err) {
4259 zfs_ioc_recv_inject_err = B_FALSE;
4260 error = 1;
4261 }
4262 #endif
4263 /*
4264 * On error, restore the original props.
4265 */
4266 if (error != 0 && props != NULL && !drc.drc_newfs) {
4267 if (clear_received_props(tofs, props, NULL) != 0) {
4268 /*
4269 * We failed to clear the received properties.
4270 * Since we may have left a $recvd value on the
4271 * system, we can't clear the $hasrecvd flag.
4272 */
4273 zc->zc_obj |= ZPROP_ERR_NORESTORE;
4274 } else if (first_recvd_props) {
4275 dsl_prop_unset_hasrecvd(tofs);
4276 }
4277
4278 if (origprops == NULL && !drc.drc_newfs) {
4279 /* We failed to stash the original properties. */
4280 zc->zc_obj |= ZPROP_ERR_NORESTORE;
4281 }
4282
4283 /*
4284 * dsl_props_set() will not convert RECEIVED to LOCAL on or
4285 * after SPA_VERSION_RECVD_PROPS, so we need to specify LOCAL
4286 * explictly if we're restoring local properties cleared in the
4287 * first new-style receive.
4288 */
4289 if (origprops != NULL &&
4290 zfs_set_prop_nvlist(tofs, (first_recvd_props ?
4291 ZPROP_SRC_LOCAL : ZPROP_SRC_RECEIVED),
4292 origprops, NULL) != 0) {
4293 /*
4294 * We stashed the original properties but failed to
4295 * restore them.
4296 */
4297 zc->zc_obj |= ZPROP_ERR_NORESTORE;
4298 }
4299 }
4300 out:
4301 nvlist_free(props);
4302 nvlist_free(origprops);
4303 nvlist_free(errors);
4304 releasef(fd);
4305
4306 if (error == 0)
4307 error = props_error;
4308
4309 return (error);
4310 }
4311
4312 /*
4313 * inputs:
4314 * zc_name name of snapshot to send
4315 * zc_cookie file descriptor to send stream to
4316 * zc_obj fromorigin flag (mutually exclusive with zc_fromobj)
4317 * zc_sendobj objsetid of snapshot to send
4318 * zc_fromobj objsetid of incremental fromsnap (may be zero)
4319 * zc_guid if set, estimate size of stream only. zc_cookie is ignored.
4320 * output size in zc_objset_type.
4321 * zc_flags lzc_send_flags
4322 *
4323 * outputs:
4324 * zc_objset_type estimated size, if zc_guid is set
4325 */
4326 static int
4327 zfs_ioc_send(zfs_cmd_t *zc)
4328 {
4329 int error;
4330 offset_t off;
4331 boolean_t estimate = (zc->zc_guid != 0);
4332 boolean_t embedok = (zc->zc_flags & 0x1);
4333 boolean_t large_block_ok = (zc->zc_flags & 0x2);
4334
4335 if (zc->zc_obj != 0) {
4336 dsl_pool_t *dp;
4337 dsl_dataset_t *tosnap;
4338
4339 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
4340 if (error != 0)
4341 return (error);
4342
4343 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &tosnap);
4344 if (error != 0) {
4345 dsl_pool_rele(dp, FTAG);
4346 return (error);
4347 }
4348
4349 if (dsl_dir_is_clone(tosnap->ds_dir))
4350 zc->zc_fromobj =
4351 dsl_dir_phys(tosnap->ds_dir)->dd_origin_obj;
4352 dsl_dataset_rele(tosnap, FTAG);
4353 dsl_pool_rele(dp, FTAG);
4354 }
4355
4356 if (estimate) {
4357 dsl_pool_t *dp;
4358 dsl_dataset_t *tosnap;
4359 dsl_dataset_t *fromsnap = NULL;
4360
4361 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
4362 if (error != 0)
4363 return (error);
4364
4365 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &tosnap);
4366 if (error != 0) {
4367 dsl_pool_rele(dp, FTAG);
4368 return (error);
4369 }
4370
4371 if (zc->zc_fromobj != 0) {
4372 error = dsl_dataset_hold_obj(dp, zc->zc_fromobj,
4373 FTAG, &fromsnap);
4374 if (error != 0) {
4375 dsl_dataset_rele(tosnap, FTAG);
4376 dsl_pool_rele(dp, FTAG);
4377 return (error);
4378 }
4379 }
4380
4381 error = dmu_send_estimate(tosnap, fromsnap,
4382 &zc->zc_objset_type);
4383
4384 if (fromsnap != NULL)
4385 dsl_dataset_rele(fromsnap, FTAG);
4386 dsl_dataset_rele(tosnap, FTAG);
4387 dsl_pool_rele(dp, FTAG);
4388 } else {
4389 file_t *fp = getf(zc->zc_cookie);
4390 if (fp == NULL)
4391 return (SET_ERROR(EBADF));
4392
4393 off = fp->f_offset;
4394 error = dmu_send_obj(zc->zc_name, zc->zc_sendobj,
4395 zc->zc_fromobj, embedok, large_block_ok,
4396 zc->zc_cookie, fp->f_vnode, &off);
4397
4398 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
4399 fp->f_offset = off;
4400 releasef(zc->zc_cookie);
4401 }
4402 return (error);
4403 }
4404
4405 /*
4406 * inputs:
4407 * zc_name name of snapshot on which to report progress
4408 * zc_cookie file descriptor of send stream
4409 *
4410 * outputs:
4411 * zc_cookie number of bytes written in send stream thus far
4412 */
4413 static int
4414 zfs_ioc_send_progress(zfs_cmd_t *zc)
4415 {
4416 dsl_pool_t *dp;
4417 dsl_dataset_t *ds;
4418 dmu_sendarg_t *dsp = NULL;
4419 int error;
4420
4421 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
4422 if (error != 0)
4423 return (error);
4424
4425 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &ds);
4426 if (error != 0) {
4427 dsl_pool_rele(dp, FTAG);
4428 return (error);
4429 }
4430
4431 mutex_enter(&ds->ds_sendstream_lock);
4432
4433 /*
4434 * Iterate over all the send streams currently active on this dataset.
4435 * If there's one which matches the specified file descriptor _and_ the
4436 * stream was started by the current process, return the progress of
4437 * that stream.
4438 */
4439 for (dsp = list_head(&ds->ds_sendstreams); dsp != NULL;
4440 dsp = list_next(&ds->ds_sendstreams, dsp)) {
4441 if (dsp->dsa_outfd == zc->zc_cookie &&
4442 dsp->dsa_proc == curproc)
4443 break;
4444 }
4445
4446 if (dsp != NULL)
4447 zc->zc_cookie = *(dsp->dsa_off);
4448 else
4449 error = SET_ERROR(ENOENT);
4450
4451 mutex_exit(&ds->ds_sendstream_lock);
4452 dsl_dataset_rele(ds, FTAG);
4453 dsl_pool_rele(dp, FTAG);
4454 return (error);
4455 }
4456
4457 static int
4458 zfs_ioc_inject_fault(zfs_cmd_t *zc)
4459 {
4460 int id, error;
4461
4462 error = zio_inject_fault(zc->zc_name, (int)zc->zc_guid, &id,
4463 &zc->zc_inject_record);
4464
4465 if (error == 0)
4466 zc->zc_guid = (uint64_t)id;
4467
4468 return (error);
4469 }
4470
4471 static int
4472 zfs_ioc_clear_fault(zfs_cmd_t *zc)
4473 {
4474 return (zio_clear_fault((int)zc->zc_guid));
4475 }
4476
4477 static int
4478 zfs_ioc_inject_list_next(zfs_cmd_t *zc)
4479 {
4480 int id = (int)zc->zc_guid;
4481 int error;
4482
4483 error = zio_inject_list_next(&id, zc->zc_name, sizeof (zc->zc_name),
4484 &zc->zc_inject_record);
4485
4486 zc->zc_guid = id;
4487
4488 return (error);
4489 }
4490
4491 static int
4492 zfs_ioc_error_log(zfs_cmd_t *zc)
4493 {
4494 spa_t *spa;
4495 int error;
4496 size_t count = (size_t)zc->zc_nvlist_dst_size;
4497
4498 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
4499 return (error);
4500
4501 error = spa_get_errlog(spa, (void *)(uintptr_t)zc->zc_nvlist_dst,
4502 &count);
4503 if (error == 0)
4504 zc->zc_nvlist_dst_size = count;
4505 else
4506 zc->zc_nvlist_dst_size = spa_get_errlog_size(spa);
4507
4508 spa_close(spa, FTAG);
4509
4510 return (error);
4511 }
4512
4513 static int
4514 zfs_ioc_clear(zfs_cmd_t *zc)
4515 {
4516 spa_t *spa;
4517 vdev_t *vd;
4518 int error;
4519
4520 /*
4521 * On zpool clear we also fix up missing slogs
4522 */
4523 mutex_enter(&spa_namespace_lock);
4524 spa = spa_lookup(zc->zc_name);
4525 if (spa == NULL) {
4526 mutex_exit(&spa_namespace_lock);
4527 return (SET_ERROR(EIO));
4528 }
4529 if (spa_get_log_state(spa) == SPA_LOG_MISSING) {
4530 /* we need to let spa_open/spa_load clear the chains */
4531 spa_set_log_state(spa, SPA_LOG_CLEAR);
4532 }
4533 spa->spa_last_open_failed = 0;
4534 mutex_exit(&spa_namespace_lock);
4535
4536 if (zc->zc_cookie & ZPOOL_NO_REWIND) {
4537 error = spa_open(zc->zc_name, &spa, FTAG);
4538 } else {
4539 nvlist_t *policy;
4540 nvlist_t *config = NULL;
4541
4542 if (zc->zc_nvlist_src == NULL)
4543 return (SET_ERROR(EINVAL));
4544
4545 if ((error = get_nvlist(zc->zc_nvlist_src,
4546 zc->zc_nvlist_src_size, zc->zc_iflags, &policy)) == 0) {
4547 error = spa_open_rewind(zc->zc_name, &spa, FTAG,
4548 policy, &config);
4549 if (config != NULL) {
4550 int err;
4551
4552 if ((err = put_nvlist(zc, config)) != 0)
4553 error = err;
4554 nvlist_free(config);
4555 }
4556 nvlist_free(policy);
4557 }
4558 }
4559
4560 if (error != 0)
4561 return (error);
4562
4563 spa_vdev_state_enter(spa, SCL_NONE);
4564
4565 if (zc->zc_guid == 0) {
4566 vd = NULL;
4567 } else {
4568 vd = spa_lookup_by_guid(spa, zc->zc_guid, B_TRUE);
4569 if (vd == NULL) {
4570 (void) spa_vdev_state_exit(spa, NULL, ENODEV);
4571 spa_close(spa, FTAG);
4572 return (SET_ERROR(ENODEV));
4573 }
4574 }
4575
4576 vdev_clear(spa, vd);
4577
4578 (void) spa_vdev_state_exit(spa, NULL, 0);
4579
4580 /*
4581 * Resume any suspended I/Os.
4582 */
4583 if (zio_resume(spa) != 0)
4584 error = SET_ERROR(EIO);
4585
4586 spa_close(spa, FTAG);
4587
4588 return (error);
4589 }
4590
4591 static int
4592 zfs_ioc_pool_reopen(zfs_cmd_t *zc)
4593 {
4594 spa_t *spa;
4595 int error;
4596
4597 error = spa_open(zc->zc_name, &spa, FTAG);
4598 if (error != 0)
4599 return (error);
4600
4601 spa_vdev_state_enter(spa, SCL_NONE);
4602
4603 /*
4604 * If a resilver is already in progress then set the
4605 * spa_scrub_reopen flag to B_TRUE so that we don't restart
4606 * the scan as a side effect of the reopen. Otherwise, let
4607 * vdev_open() decided if a resilver is required.
4608 */
4609 spa->spa_scrub_reopen = dsl_scan_resilvering(spa->spa_dsl_pool);
4610 vdev_reopen(spa->spa_root_vdev);
4611 spa->spa_scrub_reopen = B_FALSE;
4612
4613 (void) spa_vdev_state_exit(spa, NULL, 0);
4614 spa_close(spa, FTAG);
4615 return (0);
4616 }
4617 /*
4618 * inputs:
4619 * zc_name name of filesystem
4620 * zc_value name of origin snapshot
4621 *
4622 * outputs:
4623 * zc_string name of conflicting snapshot, if there is one
4624 */
4625 static int
4626 zfs_ioc_promote(zfs_cmd_t *zc)
4627 {
4628 char *cp;
4629
4630 /*
4631 * We don't need to unmount *all* the origin fs's snapshots, but
4632 * it's easier.
4633 */
4634 cp = strchr(zc->zc_value, '@');
4635 if (cp)
4636 *cp = '\0';
4637 (void) dmu_objset_find(zc->zc_value,
4638 zfs_unmount_snap_cb, NULL, DS_FIND_SNAPSHOTS);
4639 return (dsl_dataset_promote(zc->zc_name, zc->zc_string));
4640 }
4641
4642 /*
4643 * Retrieve a single {user|group}{used|quota}@... property.
4644 *
4645 * inputs:
4646 * zc_name name of filesystem
4647 * zc_objset_type zfs_userquota_prop_t
4648 * zc_value domain name (eg. "S-1-234-567-89")
4649 * zc_guid RID/UID/GID
4650 *
4651 * outputs:
4652 * zc_cookie property value
4653 */
4654 static int
4655 zfs_ioc_userspace_one(zfs_cmd_t *zc)
4656 {
4657 zfsvfs_t *zfsvfs;
4658 int error;
4659
4660 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
4661 return (SET_ERROR(EINVAL));
4662
4663 error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE);
4664 if (error != 0)
4665 return (error);
4666
4667 error = zfs_userspace_one(zfsvfs,
4668 zc->zc_objset_type, zc->zc_value, zc->zc_guid, &zc->zc_cookie);
4669 zfsvfs_rele(zfsvfs, FTAG);
4670
4671 return (error);
4672 }
4673
4674 /*
4675 * inputs:
4676 * zc_name name of filesystem
4677 * zc_cookie zap cursor
4678 * zc_objset_type zfs_userquota_prop_t
4679 * zc_nvlist_dst[_size] buffer to fill (not really an nvlist)
4680 *
4681 * outputs:
4682 * zc_nvlist_dst[_size] data buffer (array of zfs_useracct_t)
4683 * zc_cookie zap cursor
4684 */
4685 static int
4686 zfs_ioc_userspace_many(zfs_cmd_t *zc)
4687 {
4688 zfsvfs_t *zfsvfs;
4689 int bufsize = zc->zc_nvlist_dst_size;
4690
4691 if (bufsize <= 0)
4692 return (SET_ERROR(ENOMEM));
4693
4694 int error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE);
4695 if (error != 0)
4696 return (error);
4697
4698 void *buf = kmem_alloc(bufsize, KM_SLEEP);
4699
4700 error = zfs_userspace_many(zfsvfs, zc->zc_objset_type, &zc->zc_cookie,
4701 buf, &zc->zc_nvlist_dst_size);
4702
4703 if (error == 0) {
4704 error = xcopyout(buf,
4705 (void *)(uintptr_t)zc->zc_nvlist_dst,
4706 zc->zc_nvlist_dst_size);
4707 }
4708 kmem_free(buf, bufsize);
4709 zfsvfs_rele(zfsvfs, FTAG);
4710
4711 return (error);
4712 }
4713
4714 /*
4715 * inputs:
4716 * zc_name name of filesystem
4717 *
4718 * outputs:
4719 * none
4720 */
4721 static int
4722 zfs_ioc_userspace_upgrade(zfs_cmd_t *zc)
4723 {
4724 objset_t *os;
4725 int error = 0;
4726 zfsvfs_t *zfsvfs;
4727
4728 if (getzfsvfs(zc->zc_name, &zfsvfs) == 0) {
4729 if (!dmu_objset_userused_enabled(zfsvfs->z_os)) {
4730 /*
4731 * If userused is not enabled, it may be because the
4732 * objset needs to be closed & reopened (to grow the
4733 * objset_phys_t). Suspend/resume the fs will do that.
4734 */
4735 error = zfs_suspend_fs(zfsvfs);
4736 if (error == 0) {
4737 dmu_objset_refresh_ownership(zfsvfs->z_os,
4738 zfsvfs);
4739 error = zfs_resume_fs(zfsvfs, zc->zc_name);
4740 }
4741 }
4742 if (error == 0)
4743 error = dmu_objset_userspace_upgrade(zfsvfs->z_os);
4744 VFS_RELE(zfsvfs->z_vfs);
4745 } else {
4746 /* XXX kind of reading contents without owning */
4747 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
4748 if (error != 0)
4749 return (error);
4750
4751 error = dmu_objset_userspace_upgrade(os);
4752 dmu_objset_rele(os, FTAG);
4753 }
4754
4755 return (error);
4756 }
4757
4758 /*
4759 * We don't want to have a hard dependency
4760 * against some special symbols in sharefs
4761 * nfs, and smbsrv. Determine them if needed when
4762 * the first file system is shared.
4763 * Neither sharefs, nfs or smbsrv are unloadable modules.
4764 */
4765 int (*znfsexport_fs)(void *arg);
4766 int (*zshare_fs)(enum sharefs_sys_op, share_t *, uint32_t);
4767 int (*zsmbexport_fs)(void *arg, boolean_t add_share);
4768
4769 int zfs_nfsshare_inited;
4770 int zfs_smbshare_inited;
4771
4772 ddi_modhandle_t nfs_mod;
4773 ddi_modhandle_t sharefs_mod;
4774 ddi_modhandle_t smbsrv_mod;
4775 kmutex_t zfs_share_lock;
4776
4777 static int
4778 zfs_init_sharefs()
4779 {
4780 int error;
4781
4782 ASSERT(MUTEX_HELD(&zfs_share_lock));
4783 /* Both NFS and SMB shares also require sharetab support. */
4784 if (sharefs_mod == NULL && ((sharefs_mod =
4785 ddi_modopen("fs/sharefs",
4786 KRTLD_MODE_FIRST, &error)) == NULL)) {
4787 return (SET_ERROR(ENOSYS));
4788 }
4789 if (zshare_fs == NULL && ((zshare_fs =
4790 (int (*)(enum sharefs_sys_op, share_t *, uint32_t))
4791 ddi_modsym(sharefs_mod, "sharefs_impl", &error)) == NULL)) {
4792 return (SET_ERROR(ENOSYS));
4793 }
4794 return (0);
4795 }
4796
4797 static int
4798 zfs_ioc_share(zfs_cmd_t *zc)
4799 {
4800 int error;
4801 int opcode;
4802
4803 switch (zc->zc_share.z_sharetype) {
4804 case ZFS_SHARE_NFS:
4805 case ZFS_UNSHARE_NFS:
4806 if (zfs_nfsshare_inited == 0) {
4807 mutex_enter(&zfs_share_lock);
4808 if (nfs_mod == NULL && ((nfs_mod = ddi_modopen("fs/nfs",
4809 KRTLD_MODE_FIRST, &error)) == NULL)) {
4810 mutex_exit(&zfs_share_lock);
4811 return (SET_ERROR(ENOSYS));
4812 }
4813 if (znfsexport_fs == NULL &&
4814 ((znfsexport_fs = (int (*)(void *))
4815 ddi_modsym(nfs_mod,
4816 "nfs_export", &error)) == NULL)) {
4817 mutex_exit(&zfs_share_lock);
4818 return (SET_ERROR(ENOSYS));
4819 }
4820 error = zfs_init_sharefs();
4821 if (error != 0) {
4822 mutex_exit(&zfs_share_lock);
4823 return (SET_ERROR(ENOSYS));
4824 }
4825 zfs_nfsshare_inited = 1;
4826 mutex_exit(&zfs_share_lock);
4827 }
4828 break;
4829 case ZFS_SHARE_SMB:
4830 case ZFS_UNSHARE_SMB:
4831 if (zfs_smbshare_inited == 0) {
4832 mutex_enter(&zfs_share_lock);
4833 if (smbsrv_mod == NULL && ((smbsrv_mod =
4834 ddi_modopen("drv/smbsrv",
4835 KRTLD_MODE_FIRST, &error)) == NULL)) {
4836 mutex_exit(&zfs_share_lock);
4837 return (SET_ERROR(ENOSYS));
4838 }
4839 if (zsmbexport_fs == NULL && ((zsmbexport_fs =
4840 (int (*)(void *, boolean_t))ddi_modsym(smbsrv_mod,
4841 "smb_server_share", &error)) == NULL)) {
4842 mutex_exit(&zfs_share_lock);
4843 return (SET_ERROR(ENOSYS));
4844 }
4845 error = zfs_init_sharefs();
4846 if (error != 0) {
4847 mutex_exit(&zfs_share_lock);
4848 return (SET_ERROR(ENOSYS));
4849 }
4850 zfs_smbshare_inited = 1;
4851 mutex_exit(&zfs_share_lock);
4852 }
4853 break;
4854 default:
4855 return (SET_ERROR(EINVAL));
4856 }
4857
4858 switch (zc->zc_share.z_sharetype) {
4859 case ZFS_SHARE_NFS:
4860 case ZFS_UNSHARE_NFS:
4861 if (error =
4862 znfsexport_fs((void *)
4863 (uintptr_t)zc->zc_share.z_exportdata))
4864 return (error);
4865 break;
4866 case ZFS_SHARE_SMB:
4867 case ZFS_UNSHARE_SMB:
4868 if (error = zsmbexport_fs((void *)
4869 (uintptr_t)zc->zc_share.z_exportdata,
4870 zc->zc_share.z_sharetype == ZFS_SHARE_SMB ?
4871 B_TRUE: B_FALSE)) {
4872 return (error);
4873 }
4874 break;
4875 }
4876
4877 opcode = (zc->zc_share.z_sharetype == ZFS_SHARE_NFS ||
4878 zc->zc_share.z_sharetype == ZFS_SHARE_SMB) ?
4879 SHAREFS_ADD : SHAREFS_REMOVE;
4880
4881 /*
4882 * Add or remove share from sharetab
4883 */
4884 error = zshare_fs(opcode,
4885 (void *)(uintptr_t)zc->zc_share.z_sharedata,
4886 zc->zc_share.z_sharemax);
4887
4888 return (error);
4889
4890 }
4891
4892 ace_t full_access[] = {
4893 {(uid_t)-1, ACE_ALL_PERMS, ACE_EVERYONE, 0}
4894 };
4895
4896 /*
4897 * inputs:
4898 * zc_name name of containing filesystem
4899 * zc_obj object # beyond which we want next in-use object #
4900 *
4901 * outputs:
4902 * zc_obj next in-use object #
4903 */
4904 static int
4905 zfs_ioc_next_obj(zfs_cmd_t *zc)
4906 {
4907 objset_t *os = NULL;
4908 int error;
4909
4910 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
4911 if (error != 0)
4912 return (error);
4913
4914 error = dmu_object_next(os, &zc->zc_obj, B_FALSE,
4915 dsl_dataset_phys(os->os_dsl_dataset)->ds_prev_snap_txg);
4916
4917 dmu_objset_rele(os, FTAG);
4918 return (error);
4919 }
4920
4921 /*
4922 * inputs:
4923 * zc_name name of filesystem
4924 * zc_value prefix name for snapshot
4925 * zc_cleanup_fd cleanup-on-exit file descriptor for calling process
4926 *
4927 * outputs:
4928 * zc_value short name of new snapshot
4929 */
4930 static int
4931 zfs_ioc_tmp_snapshot(zfs_cmd_t *zc)
4932 {
4933 char *snap_name;
4934 char *hold_name;
4935 int error;
4936 minor_t minor;
4937
4938 error = zfs_onexit_fd_hold(zc->zc_cleanup_fd, &minor);
4939 if (error != 0)
4940 return (error);
4941
4942 snap_name = kmem_asprintf("%s-%016llx", zc->zc_value,
4943 (u_longlong_t)ddi_get_lbolt64());
4944 hold_name = kmem_asprintf("%%%s", zc->zc_value);
4945
4946 error = dsl_dataset_snapshot_tmp(zc->zc_name, snap_name, minor,
4947 hold_name);
4948 if (error == 0)
4949 (void) strcpy(zc->zc_value, snap_name);
4950 strfree(snap_name);
4951 strfree(hold_name);
4952 zfs_onexit_fd_rele(zc->zc_cleanup_fd);
4953 return (error);
4954 }
4955
4956 /*
4957 * inputs:
4958 * zc_name name of "to" snapshot
4959 * zc_value name of "from" snapshot
4960 * zc_cookie file descriptor to write diff data on
4961 *
4962 * outputs:
4963 * dmu_diff_record_t's to the file descriptor
4964 */
4965 static int
4966 zfs_ioc_diff(zfs_cmd_t *zc)
4967 {
4968 file_t *fp;
4969 offset_t off;
4970 int error;
4971
4972 fp = getf(zc->zc_cookie);
4973 if (fp == NULL)
4974 return (SET_ERROR(EBADF));
4975
4976 off = fp->f_offset;
4977
4978 error = dmu_diff(zc->zc_name, zc->zc_value, fp->f_vnode, &off);
4979
4980 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
4981 fp->f_offset = off;
4982 releasef(zc->zc_cookie);
4983
4984 return (error);
4985 }
4986
4987 /*
4988 * Remove all ACL files in shares dir
4989 */
4990 static int
4991 zfs_smb_acl_purge(znode_t *dzp)
4992 {
4993 zap_cursor_t zc;
4994 zap_attribute_t zap;
4995 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
4996 int error;
4997
4998 for (zap_cursor_init(&zc, zfsvfs->z_os, dzp->z_id);
4999 (error = zap_cursor_retrieve(&zc, &zap)) == 0;
5000 zap_cursor_advance(&zc)) {
5001 if ((error = VOP_REMOVE(ZTOV(dzp), zap.za_name, kcred,
5002 NULL, 0)) != 0)
5003 break;
5004 }
5005 zap_cursor_fini(&zc);
5006 return (error);
5007 }
5008
5009 static int
5010 zfs_ioc_smb_acl(zfs_cmd_t *zc)
5011 {
5012 vnode_t *vp;
5013 znode_t *dzp;
5014 vnode_t *resourcevp = NULL;
5015 znode_t *sharedir;
5016 zfsvfs_t *zfsvfs;
5017 nvlist_t *nvlist;
5018 char *src, *target;
5019 vattr_t vattr;
5020 vsecattr_t vsec;
5021 int error = 0;
5022
5023 if ((error = lookupname(zc->zc_value, UIO_SYSSPACE,
5024 NO_FOLLOW, NULL, &vp)) != 0)
5025 return (error);
5026
5027 /* Now make sure mntpnt and dataset are ZFS */
5028
5029 if (vp->v_vfsp->vfs_fstype != zfsfstype ||
5030 (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource),
5031 zc->zc_name) != 0)) {
5032 VN_RELE(vp);
5033 return (SET_ERROR(EINVAL));
5034 }
5035
5036 dzp = VTOZ(vp);
5037 zfsvfs = dzp->z_zfsvfs;
5038 ZFS_ENTER(zfsvfs);
5039
5040 /*
5041 * Create share dir if its missing.
5042 */
5043 mutex_enter(&zfsvfs->z_lock);
5044 if (zfsvfs->z_shares_dir == 0) {
5045 dmu_tx_t *tx;
5046
5047 tx = dmu_tx_create(zfsvfs->z_os);
5048 dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, TRUE,
5049 ZFS_SHARES_DIR);
5050 dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL);
5051 error = dmu_tx_assign(tx, TXG_WAIT);
5052 if (error != 0) {
5053 dmu_tx_abort(tx);
5054 } else {
5055 error = zfs_create_share_dir(zfsvfs, tx);
5056 dmu_tx_commit(tx);
5057 }
5058 if (error != 0) {
5059 mutex_exit(&zfsvfs->z_lock);
5060 VN_RELE(vp);
5061 ZFS_EXIT(zfsvfs);
5062 return (error);
5063 }
5064 }
5065 mutex_exit(&zfsvfs->z_lock);
5066
5067 ASSERT(zfsvfs->z_shares_dir);
5068 if ((error = zfs_zget(zfsvfs, zfsvfs->z_shares_dir, &sharedir)) != 0) {
5069 VN_RELE(vp);
5070 ZFS_EXIT(zfsvfs);
5071 return (error);
5072 }
5073
5074 switch (zc->zc_cookie) {
5075 case ZFS_SMB_ACL_ADD:
5076 vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE;
5077 vattr.va_type = VREG;
5078 vattr.va_mode = S_IFREG|0777;
5079 vattr.va_uid = 0;
5080 vattr.va_gid = 0;
5081
5082 vsec.vsa_mask = VSA_ACE;
5083 vsec.vsa_aclentp = &full_access;
5084 vsec.vsa_aclentsz = sizeof (full_access);
5085 vsec.vsa_aclcnt = 1;
5086
5087 error = VOP_CREATE(ZTOV(sharedir), zc->zc_string,
5088 &vattr, EXCL, 0, &resourcevp, kcred, 0, NULL, &vsec);
5089 if (resourcevp)
5090 VN_RELE(resourcevp);
5091 break;
5092
5093 case ZFS_SMB_ACL_REMOVE:
5094 error = VOP_REMOVE(ZTOV(sharedir), zc->zc_string, kcred,
5095 NULL, 0);
5096 break;
5097
5098 case ZFS_SMB_ACL_RENAME:
5099 if ((error = get_nvlist(zc->zc_nvlist_src,
5100 zc->zc_nvlist_src_size, zc->zc_iflags, &nvlist)) != 0) {
5101 VN_RELE(vp);
5102 VN_RELE(ZTOV(sharedir));
5103 ZFS_EXIT(zfsvfs);
5104 return (error);
5105 }
5106 if (nvlist_lookup_string(nvlist, ZFS_SMB_ACL_SRC, &src) ||
5107 nvlist_lookup_string(nvlist, ZFS_SMB_ACL_TARGET,
5108 &target)) {
5109 VN_RELE(vp);
5110 VN_RELE(ZTOV(sharedir));
5111 ZFS_EXIT(zfsvfs);
5112 nvlist_free(nvlist);
5113 return (error);
5114 }
5115 error = VOP_RENAME(ZTOV(sharedir), src, ZTOV(sharedir), target,
5116 kcred, NULL, 0);
5117 nvlist_free(nvlist);
5118 break;
5119
5120 case ZFS_SMB_ACL_PURGE:
5121 error = zfs_smb_acl_purge(sharedir);
5122 break;
5123
5124 default:
5125 error = SET_ERROR(EINVAL);
5126 break;
5127 }
5128
5129 VN_RELE(vp);
5130 VN_RELE(ZTOV(sharedir));
5131
5132 ZFS_EXIT(zfsvfs);
5133
5134 return (error);
5135 }
5136
5137 /*
5138 * innvl: {
5139 * "holds" -> { snapname -> holdname (string), ... }
5140 * (optional) "cleanup_fd" -> fd (int32)
5141 * }
5142 *
5143 * outnvl: {
5144 * snapname -> error value (int32)
5145 * ...
5146 * }
5147 */
5148 /* ARGSUSED */
5149 static int
5150 zfs_ioc_hold(const char *pool, nvlist_t *args, nvlist_t *errlist)
5151 {
5152 nvpair_t *pair;
5153 nvlist_t *holds;
5154 int cleanup_fd = -1;
5155 int error;
5156 minor_t minor = 0;
5157
5158 error = nvlist_lookup_nvlist(args, "holds", &holds);
5159 if (error != 0)
5160 return (SET_ERROR(EINVAL));
5161
5162 /* make sure the user didn't pass us any invalid (empty) tags */
5163 for (pair = nvlist_next_nvpair(holds, NULL); pair != NULL;
5164 pair = nvlist_next_nvpair(holds, pair)) {
5165 char *htag;
5166
5167 error = nvpair_value_string(pair, &htag);
5168 if (error != 0)
5169 return (SET_ERROR(error));
5170
5171 if (strlen(htag) == 0)
5172 return (SET_ERROR(EINVAL));
5173 }
5174
5175 if (nvlist_lookup_int32(args, "cleanup_fd", &cleanup_fd) == 0) {
5176 error = zfs_onexit_fd_hold(cleanup_fd, &minor);
5177 if (error != 0)
5178 return (error);
5179 }
5180
5181 error = dsl_dataset_user_hold(holds, minor, errlist);
5182 if (minor != 0)
5183 zfs_onexit_fd_rele(cleanup_fd);
5184 return (error);
5185 }
5186
5187 /*
5188 * innvl is not used.
5189 *
5190 * outnvl: {
5191 * holdname -> time added (uint64 seconds since epoch)
5192 * ...
5193 * }
5194 */
5195 /* ARGSUSED */
5196 static int
5197 zfs_ioc_get_holds(const char *snapname, nvlist_t *args, nvlist_t *outnvl)
5198 {
5199 return (dsl_dataset_get_holds(snapname, outnvl));
5200 }
5201
5202 /*
5203 * innvl: {
5204 * snapname -> { holdname, ... }
5205 * ...
5206 * }
5207 *
5208 * outnvl: {
5209 * snapname -> error value (int32)
5210 * ...
5211 * }
5212 */
5213 /* ARGSUSED */
5214 static int
5215 zfs_ioc_release(const char *pool, nvlist_t *holds, nvlist_t *errlist)
5216 {
5217 return (dsl_dataset_user_release(holds, errlist));
5218 }
5219
5220 /*
5221 * inputs:
5222 * zc_name name of new filesystem or snapshot
5223 * zc_value full name of old snapshot
5224 *
5225 * outputs:
5226 * zc_cookie space in bytes
5227 * zc_objset_type compressed space in bytes
5228 * zc_perm_action uncompressed space in bytes
5229 */
5230 static int
5231 zfs_ioc_space_written(zfs_cmd_t *zc)
5232 {
5233 int error;
5234 dsl_pool_t *dp;
5235 dsl_dataset_t *new, *old;
5236
5237 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
5238 if (error != 0)
5239 return (error);
5240 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &new);
5241 if (error != 0) {
5242 dsl_pool_rele(dp, FTAG);
5243 return (error);
5244 }
5245 error = dsl_dataset_hold(dp, zc->zc_value, FTAG, &old);
5246 if (error != 0) {
5247 dsl_dataset_rele(new, FTAG);
5248 dsl_pool_rele(dp, FTAG);
5249 return (error);
5250 }
5251
5252 error = dsl_dataset_space_written(old, new, &zc->zc_cookie,
5253 &zc->zc_objset_type, &zc->zc_perm_action);
5254 dsl_dataset_rele(old, FTAG);
5255 dsl_dataset_rele(new, FTAG);
5256 dsl_pool_rele(dp, FTAG);
5257 return (error);
5258 }
5259
5260 /*
5261 * innvl: {
5262 * "firstsnap" -> snapshot name
5263 * }
5264 *
5265 * outnvl: {
5266 * "used" -> space in bytes
5267 * "compressed" -> compressed space in bytes
5268 * "uncompressed" -> uncompressed space in bytes
5269 * }
5270 */
5271 static int
5272 zfs_ioc_space_snaps(const char *lastsnap, nvlist_t *innvl, nvlist_t *outnvl)
5273 {
5274 int error;
5275 dsl_pool_t *dp;
5276 dsl_dataset_t *new, *old;
5277 char *firstsnap;
5278 uint64_t used, comp, uncomp;
5279
5280 if (nvlist_lookup_string(innvl, "firstsnap", &firstsnap) != 0)
5281 return (SET_ERROR(EINVAL));
5282
5283 error = dsl_pool_hold(lastsnap, FTAG, &dp);
5284 if (error != 0)
5285 return (error);
5286
5287 error = dsl_dataset_hold(dp, lastsnap, FTAG, &new);
5288 if (error == 0 && !new->ds_is_snapshot) {
5289 dsl_dataset_rele(new, FTAG);
5290 error = SET_ERROR(EINVAL);
5291 }
5292 if (error != 0) {
5293 dsl_pool_rele(dp, FTAG);
5294 return (error);
5295 }
5296 error = dsl_dataset_hold(dp, firstsnap, FTAG, &old);
5297 if (error == 0 && !old->ds_is_snapshot) {
5298 dsl_dataset_rele(old, FTAG);
5299 error = SET_ERROR(EINVAL);
5300 }
5301 if (error != 0) {
5302 dsl_dataset_rele(new, FTAG);
5303 dsl_pool_rele(dp, FTAG);
5304 return (error);
5305 }
5306
5307 error = dsl_dataset_space_wouldfree(old, new, &used, &comp, &uncomp);
5308 dsl_dataset_rele(old, FTAG);
5309 dsl_dataset_rele(new, FTAG);
5310 dsl_pool_rele(dp, FTAG);
5311 fnvlist_add_uint64(outnvl, "used", used);
5312 fnvlist_add_uint64(outnvl, "compressed", comp);
5313 fnvlist_add_uint64(outnvl, "uncompressed", uncomp);
5314 return (error);
5315 }
5316
5317 /*
5318 * innvl: {
5319 * "fd" -> file descriptor to write stream to (int32)
5320 * (optional) "fromsnap" -> full snap name to send an incremental from
5321 * (optional) "largeblockok" -> (value ignored)
5322 * indicates that blocks > 128KB are permitted
5323 * (optional) "embedok" -> (value ignored)
5324 * presence indicates DRR_WRITE_EMBEDDED records are permitted
5325 * (optional) "resume_object" and "resume_offset" -> (uint64)
5326 * if present, resume send stream from specified object and offset.
5327 * }
5328 *
5329 * outnvl is unused
5330 */
5331 /* ARGSUSED */
5332 static int
5333 zfs_ioc_send_new(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl)
5334 {
5335 int error;
5336 offset_t off;
5337 char *fromname = NULL;
5338 int fd;
5339 boolean_t largeblockok;
5340 boolean_t embedok;
5341 uint64_t resumeobj = 0;
5342 uint64_t resumeoff = 0;
5343
5344 error = nvlist_lookup_int32(innvl, "fd", &fd);
5345 if (error != 0)
5346 return (SET_ERROR(EINVAL));
5347
5348 (void) nvlist_lookup_string(innvl, "fromsnap", &fromname);
5349
5350 largeblockok = nvlist_exists(innvl, "largeblockok");
5351 embedok = nvlist_exists(innvl, "embedok");
5352
5353 (void) nvlist_lookup_uint64(innvl, "resume_object", &resumeobj);
5354 (void) nvlist_lookup_uint64(innvl, "resume_offset", &resumeoff);
5355
5356 file_t *fp = getf(fd);
5357 if (fp == NULL)
5358 return (SET_ERROR(EBADF));
5359
5360 off = fp->f_offset;
5361 error = dmu_send(snapname, fromname, embedok, largeblockok, fd,
5362 resumeobj, resumeoff, fp->f_vnode, &off);
5363
5364 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
5365 fp->f_offset = off;
5366 releasef(fd);
5367 return (error);
5368 }
5369
5370 /*
5371 * Determine approximately how large a zfs send stream will be -- the number
5372 * of bytes that will be written to the fd supplied to zfs_ioc_send_new().
5373 *
5374 * innvl: {
5375 * (optional) "from" -> full snap or bookmark name to send an incremental
5376 * from
5377 * }
5378 *
5379 * outnvl: {
5380 * "space" -> bytes of space (uint64)
5381 * }
5382 */
5383 static int
5384 zfs_ioc_send_space(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl)
5385 {
5386 dsl_pool_t *dp;
5387 dsl_dataset_t *tosnap;
5388 int error;
5389 char *fromname;
5390 uint64_t space;
5391
5392 error = dsl_pool_hold(snapname, FTAG, &dp);
5393 if (error != 0)
5394 return (error);
5395
5396 error = dsl_dataset_hold(dp, snapname, FTAG, &tosnap);
5397 if (error != 0) {
5398 dsl_pool_rele(dp, FTAG);
5399 return (error);
5400 }
5401
5402 error = nvlist_lookup_string(innvl, "from", &fromname);
5403 if (error == 0) {
5404 if (strchr(fromname, '@') != NULL) {
5405 /*
5406 * If from is a snapshot, hold it and use the more
5407 * efficient dmu_send_estimate to estimate send space
5408 * size using deadlists.
5409 */
5410 dsl_dataset_t *fromsnap;
5411 error = dsl_dataset_hold(dp, fromname, FTAG, &fromsnap);
5412 if (error != 0)
5413 goto out;
5414 error = dmu_send_estimate(tosnap, fromsnap, &space);
5415 dsl_dataset_rele(fromsnap, FTAG);
5416 } else if (strchr(fromname, '#') != NULL) {
5417 /*
5418 * If from is a bookmark, fetch the creation TXG of the
5419 * snapshot it was created from and use that to find
5420 * blocks that were born after it.
5421 */
5422 zfs_bookmark_phys_t frombm;
5423
5424 error = dsl_bookmark_lookup(dp, fromname, tosnap,
5425 &frombm);
5426 if (error != 0)
5427 goto out;
5428 error = dmu_send_estimate_from_txg(tosnap,
5429 frombm.zbm_creation_txg, &space);
5430 } else {
5431 /*
5432 * from is not properly formatted as a snapshot or
5433 * bookmark
5434 */
5435 error = SET_ERROR(EINVAL);
5436 goto out;
5437 }
5438 } else {
5439 // If estimating the size of a full send, use dmu_send_estimate
5440 error = dmu_send_estimate(tosnap, NULL, &space);
5441 }
5442
5443 fnvlist_add_uint64(outnvl, "space", space);
5444
5445 out:
5446 dsl_dataset_rele(tosnap, FTAG);
5447 dsl_pool_rele(dp, FTAG);
5448 return (error);
5449 }
5450
5451 static zfs_ioc_vec_t zfs_ioc_vec[ZFS_IOC_LAST - ZFS_IOC_FIRST];
5452
5453 static void
5454 zfs_ioctl_register_legacy(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5455 zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck,
5456 boolean_t log_history, zfs_ioc_poolcheck_t pool_check)
5457 {
5458 zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST];
5459
5460 ASSERT3U(ioc, >=, ZFS_IOC_FIRST);
5461 ASSERT3U(ioc, <, ZFS_IOC_LAST);
5462 ASSERT3P(vec->zvec_legacy_func, ==, NULL);
5463 ASSERT3P(vec->zvec_func, ==, NULL);
5464
5465 vec->zvec_legacy_func = func;
5466 vec->zvec_secpolicy = secpolicy;
5467 vec->zvec_namecheck = namecheck;
5468 vec->zvec_allow_log = log_history;
5469 vec->zvec_pool_check = pool_check;
5470 }
5471
5472 /*
5473 * See the block comment at the beginning of this file for details on
5474 * each argument to this function.
5475 */
5476 static void
5477 zfs_ioctl_register(const char *name, zfs_ioc_t ioc, zfs_ioc_func_t *func,
5478 zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck,
5479 zfs_ioc_poolcheck_t pool_check, boolean_t smush_outnvlist,
5480 boolean_t allow_log)
5481 {
5482 zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST];
5483
5484 ASSERT3U(ioc, >=, ZFS_IOC_FIRST);
5485 ASSERT3U(ioc, <, ZFS_IOC_LAST);
5486 ASSERT3P(vec->zvec_legacy_func, ==, NULL);
5487 ASSERT3P(vec->zvec_func, ==, NULL);
5488
5489 /* if we are logging, the name must be valid */
5490 ASSERT(!allow_log || namecheck != NO_NAME);
5491
5492 vec->zvec_name = name;
5493 vec->zvec_func = func;
5494 vec->zvec_secpolicy = secpolicy;
5495 vec->zvec_namecheck = namecheck;
5496 vec->zvec_pool_check = pool_check;
5497 vec->zvec_smush_outnvlist = smush_outnvlist;
5498 vec->zvec_allow_log = allow_log;
5499 }
5500
5501 static void
5502 zfs_ioctl_register_pool(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5503 zfs_secpolicy_func_t *secpolicy, boolean_t log_history,
5504 zfs_ioc_poolcheck_t pool_check)
5505 {
5506 zfs_ioctl_register_legacy(ioc, func, secpolicy,
5507 POOL_NAME, log_history, pool_check);
5508 }
5509
5510 static void
5511 zfs_ioctl_register_dataset_nolog(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5512 zfs_secpolicy_func_t *secpolicy, zfs_ioc_poolcheck_t pool_check)
5513 {
5514 zfs_ioctl_register_legacy(ioc, func, secpolicy,
5515 DATASET_NAME, B_FALSE, pool_check);
5516 }
5517
5518 static void
5519 zfs_ioctl_register_pool_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func)
5520 {
5521 zfs_ioctl_register_legacy(ioc, func, zfs_secpolicy_config,
5522 POOL_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5523 }
5524
5525 static void
5526 zfs_ioctl_register_pool_meta(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5527 zfs_secpolicy_func_t *secpolicy)
5528 {
5529 zfs_ioctl_register_legacy(ioc, func, secpolicy,
5530 NO_NAME, B_FALSE, POOL_CHECK_NONE);
5531 }
5532
5533 static void
5534 zfs_ioctl_register_dataset_read_secpolicy(zfs_ioc_t ioc,
5535 zfs_ioc_legacy_func_t *func, zfs_secpolicy_func_t *secpolicy)
5536 {
5537 zfs_ioctl_register_legacy(ioc, func, secpolicy,
5538 DATASET_NAME, B_FALSE, POOL_CHECK_SUSPENDED);
5539 }
5540
5541 static void
5542 zfs_ioctl_register_dataset_read(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func)
5543 {
5544 zfs_ioctl_register_dataset_read_secpolicy(ioc, func,
5545 zfs_secpolicy_read);
5546 }
5547
5548 static void
5549 zfs_ioctl_register_dataset_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5550 zfs_secpolicy_func_t *secpolicy)
5551 {
5552 zfs_ioctl_register_legacy(ioc, func, secpolicy,
5553 DATASET_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5554 }
5555
5556 static void
5557 zfs_ioctl_init(void)
5558 {
5559 zfs_ioctl_register("snapshot", ZFS_IOC_SNAPSHOT,
5560 zfs_ioc_snapshot, zfs_secpolicy_snapshot, POOL_NAME,
5561 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5562
5563 zfs_ioctl_register("log_history", ZFS_IOC_LOG_HISTORY,
5564 zfs_ioc_log_history, zfs_secpolicy_log_history, NO_NAME,
5565 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_FALSE);
5566
5567 zfs_ioctl_register("space_snaps", ZFS_IOC_SPACE_SNAPS,
5568 zfs_ioc_space_snaps, zfs_secpolicy_read, DATASET_NAME,
5569 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5570
5571 zfs_ioctl_register("send", ZFS_IOC_SEND_NEW,
5572 zfs_ioc_send_new, zfs_secpolicy_send_new, DATASET_NAME,
5573 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5574
5575 zfs_ioctl_register("send_space", ZFS_IOC_SEND_SPACE,
5576 zfs_ioc_send_space, zfs_secpolicy_read, DATASET_NAME,
5577 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5578
5579 zfs_ioctl_register("create", ZFS_IOC_CREATE,
5580 zfs_ioc_create, zfs_secpolicy_create_clone, DATASET_NAME,
5581 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5582
5583 zfs_ioctl_register("clone", ZFS_IOC_CLONE,
5584 zfs_ioc_clone, zfs_secpolicy_create_clone, DATASET_NAME,
5585 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5586
5587 zfs_ioctl_register("destroy_snaps", ZFS_IOC_DESTROY_SNAPS,
5588 zfs_ioc_destroy_snaps, zfs_secpolicy_destroy_snaps, POOL_NAME,
5589 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5590
5591 zfs_ioctl_register("hold", ZFS_IOC_HOLD,
5592 zfs_ioc_hold, zfs_secpolicy_hold, POOL_NAME,
5593 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5594 zfs_ioctl_register("release", ZFS_IOC_RELEASE,
5595 zfs_ioc_release, zfs_secpolicy_release, POOL_NAME,
5596 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5597
5598 zfs_ioctl_register("get_holds", ZFS_IOC_GET_HOLDS,
5599 zfs_ioc_get_holds, zfs_secpolicy_read, DATASET_NAME,
5600 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5601
5602 zfs_ioctl_register("rollback", ZFS_IOC_ROLLBACK,
5603 zfs_ioc_rollback, zfs_secpolicy_rollback, DATASET_NAME,
5604 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_TRUE);
5605
5606 zfs_ioctl_register("bookmark", ZFS_IOC_BOOKMARK,
5607 zfs_ioc_bookmark, zfs_secpolicy_bookmark, POOL_NAME,
5608 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5609
5610 zfs_ioctl_register("get_bookmarks", ZFS_IOC_GET_BOOKMARKS,
5611 zfs_ioc_get_bookmarks, zfs_secpolicy_read, DATASET_NAME,
5612 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5613
5614 zfs_ioctl_register("destroy_bookmarks", ZFS_IOC_DESTROY_BOOKMARKS,
5615 zfs_ioc_destroy_bookmarks, zfs_secpolicy_destroy_bookmarks,
5616 POOL_NAME,
5617 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5618
5619 /* IOCTLS that use the legacy function signature */
5620
5621 zfs_ioctl_register_legacy(ZFS_IOC_POOL_FREEZE, zfs_ioc_pool_freeze,
5622 zfs_secpolicy_config, NO_NAME, B_FALSE, POOL_CHECK_READONLY);
5623
5624 zfs_ioctl_register_pool(ZFS_IOC_POOL_CREATE, zfs_ioc_pool_create,
5625 zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE);
5626 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SCAN,
5627 zfs_ioc_pool_scan);
5628 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_UPGRADE,
5629 zfs_ioc_pool_upgrade);
5630 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ADD,
5631 zfs_ioc_vdev_add);
5632 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_REMOVE,
5633 zfs_ioc_vdev_remove);
5634 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SET_STATE,
5635 zfs_ioc_vdev_set_state);
5636 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ATTACH,
5637 zfs_ioc_vdev_attach);
5638 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_DETACH,
5639 zfs_ioc_vdev_detach);
5640 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETPATH,
5641 zfs_ioc_vdev_setpath);
5642 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETFRU,
5643 zfs_ioc_vdev_setfru);
5644 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SET_PROPS,
5645 zfs_ioc_pool_set_props);
5646 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SPLIT,
5647 zfs_ioc_vdev_split);
5648 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_REGUID,
5649 zfs_ioc_pool_reguid);
5650
5651 zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_CONFIGS,
5652 zfs_ioc_pool_configs, zfs_secpolicy_none);
5653 zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_TRYIMPORT,
5654 zfs_ioc_pool_tryimport, zfs_secpolicy_config);
5655 zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_FAULT,
5656 zfs_ioc_inject_fault, zfs_secpolicy_inject);
5657 zfs_ioctl_register_pool_meta(ZFS_IOC_CLEAR_FAULT,
5658 zfs_ioc_clear_fault, zfs_secpolicy_inject);
5659 zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_LIST_NEXT,
5660 zfs_ioc_inject_list_next, zfs_secpolicy_inject);
5661
5662 /*
5663 * pool destroy, and export don't log the history as part of
5664 * zfsdev_ioctl, but rather zfs_ioc_pool_export
5665 * does the logging of those commands.
5666 */
5667 zfs_ioctl_register_pool(ZFS_IOC_POOL_DESTROY, zfs_ioc_pool_destroy,
5668 zfs_secpolicy_config, B_FALSE, POOL_CHECK_NONE);
5669 zfs_ioctl_register_pool(ZFS_IOC_POOL_EXPORT, zfs_ioc_pool_export,
5670 zfs_secpolicy_config, B_FALSE, POOL_CHECK_NONE);
5671
5672 zfs_ioctl_register_pool(ZFS_IOC_POOL_STATS, zfs_ioc_pool_stats,
5673 zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE);
5674 zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_PROPS, zfs_ioc_pool_get_props,
5675 zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE);
5676
5677 zfs_ioctl_register_pool(ZFS_IOC_ERROR_LOG, zfs_ioc_error_log,
5678 zfs_secpolicy_inject, B_FALSE, POOL_CHECK_SUSPENDED);
5679 zfs_ioctl_register_pool(ZFS_IOC_DSOBJ_TO_DSNAME,
5680 zfs_ioc_dsobj_to_dsname,
5681 zfs_secpolicy_diff, B_FALSE, POOL_CHECK_SUSPENDED);
5682 zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_HISTORY,
5683 zfs_ioc_pool_get_history,
5684 zfs_secpolicy_config, B_FALSE, POOL_CHECK_SUSPENDED);
5685
5686 zfs_ioctl_register_pool(ZFS_IOC_POOL_IMPORT, zfs_ioc_pool_import,
5687 zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE);
5688
5689 zfs_ioctl_register_pool(ZFS_IOC_CLEAR, zfs_ioc_clear,
5690 zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE);
5691 zfs_ioctl_register_pool(ZFS_IOC_POOL_REOPEN, zfs_ioc_pool_reopen,
5692 zfs_secpolicy_config, B_TRUE, POOL_CHECK_SUSPENDED);
5693
5694 zfs_ioctl_register_dataset_read(ZFS_IOC_SPACE_WRITTEN,
5695 zfs_ioc_space_written);
5696 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_RECVD_PROPS,
5697 zfs_ioc_objset_recvd_props);
5698 zfs_ioctl_register_dataset_read(ZFS_IOC_NEXT_OBJ,
5699 zfs_ioc_next_obj);
5700 zfs_ioctl_register_dataset_read(ZFS_IOC_GET_FSACL,
5701 zfs_ioc_get_fsacl);
5702 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_STATS,
5703 zfs_ioc_objset_stats);
5704 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_ZPLPROPS,
5705 zfs_ioc_objset_zplprops);
5706 zfs_ioctl_register_dataset_read(ZFS_IOC_DATASET_LIST_NEXT,
5707 zfs_ioc_dataset_list_next);
5708 zfs_ioctl_register_dataset_read(ZFS_IOC_SNAPSHOT_LIST_NEXT,
5709 zfs_ioc_snapshot_list_next);
5710 zfs_ioctl_register_dataset_read(ZFS_IOC_SEND_PROGRESS,
5711 zfs_ioc_send_progress);
5712
5713 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_DIFF,
5714 zfs_ioc_diff, zfs_secpolicy_diff);
5715 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_STATS,
5716 zfs_ioc_obj_to_stats, zfs_secpolicy_diff);
5717 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_PATH,
5718 zfs_ioc_obj_to_path, zfs_secpolicy_diff);
5719 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_ONE,
5720 zfs_ioc_userspace_one, zfs_secpolicy_userspace_one);
5721 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_MANY,
5722 zfs_ioc_userspace_many, zfs_secpolicy_userspace_many);
5723 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_SEND,
5724 zfs_ioc_send, zfs_secpolicy_send);
5725
5726 zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_PROP, zfs_ioc_set_prop,
5727 zfs_secpolicy_none);
5728 zfs_ioctl_register_dataset_modify(ZFS_IOC_DESTROY, zfs_ioc_destroy,
5729 zfs_secpolicy_destroy);
5730 zfs_ioctl_register_dataset_modify(ZFS_IOC_RENAME, zfs_ioc_rename,
5731 zfs_secpolicy_rename);
5732 zfs_ioctl_register_dataset_modify(ZFS_IOC_RECV, zfs_ioc_recv,
5733 zfs_secpolicy_recv);
5734 zfs_ioctl_register_dataset_modify(ZFS_IOC_PROMOTE, zfs_ioc_promote,
5735 zfs_secpolicy_promote);
5736 zfs_ioctl_register_dataset_modify(ZFS_IOC_INHERIT_PROP,
5737 zfs_ioc_inherit_prop, zfs_secpolicy_inherit_prop);
5738 zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_FSACL, zfs_ioc_set_fsacl,
5739 zfs_secpolicy_set_fsacl);
5740
5741 zfs_ioctl_register_dataset_nolog(ZFS_IOC_SHARE, zfs_ioc_share,
5742 zfs_secpolicy_share, POOL_CHECK_NONE);
5743 zfs_ioctl_register_dataset_nolog(ZFS_IOC_SMB_ACL, zfs_ioc_smb_acl,
5744 zfs_secpolicy_smb_acl, POOL_CHECK_NONE);
5745 zfs_ioctl_register_dataset_nolog(ZFS_IOC_USERSPACE_UPGRADE,
5746 zfs_ioc_userspace_upgrade, zfs_secpolicy_userspace_upgrade,
5747 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5748 zfs_ioctl_register_dataset_nolog(ZFS_IOC_TMP_SNAPSHOT,
5749 zfs_ioc_tmp_snapshot, zfs_secpolicy_tmp_snapshot,
5750 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5751 }
5752
5753 int
5754 pool_status_check(const char *name, zfs_ioc_namecheck_t type,
5755 zfs_ioc_poolcheck_t check)
5756 {
5757 spa_t *spa;
5758 int error;
5759
5760 ASSERT(type == POOL_NAME || type == DATASET_NAME);
5761
5762 if (check & POOL_CHECK_NONE)
5763 return (0);
5764
5765 error = spa_open(name, &spa, FTAG);
5766 if (error == 0) {
5767 if ((check & POOL_CHECK_SUSPENDED) && spa_suspended(spa))
5768 error = SET_ERROR(EAGAIN);
5769 else if ((check & POOL_CHECK_READONLY) && !spa_writeable(spa))
5770 error = SET_ERROR(EROFS);
5771 spa_close(spa, FTAG);
5772 }
5773 return (error);
5774 }
5775
5776 /*
5777 * Find a free minor number.
5778 */
5779 minor_t
5780 zfsdev_minor_alloc(void)
5781 {
5782 static minor_t last_minor;
5783 minor_t m;
5784
5785 ASSERT(MUTEX_HELD(&zfsdev_state_lock));
5786
5787 for (m = last_minor + 1; m != last_minor; m++) {
5788 if (m > ZFSDEV_MAX_MINOR)
5789 m = 1;
5790 if (ddi_get_soft_state(zfsdev_state, m) == NULL) {
5791 last_minor = m;
5792 return (m);
5793 }
5794 }
5795
5796 return (0);
5797 }
5798
5799 static int
5800 zfs_ctldev_init(dev_t *devp)
5801 {
5802 minor_t minor;
5803 zfs_soft_state_t *zs;
5804
5805 ASSERT(MUTEX_HELD(&zfsdev_state_lock));
5806 ASSERT(getminor(*devp) == 0);
5807
5808 minor = zfsdev_minor_alloc();
5809 if (minor == 0)
5810 return (SET_ERROR(ENXIO));
5811
5812 if (ddi_soft_state_zalloc(zfsdev_state, minor) != DDI_SUCCESS)
5813 return (SET_ERROR(EAGAIN));
5814
5815 *devp = makedevice(getemajor(*devp), minor);
5816
5817 zs = ddi_get_soft_state(zfsdev_state, minor);
5818 zs->zss_type = ZSST_CTLDEV;
5819 zfs_onexit_init((zfs_onexit_t **)&zs->zss_data);
5820
5821 return (0);
5822 }
5823
5824 static void
5825 zfs_ctldev_destroy(zfs_onexit_t *zo, minor_t minor)
5826 {
5827 ASSERT(MUTEX_HELD(&zfsdev_state_lock));
5828
5829 zfs_onexit_destroy(zo);
5830 ddi_soft_state_free(zfsdev_state, minor);
5831 }
5832
5833 void *
5834 zfsdev_get_soft_state(minor_t minor, enum zfs_soft_state_type which)
5835 {
5836 zfs_soft_state_t *zp;
5837
5838 zp = ddi_get_soft_state(zfsdev_state, minor);
5839 if (zp == NULL || zp->zss_type != which)
5840 return (NULL);
5841
5842 return (zp->zss_data);
5843 }
5844
5845 static int
5846 zfsdev_open(dev_t *devp, int flag, int otyp, cred_t *cr)
5847 {
5848 int error = 0;
5849
5850 if (getminor(*devp) != 0)
5851 return (zvol_open(devp, flag, otyp, cr));
5852
5853 /* This is the control device. Allocate a new minor if requested. */
5854 if (flag & FEXCL) {
5855 mutex_enter(&zfsdev_state_lock);
5856 error = zfs_ctldev_init(devp);
5857 mutex_exit(&zfsdev_state_lock);
5858 }
5859
5860 return (error);
5861 }
5862
5863 static int
5864 zfsdev_close(dev_t dev, int flag, int otyp, cred_t *cr)
5865 {
5866 zfs_onexit_t *zo;
5867 minor_t minor = getminor(dev);
5868
5869 if (minor == 0)
5870 return (0);
5871
5872 mutex_enter(&zfsdev_state_lock);
5873 zo = zfsdev_get_soft_state(minor, ZSST_CTLDEV);
5874 if (zo == NULL) {
5875 mutex_exit(&zfsdev_state_lock);
5876 return (zvol_close(dev, flag, otyp, cr));
5877 }
5878 zfs_ctldev_destroy(zo, minor);
5879 mutex_exit(&zfsdev_state_lock);
5880
5881 return (0);
5882 }
5883
5884 static int
5885 zfsdev_ioctl(dev_t dev, int cmd, intptr_t arg, int flag, cred_t *cr, int *rvalp)
5886 {
5887 zfs_cmd_t *zc;
5888 uint_t vecnum;
5889 int error, rc, len;
5890 minor_t minor = getminor(dev);
5891 const zfs_ioc_vec_t *vec;
5892 char *saved_poolname = NULL;
5893 nvlist_t *innvl = NULL;
5894
5895 if (minor != 0 &&
5896 zfsdev_get_soft_state(minor, ZSST_CTLDEV) == NULL)
5897 return (zvol_ioctl(dev, cmd, arg, flag, cr, rvalp));
5898
5899 vecnum = cmd - ZFS_IOC_FIRST;
5900 ASSERT3U(getmajor(dev), ==, ddi_driver_major(zfs_dip));
5901
5902 if (vecnum >= sizeof (zfs_ioc_vec) / sizeof (zfs_ioc_vec[0]))
5903 return (SET_ERROR(EINVAL));
5904 vec = &zfs_ioc_vec[vecnum];
5905
5906 zc = kmem_zalloc(sizeof (zfs_cmd_t), KM_SLEEP);
5907
5908 error = ddi_copyin((void *)arg, zc, sizeof (zfs_cmd_t), flag);
5909 if (error != 0) {
5910 error = SET_ERROR(EFAULT);
5911 goto out;
5912 }
5913
5914 zc->zc_iflags = flag & FKIOCTL;
5915 if (zc->zc_nvlist_src_size != 0) {
5916 error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
5917 zc->zc_iflags, &innvl);
5918 if (error != 0)
5919 goto out;
5920 }
5921
5922 /*
5923 * Ensure that all pool/dataset names are valid before we pass down to
5924 * the lower layers.
5925 */
5926 zc->zc_name[sizeof (zc->zc_name) - 1] = '\0';
5927 switch (vec->zvec_namecheck) {
5928 case POOL_NAME:
5929 if (pool_namecheck(zc->zc_name, NULL, NULL) != 0)
5930 error = SET_ERROR(EINVAL);
5931 else
5932 error = pool_status_check(zc->zc_name,
5933 vec->zvec_namecheck, vec->zvec_pool_check);
5934 break;
5935
5936 case DATASET_NAME:
5937 if (dataset_namecheck(zc->zc_name, NULL, NULL) != 0)
5938 error = SET_ERROR(EINVAL);
5939 else
5940 error = pool_status_check(zc->zc_name,
5941 vec->zvec_namecheck, vec->zvec_pool_check);
5942 break;
5943
5944 case NO_NAME:
5945 break;
5946 }
5947
5948
5949 if (error == 0 && !(flag & FKIOCTL))
5950 error = vec->zvec_secpolicy(zc, innvl, cr);
5951
5952 if (error != 0)
5953 goto out;
5954
5955 /* legacy ioctls can modify zc_name */
5956 len = strcspn(zc->zc_name, "/@#") + 1;
5957 saved_poolname = kmem_alloc(len, KM_SLEEP);
5958 (void) strlcpy(saved_poolname, zc->zc_name, len);
5959
5960 if (vec->zvec_func != NULL) {
5961 nvlist_t *outnvl;
5962 int puterror = 0;
5963 spa_t *spa;
5964 nvlist_t *lognv = NULL;
5965
5966 ASSERT(vec->zvec_legacy_func == NULL);
5967
5968 /*
5969 * Add the innvl to the lognv before calling the func,
5970 * in case the func changes the innvl.
5971 */
5972 if (vec->zvec_allow_log) {
5973 lognv = fnvlist_alloc();
5974 fnvlist_add_string(lognv, ZPOOL_HIST_IOCTL,
5975 vec->zvec_name);
5976 if (!nvlist_empty(innvl)) {
5977 fnvlist_add_nvlist(lognv, ZPOOL_HIST_INPUT_NVL,
5978 innvl);
5979 }
5980 }
5981
5982 outnvl = fnvlist_alloc();
5983 error = vec->zvec_func(zc->zc_name, innvl, outnvl);
5984
5985 if (error == 0 && vec->zvec_allow_log &&
5986 spa_open(zc->zc_name, &spa, FTAG) == 0) {
5987 if (!nvlist_empty(outnvl)) {
5988 fnvlist_add_nvlist(lognv, ZPOOL_HIST_OUTPUT_NVL,
5989 outnvl);
5990 }
5991 (void) spa_history_log_nvl(spa, lognv);
5992 spa_close(spa, FTAG);
5993 }
5994 fnvlist_free(lognv);
5995
5996 if (!nvlist_empty(outnvl) || zc->zc_nvlist_dst_size != 0) {
5997 int smusherror = 0;
5998 if (vec->zvec_smush_outnvlist) {
5999 smusherror = nvlist_smush(outnvl,
6000 zc->zc_nvlist_dst_size);
6001 }
6002 if (smusherror == 0)
6003 puterror = put_nvlist(zc, outnvl);
6004 }
6005
6006 if (puterror != 0)
6007 error = puterror;
6008
6009 nvlist_free(outnvl);
6010 } else {
6011 error = vec->zvec_legacy_func(zc);
6012 }
6013
6014 out:
6015 nvlist_free(innvl);
6016 rc = ddi_copyout(zc, (void *)arg, sizeof (zfs_cmd_t), flag);
6017 if (error == 0 && rc != 0)
6018 error = SET_ERROR(EFAULT);
6019 if (error == 0 && vec->zvec_allow_log) {
6020 char *s = tsd_get(zfs_allow_log_key);
6021 if (s != NULL)
6022 strfree(s);
6023 (void) tsd_set(zfs_allow_log_key, saved_poolname);
6024 } else {
6025 if (saved_poolname != NULL)
6026 strfree(saved_poolname);
6027 }
6028
6029 kmem_free(zc, sizeof (zfs_cmd_t));
6030 return (error);
6031 }
6032
6033 static int
6034 zfs_attach(dev_info_t *dip, ddi_attach_cmd_t cmd)
6035 {
6036 if (cmd != DDI_ATTACH)
6037 return (DDI_FAILURE);
6038
6039 if (ddi_create_minor_node(dip, "zfs", S_IFCHR, 0,
6040 DDI_PSEUDO, 0) == DDI_FAILURE)
6041 return (DDI_FAILURE);
6042
6043 zfs_dip = dip;
6044
6045 ddi_report_dev(dip);
6046
6047 return (DDI_SUCCESS);
6048 }
6049
6050 static int
6051 zfs_detach(dev_info_t *dip, ddi_detach_cmd_t cmd)
6052 {
6053 if (spa_busy() || zfs_busy() || zvol_busy())
6054 return (DDI_FAILURE);
6055
6056 if (cmd != DDI_DETACH)
6057 return (DDI_FAILURE);
6058
6059 zfs_dip = NULL;
6060
6061 ddi_prop_remove_all(dip);
6062 ddi_remove_minor_node(dip, NULL);
6063
6064 return (DDI_SUCCESS);
6065 }
6066
6067 /*ARGSUSED*/
6068 static int
6069 zfs_info(dev_info_t *dip, ddi_info_cmd_t infocmd, void *arg, void **result)
6070 {
6071 switch (infocmd) {
6072 case DDI_INFO_DEVT2DEVINFO:
6073 *result = zfs_dip;
6074 return (DDI_SUCCESS);
6075
6076 case DDI_INFO_DEVT2INSTANCE:
6077 *result = (void *)0;
6078 return (DDI_SUCCESS);
6079 }
6080
6081 return (DDI_FAILURE);
6082 }
6083
6084 /*
6085 * OK, so this is a little weird.
6086 *
6087 * /dev/zfs is the control node, i.e. minor 0.
6088 * /dev/zvol/[r]dsk/pool/dataset are the zvols, minor > 0.
6089 *
6090 * /dev/zfs has basically nothing to do except serve up ioctls,
6091 * so most of the standard driver entry points are in zvol.c.
6092 */
6093 static struct cb_ops zfs_cb_ops = {
6094 zfsdev_open, /* open */
6095 zfsdev_close, /* close */
6096 zvol_strategy, /* strategy */
6097 nodev, /* print */
6098 zvol_dump, /* dump */
6099 zvol_read, /* read */
6100 zvol_write, /* write */
6101 zfsdev_ioctl, /* ioctl */
6102 nodev, /* devmap */
6103 nodev, /* mmap */
6104 nodev, /* segmap */
6105 nochpoll, /* poll */
6106 ddi_prop_op, /* prop_op */
6107 NULL, /* streamtab */
6108 D_NEW | D_MP | D_64BIT, /* Driver compatibility flag */
6109 CB_REV, /* version */
6110 nodev, /* async read */
6111 nodev, /* async write */
6112 };
6113
6114 static struct dev_ops zfs_dev_ops = {
6115 DEVO_REV, /* version */
6116 0, /* refcnt */
6117 zfs_info, /* info */
6118 nulldev, /* identify */
6119 nulldev, /* probe */
6120 zfs_attach, /* attach */
6121 zfs_detach, /* detach */
6122 nodev, /* reset */
6123 &zfs_cb_ops, /* driver operations */
6124 NULL, /* no bus operations */
6125 NULL, /* power */
6126 ddi_quiesce_not_needed, /* quiesce */
6127 };
6128
6129 static struct modldrv zfs_modldrv = {
6130 &mod_driverops,
6131 "ZFS storage pool",
6132 &zfs_dev_ops
6133 };
6134
6135 static struct modlinkage modlinkage = {
6136 MODREV_1,
6137 (void *)&zfs_modlfs,
6138 (void *)&zfs_modldrv,
6139 NULL
6140 };
6141
6142 static void
6143 zfs_allow_log_destroy(void *arg)
6144 {
6145 char *poolname = arg;
6146 strfree(poolname);
6147 }
6148
6149 int
6150 _init(void)
6151 {
6152 int error;
6153
6154 spa_init(FREAD | FWRITE);
6155 zfs_init();
6156 zvol_init();
6157 zfs_ioctl_init();
6158
6159 if ((error = mod_install(&modlinkage)) != 0) {
6160 zvol_fini();
6161 zfs_fini();
6162 spa_fini();
6163 return (error);
6164 }
6165
6166 tsd_create(&zfs_fsyncer_key, NULL);
6167 tsd_create(&rrw_tsd_key, rrw_tsd_destroy);
6168 tsd_create(&zfs_allow_log_key, zfs_allow_log_destroy);
6169
6170 error = ldi_ident_from_mod(&modlinkage, &zfs_li);
6171 ASSERT(error == 0);
6172 mutex_init(&zfs_share_lock, NULL, MUTEX_DEFAULT, NULL);
6173
6174 return (0);
6175 }
6176
6177 int
6178 _fini(void)
6179 {
6180 int error;
6181
6182 if (spa_busy() || zfs_busy() || zvol_busy() || zio_injection_enabled)
6183 return (SET_ERROR(EBUSY));
6184
6185 if ((error = mod_remove(&modlinkage)) != 0)
6186 return (error);
6187
6188 zvol_fini();
6189 zfs_fini();
6190 spa_fini();
6191 if (zfs_nfsshare_inited)
6192 (void) ddi_modclose(nfs_mod);
6193 if (zfs_smbshare_inited)
6194 (void) ddi_modclose(smbsrv_mod);
6195 if (zfs_nfsshare_inited || zfs_smbshare_inited)
6196 (void) ddi_modclose(sharefs_mod);
6197
6198 tsd_destroy(&zfs_fsyncer_key);
6199 ldi_ident_release(zfs_li);
6200 zfs_li = NULL;
6201 mutex_destroy(&zfs_share_lock);
6202
6203 return (error);
6204 }
6205
6206 int
6207 _info(struct modinfo *modinfop)
6208 {
6209 return (mod_info(&modlinkage, modinfop));
6210 }