1 /*
2 * CDDL HEADER START
3 *
4 * The contents of this file are subject to the terms of the
5 * Common Development and Distribution License (the "License").
6 * You may not use this file except in compliance with the License.
7 *
8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9 * or http://www.opensolaris.org/os/licensing.
10 * See the License for the specific language governing permissions
11 * and limitations under the License.
12 *
13 * When distributing Covered Code, include this CDDL HEADER in each
14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15 * If applicable, add the following below this CDDL HEADER, with the
16 * fields enclosed by brackets "[]" replaced with your own identifying
17 * information: Portions Copyright [yyyy] [name of copyright owner]
18 *
19 * CDDL HEADER END
20 */
21 /*
22 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
23 */
24
25 #include <sys/types.h>
26 #include <sys/param.h>
27 #include <sys/errno.h>
28 #include <sys/uio.h>
29 #include <sys/buf.h>
30 #include <sys/modctl.h>
31 #include <sys/open.h>
32 #include <sys/file.h>
33 #include <sys/kmem.h>
34 #include <sys/conf.h>
35 #include <sys/cmn_err.h>
36 #include <sys/stat.h>
37 #include <sys/zfs_ioctl.h>
38 #include <sys/zfs_vfsops.h>
39 #include <sys/zfs_znode.h>
40 #include <sys/zap.h>
41 #include <sys/spa.h>
42 #include <sys/spa_impl.h>
43 #include <sys/vdev.h>
44 #include <sys/priv_impl.h>
45 #include <sys/dmu.h>
46 #include <sys/dsl_dir.h>
47 #include <sys/dsl_dataset.h>
48 #include <sys/dsl_prop.h>
49 #include <sys/dsl_deleg.h>
50 #include <sys/dmu_objset.h>
51 #include <sys/ddi.h>
52 #include <sys/sunddi.h>
53 #include <sys/sunldi.h>
54 #include <sys/policy.h>
55 #include <sys/zone.h>
56 #include <sys/nvpair.h>
57 #include <sys/pathname.h>
58 #include <sys/mount.h>
59 #include <sys/sdt.h>
60 #include <sys/fs/zfs.h>
61 #include <sys/zfs_ctldir.h>
62 #include <sys/zfs_dir.h>
63 #include <sys/zfs_onexit.h>
64 #include <sys/zvol.h>
65 #include <sys/dsl_scan.h>
66 #include <sharefs/share.h>
67 #include <sys/dmu_objset.h>
68
69 #include "zfs_namecheck.h"
70 #include "zfs_prop.h"
71 #include "zfs_deleg.h"
72 #include "zfs_comutil.h"
73
74 extern struct modlfs zfs_modlfs;
75
76 extern void zfs_init(void);
77 extern void zfs_fini(void);
78
79 ldi_ident_t zfs_li = NULL;
80 dev_info_t *zfs_dip;
81
82 typedef int zfs_ioc_func_t(zfs_cmd_t *);
83 typedef int zfs_secpolicy_func_t(zfs_cmd_t *, cred_t *);
84
85 typedef enum {
86 NO_NAME,
87 POOL_NAME,
88 DATASET_NAME
89 } zfs_ioc_namecheck_t;
90
91 typedef enum {
92 POOL_CHECK_NONE = 1 << 0,
93 POOL_CHECK_SUSPENDED = 1 << 1,
94 POOL_CHECK_READONLY = 1 << 2
95 } zfs_ioc_poolcheck_t;
96
97 typedef struct zfs_ioc_vec {
98 zfs_ioc_func_t *zvec_func;
99 zfs_secpolicy_func_t *zvec_secpolicy;
100 zfs_ioc_namecheck_t zvec_namecheck;
101 boolean_t zvec_his_log;
102 zfs_ioc_poolcheck_t zvec_pool_check;
103 } zfs_ioc_vec_t;
104
105 /* This array is indexed by zfs_userquota_prop_t */
106 static const char *userquota_perms[] = {
107 ZFS_DELEG_PERM_USERUSED,
108 ZFS_DELEG_PERM_USERQUOTA,
109 ZFS_DELEG_PERM_GROUPUSED,
110 ZFS_DELEG_PERM_GROUPQUOTA,
111 };
112
113 static int zfs_ioc_userspace_upgrade(zfs_cmd_t *zc);
114 static int zfs_check_settable(const char *name, nvpair_t *property,
115 cred_t *cr);
116 static int zfs_check_clearable(char *dataset, nvlist_t *props,
117 nvlist_t **errors);
118 static int zfs_fill_zplprops_root(uint64_t, nvlist_t *, nvlist_t *,
119 boolean_t *);
120 int zfs_set_prop_nvlist(const char *, zprop_source_t, nvlist_t *, nvlist_t **);
121
122 /* _NOTE(PRINTFLIKE(4)) - this is printf-like, but lint is too whiney */
123 void
124 __dprintf(const char *file, const char *func, int line, const char *fmt, ...)
125 {
126 const char *newfile;
127 char buf[512];
128 va_list adx;
129
130 /*
131 * Get rid of annoying "../common/" prefix to filename.
132 */
133 newfile = strrchr(file, '/');
134 if (newfile != NULL) {
135 newfile = newfile + 1; /* Get rid of leading / */
136 } else {
137 newfile = file;
138 }
139
140 va_start(adx, fmt);
141 (void) vsnprintf(buf, sizeof (buf), fmt, adx);
142 va_end(adx);
143
144 /*
145 * To get this data, use the zfs-dprintf probe as so:
146 * dtrace -q -n 'zfs-dprintf \
147 * /stringof(arg0) == "dbuf.c"/ \
148 * {printf("%s: %s", stringof(arg1), stringof(arg3))}'
149 * arg0 = file name
150 * arg1 = function name
151 * arg2 = line number
152 * arg3 = message
153 */
154 DTRACE_PROBE4(zfs__dprintf,
155 char *, newfile, char *, func, int, line, char *, buf);
156 }
157
158 static void
159 history_str_free(char *buf)
160 {
161 kmem_free(buf, HIS_MAX_RECORD_LEN);
162 }
163
164 static char *
165 history_str_get(zfs_cmd_t *zc)
166 {
167 char *buf;
168
169 if (zc->zc_history == NULL)
170 return (NULL);
171
172 buf = kmem_alloc(HIS_MAX_RECORD_LEN, KM_SLEEP);
173 if (copyinstr((void *)(uintptr_t)zc->zc_history,
174 buf, HIS_MAX_RECORD_LEN, NULL) != 0) {
175 history_str_free(buf);
176 return (NULL);
177 }
178
179 buf[HIS_MAX_RECORD_LEN -1] = '\0';
180
181 return (buf);
182 }
183
184 /*
185 * Check to see if the named dataset is currently defined as bootable
186 */
187 static boolean_t
188 zfs_is_bootfs(const char *name)
189 {
190 objset_t *os;
191
192 if (dmu_objset_hold(name, FTAG, &os) == 0) {
193 boolean_t ret;
194 ret = (dmu_objset_id(os) == spa_bootfs(dmu_objset_spa(os)));
195 dmu_objset_rele(os, FTAG);
196 return (ret);
197 }
198 return (B_FALSE);
199 }
200
201 /*
202 * zfs_earlier_version
203 *
204 * Return non-zero if the spa version is less than requested version.
205 */
206 static int
207 zfs_earlier_version(const char *name, int version)
208 {
209 spa_t *spa;
210
211 if (spa_open(name, &spa, FTAG) == 0) {
212 if (spa_version(spa) < version) {
213 spa_close(spa, FTAG);
214 return (1);
215 }
216 spa_close(spa, FTAG);
217 }
218 return (0);
219 }
220
221 /*
222 * zpl_earlier_version
223 *
224 * Return TRUE if the ZPL version is less than requested version.
225 */
226 static boolean_t
227 zpl_earlier_version(const char *name, int version)
228 {
229 objset_t *os;
230 boolean_t rc = B_TRUE;
231
232 if (dmu_objset_hold(name, FTAG, &os) == 0) {
233 uint64_t zplversion;
234
235 if (dmu_objset_type(os) != DMU_OST_ZFS) {
236 dmu_objset_rele(os, FTAG);
237 return (B_TRUE);
238 }
239 /* XXX reading from non-owned objset */
240 if (zfs_get_zplprop(os, ZFS_PROP_VERSION, &zplversion) == 0)
241 rc = zplversion < version;
242 dmu_objset_rele(os, FTAG);
243 }
244 return (rc);
245 }
246
247 static void
248 zfs_log_history(zfs_cmd_t *zc)
249 {
250 spa_t *spa;
251 char *buf;
252
253 if ((buf = history_str_get(zc)) == NULL)
254 return;
255
256 if (spa_open(zc->zc_name, &spa, FTAG) == 0) {
257 if (spa_version(spa) >= SPA_VERSION_ZPOOL_HISTORY)
258 (void) spa_history_log(spa, buf, LOG_CMD_NORMAL);
259 spa_close(spa, FTAG);
260 }
261 history_str_free(buf);
262 }
263
264 /*
265 * Policy for top-level read operations (list pools). Requires no privileges,
266 * and can be used in the local zone, as there is no associated dataset.
267 */
268 /* ARGSUSED */
269 static int
270 zfs_secpolicy_none(zfs_cmd_t *zc, cred_t *cr)
271 {
272 return (0);
273 }
274
275 /*
276 * Policy for dataset read operations (list children, get statistics). Requires
277 * no privileges, but must be visible in the local zone.
278 */
279 /* ARGSUSED */
280 static int
281 zfs_secpolicy_read(zfs_cmd_t *zc, cred_t *cr)
282 {
283 if (INGLOBALZONE(curproc) ||
284 zone_dataset_visible(zc->zc_name, NULL))
285 return (0);
286
287 return (ENOENT);
288 }
289
290 static int
291 zfs_dozonecheck_impl(const char *dataset, uint64_t zoned, cred_t *cr)
292 {
293 int writable = 1;
294
295 /*
296 * The dataset must be visible by this zone -- check this first
297 * so they don't see EPERM on something they shouldn't know about.
298 */
299 if (!INGLOBALZONE(curproc) &&
300 !zone_dataset_visible(dataset, &writable))
301 return (ENOENT);
302
303 if (INGLOBALZONE(curproc)) {
304 /*
305 * If the fs is zoned, only root can access it from the
306 * global zone.
307 */
308 if (secpolicy_zfs(cr) && zoned)
309 return (EPERM);
310 } else {
311 /*
312 * If we are in a local zone, the 'zoned' property must be set.
313 */
314 if (!zoned)
315 return (EPERM);
316
317 /* must be writable by this zone */
318 if (!writable)
319 return (EPERM);
320 }
321 return (0);
322 }
323
324 static int
325 zfs_dozonecheck(const char *dataset, cred_t *cr)
326 {
327 uint64_t zoned;
328
329 if (dsl_prop_get_integer(dataset, "zoned", &zoned, NULL))
330 return (ENOENT);
331
332 return (zfs_dozonecheck_impl(dataset, zoned, cr));
333 }
334
335 static int
336 zfs_dozonecheck_ds(const char *dataset, dsl_dataset_t *ds, cred_t *cr)
337 {
338 uint64_t zoned;
339
340 rw_enter(&ds->ds_dir->dd_pool->dp_config_rwlock, RW_READER);
341 if (dsl_prop_get_ds(ds, "zoned", 8, 1, &zoned, NULL)) {
342 rw_exit(&ds->ds_dir->dd_pool->dp_config_rwlock);
343 return (ENOENT);
344 }
345 rw_exit(&ds->ds_dir->dd_pool->dp_config_rwlock);
346
347 return (zfs_dozonecheck_impl(dataset, zoned, cr));
348 }
349
350 int
351 zfs_secpolicy_write_perms(const char *name, const char *perm, cred_t *cr)
352 {
353 int error;
354
355 error = zfs_dozonecheck(name, cr);
356 if (error == 0) {
357 error = secpolicy_zfs(cr);
358 if (error)
359 error = dsl_deleg_access(name, perm, cr);
360 }
361 return (error);
362 }
363
364 int
365 zfs_secpolicy_write_perms_ds(const char *name, dsl_dataset_t *ds,
366 const char *perm, cred_t *cr)
367 {
368 int error;
369
370 error = zfs_dozonecheck_ds(name, ds, cr);
371 if (error == 0) {
372 error = secpolicy_zfs(cr);
373 if (error)
374 error = dsl_deleg_access_impl(ds, perm, cr);
375 }
376 return (error);
377 }
378
379 /*
380 * Policy for setting the security label property.
381 *
382 * Returns 0 for success, non-zero for access and other errors.
383 */
384 static int
385 zfs_set_slabel_policy(const char *name, char *strval, cred_t *cr)
386 {
387 char ds_hexsl[MAXNAMELEN];
388 bslabel_t ds_sl, new_sl;
389 boolean_t new_default = FALSE;
390 uint64_t zoned;
391 int needed_priv = -1;
392 int error;
393
394 /* First get the existing dataset label. */
395 error = dsl_prop_get(name, zfs_prop_to_name(ZFS_PROP_MLSLABEL),
396 1, sizeof (ds_hexsl), &ds_hexsl, NULL);
397 if (error)
398 return (EPERM);
399
400 if (strcasecmp(strval, ZFS_MLSLABEL_DEFAULT) == 0)
401 new_default = TRUE;
402
403 /* The label must be translatable */
404 if (!new_default && (hexstr_to_label(strval, &new_sl) != 0))
405 return (EINVAL);
406
407 /*
408 * In a non-global zone, disallow attempts to set a label that
409 * doesn't match that of the zone; otherwise no other checks
410 * are needed.
411 */
412 if (!INGLOBALZONE(curproc)) {
413 if (new_default || !blequal(&new_sl, CR_SL(CRED())))
414 return (EPERM);
415 return (0);
416 }
417
418 /*
419 * For global-zone datasets (i.e., those whose zoned property is
420 * "off", verify that the specified new label is valid for the
421 * global zone.
422 */
423 if (dsl_prop_get_integer(name,
424 zfs_prop_to_name(ZFS_PROP_ZONED), &zoned, NULL))
425 return (EPERM);
426 if (!zoned) {
427 if (zfs_check_global_label(name, strval) != 0)
428 return (EPERM);
429 }
430
431 /*
432 * If the existing dataset label is nondefault, check if the
433 * dataset is mounted (label cannot be changed while mounted).
434 * Get the zfsvfs; if there isn't one, then the dataset isn't
435 * mounted (or isn't a dataset, doesn't exist, ...).
436 */
437 if (strcasecmp(ds_hexsl, ZFS_MLSLABEL_DEFAULT) != 0) {
438 objset_t *os;
439 static char *setsl_tag = "setsl_tag";
440
441 /*
442 * Try to own the dataset; abort if there is any error,
443 * (e.g., already mounted, in use, or other error).
444 */
445 error = dmu_objset_own(name, DMU_OST_ZFS, B_TRUE,
446 setsl_tag, &os);
447 if (error)
448 return (EPERM);
449
450 dmu_objset_disown(os, setsl_tag);
451
452 if (new_default) {
453 needed_priv = PRIV_FILE_DOWNGRADE_SL;
454 goto out_check;
455 }
456
457 if (hexstr_to_label(strval, &new_sl) != 0)
458 return (EPERM);
459
460 if (blstrictdom(&ds_sl, &new_sl))
461 needed_priv = PRIV_FILE_DOWNGRADE_SL;
462 else if (blstrictdom(&new_sl, &ds_sl))
463 needed_priv = PRIV_FILE_UPGRADE_SL;
464 } else {
465 /* dataset currently has a default label */
466 if (!new_default)
467 needed_priv = PRIV_FILE_UPGRADE_SL;
468 }
469
470 out_check:
471 if (needed_priv != -1)
472 return (PRIV_POLICY(cr, needed_priv, B_FALSE, EPERM, NULL));
473 return (0);
474 }
475
476 static int
477 zfs_secpolicy_setprop(const char *dsname, zfs_prop_t prop, nvpair_t *propval,
478 cred_t *cr)
479 {
480 char *strval;
481
482 /*
483 * Check permissions for special properties.
484 */
485 switch (prop) {
486 case ZFS_PROP_ZONED:
487 /*
488 * Disallow setting of 'zoned' from within a local zone.
489 */
490 if (!INGLOBALZONE(curproc))
491 return (EPERM);
492 break;
493
494 case ZFS_PROP_QUOTA:
495 if (!INGLOBALZONE(curproc)) {
496 uint64_t zoned;
497 char setpoint[MAXNAMELEN];
498 /*
499 * Unprivileged users are allowed to modify the
500 * quota on things *under* (ie. contained by)
501 * the thing they own.
502 */
503 if (dsl_prop_get_integer(dsname, "zoned", &zoned,
504 setpoint))
505 return (EPERM);
506 if (!zoned || strlen(dsname) <= strlen(setpoint))
507 return (EPERM);
508 }
509 break;
510
511 case ZFS_PROP_MLSLABEL:
512 if (!is_system_labeled())
513 return (EPERM);
514
515 if (nvpair_value_string(propval, &strval) == 0) {
516 int err;
517
518 err = zfs_set_slabel_policy(dsname, strval, CRED());
519 if (err != 0)
520 return (err);
521 }
522 break;
523 }
524
525 return (zfs_secpolicy_write_perms(dsname, zfs_prop_to_name(prop), cr));
526 }
527
528 int
529 zfs_secpolicy_fsacl(zfs_cmd_t *zc, cred_t *cr)
530 {
531 int error;
532
533 error = zfs_dozonecheck(zc->zc_name, cr);
534 if (error)
535 return (error);
536
537 /*
538 * permission to set permissions will be evaluated later in
539 * dsl_deleg_can_allow()
540 */
541 return (0);
542 }
543
544 int
545 zfs_secpolicy_rollback(zfs_cmd_t *zc, cred_t *cr)
546 {
547 return (zfs_secpolicy_write_perms(zc->zc_name,
548 ZFS_DELEG_PERM_ROLLBACK, cr));
549 }
550
551 int
552 zfs_secpolicy_send(zfs_cmd_t *zc, cred_t *cr)
553 {
554 spa_t *spa;
555 dsl_pool_t *dp;
556 dsl_dataset_t *ds;
557 char *cp;
558 int error;
559
560 /*
561 * Generate the current snapshot name from the given objsetid, then
562 * use that name for the secpolicy/zone checks.
563 */
564 cp = strchr(zc->zc_name, '@');
565 if (cp == NULL)
566 return (EINVAL);
567 error = spa_open(zc->zc_name, &spa, FTAG);
568 if (error)
569 return (error);
570
571 dp = spa_get_dsl(spa);
572 rw_enter(&dp->dp_config_rwlock, RW_READER);
573 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &ds);
574 rw_exit(&dp->dp_config_rwlock);
575 spa_close(spa, FTAG);
576 if (error)
577 return (error);
578
579 dsl_dataset_name(ds, zc->zc_name);
580
581 error = zfs_secpolicy_write_perms_ds(zc->zc_name, ds,
582 ZFS_DELEG_PERM_SEND, cr);
583 dsl_dataset_rele(ds, FTAG);
584
585 return (error);
586 }
587
588 static int
589 zfs_secpolicy_deleg_share(zfs_cmd_t *zc, cred_t *cr)
590 {
591 vnode_t *vp;
592 int error;
593
594 if ((error = lookupname(zc->zc_value, UIO_SYSSPACE,
595 NO_FOLLOW, NULL, &vp)) != 0)
596 return (error);
597
598 /* Now make sure mntpnt and dataset are ZFS */
599
600 if (vp->v_vfsp->vfs_fstype != zfsfstype ||
601 (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource),
602 zc->zc_name) != 0)) {
603 VN_RELE(vp);
604 return (EPERM);
605 }
606
607 VN_RELE(vp);
608 return (dsl_deleg_access(zc->zc_name,
609 ZFS_DELEG_PERM_SHARE, cr));
610 }
611
612 int
613 zfs_secpolicy_share(zfs_cmd_t *zc, cred_t *cr)
614 {
615 if (!INGLOBALZONE(curproc))
616 return (EPERM);
617
618 if (secpolicy_nfs(cr) == 0) {
619 return (0);
620 } else {
621 return (zfs_secpolicy_deleg_share(zc, cr));
622 }
623 }
624
625 int
626 zfs_secpolicy_smb_acl(zfs_cmd_t *zc, cred_t *cr)
627 {
628 if (!INGLOBALZONE(curproc))
629 return (EPERM);
630
631 if (secpolicy_smb(cr) == 0) {
632 return (0);
633 } else {
634 return (zfs_secpolicy_deleg_share(zc, cr));
635 }
636 }
637
638 static int
639 zfs_get_parent(const char *datasetname, char *parent, int parentsize)
640 {
641 char *cp;
642
643 /*
644 * Remove the @bla or /bla from the end of the name to get the parent.
645 */
646 (void) strncpy(parent, datasetname, parentsize);
647 cp = strrchr(parent, '@');
648 if (cp != NULL) {
649 cp[0] = '\0';
650 } else {
651 cp = strrchr(parent, '/');
652 if (cp == NULL)
653 return (ENOENT);
654 cp[0] = '\0';
655 }
656
657 return (0);
658 }
659
660 int
661 zfs_secpolicy_destroy_perms(const char *name, cred_t *cr)
662 {
663 int error;
664
665 if ((error = zfs_secpolicy_write_perms(name,
666 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
667 return (error);
668
669 return (zfs_secpolicy_write_perms(name, ZFS_DELEG_PERM_DESTROY, cr));
670 }
671
672 static int
673 zfs_secpolicy_destroy(zfs_cmd_t *zc, cred_t *cr)
674 {
675 return (zfs_secpolicy_destroy_perms(zc->zc_name, cr));
676 }
677
678 /*
679 * Destroying snapshots with delegated permissions requires
680 * descendent mount and destroy permissions.
681 * Reassemble the full filesystem@snap name so dsl_deleg_access()
682 * can do the correct permission check.
683 *
684 * Since this routine is used when doing a recursive destroy of snapshots
685 * and destroying snapshots requires descendent permissions, a successfull
686 * check of the top level snapshot applies to snapshots of all descendent
687 * datasets as well.
688 */
689 static int
690 zfs_secpolicy_destroy_snaps(zfs_cmd_t *zc, cred_t *cr)
691 {
692 int error;
693 char *dsname;
694
695 dsname = kmem_asprintf("%s@%s", zc->zc_name, zc->zc_value);
696
697 error = zfs_secpolicy_destroy_perms(dsname, cr);
698
699 strfree(dsname);
700 return (error);
701 }
702
703 int
704 zfs_secpolicy_rename_perms(const char *from, const char *to, cred_t *cr)
705 {
706 char parentname[MAXNAMELEN];
707 int error;
708
709 if ((error = zfs_secpolicy_write_perms(from,
710 ZFS_DELEG_PERM_RENAME, cr)) != 0)
711 return (error);
712
713 if ((error = zfs_secpolicy_write_perms(from,
714 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
715 return (error);
716
717 if ((error = zfs_get_parent(to, parentname,
718 sizeof (parentname))) != 0)
719 return (error);
720
721 if ((error = zfs_secpolicy_write_perms(parentname,
722 ZFS_DELEG_PERM_CREATE, cr)) != 0)
723 return (error);
724
725 if ((error = zfs_secpolicy_write_perms(parentname,
726 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
727 return (error);
728
729 return (error);
730 }
731
732 static int
733 zfs_secpolicy_rename(zfs_cmd_t *zc, cred_t *cr)
734 {
735 return (zfs_secpolicy_rename_perms(zc->zc_name, zc->zc_value, cr));
736 }
737
738 static int
739 zfs_secpolicy_promote(zfs_cmd_t *zc, cred_t *cr)
740 {
741 char parentname[MAXNAMELEN];
742 objset_t *clone;
743 int error;
744
745 error = zfs_secpolicy_write_perms(zc->zc_name,
746 ZFS_DELEG_PERM_PROMOTE, cr);
747 if (error)
748 return (error);
749
750 error = dmu_objset_hold(zc->zc_name, FTAG, &clone);
751
752 if (error == 0) {
753 dsl_dataset_t *pclone = NULL;
754 dsl_dir_t *dd;
755 dd = clone->os_dsl_dataset->ds_dir;
756
757 rw_enter(&dd->dd_pool->dp_config_rwlock, RW_READER);
758 error = dsl_dataset_hold_obj(dd->dd_pool,
759 dd->dd_phys->dd_origin_obj, FTAG, &pclone);
760 rw_exit(&dd->dd_pool->dp_config_rwlock);
761 if (error) {
762 dmu_objset_rele(clone, FTAG);
763 return (error);
764 }
765
766 error = zfs_secpolicy_write_perms(zc->zc_name,
767 ZFS_DELEG_PERM_MOUNT, cr);
768
769 dsl_dataset_name(pclone, parentname);
770 dmu_objset_rele(clone, FTAG);
771 dsl_dataset_rele(pclone, FTAG);
772 if (error == 0)
773 error = zfs_secpolicy_write_perms(parentname,
774 ZFS_DELEG_PERM_PROMOTE, cr);
775 }
776 return (error);
777 }
778
779 static int
780 zfs_secpolicy_receive(zfs_cmd_t *zc, cred_t *cr)
781 {
782 int error;
783
784 if ((error = zfs_secpolicy_write_perms(zc->zc_name,
785 ZFS_DELEG_PERM_RECEIVE, cr)) != 0)
786 return (error);
787
788 if ((error = zfs_secpolicy_write_perms(zc->zc_name,
789 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
790 return (error);
791
792 return (zfs_secpolicy_write_perms(zc->zc_name,
793 ZFS_DELEG_PERM_CREATE, cr));
794 }
795
796 int
797 zfs_secpolicy_snapshot_perms(const char *name, cred_t *cr)
798 {
799 return (zfs_secpolicy_write_perms(name,
800 ZFS_DELEG_PERM_SNAPSHOT, cr));
801 }
802
803 static int
804 zfs_secpolicy_snapshot(zfs_cmd_t *zc, cred_t *cr)
805 {
806
807 return (zfs_secpolicy_snapshot_perms(zc->zc_name, cr));
808 }
809
810 static int
811 zfs_secpolicy_create(zfs_cmd_t *zc, cred_t *cr)
812 {
813 char parentname[MAXNAMELEN];
814 int error;
815
816 if ((error = zfs_get_parent(zc->zc_name, parentname,
817 sizeof (parentname))) != 0)
818 return (error);
819
820 if (zc->zc_value[0] != '\0') {
821 if ((error = zfs_secpolicy_write_perms(zc->zc_value,
822 ZFS_DELEG_PERM_CLONE, cr)) != 0)
823 return (error);
824 }
825
826 if ((error = zfs_secpolicy_write_perms(parentname,
827 ZFS_DELEG_PERM_CREATE, cr)) != 0)
828 return (error);
829
830 error = zfs_secpolicy_write_perms(parentname,
831 ZFS_DELEG_PERM_MOUNT, cr);
832
833 return (error);
834 }
835
836 static int
837 zfs_secpolicy_umount(zfs_cmd_t *zc, cred_t *cr)
838 {
839 int error;
840
841 error = secpolicy_fs_unmount(cr, NULL);
842 if (error) {
843 error = dsl_deleg_access(zc->zc_name, ZFS_DELEG_PERM_MOUNT, cr);
844 }
845 return (error);
846 }
847
848 /*
849 * Policy for pool operations - create/destroy pools, add vdevs, etc. Requires
850 * SYS_CONFIG privilege, which is not available in a local zone.
851 */
852 /* ARGSUSED */
853 static int
854 zfs_secpolicy_config(zfs_cmd_t *zc, cred_t *cr)
855 {
856 if (secpolicy_sys_config(cr, B_FALSE) != 0)
857 return (EPERM);
858
859 return (0);
860 }
861
862 /*
863 * Policy for object to name lookups.
864 */
865 /* ARGSUSED */
866 static int
867 zfs_secpolicy_diff(zfs_cmd_t *zc, cred_t *cr)
868 {
869 int error;
870
871 if ((error = secpolicy_sys_config(cr, B_FALSE)) == 0)
872 return (0);
873
874 error = zfs_secpolicy_write_perms(zc->zc_name, ZFS_DELEG_PERM_DIFF, cr);
875 return (error);
876 }
877
878 /*
879 * Policy for fault injection. Requires all privileges.
880 */
881 /* ARGSUSED */
882 static int
883 zfs_secpolicy_inject(zfs_cmd_t *zc, cred_t *cr)
884 {
885 return (secpolicy_zinject(cr));
886 }
887
888 static int
889 zfs_secpolicy_inherit(zfs_cmd_t *zc, cred_t *cr)
890 {
891 zfs_prop_t prop = zfs_name_to_prop(zc->zc_value);
892
893 if (prop == ZPROP_INVAL) {
894 if (!zfs_prop_user(zc->zc_value))
895 return (EINVAL);
896 return (zfs_secpolicy_write_perms(zc->zc_name,
897 ZFS_DELEG_PERM_USERPROP, cr));
898 } else {
899 return (zfs_secpolicy_setprop(zc->zc_name, prop,
900 NULL, cr));
901 }
902 }
903
904 static int
905 zfs_secpolicy_userspace_one(zfs_cmd_t *zc, cred_t *cr)
906 {
907 int err = zfs_secpolicy_read(zc, cr);
908 if (err)
909 return (err);
910
911 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
912 return (EINVAL);
913
914 if (zc->zc_value[0] == 0) {
915 /*
916 * They are asking about a posix uid/gid. If it's
917 * themself, allow it.
918 */
919 if (zc->zc_objset_type == ZFS_PROP_USERUSED ||
920 zc->zc_objset_type == ZFS_PROP_USERQUOTA) {
921 if (zc->zc_guid == crgetuid(cr))
922 return (0);
923 } else {
924 if (groupmember(zc->zc_guid, cr))
925 return (0);
926 }
927 }
928
929 return (zfs_secpolicy_write_perms(zc->zc_name,
930 userquota_perms[zc->zc_objset_type], cr));
931 }
932
933 static int
934 zfs_secpolicy_userspace_many(zfs_cmd_t *zc, cred_t *cr)
935 {
936 int err = zfs_secpolicy_read(zc, cr);
937 if (err)
938 return (err);
939
940 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
941 return (EINVAL);
942
943 return (zfs_secpolicy_write_perms(zc->zc_name,
944 userquota_perms[zc->zc_objset_type], cr));
945 }
946
947 static int
948 zfs_secpolicy_userspace_upgrade(zfs_cmd_t *zc, cred_t *cr)
949 {
950 return (zfs_secpolicy_setprop(zc->zc_name, ZFS_PROP_VERSION,
951 NULL, cr));
952 }
953
954 static int
955 zfs_secpolicy_hold(zfs_cmd_t *zc, cred_t *cr)
956 {
957 return (zfs_secpolicy_write_perms(zc->zc_name,
958 ZFS_DELEG_PERM_HOLD, cr));
959 }
960
961 static int
962 zfs_secpolicy_release(zfs_cmd_t *zc, cred_t *cr)
963 {
964 return (zfs_secpolicy_write_perms(zc->zc_name,
965 ZFS_DELEG_PERM_RELEASE, cr));
966 }
967
968 /*
969 * Policy for allowing temporary snapshots to be taken or released
970 */
971 static int
972 zfs_secpolicy_tmp_snapshot(zfs_cmd_t *zc, cred_t *cr)
973 {
974 /*
975 * A temporary snapshot is the same as a snapshot,
976 * hold, destroy and release all rolled into one.
977 * Delegated diff alone is sufficient that we allow this.
978 */
979 int error;
980
981 if ((error = zfs_secpolicy_write_perms(zc->zc_name,
982 ZFS_DELEG_PERM_DIFF, cr)) == 0)
983 return (0);
984
985 error = zfs_secpolicy_snapshot(zc, cr);
986 if (!error)
987 error = zfs_secpolicy_hold(zc, cr);
988 if (!error)
989 error = zfs_secpolicy_release(zc, cr);
990 if (!error)
991 error = zfs_secpolicy_destroy(zc, cr);
992 return (error);
993 }
994
995 /*
996 * Returns the nvlist as specified by the user in the zfs_cmd_t.
997 */
998 static int
999 get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp)
1000 {
1001 char *packed;
1002 int error;
1003 nvlist_t *list = NULL;
1004
1005 /*
1006 * Read in and unpack the user-supplied nvlist.
1007 */
1008 if (size == 0)
1009 return (EINVAL);
1010
1011 packed = kmem_alloc(size, KM_SLEEP);
1012
1013 if ((error = ddi_copyin((void *)(uintptr_t)nvl, packed, size,
1014 iflag)) != 0) {
1015 kmem_free(packed, size);
1016 return (error);
1017 }
1018
1019 if ((error = nvlist_unpack(packed, size, &list, 0)) != 0) {
1020 kmem_free(packed, size);
1021 return (error);
1022 }
1023
1024 kmem_free(packed, size);
1025
1026 *nvp = list;
1027 return (0);
1028 }
1029
1030 static int
1031 fit_error_list(zfs_cmd_t *zc, nvlist_t **errors)
1032 {
1033 size_t size;
1034
1035 VERIFY(nvlist_size(*errors, &size, NV_ENCODE_NATIVE) == 0);
1036
1037 if (size > zc->zc_nvlist_dst_size) {
1038 nvpair_t *more_errors;
1039 int n = 0;
1040
1041 if (zc->zc_nvlist_dst_size < 1024)
1042 return (ENOMEM);
1043
1044 VERIFY(nvlist_add_int32(*errors, ZPROP_N_MORE_ERRORS, 0) == 0);
1045 more_errors = nvlist_prev_nvpair(*errors, NULL);
1046
1047 do {
1048 nvpair_t *pair = nvlist_prev_nvpair(*errors,
1049 more_errors);
1050 VERIFY(nvlist_remove_nvpair(*errors, pair) == 0);
1051 n++;
1052 VERIFY(nvlist_size(*errors, &size,
1053 NV_ENCODE_NATIVE) == 0);
1054 } while (size > zc->zc_nvlist_dst_size);
1055
1056 VERIFY(nvlist_remove_nvpair(*errors, more_errors) == 0);
1057 VERIFY(nvlist_add_int32(*errors, ZPROP_N_MORE_ERRORS, n) == 0);
1058 ASSERT(nvlist_size(*errors, &size, NV_ENCODE_NATIVE) == 0);
1059 ASSERT(size <= zc->zc_nvlist_dst_size);
1060 }
1061
1062 return (0);
1063 }
1064
1065 static int
1066 put_nvlist(zfs_cmd_t *zc, nvlist_t *nvl)
1067 {
1068 char *packed = NULL;
1069 int error = 0;
1070 size_t size;
1071
1072 VERIFY(nvlist_size(nvl, &size, NV_ENCODE_NATIVE) == 0);
1073
1074 if (size > zc->zc_nvlist_dst_size) {
1075 error = ENOMEM;
1076 } else {
1077 packed = kmem_alloc(size, KM_SLEEP);
1078 VERIFY(nvlist_pack(nvl, &packed, &size, NV_ENCODE_NATIVE,
1079 KM_SLEEP) == 0);
1080 if (ddi_copyout(packed, (void *)(uintptr_t)zc->zc_nvlist_dst,
1081 size, zc->zc_iflags) != 0)
1082 error = EFAULT;
1083 kmem_free(packed, size);
1084 }
1085
1086 zc->zc_nvlist_dst_size = size;
1087 return (error);
1088 }
1089
1090 static int
1091 getzfsvfs(const char *dsname, zfsvfs_t **zfvp)
1092 {
1093 objset_t *os;
1094 int error;
1095
1096 error = dmu_objset_hold(dsname, FTAG, &os);
1097 if (error)
1098 return (error);
1099 if (dmu_objset_type(os) != DMU_OST_ZFS) {
1100 dmu_objset_rele(os, FTAG);
1101 return (EINVAL);
1102 }
1103
1104 mutex_enter(&os->os_user_ptr_lock);
1105 *zfvp = dmu_objset_get_user(os);
1106 if (*zfvp) {
1107 VFS_HOLD((*zfvp)->z_vfs);
1108 } else {
1109 error = ESRCH;
1110 }
1111 mutex_exit(&os->os_user_ptr_lock);
1112 dmu_objset_rele(os, FTAG);
1113 return (error);
1114 }
1115
1116 /*
1117 * Find a zfsvfs_t for a mounted filesystem, or create our own, in which
1118 * case its z_vfs will be NULL, and it will be opened as the owner.
1119 */
1120 static int
1121 zfsvfs_hold(const char *name, void *tag, zfsvfs_t **zfvp, boolean_t writer)
1122 {
1123 int error = 0;
1124
1125 if (getzfsvfs(name, zfvp) != 0)
1126 error = zfsvfs_create(name, zfvp);
1127 if (error == 0) {
1128 rrw_enter(&(*zfvp)->z_teardown_lock, (writer) ? RW_WRITER :
1129 RW_READER, tag);
1130 if ((*zfvp)->z_unmounted) {
1131 /*
1132 * XXX we could probably try again, since the unmounting
1133 * thread should be just about to disassociate the
1134 * objset from the zfsvfs.
1135 */
1136 rrw_exit(&(*zfvp)->z_teardown_lock, tag);
1137 return (EBUSY);
1138 }
1139 }
1140 return (error);
1141 }
1142
1143 static void
1144 zfsvfs_rele(zfsvfs_t *zfsvfs, void *tag)
1145 {
1146 rrw_exit(&zfsvfs->z_teardown_lock, tag);
1147
1148 if (zfsvfs->z_vfs) {
1149 VFS_RELE(zfsvfs->z_vfs);
1150 } else {
1151 dmu_objset_disown(zfsvfs->z_os, zfsvfs);
1152 zfsvfs_free(zfsvfs);
1153 }
1154 }
1155
1156 static int
1157 zfs_ioc_pool_create(zfs_cmd_t *zc)
1158 {
1159 int error;
1160 nvlist_t *config, *props = NULL;
1161 nvlist_t *rootprops = NULL;
1162 nvlist_t *zplprops = NULL;
1163 char *buf;
1164
1165 if (error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1166 zc->zc_iflags, &config))
1167 return (error);
1168
1169 if (zc->zc_nvlist_src_size != 0 && (error =
1170 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1171 zc->zc_iflags, &props))) {
1172 nvlist_free(config);
1173 return (error);
1174 }
1175
1176 if (props) {
1177 nvlist_t *nvl = NULL;
1178 uint64_t version = SPA_VERSION;
1179
1180 (void) nvlist_lookup_uint64(props,
1181 zpool_prop_to_name(ZPOOL_PROP_VERSION), &version);
1182 if (version < SPA_VERSION_INITIAL || version > SPA_VERSION) {
1183 error = EINVAL;
1184 goto pool_props_bad;
1185 }
1186 (void) nvlist_lookup_nvlist(props, ZPOOL_ROOTFS_PROPS, &nvl);
1187 if (nvl) {
1188 error = nvlist_dup(nvl, &rootprops, KM_SLEEP);
1189 if (error != 0) {
1190 nvlist_free(config);
1191 nvlist_free(props);
1192 return (error);
1193 }
1194 (void) nvlist_remove_all(props, ZPOOL_ROOTFS_PROPS);
1195 }
1196 VERIFY(nvlist_alloc(&zplprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
1197 error = zfs_fill_zplprops_root(version, rootprops,
1198 zplprops, NULL);
1199 if (error)
1200 goto pool_props_bad;
1201 }
1202
1203 buf = history_str_get(zc);
1204
1205 error = spa_create(zc->zc_name, config, props, buf, zplprops);
1206
1207 /*
1208 * Set the remaining root properties
1209 */
1210 if (!error && (error = zfs_set_prop_nvlist(zc->zc_name,
1211 ZPROP_SRC_LOCAL, rootprops, NULL)) != 0)
1212 (void) spa_destroy(zc->zc_name);
1213
1214 if (buf != NULL)
1215 history_str_free(buf);
1216
1217 pool_props_bad:
1218 nvlist_free(rootprops);
1219 nvlist_free(zplprops);
1220 nvlist_free(config);
1221 nvlist_free(props);
1222
1223 return (error);
1224 }
1225
1226 static int
1227 zfs_ioc_pool_destroy(zfs_cmd_t *zc)
1228 {
1229 int error;
1230 zfs_log_history(zc);
1231 error = spa_destroy(zc->zc_name);
1232 if (error == 0)
1233 zvol_remove_minors(zc->zc_name);
1234 return (error);
1235 }
1236
1237 static int
1238 zfs_ioc_pool_import(zfs_cmd_t *zc)
1239 {
1240 nvlist_t *config, *props = NULL;
1241 uint64_t guid;
1242 int error;
1243
1244 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1245 zc->zc_iflags, &config)) != 0)
1246 return (error);
1247
1248 if (zc->zc_nvlist_src_size != 0 && (error =
1249 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1250 zc->zc_iflags, &props))) {
1251 nvlist_free(config);
1252 return (error);
1253 }
1254
1255 if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, &guid) != 0 ||
1256 guid != zc->zc_guid)
1257 error = EINVAL;
1258 else
1259 error = spa_import(zc->zc_name, config, props, zc->zc_cookie);
1260
1261 if (zc->zc_nvlist_dst != 0) {
1262 int err;
1263
1264 if ((err = put_nvlist(zc, config)) != 0)
1265 error = err;
1266 }
1267
1268 nvlist_free(config);
1269
1270 if (props)
1271 nvlist_free(props);
1272
1273 return (error);
1274 }
1275
1276 static int
1277 zfs_ioc_pool_export(zfs_cmd_t *zc)
1278 {
1279 int error;
1280 boolean_t force = (boolean_t)zc->zc_cookie;
1281 boolean_t hardforce = (boolean_t)zc->zc_guid;
1282
1283 zfs_log_history(zc);
1284 error = spa_export(zc->zc_name, NULL, force, hardforce);
1285 if (error == 0)
1286 zvol_remove_minors(zc->zc_name);
1287 return (error);
1288 }
1289
1290 static int
1291 zfs_ioc_pool_configs(zfs_cmd_t *zc)
1292 {
1293 nvlist_t *configs;
1294 int error;
1295
1296 if ((configs = spa_all_configs(&zc->zc_cookie)) == NULL)
1297 return (EEXIST);
1298
1299 error = put_nvlist(zc, configs);
1300
1301 nvlist_free(configs);
1302
1303 return (error);
1304 }
1305
1306 static int
1307 zfs_ioc_pool_stats(zfs_cmd_t *zc)
1308 {
1309 nvlist_t *config;
1310 int error;
1311 int ret = 0;
1312
1313 error = spa_get_stats(zc->zc_name, &config, zc->zc_value,
1314 sizeof (zc->zc_value));
1315
1316 if (config != NULL) {
1317 ret = put_nvlist(zc, config);
1318 nvlist_free(config);
1319
1320 /*
1321 * The config may be present even if 'error' is non-zero.
1322 * In this case we return success, and preserve the real errno
1323 * in 'zc_cookie'.
1324 */
1325 zc->zc_cookie = error;
1326 } else {
1327 ret = error;
1328 }
1329
1330 return (ret);
1331 }
1332
1333 /*
1334 * Try to import the given pool, returning pool stats as appropriate so that
1335 * user land knows which devices are available and overall pool health.
1336 */
1337 static int
1338 zfs_ioc_pool_tryimport(zfs_cmd_t *zc)
1339 {
1340 nvlist_t *tryconfig, *config;
1341 int error;
1342
1343 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1344 zc->zc_iflags, &tryconfig)) != 0)
1345 return (error);
1346
1347 config = spa_tryimport(tryconfig);
1348
1349 nvlist_free(tryconfig);
1350
1351 if (config == NULL)
1352 return (EINVAL);
1353
1354 error = put_nvlist(zc, config);
1355 nvlist_free(config);
1356
1357 return (error);
1358 }
1359
1360 /*
1361 * inputs:
1362 * zc_name name of the pool
1363 * zc_cookie scan func (pool_scan_func_t)
1364 */
1365 static int
1366 zfs_ioc_pool_scan(zfs_cmd_t *zc)
1367 {
1368 spa_t *spa;
1369 int error;
1370
1371 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1372 return (error);
1373
1374 if (zc->zc_cookie == POOL_SCAN_NONE)
1375 error = spa_scan_stop(spa);
1376 else
1377 error = spa_scan(spa, zc->zc_cookie);
1378
1379 spa_close(spa, FTAG);
1380
1381 return (error);
1382 }
1383
1384 static int
1385 zfs_ioc_pool_freeze(zfs_cmd_t *zc)
1386 {
1387 spa_t *spa;
1388 int error;
1389
1390 error = spa_open(zc->zc_name, &spa, FTAG);
1391 if (error == 0) {
1392 spa_freeze(spa);
1393 spa_close(spa, FTAG);
1394 }
1395 return (error);
1396 }
1397
1398 static int
1399 zfs_ioc_pool_upgrade(zfs_cmd_t *zc)
1400 {
1401 spa_t *spa;
1402 int error;
1403
1404 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1405 return (error);
1406
1407 if (zc->zc_cookie < spa_version(spa) || zc->zc_cookie > SPA_VERSION) {
1408 spa_close(spa, FTAG);
1409 return (EINVAL);
1410 }
1411
1412 spa_upgrade(spa, zc->zc_cookie);
1413 spa_close(spa, FTAG);
1414
1415 return (error);
1416 }
1417
1418 static int
1419 zfs_ioc_pool_get_history(zfs_cmd_t *zc)
1420 {
1421 spa_t *spa;
1422 char *hist_buf;
1423 uint64_t size;
1424 int error;
1425
1426 if ((size = zc->zc_history_len) == 0)
1427 return (EINVAL);
1428
1429 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1430 return (error);
1431
1432 if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) {
1433 spa_close(spa, FTAG);
1434 return (ENOTSUP);
1435 }
1436
1437 hist_buf = kmem_alloc(size, KM_SLEEP);
1438 if ((error = spa_history_get(spa, &zc->zc_history_offset,
1439 &zc->zc_history_len, hist_buf)) == 0) {
1440 error = ddi_copyout(hist_buf,
1441 (void *)(uintptr_t)zc->zc_history,
1442 zc->zc_history_len, zc->zc_iflags);
1443 }
1444
1445 spa_close(spa, FTAG);
1446 kmem_free(hist_buf, size);
1447 return (error);
1448 }
1449
1450 static int
1451 zfs_ioc_dsobj_to_dsname(zfs_cmd_t *zc)
1452 {
1453 int error;
1454
1455 if (error = dsl_dsobj_to_dsname(zc->zc_name, zc->zc_obj, zc->zc_value))
1456 return (error);
1457
1458 return (0);
1459 }
1460
1461 /*
1462 * inputs:
1463 * zc_name name of filesystem
1464 * zc_obj object to find
1465 *
1466 * outputs:
1467 * zc_value name of object
1468 */
1469 static int
1470 zfs_ioc_obj_to_path(zfs_cmd_t *zc)
1471 {
1472 objset_t *os;
1473 int error;
1474
1475 /* XXX reading from objset not owned */
1476 if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os)) != 0)
1477 return (error);
1478 if (dmu_objset_type(os) != DMU_OST_ZFS) {
1479 dmu_objset_rele(os, FTAG);
1480 return (EINVAL);
1481 }
1482 error = zfs_obj_to_path(os, zc->zc_obj, zc->zc_value,
1483 sizeof (zc->zc_value));
1484 dmu_objset_rele(os, FTAG);
1485
1486 return (error);
1487 }
1488
1489 /*
1490 * inputs:
1491 * zc_name name of filesystem
1492 * zc_obj object to find
1493 *
1494 * outputs:
1495 * zc_stat stats on object
1496 * zc_value path to object
1497 */
1498 static int
1499 zfs_ioc_obj_to_stats(zfs_cmd_t *zc)
1500 {
1501 objset_t *os;
1502 int error;
1503
1504 /* XXX reading from objset not owned */
1505 if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os)) != 0)
1506 return (error);
1507 if (dmu_objset_type(os) != DMU_OST_ZFS) {
1508 dmu_objset_rele(os, FTAG);
1509 return (EINVAL);
1510 }
1511 error = zfs_obj_to_stats(os, zc->zc_obj, &zc->zc_stat, zc->zc_value,
1512 sizeof (zc->zc_value));
1513 dmu_objset_rele(os, FTAG);
1514
1515 return (error);
1516 }
1517
1518 static int
1519 zfs_ioc_vdev_add(zfs_cmd_t *zc)
1520 {
1521 spa_t *spa;
1522 int error;
1523 nvlist_t *config, **l2cache, **spares;
1524 uint_t nl2cache = 0, nspares = 0;
1525
1526 error = spa_open(zc->zc_name, &spa, FTAG);
1527 if (error != 0)
1528 return (error);
1529
1530 error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1531 zc->zc_iflags, &config);
1532 (void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_L2CACHE,
1533 &l2cache, &nl2cache);
1534
1535 (void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_SPARES,
1536 &spares, &nspares);
1537
1538 /*
1539 * A root pool with concatenated devices is not supported.
1540 * Thus, can not add a device to a root pool.
1541 *
1542 * Intent log device can not be added to a rootpool because
1543 * during mountroot, zil is replayed, a seperated log device
1544 * can not be accessed during the mountroot time.
1545 *
1546 * l2cache and spare devices are ok to be added to a rootpool.
1547 */
1548 if (spa_bootfs(spa) != 0 && nl2cache == 0 && nspares == 0) {
1549 nvlist_free(config);
1550 spa_close(spa, FTAG);
1551 return (EDOM);
1552 }
1553
1554 if (error == 0) {
1555 error = spa_vdev_add(spa, config);
1556 nvlist_free(config);
1557 }
1558 spa_close(spa, FTAG);
1559 return (error);
1560 }
1561
1562 /*
1563 * inputs:
1564 * zc_name name of the pool
1565 * zc_nvlist_conf nvlist of devices to remove
1566 * zc_cookie to stop the remove?
1567 */
1568 static int
1569 zfs_ioc_vdev_remove(zfs_cmd_t *zc)
1570 {
1571 spa_t *spa;
1572 int error;
1573
1574 error = spa_open(zc->zc_name, &spa, FTAG);
1575 if (error != 0)
1576 return (error);
1577 error = spa_vdev_remove(spa, zc->zc_guid, B_FALSE);
1578 spa_close(spa, FTAG);
1579 return (error);
1580 }
1581
1582 static int
1583 zfs_ioc_vdev_set_state(zfs_cmd_t *zc)
1584 {
1585 spa_t *spa;
1586 int error;
1587 vdev_state_t newstate = VDEV_STATE_UNKNOWN;
1588
1589 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1590 return (error);
1591 switch (zc->zc_cookie) {
1592 case VDEV_STATE_ONLINE:
1593 error = vdev_online(spa, zc->zc_guid, zc->zc_obj, &newstate);
1594 break;
1595
1596 case VDEV_STATE_OFFLINE:
1597 error = vdev_offline(spa, zc->zc_guid, zc->zc_obj);
1598 break;
1599
1600 case VDEV_STATE_FAULTED:
1601 if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED &&
1602 zc->zc_obj != VDEV_AUX_EXTERNAL)
1603 zc->zc_obj = VDEV_AUX_ERR_EXCEEDED;
1604
1605 error = vdev_fault(spa, zc->zc_guid, zc->zc_obj);
1606 break;
1607
1608 case VDEV_STATE_DEGRADED:
1609 if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED &&
1610 zc->zc_obj != VDEV_AUX_EXTERNAL)
1611 zc->zc_obj = VDEV_AUX_ERR_EXCEEDED;
1612
1613 error = vdev_degrade(spa, zc->zc_guid, zc->zc_obj);
1614 break;
1615
1616 default:
1617 error = EINVAL;
1618 }
1619 zc->zc_cookie = newstate;
1620 spa_close(spa, FTAG);
1621 return (error);
1622 }
1623
1624 static int
1625 zfs_ioc_vdev_attach(zfs_cmd_t *zc)
1626 {
1627 spa_t *spa;
1628 int replacing = zc->zc_cookie;
1629 nvlist_t *config;
1630 int error;
1631
1632 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1633 return (error);
1634
1635 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1636 zc->zc_iflags, &config)) == 0) {
1637 error = spa_vdev_attach(spa, zc->zc_guid, config, replacing);
1638 nvlist_free(config);
1639 }
1640
1641 spa_close(spa, FTAG);
1642 return (error);
1643 }
1644
1645 static int
1646 zfs_ioc_vdev_detach(zfs_cmd_t *zc)
1647 {
1648 spa_t *spa;
1649 int error;
1650
1651 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1652 return (error);
1653
1654 error = spa_vdev_detach(spa, zc->zc_guid, 0, B_FALSE);
1655
1656 spa_close(spa, FTAG);
1657 return (error);
1658 }
1659
1660 static int
1661 zfs_ioc_vdev_split(zfs_cmd_t *zc)
1662 {
1663 spa_t *spa;
1664 nvlist_t *config, *props = NULL;
1665 int error;
1666 boolean_t exp = !!(zc->zc_cookie & ZPOOL_EXPORT_AFTER_SPLIT);
1667
1668 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1669 return (error);
1670
1671 if (error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1672 zc->zc_iflags, &config)) {
1673 spa_close(spa, FTAG);
1674 return (error);
1675 }
1676
1677 if (zc->zc_nvlist_src_size != 0 && (error =
1678 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1679 zc->zc_iflags, &props))) {
1680 spa_close(spa, FTAG);
1681 nvlist_free(config);
1682 return (error);
1683 }
1684
1685 error = spa_vdev_split_mirror(spa, zc->zc_string, config, props, exp);
1686
1687 spa_close(spa, FTAG);
1688
1689 nvlist_free(config);
1690 nvlist_free(props);
1691
1692 return (error);
1693 }
1694
1695 static int
1696 zfs_ioc_vdev_setpath(zfs_cmd_t *zc)
1697 {
1698 spa_t *spa;
1699 char *path = zc->zc_value;
1700 uint64_t guid = zc->zc_guid;
1701 int error;
1702
1703 error = spa_open(zc->zc_name, &spa, FTAG);
1704 if (error != 0)
1705 return (error);
1706
1707 error = spa_vdev_setpath(spa, guid, path);
1708 spa_close(spa, FTAG);
1709 return (error);
1710 }
1711
1712 static int
1713 zfs_ioc_vdev_setfru(zfs_cmd_t *zc)
1714 {
1715 spa_t *spa;
1716 char *fru = zc->zc_value;
1717 uint64_t guid = zc->zc_guid;
1718 int error;
1719
1720 error = spa_open(zc->zc_name, &spa, FTAG);
1721 if (error != 0)
1722 return (error);
1723
1724 error = spa_vdev_setfru(spa, guid, fru);
1725 spa_close(spa, FTAG);
1726 return (error);
1727 }
1728
1729 static int
1730 zfs_ioc_objset_stats_impl(zfs_cmd_t *zc, objset_t *os)
1731 {
1732 int error = 0;
1733 nvlist_t *nv;
1734
1735 dmu_objset_fast_stat(os, &zc->zc_objset_stats);
1736
1737 if (zc->zc_nvlist_dst != 0 &&
1738 (error = dsl_prop_get_all(os, &nv)) == 0) {
1739 dmu_objset_stats(os, nv);
1740 /*
1741 * NB: zvol_get_stats() will read the objset contents,
1742 * which we aren't supposed to do with a
1743 * DS_MODE_USER hold, because it could be
1744 * inconsistent. So this is a bit of a workaround...
1745 * XXX reading with out owning
1746 */
1747 if (!zc->zc_objset_stats.dds_inconsistent) {
1748 if (dmu_objset_type(os) == DMU_OST_ZVOL)
1749 VERIFY(zvol_get_stats(os, nv) == 0);
1750 }
1751 error = put_nvlist(zc, nv);
1752 nvlist_free(nv);
1753 }
1754
1755 return (error);
1756 }
1757
1758 /*
1759 * inputs:
1760 * zc_name name of filesystem
1761 * zc_nvlist_dst_size size of buffer for property nvlist
1762 *
1763 * outputs:
1764 * zc_objset_stats stats
1765 * zc_nvlist_dst property nvlist
1766 * zc_nvlist_dst_size size of property nvlist
1767 */
1768 static int
1769 zfs_ioc_objset_stats(zfs_cmd_t *zc)
1770 {
1771 objset_t *os = NULL;
1772 int error;
1773
1774 if (error = dmu_objset_hold(zc->zc_name, FTAG, &os))
1775 return (error);
1776
1777 error = zfs_ioc_objset_stats_impl(zc, os);
1778
1779 dmu_objset_rele(os, FTAG);
1780
1781 return (error);
1782 }
1783
1784 /*
1785 * inputs:
1786 * zc_name name of filesystem
1787 * zc_nvlist_dst_size size of buffer for property nvlist
1788 *
1789 * outputs:
1790 * zc_nvlist_dst received property nvlist
1791 * zc_nvlist_dst_size size of received property nvlist
1792 *
1793 * Gets received properties (distinct from local properties on or after
1794 * SPA_VERSION_RECVD_PROPS) for callers who want to differentiate received from
1795 * local property values.
1796 */
1797 static int
1798 zfs_ioc_objset_recvd_props(zfs_cmd_t *zc)
1799 {
1800 objset_t *os = NULL;
1801 int error;
1802 nvlist_t *nv;
1803
1804 if (error = dmu_objset_hold(zc->zc_name, FTAG, &os))
1805 return (error);
1806
1807 /*
1808 * Without this check, we would return local property values if the
1809 * caller has not already received properties on or after
1810 * SPA_VERSION_RECVD_PROPS.
1811 */
1812 if (!dsl_prop_get_hasrecvd(os)) {
1813 dmu_objset_rele(os, FTAG);
1814 return (ENOTSUP);
1815 }
1816
1817 if (zc->zc_nvlist_dst != 0 &&
1818 (error = dsl_prop_get_received(os, &nv)) == 0) {
1819 error = put_nvlist(zc, nv);
1820 nvlist_free(nv);
1821 }
1822
1823 dmu_objset_rele(os, FTAG);
1824 return (error);
1825 }
1826
1827 static int
1828 nvl_add_zplprop(objset_t *os, nvlist_t *props, zfs_prop_t prop)
1829 {
1830 uint64_t value;
1831 int error;
1832
1833 /*
1834 * zfs_get_zplprop() will either find a value or give us
1835 * the default value (if there is one).
1836 */
1837 if ((error = zfs_get_zplprop(os, prop, &value)) != 0)
1838 return (error);
1839 VERIFY(nvlist_add_uint64(props, zfs_prop_to_name(prop), value) == 0);
1840 return (0);
1841 }
1842
1843 /*
1844 * inputs:
1845 * zc_name name of filesystem
1846 * zc_nvlist_dst_size size of buffer for zpl property nvlist
1847 *
1848 * outputs:
1849 * zc_nvlist_dst zpl property nvlist
1850 * zc_nvlist_dst_size size of zpl property nvlist
1851 */
1852 static int
1853 zfs_ioc_objset_zplprops(zfs_cmd_t *zc)
1854 {
1855 objset_t *os;
1856 int err;
1857
1858 /* XXX reading without owning */
1859 if (err = dmu_objset_hold(zc->zc_name, FTAG, &os))
1860 return (err);
1861
1862 dmu_objset_fast_stat(os, &zc->zc_objset_stats);
1863
1864 /*
1865 * NB: nvl_add_zplprop() will read the objset contents,
1866 * which we aren't supposed to do with a DS_MODE_USER
1867 * hold, because it could be inconsistent.
1868 */
1869 if (zc->zc_nvlist_dst != NULL &&
1870 !zc->zc_objset_stats.dds_inconsistent &&
1871 dmu_objset_type(os) == DMU_OST_ZFS) {
1872 nvlist_t *nv;
1873
1874 VERIFY(nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) == 0);
1875 if ((err = nvl_add_zplprop(os, nv, ZFS_PROP_VERSION)) == 0 &&
1876 (err = nvl_add_zplprop(os, nv, ZFS_PROP_NORMALIZE)) == 0 &&
1877 (err = nvl_add_zplprop(os, nv, ZFS_PROP_UTF8ONLY)) == 0 &&
1878 (err = nvl_add_zplprop(os, nv, ZFS_PROP_CASE)) == 0)
1879 err = put_nvlist(zc, nv);
1880 nvlist_free(nv);
1881 } else {
1882 err = ENOENT;
1883 }
1884 dmu_objset_rele(os, FTAG);
1885 return (err);
1886 }
1887
1888 static boolean_t
1889 dataset_name_hidden(const char *name)
1890 {
1891 /*
1892 * Skip over datasets that are not visible in this zone,
1893 * internal datasets (which have a $ in their name), and
1894 * temporary datasets (which have a % in their name).
1895 */
1896 if (strchr(name, '$') != NULL)
1897 return (B_TRUE);
1898 if (strchr(name, '%') != NULL)
1899 return (B_TRUE);
1900 if (!INGLOBALZONE(curproc) && !zone_dataset_visible(name, NULL))
1901 return (B_TRUE);
1902 return (B_FALSE);
1903 }
1904
1905 /*
1906 * inputs:
1907 * zc_name name of filesystem
1908 * zc_cookie zap cursor
1909 * zc_nvlist_dst_size size of buffer for property nvlist
1910 *
1911 * outputs:
1912 * zc_name name of next filesystem
1913 * zc_cookie zap cursor
1914 * zc_objset_stats stats
1915 * zc_nvlist_dst property nvlist
1916 * zc_nvlist_dst_size size of property nvlist
1917 */
1918 static int
1919 zfs_ioc_dataset_list_next(zfs_cmd_t *zc)
1920 {
1921 objset_t *os;
1922 int error;
1923 char *p;
1924 size_t orig_len = strlen(zc->zc_name);
1925
1926 top:
1927 if (error = dmu_objset_hold(zc->zc_name, FTAG, &os)) {
1928 if (error == ENOENT)
1929 error = ESRCH;
1930 return (error);
1931 }
1932
1933 p = strrchr(zc->zc_name, '/');
1934 if (p == NULL || p[1] != '\0')
1935 (void) strlcat(zc->zc_name, "/", sizeof (zc->zc_name));
1936 p = zc->zc_name + strlen(zc->zc_name);
1937
1938 /*
1939 * Pre-fetch the datasets. dmu_objset_prefetch() always returns 0
1940 * but is not declared void because its called by dmu_objset_find().
1941 */
1942 if (zc->zc_cookie == 0) {
1943 uint64_t cookie = 0;
1944 int len = sizeof (zc->zc_name) - (p - zc->zc_name);
1945
1946 while (dmu_dir_list_next(os, len, p, NULL, &cookie) == 0)
1947 (void) dmu_objset_prefetch(p, NULL);
1948 }
1949
1950 do {
1951 error = dmu_dir_list_next(os,
1952 sizeof (zc->zc_name) - (p - zc->zc_name), p,
1953 NULL, &zc->zc_cookie);
1954 if (error == ENOENT)
1955 error = ESRCH;
1956 } while (error == 0 && dataset_name_hidden(zc->zc_name) &&
1957 !(zc->zc_iflags & FKIOCTL));
1958 dmu_objset_rele(os, FTAG);
1959
1960 /*
1961 * If it's an internal dataset (ie. with a '$' in its name),
1962 * don't try to get stats for it, otherwise we'll return ENOENT.
1963 */
1964 if (error == 0 && strchr(zc->zc_name, '$') == NULL) {
1965 error = zfs_ioc_objset_stats(zc); /* fill in the stats */
1966 if (error == ENOENT) {
1967 /* We lost a race with destroy, get the next one. */
1968 zc->zc_name[orig_len] = '\0';
1969 goto top;
1970 }
1971 }
1972 return (error);
1973 }
1974
1975 /*
1976 * inputs:
1977 * zc_name name of filesystem
1978 * zc_cookie zap cursor
1979 * zc_nvlist_dst_size size of buffer for property nvlist
1980 *
1981 * outputs:
1982 * zc_name name of next snapshot
1983 * zc_objset_stats stats
1984 * zc_nvlist_dst property nvlist
1985 * zc_nvlist_dst_size size of property nvlist
1986 */
1987 static int
1988 zfs_ioc_snapshot_list_next(zfs_cmd_t *zc)
1989 {
1990 objset_t *os;
1991 int error;
1992
1993 top:
1994 if (zc->zc_cookie == 0)
1995 (void) dmu_objset_find(zc->zc_name, dmu_objset_prefetch,
1996 NULL, DS_FIND_SNAPSHOTS);
1997
1998 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
1999 if (error)
2000 return (error == ENOENT ? ESRCH : error);
2001
2002 /*
2003 * A dataset name of maximum length cannot have any snapshots,
2004 * so exit immediately.
2005 */
2006 if (strlcat(zc->zc_name, "@", sizeof (zc->zc_name)) >= MAXNAMELEN) {
2007 dmu_objset_rele(os, FTAG);
2008 return (ESRCH);
2009 }
2010
2011 error = dmu_snapshot_list_next(os,
2012 sizeof (zc->zc_name) - strlen(zc->zc_name),
2013 zc->zc_name + strlen(zc->zc_name), &zc->zc_obj, &zc->zc_cookie,
2014 NULL);
2015
2016 if (error == 0) {
2017 dsl_dataset_t *ds;
2018 dsl_pool_t *dp = os->os_dsl_dataset->ds_dir->dd_pool;
2019
2020 /*
2021 * Since we probably don't have a hold on this snapshot,
2022 * it's possible that the objsetid could have been destroyed
2023 * and reused for a new objset. It's OK if this happens during
2024 * a zfs send operation, since the new createtxg will be
2025 * beyond the range we're interested in.
2026 */
2027 rw_enter(&dp->dp_config_rwlock, RW_READER);
2028 error = dsl_dataset_hold_obj(dp, zc->zc_obj, FTAG, &ds);
2029 rw_exit(&dp->dp_config_rwlock);
2030 if (error) {
2031 if (error == ENOENT) {
2032 /* Racing with destroy, get the next one. */
2033 *strchr(zc->zc_name, '@') = '\0';
2034 dmu_objset_rele(os, FTAG);
2035 goto top;
2036 }
2037 } else {
2038 objset_t *ossnap;
2039
2040 error = dmu_objset_from_ds(ds, &ossnap);
2041 if (error == 0)
2042 error = zfs_ioc_objset_stats_impl(zc, ossnap);
2043 dsl_dataset_rele(ds, FTAG);
2044 }
2045 } else if (error == ENOENT) {
2046 error = ESRCH;
2047 }
2048
2049 dmu_objset_rele(os, FTAG);
2050 /* if we failed, undo the @ that we tacked on to zc_name */
2051 if (error)
2052 *strchr(zc->zc_name, '@') = '\0';
2053 return (error);
2054 }
2055
2056 static int
2057 zfs_prop_set_userquota(const char *dsname, nvpair_t *pair)
2058 {
2059 const char *propname = nvpair_name(pair);
2060 uint64_t *valary;
2061 unsigned int vallen;
2062 const char *domain;
2063 char *dash;
2064 zfs_userquota_prop_t type;
2065 uint64_t rid;
2066 uint64_t quota;
2067 zfsvfs_t *zfsvfs;
2068 int err;
2069
2070 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2071 nvlist_t *attrs;
2072 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
2073 if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2074 &pair) != 0)
2075 return (EINVAL);
2076 }
2077
2078 /*
2079 * A correctly constructed propname is encoded as
2080 * userquota@<rid>-<domain>.
2081 */
2082 if ((dash = strchr(propname, '-')) == NULL ||
2083 nvpair_value_uint64_array(pair, &valary, &vallen) != 0 ||
2084 vallen != 3)
2085 return (EINVAL);
2086
2087 domain = dash + 1;
2088 type = valary[0];
2089 rid = valary[1];
2090 quota = valary[2];
2091
2092 err = zfsvfs_hold(dsname, FTAG, &zfsvfs, B_FALSE);
2093 if (err == 0) {
2094 err = zfs_set_userquota(zfsvfs, type, domain, rid, quota);
2095 zfsvfs_rele(zfsvfs, FTAG);
2096 }
2097
2098 return (err);
2099 }
2100
2101 /*
2102 * If the named property is one that has a special function to set its value,
2103 * return 0 on success and a positive error code on failure; otherwise if it is
2104 * not one of the special properties handled by this function, return -1.
2105 *
2106 * XXX: It would be better for callers of the property interface if we handled
2107 * these special cases in dsl_prop.c (in the dsl layer).
2108 */
2109 static int
2110 zfs_prop_set_special(const char *dsname, zprop_source_t source,
2111 nvpair_t *pair)
2112 {
2113 const char *propname = nvpair_name(pair);
2114 zfs_prop_t prop = zfs_name_to_prop(propname);
2115 uint64_t intval;
2116 int err;
2117
2118 if (prop == ZPROP_INVAL) {
2119 if (zfs_prop_userquota(propname))
2120 return (zfs_prop_set_userquota(dsname, pair));
2121 return (-1);
2122 }
2123
2124 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2125 nvlist_t *attrs;
2126 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
2127 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2128 &pair) == 0);
2129 }
2130
2131 if (zfs_prop_get_type(prop) == PROP_TYPE_STRING)
2132 return (-1);
2133
2134 VERIFY(0 == nvpair_value_uint64(pair, &intval));
2135
2136 switch (prop) {
2137 case ZFS_PROP_QUOTA:
2138 err = dsl_dir_set_quota(dsname, source, intval);
2139 break;
2140 case ZFS_PROP_REFQUOTA:
2141 err = dsl_dataset_set_quota(dsname, source, intval);
2142 break;
2143 case ZFS_PROP_RESERVATION:
2144 err = dsl_dir_set_reservation(dsname, source, intval);
2145 break;
2146 case ZFS_PROP_REFRESERVATION:
2147 err = dsl_dataset_set_reservation(dsname, source, intval);
2148 break;
2149 case ZFS_PROP_VOLSIZE:
2150 err = zvol_set_volsize(dsname, ddi_driver_major(zfs_dip),
2151 intval);
2152 break;
2153 case ZFS_PROP_VERSION:
2154 {
2155 zfsvfs_t *zfsvfs;
2156
2157 if ((err = zfsvfs_hold(dsname, FTAG, &zfsvfs, B_TRUE)) != 0)
2158 break;
2159
2160 err = zfs_set_version(zfsvfs, intval);
2161 zfsvfs_rele(zfsvfs, FTAG);
2162
2163 if (err == 0 && intval >= ZPL_VERSION_USERSPACE) {
2164 zfs_cmd_t *zc;
2165
2166 zc = kmem_zalloc(sizeof (zfs_cmd_t), KM_SLEEP);
2167 (void) strcpy(zc->zc_name, dsname);
2168 (void) zfs_ioc_userspace_upgrade(zc);
2169 kmem_free(zc, sizeof (zfs_cmd_t));
2170 }
2171 break;
2172 }
2173
2174 default:
2175 err = -1;
2176 }
2177
2178 return (err);
2179 }
2180
2181 /*
2182 * This function is best effort. If it fails to set any of the given properties,
2183 * it continues to set as many as it can and returns the first error
2184 * encountered. If the caller provides a non-NULL errlist, it also gives the
2185 * complete list of names of all the properties it failed to set along with the
2186 * corresponding error numbers. The caller is responsible for freeing the
2187 * returned errlist.
2188 *
2189 * If every property is set successfully, zero is returned and the list pointed
2190 * at by errlist is NULL.
2191 */
2192 int
2193 zfs_set_prop_nvlist(const char *dsname, zprop_source_t source, nvlist_t *nvl,
2194 nvlist_t **errlist)
2195 {
2196 nvpair_t *pair;
2197 nvpair_t *propval;
2198 int rv = 0;
2199 uint64_t intval;
2200 char *strval;
2201 nvlist_t *genericnvl;
2202 nvlist_t *errors;
2203 nvlist_t *retrynvl;
2204
2205 VERIFY(nvlist_alloc(&genericnvl, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2206 VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2207 VERIFY(nvlist_alloc(&retrynvl, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2208
2209 retry:
2210 pair = NULL;
2211 while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) {
2212 const char *propname = nvpair_name(pair);
2213 zfs_prop_t prop = zfs_name_to_prop(propname);
2214 int err = 0;
2215
2216 /* decode the property value */
2217 propval = pair;
2218 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2219 nvlist_t *attrs;
2220 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
2221 if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2222 &propval) != 0)
2223 err = EINVAL;
2224 }
2225
2226 /* Validate value type */
2227 if (err == 0 && prop == ZPROP_INVAL) {
2228 if (zfs_prop_user(propname)) {
2229 if (nvpair_type(propval) != DATA_TYPE_STRING)
2230 err = EINVAL;
2231 } else if (zfs_prop_userquota(propname)) {
2232 if (nvpair_type(propval) !=
2233 DATA_TYPE_UINT64_ARRAY)
2234 err = EINVAL;
2235 }
2236 } else if (err == 0) {
2237 if (nvpair_type(propval) == DATA_TYPE_STRING) {
2238 if (zfs_prop_get_type(prop) != PROP_TYPE_STRING)
2239 err = EINVAL;
2240 } else if (nvpair_type(propval) == DATA_TYPE_UINT64) {
2241 const char *unused;
2242
2243 VERIFY(nvpair_value_uint64(propval,
2244 &intval) == 0);
2245
2246 switch (zfs_prop_get_type(prop)) {
2247 case PROP_TYPE_NUMBER:
2248 break;
2249 case PROP_TYPE_STRING:
2250 err = EINVAL;
2251 break;
2252 case PROP_TYPE_INDEX:
2253 if (zfs_prop_index_to_string(prop,
2254 intval, &unused) != 0)
2255 err = EINVAL;
2256 break;
2257 default:
2258 cmn_err(CE_PANIC,
2259 "unknown property type");
2260 }
2261 } else {
2262 err = EINVAL;
2263 }
2264 }
2265
2266 /* Validate permissions */
2267 if (err == 0)
2268 err = zfs_check_settable(dsname, pair, CRED());
2269
2270 if (err == 0) {
2271 err = zfs_prop_set_special(dsname, source, pair);
2272 if (err == -1) {
2273 /*
2274 * For better performance we build up a list of
2275 * properties to set in a single transaction.
2276 */
2277 err = nvlist_add_nvpair(genericnvl, pair);
2278 } else if (err != 0 && nvl != retrynvl) {
2279 /*
2280 * This may be a spurious error caused by
2281 * receiving quota and reservation out of order.
2282 * Try again in a second pass.
2283 */
2284 err = nvlist_add_nvpair(retrynvl, pair);
2285 }
2286 }
2287
2288 if (err != 0)
2289 VERIFY(nvlist_add_int32(errors, propname, err) == 0);
2290 }
2291
2292 if (nvl != retrynvl && !nvlist_empty(retrynvl)) {
2293 nvl = retrynvl;
2294 goto retry;
2295 }
2296
2297 if (!nvlist_empty(genericnvl) &&
2298 dsl_props_set(dsname, source, genericnvl) != 0) {
2299 /*
2300 * If this fails, we still want to set as many properties as we
2301 * can, so try setting them individually.
2302 */
2303 pair = NULL;
2304 while ((pair = nvlist_next_nvpair(genericnvl, pair)) != NULL) {
2305 const char *propname = nvpair_name(pair);
2306 int err = 0;
2307
2308 propval = pair;
2309 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2310 nvlist_t *attrs;
2311 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
2312 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2313 &propval) == 0);
2314 }
2315
2316 if (nvpair_type(propval) == DATA_TYPE_STRING) {
2317 VERIFY(nvpair_value_string(propval,
2318 &strval) == 0);
2319 err = dsl_prop_set(dsname, propname, source, 1,
2320 strlen(strval) + 1, strval);
2321 } else {
2322 VERIFY(nvpair_value_uint64(propval,
2323 &intval) == 0);
2324 err = dsl_prop_set(dsname, propname, source, 8,
2325 1, &intval);
2326 }
2327
2328 if (err != 0) {
2329 VERIFY(nvlist_add_int32(errors, propname,
2330 err) == 0);
2331 }
2332 }
2333 }
2334 nvlist_free(genericnvl);
2335 nvlist_free(retrynvl);
2336
2337 if ((pair = nvlist_next_nvpair(errors, NULL)) == NULL) {
2338 nvlist_free(errors);
2339 errors = NULL;
2340 } else {
2341 VERIFY(nvpair_value_int32(pair, &rv) == 0);
2342 }
2343
2344 if (errlist == NULL)
2345 nvlist_free(errors);
2346 else
2347 *errlist = errors;
2348
2349 return (rv);
2350 }
2351
2352 /*
2353 * Check that all the properties are valid user properties.
2354 */
2355 static int
2356 zfs_check_userprops(char *fsname, nvlist_t *nvl)
2357 {
2358 nvpair_t *pair = NULL;
2359 int error = 0;
2360
2361 while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) {
2362 const char *propname = nvpair_name(pair);
2363 char *valstr;
2364
2365 if (!zfs_prop_user(propname) ||
2366 nvpair_type(pair) != DATA_TYPE_STRING)
2367 return (EINVAL);
2368
2369 if (error = zfs_secpolicy_write_perms(fsname,
2370 ZFS_DELEG_PERM_USERPROP, CRED()))
2371 return (error);
2372
2373 if (strlen(propname) >= ZAP_MAXNAMELEN)
2374 return (ENAMETOOLONG);
2375
2376 VERIFY(nvpair_value_string(pair, &valstr) == 0);
2377 if (strlen(valstr) >= ZAP_MAXVALUELEN)
2378 return (E2BIG);
2379 }
2380 return (0);
2381 }
2382
2383 static void
2384 props_skip(nvlist_t *props, nvlist_t *skipped, nvlist_t **newprops)
2385 {
2386 nvpair_t *pair;
2387
2388 VERIFY(nvlist_alloc(newprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2389
2390 pair = NULL;
2391 while ((pair = nvlist_next_nvpair(props, pair)) != NULL) {
2392 if (nvlist_exists(skipped, nvpair_name(pair)))
2393 continue;
2394
2395 VERIFY(nvlist_add_nvpair(*newprops, pair) == 0);
2396 }
2397 }
2398
2399 static int
2400 clear_received_props(objset_t *os, const char *fs, nvlist_t *props,
2401 nvlist_t *skipped)
2402 {
2403 int err = 0;
2404 nvlist_t *cleared_props = NULL;
2405 props_skip(props, skipped, &cleared_props);
2406 if (!nvlist_empty(cleared_props)) {
2407 /*
2408 * Acts on local properties until the dataset has received
2409 * properties at least once on or after SPA_VERSION_RECVD_PROPS.
2410 */
2411 zprop_source_t flags = (ZPROP_SRC_NONE |
2412 (dsl_prop_get_hasrecvd(os) ? ZPROP_SRC_RECEIVED : 0));
2413 err = zfs_set_prop_nvlist(fs, flags, cleared_props, NULL);
2414 }
2415 nvlist_free(cleared_props);
2416 return (err);
2417 }
2418
2419 /*
2420 * inputs:
2421 * zc_name name of filesystem
2422 * zc_value name of property to set
2423 * zc_nvlist_src{_size} nvlist of properties to apply
2424 * zc_cookie received properties flag
2425 *
2426 * outputs:
2427 * zc_nvlist_dst{_size} error for each unapplied received property
2428 */
2429 static int
2430 zfs_ioc_set_prop(zfs_cmd_t *zc)
2431 {
2432 nvlist_t *nvl;
2433 boolean_t received = zc->zc_cookie;
2434 zprop_source_t source = (received ? ZPROP_SRC_RECEIVED :
2435 ZPROP_SRC_LOCAL);
2436 nvlist_t *errors = NULL;
2437 int error;
2438
2439 if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2440 zc->zc_iflags, &nvl)) != 0)
2441 return (error);
2442
2443 if (received) {
2444 nvlist_t *origprops;
2445 objset_t *os;
2446
2447 if (dmu_objset_hold(zc->zc_name, FTAG, &os) == 0) {
2448 if (dsl_prop_get_received(os, &origprops) == 0) {
2449 (void) clear_received_props(os,
2450 zc->zc_name, origprops, nvl);
2451 nvlist_free(origprops);
2452 }
2453
2454 dsl_prop_set_hasrecvd(os);
2455 dmu_objset_rele(os, FTAG);
2456 }
2457 }
2458
2459 error = zfs_set_prop_nvlist(zc->zc_name, source, nvl, &errors);
2460
2461 if (zc->zc_nvlist_dst != NULL && errors != NULL) {
2462 (void) put_nvlist(zc, errors);
2463 }
2464
2465 nvlist_free(errors);
2466 nvlist_free(nvl);
2467 return (error);
2468 }
2469
2470 /*
2471 * inputs:
2472 * zc_name name of filesystem
2473 * zc_value name of property to inherit
2474 * zc_cookie revert to received value if TRUE
2475 *
2476 * outputs: none
2477 */
2478 static int
2479 zfs_ioc_inherit_prop(zfs_cmd_t *zc)
2480 {
2481 const char *propname = zc->zc_value;
2482 zfs_prop_t prop = zfs_name_to_prop(propname);
2483 boolean_t received = zc->zc_cookie;
2484 zprop_source_t source = (received
2485 ? ZPROP_SRC_NONE /* revert to received value, if any */
2486 : ZPROP_SRC_INHERITED); /* explicitly inherit */
2487
2488 if (received) {
2489 nvlist_t *dummy;
2490 nvpair_t *pair;
2491 zprop_type_t type;
2492 int err;
2493
2494 /*
2495 * zfs_prop_set_special() expects properties in the form of an
2496 * nvpair with type info.
2497 */
2498 if (prop == ZPROP_INVAL) {
2499 if (!zfs_prop_user(propname))
2500 return (EINVAL);
2501
2502 type = PROP_TYPE_STRING;
2503 } else if (prop == ZFS_PROP_VOLSIZE ||
2504 prop == ZFS_PROP_VERSION) {
2505 return (EINVAL);
2506 } else {
2507 type = zfs_prop_get_type(prop);
2508 }
2509
2510 VERIFY(nvlist_alloc(&dummy, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2511
2512 switch (type) {
2513 case PROP_TYPE_STRING:
2514 VERIFY(0 == nvlist_add_string(dummy, propname, ""));
2515 break;
2516 case PROP_TYPE_NUMBER:
2517 case PROP_TYPE_INDEX:
2518 VERIFY(0 == nvlist_add_uint64(dummy, propname, 0));
2519 break;
2520 default:
2521 nvlist_free(dummy);
2522 return (EINVAL);
2523 }
2524
2525 pair = nvlist_next_nvpair(dummy, NULL);
2526 err = zfs_prop_set_special(zc->zc_name, source, pair);
2527 nvlist_free(dummy);
2528 if (err != -1)
2529 return (err); /* special property already handled */
2530 } else {
2531 /*
2532 * Only check this in the non-received case. We want to allow
2533 * 'inherit -S' to revert non-inheritable properties like quota
2534 * and reservation to the received or default values even though
2535 * they are not considered inheritable.
2536 */
2537 if (prop != ZPROP_INVAL && !zfs_prop_inheritable(prop))
2538 return (EINVAL);
2539 }
2540
2541 /* the property name has been validated by zfs_secpolicy_inherit() */
2542 return (dsl_prop_set(zc->zc_name, zc->zc_value, source, 0, 0, NULL));
2543 }
2544
2545 static int
2546 zfs_ioc_pool_set_props(zfs_cmd_t *zc)
2547 {
2548 nvlist_t *props;
2549 spa_t *spa;
2550 int error;
2551 nvpair_t *pair;
2552
2553 if (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2554 zc->zc_iflags, &props))
2555 return (error);
2556
2557 /*
2558 * If the only property is the configfile, then just do a spa_lookup()
2559 * to handle the faulted case.
2560 */
2561 pair = nvlist_next_nvpair(props, NULL);
2562 if (pair != NULL && strcmp(nvpair_name(pair),
2563 zpool_prop_to_name(ZPOOL_PROP_CACHEFILE)) == 0 &&
2564 nvlist_next_nvpair(props, pair) == NULL) {
2565 mutex_enter(&spa_namespace_lock);
2566 if ((spa = spa_lookup(zc->zc_name)) != NULL) {
2567 spa_configfile_set(spa, props, B_FALSE);
2568 spa_config_sync(spa, B_FALSE, B_TRUE);
2569 }
2570 mutex_exit(&spa_namespace_lock);
2571 if (spa != NULL) {
2572 nvlist_free(props);
2573 return (0);
2574 }
2575 }
2576
2577 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) {
2578 nvlist_free(props);
2579 return (error);
2580 }
2581
2582 error = spa_prop_set(spa, props);
2583
2584 nvlist_free(props);
2585 spa_close(spa, FTAG);
2586
2587 return (error);
2588 }
2589
2590 static int
2591 zfs_ioc_pool_get_props(zfs_cmd_t *zc)
2592 {
2593 spa_t *spa;
2594 int error;
2595 nvlist_t *nvp = NULL;
2596
2597 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) {
2598 /*
2599 * If the pool is faulted, there may be properties we can still
2600 * get (such as altroot and cachefile), so attempt to get them
2601 * anyway.
2602 */
2603 mutex_enter(&spa_namespace_lock);
2604 if ((spa = spa_lookup(zc->zc_name)) != NULL)
2605 error = spa_prop_get(spa, &nvp);
2606 mutex_exit(&spa_namespace_lock);
2607 } else {
2608 error = spa_prop_get(spa, &nvp);
2609 spa_close(spa, FTAG);
2610 }
2611
2612 if (error == 0 && zc->zc_nvlist_dst != NULL)
2613 error = put_nvlist(zc, nvp);
2614 else
2615 error = EFAULT;
2616
2617 nvlist_free(nvp);
2618 return (error);
2619 }
2620
2621 /*
2622 * inputs:
2623 * zc_name name of filesystem
2624 * zc_nvlist_src{_size} nvlist of delegated permissions
2625 * zc_perm_action allow/unallow flag
2626 *
2627 * outputs: none
2628 */
2629 static int
2630 zfs_ioc_set_fsacl(zfs_cmd_t *zc)
2631 {
2632 int error;
2633 nvlist_t *fsaclnv = NULL;
2634
2635 if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2636 zc->zc_iflags, &fsaclnv)) != 0)
2637 return (error);
2638
2639 /*
2640 * Verify nvlist is constructed correctly
2641 */
2642 if ((error = zfs_deleg_verify_nvlist(fsaclnv)) != 0) {
2643 nvlist_free(fsaclnv);
2644 return (EINVAL);
2645 }
2646
2647 /*
2648 * If we don't have PRIV_SYS_MOUNT, then validate
2649 * that user is allowed to hand out each permission in
2650 * the nvlist(s)
2651 */
2652
2653 error = secpolicy_zfs(CRED());
2654 if (error) {
2655 if (zc->zc_perm_action == B_FALSE) {
2656 error = dsl_deleg_can_allow(zc->zc_name,
2657 fsaclnv, CRED());
2658 } else {
2659 error = dsl_deleg_can_unallow(zc->zc_name,
2660 fsaclnv, CRED());
2661 }
2662 }
2663
2664 if (error == 0)
2665 error = dsl_deleg_set(zc->zc_name, fsaclnv, zc->zc_perm_action);
2666
2667 nvlist_free(fsaclnv);
2668 return (error);
2669 }
2670
2671 /*
2672 * inputs:
2673 * zc_name name of filesystem
2674 *
2675 * outputs:
2676 * zc_nvlist_src{_size} nvlist of delegated permissions
2677 */
2678 static int
2679 zfs_ioc_get_fsacl(zfs_cmd_t *zc)
2680 {
2681 nvlist_t *nvp;
2682 int error;
2683
2684 if ((error = dsl_deleg_get(zc->zc_name, &nvp)) == 0) {
2685 error = put_nvlist(zc, nvp);
2686 nvlist_free(nvp);
2687 }
2688
2689 return (error);
2690 }
2691
2692 /*
2693 * Search the vfs list for a specified resource. Returns a pointer to it
2694 * or NULL if no suitable entry is found. The caller of this routine
2695 * is responsible for releasing the returned vfs pointer.
2696 */
2697 static vfs_t *
2698 zfs_get_vfs(const char *resource)
2699 {
2700 struct vfs *vfsp;
2701 struct vfs *vfs_found = NULL;
2702
2703 vfs_list_read_lock();
2704 vfsp = rootvfs;
2705 do {
2706 if (strcmp(refstr_value(vfsp->vfs_resource), resource) == 0) {
2707 VFS_HOLD(vfsp);
2708 vfs_found = vfsp;
2709 break;
2710 }
2711 vfsp = vfsp->vfs_next;
2712 } while (vfsp != rootvfs);
2713 vfs_list_unlock();
2714 return (vfs_found);
2715 }
2716
2717 /* ARGSUSED */
2718 static void
2719 zfs_create_cb(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx)
2720 {
2721 zfs_creat_t *zct = arg;
2722
2723 zfs_create_fs(os, cr, zct->zct_zplprops, tx);
2724 }
2725
2726 #define ZFS_PROP_UNDEFINED ((uint64_t)-1)
2727
2728 /*
2729 * inputs:
2730 * createprops list of properties requested by creator
2731 * default_zplver zpl version to use if unspecified in createprops
2732 * fuids_ok fuids allowed in this version of the spa?
2733 * os parent objset pointer (NULL if root fs)
2734 *
2735 * outputs:
2736 * zplprops values for the zplprops we attach to the master node object
2737 * is_ci true if requested file system will be purely case-insensitive
2738 *
2739 * Determine the settings for utf8only, normalization and
2740 * casesensitivity. Specific values may have been requested by the
2741 * creator and/or we can inherit values from the parent dataset. If
2742 * the file system is of too early a vintage, a creator can not
2743 * request settings for these properties, even if the requested
2744 * setting is the default value. We don't actually want to create dsl
2745 * properties for these, so remove them from the source nvlist after
2746 * processing.
2747 */
2748 static int
2749 zfs_fill_zplprops_impl(objset_t *os, uint64_t zplver,
2750 boolean_t fuids_ok, boolean_t sa_ok, nvlist_t *createprops,
2751 nvlist_t *zplprops, boolean_t *is_ci)
2752 {
2753 uint64_t sense = ZFS_PROP_UNDEFINED;
2754 uint64_t norm = ZFS_PROP_UNDEFINED;
2755 uint64_t u8 = ZFS_PROP_UNDEFINED;
2756
2757 ASSERT(zplprops != NULL);
2758
2759 /*
2760 * Pull out creator prop choices, if any.
2761 */
2762 if (createprops) {
2763 (void) nvlist_lookup_uint64(createprops,
2764 zfs_prop_to_name(ZFS_PROP_VERSION), &zplver);
2765 (void) nvlist_lookup_uint64(createprops,
2766 zfs_prop_to_name(ZFS_PROP_NORMALIZE), &norm);
2767 (void) nvlist_remove_all(createprops,
2768 zfs_prop_to_name(ZFS_PROP_NORMALIZE));
2769 (void) nvlist_lookup_uint64(createprops,
2770 zfs_prop_to_name(ZFS_PROP_UTF8ONLY), &u8);
2771 (void) nvlist_remove_all(createprops,
2772 zfs_prop_to_name(ZFS_PROP_UTF8ONLY));
2773 (void) nvlist_lookup_uint64(createprops,
2774 zfs_prop_to_name(ZFS_PROP_CASE), &sense);
2775 (void) nvlist_remove_all(createprops,
2776 zfs_prop_to_name(ZFS_PROP_CASE));
2777 }
2778
2779 /*
2780 * If the zpl version requested is whacky or the file system
2781 * or pool is version is too "young" to support normalization
2782 * and the creator tried to set a value for one of the props,
2783 * error out.
2784 */
2785 if ((zplver < ZPL_VERSION_INITIAL || zplver > ZPL_VERSION) ||
2786 (zplver >= ZPL_VERSION_FUID && !fuids_ok) ||
2787 (zplver >= ZPL_VERSION_SA && !sa_ok) ||
2788 (zplver < ZPL_VERSION_NORMALIZATION &&
2789 (norm != ZFS_PROP_UNDEFINED || u8 != ZFS_PROP_UNDEFINED ||
2790 sense != ZFS_PROP_UNDEFINED)))
2791 return (ENOTSUP);
2792
2793 /*
2794 * Put the version in the zplprops
2795 */
2796 VERIFY(nvlist_add_uint64(zplprops,
2797 zfs_prop_to_name(ZFS_PROP_VERSION), zplver) == 0);
2798
2799 if (norm == ZFS_PROP_UNDEFINED)
2800 VERIFY(zfs_get_zplprop(os, ZFS_PROP_NORMALIZE, &norm) == 0);
2801 VERIFY(nvlist_add_uint64(zplprops,
2802 zfs_prop_to_name(ZFS_PROP_NORMALIZE), norm) == 0);
2803
2804 /*
2805 * If we're normalizing, names must always be valid UTF-8 strings.
2806 */
2807 if (norm)
2808 u8 = 1;
2809 if (u8 == ZFS_PROP_UNDEFINED)
2810 VERIFY(zfs_get_zplprop(os, ZFS_PROP_UTF8ONLY, &u8) == 0);
2811 VERIFY(nvlist_add_uint64(zplprops,
2812 zfs_prop_to_name(ZFS_PROP_UTF8ONLY), u8) == 0);
2813
2814 if (sense == ZFS_PROP_UNDEFINED)
2815 VERIFY(zfs_get_zplprop(os, ZFS_PROP_CASE, &sense) == 0);
2816 VERIFY(nvlist_add_uint64(zplprops,
2817 zfs_prop_to_name(ZFS_PROP_CASE), sense) == 0);
2818
2819 if (is_ci)
2820 *is_ci = (sense == ZFS_CASE_INSENSITIVE);
2821
2822 return (0);
2823 }
2824
2825 static int
2826 zfs_fill_zplprops(const char *dataset, nvlist_t *createprops,
2827 nvlist_t *zplprops, boolean_t *is_ci)
2828 {
2829 boolean_t fuids_ok, sa_ok;
2830 uint64_t zplver = ZPL_VERSION;
2831 objset_t *os = NULL;
2832 char parentname[MAXNAMELEN];
2833 char *cp;
2834 spa_t *spa;
2835 uint64_t spa_vers;
2836 int error;
2837
2838 (void) strlcpy(parentname, dataset, sizeof (parentname));
2839 cp = strrchr(parentname, '/');
2840 ASSERT(cp != NULL);
2841 cp[0] = '\0';
2842
2843 if ((error = spa_open(dataset, &spa, FTAG)) != 0)
2844 return (error);
2845
2846 spa_vers = spa_version(spa);
2847 spa_close(spa, FTAG);
2848
2849 zplver = zfs_zpl_version_map(spa_vers);
2850 fuids_ok = (zplver >= ZPL_VERSION_FUID);
2851 sa_ok = (zplver >= ZPL_VERSION_SA);
2852
2853 /*
2854 * Open parent object set so we can inherit zplprop values.
2855 */
2856 if ((error = dmu_objset_hold(parentname, FTAG, &os)) != 0)
2857 return (error);
2858
2859 error = zfs_fill_zplprops_impl(os, zplver, fuids_ok, sa_ok, createprops,
2860 zplprops, is_ci);
2861 dmu_objset_rele(os, FTAG);
2862 return (error);
2863 }
2864
2865 static int
2866 zfs_fill_zplprops_root(uint64_t spa_vers, nvlist_t *createprops,
2867 nvlist_t *zplprops, boolean_t *is_ci)
2868 {
2869 boolean_t fuids_ok;
2870 boolean_t sa_ok;
2871 uint64_t zplver = ZPL_VERSION;
2872 int error;
2873
2874 zplver = zfs_zpl_version_map(spa_vers);
2875 fuids_ok = (zplver >= ZPL_VERSION_FUID);
2876 sa_ok = (zplver >= ZPL_VERSION_SA);
2877
2878 error = zfs_fill_zplprops_impl(NULL, zplver, fuids_ok, sa_ok,
2879 createprops, zplprops, is_ci);
2880 return (error);
2881 }
2882
2883 /*
2884 * inputs:
2885 * zc_objset_type type of objset to create (fs vs zvol)
2886 * zc_name name of new objset
2887 * zc_value name of snapshot to clone from (may be empty)
2888 * zc_nvlist_src{_size} nvlist of properties to apply
2889 *
2890 * outputs: none
2891 */
2892 static int
2893 zfs_ioc_create(zfs_cmd_t *zc)
2894 {
2895 objset_t *clone;
2896 int error = 0;
2897 zfs_creat_t zct;
2898 nvlist_t *nvprops = NULL;
2899 void (*cbfunc)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx);
2900 dmu_objset_type_t type = zc->zc_objset_type;
2901
2902 switch (type) {
2903
2904 case DMU_OST_ZFS:
2905 cbfunc = zfs_create_cb;
2906 break;
2907
2908 case DMU_OST_ZVOL:
2909 cbfunc = zvol_create_cb;
2910 break;
2911
2912 default:
2913 cbfunc = NULL;
2914 break;
2915 }
2916 if (strchr(zc->zc_name, '@') ||
2917 strchr(zc->zc_name, '%'))
2918 return (EINVAL);
2919
2920 if (zc->zc_nvlist_src != NULL &&
2921 (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2922 zc->zc_iflags, &nvprops)) != 0)
2923 return (error);
2924
2925 zct.zct_zplprops = NULL;
2926 zct.zct_props = nvprops;
2927
2928 if (zc->zc_value[0] != '\0') {
2929 /*
2930 * We're creating a clone of an existing snapshot.
2931 */
2932 zc->zc_value[sizeof (zc->zc_value) - 1] = '\0';
2933 if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0) {
2934 nvlist_free(nvprops);
2935 return (EINVAL);
2936 }
2937
2938 error = dmu_objset_hold(zc->zc_value, FTAG, &clone);
2939 if (error) {
2940 nvlist_free(nvprops);
2941 return (error);
2942 }
2943
2944 error = dmu_objset_clone(zc->zc_name, dmu_objset_ds(clone), 0);
2945 dmu_objset_rele(clone, FTAG);
2946 if (error) {
2947 nvlist_free(nvprops);
2948 return (error);
2949 }
2950 } else {
2951 boolean_t is_insensitive = B_FALSE;
2952
2953 if (cbfunc == NULL) {
2954 nvlist_free(nvprops);
2955 return (EINVAL);
2956 }
2957
2958 if (type == DMU_OST_ZVOL) {
2959 uint64_t volsize, volblocksize;
2960
2961 if (nvprops == NULL ||
2962 nvlist_lookup_uint64(nvprops,
2963 zfs_prop_to_name(ZFS_PROP_VOLSIZE),
2964 &volsize) != 0) {
2965 nvlist_free(nvprops);
2966 return (EINVAL);
2967 }
2968
2969 if ((error = nvlist_lookup_uint64(nvprops,
2970 zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE),
2971 &volblocksize)) != 0 && error != ENOENT) {
2972 nvlist_free(nvprops);
2973 return (EINVAL);
2974 }
2975
2976 if (error != 0)
2977 volblocksize = zfs_prop_default_numeric(
2978 ZFS_PROP_VOLBLOCKSIZE);
2979
2980 if ((error = zvol_check_volblocksize(
2981 volblocksize)) != 0 ||
2982 (error = zvol_check_volsize(volsize,
2983 volblocksize)) != 0) {
2984 nvlist_free(nvprops);
2985 return (error);
2986 }
2987 } else if (type == DMU_OST_ZFS) {
2988 int error;
2989
2990 /*
2991 * We have to have normalization and
2992 * case-folding flags correct when we do the
2993 * file system creation, so go figure them out
2994 * now.
2995 */
2996 VERIFY(nvlist_alloc(&zct.zct_zplprops,
2997 NV_UNIQUE_NAME, KM_SLEEP) == 0);
2998 error = zfs_fill_zplprops(zc->zc_name, nvprops,
2999 zct.zct_zplprops, &is_insensitive);
3000 if (error != 0) {
3001 nvlist_free(nvprops);
3002 nvlist_free(zct.zct_zplprops);
3003 return (error);
3004 }
3005 }
3006 error = dmu_objset_create(zc->zc_name, type,
3007 is_insensitive ? DS_FLAG_CI_DATASET : 0, cbfunc, &zct);
3008 nvlist_free(zct.zct_zplprops);
3009 }
3010
3011 /*
3012 * It would be nice to do this atomically.
3013 */
3014 if (error == 0) {
3015 error = zfs_set_prop_nvlist(zc->zc_name, ZPROP_SRC_LOCAL,
3016 nvprops, NULL);
3017 if (error != 0)
3018 (void) dmu_objset_destroy(zc->zc_name, B_FALSE);
3019 }
3020 nvlist_free(nvprops);
3021 return (error);
3022 }
3023
3024 /*
3025 * inputs:
3026 * zc_name name of filesystem
3027 * zc_value short name of snapshot
3028 * zc_cookie recursive flag
3029 * zc_nvlist_src[_size] property list
3030 *
3031 * outputs:
3032 * zc_value short snapname (i.e. part after the '@')
3033 */
3034 static int
3035 zfs_ioc_snapshot(zfs_cmd_t *zc)
3036 {
3037 nvlist_t *nvprops = NULL;
3038 int error;
3039 boolean_t recursive = zc->zc_cookie;
3040
3041 if (snapshot_namecheck(zc->zc_value, NULL, NULL) != 0)
3042 return (EINVAL);
3043
3044 if (zc->zc_nvlist_src != NULL &&
3045 (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
3046 zc->zc_iflags, &nvprops)) != 0)
3047 return (error);
3048
3049 error = zfs_check_userprops(zc->zc_name, nvprops);
3050 if (error)
3051 goto out;
3052
3053 if (!nvlist_empty(nvprops) &&
3054 zfs_earlier_version(zc->zc_name, SPA_VERSION_SNAP_PROPS)) {
3055 error = ENOTSUP;
3056 goto out;
3057 }
3058
3059 error = dmu_objset_snapshot(zc->zc_name, zc->zc_value, NULL,
3060 nvprops, recursive, B_FALSE, -1);
3061
3062 out:
3063 nvlist_free(nvprops);
3064 return (error);
3065 }
3066
3067 int
3068 zfs_unmount_snap(const char *name, void *arg)
3069 {
3070 vfs_t *vfsp = NULL;
3071
3072 if (arg) {
3073 char *snapname = arg;
3074 char *fullname = kmem_asprintf("%s@%s", name, snapname);
3075 vfsp = zfs_get_vfs(fullname);
3076 strfree(fullname);
3077 } else if (strchr(name, '@')) {
3078 vfsp = zfs_get_vfs(name);
3079 }
3080
3081 if (vfsp) {
3082 /*
3083 * Always force the unmount for snapshots.
3084 */
3085 int flag = MS_FORCE;
3086 int err;
3087
3088 if ((err = vn_vfswlock(vfsp->vfs_vnodecovered)) != 0) {
3089 VFS_RELE(vfsp);
3090 return (err);
3091 }
3092 VFS_RELE(vfsp);
3093 if ((err = dounmount(vfsp, flag, kcred)) != 0)
3094 return (err);
3095 }
3096 return (0);
3097 }
3098
3099 /*
3100 * inputs:
3101 * zc_name name of filesystem
3102 * zc_value short name of snapshot
3103 * zc_defer_destroy mark for deferred destroy
3104 *
3105 * outputs: none
3106 */
3107 static int
3108 zfs_ioc_destroy_snaps(zfs_cmd_t *zc)
3109 {
3110 int err;
3111
3112 if (snapshot_namecheck(zc->zc_value, NULL, NULL) != 0)
3113 return (EINVAL);
3114 err = dmu_objset_find(zc->zc_name,
3115 zfs_unmount_snap, zc->zc_value, DS_FIND_CHILDREN);
3116 if (err)
3117 return (err);
3118 return (dmu_snapshots_destroy(zc->zc_name, zc->zc_value,
3119 zc->zc_defer_destroy));
3120 }
3121
3122 /*
3123 * inputs:
3124 * zc_name name of dataset to destroy
3125 * zc_objset_type type of objset
3126 * zc_defer_destroy mark for deferred destroy
3127 *
3128 * outputs: none
3129 */
3130 static int
3131 zfs_ioc_destroy(zfs_cmd_t *zc)
3132 {
3133 int err;
3134 if (strchr(zc->zc_name, '@') && zc->zc_objset_type == DMU_OST_ZFS) {
3135 err = zfs_unmount_snap(zc->zc_name, NULL);
3136 if (err)
3137 return (err);
3138 }
3139
3140 err = dmu_objset_destroy(zc->zc_name, zc->zc_defer_destroy);
3141 if (zc->zc_objset_type == DMU_OST_ZVOL && err == 0)
3142 (void) zvol_remove_minor(zc->zc_name);
3143 return (err);
3144 }
3145
3146 /*
3147 * inputs:
3148 * zc_name name of dataset to rollback (to most recent snapshot)
3149 *
3150 * outputs: none
3151 */
3152 static int
3153 zfs_ioc_rollback(zfs_cmd_t *zc)
3154 {
3155 dsl_dataset_t *ds, *clone;
3156 int error;
3157 zfsvfs_t *zfsvfs;
3158 char *clone_name;
3159
3160 error = dsl_dataset_hold(zc->zc_name, FTAG, &ds);
3161 if (error)
3162 return (error);
3163
3164 /* must not be a snapshot */
3165 if (dsl_dataset_is_snapshot(ds)) {
3166 dsl_dataset_rele(ds, FTAG);
3167 return (EINVAL);
3168 }
3169
3170 /* must have a most recent snapshot */
3171 if (ds->ds_phys->ds_prev_snap_txg < TXG_INITIAL) {
3172 dsl_dataset_rele(ds, FTAG);
3173 return (EINVAL);
3174 }
3175
3176 /*
3177 * Create clone of most recent snapshot.
3178 */
3179 clone_name = kmem_asprintf("%s/%%rollback", zc->zc_name);
3180 error = dmu_objset_clone(clone_name, ds->ds_prev, DS_FLAG_INCONSISTENT);
3181 if (error)
3182 goto out;
3183
3184 error = dsl_dataset_own(clone_name, B_TRUE, FTAG, &clone);
3185 if (error)
3186 goto out;
3187
3188 /*
3189 * Do clone swap.
3190 */
3191 if (getzfsvfs(zc->zc_name, &zfsvfs) == 0) {
3192 error = zfs_suspend_fs(zfsvfs);
3193 if (error == 0) {
3194 int resume_err;
3195
3196 if (dsl_dataset_tryown(ds, B_FALSE, FTAG)) {
3197 error = dsl_dataset_clone_swap(clone, ds,
3198 B_TRUE);
3199 dsl_dataset_disown(ds, FTAG);
3200 ds = NULL;
3201 } else {
3202 error = EBUSY;
3203 }
3204 resume_err = zfs_resume_fs(zfsvfs, zc->zc_name);
3205 error = error ? error : resume_err;
3206 }
3207 VFS_RELE(zfsvfs->z_vfs);
3208 } else {
3209 if (dsl_dataset_tryown(ds, B_FALSE, FTAG)) {
3210 error = dsl_dataset_clone_swap(clone, ds, B_TRUE);
3211 dsl_dataset_disown(ds, FTAG);
3212 ds = NULL;
3213 } else {
3214 error = EBUSY;
3215 }
3216 }
3217
3218 /*
3219 * Destroy clone (which also closes it).
3220 */
3221 (void) dsl_dataset_destroy(clone, FTAG, B_FALSE);
3222
3223 out:
3224 strfree(clone_name);
3225 if (ds)
3226 dsl_dataset_rele(ds, FTAG);
3227 return (error);
3228 }
3229
3230 /*
3231 * inputs:
3232 * zc_name old name of dataset
3233 * zc_value new name of dataset
3234 * zc_cookie recursive flag (only valid for snapshots)
3235 *
3236 * outputs: none
3237 */
3238 static int
3239 zfs_ioc_rename(zfs_cmd_t *zc)
3240 {
3241 boolean_t recursive = zc->zc_cookie & 1;
3242
3243 zc->zc_value[sizeof (zc->zc_value) - 1] = '\0';
3244 if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
3245 strchr(zc->zc_value, '%'))
3246 return (EINVAL);
3247
3248 /*
3249 * Unmount snapshot unless we're doing a recursive rename,
3250 * in which case the dataset code figures out which snapshots
3251 * to unmount.
3252 */
3253 if (!recursive && strchr(zc->zc_name, '@') != NULL &&
3254 zc->zc_objset_type == DMU_OST_ZFS) {
3255 int err = zfs_unmount_snap(zc->zc_name, NULL);
3256 if (err)
3257 return (err);
3258 }
3259 if (zc->zc_objset_type == DMU_OST_ZVOL)
3260 (void) zvol_remove_minor(zc->zc_name);
3261 return (dmu_objset_rename(zc->zc_name, zc->zc_value, recursive));
3262 }
3263
3264 static int
3265 zfs_check_settable(const char *dsname, nvpair_t *pair, cred_t *cr)
3266 {
3267 const char *propname = nvpair_name(pair);
3268 boolean_t issnap = (strchr(dsname, '@') != NULL);
3269 zfs_prop_t prop = zfs_name_to_prop(propname);
3270 uint64_t intval;
3271 int err;
3272
3273 if (prop == ZPROP_INVAL) {
3274 if (zfs_prop_user(propname)) {
3275 if (err = zfs_secpolicy_write_perms(dsname,
3276 ZFS_DELEG_PERM_USERPROP, cr))
3277 return (err);
3278 return (0);
3279 }
3280
3281 if (!issnap && zfs_prop_userquota(propname)) {
3282 const char *perm = NULL;
3283 const char *uq_prefix =
3284 zfs_userquota_prop_prefixes[ZFS_PROP_USERQUOTA];
3285 const char *gq_prefix =
3286 zfs_userquota_prop_prefixes[ZFS_PROP_GROUPQUOTA];
3287
3288 if (strncmp(propname, uq_prefix,
3289 strlen(uq_prefix)) == 0) {
3290 perm = ZFS_DELEG_PERM_USERQUOTA;
3291 } else if (strncmp(propname, gq_prefix,
3292 strlen(gq_prefix)) == 0) {
3293 perm = ZFS_DELEG_PERM_GROUPQUOTA;
3294 } else {
3295 /* USERUSED and GROUPUSED are read-only */
3296 return (EINVAL);
3297 }
3298
3299 if (err = zfs_secpolicy_write_perms(dsname, perm, cr))
3300 return (err);
3301 return (0);
3302 }
3303
3304 return (EINVAL);
3305 }
3306
3307 if (issnap)
3308 return (EINVAL);
3309
3310 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
3311 /*
3312 * dsl_prop_get_all_impl() returns properties in this
3313 * format.
3314 */
3315 nvlist_t *attrs;
3316 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
3317 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
3318 &pair) == 0);
3319 }
3320
3321 /*
3322 * Check that this value is valid for this pool version
3323 */
3324 switch (prop) {
3325 case ZFS_PROP_COMPRESSION:
3326 /*
3327 * If the user specified gzip compression, make sure
3328 * the SPA supports it. We ignore any errors here since
3329 * we'll catch them later.
3330 */
3331 if (nvpair_type(pair) == DATA_TYPE_UINT64 &&
3332 nvpair_value_uint64(pair, &intval) == 0) {
3333 if (intval >= ZIO_COMPRESS_GZIP_1 &&
3334 intval <= ZIO_COMPRESS_GZIP_9 &&
3335 zfs_earlier_version(dsname,
3336 SPA_VERSION_GZIP_COMPRESSION)) {
3337 return (ENOTSUP);
3338 }
3339
3340 if (intval == ZIO_COMPRESS_ZLE &&
3341 zfs_earlier_version(dsname,
3342 SPA_VERSION_ZLE_COMPRESSION))
3343 return (ENOTSUP);
3344
3345 /*
3346 * If this is a bootable dataset then
3347 * verify that the compression algorithm
3348 * is supported for booting. We must return
3349 * something other than ENOTSUP since it
3350 * implies a downrev pool version.
3351 */
3352 if (zfs_is_bootfs(dsname) &&
3353 !BOOTFS_COMPRESS_VALID(intval)) {
3354 return (ERANGE);
3355 }
3356 }
3357 break;
3358
3359 case ZFS_PROP_COPIES:
3360 if (zfs_earlier_version(dsname, SPA_VERSION_DITTO_BLOCKS))
3361 return (ENOTSUP);
3362 break;
3363
3364 case ZFS_PROP_DEDUP:
3365 if (zfs_earlier_version(dsname, SPA_VERSION_DEDUP))
3366 return (ENOTSUP);
3367 break;
3368
3369 case ZFS_PROP_SHARESMB:
3370 if (zpl_earlier_version(dsname, ZPL_VERSION_FUID))
3371 return (ENOTSUP);
3372 break;
3373
3374 case ZFS_PROP_ACLINHERIT:
3375 if (nvpair_type(pair) == DATA_TYPE_UINT64 &&
3376 nvpair_value_uint64(pair, &intval) == 0) {
3377 if (intval == ZFS_ACL_PASSTHROUGH_X &&
3378 zfs_earlier_version(dsname,
3379 SPA_VERSION_PASSTHROUGH_X))
3380 return (ENOTSUP);
3381 }
3382 break;
3383 }
3384
3385 return (zfs_secpolicy_setprop(dsname, prop, pair, CRED()));
3386 }
3387
3388 /*
3389 * Removes properties from the given props list that fail permission checks
3390 * needed to clear them and to restore them in case of a receive error. For each
3391 * property, make sure we have both set and inherit permissions.
3392 *
3393 * Returns the first error encountered if any permission checks fail. If the
3394 * caller provides a non-NULL errlist, it also gives the complete list of names
3395 * of all the properties that failed a permission check along with the
3396 * corresponding error numbers. The caller is responsible for freeing the
3397 * returned errlist.
3398 *
3399 * If every property checks out successfully, zero is returned and the list
3400 * pointed at by errlist is NULL.
3401 */
3402 static int
3403 zfs_check_clearable(char *dataset, nvlist_t *props, nvlist_t **errlist)
3404 {
3405 zfs_cmd_t *zc;
3406 nvpair_t *pair, *next_pair;
3407 nvlist_t *errors;
3408 int err, rv = 0;
3409
3410 if (props == NULL)
3411 return (0);
3412
3413 VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0);
3414
3415 zc = kmem_alloc(sizeof (zfs_cmd_t), KM_SLEEP);
3416 (void) strcpy(zc->zc_name, dataset);
3417 pair = nvlist_next_nvpair(props, NULL);
3418 while (pair != NULL) {
3419 next_pair = nvlist_next_nvpair(props, pair);
3420
3421 (void) strcpy(zc->zc_value, nvpair_name(pair));
3422 if ((err = zfs_check_settable(dataset, pair, CRED())) != 0 ||
3423 (err = zfs_secpolicy_inherit(zc, CRED())) != 0) {
3424 VERIFY(nvlist_remove_nvpair(props, pair) == 0);
3425 VERIFY(nvlist_add_int32(errors,
3426 zc->zc_value, err) == 0);
3427 }
3428 pair = next_pair;
3429 }
3430 kmem_free(zc, sizeof (zfs_cmd_t));
3431
3432 if ((pair = nvlist_next_nvpair(errors, NULL)) == NULL) {
3433 nvlist_free(errors);
3434 errors = NULL;
3435 } else {
3436 VERIFY(nvpair_value_int32(pair, &rv) == 0);
3437 }
3438
3439 if (errlist == NULL)
3440 nvlist_free(errors);
3441 else
3442 *errlist = errors;
3443
3444 return (rv);
3445 }
3446
3447 static boolean_t
3448 propval_equals(nvpair_t *p1, nvpair_t *p2)
3449 {
3450 if (nvpair_type(p1) == DATA_TYPE_NVLIST) {
3451 /* dsl_prop_get_all_impl() format */
3452 nvlist_t *attrs;
3453 VERIFY(nvpair_value_nvlist(p1, &attrs) == 0);
3454 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
3455 &p1) == 0);
3456 }
3457
3458 if (nvpair_type(p2) == DATA_TYPE_NVLIST) {
3459 nvlist_t *attrs;
3460 VERIFY(nvpair_value_nvlist(p2, &attrs) == 0);
3461 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
3462 &p2) == 0);
3463 }
3464
3465 if (nvpair_type(p1) != nvpair_type(p2))
3466 return (B_FALSE);
3467
3468 if (nvpair_type(p1) == DATA_TYPE_STRING) {
3469 char *valstr1, *valstr2;
3470
3471 VERIFY(nvpair_value_string(p1, (char **)&valstr1) == 0);
3472 VERIFY(nvpair_value_string(p2, (char **)&valstr2) == 0);
3473 return (strcmp(valstr1, valstr2) == 0);
3474 } else {
3475 uint64_t intval1, intval2;
3476
3477 VERIFY(nvpair_value_uint64(p1, &intval1) == 0);
3478 VERIFY(nvpair_value_uint64(p2, &intval2) == 0);
3479 return (intval1 == intval2);
3480 }
3481 }
3482
3483 /*
3484 * Remove properties from props if they are not going to change (as determined
3485 * by comparison with origprops). Remove them from origprops as well, since we
3486 * do not need to clear or restore properties that won't change.
3487 */
3488 static void
3489 props_reduce(nvlist_t *props, nvlist_t *origprops)
3490 {
3491 nvpair_t *pair, *next_pair;
3492
3493 if (origprops == NULL)
3494 return; /* all props need to be received */
3495
3496 pair = nvlist_next_nvpair(props, NULL);
3497 while (pair != NULL) {
3498 const char *propname = nvpair_name(pair);
3499 nvpair_t *match;
3500
3501 next_pair = nvlist_next_nvpair(props, pair);
3502
3503 if ((nvlist_lookup_nvpair(origprops, propname,
3504 &match) != 0) || !propval_equals(pair, match))
3505 goto next; /* need to set received value */
3506
3507 /* don't clear the existing received value */
3508 (void) nvlist_remove_nvpair(origprops, match);
3509 /* don't bother receiving the property */
3510 (void) nvlist_remove_nvpair(props, pair);
3511 next:
3512 pair = next_pair;
3513 }
3514 }
3515
3516 #ifdef DEBUG
3517 static boolean_t zfs_ioc_recv_inject_err;
3518 #endif
3519
3520 /*
3521 * inputs:
3522 * zc_name name of containing filesystem
3523 * zc_nvlist_src{_size} nvlist of properties to apply
3524 * zc_value name of snapshot to create
3525 * zc_string name of clone origin (if DRR_FLAG_CLONE)
3526 * zc_cookie file descriptor to recv from
3527 * zc_begin_record the BEGIN record of the stream (not byteswapped)
3528 * zc_guid force flag
3529 * zc_cleanup_fd cleanup-on-exit file descriptor
3530 * zc_action_handle handle for this guid/ds mapping (or zero on first call)
3531 *
3532 * outputs:
3533 * zc_cookie number of bytes read
3534 * zc_nvlist_dst{_size} error for each unapplied received property
3535 * zc_obj zprop_errflags_t
3536 * zc_action_handle handle for this guid/ds mapping
3537 */
3538 static int
3539 zfs_ioc_recv(zfs_cmd_t *zc)
3540 {
3541 file_t *fp;
3542 objset_t *os;
3543 dmu_recv_cookie_t drc;
3544 boolean_t force = (boolean_t)zc->zc_guid;
3545 int fd;
3546 int error = 0;
3547 int props_error = 0;
3548 nvlist_t *errors;
3549 offset_t off;
3550 nvlist_t *props = NULL; /* sent properties */
3551 nvlist_t *origprops = NULL; /* existing properties */
3552 objset_t *origin = NULL;
3553 char *tosnap;
3554 char tofs[ZFS_MAXNAMELEN];
3555 boolean_t first_recvd_props = B_FALSE;
3556
3557 if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
3558 strchr(zc->zc_value, '@') == NULL ||
3559 strchr(zc->zc_value, '%'))
3560 return (EINVAL);
3561
3562 (void) strcpy(tofs, zc->zc_value);
3563 tosnap = strchr(tofs, '@');
3564 *tosnap++ = '\0';
3565
3566 if (zc->zc_nvlist_src != NULL &&
3567 (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
3568 zc->zc_iflags, &props)) != 0)
3569 return (error);
3570
3571 fd = zc->zc_cookie;
3572 fp = getf(fd);
3573 if (fp == NULL) {
3574 nvlist_free(props);
3575 return (EBADF);
3576 }
3577
3578 VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0);
3579
3580 if (props && dmu_objset_hold(tofs, FTAG, &os) == 0) {
3581 if ((spa_version(os->os_spa) >= SPA_VERSION_RECVD_PROPS) &&
3582 !dsl_prop_get_hasrecvd(os)) {
3583 first_recvd_props = B_TRUE;
3584 }
3585
3586 /*
3587 * If new received properties are supplied, they are to
3588 * completely replace the existing received properties, so stash
3589 * away the existing ones.
3590 */
3591 if (dsl_prop_get_received(os, &origprops) == 0) {
3592 nvlist_t *errlist = NULL;
3593 /*
3594 * Don't bother writing a property if its value won't
3595 * change (and avoid the unnecessary security checks).
3596 *
3597 * The first receive after SPA_VERSION_RECVD_PROPS is a
3598 * special case where we blow away all local properties
3599 * regardless.
3600 */
3601 if (!first_recvd_props)
3602 props_reduce(props, origprops);
3603 if (zfs_check_clearable(tofs, origprops,
3604 &errlist) != 0)
3605 (void) nvlist_merge(errors, errlist, 0);
3606 nvlist_free(errlist);
3607 }
3608
3609 dmu_objset_rele(os, FTAG);
3610 }
3611
3612 if (zc->zc_string[0]) {
3613 error = dmu_objset_hold(zc->zc_string, FTAG, &origin);
3614 if (error)
3615 goto out;
3616 }
3617
3618 error = dmu_recv_begin(tofs, tosnap, zc->zc_top_ds,
3619 &zc->zc_begin_record, force, origin, &drc);
3620 if (origin)
3621 dmu_objset_rele(origin, FTAG);
3622 if (error)
3623 goto out;
3624
3625 /*
3626 * Set properties before we receive the stream so that they are applied
3627 * to the new data. Note that we must call dmu_recv_stream() if
3628 * dmu_recv_begin() succeeds.
3629 */
3630 if (props) {
3631 nvlist_t *errlist;
3632
3633 if (dmu_objset_from_ds(drc.drc_logical_ds, &os) == 0) {
3634 if (drc.drc_newfs) {
3635 if (spa_version(os->os_spa) >=
3636 SPA_VERSION_RECVD_PROPS)
3637 first_recvd_props = B_TRUE;
3638 } else if (origprops != NULL) {
3639 if (clear_received_props(os, tofs, origprops,
3640 first_recvd_props ? NULL : props) != 0)
3641 zc->zc_obj |= ZPROP_ERR_NOCLEAR;
3642 } else {
3643 zc->zc_obj |= ZPROP_ERR_NOCLEAR;
3644 }
3645 dsl_prop_set_hasrecvd(os);
3646 } else if (!drc.drc_newfs) {
3647 zc->zc_obj |= ZPROP_ERR_NOCLEAR;
3648 }
3649
3650 (void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_RECEIVED,
3651 props, &errlist);
3652 (void) nvlist_merge(errors, errlist, 0);
3653 nvlist_free(errlist);
3654 }
3655
3656 if (fit_error_list(zc, &errors) != 0 || put_nvlist(zc, errors) != 0) {
3657 /*
3658 * Caller made zc->zc_nvlist_dst less than the minimum expected
3659 * size or supplied an invalid address.
3660 */
3661 props_error = EINVAL;
3662 }
3663
3664 off = fp->f_offset;
3665 error = dmu_recv_stream(&drc, fp->f_vnode, &off, zc->zc_cleanup_fd,
3666 &zc->zc_action_handle);
3667
3668 if (error == 0) {
3669 zfsvfs_t *zfsvfs = NULL;
3670
3671 if (getzfsvfs(tofs, &zfsvfs) == 0) {
3672 /* online recv */
3673 int end_err;
3674
3675 error = zfs_suspend_fs(zfsvfs);
3676 /*
3677 * If the suspend fails, then the recv_end will
3678 * likely also fail, and clean up after itself.
3679 */
3680 end_err = dmu_recv_end(&drc);
3681 if (error == 0)
3682 error = zfs_resume_fs(zfsvfs, tofs);
3683 error = error ? error : end_err;
3684 VFS_RELE(zfsvfs->z_vfs);
3685 } else {
3686 error = dmu_recv_end(&drc);
3687 }
3688 }
3689
3690 zc->zc_cookie = off - fp->f_offset;
3691 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
3692 fp->f_offset = off;
3693
3694 #ifdef DEBUG
3695 if (zfs_ioc_recv_inject_err) {
3696 zfs_ioc_recv_inject_err = B_FALSE;
3697 error = 1;
3698 }
3699 #endif
3700 /*
3701 * On error, restore the original props.
3702 */
3703 if (error && props) {
3704 if (dmu_objset_hold(tofs, FTAG, &os) == 0) {
3705 if (clear_received_props(os, tofs, props, NULL) != 0) {
3706 /*
3707 * We failed to clear the received properties.
3708 * Since we may have left a $recvd value on the
3709 * system, we can't clear the $hasrecvd flag.
3710 */
3711 zc->zc_obj |= ZPROP_ERR_NORESTORE;
3712 } else if (first_recvd_props) {
3713 dsl_prop_unset_hasrecvd(os);
3714 }
3715 dmu_objset_rele(os, FTAG);
3716 } else if (!drc.drc_newfs) {
3717 /* We failed to clear the received properties. */
3718 zc->zc_obj |= ZPROP_ERR_NORESTORE;
3719 }
3720
3721 if (origprops == NULL && !drc.drc_newfs) {
3722 /* We failed to stash the original properties. */
3723 zc->zc_obj |= ZPROP_ERR_NORESTORE;
3724 }
3725
3726 /*
3727 * dsl_props_set() will not convert RECEIVED to LOCAL on or
3728 * after SPA_VERSION_RECVD_PROPS, so we need to specify LOCAL
3729 * explictly if we're restoring local properties cleared in the
3730 * first new-style receive.
3731 */
3732 if (origprops != NULL &&
3733 zfs_set_prop_nvlist(tofs, (first_recvd_props ?
3734 ZPROP_SRC_LOCAL : ZPROP_SRC_RECEIVED),
3735 origprops, NULL) != 0) {
3736 /*
3737 * We stashed the original properties but failed to
3738 * restore them.
3739 */
3740 zc->zc_obj |= ZPROP_ERR_NORESTORE;
3741 }
3742 }
3743 out:
3744 nvlist_free(props);
3745 nvlist_free(origprops);
3746 nvlist_free(errors);
3747 releasef(fd);
3748
3749 if (error == 0)
3750 error = props_error;
3751
3752 return (error);
3753 }
3754
3755 /*
3756 * inputs:
3757 * zc_name name of snapshot to send
3758 * zc_cookie file descriptor to send stream to
3759 * zc_obj fromorigin flag (mutually exclusive with zc_fromobj)
3760 * zc_sendobj objsetid of snapshot to send
3761 * zc_fromobj objsetid of incremental fromsnap (may be zero)
3762 *
3763 * outputs: none
3764 */
3765 static int
3766 zfs_ioc_send(zfs_cmd_t *zc)
3767 {
3768 objset_t *fromsnap = NULL;
3769 objset_t *tosnap;
3770 file_t *fp;
3771 int error;
3772 offset_t off;
3773 dsl_dataset_t *ds;
3774 dsl_dataset_t *dsfrom = NULL;
3775 spa_t *spa;
3776 dsl_pool_t *dp;
3777
3778 error = spa_open(zc->zc_name, &spa, FTAG);
3779 if (error)
3780 return (error);
3781
3782 dp = spa_get_dsl(spa);
3783 rw_enter(&dp->dp_config_rwlock, RW_READER);
3784 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &ds);
3785 rw_exit(&dp->dp_config_rwlock);
3786 if (error) {
3787 spa_close(spa, FTAG);
3788 return (error);
3789 }
3790
3791 error = dmu_objset_from_ds(ds, &tosnap);
3792 if (error) {
3793 dsl_dataset_rele(ds, FTAG);
3794 spa_close(spa, FTAG);
3795 return (error);
3796 }
3797
3798 if (zc->zc_fromobj != 0) {
3799 rw_enter(&dp->dp_config_rwlock, RW_READER);
3800 error = dsl_dataset_hold_obj(dp, zc->zc_fromobj, FTAG, &dsfrom);
3801 rw_exit(&dp->dp_config_rwlock);
3802 spa_close(spa, FTAG);
3803 if (error) {
3804 dsl_dataset_rele(ds, FTAG);
3805 return (error);
3806 }
3807 error = dmu_objset_from_ds(dsfrom, &fromsnap);
3808 if (error) {
3809 dsl_dataset_rele(dsfrom, FTAG);
3810 dsl_dataset_rele(ds, FTAG);
3811 return (error);
3812 }
3813 } else {
3814 spa_close(spa, FTAG);
3815 }
3816
3817 fp = getf(zc->zc_cookie);
3818 if (fp == NULL) {
3819 dsl_dataset_rele(ds, FTAG);
3820 if (dsfrom)
3821 dsl_dataset_rele(dsfrom, FTAG);
3822 return (EBADF);
3823 }
3824
3825 off = fp->f_offset;
3826 error = dmu_sendbackup(tosnap, fromsnap, zc->zc_obj, fp->f_vnode, &off);
3827
3828 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
3829 fp->f_offset = off;
3830 releasef(zc->zc_cookie);
3831 if (dsfrom)
3832 dsl_dataset_rele(dsfrom, FTAG);
3833 dsl_dataset_rele(ds, FTAG);
3834 return (error);
3835 }
3836
3837 static int
3838 zfs_ioc_inject_fault(zfs_cmd_t *zc)
3839 {
3840 int id, error;
3841
3842 error = zio_inject_fault(zc->zc_name, (int)zc->zc_guid, &id,
3843 &zc->zc_inject_record);
3844
3845 if (error == 0)
3846 zc->zc_guid = (uint64_t)id;
3847
3848 return (error);
3849 }
3850
3851 static int
3852 zfs_ioc_clear_fault(zfs_cmd_t *zc)
3853 {
3854 return (zio_clear_fault((int)zc->zc_guid));
3855 }
3856
3857 static int
3858 zfs_ioc_inject_list_next(zfs_cmd_t *zc)
3859 {
3860 int id = (int)zc->zc_guid;
3861 int error;
3862
3863 error = zio_inject_list_next(&id, zc->zc_name, sizeof (zc->zc_name),
3864 &zc->zc_inject_record);
3865
3866 zc->zc_guid = id;
3867
3868 return (error);
3869 }
3870
3871 static int
3872 zfs_ioc_error_log(zfs_cmd_t *zc)
3873 {
3874 spa_t *spa;
3875 int error;
3876 size_t count = (size_t)zc->zc_nvlist_dst_size;
3877
3878 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
3879 return (error);
3880
3881 error = spa_get_errlog(spa, (void *)(uintptr_t)zc->zc_nvlist_dst,
3882 &count);
3883 if (error == 0)
3884 zc->zc_nvlist_dst_size = count;
3885 else
3886 zc->zc_nvlist_dst_size = spa_get_errlog_size(spa);
3887
3888 spa_close(spa, FTAG);
3889
3890 return (error);
3891 }
3892
3893 static int
3894 zfs_ioc_clear(zfs_cmd_t *zc)
3895 {
3896 spa_t *spa;
3897 vdev_t *vd;
3898 int error;
3899
3900 /*
3901 * On zpool clear we also fix up missing slogs
3902 */
3903 mutex_enter(&spa_namespace_lock);
3904 spa = spa_lookup(zc->zc_name);
3905 if (spa == NULL) {
3906 mutex_exit(&spa_namespace_lock);
3907 return (EIO);
3908 }
3909 if (spa_get_log_state(spa) == SPA_LOG_MISSING) {
3910 /* we need to let spa_open/spa_load clear the chains */
3911 spa_set_log_state(spa, SPA_LOG_CLEAR);
3912 }
3913 spa->spa_last_open_failed = 0;
3914 mutex_exit(&spa_namespace_lock);
3915
3916 if (zc->zc_cookie & ZPOOL_NO_REWIND) {
3917 error = spa_open(zc->zc_name, &spa, FTAG);
3918 } else {
3919 nvlist_t *policy;
3920 nvlist_t *config = NULL;
3921
3922 if (zc->zc_nvlist_src == NULL)
3923 return (EINVAL);
3924
3925 if ((error = get_nvlist(zc->zc_nvlist_src,
3926 zc->zc_nvlist_src_size, zc->zc_iflags, &policy)) == 0) {
3927 error = spa_open_rewind(zc->zc_name, &spa, FTAG,
3928 policy, &config);
3929 if (config != NULL) {
3930 int err;
3931
3932 if ((err = put_nvlist(zc, config)) != 0)
3933 error = err;
3934 nvlist_free(config);
3935 }
3936 nvlist_free(policy);
3937 }
3938 }
3939
3940 if (error)
3941 return (error);
3942
3943 spa_vdev_state_enter(spa, SCL_NONE);
3944
3945 if (zc->zc_guid == 0) {
3946 vd = NULL;
3947 } else {
3948 vd = spa_lookup_by_guid(spa, zc->zc_guid, B_TRUE);
3949 if (vd == NULL) {
3950 (void) spa_vdev_state_exit(spa, NULL, ENODEV);
3951 spa_close(spa, FTAG);
3952 return (ENODEV);
3953 }
3954 }
3955
3956 vdev_clear(spa, vd);
3957
3958 (void) spa_vdev_state_exit(spa, NULL, 0);
3959
3960 /*
3961 * Resume any suspended I/Os.
3962 */
3963 if (zio_resume(spa) != 0)
3964 error = EIO;
3965
3966 spa_close(spa, FTAG);
3967
3968 return (error);
3969 }
3970
3971 /*
3972 * inputs:
3973 * zc_name name of filesystem
3974 * zc_value name of origin snapshot
3975 *
3976 * outputs:
3977 * zc_string name of conflicting snapshot, if there is one
3978 */
3979 static int
3980 zfs_ioc_promote(zfs_cmd_t *zc)
3981 {
3982 char *cp;
3983
3984 /*
3985 * We don't need to unmount *all* the origin fs's snapshots, but
3986 * it's easier.
3987 */
3988 cp = strchr(zc->zc_value, '@');
3989 if (cp)
3990 *cp = '\0';
3991 (void) dmu_objset_find(zc->zc_value,
3992 zfs_unmount_snap, NULL, DS_FIND_SNAPSHOTS);
3993 return (dsl_dataset_promote(zc->zc_name, zc->zc_string));
3994 }
3995
3996 /*
3997 * Retrieve a single {user|group}{used|quota}@... property.
3998 *
3999 * inputs:
4000 * zc_name name of filesystem
4001 * zc_objset_type zfs_userquota_prop_t
4002 * zc_value domain name (eg. "S-1-234-567-89")
4003 * zc_guid RID/UID/GID
4004 *
4005 * outputs:
4006 * zc_cookie property value
4007 */
4008 static int
4009 zfs_ioc_userspace_one(zfs_cmd_t *zc)
4010 {
4011 zfsvfs_t *zfsvfs;
4012 int error;
4013
4014 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
4015 return (EINVAL);
4016
4017 error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE);
4018 if (error)
4019 return (error);
4020
4021 error = zfs_userspace_one(zfsvfs,
4022 zc->zc_objset_type, zc->zc_value, zc->zc_guid, &zc->zc_cookie);
4023 zfsvfs_rele(zfsvfs, FTAG);
4024
4025 return (error);
4026 }
4027
4028 /*
4029 * inputs:
4030 * zc_name name of filesystem
4031 * zc_cookie zap cursor
4032 * zc_objset_type zfs_userquota_prop_t
4033 * zc_nvlist_dst[_size] buffer to fill (not really an nvlist)
4034 *
4035 * outputs:
4036 * zc_nvlist_dst[_size] data buffer (array of zfs_useracct_t)
4037 * zc_cookie zap cursor
4038 */
4039 static int
4040 zfs_ioc_userspace_many(zfs_cmd_t *zc)
4041 {
4042 zfsvfs_t *zfsvfs;
4043 int bufsize = zc->zc_nvlist_dst_size;
4044
4045 if (bufsize <= 0)
4046 return (ENOMEM);
4047
4048 int error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE);
4049 if (error)
4050 return (error);
4051
4052 void *buf = kmem_alloc(bufsize, KM_SLEEP);
4053
4054 error = zfs_userspace_many(zfsvfs, zc->zc_objset_type, &zc->zc_cookie,
4055 buf, &zc->zc_nvlist_dst_size);
4056
4057 if (error == 0) {
4058 error = xcopyout(buf,
4059 (void *)(uintptr_t)zc->zc_nvlist_dst,
4060 zc->zc_nvlist_dst_size);
4061 }
4062 kmem_free(buf, bufsize);
4063 zfsvfs_rele(zfsvfs, FTAG);
4064
4065 return (error);
4066 }
4067
4068 /*
4069 * inputs:
4070 * zc_name name of filesystem
4071 *
4072 * outputs:
4073 * none
4074 */
4075 static int
4076 zfs_ioc_userspace_upgrade(zfs_cmd_t *zc)
4077 {
4078 objset_t *os;
4079 int error = 0;
4080 zfsvfs_t *zfsvfs;
4081
4082 if (getzfsvfs(zc->zc_name, &zfsvfs) == 0) {
4083 if (!dmu_objset_userused_enabled(zfsvfs->z_os)) {
4084 /*
4085 * If userused is not enabled, it may be because the
4086 * objset needs to be closed & reopened (to grow the
4087 * objset_phys_t). Suspend/resume the fs will do that.
4088 */
4089 error = zfs_suspend_fs(zfsvfs);
4090 if (error == 0)
4091 error = zfs_resume_fs(zfsvfs, zc->zc_name);
4092 }
4093 if (error == 0)
4094 error = dmu_objset_userspace_upgrade(zfsvfs->z_os);
4095 VFS_RELE(zfsvfs->z_vfs);
4096 } else {
4097 /* XXX kind of reading contents without owning */
4098 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
4099 if (error)
4100 return (error);
4101
4102 error = dmu_objset_userspace_upgrade(os);
4103 dmu_objset_rele(os, FTAG);
4104 }
4105
4106 return (error);
4107 }
4108
4109 /*
4110 * We don't want to have a hard dependency
4111 * against some special symbols in sharefs
4112 * nfs, and smbsrv. Determine them if needed when
4113 * the first file system is shared.
4114 * Neither sharefs, nfs or smbsrv are unloadable modules.
4115 */
4116 int (*znfsexport_fs)(void *arg);
4117 int (*zshare_fs)(enum sharefs_sys_op, share_t *, uint32_t);
4118 int (*zsmbexport_fs)(void *arg, boolean_t add_share);
4119
4120 int zfs_nfsshare_inited;
4121 int zfs_smbshare_inited;
4122
4123 ddi_modhandle_t nfs_mod;
4124 ddi_modhandle_t sharefs_mod;
4125 ddi_modhandle_t smbsrv_mod;
4126 kmutex_t zfs_share_lock;
4127
4128 static int
4129 zfs_init_sharefs()
4130 {
4131 int error;
4132
4133 ASSERT(MUTEX_HELD(&zfs_share_lock));
4134 /* Both NFS and SMB shares also require sharetab support. */
4135 if (sharefs_mod == NULL && ((sharefs_mod =
4136 ddi_modopen("fs/sharefs",
4137 KRTLD_MODE_FIRST, &error)) == NULL)) {
4138 return (ENOSYS);
4139 }
4140 if (zshare_fs == NULL && ((zshare_fs =
4141 (int (*)(enum sharefs_sys_op, share_t *, uint32_t))
4142 ddi_modsym(sharefs_mod, "sharefs_impl", &error)) == NULL)) {
4143 return (ENOSYS);
4144 }
4145 return (0);
4146 }
4147
4148 static int
4149 zfs_ioc_share(zfs_cmd_t *zc)
4150 {
4151 int error;
4152 int opcode;
4153
4154 switch (zc->zc_share.z_sharetype) {
4155 case ZFS_SHARE_NFS:
4156 case ZFS_UNSHARE_NFS:
4157 if (zfs_nfsshare_inited == 0) {
4158 mutex_enter(&zfs_share_lock);
4159 if (nfs_mod == NULL && ((nfs_mod = ddi_modopen("fs/nfs",
4160 KRTLD_MODE_FIRST, &error)) == NULL)) {
4161 mutex_exit(&zfs_share_lock);
4162 return (ENOSYS);
4163 }
4164 if (znfsexport_fs == NULL &&
4165 ((znfsexport_fs = (int (*)(void *))
4166 ddi_modsym(nfs_mod,
4167 "nfs_export", &error)) == NULL)) {
4168 mutex_exit(&zfs_share_lock);
4169 return (ENOSYS);
4170 }
4171 error = zfs_init_sharefs();
4172 if (error) {
4173 mutex_exit(&zfs_share_lock);
4174 return (ENOSYS);
4175 }
4176 zfs_nfsshare_inited = 1;
4177 mutex_exit(&zfs_share_lock);
4178 }
4179 break;
4180 case ZFS_SHARE_SMB:
4181 case ZFS_UNSHARE_SMB:
4182 if (zfs_smbshare_inited == 0) {
4183 mutex_enter(&zfs_share_lock);
4184 if (smbsrv_mod == NULL && ((smbsrv_mod =
4185 ddi_modopen("drv/smbsrv",
4186 KRTLD_MODE_FIRST, &error)) == NULL)) {
4187 mutex_exit(&zfs_share_lock);
4188 return (ENOSYS);
4189 }
4190 if (zsmbexport_fs == NULL && ((zsmbexport_fs =
4191 (int (*)(void *, boolean_t))ddi_modsym(smbsrv_mod,
4192 "smb_server_share", &error)) == NULL)) {
4193 mutex_exit(&zfs_share_lock);
4194 return (ENOSYS);
4195 }
4196 error = zfs_init_sharefs();
4197 if (error) {
4198 mutex_exit(&zfs_share_lock);
4199 return (ENOSYS);
4200 }
4201 zfs_smbshare_inited = 1;
4202 mutex_exit(&zfs_share_lock);
4203 }
4204 break;
4205 default:
4206 return (EINVAL);
4207 }
4208
4209 switch (zc->zc_share.z_sharetype) {
4210 case ZFS_SHARE_NFS:
4211 case ZFS_UNSHARE_NFS:
4212 if (error =
4213 znfsexport_fs((void *)
4214 (uintptr_t)zc->zc_share.z_exportdata))
4215 return (error);
4216 break;
4217 case ZFS_SHARE_SMB:
4218 case ZFS_UNSHARE_SMB:
4219 if (error = zsmbexport_fs((void *)
4220 (uintptr_t)zc->zc_share.z_exportdata,
4221 zc->zc_share.z_sharetype == ZFS_SHARE_SMB ?
4222 B_TRUE: B_FALSE)) {
4223 return (error);
4224 }
4225 break;
4226 }
4227
4228 opcode = (zc->zc_share.z_sharetype == ZFS_SHARE_NFS ||
4229 zc->zc_share.z_sharetype == ZFS_SHARE_SMB) ?
4230 SHAREFS_ADD : SHAREFS_REMOVE;
4231
4232 /*
4233 * Add or remove share from sharetab
4234 */
4235 error = zshare_fs(opcode,
4236 (void *)(uintptr_t)zc->zc_share.z_sharedata,
4237 zc->zc_share.z_sharemax);
4238
4239 return (error);
4240
4241 }
4242
4243 ace_t full_access[] = {
4244 {(uid_t)-1, ACE_ALL_PERMS, ACE_EVERYONE, 0}
4245 };
4246
4247 /*
4248 * inputs:
4249 * zc_name name of containing filesystem
4250 * zc_obj object # beyond which we want next in-use object #
4251 *
4252 * outputs:
4253 * zc_obj next in-use object #
4254 */
4255 static int
4256 zfs_ioc_next_obj(zfs_cmd_t *zc)
4257 {
4258 objset_t *os = NULL;
4259 int error;
4260
4261 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
4262 if (error)
4263 return (error);
4264
4265 error = dmu_object_next(os, &zc->zc_obj, B_FALSE,
4266 os->os_dsl_dataset->ds_phys->ds_prev_snap_txg);
4267
4268 dmu_objset_rele(os, FTAG);
4269 return (error);
4270 }
4271
4272 /*
4273 * inputs:
4274 * zc_name name of filesystem
4275 * zc_value prefix name for snapshot
4276 * zc_cleanup_fd cleanup-on-exit file descriptor for calling process
4277 *
4278 * outputs:
4279 */
4280 static int
4281 zfs_ioc_tmp_snapshot(zfs_cmd_t *zc)
4282 {
4283 char *snap_name;
4284 int error;
4285
4286 snap_name = kmem_asprintf("%s-%016llx", zc->zc_value,
4287 (u_longlong_t)ddi_get_lbolt64());
4288
4289 if (strlen(snap_name) >= MAXNAMELEN) {
4290 strfree(snap_name);
4291 return (E2BIG);
4292 }
4293
4294 error = dmu_objset_snapshot(zc->zc_name, snap_name, snap_name,
4295 NULL, B_FALSE, B_TRUE, zc->zc_cleanup_fd);
4296 if (error != 0) {
4297 strfree(snap_name);
4298 return (error);
4299 }
4300
4301 (void) strcpy(zc->zc_value, snap_name);
4302 strfree(snap_name);
4303 return (0);
4304 }
4305
4306 /*
4307 * inputs:
4308 * zc_name name of "to" snapshot
4309 * zc_value name of "from" snapshot
4310 * zc_cookie file descriptor to write diff data on
4311 *
4312 * outputs:
4313 * dmu_diff_record_t's to the file descriptor
4314 */
4315 static int
4316 zfs_ioc_diff(zfs_cmd_t *zc)
4317 {
4318 objset_t *fromsnap;
4319 objset_t *tosnap;
4320 file_t *fp;
4321 offset_t off;
4322 int error;
4323
4324 error = dmu_objset_hold(zc->zc_name, FTAG, &tosnap);
4325 if (error)
4326 return (error);
4327
4328 error = dmu_objset_hold(zc->zc_value, FTAG, &fromsnap);
4329 if (error) {
4330 dmu_objset_rele(tosnap, FTAG);
4331 return (error);
4332 }
4333
4334 fp = getf(zc->zc_cookie);
4335 if (fp == NULL) {
4336 dmu_objset_rele(fromsnap, FTAG);
4337 dmu_objset_rele(tosnap, FTAG);
4338 return (EBADF);
4339 }
4340
4341 off = fp->f_offset;
4342
4343 error = dmu_diff(tosnap, fromsnap, fp->f_vnode, &off);
4344
4345 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
4346 fp->f_offset = off;
4347 releasef(zc->zc_cookie);
4348
4349 dmu_objset_rele(fromsnap, FTAG);
4350 dmu_objset_rele(tosnap, FTAG);
4351 return (error);
4352 }
4353
4354 /*
4355 * Remove all ACL files in shares dir
4356 */
4357 static int
4358 zfs_smb_acl_purge(znode_t *dzp)
4359 {
4360 zap_cursor_t zc;
4361 zap_attribute_t zap;
4362 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
4363 int error;
4364
4365 for (zap_cursor_init(&zc, zfsvfs->z_os, dzp->z_id);
4366 (error = zap_cursor_retrieve(&zc, &zap)) == 0;
4367 zap_cursor_advance(&zc)) {
4368 if ((error = VOP_REMOVE(ZTOV(dzp), zap.za_name, kcred,
4369 NULL, 0)) != 0)
4370 break;
4371 }
4372 zap_cursor_fini(&zc);
4373 return (error);
4374 }
4375
4376 static int
4377 zfs_ioc_smb_acl(zfs_cmd_t *zc)
4378 {
4379 vnode_t *vp;
4380 znode_t *dzp;
4381 vnode_t *resourcevp = NULL;
4382 znode_t *sharedir;
4383 zfsvfs_t *zfsvfs;
4384 nvlist_t *nvlist;
4385 char *src, *target;
4386 vattr_t vattr;
4387 vsecattr_t vsec;
4388 int error = 0;
4389
4390 if ((error = lookupname(zc->zc_value, UIO_SYSSPACE,
4391 NO_FOLLOW, NULL, &vp)) != 0)
4392 return (error);
4393
4394 /* Now make sure mntpnt and dataset are ZFS */
4395
4396 if (vp->v_vfsp->vfs_fstype != zfsfstype ||
4397 (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource),
4398 zc->zc_name) != 0)) {
4399 VN_RELE(vp);
4400 return (EINVAL);
4401 }
4402
4403 dzp = VTOZ(vp);
4404 zfsvfs = dzp->z_zfsvfs;
4405 ZFS_ENTER(zfsvfs);
4406
4407 /*
4408 * Create share dir if its missing.
4409 */
4410 mutex_enter(&zfsvfs->z_lock);
4411 if (zfsvfs->z_shares_dir == 0) {
4412 dmu_tx_t *tx;
4413
4414 tx = dmu_tx_create(zfsvfs->z_os);
4415 dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, TRUE,
4416 ZFS_SHARES_DIR);
4417 dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL);
4418 error = dmu_tx_assign(tx, TXG_WAIT);
4419 if (error) {
4420 dmu_tx_abort(tx);
4421 } else {
4422 error = zfs_create_share_dir(zfsvfs, tx);
4423 dmu_tx_commit(tx);
4424 }
4425 if (error) {
4426 mutex_exit(&zfsvfs->z_lock);
4427 VN_RELE(vp);
4428 ZFS_EXIT(zfsvfs);
4429 return (error);
4430 }
4431 }
4432 mutex_exit(&zfsvfs->z_lock);
4433
4434 ASSERT(zfsvfs->z_shares_dir);
4435 if ((error = zfs_zget(zfsvfs, zfsvfs->z_shares_dir, &sharedir)) != 0) {
4436 VN_RELE(vp);
4437 ZFS_EXIT(zfsvfs);
4438 return (error);
4439 }
4440
4441 switch (zc->zc_cookie) {
4442 case ZFS_SMB_ACL_ADD:
4443 vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE;
4444 vattr.va_type = VREG;
4445 vattr.va_mode = S_IFREG|0777;
4446 vattr.va_uid = 0;
4447 vattr.va_gid = 0;
4448
4449 vsec.vsa_mask = VSA_ACE;
4450 vsec.vsa_aclentp = &full_access;
4451 vsec.vsa_aclentsz = sizeof (full_access);
4452 vsec.vsa_aclcnt = 1;
4453
4454 error = VOP_CREATE(ZTOV(sharedir), zc->zc_string,
4455 &vattr, EXCL, 0, &resourcevp, kcred, 0, NULL, &vsec);
4456 if (resourcevp)
4457 VN_RELE(resourcevp);
4458 break;
4459
4460 case ZFS_SMB_ACL_REMOVE:
4461 error = VOP_REMOVE(ZTOV(sharedir), zc->zc_string, kcred,
4462 NULL, 0);
4463 break;
4464
4465 case ZFS_SMB_ACL_RENAME:
4466 if ((error = get_nvlist(zc->zc_nvlist_src,
4467 zc->zc_nvlist_src_size, zc->zc_iflags, &nvlist)) != 0) {
4468 VN_RELE(vp);
4469 ZFS_EXIT(zfsvfs);
4470 return (error);
4471 }
4472 if (nvlist_lookup_string(nvlist, ZFS_SMB_ACL_SRC, &src) ||
4473 nvlist_lookup_string(nvlist, ZFS_SMB_ACL_TARGET,
4474 &target)) {
4475 VN_RELE(vp);
4476 VN_RELE(ZTOV(sharedir));
4477 ZFS_EXIT(zfsvfs);
4478 nvlist_free(nvlist);
4479 return (error);
4480 }
4481 error = VOP_RENAME(ZTOV(sharedir), src, ZTOV(sharedir), target,
4482 kcred, NULL, 0);
4483 nvlist_free(nvlist);
4484 break;
4485
4486 case ZFS_SMB_ACL_PURGE:
4487 error = zfs_smb_acl_purge(sharedir);
4488 break;
4489
4490 default:
4491 error = EINVAL;
4492 break;
4493 }
4494
4495 VN_RELE(vp);
4496 VN_RELE(ZTOV(sharedir));
4497
4498 ZFS_EXIT(zfsvfs);
4499
4500 return (error);
4501 }
4502
4503 /*
4504 * inputs:
4505 * zc_name name of filesystem
4506 * zc_value short name of snap
4507 * zc_string user-supplied tag for this hold
4508 * zc_cookie recursive flag
4509 * zc_temphold set if hold is temporary
4510 * zc_cleanup_fd cleanup-on-exit file descriptor for calling process
4511 * zc_sendobj if non-zero, the objid for zc_name@zc_value
4512 * zc_createtxg if zc_sendobj is non-zero, snap must have zc_createtxg
4513 *
4514 * outputs: none
4515 */
4516 static int
4517 zfs_ioc_hold(zfs_cmd_t *zc)
4518 {
4519 boolean_t recursive = zc->zc_cookie;
4520 spa_t *spa;
4521 dsl_pool_t *dp;
4522 dsl_dataset_t *ds;
4523 int error;
4524 minor_t minor = 0;
4525
4526 if (snapshot_namecheck(zc->zc_value, NULL, NULL) != 0)
4527 return (EINVAL);
4528
4529 if (zc->zc_sendobj == 0) {
4530 return (dsl_dataset_user_hold(zc->zc_name, zc->zc_value,
4531 zc->zc_string, recursive, zc->zc_temphold,
4532 zc->zc_cleanup_fd));
4533 }
4534
4535 if (recursive)
4536 return (EINVAL);
4537
4538 error = spa_open(zc->zc_name, &spa, FTAG);
4539 if (error)
4540 return (error);
4541
4542 dp = spa_get_dsl(spa);
4543 rw_enter(&dp->dp_config_rwlock, RW_READER);
4544 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &ds);
4545 rw_exit(&dp->dp_config_rwlock);
4546 spa_close(spa, FTAG);
4547 if (error)
4548 return (error);
4549
4550 /*
4551 * Until we have a hold on this snapshot, it's possible that
4552 * zc_sendobj could've been destroyed and reused as part
4553 * of a later txg. Make sure we're looking at the right object.
4554 */
4555 if (zc->zc_createtxg != ds->ds_phys->ds_creation_txg) {
4556 dsl_dataset_rele(ds, FTAG);
4557 return (ENOENT);
4558 }
4559
4560 if (zc->zc_cleanup_fd != -1 && zc->zc_temphold) {
4561 error = zfs_onexit_fd_hold(zc->zc_cleanup_fd, &minor);
4562 if (error) {
4563 dsl_dataset_rele(ds, FTAG);
4564 return (error);
4565 }
4566 }
4567
4568 error = dsl_dataset_user_hold_for_send(ds, zc->zc_string,
4569 zc->zc_temphold);
4570 if (minor != 0) {
4571 if (error == 0) {
4572 dsl_register_onexit_hold_cleanup(ds, zc->zc_string,
4573 minor);
4574 }
4575 zfs_onexit_fd_rele(zc->zc_cleanup_fd);
4576 }
4577 dsl_dataset_rele(ds, FTAG);
4578
4579 return (error);
4580 }
4581
4582 /*
4583 * inputs:
4584 * zc_name name of dataset from which we're releasing a user hold
4585 * zc_value short name of snap
4586 * zc_string user-supplied tag for this hold
4587 * zc_cookie recursive flag
4588 *
4589 * outputs: none
4590 */
4591 static int
4592 zfs_ioc_release(zfs_cmd_t *zc)
4593 {
4594 boolean_t recursive = zc->zc_cookie;
4595
4596 if (snapshot_namecheck(zc->zc_value, NULL, NULL) != 0)
4597 return (EINVAL);
4598
4599 return (dsl_dataset_user_release(zc->zc_name, zc->zc_value,
4600 zc->zc_string, recursive));
4601 }
4602
4603 /*
4604 * inputs:
4605 * zc_name name of filesystem
4606 *
4607 * outputs:
4608 * zc_nvlist_src{_size} nvlist of snapshot holds
4609 */
4610 static int
4611 zfs_ioc_get_holds(zfs_cmd_t *zc)
4612 {
4613 nvlist_t *nvp;
4614 int error;
4615
4616 if ((error = dsl_dataset_get_holds(zc->zc_name, &nvp)) == 0) {
4617 error = put_nvlist(zc, nvp);
4618 nvlist_free(nvp);
4619 }
4620
4621 return (error);
4622 }
4623
4624 /*
4625 * pool create, destroy, and export don't log the history as part of
4626 * zfsdev_ioctl, but rather zfs_ioc_pool_create, and zfs_ioc_pool_export
4627 * do the logging of those commands.
4628 */
4629 static zfs_ioc_vec_t zfs_ioc_vec[] = {
4630 { zfs_ioc_pool_create, zfs_secpolicy_config, POOL_NAME, B_FALSE,
4631 POOL_CHECK_NONE },
4632 { zfs_ioc_pool_destroy, zfs_secpolicy_config, POOL_NAME, B_FALSE,
4633 POOL_CHECK_NONE },
4634 { zfs_ioc_pool_import, zfs_secpolicy_config, POOL_NAME, B_TRUE,
4635 POOL_CHECK_NONE },
4636 { zfs_ioc_pool_export, zfs_secpolicy_config, POOL_NAME, B_FALSE,
4637 POOL_CHECK_NONE },
4638 { zfs_ioc_pool_configs, zfs_secpolicy_none, NO_NAME, B_FALSE,
4639 POOL_CHECK_NONE },
4640 { zfs_ioc_pool_stats, zfs_secpolicy_read, POOL_NAME, B_FALSE,
4641 POOL_CHECK_NONE },
4642 { zfs_ioc_pool_tryimport, zfs_secpolicy_config, NO_NAME, B_FALSE,
4643 POOL_CHECK_NONE },
4644 { zfs_ioc_pool_scan, zfs_secpolicy_config, POOL_NAME, B_TRUE,
4645 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4646 { zfs_ioc_pool_freeze, zfs_secpolicy_config, NO_NAME, B_FALSE,
4647 POOL_CHECK_READONLY },
4648 { zfs_ioc_pool_upgrade, zfs_secpolicy_config, POOL_NAME, B_TRUE,
4649 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4650 { zfs_ioc_pool_get_history, zfs_secpolicy_config, POOL_NAME, B_FALSE,
4651 POOL_CHECK_NONE },
4652 { zfs_ioc_vdev_add, zfs_secpolicy_config, POOL_NAME, B_TRUE,
4653 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4654 { zfs_ioc_vdev_remove, zfs_secpolicy_config, POOL_NAME, B_TRUE,
4655 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4656 { zfs_ioc_vdev_set_state, zfs_secpolicy_config, POOL_NAME, B_TRUE,
4657 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4658 { zfs_ioc_vdev_attach, zfs_secpolicy_config, POOL_NAME, B_TRUE,
4659 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4660 { zfs_ioc_vdev_detach, zfs_secpolicy_config, POOL_NAME, B_TRUE,
4661 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4662 { zfs_ioc_vdev_setpath, zfs_secpolicy_config, POOL_NAME, B_FALSE,
4663 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4664 { zfs_ioc_vdev_setfru, zfs_secpolicy_config, POOL_NAME, B_FALSE,
4665 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4666 { zfs_ioc_objset_stats, zfs_secpolicy_read, DATASET_NAME, B_FALSE,
4667 POOL_CHECK_SUSPENDED },
4668 { zfs_ioc_objset_zplprops, zfs_secpolicy_read, DATASET_NAME, B_FALSE,
4669 POOL_CHECK_NONE },
4670 { zfs_ioc_dataset_list_next, zfs_secpolicy_read, DATASET_NAME, B_FALSE,
4671 POOL_CHECK_SUSPENDED },
4672 { zfs_ioc_snapshot_list_next, zfs_secpolicy_read, DATASET_NAME, B_FALSE,
4673 POOL_CHECK_SUSPENDED },
4674 { zfs_ioc_set_prop, zfs_secpolicy_none, DATASET_NAME, B_TRUE,
4675 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4676 { zfs_ioc_create, zfs_secpolicy_create, DATASET_NAME, B_TRUE,
4677 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4678 { zfs_ioc_destroy, zfs_secpolicy_destroy, DATASET_NAME, B_TRUE,
4679 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4680 { zfs_ioc_rollback, zfs_secpolicy_rollback, DATASET_NAME, B_TRUE,
4681 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4682 { zfs_ioc_rename, zfs_secpolicy_rename, DATASET_NAME, B_TRUE,
4683 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4684 { zfs_ioc_recv, zfs_secpolicy_receive, DATASET_NAME, B_TRUE,
4685 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4686 { zfs_ioc_send, zfs_secpolicy_send, DATASET_NAME, B_TRUE,
4687 POOL_CHECK_NONE },
4688 { zfs_ioc_inject_fault, zfs_secpolicy_inject, NO_NAME, B_FALSE,
4689 POOL_CHECK_NONE },
4690 { zfs_ioc_clear_fault, zfs_secpolicy_inject, NO_NAME, B_FALSE,
4691 POOL_CHECK_NONE },
4692 { zfs_ioc_inject_list_next, zfs_secpolicy_inject, NO_NAME, B_FALSE,
4693 POOL_CHECK_NONE },
4694 { zfs_ioc_error_log, zfs_secpolicy_inject, POOL_NAME, B_FALSE,
4695 POOL_CHECK_NONE },
4696 { zfs_ioc_clear, zfs_secpolicy_config, POOL_NAME, B_TRUE,
4697 POOL_CHECK_NONE },
4698 { zfs_ioc_promote, zfs_secpolicy_promote, DATASET_NAME, B_TRUE,
4699 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4700 { zfs_ioc_destroy_snaps, zfs_secpolicy_destroy_snaps, DATASET_NAME,
4701 B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4702 { zfs_ioc_snapshot, zfs_secpolicy_snapshot, DATASET_NAME, B_TRUE,
4703 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4704 { zfs_ioc_dsobj_to_dsname, zfs_secpolicy_diff, POOL_NAME, B_FALSE,
4705 POOL_CHECK_NONE },
4706 { zfs_ioc_obj_to_path, zfs_secpolicy_diff, DATASET_NAME, B_FALSE,
4707 POOL_CHECK_SUSPENDED },
4708 { zfs_ioc_pool_set_props, zfs_secpolicy_config, POOL_NAME, B_TRUE,
4709 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4710 { zfs_ioc_pool_get_props, zfs_secpolicy_read, POOL_NAME, B_FALSE,
4711 POOL_CHECK_NONE },
4712 { zfs_ioc_set_fsacl, zfs_secpolicy_fsacl, DATASET_NAME, B_TRUE,
4713 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4714 { zfs_ioc_get_fsacl, zfs_secpolicy_read, DATASET_NAME, B_FALSE,
4715 POOL_CHECK_NONE },
4716 { zfs_ioc_share, zfs_secpolicy_share, DATASET_NAME, B_FALSE,
4717 POOL_CHECK_NONE },
4718 { zfs_ioc_inherit_prop, zfs_secpolicy_inherit, DATASET_NAME, B_TRUE,
4719 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4720 { zfs_ioc_smb_acl, zfs_secpolicy_smb_acl, DATASET_NAME, B_FALSE,
4721 POOL_CHECK_NONE },
4722 { zfs_ioc_userspace_one, zfs_secpolicy_userspace_one, DATASET_NAME,
4723 B_FALSE, POOL_CHECK_NONE },
4724 { zfs_ioc_userspace_many, zfs_secpolicy_userspace_many, DATASET_NAME,
4725 B_FALSE, POOL_CHECK_NONE },
4726 { zfs_ioc_userspace_upgrade, zfs_secpolicy_userspace_upgrade,
4727 DATASET_NAME, B_FALSE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4728 { zfs_ioc_hold, zfs_secpolicy_hold, DATASET_NAME, B_TRUE,
4729 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4730 { zfs_ioc_release, zfs_secpolicy_release, DATASET_NAME, B_TRUE,
4731 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4732 { zfs_ioc_get_holds, zfs_secpolicy_read, DATASET_NAME, B_FALSE,
4733 POOL_CHECK_SUSPENDED },
4734 { zfs_ioc_objset_recvd_props, zfs_secpolicy_read, DATASET_NAME, B_FALSE,
4735 POOL_CHECK_NONE },
4736 { zfs_ioc_vdev_split, zfs_secpolicy_config, POOL_NAME, B_TRUE,
4737 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4738 { zfs_ioc_next_obj, zfs_secpolicy_read, DATASET_NAME, B_FALSE,
4739 POOL_CHECK_NONE },
4740 { zfs_ioc_diff, zfs_secpolicy_diff, DATASET_NAME, B_FALSE,
4741 POOL_CHECK_NONE },
4742 { zfs_ioc_tmp_snapshot, zfs_secpolicy_tmp_snapshot, DATASET_NAME,
4743 B_FALSE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4744 { zfs_ioc_obj_to_stats, zfs_secpolicy_diff, DATASET_NAME, B_FALSE,
4745 POOL_CHECK_SUSPENDED }
4746 };
4747
4748 int
4749 pool_status_check(const char *name, zfs_ioc_namecheck_t type,
4750 zfs_ioc_poolcheck_t check)
4751 {
4752 spa_t *spa;
4753 int error;
4754
4755 ASSERT(type == POOL_NAME || type == DATASET_NAME);
4756
4757 if (check & POOL_CHECK_NONE)
4758 return (0);
4759
4760 error = spa_open(name, &spa, FTAG);
4761 if (error == 0) {
4762 if ((check & POOL_CHECK_SUSPENDED) && spa_suspended(spa))
4763 error = EAGAIN;
4764 else if ((check & POOL_CHECK_READONLY) && !spa_writeable(spa))
4765 error = EROFS;
4766 spa_close(spa, FTAG);
4767 }
4768 return (error);
4769 }
4770
4771 /*
4772 * Find a free minor number.
4773 */
4774 minor_t
4775 zfsdev_minor_alloc(void)
4776 {
4777 static minor_t last_minor;
4778 minor_t m;
4779
4780 ASSERT(MUTEX_HELD(&zfsdev_state_lock));
4781
4782 for (m = last_minor + 1; m != last_minor; m++) {
4783 if (m > ZFSDEV_MAX_MINOR)
4784 m = 1;
4785 if (ddi_get_soft_state(zfsdev_state, m) == NULL) {
4786 last_minor = m;
4787 return (m);
4788 }
4789 }
4790
4791 return (0);
4792 }
4793
4794 static int
4795 zfs_ctldev_init(dev_t *devp)
4796 {
4797 minor_t minor;
4798 zfs_soft_state_t *zs;
4799
4800 ASSERT(MUTEX_HELD(&zfsdev_state_lock));
4801 ASSERT(getminor(*devp) == 0);
4802
4803 minor = zfsdev_minor_alloc();
4804 if (minor == 0)
4805 return (ENXIO);
4806
4807 if (ddi_soft_state_zalloc(zfsdev_state, minor) != DDI_SUCCESS)
4808 return (EAGAIN);
4809
4810 *devp = makedevice(getemajor(*devp), minor);
4811
4812 zs = ddi_get_soft_state(zfsdev_state, minor);
4813 zs->zss_type = ZSST_CTLDEV;
4814 zfs_onexit_init((zfs_onexit_t **)&zs->zss_data);
4815
4816 return (0);
4817 }
4818
4819 static void
4820 zfs_ctldev_destroy(zfs_onexit_t *zo, minor_t minor)
4821 {
4822 ASSERT(MUTEX_HELD(&zfsdev_state_lock));
4823
4824 zfs_onexit_destroy(zo);
4825 ddi_soft_state_free(zfsdev_state, minor);
4826 }
4827
4828 void *
4829 zfsdev_get_soft_state(minor_t minor, enum zfs_soft_state_type which)
4830 {
4831 zfs_soft_state_t *zp;
4832
4833 zp = ddi_get_soft_state(zfsdev_state, minor);
4834 if (zp == NULL || zp->zss_type != which)
4835 return (NULL);
4836
4837 return (zp->zss_data);
4838 }
4839
4840 static int
4841 zfsdev_open(dev_t *devp, int flag, int otyp, cred_t *cr)
4842 {
4843 int error = 0;
4844
4845 if (getminor(*devp) != 0)
4846 return (zvol_open(devp, flag, otyp, cr));
4847
4848 /* This is the control device. Allocate a new minor if requested. */
4849 if (flag & FEXCL) {
4850 mutex_enter(&zfsdev_state_lock);
4851 error = zfs_ctldev_init(devp);
4852 mutex_exit(&zfsdev_state_lock);
4853 }
4854
4855 return (error);
4856 }
4857
4858 static int
4859 zfsdev_close(dev_t dev, int flag, int otyp, cred_t *cr)
4860 {
4861 zfs_onexit_t *zo;
4862 minor_t minor = getminor(dev);
4863
4864 if (minor == 0)
4865 return (0);
4866
4867 mutex_enter(&zfsdev_state_lock);
4868 zo = zfsdev_get_soft_state(minor, ZSST_CTLDEV);
4869 if (zo == NULL) {
4870 mutex_exit(&zfsdev_state_lock);
4871 return (zvol_close(dev, flag, otyp, cr));
4872 }
4873 zfs_ctldev_destroy(zo, minor);
4874 mutex_exit(&zfsdev_state_lock);
4875
4876 return (0);
4877 }
4878
4879 static int
4880 zfsdev_ioctl(dev_t dev, int cmd, intptr_t arg, int flag, cred_t *cr, int *rvalp)
4881 {
4882 zfs_cmd_t *zc;
4883 uint_t vec;
4884 int error, rc;
4885 minor_t minor = getminor(dev);
4886
4887 if (minor != 0 &&
4888 zfsdev_get_soft_state(minor, ZSST_CTLDEV) == NULL)
4889 return (zvol_ioctl(dev, cmd, arg, flag, cr, rvalp));
4890
4891 vec = cmd - ZFS_IOC;
4892 ASSERT3U(getmajor(dev), ==, ddi_driver_major(zfs_dip));
4893
4894 if (vec >= sizeof (zfs_ioc_vec) / sizeof (zfs_ioc_vec[0]))
4895 return (EINVAL);
4896
4897 zc = kmem_zalloc(sizeof (zfs_cmd_t), KM_SLEEP);
4898
4899 error = ddi_copyin((void *)arg, zc, sizeof (zfs_cmd_t), flag);
4900 if (error != 0)
4901 error = EFAULT;
4902
4903 if ((error == 0) && !(flag & FKIOCTL))
4904 error = zfs_ioc_vec[vec].zvec_secpolicy(zc, cr);
4905
4906 /*
4907 * Ensure that all pool/dataset names are valid before we pass down to
4908 * the lower layers.
4909 */
4910 if (error == 0) {
4911 zc->zc_name[sizeof (zc->zc_name) - 1] = '\0';
4912 zc->zc_iflags = flag & FKIOCTL;
4913 switch (zfs_ioc_vec[vec].zvec_namecheck) {
4914 case POOL_NAME:
4915 if (pool_namecheck(zc->zc_name, NULL, NULL) != 0)
4916 error = EINVAL;
4917 error = pool_status_check(zc->zc_name,
4918 zfs_ioc_vec[vec].zvec_namecheck,
4919 zfs_ioc_vec[vec].zvec_pool_check);
4920 break;
4921
4922 case DATASET_NAME:
4923 if (dataset_namecheck(zc->zc_name, NULL, NULL) != 0)
4924 error = EINVAL;
4925 error = pool_status_check(zc->zc_name,
4926 zfs_ioc_vec[vec].zvec_namecheck,
4927 zfs_ioc_vec[vec].zvec_pool_check);
4928 break;
4929
4930 case NO_NAME:
4931 break;
4932 }
4933 }
4934
4935 if (error == 0)
4936 error = zfs_ioc_vec[vec].zvec_func(zc);
4937
4938 rc = ddi_copyout(zc, (void *)arg, sizeof (zfs_cmd_t), flag);
4939 if (error == 0) {
4940 if (rc != 0)
4941 error = EFAULT;
4942 if (zfs_ioc_vec[vec].zvec_his_log)
4943 zfs_log_history(zc);
4944 }
4945
4946 kmem_free(zc, sizeof (zfs_cmd_t));
4947 return (error);
4948 }
4949
4950 static int
4951 zfs_attach(dev_info_t *dip, ddi_attach_cmd_t cmd)
4952 {
4953 if (cmd != DDI_ATTACH)
4954 return (DDI_FAILURE);
4955
4956 if (ddi_create_minor_node(dip, "zfs", S_IFCHR, 0,
4957 DDI_PSEUDO, 0) == DDI_FAILURE)
4958 return (DDI_FAILURE);
4959
4960 zfs_dip = dip;
4961
4962 ddi_report_dev(dip);
4963
4964 return (DDI_SUCCESS);
4965 }
4966
4967 static int
4968 zfs_detach(dev_info_t *dip, ddi_detach_cmd_t cmd)
4969 {
4970 if (spa_busy() || zfs_busy() || zvol_busy())
4971 return (DDI_FAILURE);
4972
4973 if (cmd != DDI_DETACH)
4974 return (DDI_FAILURE);
4975
4976 zfs_dip = NULL;
4977
4978 ddi_prop_remove_all(dip);
4979 ddi_remove_minor_node(dip, NULL);
4980
4981 return (DDI_SUCCESS);
4982 }
4983
4984 /*ARGSUSED*/
4985 static int
4986 zfs_info(dev_info_t *dip, ddi_info_cmd_t infocmd, void *arg, void **result)
4987 {
4988 switch (infocmd) {
4989 case DDI_INFO_DEVT2DEVINFO:
4990 *result = zfs_dip;
4991 return (DDI_SUCCESS);
4992
4993 case DDI_INFO_DEVT2INSTANCE:
4994 *result = (void *)0;
4995 return (DDI_SUCCESS);
4996 }
4997
4998 return (DDI_FAILURE);
4999 }
5000
5001 /*
5002 * OK, so this is a little weird.
5003 *
5004 * /dev/zfs is the control node, i.e. minor 0.
5005 * /dev/zvol/[r]dsk/pool/dataset are the zvols, minor > 0.
5006 *
5007 * /dev/zfs has basically nothing to do except serve up ioctls,
5008 * so most of the standard driver entry points are in zvol.c.
5009 */
5010 static struct cb_ops zfs_cb_ops = {
5011 zfsdev_open, /* open */
5012 zfsdev_close, /* close */
5013 zvol_strategy, /* strategy */
5014 nodev, /* print */
5015 zvol_dump, /* dump */
5016 zvol_read, /* read */
5017 zvol_write, /* write */
5018 zfsdev_ioctl, /* ioctl */
5019 nodev, /* devmap */
5020 nodev, /* mmap */
5021 nodev, /* segmap */
5022 nochpoll, /* poll */
5023 ddi_prop_op, /* prop_op */
5024 NULL, /* streamtab */
5025 D_NEW | D_MP | D_64BIT, /* Driver compatibility flag */
5026 CB_REV, /* version */
5027 nodev, /* async read */
5028 nodev, /* async write */
5029 };
5030
5031 static struct dev_ops zfs_dev_ops = {
5032 DEVO_REV, /* version */
5033 0, /* refcnt */
5034 zfs_info, /* info */
5035 nulldev, /* identify */
5036 nulldev, /* probe */
5037 zfs_attach, /* attach */
5038 zfs_detach, /* detach */
5039 nodev, /* reset */
5040 &zfs_cb_ops, /* driver operations */
5041 NULL, /* no bus operations */
5042 NULL, /* power */
5043 ddi_quiesce_not_needed, /* quiesce */
5044 };
5045
5046 static struct modldrv zfs_modldrv = {
5047 &mod_driverops,
5048 "ZFS storage pool",
5049 &zfs_dev_ops
5050 };
5051
5052 static struct modlinkage modlinkage = {
5053 MODREV_1,
5054 (void *)&zfs_modlfs,
5055 (void *)&zfs_modldrv,
5056 NULL
5057 };
5058
5059
5060 uint_t zfs_fsyncer_key;
5061 extern uint_t rrw_tsd_key;
5062
5063 int
5064 _init(void)
5065 {
5066 int error;
5067
5068 spa_init(FREAD | FWRITE);
5069 zfs_init();
5070 zvol_init();
5071
5072 if ((error = mod_install(&modlinkage)) != 0) {
5073 zvol_fini();
5074 zfs_fini();
5075 spa_fini();
5076 return (error);
5077 }
5078
5079 tsd_create(&zfs_fsyncer_key, NULL);
5080 tsd_create(&rrw_tsd_key, NULL);
5081
5082 error = ldi_ident_from_mod(&modlinkage, &zfs_li);
5083 ASSERT(error == 0);
5084 mutex_init(&zfs_share_lock, NULL, MUTEX_DEFAULT, NULL);
5085
5086 return (0);
5087 }
5088
5089 int
5090 _fini(void)
5091 {
5092 int error;
5093
5094 if (spa_busy() || zfs_busy() || zvol_busy() || zio_injection_enabled)
5095 return (EBUSY);
5096
5097 if ((error = mod_remove(&modlinkage)) != 0)
5098 return (error);
5099
5100 zvol_fini();
5101 zfs_fini();
5102 spa_fini();
5103 if (zfs_nfsshare_inited)
5104 (void) ddi_modclose(nfs_mod);
5105 if (zfs_smbshare_inited)
5106 (void) ddi_modclose(smbsrv_mod);
5107 if (zfs_nfsshare_inited || zfs_smbshare_inited)
5108 (void) ddi_modclose(sharefs_mod);
5109
5110 tsd_destroy(&zfs_fsyncer_key);
5111 ldi_ident_release(zfs_li);
5112 zfs_li = NULL;
5113 mutex_destroy(&zfs_share_lock);
5114
5115 return (error);
5116 }
5117
5118 int
5119 _info(struct modinfo *modinfop)
5120 {
5121 return (mod_info(&modlinkage, modinfop));
5122 }