]> git.proxmox.com Git - mirror_zfs.git/blob - module/zfs/zfs_ioctl.c
projects / mirror_zfs.git / blob
? search:


e378d8e58e49c91705fe0bc1c542141b71194134
[mirror_zfs.git] / module / zfs / zfs_ioctl.c
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_dir.h>
62 #include <sys/zfs_onexit.h>
63 #include <sys/zvol.h>
64 #include <sys/dsl_scan.h>
65 #include <sharefs/share.h>
66 #include <sys/dmu_objset.h>
67 #include <sys/fm/util.h>
68
69 #include <linux/miscdevice.h>
70
71 #include "zfs_namecheck.h"
72 #include "zfs_prop.h"
73 #include "zfs_deleg.h"
74 #include "zfs_comutil.h"
75
76 kmutex_t zfsdev_state_lock;
77 list_t zfsdev_state_list;
78
79 extern void zfs_init(void);
80 extern void zfs_fini(void);
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 == 0)
170 return (NULL);
171
172 buf = kmem_alloc(HIS_MAX_RECORD_LEN, KM_SLEEP | KM_NODEBUG);
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 #ifdef HAVE_MLSLABEL
388 char ds_hexsl[MAXNAMELEN];
389 bslabel_t ds_sl, new_sl;
390 boolean_t new_default = FALSE;
391 uint64_t zoned;
392 int needed_priv = -1;
393 int error;
394
395 /* First get the existing dataset label. */
396 error = dsl_prop_get(name, zfs_prop_to_name(ZFS_PROP_MLSLABEL),
397 1, sizeof (ds_hexsl), &ds_hexsl, NULL);
398 if (error)
399 return (EPERM);
400
401 if (strcasecmp(strval, ZFS_MLSLABEL_DEFAULT) == 0)
402 new_default = TRUE;
403
404 /* The label must be translatable */
405 if (!new_default && (hexstr_to_label(strval, &new_sl) != 0))
406 return (EINVAL);
407
408 /*
409 * In a non-global zone, disallow attempts to set a label that
410 * doesn't match that of the zone; otherwise no other checks
411 * are needed.
412 */
413 if (!INGLOBALZONE(curproc)) {
414 if (new_default || !blequal(&new_sl, CR_SL(CRED())))
415 return (EPERM);
416 return (0);
417 }
418
419 /*
420 * For global-zone datasets (i.e., those whose zoned property is
421 * "off", verify that the specified new label is valid for the
422 * global zone.
423 */
424 if (dsl_prop_get_integer(name,
425 zfs_prop_to_name(ZFS_PROP_ZONED), &zoned, NULL))
426 return (EPERM);
427 if (!zoned) {
428 if (zfs_check_global_label(name, strval) != 0)
429 return (EPERM);
430 }
431
432 /*
433 * If the existing dataset label is nondefault, check if the
434 * dataset is mounted (label cannot be changed while mounted).
435 * Get the zfs_sb_t; if there isn't one, then the dataset isn't
436 * mounted (or isn't a dataset, doesn't exist, ...).
437 */
438 if (strcasecmp(ds_hexsl, ZFS_MLSLABEL_DEFAULT) != 0) {
439 objset_t *os;
440 static char *setsl_tag = "setsl_tag";
441
442 /*
443 * Try to own the dataset; abort if there is any error,
444 * (e.g., already mounted, in use, or other error).
445 */
446 error = dmu_objset_own(name, DMU_OST_ZFS, B_TRUE,
447 setsl_tag, &os);
448 if (error)
449 return (EPERM);
450
451 dmu_objset_disown(os, setsl_tag);
452
453 if (new_default) {
454 needed_priv = PRIV_FILE_DOWNGRADE_SL;
455 goto out_check;
456 }
457
458 if (hexstr_to_label(strval, &new_sl) != 0)
459 return (EPERM);
460
461 if (blstrictdom(&ds_sl, &new_sl))
462 needed_priv = PRIV_FILE_DOWNGRADE_SL;
463 else if (blstrictdom(&new_sl, &ds_sl))
464 needed_priv = PRIV_FILE_UPGRADE_SL;
465 } else {
466 /* dataset currently has a default label */
467 if (!new_default)
468 needed_priv = PRIV_FILE_UPGRADE_SL;
469 }
470
471 out_check:
472 if (needed_priv != -1)
473 return (PRIV_POLICY(cr, needed_priv, B_FALSE, EPERM, NULL));
474 return (0);
475 #else
476 return ENOTSUP;
477 #endif /* HAVE_MLSLABEL */
478 }
479
480 static int
481 zfs_secpolicy_setprop(const char *dsname, zfs_prop_t prop, nvpair_t *propval,
482 cred_t *cr)
483 {
484 char *strval;
485
486 /*
487 * Check permissions for special properties.
488 */
489 switch (prop) {
490 default:
491 break;
492 case ZFS_PROP_ZONED:
493 /*
494 * Disallow setting of 'zoned' from within a local zone.
495 */
496 if (!INGLOBALZONE(curproc))
497 return (EPERM);
498 break;
499
500 case ZFS_PROP_QUOTA:
501 if (!INGLOBALZONE(curproc)) {
502 uint64_t zoned;
503 char setpoint[MAXNAMELEN];
504 /*
505 * Unprivileged users are allowed to modify the
506 * quota on things *under* (ie. contained by)
507 * the thing they own.
508 */
509 if (dsl_prop_get_integer(dsname, "zoned", &zoned,
510 setpoint))
511 return (EPERM);
512 if (!zoned || strlen(dsname) <= strlen(setpoint))
513 return (EPERM);
514 }
515 break;
516
517 case ZFS_PROP_MLSLABEL:
518 if (!is_system_labeled())
519 return (EPERM);
520
521 if (nvpair_value_string(propval, &strval) == 0) {
522 int err;
523
524 err = zfs_set_slabel_policy(dsname, strval, CRED());
525 if (err != 0)
526 return (err);
527 }
528 break;
529 }
530
531 return (zfs_secpolicy_write_perms(dsname, zfs_prop_to_name(prop), cr));
532 }
533
534 int
535 zfs_secpolicy_fsacl(zfs_cmd_t *zc, cred_t *cr)
536 {
537 int error;
538
539 error = zfs_dozonecheck(zc->zc_name, cr);
540 if (error)
541 return (error);
542
543 /*
544 * permission to set permissions will be evaluated later in
545 * dsl_deleg_can_allow()
546 */
547 return (0);
548 }
549
550 int
551 zfs_secpolicy_rollback(zfs_cmd_t *zc, cred_t *cr)
552 {
553 return (zfs_secpolicy_write_perms(zc->zc_name,
554 ZFS_DELEG_PERM_ROLLBACK, cr));
555 }
556
557 int
558 zfs_secpolicy_send(zfs_cmd_t *zc, cred_t *cr)
559 {
560 spa_t *spa;
561 dsl_pool_t *dp;
562 dsl_dataset_t *ds;
563 char *cp;
564 int error;
565
566 /*
567 * Generate the current snapshot name from the given objsetid, then
568 * use that name for the secpolicy/zone checks.
569 */
570 cp = strchr(zc->zc_name, '@');
571 if (cp == NULL)
572 return (EINVAL);
573 error = spa_open(zc->zc_name, &spa, FTAG);
574 if (error)
575 return (error);
576
577 dp = spa_get_dsl(spa);
578 rw_enter(&dp->dp_config_rwlock, RW_READER);
579 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &ds);
580 rw_exit(&dp->dp_config_rwlock);
581 spa_close(spa, FTAG);
582 if (error)
583 return (error);
584
585 dsl_dataset_name(ds, zc->zc_name);
586
587 error = zfs_secpolicy_write_perms_ds(zc->zc_name, ds,
588 ZFS_DELEG_PERM_SEND, cr);
589 dsl_dataset_rele(ds, FTAG);
590
591 return (error);
592 }
593
594 #ifdef HAVE_SMB_SHARE
595 static int
596 zfs_secpolicy_deleg_share(zfs_cmd_t *zc, cred_t *cr)
597 {
598 vnode_t *vp;
599 int error;
600
601 if ((error = lookupname(zc->zc_value, UIO_SYSSPACE,
602 NO_FOLLOW, NULL, &vp)) != 0)
603 return (error);
604
605 /* Now make sure mntpnt and dataset are ZFS */
606
607 if (vp->v_vfsp->vfs_fstype != zfsfstype ||
608 (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource),
609 zc->zc_name) != 0)) {
610 VN_RELE(vp);
611 return (EPERM);
612 }
613
614 VN_RELE(vp);
615 return (dsl_deleg_access(zc->zc_name,
616 ZFS_DELEG_PERM_SHARE, cr));
617 }
618 #endif /* HAVE_SMB_SHARE */
619
620 int
621 zfs_secpolicy_share(zfs_cmd_t *zc, cred_t *cr)
622 {
623 #ifdef HAVE_SMB_SHARE
624 if (!INGLOBALZONE(curproc))
625 return (EPERM);
626
627 if (secpolicy_nfs(cr) == 0) {
628 return (0);
629 } else {
630 return (zfs_secpolicy_deleg_share(zc, cr));
631 }
632 #else
633 return (ENOTSUP);
634 #endif /* HAVE_SMB_SHARE */
635 }
636
637 int
638 zfs_secpolicy_smb_acl(zfs_cmd_t *zc, cred_t *cr)
639 {
640 #ifdef HAVE_SMB_SHARE
641 if (!INGLOBALZONE(curproc))
642 return (EPERM);
643
644 if (secpolicy_smb(cr) == 0) {
645 return (0);
646 } else {
647 return (zfs_secpolicy_deleg_share(zc, cr));
648 }
649 #else
650 return (ENOTSUP);
651 #endif /* HAVE_SMB_SHARE */
652 }
653
654 static int
655 zfs_get_parent(const char *datasetname, char *parent, int parentsize)
656 {
657 char *cp;
658
659 /*
660 * Remove the @bla or /bla from the end of the name to get the parent.
661 */
662 (void) strncpy(parent, datasetname, parentsize);
663 cp = strrchr(parent, '@');
664 if (cp != NULL) {
665 cp[0] = '\0';
666 } else {
667 cp = strrchr(parent, '/');
668 if (cp == NULL)
669 return (ENOENT);
670 cp[0] = '\0';
671 }
672
673 return (0);
674 }
675
676 int
677 zfs_secpolicy_destroy_perms(const char *name, cred_t *cr)
678 {
679 int error;
680
681 if ((error = zfs_secpolicy_write_perms(name,
682 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
683 return (error);
684
685 return (zfs_secpolicy_write_perms(name, ZFS_DELEG_PERM_DESTROY, cr));
686 }
687
688 static int
689 zfs_secpolicy_destroy(zfs_cmd_t *zc, cred_t *cr)
690 {
691 return (zfs_secpolicy_destroy_perms(zc->zc_name, cr));
692 }
693
694 /*
695 * Destroying snapshots with delegated permissions requires
696 * descendent mount and destroy permissions.
697 * Reassemble the full filesystem@snap name so dsl_deleg_access()
698 * can do the correct permission check.
699 *
700 * Since this routine is used when doing a recursive destroy of snapshots
701 * and destroying snapshots requires descendent permissions, a successfull
702 * check of the top level snapshot applies to snapshots of all descendent
703 * datasets as well.
704 */
705 static int
706 zfs_secpolicy_destroy_snaps(zfs_cmd_t *zc, cred_t *cr)
707 {
708 int error;
709 char *dsname;
710
711 dsname = kmem_asprintf("%s@%s", zc->zc_name, zc->zc_value);
712
713 error = zfs_secpolicy_destroy_perms(dsname, cr);
714
715 strfree(dsname);
716 return (error);
717 }
718
719 int
720 zfs_secpolicy_rename_perms(const char *from, const char *to, cred_t *cr)
721 {
722 char parentname[MAXNAMELEN];
723 int error;
724
725 if ((error = zfs_secpolicy_write_perms(from,
726 ZFS_DELEG_PERM_RENAME, cr)) != 0)
727 return (error);
728
729 if ((error = zfs_secpolicy_write_perms(from,
730 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
731 return (error);
732
733 if ((error = zfs_get_parent(to, parentname,
734 sizeof (parentname))) != 0)
735 return (error);
736
737 if ((error = zfs_secpolicy_write_perms(parentname,
738 ZFS_DELEG_PERM_CREATE, cr)) != 0)
739 return (error);
740
741 if ((error = zfs_secpolicy_write_perms(parentname,
742 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
743 return (error);
744
745 return (error);
746 }
747
748 static int
749 zfs_secpolicy_rename(zfs_cmd_t *zc, cred_t *cr)
750 {
751 return (zfs_secpolicy_rename_perms(zc->zc_name, zc->zc_value, cr));
752 }
753
754 static int
755 zfs_secpolicy_promote(zfs_cmd_t *zc, cred_t *cr)
756 {
757 char parentname[MAXNAMELEN];
758 objset_t *clone;
759 int error;
760
761 error = zfs_secpolicy_write_perms(zc->zc_name,
762 ZFS_DELEG_PERM_PROMOTE, cr);
763 if (error)
764 return (error);
765
766 error = dmu_objset_hold(zc->zc_name, FTAG, &clone);
767
768 if (error == 0) {
769 dsl_dataset_t *pclone = NULL;
770 dsl_dir_t *dd;
771 dd = clone->os_dsl_dataset->ds_dir;
772
773 rw_enter(&dd->dd_pool->dp_config_rwlock, RW_READER);
774 error = dsl_dataset_hold_obj(dd->dd_pool,
775 dd->dd_phys->dd_origin_obj, FTAG, &pclone);
776 rw_exit(&dd->dd_pool->dp_config_rwlock);
777 if (error) {
778 dmu_objset_rele(clone, FTAG);
779 return (error);
780 }
781
782 error = zfs_secpolicy_write_perms(zc->zc_name,
783 ZFS_DELEG_PERM_MOUNT, cr);
784
785 dsl_dataset_name(pclone, parentname);
786 dmu_objset_rele(clone, FTAG);
787 dsl_dataset_rele(pclone, FTAG);
788 if (error == 0)
789 error = zfs_secpolicy_write_perms(parentname,
790 ZFS_DELEG_PERM_PROMOTE, cr);
791 }
792 return (error);
793 }
794
795 static int
796 zfs_secpolicy_receive(zfs_cmd_t *zc, cred_t *cr)
797 {
798 int error;
799
800 if ((error = zfs_secpolicy_write_perms(zc->zc_name,
801 ZFS_DELEG_PERM_RECEIVE, cr)) != 0)
802 return (error);
803
804 if ((error = zfs_secpolicy_write_perms(zc->zc_name,
805 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
806 return (error);
807
808 return (zfs_secpolicy_write_perms(zc->zc_name,
809 ZFS_DELEG_PERM_CREATE, cr));
810 }
811
812 int
813 zfs_secpolicy_snapshot_perms(const char *name, cred_t *cr)
814 {
815 return (zfs_secpolicy_write_perms(name,
816 ZFS_DELEG_PERM_SNAPSHOT, cr));
817 }
818
819 static int
820 zfs_secpolicy_snapshot(zfs_cmd_t *zc, cred_t *cr)
821 {
822
823 return (zfs_secpolicy_snapshot_perms(zc->zc_name, cr));
824 }
825
826 static int
827 zfs_secpolicy_create(zfs_cmd_t *zc, cred_t *cr)
828 {
829 char parentname[MAXNAMELEN];
830 int error;
831
832 if ((error = zfs_get_parent(zc->zc_name, parentname,
833 sizeof (parentname))) != 0)
834 return (error);
835
836 if (zc->zc_value[0] != '\0') {
837 if ((error = zfs_secpolicy_write_perms(zc->zc_value,
838 ZFS_DELEG_PERM_CLONE, cr)) != 0)
839 return (error);
840 }
841
842 if ((error = zfs_secpolicy_write_perms(parentname,
843 ZFS_DELEG_PERM_CREATE, cr)) != 0)
844 return (error);
845
846 error = zfs_secpolicy_write_perms(parentname,
847 ZFS_DELEG_PERM_MOUNT, cr);
848
849 return (error);
850 }
851
852 /*
853 * Policy for pool operations - create/destroy pools, add vdevs, etc. Requires
854 * SYS_CONFIG privilege, which is not available in a local zone.
855 */
856 /* ARGSUSED */
857 static int
858 zfs_secpolicy_config(zfs_cmd_t *zc, cred_t *cr)
859 {
860 if (secpolicy_sys_config(cr, B_FALSE) != 0)
861 return (EPERM);
862
863 return (0);
864 }
865
866 /*
867 * Policy for object to name lookups.
868 */
869 /* ARGSUSED */
870 static int
871 zfs_secpolicy_diff(zfs_cmd_t *zc, cred_t *cr)
872 {
873 int error;
874
875 if ((error = secpolicy_sys_config(cr, B_FALSE)) == 0)
876 return (0);
877
878 error = zfs_secpolicy_write_perms(zc->zc_name, ZFS_DELEG_PERM_DIFF, cr);
879 return (error);
880 }
881
882 /*
883 * Policy for fault injection. Requires all privileges.
884 */
885 /* ARGSUSED */
886 static int
887 zfs_secpolicy_inject(zfs_cmd_t *zc, cred_t *cr)
888 {
889 return (secpolicy_zinject(cr));
890 }
891
892 static int
893 zfs_secpolicy_inherit(zfs_cmd_t *zc, cred_t *cr)
894 {
895 zfs_prop_t prop = zfs_name_to_prop(zc->zc_value);
896
897 if (prop == ZPROP_INVAL) {
898 if (!zfs_prop_user(zc->zc_value))
899 return (EINVAL);
900 return (zfs_secpolicy_write_perms(zc->zc_name,
901 ZFS_DELEG_PERM_USERPROP, cr));
902 } else {
903 return (zfs_secpolicy_setprop(zc->zc_name, prop,
904 NULL, cr));
905 }
906 }
907
908 static int
909 zfs_secpolicy_userspace_one(zfs_cmd_t *zc, cred_t *cr)
910 {
911 int err = zfs_secpolicy_read(zc, cr);
912 if (err)
913 return (err);
914
915 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
916 return (EINVAL);
917
918 if (zc->zc_value[0] == 0) {
919 /*
920 * They are asking about a posix uid/gid. If it's
921 * themself, allow it.
922 */
923 if (zc->zc_objset_type == ZFS_PROP_USERUSED ||
924 zc->zc_objset_type == ZFS_PROP_USERQUOTA) {
925 if (zc->zc_guid == crgetuid(cr))
926 return (0);
927 } else {
928 if (groupmember(zc->zc_guid, cr))
929 return (0);
930 }
931 }
932
933 return (zfs_secpolicy_write_perms(zc->zc_name,
934 userquota_perms[zc->zc_objset_type], cr));
935 }
936
937 static int
938 zfs_secpolicy_userspace_many(zfs_cmd_t *zc, cred_t *cr)
939 {
940 int err = zfs_secpolicy_read(zc, cr);
941 if (err)
942 return (err);
943
944 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
945 return (EINVAL);
946
947 return (zfs_secpolicy_write_perms(zc->zc_name,
948 userquota_perms[zc->zc_objset_type], cr));
949 }
950
951 static int
952 zfs_secpolicy_userspace_upgrade(zfs_cmd_t *zc, cred_t *cr)
953 {
954 return (zfs_secpolicy_setprop(zc->zc_name, ZFS_PROP_VERSION,
955 NULL, cr));
956 }
957
958 static int
959 zfs_secpolicy_hold(zfs_cmd_t *zc, cred_t *cr)
960 {
961 return (zfs_secpolicy_write_perms(zc->zc_name,
962 ZFS_DELEG_PERM_HOLD, cr));
963 }
964
965 static int
966 zfs_secpolicy_release(zfs_cmd_t *zc, cred_t *cr)
967 {
968 return (zfs_secpolicy_write_perms(zc->zc_name,
969 ZFS_DELEG_PERM_RELEASE, cr));
970 }
971
972 /*
973 * Policy for allowing temporary snapshots to be taken or released
974 */
975 static int
976 zfs_secpolicy_tmp_snapshot(zfs_cmd_t *zc, cred_t *cr)
977 {
978 /*
979 * A temporary snapshot is the same as a snapshot,
980 * hold, destroy and release all rolled into one.
981 * Delegated diff alone is sufficient that we allow this.
982 */
983 int error;
984
985 if ((error = zfs_secpolicy_write_perms(zc->zc_name,
986 ZFS_DELEG_PERM_DIFF, cr)) == 0)
987 return (0);
988
989 error = zfs_secpolicy_snapshot(zc, cr);
990 if (!error)
991 error = zfs_secpolicy_hold(zc, cr);
992 if (!error)
993 error = zfs_secpolicy_release(zc, cr);
994 if (!error)
995 error = zfs_secpolicy_destroy(zc, cr);
996 return (error);
997 }
998
999 /*
1000 * Returns the nvlist as specified by the user in the zfs_cmd_t.
1001 */
1002 static int
1003 get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp)
1004 {
1005 char *packed;
1006 int error;
1007 nvlist_t *list = NULL;
1008
1009 /*
1010 * Read in and unpack the user-supplied nvlist.
1011 */
1012 if (size == 0)
1013 return (EINVAL);
1014
1015 packed = kmem_alloc(size, KM_SLEEP | KM_NODEBUG);
1016
1017 if ((error = ddi_copyin((void *)(uintptr_t)nvl, packed, size,
1018 iflag)) != 0) {
1019 kmem_free(packed, size);
1020 return (error);
1021 }
1022
1023 if ((error = nvlist_unpack(packed, size, &list, 0)) != 0) {
1024 kmem_free(packed, size);
1025 return (error);
1026 }
1027
1028 kmem_free(packed, size);
1029
1030 *nvp = list;
1031 return (0);
1032 }
1033
1034 static int
1035 fit_error_list(zfs_cmd_t *zc, nvlist_t **errors)
1036 {
1037 size_t size;
1038
1039 VERIFY(nvlist_size(*errors, &size, NV_ENCODE_NATIVE) == 0);
1040
1041 if (size > zc->zc_nvlist_dst_size) {
1042 nvpair_t *more_errors;
1043 int n = 0;
1044
1045 if (zc->zc_nvlist_dst_size < 1024)
1046 return (ENOMEM);
1047
1048 VERIFY(nvlist_add_int32(*errors, ZPROP_N_MORE_ERRORS, 0) == 0);
1049 more_errors = nvlist_prev_nvpair(*errors, NULL);
1050
1051 do {
1052 nvpair_t *pair = nvlist_prev_nvpair(*errors,
1053 more_errors);
1054 VERIFY(nvlist_remove_nvpair(*errors, pair) == 0);
1055 n++;
1056 VERIFY(nvlist_size(*errors, &size,
1057 NV_ENCODE_NATIVE) == 0);
1058 } while (size > zc->zc_nvlist_dst_size);
1059
1060 VERIFY(nvlist_remove_nvpair(*errors, more_errors) == 0);
1061 VERIFY(nvlist_add_int32(*errors, ZPROP_N_MORE_ERRORS, n) == 0);
1062 ASSERT(nvlist_size(*errors, &size, NV_ENCODE_NATIVE) == 0);
1063 ASSERT(size <= zc->zc_nvlist_dst_size);
1064 }
1065
1066 return (0);
1067 }
1068
1069 static int
1070 put_nvlist(zfs_cmd_t *zc, nvlist_t *nvl)
1071 {
1072 char *packed = NULL;
1073 int error = 0;
1074 size_t size;
1075
1076 VERIFY(nvlist_size(nvl, &size, NV_ENCODE_NATIVE) == 0);
1077
1078 if (size > zc->zc_nvlist_dst_size) {
1079 error = ENOMEM;
1080 } else {
1081 packed = kmem_alloc(size, KM_SLEEP | KM_NODEBUG);
1082 VERIFY(nvlist_pack(nvl, &packed, &size, NV_ENCODE_NATIVE,
1083 KM_SLEEP) == 0);
1084 if (ddi_copyout(packed, (void *)(uintptr_t)zc->zc_nvlist_dst,
1085 size, zc->zc_iflags) != 0)
1086 error = EFAULT;
1087 kmem_free(packed, size);
1088 }
1089
1090 zc->zc_nvlist_dst_size = size;
1091 return (error);
1092 }
1093
1094 static int
1095 get_zfs_sb(const char *dsname, zfs_sb_t **zsbp)
1096 {
1097 objset_t *os;
1098 int error;
1099
1100 error = dmu_objset_hold(dsname, FTAG, &os);
1101 if (error)
1102 return (error);
1103 if (dmu_objset_type(os) != DMU_OST_ZFS) {
1104 dmu_objset_rele(os, FTAG);
1105 return (EINVAL);
1106 }
1107
1108 mutex_enter(&os->os_user_ptr_lock);
1109 *zsbp = dmu_objset_get_user(os);
1110 if (*zsbp && (*zsbp)->z_sb) {
1111 if (atomic_inc_not_zero(&((*zsbp)->z_sb->s_active)))
1112 error = ESRCH;
1113 } else {
1114 error = ESRCH;
1115 }
1116 mutex_exit(&os->os_user_ptr_lock);
1117 dmu_objset_rele(os, FTAG);
1118 return (error);
1119 }
1120
1121 /*
1122 * Find a zfs_sb_t for a mounted filesystem, or create our own, in which
1123 * case its z_sb will be NULL, and it will be opened as the owner.
1124 */
1125 static int
1126 zfs_sb_hold(const char *name, void *tag, zfs_sb_t **zsbp, boolean_t writer)
1127 {
1128 int error = 0;
1129
1130 if (get_zfs_sb(name, zsbp) != 0)
1131 error = zfs_sb_create(name, zsbp);
1132 if (error == 0) {
1133 rrw_enter(&(*zsbp)->z_teardown_lock, (writer) ? RW_WRITER :
1134 RW_READER, tag);
1135 if ((*zsbp)->z_unmounted) {
1136 /*
1137 * XXX we could probably try again, since the unmounting
1138 * thread should be just about to disassociate the
1139 * objset from the zfsvfs.
1140 */
1141 rrw_exit(&(*zsbp)->z_teardown_lock, tag);
1142 return (EBUSY);
1143 }
1144 }
1145 return (error);
1146 }
1147
1148 static void
1149 zfs_sb_rele(zfs_sb_t *zsb, void *tag)
1150 {
1151 rrw_exit(&zsb->z_teardown_lock, tag);
1152
1153 if (zsb->z_sb) {
1154 deactivate_super(zsb->z_sb);
1155 } else {
1156 dmu_objset_disown(zsb->z_os, zsb);
1157 zfs_sb_free(zsb);
1158 }
1159 }
1160
1161 static int
1162 zfs_ioc_pool_create(zfs_cmd_t *zc)
1163 {
1164 int error;
1165 nvlist_t *config, *props = NULL;
1166 nvlist_t *rootprops = NULL;
1167 nvlist_t *zplprops = NULL;
1168 char *buf;
1169
1170 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1171 zc->zc_iflags, &config)))
1172 return (error);
1173
1174 if (zc->zc_nvlist_src_size != 0 && (error =
1175 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1176 zc->zc_iflags, &props))) {
1177 nvlist_free(config);
1178 return (error);
1179 }
1180
1181 if (props) {
1182 nvlist_t *nvl = NULL;
1183 uint64_t version = SPA_VERSION;
1184
1185 (void) nvlist_lookup_uint64(props,
1186 zpool_prop_to_name(ZPOOL_PROP_VERSION), &version);
1187 if (version < SPA_VERSION_INITIAL || version > SPA_VERSION) {
1188 error = EINVAL;
1189 goto pool_props_bad;
1190 }
1191 (void) nvlist_lookup_nvlist(props, ZPOOL_ROOTFS_PROPS, &nvl);
1192 if (nvl) {
1193 error = nvlist_dup(nvl, &rootprops, KM_SLEEP);
1194 if (error != 0) {
1195 nvlist_free(config);
1196 nvlist_free(props);
1197 return (error);
1198 }
1199 (void) nvlist_remove_all(props, ZPOOL_ROOTFS_PROPS);
1200 }
1201 VERIFY(nvlist_alloc(&zplprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
1202 error = zfs_fill_zplprops_root(version, rootprops,
1203 zplprops, NULL);
1204 if (error)
1205 goto pool_props_bad;
1206 }
1207
1208 buf = history_str_get(zc);
1209
1210 error = spa_create(zc->zc_name, config, props, buf, zplprops);
1211
1212 /*
1213 * Set the remaining root properties
1214 */
1215 if (!error && (error = zfs_set_prop_nvlist(zc->zc_name,
1216 ZPROP_SRC_LOCAL, rootprops, NULL)) != 0)
1217 (void) spa_destroy(zc->zc_name);
1218
1219 if (buf != NULL)
1220 history_str_free(buf);
1221
1222 pool_props_bad:
1223 nvlist_free(rootprops);
1224 nvlist_free(zplprops);
1225 nvlist_free(config);
1226 nvlist_free(props);
1227
1228 return (error);
1229 }
1230
1231 static int
1232 zfs_ioc_pool_destroy(zfs_cmd_t *zc)
1233 {
1234 int error;
1235 zfs_log_history(zc);
1236 error = spa_destroy(zc->zc_name);
1237 if (error == 0)
1238 zvol_remove_minors(zc->zc_name);
1239 return (error);
1240 }
1241
1242 static int
1243 zfs_ioc_pool_import(zfs_cmd_t *zc)
1244 {
1245 nvlist_t *config, *props = NULL;
1246 uint64_t guid;
1247 int error;
1248
1249 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1250 zc->zc_iflags, &config)) != 0)
1251 return (error);
1252
1253 if (zc->zc_nvlist_src_size != 0 && (error =
1254 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1255 zc->zc_iflags, &props))) {
1256 nvlist_free(config);
1257 return (error);
1258 }
1259
1260 if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, &guid) != 0 ||
1261 guid != zc->zc_guid)
1262 error = EINVAL;
1263 else
1264 error = spa_import(zc->zc_name, config, props, zc->zc_cookie);
1265
1266 if (zc->zc_nvlist_dst != 0) {
1267 int err;
1268
1269 if ((err = put_nvlist(zc, config)) != 0)
1270 error = err;
1271 }
1272
1273 if (error == 0)
1274 zvol_create_minors(zc->zc_name);
1275
1276 nvlist_free(config);
1277
1278 if (props)
1279 nvlist_free(props);
1280
1281 return (error);
1282 }
1283
1284 static int
1285 zfs_ioc_pool_export(zfs_cmd_t *zc)
1286 {
1287 int error;
1288 boolean_t force = (boolean_t)zc->zc_cookie;
1289 boolean_t hardforce = (boolean_t)zc->zc_guid;
1290
1291 zfs_log_history(zc);
1292 error = spa_export(zc->zc_name, NULL, force, hardforce);
1293 if (error == 0)
1294 zvol_remove_minors(zc->zc_name);
1295 return (error);
1296 }
1297
1298 static int
1299 zfs_ioc_pool_configs(zfs_cmd_t *zc)
1300 {
1301 nvlist_t *configs;
1302 int error;
1303
1304 if ((configs = spa_all_configs(&zc->zc_cookie)) == NULL)
1305 return (EEXIST);
1306
1307 error = put_nvlist(zc, configs);
1308
1309 nvlist_free(configs);
1310
1311 return (error);
1312 }
1313
1314 static int
1315 zfs_ioc_pool_stats(zfs_cmd_t *zc)
1316 {
1317 nvlist_t *config;
1318 int error;
1319 int ret = 0;
1320
1321 error = spa_get_stats(zc->zc_name, &config, zc->zc_value,
1322 sizeof (zc->zc_value));
1323
1324 if (config != NULL) {
1325 ret = put_nvlist(zc, config);
1326 nvlist_free(config);
1327
1328 /*
1329 * The config may be present even if 'error' is non-zero.
1330 * In this case we return success, and preserve the real errno
1331 * in 'zc_cookie'.
1332 */
1333 zc->zc_cookie = error;
1334 } else {
1335 ret = error;
1336 }
1337
1338 return (ret);
1339 }
1340
1341 /*
1342 * Try to import the given pool, returning pool stats as appropriate so that
1343 * user land knows which devices are available and overall pool health.
1344 */
1345 static int
1346 zfs_ioc_pool_tryimport(zfs_cmd_t *zc)
1347 {
1348 nvlist_t *tryconfig, *config;
1349 int error;
1350
1351 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1352 zc->zc_iflags, &tryconfig)) != 0)
1353 return (error);
1354
1355 config = spa_tryimport(tryconfig);
1356
1357 nvlist_free(tryconfig);
1358
1359 if (config == NULL)
1360 return (EINVAL);
1361
1362 error = put_nvlist(zc, config);
1363 nvlist_free(config);
1364
1365 return (error);
1366 }
1367
1368 /*
1369 * inputs:
1370 * zc_name name of the pool
1371 * zc_cookie scan func (pool_scan_func_t)
1372 */
1373 static int
1374 zfs_ioc_pool_scan(zfs_cmd_t *zc)
1375 {
1376 spa_t *spa;
1377 int error;
1378
1379 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1380 return (error);
1381
1382 if (zc->zc_cookie == POOL_SCAN_NONE)
1383 error = spa_scan_stop(spa);
1384 else
1385 error = spa_scan(spa, zc->zc_cookie);
1386
1387 spa_close(spa, FTAG);
1388
1389 return (error);
1390 }
1391
1392 static int
1393 zfs_ioc_pool_freeze(zfs_cmd_t *zc)
1394 {
1395 spa_t *spa;
1396 int error;
1397
1398 error = spa_open(zc->zc_name, &spa, FTAG);
1399 if (error == 0) {
1400 spa_freeze(spa);
1401 spa_close(spa, FTAG);
1402 }
1403 return (error);
1404 }
1405
1406 static int
1407 zfs_ioc_pool_upgrade(zfs_cmd_t *zc)
1408 {
1409 spa_t *spa;
1410 int error;
1411
1412 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1413 return (error);
1414
1415 if (zc->zc_cookie < spa_version(spa) || zc->zc_cookie > SPA_VERSION) {
1416 spa_close(spa, FTAG);
1417 return (EINVAL);
1418 }
1419
1420 spa_upgrade(spa, zc->zc_cookie);
1421 spa_close(spa, FTAG);
1422
1423 return (error);
1424 }
1425
1426 static int
1427 zfs_ioc_pool_get_history(zfs_cmd_t *zc)
1428 {
1429 spa_t *spa;
1430 char *hist_buf;
1431 uint64_t size;
1432 int error;
1433
1434 if ((size = zc->zc_history_len) == 0)
1435 return (EINVAL);
1436
1437 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1438 return (error);
1439
1440 if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) {
1441 spa_close(spa, FTAG);
1442 return (ENOTSUP);
1443 }
1444
1445 hist_buf = vmem_alloc(size, KM_SLEEP);
1446 if ((error = spa_history_get(spa, &zc->zc_history_offset,
1447 &zc->zc_history_len, hist_buf)) == 0) {
1448 error = ddi_copyout(hist_buf,
1449 (void *)(uintptr_t)zc->zc_history,
1450 zc->zc_history_len, zc->zc_iflags);
1451 }
1452
1453 spa_close(spa, FTAG);
1454 vmem_free(hist_buf, size);
1455 return (error);
1456 }
1457
1458 static int
1459 zfs_ioc_dsobj_to_dsname(zfs_cmd_t *zc)
1460 {
1461 int error;
1462
1463 if ((error = dsl_dsobj_to_dsname(zc->zc_name,zc->zc_obj,zc->zc_value)))
1464 return (error);
1465
1466 return (0);
1467 }
1468
1469 /*
1470 * inputs:
1471 * zc_name name of filesystem
1472 * zc_obj object to find
1473 *
1474 * outputs:
1475 * zc_value name of object
1476 */
1477 static int
1478 zfs_ioc_obj_to_path(zfs_cmd_t *zc)
1479 {
1480 objset_t *os;
1481 int error;
1482
1483 /* XXX reading from objset not owned */
1484 if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os)) != 0)
1485 return (error);
1486 if (dmu_objset_type(os) != DMU_OST_ZFS) {
1487 dmu_objset_rele(os, FTAG);
1488 return (EINVAL);
1489 }
1490 error = zfs_obj_to_path(os, zc->zc_obj, zc->zc_value,
1491 sizeof (zc->zc_value));
1492 dmu_objset_rele(os, FTAG);
1493
1494 return (error);
1495 }
1496
1497 /*
1498 * inputs:
1499 * zc_name name of filesystem
1500 * zc_obj object to find
1501 *
1502 * outputs:
1503 * zc_stat stats on object
1504 * zc_value path to object
1505 */
1506 static int
1507 zfs_ioc_obj_to_stats(zfs_cmd_t *zc)
1508 {
1509 objset_t *os;
1510 int error;
1511
1512 /* XXX reading from objset not owned */
1513 if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os)) != 0)
1514 return (error);
1515 if (dmu_objset_type(os) != DMU_OST_ZFS) {
1516 dmu_objset_rele(os, FTAG);
1517 return (EINVAL);
1518 }
1519 error = zfs_obj_to_stats(os, zc->zc_obj, &zc->zc_stat, zc->zc_value,
1520 sizeof (zc->zc_value));
1521 dmu_objset_rele(os, FTAG);
1522
1523 return (error);
1524 }
1525
1526 static int
1527 zfs_ioc_vdev_add(zfs_cmd_t *zc)
1528 {
1529 spa_t *spa;
1530 int error;
1531 nvlist_t *config, **l2cache, **spares;
1532 uint_t nl2cache = 0, nspares = 0;
1533
1534 error = spa_open(zc->zc_name, &spa, FTAG);
1535 if (error != 0)
1536 return (error);
1537
1538 error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1539 zc->zc_iflags, &config);
1540 (void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_L2CACHE,
1541 &l2cache, &nl2cache);
1542
1543 (void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_SPARES,
1544 &spares, &nspares);
1545
1546 /*
1547 * A root pool with concatenated devices is not supported.
1548 * Thus, can not add a device to a root pool.
1549 *
1550 * Intent log device can not be added to a rootpool because
1551 * during mountroot, zil is replayed, a seperated log device
1552 * can not be accessed during the mountroot time.
1553 *
1554 * l2cache and spare devices are ok to be added to a rootpool.
1555 */
1556 if (spa_bootfs(spa) != 0 && nl2cache == 0 && nspares == 0) {
1557 nvlist_free(config);
1558 spa_close(spa, FTAG);
1559 return (EDOM);
1560 }
1561
1562 if (error == 0) {
1563 error = spa_vdev_add(spa, config);
1564 nvlist_free(config);
1565 }
1566 spa_close(spa, FTAG);
1567 return (error);
1568 }
1569
1570 /*
1571 * inputs:
1572 * zc_name name of the pool
1573 * zc_nvlist_conf nvlist of devices to remove
1574 * zc_cookie to stop the remove?
1575 */
1576 static int
1577 zfs_ioc_vdev_remove(zfs_cmd_t *zc)
1578 {
1579 spa_t *spa;
1580 int error;
1581
1582 error = spa_open(zc->zc_name, &spa, FTAG);
1583 if (error != 0)
1584 return (error);
1585 error = spa_vdev_remove(spa, zc->zc_guid, B_FALSE);
1586 spa_close(spa, FTAG);
1587 return (error);
1588 }
1589
1590 static int
1591 zfs_ioc_vdev_set_state(zfs_cmd_t *zc)
1592 {
1593 spa_t *spa;
1594 int error;
1595 vdev_state_t newstate = VDEV_STATE_UNKNOWN;
1596
1597 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1598 return (error);
1599 switch (zc->zc_cookie) {
1600 case VDEV_STATE_ONLINE:
1601 error = vdev_online(spa, zc->zc_guid, zc->zc_obj, &newstate);
1602 break;
1603
1604 case VDEV_STATE_OFFLINE:
1605 error = vdev_offline(spa, zc->zc_guid, zc->zc_obj);
1606 break;
1607
1608 case VDEV_STATE_FAULTED:
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_fault(spa, zc->zc_guid, zc->zc_obj);
1614 break;
1615
1616 case VDEV_STATE_DEGRADED:
1617 if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED &&
1618 zc->zc_obj != VDEV_AUX_EXTERNAL)
1619 zc->zc_obj = VDEV_AUX_ERR_EXCEEDED;
1620
1621 error = vdev_degrade(spa, zc->zc_guid, zc->zc_obj);
1622 break;
1623
1624 default:
1625 error = EINVAL;
1626 }
1627 zc->zc_cookie = newstate;
1628 spa_close(spa, FTAG);
1629 return (error);
1630 }
1631
1632 static int
1633 zfs_ioc_vdev_attach(zfs_cmd_t *zc)
1634 {
1635 spa_t *spa;
1636 int replacing = zc->zc_cookie;
1637 nvlist_t *config;
1638 int error;
1639
1640 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1641 return (error);
1642
1643 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1644 zc->zc_iflags, &config)) == 0) {
1645 error = spa_vdev_attach(spa, zc->zc_guid, config, replacing);
1646 nvlist_free(config);
1647 }
1648
1649 spa_close(spa, FTAG);
1650 return (error);
1651 }
1652
1653 static int
1654 zfs_ioc_vdev_detach(zfs_cmd_t *zc)
1655 {
1656 spa_t *spa;
1657 int error;
1658
1659 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1660 return (error);
1661
1662 error = spa_vdev_detach(spa, zc->zc_guid, 0, B_FALSE);
1663
1664 spa_close(spa, FTAG);
1665 return (error);
1666 }
1667
1668 static int
1669 zfs_ioc_vdev_split(zfs_cmd_t *zc)
1670 {
1671 spa_t *spa;
1672 nvlist_t *config, *props = NULL;
1673 int error;
1674 boolean_t exp = !!(zc->zc_cookie & ZPOOL_EXPORT_AFTER_SPLIT);
1675
1676 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1677 return (error);
1678
1679 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1680 zc->zc_iflags, &config))) {
1681 spa_close(spa, FTAG);
1682 return (error);
1683 }
1684
1685 if (zc->zc_nvlist_src_size != 0 && (error =
1686 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1687 zc->zc_iflags, &props))) {
1688 spa_close(spa, FTAG);
1689 nvlist_free(config);
1690 return (error);
1691 }
1692
1693 error = spa_vdev_split_mirror(spa, zc->zc_string, config, props, exp);
1694
1695 spa_close(spa, FTAG);
1696
1697 nvlist_free(config);
1698 nvlist_free(props);
1699
1700 return (error);
1701 }
1702
1703 static int
1704 zfs_ioc_vdev_setpath(zfs_cmd_t *zc)
1705 {
1706 spa_t *spa;
1707 char *path = zc->zc_value;
1708 uint64_t guid = zc->zc_guid;
1709 int error;
1710
1711 error = spa_open(zc->zc_name, &spa, FTAG);
1712 if (error != 0)
1713 return (error);
1714
1715 error = spa_vdev_setpath(spa, guid, path);
1716 spa_close(spa, FTAG);
1717 return (error);
1718 }
1719
1720 static int
1721 zfs_ioc_vdev_setfru(zfs_cmd_t *zc)
1722 {
1723 spa_t *spa;
1724 char *fru = zc->zc_value;
1725 uint64_t guid = zc->zc_guid;
1726 int error;
1727
1728 error = spa_open(zc->zc_name, &spa, FTAG);
1729 if (error != 0)
1730 return (error);
1731
1732 error = spa_vdev_setfru(spa, guid, fru);
1733 spa_close(spa, FTAG);
1734 return (error);
1735 }
1736
1737 static int
1738 zfs_ioc_objset_stats_impl(zfs_cmd_t *zc, objset_t *os)
1739 {
1740 int error = 0;
1741 nvlist_t *nv;
1742
1743 dmu_objset_fast_stat(os, &zc->zc_objset_stats);
1744
1745 if (zc->zc_nvlist_dst != 0 &&
1746 (error = dsl_prop_get_all(os, &nv)) == 0) {
1747 dmu_objset_stats(os, nv);
1748 /*
1749 * NB: zvol_get_stats() will read the objset contents,
1750 * which we aren't supposed to do with a
1751 * DS_MODE_USER hold, because it could be
1752 * inconsistent. So this is a bit of a workaround...
1753 * XXX reading with out owning
1754 */
1755 if (!zc->zc_objset_stats.dds_inconsistent) {
1756 if (dmu_objset_type(os) == DMU_OST_ZVOL)
1757 error = zvol_get_stats(os, nv);
1758 }
1759 if (error == 0)
1760 error = put_nvlist(zc, nv);
1761 nvlist_free(nv);
1762 }
1763
1764 return (error);
1765 }
1766
1767 /*
1768 * inputs:
1769 * zc_name name of filesystem
1770 * zc_nvlist_dst_size size of buffer for property nvlist
1771 *
1772 * outputs:
1773 * zc_objset_stats stats
1774 * zc_nvlist_dst property nvlist
1775 * zc_nvlist_dst_size size of property nvlist
1776 */
1777 static int
1778 zfs_ioc_objset_stats(zfs_cmd_t *zc)
1779 {
1780 objset_t *os = NULL;
1781 int error;
1782
1783 if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os)))
1784 return (error);
1785
1786 error = zfs_ioc_objset_stats_impl(zc, os);
1787
1788 dmu_objset_rele(os, FTAG);
1789
1790 return (error);
1791 }
1792
1793 /*
1794 * inputs:
1795 * zc_name name of filesystem
1796 * zc_nvlist_dst_size size of buffer for property nvlist
1797 *
1798 * outputs:
1799 * zc_nvlist_dst received property nvlist
1800 * zc_nvlist_dst_size size of received property nvlist
1801 *
1802 * Gets received properties (distinct from local properties on or after
1803 * SPA_VERSION_RECVD_PROPS) for callers who want to differentiate received from
1804 * local property values.
1805 */
1806 static int
1807 zfs_ioc_objset_recvd_props(zfs_cmd_t *zc)
1808 {
1809 objset_t *os = NULL;
1810 int error;
1811 nvlist_t *nv;
1812
1813 if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os)))
1814 return (error);
1815
1816 /*
1817 * Without this check, we would return local property values if the
1818 * caller has not already received properties on or after
1819 * SPA_VERSION_RECVD_PROPS.
1820 */
1821 if (!dsl_prop_get_hasrecvd(os)) {
1822 dmu_objset_rele(os, FTAG);
1823 return (ENOTSUP);
1824 }
1825
1826 if (zc->zc_nvlist_dst != 0 &&
1827 (error = dsl_prop_get_received(os, &nv)) == 0) {
1828 error = put_nvlist(zc, nv);
1829 nvlist_free(nv);
1830 }
1831
1832 dmu_objset_rele(os, FTAG);
1833 return (error);
1834 }
1835
1836 static int
1837 nvl_add_zplprop(objset_t *os, nvlist_t *props, zfs_prop_t prop)
1838 {
1839 uint64_t value;
1840 int error;
1841
1842 /*
1843 * zfs_get_zplprop() will either find a value or give us
1844 * the default value (if there is one).
1845 */
1846 if ((error = zfs_get_zplprop(os, prop, &value)) != 0)
1847 return (error);
1848 VERIFY(nvlist_add_uint64(props, zfs_prop_to_name(prop), value) == 0);
1849 return (0);
1850 }
1851
1852 /*
1853 * inputs:
1854 * zc_name name of filesystem
1855 * zc_nvlist_dst_size size of buffer for zpl property nvlist
1856 *
1857 * outputs:
1858 * zc_nvlist_dst zpl property nvlist
1859 * zc_nvlist_dst_size size of zpl property nvlist
1860 */
1861 static int
1862 zfs_ioc_objset_zplprops(zfs_cmd_t *zc)
1863 {
1864 objset_t *os;
1865 int err;
1866
1867 /* XXX reading without owning */
1868 if ((err = dmu_objset_hold(zc->zc_name, FTAG, &os)))
1869 return (err);
1870
1871 dmu_objset_fast_stat(os, &zc->zc_objset_stats);
1872
1873 /*
1874 * NB: nvl_add_zplprop() will read the objset contents,
1875 * which we aren't supposed to do with a DS_MODE_USER
1876 * hold, because it could be inconsistent.
1877 */
1878 if (zc->zc_nvlist_dst != 0 &&
1879 !zc->zc_objset_stats.dds_inconsistent &&
1880 dmu_objset_type(os) == DMU_OST_ZFS) {
1881 nvlist_t *nv;
1882
1883 VERIFY(nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) == 0);
1884 if ((err = nvl_add_zplprop(os, nv, ZFS_PROP_VERSION)) == 0 &&
1885 (err = nvl_add_zplprop(os, nv, ZFS_PROP_NORMALIZE)) == 0 &&
1886 (err = nvl_add_zplprop(os, nv, ZFS_PROP_UTF8ONLY)) == 0 &&
1887 (err = nvl_add_zplprop(os, nv, ZFS_PROP_CASE)) == 0)
1888 err = put_nvlist(zc, nv);
1889 nvlist_free(nv);
1890 } else {
1891 err = ENOENT;
1892 }
1893 dmu_objset_rele(os, FTAG);
1894 return (err);
1895 }
1896
1897 static boolean_t
1898 dataset_name_hidden(const char *name)
1899 {
1900 /*
1901 * Skip over datasets that are not visible in this zone,
1902 * internal datasets (which have a $ in their name), and
1903 * temporary datasets (which have a % in their name).
1904 */
1905 if (strchr(name, '$') != NULL)
1906 return (B_TRUE);
1907 if (strchr(name, '%') != NULL)
1908 return (B_TRUE);
1909 if (!INGLOBALZONE(curproc) && !zone_dataset_visible(name, NULL))
1910 return (B_TRUE);
1911 return (B_FALSE);
1912 }
1913
1914 /*
1915 * inputs:
1916 * zc_name name of filesystem
1917 * zc_cookie zap cursor
1918 * zc_nvlist_dst_size size of buffer for property nvlist
1919 *
1920 * outputs:
1921 * zc_name name of next filesystem
1922 * zc_cookie zap cursor
1923 * zc_objset_stats stats
1924 * zc_nvlist_dst property nvlist
1925 * zc_nvlist_dst_size size of property nvlist
1926 */
1927 static int
1928 zfs_ioc_dataset_list_next(zfs_cmd_t *zc)
1929 {
1930 objset_t *os;
1931 int error;
1932 char *p;
1933 size_t orig_len = strlen(zc->zc_name);
1934
1935 top:
1936 if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os))) {
1937 if (error == ENOENT)
1938 error = ESRCH;
1939 return (error);
1940 }
1941
1942 p = strrchr(zc->zc_name, '/');
1943 if (p == NULL || p[1] != '\0')
1944 (void) strlcat(zc->zc_name, "/", sizeof (zc->zc_name));
1945 p = zc->zc_name + strlen(zc->zc_name);
1946
1947 /*
1948 * Pre-fetch the datasets. dmu_objset_prefetch() always returns 0
1949 * but is not declared void because its called by dmu_objset_find().
1950 */
1951 if (zc->zc_cookie == 0) {
1952 uint64_t cookie = 0;
1953 int len = sizeof (zc->zc_name) - (p - zc->zc_name);
1954
1955 while (dmu_dir_list_next(os, len, p, NULL, &cookie) == 0)
1956 (void) dmu_objset_prefetch(p, NULL);
1957 }
1958
1959 do {
1960 error = dmu_dir_list_next(os,
1961 sizeof (zc->zc_name) - (p - zc->zc_name), p,
1962 NULL, &zc->zc_cookie);
1963 if (error == ENOENT)
1964 error = ESRCH;
1965 } while (error == 0 && dataset_name_hidden(zc->zc_name) &&
1966 !(zc->zc_iflags & FKIOCTL));
1967 dmu_objset_rele(os, FTAG);
1968
1969 /*
1970 * If it's an internal dataset (ie. with a '$' in its name),
1971 * don't try to get stats for it, otherwise we'll return ENOENT.
1972 */
1973 if (error == 0 && strchr(zc->zc_name, '$') == NULL) {
1974 error = zfs_ioc_objset_stats(zc); /* fill in the stats */
1975 if (error == ENOENT) {
1976 /* We lost a race with destroy, get the next one. */
1977 zc->zc_name[orig_len] = '\0';
1978 goto top;
1979 }
1980 }
1981 return (error);
1982 }
1983
1984 /*
1985 * inputs:
1986 * zc_name name of filesystem
1987 * zc_cookie zap cursor
1988 * zc_nvlist_dst_size size of buffer for property nvlist
1989 *
1990 * outputs:
1991 * zc_name name of next snapshot
1992 * zc_objset_stats stats
1993 * zc_nvlist_dst property nvlist
1994 * zc_nvlist_dst_size size of property nvlist
1995 */
1996 static int
1997 zfs_ioc_snapshot_list_next(zfs_cmd_t *zc)
1998 {
1999 objset_t *os;
2000 int error;
2001
2002 top:
2003 if (zc->zc_cookie == 0)
2004 (void) dmu_objset_find(zc->zc_name, dmu_objset_prefetch,
2005 NULL, DS_FIND_SNAPSHOTS);
2006
2007 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
2008 if (error)
2009 return (error == ENOENT ? ESRCH : error);
2010
2011 /*
2012 * A dataset name of maximum length cannot have any snapshots,
2013 * so exit immediately.
2014 */
2015 if (strlcat(zc->zc_name, "@", sizeof (zc->zc_name)) >= MAXNAMELEN) {
2016 dmu_objset_rele(os, FTAG);
2017 return (ESRCH);
2018 }
2019
2020 error = dmu_snapshot_list_next(os,
2021 sizeof (zc->zc_name) - strlen(zc->zc_name),
2022 zc->zc_name + strlen(zc->zc_name), &zc->zc_obj, &zc->zc_cookie,
2023 NULL);
2024
2025 if (error == 0) {
2026 dsl_dataset_t *ds;
2027 dsl_pool_t *dp = os->os_dsl_dataset->ds_dir->dd_pool;
2028
2029 /*
2030 * Since we probably don't have a hold on this snapshot,
2031 * it's possible that the objsetid could have been destroyed
2032 * and reused for a new objset. It's OK if this happens during
2033 * a zfs send operation, since the new createtxg will be
2034 * beyond the range we're interested in.
2035 */
2036 rw_enter(&dp->dp_config_rwlock, RW_READER);
2037 error = dsl_dataset_hold_obj(dp, zc->zc_obj, FTAG, &ds);
2038 rw_exit(&dp->dp_config_rwlock);
2039 if (error) {
2040 if (error == ENOENT) {
2041 /* Racing with destroy, get the next one. */
2042 *strchr(zc->zc_name, '@') = '\0';
2043 dmu_objset_rele(os, FTAG);
2044 goto top;
2045 }
2046 } else {
2047 objset_t *ossnap;
2048
2049 error = dmu_objset_from_ds(ds, &ossnap);
2050 if (error == 0)
2051 error = zfs_ioc_objset_stats_impl(zc, ossnap);
2052 dsl_dataset_rele(ds, FTAG);
2053 }
2054 } else if (error == ENOENT) {
2055 error = ESRCH;
2056 }
2057
2058 dmu_objset_rele(os, FTAG);
2059 /* if we failed, undo the @ that we tacked on to zc_name */
2060 if (error)
2061 *strchr(zc->zc_name, '@') = '\0';
2062 return (error);
2063 }
2064
2065 static int
2066 zfs_prop_set_userquota(const char *dsname, nvpair_t *pair)
2067 {
2068 const char *propname = nvpair_name(pair);
2069 uint64_t *valary;
2070 unsigned int vallen;
2071 const char *domain;
2072 char *dash;
2073 zfs_userquota_prop_t type;
2074 uint64_t rid;
2075 uint64_t quota;
2076 zfs_sb_t *zsb;
2077 int err;
2078
2079 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2080 nvlist_t *attrs;
2081 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
2082 if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2083 &pair) != 0)
2084 return (EINVAL);
2085 }
2086
2087 /*
2088 * A correctly constructed propname is encoded as
2089 * userquota@<rid>-<domain>.
2090 */
2091 if ((dash = strchr(propname, '-')) == NULL ||
2092 nvpair_value_uint64_array(pair, &valary, &vallen) != 0 ||
2093 vallen != 3)
2094 return (EINVAL);
2095
2096 domain = dash + 1;
2097 type = valary[0];
2098 rid = valary[1];
2099 quota = valary[2];
2100
2101 err = zfs_sb_hold(dsname, FTAG, &zsb, B_FALSE);
2102 if (err == 0) {
2103 err = zfs_set_userquota(zsb, type, domain, rid, quota);
2104 zfs_sb_rele(zsb, FTAG);
2105 }
2106
2107 return (err);
2108 }
2109
2110 /*
2111 * If the named property is one that has a special function to set its value,
2112 * return 0 on success and a positive error code on failure; otherwise if it is
2113 * not one of the special properties handled by this function, return -1.
2114 *
2115 * XXX: It would be better for callers of the property interface if we handled
2116 * these special cases in dsl_prop.c (in the dsl layer).
2117 */
2118 static int
2119 zfs_prop_set_special(const char *dsname, zprop_source_t source,
2120 nvpair_t *pair)
2121 {
2122 const char *propname = nvpair_name(pair);
2123 zfs_prop_t prop = zfs_name_to_prop(propname);
2124 uint64_t intval;
2125 int err;
2126
2127 if (prop == ZPROP_INVAL) {
2128 if (zfs_prop_userquota(propname))
2129 return (zfs_prop_set_userquota(dsname, pair));
2130 return (-1);
2131 }
2132
2133 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2134 nvlist_t *attrs;
2135 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
2136 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2137 &pair) == 0);
2138 }
2139
2140 if (zfs_prop_get_type(prop) == PROP_TYPE_STRING)
2141 return (-1);
2142
2143 VERIFY(0 == nvpair_value_uint64(pair, &intval));
2144
2145 switch (prop) {
2146 case ZFS_PROP_QUOTA:
2147 err = dsl_dir_set_quota(dsname, source, intval);
2148 break;
2149 case ZFS_PROP_REFQUOTA:
2150 err = dsl_dataset_set_quota(dsname, source, intval);
2151 break;
2152 case ZFS_PROP_RESERVATION:
2153 err = dsl_dir_set_reservation(dsname, source, intval);
2154 break;
2155 case ZFS_PROP_REFRESERVATION:
2156 err = dsl_dataset_set_reservation(dsname, source, intval);
2157 break;
2158 case ZFS_PROP_VOLSIZE:
2159 err = zvol_set_volsize(dsname, intval);
2160 break;
2161 case ZFS_PROP_VERSION:
2162 {
2163 zfs_sb_t *zsb;
2164
2165 if ((err = zfs_sb_hold(dsname, FTAG, &zsb, B_TRUE)) != 0)
2166 break;
2167
2168 err = zfs_set_version(zsb, intval);
2169 zfs_sb_rele(zsb, FTAG);
2170
2171 if (err == 0 && intval >= ZPL_VERSION_USERSPACE) {
2172 zfs_cmd_t *zc;
2173
2174 zc = kmem_zalloc(sizeof (zfs_cmd_t), KM_SLEEP);
2175 (void) strcpy(zc->zc_name, dsname);
2176 (void) zfs_ioc_userspace_upgrade(zc);
2177 kmem_free(zc, sizeof (zfs_cmd_t));
2178 }
2179 break;
2180 }
2181
2182 default:
2183 err = -1;
2184 }
2185
2186 return (err);
2187 }
2188
2189 /*
2190 * This function is best effort. If it fails to set any of the given properties,
2191 * it continues to set as many as it can and returns the first error
2192 * encountered. If the caller provides a non-NULL errlist, it also gives the
2193 * complete list of names of all the properties it failed to set along with the
2194 * corresponding error numbers. The caller is responsible for freeing the
2195 * returned errlist.
2196 *
2197 * If every property is set successfully, zero is returned and the list pointed
2198 * at by errlist is NULL.
2199 */
2200 int
2201 zfs_set_prop_nvlist(const char *dsname, zprop_source_t source, nvlist_t *nvl,
2202 nvlist_t **errlist)
2203 {
2204 nvpair_t *pair;
2205 nvpair_t *propval;
2206 int rv = 0;
2207 uint64_t intval;
2208 char *strval;
2209 nvlist_t *genericnvl;
2210 nvlist_t *errors;
2211 nvlist_t *retrynvl;
2212
2213 VERIFY(nvlist_alloc(&genericnvl, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2214 VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2215 VERIFY(nvlist_alloc(&retrynvl, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2216
2217 retry:
2218 pair = NULL;
2219 while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) {
2220 const char *propname = nvpair_name(pair);
2221 zfs_prop_t prop = zfs_name_to_prop(propname);
2222 int err = 0;
2223
2224 /* decode the property value */
2225 propval = pair;
2226 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2227 nvlist_t *attrs;
2228 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
2229 if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2230 &propval) != 0)
2231 err = EINVAL;
2232 }
2233
2234 /* Validate value type */
2235 if (err == 0 && prop == ZPROP_INVAL) {
2236 if (zfs_prop_user(propname)) {
2237 if (nvpair_type(propval) != DATA_TYPE_STRING)
2238 err = EINVAL;
2239 } else if (zfs_prop_userquota(propname)) {
2240 if (nvpair_type(propval) !=
2241 DATA_TYPE_UINT64_ARRAY)
2242 err = EINVAL;
2243 }
2244 } else if (err == 0) {
2245 if (nvpair_type(propval) == DATA_TYPE_STRING) {
2246 if (zfs_prop_get_type(prop) != PROP_TYPE_STRING)
2247 err = EINVAL;
2248 } else if (nvpair_type(propval) == DATA_TYPE_UINT64) {
2249 const char *unused;
2250
2251 VERIFY(nvpair_value_uint64(propval,
2252 &intval) == 0);
2253
2254 switch (zfs_prop_get_type(prop)) {
2255 case PROP_TYPE_NUMBER:
2256 break;
2257 case PROP_TYPE_STRING:
2258 err = EINVAL;
2259 break;
2260 case PROP_TYPE_INDEX:
2261 if (zfs_prop_index_to_string(prop,
2262 intval, &unused) != 0)
2263 err = EINVAL;
2264 break;
2265 default:
2266 cmn_err(CE_PANIC,
2267 "unknown property type");
2268 }
2269 } else {
2270 err = EINVAL;
2271 }
2272 }
2273
2274 /* Validate permissions */
2275 if (err == 0)
2276 err = zfs_check_settable(dsname, pair, CRED());
2277
2278 if (err == 0) {
2279 err = zfs_prop_set_special(dsname, source, pair);
2280 if (err == -1) {
2281 /*
2282 * For better performance we build up a list of
2283 * properties to set in a single transaction.
2284 */
2285 err = nvlist_add_nvpair(genericnvl, pair);
2286 } else if (err != 0 && nvl != retrynvl) {
2287 /*
2288 * This may be a spurious error caused by
2289 * receiving quota and reservation out of order.
2290 * Try again in a second pass.
2291 */
2292 err = nvlist_add_nvpair(retrynvl, pair);
2293 }
2294 }
2295
2296 if (err != 0)
2297 VERIFY(nvlist_add_int32(errors, propname, err) == 0);
2298 }
2299
2300 if (nvl != retrynvl && !nvlist_empty(retrynvl)) {
2301 nvl = retrynvl;
2302 goto retry;
2303 }
2304
2305 if (!nvlist_empty(genericnvl) &&
2306 dsl_props_set(dsname, source, genericnvl) != 0) {
2307 /*
2308 * If this fails, we still want to set as many properties as we
2309 * can, so try setting them individually.
2310 */
2311 pair = NULL;
2312 while ((pair = nvlist_next_nvpair(genericnvl, pair)) != NULL) {
2313 const char *propname = nvpair_name(pair);
2314 int err = 0;
2315
2316 propval = pair;
2317 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2318 nvlist_t *attrs;
2319 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
2320 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2321 &propval) == 0);
2322 }
2323
2324 if (nvpair_type(propval) == DATA_TYPE_STRING) {
2325 VERIFY(nvpair_value_string(propval,
2326 &strval) == 0);
2327 err = dsl_prop_set(dsname, propname, source, 1,
2328 strlen(strval) + 1, strval);
2329 } else {
2330 VERIFY(nvpair_value_uint64(propval,
2331 &intval) == 0);
2332 err = dsl_prop_set(dsname, propname, source, 8,
2333 1, &intval);
2334 }
2335
2336 if (err != 0) {
2337 VERIFY(nvlist_add_int32(errors, propname,
2338 err) == 0);
2339 }
2340 }
2341 }
2342 nvlist_free(genericnvl);
2343 nvlist_free(retrynvl);
2344
2345 if ((pair = nvlist_next_nvpair(errors, NULL)) == NULL) {
2346 nvlist_free(errors);
2347 errors = NULL;
2348 } else {
2349 VERIFY(nvpair_value_int32(pair, &rv) == 0);
2350 }
2351
2352 if (errlist == NULL)
2353 nvlist_free(errors);
2354 else
2355 *errlist = errors;
2356
2357 return (rv);
2358 }
2359
2360 /*
2361 * Check that all the properties are valid user properties.
2362 */
2363 static int
2364 zfs_check_userprops(char *fsname, nvlist_t *nvl)
2365 {
2366 nvpair_t *pair = NULL;
2367 int error = 0;
2368
2369 while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) {
2370 const char *propname = nvpair_name(pair);
2371 char *valstr;
2372
2373 if (!zfs_prop_user(propname) ||
2374 nvpair_type(pair) != DATA_TYPE_STRING)
2375 return (EINVAL);
2376
2377 if ((error = zfs_secpolicy_write_perms(fsname,
2378 ZFS_DELEG_PERM_USERPROP, CRED())))
2379 return (error);
2380
2381 if (strlen(propname) >= ZAP_MAXNAMELEN)
2382 return (ENAMETOOLONG);
2383
2384 VERIFY(nvpair_value_string(pair, &valstr) == 0);
2385 if (strlen(valstr) >= ZAP_MAXVALUELEN)
2386 return (E2BIG);
2387 }
2388 return (0);
2389 }
2390
2391 static void
2392 props_skip(nvlist_t *props, nvlist_t *skipped, nvlist_t **newprops)
2393 {
2394 nvpair_t *pair;
2395
2396 VERIFY(nvlist_alloc(newprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2397
2398 pair = NULL;
2399 while ((pair = nvlist_next_nvpair(props, pair)) != NULL) {
2400 if (nvlist_exists(skipped, nvpair_name(pair)))
2401 continue;
2402
2403 VERIFY(nvlist_add_nvpair(*newprops, pair) == 0);
2404 }
2405 }
2406
2407 static int
2408 clear_received_props(objset_t *os, const char *fs, nvlist_t *props,
2409 nvlist_t *skipped)
2410 {
2411 int err = 0;
2412 nvlist_t *cleared_props = NULL;
2413 props_skip(props, skipped, &cleared_props);
2414 if (!nvlist_empty(cleared_props)) {
2415 /*
2416 * Acts on local properties until the dataset has received
2417 * properties at least once on or after SPA_VERSION_RECVD_PROPS.
2418 */
2419 zprop_source_t flags = (ZPROP_SRC_NONE |
2420 (dsl_prop_get_hasrecvd(os) ? ZPROP_SRC_RECEIVED : 0));
2421 err = zfs_set_prop_nvlist(fs, flags, cleared_props, NULL);
2422 }
2423 nvlist_free(cleared_props);
2424 return (err);
2425 }
2426
2427 /*
2428 * inputs:
2429 * zc_name name of filesystem
2430 * zc_value name of property to set
2431 * zc_nvlist_src{_size} nvlist of properties to apply
2432 * zc_cookie received properties flag
2433 *
2434 * outputs:
2435 * zc_nvlist_dst{_size} error for each unapplied received property
2436 */
2437 static int
2438 zfs_ioc_set_prop(zfs_cmd_t *zc)
2439 {
2440 nvlist_t *nvl;
2441 boolean_t received = zc->zc_cookie;
2442 zprop_source_t source = (received ? ZPROP_SRC_RECEIVED :
2443 ZPROP_SRC_LOCAL);
2444 nvlist_t *errors = NULL;
2445 int error;
2446
2447 if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2448 zc->zc_iflags, &nvl)) != 0)
2449 return (error);
2450
2451 if (received) {
2452 nvlist_t *origprops;
2453 objset_t *os;
2454
2455 if (dmu_objset_hold(zc->zc_name, FTAG, &os) == 0) {
2456 if (dsl_prop_get_received(os, &origprops) == 0) {
2457 (void) clear_received_props(os,
2458 zc->zc_name, origprops, nvl);
2459 nvlist_free(origprops);
2460 }
2461
2462 dsl_prop_set_hasrecvd(os);
2463 dmu_objset_rele(os, FTAG);
2464 }
2465 }
2466
2467 error = zfs_set_prop_nvlist(zc->zc_name, source, nvl, &errors);
2468
2469 if (zc->zc_nvlist_dst != 0 && errors != NULL) {
2470 (void) put_nvlist(zc, errors);
2471 }
2472
2473 nvlist_free(errors);
2474 nvlist_free(nvl);
2475 return (error);
2476 }
2477
2478 /*
2479 * inputs:
2480 * zc_name name of filesystem
2481 * zc_value name of property to inherit
2482 * zc_cookie revert to received value if TRUE
2483 *
2484 * outputs: none
2485 */
2486 static int
2487 zfs_ioc_inherit_prop(zfs_cmd_t *zc)
2488 {
2489 const char *propname = zc->zc_value;
2490 zfs_prop_t prop = zfs_name_to_prop(propname);
2491 boolean_t received = zc->zc_cookie;
2492 zprop_source_t source = (received
2493 ? ZPROP_SRC_NONE /* revert to received value, if any */
2494 : ZPROP_SRC_INHERITED); /* explicitly inherit */
2495
2496 if (received) {
2497 nvlist_t *dummy;
2498 nvpair_t *pair;
2499 zprop_type_t type;
2500 int err;
2501
2502 /*
2503 * zfs_prop_set_special() expects properties in the form of an
2504 * nvpair with type info.
2505 */
2506 if (prop == ZPROP_INVAL) {
2507 if (!zfs_prop_user(propname))
2508 return (EINVAL);
2509
2510 type = PROP_TYPE_STRING;
2511 } else if (prop == ZFS_PROP_VOLSIZE ||
2512 prop == ZFS_PROP_VERSION) {
2513 return (EINVAL);
2514 } else {
2515 type = zfs_prop_get_type(prop);
2516 }
2517
2518 VERIFY(nvlist_alloc(&dummy, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2519
2520 switch (type) {
2521 case PROP_TYPE_STRING:
2522 VERIFY(0 == nvlist_add_string(dummy, propname, ""));
2523 break;
2524 case PROP_TYPE_NUMBER:
2525 case PROP_TYPE_INDEX:
2526 VERIFY(0 == nvlist_add_uint64(dummy, propname, 0));
2527 break;
2528 default:
2529 nvlist_free(dummy);
2530 return (EINVAL);
2531 }
2532
2533 pair = nvlist_next_nvpair(dummy, NULL);
2534 err = zfs_prop_set_special(zc->zc_name, source, pair);
2535 nvlist_free(dummy);
2536 if (err != -1)
2537 return (err); /* special property already handled */
2538 } else {
2539 /*
2540 * Only check this in the non-received case. We want to allow
2541 * 'inherit -S' to revert non-inheritable properties like quota
2542 * and reservation to the received or default values even though
2543 * they are not considered inheritable.
2544 */
2545 if (prop != ZPROP_INVAL && !zfs_prop_inheritable(prop))
2546 return (EINVAL);
2547 }
2548
2549 /* the property name has been validated by zfs_secpolicy_inherit() */
2550 return (dsl_prop_set(zc->zc_name, zc->zc_value, source, 0, 0, NULL));
2551 }
2552
2553 static int
2554 zfs_ioc_pool_set_props(zfs_cmd_t *zc)
2555 {
2556 nvlist_t *props;
2557 spa_t *spa;
2558 int error;
2559 nvpair_t *pair;
2560
2561 if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2562 zc->zc_iflags, &props)))
2563 return (error);
2564
2565 /*
2566 * If the only property is the configfile, then just do a spa_lookup()
2567 * to handle the faulted case.
2568 */
2569 pair = nvlist_next_nvpair(props, NULL);
2570 if (pair != NULL && strcmp(nvpair_name(pair),
2571 zpool_prop_to_name(ZPOOL_PROP_CACHEFILE)) == 0 &&
2572 nvlist_next_nvpair(props, pair) == NULL) {
2573 mutex_enter(&spa_namespace_lock);
2574 if ((spa = spa_lookup(zc->zc_name)) != NULL) {
2575 spa_configfile_set(spa, props, B_FALSE);
2576 spa_config_sync(spa, B_FALSE, B_TRUE);
2577 }
2578 mutex_exit(&spa_namespace_lock);
2579 if (spa != NULL) {
2580 nvlist_free(props);
2581 return (0);
2582 }
2583 }
2584
2585 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) {
2586 nvlist_free(props);
2587 return (error);
2588 }
2589
2590 error = spa_prop_set(spa, props);
2591
2592 nvlist_free(props);
2593 spa_close(spa, FTAG);
2594
2595 return (error);
2596 }
2597
2598 static int
2599 zfs_ioc_pool_get_props(zfs_cmd_t *zc)
2600 {
2601 spa_t *spa;
2602 int error;
2603 nvlist_t *nvp = NULL;
2604
2605 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) {
2606 /*
2607 * If the pool is faulted, there may be properties we can still
2608 * get (such as altroot and cachefile), so attempt to get them
2609 * anyway.
2610 */
2611 mutex_enter(&spa_namespace_lock);
2612 if ((spa = spa_lookup(zc->zc_name)) != NULL)
2613 error = spa_prop_get(spa, &nvp);
2614 mutex_exit(&spa_namespace_lock);
2615 } else {
2616 error = spa_prop_get(spa, &nvp);
2617 spa_close(spa, FTAG);
2618 }
2619
2620 if (error == 0 && zc->zc_nvlist_dst != 0)
2621 error = put_nvlist(zc, nvp);
2622 else
2623 error = EFAULT;
2624
2625 nvlist_free(nvp);
2626 return (error);
2627 }
2628
2629 /*
2630 * inputs:
2631 * zc_name name of volume
2632 *
2633 * outputs: none
2634 */
2635 static int
2636 zfs_ioc_create_minor(zfs_cmd_t *zc)
2637 {
2638 return (zvol_create_minor(zc->zc_name));
2639 }
2640
2641 /*
2642 * inputs:
2643 * zc_name name of volume
2644 *
2645 * outputs: none
2646 */
2647 static int
2648 zfs_ioc_remove_minor(zfs_cmd_t *zc)
2649 {
2650 return (zvol_remove_minor(zc->zc_name));
2651 }
2652
2653 /*
2654 * inputs:
2655 * zc_name name of filesystem
2656 * zc_nvlist_src{_size} nvlist of delegated permissions
2657 * zc_perm_action allow/unallow flag
2658 *
2659 * outputs: none
2660 */
2661 static int
2662 zfs_ioc_set_fsacl(zfs_cmd_t *zc)
2663 {
2664 int error;
2665 nvlist_t *fsaclnv = NULL;
2666
2667 if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2668 zc->zc_iflags, &fsaclnv)) != 0)
2669 return (error);
2670
2671 /*
2672 * Verify nvlist is constructed correctly
2673 */
2674 if ((error = zfs_deleg_verify_nvlist(fsaclnv)) != 0) {
2675 nvlist_free(fsaclnv);
2676 return (EINVAL);
2677 }
2678
2679 /*
2680 * If we don't have PRIV_SYS_MOUNT, then validate
2681 * that user is allowed to hand out each permission in
2682 * the nvlist(s)
2683 */
2684
2685 error = secpolicy_zfs(CRED());
2686 if (error) {
2687 if (zc->zc_perm_action == B_FALSE) {
2688 error = dsl_deleg_can_allow(zc->zc_name,
2689 fsaclnv, CRED());
2690 } else {
2691 error = dsl_deleg_can_unallow(zc->zc_name,
2692 fsaclnv, CRED());
2693 }
2694 }
2695
2696 if (error == 0)
2697 error = dsl_deleg_set(zc->zc_name, fsaclnv, zc->zc_perm_action);
2698
2699 nvlist_free(fsaclnv);
2700 return (error);
2701 }
2702
2703 /*
2704 * inputs:
2705 * zc_name name of filesystem
2706 *
2707 * outputs:
2708 * zc_nvlist_src{_size} nvlist of delegated permissions
2709 */
2710 static int
2711 zfs_ioc_get_fsacl(zfs_cmd_t *zc)
2712 {
2713 nvlist_t *nvp;
2714 int error;
2715
2716 if ((error = dsl_deleg_get(zc->zc_name, &nvp)) == 0) {
2717 error = put_nvlist(zc, nvp);
2718 nvlist_free(nvp);
2719 }
2720
2721 return (error);
2722 }
2723
2724 #ifdef HAVE_SNAPSHOT
2725 /*
2726 * Search the vfs list for a specified resource. Returns a pointer to it
2727 * or NULL if no suitable entry is found. The caller of this routine
2728 * is responsible for releasing the returned vfs pointer.
2729 */
2730 static vfs_t *
2731 zfs_get_vfs(const char *resource)
2732 {
2733 struct vfs *vfsp;
2734 struct vfs *vfs_found = NULL;
2735
2736 vfs_list_read_lock();
2737 vfsp = rootvfs;
2738 do {
2739 if (strcmp(refstr_value(vfsp->vfs_resource), resource) == 0) {
2740 mntget(vfsp);
2741 vfs_found = vfsp;
2742 break;
2743 }
2744 vfsp = vfsp->vfs_next;
2745 } while (vfsp != rootvfs);
2746 vfs_list_unlock();
2747 return (vfs_found);
2748 }
2749 #endif /* HAVE_SNAPSHOT */
2750
2751 /* ARGSUSED */
2752 static void
2753 zfs_create_cb(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx)
2754 {
2755 zfs_creat_t *zct = arg;
2756
2757 zfs_create_fs(os, cr, zct->zct_zplprops, tx);
2758 }
2759
2760 #define ZFS_PROP_UNDEFINED ((uint64_t)-1)
2761
2762 /*
2763 * inputs:
2764 * createprops list of properties requested by creator
2765 * default_zplver zpl version to use if unspecified in createprops
2766 * fuids_ok fuids allowed in this version of the spa?
2767 * os parent objset pointer (NULL if root fs)
2768 *
2769 * outputs:
2770 * zplprops values for the zplprops we attach to the master node object
2771 * is_ci true if requested file system will be purely case-insensitive
2772 *
2773 * Determine the settings for utf8only, normalization and
2774 * casesensitivity. Specific values may have been requested by the
2775 * creator and/or we can inherit values from the parent dataset. If
2776 * the file system is of too early a vintage, a creator can not
2777 * request settings for these properties, even if the requested
2778 * setting is the default value. We don't actually want to create dsl
2779 * properties for these, so remove them from the source nvlist after
2780 * processing.
2781 */
2782 static int
2783 zfs_fill_zplprops_impl(objset_t *os, uint64_t zplver,
2784 boolean_t fuids_ok, boolean_t sa_ok, nvlist_t *createprops,
2785 nvlist_t *zplprops, boolean_t *is_ci)
2786 {
2787 uint64_t sense = ZFS_PROP_UNDEFINED;
2788 uint64_t norm = ZFS_PROP_UNDEFINED;
2789 uint64_t u8 = ZFS_PROP_UNDEFINED;
2790
2791 ASSERT(zplprops != NULL);
2792
2793 /*
2794 * Pull out creator prop choices, if any.
2795 */
2796 if (createprops) {
2797 (void) nvlist_lookup_uint64(createprops,
2798 zfs_prop_to_name(ZFS_PROP_VERSION), &zplver);
2799 (void) nvlist_lookup_uint64(createprops,
2800 zfs_prop_to_name(ZFS_PROP_NORMALIZE), &norm);
2801 (void) nvlist_remove_all(createprops,
2802 zfs_prop_to_name(ZFS_PROP_NORMALIZE));
2803 (void) nvlist_lookup_uint64(createprops,
2804 zfs_prop_to_name(ZFS_PROP_UTF8ONLY), &u8);
2805 (void) nvlist_remove_all(createprops,
2806 zfs_prop_to_name(ZFS_PROP_UTF8ONLY));
2807 (void) nvlist_lookup_uint64(createprops,
2808 zfs_prop_to_name(ZFS_PROP_CASE), &sense);
2809 (void) nvlist_remove_all(createprops,
2810 zfs_prop_to_name(ZFS_PROP_CASE));
2811 }
2812
2813 /*
2814 * If the zpl version requested is whacky or the file system
2815 * or pool is version is too "young" to support normalization
2816 * and the creator tried to set a value for one of the props,
2817 * error out.
2818 */
2819 if ((zplver < ZPL_VERSION_INITIAL || zplver > ZPL_VERSION) ||
2820 (zplver >= ZPL_VERSION_FUID && !fuids_ok) ||
2821 (zplver >= ZPL_VERSION_SA && !sa_ok) ||
2822 (zplver < ZPL_VERSION_NORMALIZATION &&
2823 (norm != ZFS_PROP_UNDEFINED || u8 != ZFS_PROP_UNDEFINED ||
2824 sense != ZFS_PROP_UNDEFINED)))
2825 return (ENOTSUP);
2826
2827 /*
2828 * Put the version in the zplprops
2829 */
2830 VERIFY(nvlist_add_uint64(zplprops,
2831 zfs_prop_to_name(ZFS_PROP_VERSION), zplver) == 0);
2832
2833 if (norm == ZFS_PROP_UNDEFINED)
2834 VERIFY(zfs_get_zplprop(os, ZFS_PROP_NORMALIZE, &norm) == 0);
2835 VERIFY(nvlist_add_uint64(zplprops,
2836 zfs_prop_to_name(ZFS_PROP_NORMALIZE), norm) == 0);
2837
2838 /*
2839 * If we're normalizing, names must always be valid UTF-8 strings.
2840 */
2841 if (norm)
2842 u8 = 1;
2843 if (u8 == ZFS_PROP_UNDEFINED)
2844 VERIFY(zfs_get_zplprop(os, ZFS_PROP_UTF8ONLY, &u8) == 0);
2845 VERIFY(nvlist_add_uint64(zplprops,
2846 zfs_prop_to_name(ZFS_PROP_UTF8ONLY), u8) == 0);
2847
2848 if (sense == ZFS_PROP_UNDEFINED)
2849 VERIFY(zfs_get_zplprop(os, ZFS_PROP_CASE, &sense) == 0);
2850 VERIFY(nvlist_add_uint64(zplprops,
2851 zfs_prop_to_name(ZFS_PROP_CASE), sense) == 0);
2852
2853 if (is_ci)
2854 *is_ci = (sense == ZFS_CASE_INSENSITIVE);
2855
2856 return (0);
2857 }
2858
2859 static int
2860 zfs_fill_zplprops(const char *dataset, nvlist_t *createprops,
2861 nvlist_t *zplprops, boolean_t *is_ci)
2862 {
2863 boolean_t fuids_ok, sa_ok;
2864 uint64_t zplver = ZPL_VERSION;
2865 objset_t *os = NULL;
2866 char parentname[MAXNAMELEN];
2867 char *cp;
2868 spa_t *spa;
2869 uint64_t spa_vers;
2870 int error;
2871
2872 (void) strlcpy(parentname, dataset, sizeof (parentname));
2873 cp = strrchr(parentname, '/');
2874 ASSERT(cp != NULL);
2875 cp[0] = '\0';
2876
2877 if ((error = spa_open(dataset, &spa, FTAG)) != 0)
2878 return (error);
2879
2880 spa_vers = spa_version(spa);
2881 spa_close(spa, FTAG);
2882
2883 zplver = zfs_zpl_version_map(spa_vers);
2884 fuids_ok = (zplver >= ZPL_VERSION_FUID);
2885 sa_ok = (zplver >= ZPL_VERSION_SA);
2886
2887 /*
2888 * Open parent object set so we can inherit zplprop values.
2889 */
2890 if ((error = dmu_objset_hold(parentname, FTAG, &os)) != 0)
2891 return (error);
2892
2893 error = zfs_fill_zplprops_impl(os, zplver, fuids_ok, sa_ok, createprops,
2894 zplprops, is_ci);
2895 dmu_objset_rele(os, FTAG);
2896 return (error);
2897 }
2898
2899 static int
2900 zfs_fill_zplprops_root(uint64_t spa_vers, nvlist_t *createprops,
2901 nvlist_t *zplprops, boolean_t *is_ci)
2902 {
2903 boolean_t fuids_ok;
2904 boolean_t sa_ok;
2905 uint64_t zplver = ZPL_VERSION;
2906 int error;
2907
2908 zplver = zfs_zpl_version_map(spa_vers);
2909 fuids_ok = (zplver >= ZPL_VERSION_FUID);
2910 sa_ok = (zplver >= ZPL_VERSION_SA);
2911
2912 error = zfs_fill_zplprops_impl(NULL, zplver, fuids_ok, sa_ok,
2913 createprops, zplprops, is_ci);
2914 return (error);
2915 }
2916
2917 /*
2918 * inputs:
2919 * zc_objset_type type of objset to create (fs vs zvol)
2920 * zc_name name of new objset
2921 * zc_value name of snapshot to clone from (may be empty)
2922 * zc_nvlist_src{_size} nvlist of properties to apply
2923 *
2924 * outputs: none
2925 */
2926 static int
2927 zfs_ioc_create(zfs_cmd_t *zc)
2928 {
2929 objset_t *clone;
2930 int error = 0;
2931 zfs_creat_t zct;
2932 nvlist_t *nvprops = NULL;
2933 void (*cbfunc)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx);
2934 dmu_objset_type_t type = zc->zc_objset_type;
2935
2936 switch (type) {
2937
2938 case DMU_OST_ZFS:
2939 cbfunc = zfs_create_cb;
2940 break;
2941
2942 case DMU_OST_ZVOL:
2943 cbfunc = zvol_create_cb;
2944 break;
2945
2946 default:
2947 cbfunc = NULL;
2948 break;
2949 }
2950 if (strchr(zc->zc_name, '@') ||
2951 strchr(zc->zc_name, '%'))
2952 return (EINVAL);
2953
2954 if (zc->zc_nvlist_src != 0 &&
2955 (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2956 zc->zc_iflags, &nvprops)) != 0)
2957 return (error);
2958
2959 zct.zct_zplprops = NULL;
2960 zct.zct_props = nvprops;
2961
2962 if (zc->zc_value[0] != '\0') {
2963 /*
2964 * We're creating a clone of an existing snapshot.
2965 */
2966 zc->zc_value[sizeof (zc->zc_value) - 1] = '\0';
2967 if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0) {
2968 nvlist_free(nvprops);
2969 return (EINVAL);
2970 }
2971
2972 error = dmu_objset_hold(zc->zc_value, FTAG, &clone);
2973 if (error) {
2974 nvlist_free(nvprops);
2975 return (error);
2976 }
2977
2978 error = dmu_objset_clone(zc->zc_name, dmu_objset_ds(clone), 0);
2979 dmu_objset_rele(clone, FTAG);
2980 if (error) {
2981 nvlist_free(nvprops);
2982 return (error);
2983 }
2984 } else {
2985 boolean_t is_insensitive = B_FALSE;
2986
2987 if (cbfunc == NULL) {
2988 nvlist_free(nvprops);
2989 return (EINVAL);
2990 }
2991
2992 if (type == DMU_OST_ZVOL) {
2993 uint64_t volsize, volblocksize;
2994
2995 if (nvprops == NULL ||
2996 nvlist_lookup_uint64(nvprops,
2997 zfs_prop_to_name(ZFS_PROP_VOLSIZE),
2998 &volsize) != 0) {
2999 nvlist_free(nvprops);
3000 return (EINVAL);
3001 }
3002
3003 if ((error = nvlist_lookup_uint64(nvprops,
3004 zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE),
3005 &volblocksize)) != 0 && error != ENOENT) {
3006 nvlist_free(nvprops);
3007 return (EINVAL);
3008 }
3009
3010 if (error != 0)
3011 volblocksize = zfs_prop_default_numeric(
3012 ZFS_PROP_VOLBLOCKSIZE);
3013
3014 if ((error = zvol_check_volblocksize(
3015 volblocksize)) != 0 ||
3016 (error = zvol_check_volsize(volsize,
3017 volblocksize)) != 0) {
3018 nvlist_free(nvprops);
3019 return (error);
3020 }
3021 } else if (type == DMU_OST_ZFS) {
3022 int error;
3023
3024 /*
3025 * We have to have normalization and
3026 * case-folding flags correct when we do the
3027 * file system creation, so go figure them out
3028 * now.
3029 */
3030 VERIFY(nvlist_alloc(&zct.zct_zplprops,
3031 NV_UNIQUE_NAME, KM_SLEEP) == 0);
3032 error = zfs_fill_zplprops(zc->zc_name, nvprops,
3033 zct.zct_zplprops, &is_insensitive);
3034 if (error != 0) {
3035 nvlist_free(nvprops);
3036 nvlist_free(zct.zct_zplprops);
3037 return (error);
3038 }
3039 }
3040 error = dmu_objset_create(zc->zc_name, type,
3041 is_insensitive ? DS_FLAG_CI_DATASET : 0, cbfunc, &zct);
3042 nvlist_free(zct.zct_zplprops);
3043 }
3044
3045 /*
3046 * It would be nice to do this atomically.
3047 */
3048 if (error == 0) {
3049 error = zfs_set_prop_nvlist(zc->zc_name, ZPROP_SRC_LOCAL,
3050 nvprops, NULL);
3051 if (error != 0)
3052 (void) dmu_objset_destroy(zc->zc_name, B_FALSE);
3053 }
3054 nvlist_free(nvprops);
3055 return (error);
3056 }
3057
3058 /*
3059 * inputs:
3060 * zc_name name of filesystem
3061 * zc_value short name of snapshot
3062 * zc_cookie recursive flag
3063 * zc_nvlist_src[_size] property list
3064 *
3065 * outputs:
3066 * zc_value short snapname (i.e. part after the '@')
3067 */
3068 static int
3069 zfs_ioc_snapshot(zfs_cmd_t *zc)
3070 {
3071 nvlist_t *nvprops = NULL;
3072 int error;
3073 boolean_t recursive = zc->zc_cookie;
3074
3075 if (snapshot_namecheck(zc->zc_value, NULL, NULL) != 0)
3076 return (EINVAL);
3077
3078 if (zc->zc_nvlist_src != 0 &&
3079 (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
3080 zc->zc_iflags, &nvprops)) != 0)
3081 return (error);
3082
3083 error = zfs_check_userprops(zc->zc_name, nvprops);
3084 if (error)
3085 goto out;
3086
3087 if (!nvlist_empty(nvprops) &&
3088 zfs_earlier_version(zc->zc_name, SPA_VERSION_SNAP_PROPS)) {
3089 error = ENOTSUP;
3090 goto out;
3091 }
3092
3093 error = dmu_objset_snapshot(zc->zc_name, zc->zc_value, NULL,
3094 nvprops, recursive, B_FALSE, -1);
3095
3096 out:
3097 nvlist_free(nvprops);
3098 return (error);
3099 }
3100
3101 int
3102 zfs_unmount_snap(const char *name, void *arg)
3103 {
3104 #ifdef HAVE_SNAPSHOT
3105 vfs_t *vfsp = NULL;
3106
3107 if (arg) {
3108 char *snapname = arg;
3109 char *fullname = kmem_asprintf("%s@%s", name, snapname);
3110 vfsp = zfs_get_vfs(fullname);
3111 strfree(fullname);
3112 } else if (strchr(name, '@')) {
3113 vfsp = zfs_get_vfs(name);
3114 }
3115
3116 if (vfsp) {
3117 /*
3118 * Always force the unmount for snapshots.
3119 */
3120 int flag = MS_FORCE;
3121 int err;
3122
3123 if ((err = vn_vfswlock(vfsp->vfs_vnodecovered)) != 0) {
3124 mntput(vfsp);
3125 return (err);
3126 }
3127 mntput(vfsp);
3128 if ((err = dounmount(vfsp, flag, kcred)) != 0)
3129 return (err);
3130 }
3131 #endif /* HAVE_SNAPSHOT */
3132 return (0);
3133 }
3134
3135 /*
3136 * inputs:
3137 * zc_name name of filesystem
3138 * zc_value short name of snapshot
3139 * zc_defer_destroy mark for deferred destroy
3140 *
3141 * outputs: none
3142 */
3143 static int
3144 zfs_ioc_destroy_snaps(zfs_cmd_t *zc)
3145 {
3146 int err;
3147
3148 if (snapshot_namecheck(zc->zc_value, NULL, NULL) != 0)
3149 return (EINVAL);
3150 err = dmu_objset_find(zc->zc_name,
3151 zfs_unmount_snap, zc->zc_value, DS_FIND_CHILDREN);
3152 if (err)
3153 return (err);
3154 return (dmu_snapshots_destroy(zc->zc_name, zc->zc_value,
3155 zc->zc_defer_destroy));
3156 }
3157
3158 /*
3159 * inputs:
3160 * zc_name name of dataset to destroy
3161 * zc_objset_type type of objset
3162 * zc_defer_destroy mark for deferred destroy
3163 *
3164 * outputs: none
3165 */
3166 static int
3167 zfs_ioc_destroy(zfs_cmd_t *zc)
3168 {
3169 int err;
3170 if (strchr(zc->zc_name, '@') && zc->zc_objset_type == DMU_OST_ZFS) {
3171 err = zfs_unmount_snap(zc->zc_name, NULL);
3172 if (err)
3173 return (err);
3174 }
3175
3176 err = dmu_objset_destroy(zc->zc_name, zc->zc_defer_destroy);
3177 if (zc->zc_objset_type == DMU_OST_ZVOL && err == 0)
3178 (void) zvol_remove_minor(zc->zc_name);
3179 return (err);
3180 }
3181
3182 /*
3183 * inputs:
3184 * zc_name name of dataset to rollback (to most recent snapshot)
3185 *
3186 * outputs: none
3187 */
3188 static int
3189 zfs_ioc_rollback(zfs_cmd_t *zc)
3190 {
3191 dsl_dataset_t *ds, *clone;
3192 int error;
3193 zfs_sb_t *zsb;
3194 char *clone_name;
3195
3196 error = dsl_dataset_hold(zc->zc_name, FTAG, &ds);
3197 if (error)
3198 return (error);
3199
3200 /* must not be a snapshot */
3201 if (dsl_dataset_is_snapshot(ds)) {
3202 dsl_dataset_rele(ds, FTAG);
3203 return (EINVAL);
3204 }
3205
3206 /* must have a most recent snapshot */
3207 if (ds->ds_phys->ds_prev_snap_txg < TXG_INITIAL) {
3208 dsl_dataset_rele(ds, FTAG);
3209 return (EINVAL);
3210 }
3211
3212 /*
3213 * Create clone of most recent snapshot.
3214 */
3215 clone_name = kmem_asprintf("%s/%%rollback", zc->zc_name);
3216 error = dmu_objset_clone(clone_name, ds->ds_prev, DS_FLAG_INCONSISTENT);
3217 if (error)
3218 goto out;
3219
3220 error = dsl_dataset_own(clone_name, B_TRUE, FTAG, &clone);
3221 if (error)
3222 goto out;
3223
3224 /*
3225 * Do clone swap.
3226 */
3227 if (get_zfs_sb(zc->zc_name, &zsb) == 0) {
3228 error = zfs_suspend_fs(zsb);
3229 if (error == 0) {
3230 int resume_err;
3231
3232 if (dsl_dataset_tryown(ds, B_FALSE, FTAG)) {
3233 error = dsl_dataset_clone_swap(clone, ds,
3234 B_TRUE);
3235 dsl_dataset_disown(ds, FTAG);
3236 ds = NULL;
3237 } else {
3238 error = EBUSY;
3239 }
3240 resume_err = zfs_resume_fs(zsb, zc->zc_name);
3241 error = error ? error : resume_err;
3242 }
3243 deactivate_super(zsb->z_sb);
3244 } else {
3245 if (dsl_dataset_tryown(ds, B_FALSE, FTAG)) {
3246 error = dsl_dataset_clone_swap(clone, ds, B_TRUE);
3247 dsl_dataset_disown(ds, FTAG);
3248 ds = NULL;
3249 } else {
3250 error = EBUSY;
3251 }
3252 }
3253
3254 /*
3255 * Destroy clone (which also closes it).
3256 */
3257 (void) dsl_dataset_destroy(clone, FTAG, B_FALSE);
3258
3259 out:
3260 strfree(clone_name);
3261 if (ds)
3262 dsl_dataset_rele(ds, FTAG);
3263 return (error);
3264 }
3265
3266 /*
3267 * inputs:
3268 * zc_name old name of dataset
3269 * zc_value new name of dataset
3270 * zc_cookie recursive flag (only valid for snapshots)
3271 *
3272 * outputs: none
3273 */
3274 static int
3275 zfs_ioc_rename(zfs_cmd_t *zc)
3276 {
3277 boolean_t recursive = zc->zc_cookie & 1;
3278 int err;
3279
3280 zc->zc_value[sizeof (zc->zc_value) - 1] = '\0';
3281 if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
3282 strchr(zc->zc_value, '%'))
3283 return (EINVAL);
3284
3285 /*
3286 * Unmount snapshot unless we're doing a recursive rename,
3287 * in which case the dataset code figures out which snapshots
3288 * to unmount.
3289 */
3290 if (!recursive && strchr(zc->zc_name, '@') != NULL &&
3291 zc->zc_objset_type == DMU_OST_ZFS) {
3292 err = zfs_unmount_snap(zc->zc_name, NULL);
3293 if (err)
3294 return (err);
3295 }
3296
3297 err = dmu_objset_rename(zc->zc_name, zc->zc_value, recursive);
3298 if ((err == 0) && (zc->zc_objset_type == DMU_OST_ZVOL)) {
3299 (void) zvol_remove_minor(zc->zc_name);
3300 (void) zvol_create_minor(zc->zc_value);
3301 }
3302
3303 return (err);
3304 }
3305
3306 static int
3307 zfs_check_settable(const char *dsname, nvpair_t *pair, cred_t *cr)
3308 {
3309 const char *propname = nvpair_name(pair);
3310 boolean_t issnap = (strchr(dsname, '@') != NULL);
3311 zfs_prop_t prop = zfs_name_to_prop(propname);
3312 uint64_t intval;
3313 int err;
3314
3315 if (prop == ZPROP_INVAL) {
3316 if (zfs_prop_user(propname)) {
3317 if ((err = zfs_secpolicy_write_perms(dsname,
3318 ZFS_DELEG_PERM_USERPROP, cr)))
3319 return (err);
3320 return (0);
3321 }
3322
3323 if (!issnap && zfs_prop_userquota(propname)) {
3324 const char *perm = NULL;
3325 const char *uq_prefix =
3326 zfs_userquota_prop_prefixes[ZFS_PROP_USERQUOTA];
3327 const char *gq_prefix =
3328 zfs_userquota_prop_prefixes[ZFS_PROP_GROUPQUOTA];
3329
3330 if (strncmp(propname, uq_prefix,
3331 strlen(uq_prefix)) == 0) {
3332 perm = ZFS_DELEG_PERM_USERQUOTA;
3333 } else if (strncmp(propname, gq_prefix,
3334 strlen(gq_prefix)) == 0) {
3335 perm = ZFS_DELEG_PERM_GROUPQUOTA;
3336 } else {
3337 /* USERUSED and GROUPUSED are read-only */
3338 return (EINVAL);
3339 }
3340
3341 if ((err = zfs_secpolicy_write_perms(dsname, perm, cr)))
3342 return (err);
3343 return (0);
3344 }
3345
3346 return (EINVAL);
3347 }
3348
3349 if (issnap)
3350 return (EINVAL);
3351
3352 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
3353 /*
3354 * dsl_prop_get_all_impl() returns properties in this
3355 * format.
3356 */
3357 nvlist_t *attrs;
3358 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
3359 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
3360 &pair) == 0);
3361 }
3362
3363 /*
3364 * Check that this value is valid for this pool version
3365 */
3366 switch (prop) {
3367 case ZFS_PROP_COMPRESSION:
3368 /*
3369 * If the user specified gzip compression, make sure
3370 * the SPA supports it. We ignore any errors here since
3371 * we'll catch them later.
3372 */
3373 if (nvpair_type(pair) == DATA_TYPE_UINT64 &&
3374 nvpair_value_uint64(pair, &intval) == 0) {
3375 if (intval >= ZIO_COMPRESS_GZIP_1 &&
3376 intval <= ZIO_COMPRESS_GZIP_9 &&
3377 zfs_earlier_version(dsname,
3378 SPA_VERSION_GZIP_COMPRESSION)) {
3379 return (ENOTSUP);
3380 }
3381
3382 if (intval == ZIO_COMPRESS_ZLE &&
3383 zfs_earlier_version(dsname,
3384 SPA_VERSION_ZLE_COMPRESSION))
3385 return (ENOTSUP);
3386
3387 /*
3388 * If this is a bootable dataset then
3389 * verify that the compression algorithm
3390 * is supported for booting. We must return
3391 * something other than ENOTSUP since it
3392 * implies a downrev pool version.
3393 */
3394 if (zfs_is_bootfs(dsname) &&
3395 !BOOTFS_COMPRESS_VALID(intval)) {
3396 return (ERANGE);
3397 }
3398 }
3399 break;
3400
3401 case ZFS_PROP_COPIES:
3402 if (zfs_earlier_version(dsname, SPA_VERSION_DITTO_BLOCKS))
3403 return (ENOTSUP);
3404 break;
3405
3406 case ZFS_PROP_DEDUP:
3407 if (zfs_earlier_version(dsname, SPA_VERSION_DEDUP))
3408 return (ENOTSUP);
3409 break;
3410
3411 case ZFS_PROP_SHARESMB:
3412 if (zpl_earlier_version(dsname, ZPL_VERSION_FUID))
3413 return (ENOTSUP);
3414 break;
3415
3416 case ZFS_PROP_ACLINHERIT:
3417 if (nvpair_type(pair) == DATA_TYPE_UINT64 &&
3418 nvpair_value_uint64(pair, &intval) == 0) {
3419 if (intval == ZFS_ACL_PASSTHROUGH_X &&
3420 zfs_earlier_version(dsname,
3421 SPA_VERSION_PASSTHROUGH_X))
3422 return (ENOTSUP);
3423 }
3424 break;
3425 default:
3426 break;
3427 }
3428
3429 return (zfs_secpolicy_setprop(dsname, prop, pair, CRED()));
3430 }
3431
3432 /*
3433 * Removes properties from the given props list that fail permission checks
3434 * needed to clear them and to restore them in case of a receive error. For each
3435 * property, make sure we have both set and inherit permissions.
3436 *
3437 * Returns the first error encountered if any permission checks fail. If the
3438 * caller provides a non-NULL errlist, it also gives the complete list of names
3439 * of all the properties that failed a permission check along with the
3440 * corresponding error numbers. The caller is responsible for freeing the
3441 * returned errlist.
3442 *
3443 * If every property checks out successfully, zero is returned and the list
3444 * pointed at by errlist is NULL.
3445 */
3446 static int
3447 zfs_check_clearable(char *dataset, nvlist_t *props, nvlist_t **errlist)
3448 {
3449 zfs_cmd_t *zc;
3450 nvpair_t *pair, *next_pair;
3451 nvlist_t *errors;
3452 int err, rv = 0;
3453
3454 if (props == NULL)
3455 return (0);
3456
3457 VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0);
3458
3459 zc = kmem_alloc(sizeof (zfs_cmd_t), KM_SLEEP | KM_NODEBUG);
3460 (void) strcpy(zc->zc_name, dataset);
3461 pair = nvlist_next_nvpair(props, NULL);
3462 while (pair != NULL) {
3463 next_pair = nvlist_next_nvpair(props, pair);
3464
3465 (void) strcpy(zc->zc_value, nvpair_name(pair));
3466 if ((err = zfs_check_settable(dataset, pair, CRED())) != 0 ||
3467 (err = zfs_secpolicy_inherit(zc, CRED())) != 0) {
3468 VERIFY(nvlist_remove_nvpair(props, pair) == 0);
3469 VERIFY(nvlist_add_int32(errors,
3470 zc->zc_value, err) == 0);
3471 }
3472 pair = next_pair;
3473 }
3474 kmem_free(zc, sizeof (zfs_cmd_t));
3475
3476 if ((pair = nvlist_next_nvpair(errors, NULL)) == NULL) {
3477 nvlist_free(errors);
3478 errors = NULL;
3479 } else {
3480 VERIFY(nvpair_value_int32(pair, &rv) == 0);
3481 }
3482
3483 if (errlist == NULL)
3484 nvlist_free(errors);
3485 else
3486 *errlist = errors;
3487
3488 return (rv);
3489 }
3490
3491 static boolean_t
3492 propval_equals(nvpair_t *p1, nvpair_t *p2)
3493 {
3494 if (nvpair_type(p1) == DATA_TYPE_NVLIST) {
3495 /* dsl_prop_get_all_impl() format */
3496 nvlist_t *attrs;
3497 VERIFY(nvpair_value_nvlist(p1, &attrs) == 0);
3498 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
3499 &p1) == 0);
3500 }
3501
3502 if (nvpair_type(p2) == DATA_TYPE_NVLIST) {
3503 nvlist_t *attrs;
3504 VERIFY(nvpair_value_nvlist(p2, &attrs) == 0);
3505 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
3506 &p2) == 0);
3507 }
3508
3509 if (nvpair_type(p1) != nvpair_type(p2))
3510 return (B_FALSE);
3511
3512 if (nvpair_type(p1) == DATA_TYPE_STRING) {
3513 char *valstr1, *valstr2;
3514
3515 VERIFY(nvpair_value_string(p1, (char **)&valstr1) == 0);
3516 VERIFY(nvpair_value_string(p2, (char **)&valstr2) == 0);
3517 return (strcmp(valstr1, valstr2) == 0);
3518 } else {
3519 uint64_t intval1, intval2;
3520
3521 VERIFY(nvpair_value_uint64(p1, &intval1) == 0);
3522 VERIFY(nvpair_value_uint64(p2, &intval2) == 0);
3523 return (intval1 == intval2);
3524 }
3525 }
3526
3527 /*
3528 * Remove properties from props if they are not going to change (as determined
3529 * by comparison with origprops). Remove them from origprops as well, since we
3530 * do not need to clear or restore properties that won't change.
3531 */
3532 static void
3533 props_reduce(nvlist_t *props, nvlist_t *origprops)
3534 {
3535 nvpair_t *pair, *next_pair;
3536
3537 if (origprops == NULL)
3538 return; /* all props need to be received */
3539
3540 pair = nvlist_next_nvpair(props, NULL);
3541 while (pair != NULL) {
3542 const char *propname = nvpair_name(pair);
3543 nvpair_t *match;
3544
3545 next_pair = nvlist_next_nvpair(props, pair);
3546
3547 if ((nvlist_lookup_nvpair(origprops, propname,
3548 &match) != 0) || !propval_equals(pair, match))
3549 goto next; /* need to set received value */
3550
3551 /* don't clear the existing received value */
3552 (void) nvlist_remove_nvpair(origprops, match);
3553 /* don't bother receiving the property */
3554 (void) nvlist_remove_nvpair(props, pair);
3555 next:
3556 pair = next_pair;
3557 }
3558 }
3559
3560 #ifdef DEBUG
3561 static boolean_t zfs_ioc_recv_inject_err;
3562 #endif
3563
3564 /*
3565 * inputs:
3566 * zc_name name of containing filesystem
3567 * zc_nvlist_src{_size} nvlist of properties to apply
3568 * zc_value name of snapshot to create
3569 * zc_string name of clone origin (if DRR_FLAG_CLONE)
3570 * zc_cookie file descriptor to recv from
3571 * zc_begin_record the BEGIN record of the stream (not byteswapped)
3572 * zc_guid force flag
3573 * zc_cleanup_fd cleanup-on-exit file descriptor
3574 * zc_action_handle handle for this guid/ds mapping (or zero on first call)
3575 *
3576 * outputs:
3577 * zc_cookie number of bytes read
3578 * zc_nvlist_dst{_size} error for each unapplied received property
3579 * zc_obj zprop_errflags_t
3580 * zc_action_handle handle for this guid/ds mapping
3581 */
3582 static int
3583 zfs_ioc_recv(zfs_cmd_t *zc)
3584 {
3585 file_t *fp;
3586 objset_t *os;
3587 dmu_recv_cookie_t drc;
3588 boolean_t force = (boolean_t)zc->zc_guid;
3589 int fd;
3590 int error = 0;
3591 int props_error = 0;
3592 nvlist_t *errors;
3593 offset_t off;
3594 nvlist_t *props = NULL; /* sent properties */
3595 nvlist_t *origprops = NULL; /* existing properties */
3596 objset_t *origin = NULL;
3597 char *tosnap;
3598 char tofs[ZFS_MAXNAMELEN];
3599 boolean_t first_recvd_props = B_FALSE;
3600
3601 if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
3602 strchr(zc->zc_value, '@') == NULL ||
3603 strchr(zc->zc_value, '%'))
3604 return (EINVAL);
3605
3606 (void) strcpy(tofs, zc->zc_value);
3607 tosnap = strchr(tofs, '@');
3608 *tosnap++ = '\0';
3609
3610 if (zc->zc_nvlist_src != 0 &&
3611 (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
3612 zc->zc_iflags, &props)) != 0)
3613 return (error);
3614
3615 fd = zc->zc_cookie;
3616 fp = getf(fd);
3617 if (fp == NULL) {
3618 nvlist_free(props);
3619 return (EBADF);
3620 }
3621
3622 VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0);
3623
3624 if (props && dmu_objset_hold(tofs, FTAG, &os) == 0) {
3625 if ((spa_version(os->os_spa) >= SPA_VERSION_RECVD_PROPS) &&
3626 !dsl_prop_get_hasrecvd(os)) {
3627 first_recvd_props = B_TRUE;
3628 }
3629
3630 /*
3631 * If new received properties are supplied, they are to
3632 * completely replace the existing received properties, so stash
3633 * away the existing ones.
3634 */
3635 if (dsl_prop_get_received(os, &origprops) == 0) {
3636 nvlist_t *errlist = NULL;
3637 /*
3638 * Don't bother writing a property if its value won't
3639 * change (and avoid the unnecessary security checks).
3640 *
3641 * The first receive after SPA_VERSION_RECVD_PROPS is a
3642 * special case where we blow away all local properties
3643 * regardless.
3644 */
3645 if (!first_recvd_props)
3646 props_reduce(props, origprops);
3647 if (zfs_check_clearable(tofs, origprops,
3648 &errlist) != 0)
3649 (void) nvlist_merge(errors, errlist, 0);
3650 nvlist_free(errlist);
3651 }
3652
3653 dmu_objset_rele(os, FTAG);
3654 }
3655
3656 if (zc->zc_string[0]) {
3657 error = dmu_objset_hold(zc->zc_string, FTAG, &origin);
3658 if (error)
3659 goto out;
3660 }
3661
3662 error = dmu_recv_begin(tofs, tosnap, zc->zc_top_ds,
3663 &zc->zc_begin_record, force, origin, &drc);
3664 if (origin)
3665 dmu_objset_rele(origin, FTAG);
3666 if (error)
3667 goto out;
3668
3669 /*
3670 * Set properties before we receive the stream so that they are applied
3671 * to the new data. Note that we must call dmu_recv_stream() if
3672 * dmu_recv_begin() succeeds.
3673 */
3674 if (props) {
3675 nvlist_t *errlist;
3676
3677 if (dmu_objset_from_ds(drc.drc_logical_ds, &os) == 0) {
3678 if (drc.drc_newfs) {
3679 if (spa_version(os->os_spa) >=
3680 SPA_VERSION_RECVD_PROPS)
3681 first_recvd_props = B_TRUE;
3682 } else if (origprops != NULL) {
3683 if (clear_received_props(os, tofs, origprops,
3684 first_recvd_props ? NULL : props) != 0)
3685 zc->zc_obj |= ZPROP_ERR_NOCLEAR;
3686 } else {
3687 zc->zc_obj |= ZPROP_ERR_NOCLEAR;
3688 }
3689 dsl_prop_set_hasrecvd(os);
3690 } else if (!drc.drc_newfs) {
3691 zc->zc_obj |= ZPROP_ERR_NOCLEAR;
3692 }
3693
3694 (void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_RECEIVED,
3695 props, &errlist);
3696 (void) nvlist_merge(errors, errlist, 0);
3697 nvlist_free(errlist);
3698 }
3699
3700 if (fit_error_list(zc, &errors) != 0 || put_nvlist(zc, errors) != 0) {
3701 /*
3702 * Caller made zc->zc_nvlist_dst less than the minimum expected
3703 * size or supplied an invalid address.
3704 */
3705 props_error = EINVAL;
3706 }
3707
3708 off = fp->f_offset;
3709 error = dmu_recv_stream(&drc, fp->f_vnode, &off, zc->zc_cleanup_fd,
3710 &zc->zc_action_handle);
3711
3712 if (error == 0) {
3713 zfs_sb_t *zsb = NULL;
3714
3715 if (get_zfs_sb(tofs, &zsb) == 0) {
3716 /* online recv */
3717 int end_err;
3718
3719 error = zfs_suspend_fs(zsb);
3720 /*
3721 * If the suspend fails, then the recv_end will
3722 * likely also fail, and clean up after itself.
3723 */
3724 end_err = dmu_recv_end(&drc);
3725 if (error == 0)
3726 error = zfs_resume_fs(zsb, tofs);
3727 error = error ? error : end_err;
3728 deactivate_super(zsb->z_sb);
3729 } else {
3730 error = dmu_recv_end(&drc);
3731 }
3732 }
3733
3734 zc->zc_cookie = off - fp->f_offset;
3735 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
3736 fp->f_offset = off;
3737
3738 #ifdef DEBUG
3739 if (zfs_ioc_recv_inject_err) {
3740 zfs_ioc_recv_inject_err = B_FALSE;
3741 error = 1;
3742 }
3743 #endif
3744 /*
3745 * On error, restore the original props.
3746 */
3747 if (error && props) {
3748 if (dmu_objset_hold(tofs, FTAG, &os) == 0) {
3749 if (clear_received_props(os, tofs, props, NULL) != 0) {
3750 /*
3751 * We failed to clear the received properties.
3752 * Since we may have left a $recvd value on the
3753 * system, we can't clear the $hasrecvd flag.
3754 */
3755 zc->zc_obj |= ZPROP_ERR_NORESTORE;
3756 } else if (first_recvd_props) {
3757 dsl_prop_unset_hasrecvd(os);
3758 }
3759 dmu_objset_rele(os, FTAG);
3760 } else if (!drc.drc_newfs) {
3761 /* We failed to clear the received properties. */
3762 zc->zc_obj |= ZPROP_ERR_NORESTORE;
3763 }
3764
3765 if (origprops == NULL && !drc.drc_newfs) {
3766 /* We failed to stash the original properties. */
3767 zc->zc_obj |= ZPROP_ERR_NORESTORE;
3768 }
3769
3770 /*
3771 * dsl_props_set() will not convert RECEIVED to LOCAL on or
3772 * after SPA_VERSION_RECVD_PROPS, so we need to specify LOCAL
3773 * explictly if we're restoring local properties cleared in the
3774 * first new-style receive.
3775 */
3776 if (origprops != NULL &&
3777 zfs_set_prop_nvlist(tofs, (first_recvd_props ?
3778 ZPROP_SRC_LOCAL : ZPROP_SRC_RECEIVED),
3779 origprops, NULL) != 0) {
3780 /*
3781 * We stashed the original properties but failed to
3782 * restore them.
3783 */
3784 zc->zc_obj |= ZPROP_ERR_NORESTORE;
3785 }
3786 }
3787 out:
3788 nvlist_free(props);
3789 nvlist_free(origprops);
3790 nvlist_free(errors);
3791 releasef(fd);
3792
3793 if (error == 0)
3794 error = props_error;
3795
3796 return (error);
3797 }
3798
3799 /*
3800 * inputs:
3801 * zc_name name of snapshot to send
3802 * zc_cookie file descriptor to send stream to
3803 * zc_obj fromorigin flag (mutually exclusive with zc_fromobj)
3804 * zc_sendobj objsetid of snapshot to send
3805 * zc_fromobj objsetid of incremental fromsnap (may be zero)
3806 *
3807 * outputs: none
3808 */
3809 static int
3810 zfs_ioc_send(zfs_cmd_t *zc)
3811 {
3812 objset_t *fromsnap = NULL;
3813 objset_t *tosnap;
3814 file_t *fp;
3815 int error;
3816 offset_t off;
3817 dsl_dataset_t *ds;
3818 dsl_dataset_t *dsfrom = NULL;
3819 spa_t *spa;
3820 dsl_pool_t *dp;
3821
3822 error = spa_open(zc->zc_name, &spa, FTAG);
3823 if (error)
3824 return (error);
3825
3826 dp = spa_get_dsl(spa);
3827 rw_enter(&dp->dp_config_rwlock, RW_READER);
3828 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &ds);
3829 rw_exit(&dp->dp_config_rwlock);
3830 if (error) {
3831 spa_close(spa, FTAG);
3832 return (error);
3833 }
3834
3835 error = dmu_objset_from_ds(ds, &tosnap);
3836 if (error) {
3837 dsl_dataset_rele(ds, FTAG);
3838 spa_close(spa, FTAG);
3839 return (error);
3840 }
3841
3842 if (zc->zc_fromobj != 0) {
3843 rw_enter(&dp->dp_config_rwlock, RW_READER);
3844 error = dsl_dataset_hold_obj(dp, zc->zc_fromobj, FTAG, &dsfrom);
3845 rw_exit(&dp->dp_config_rwlock);
3846 spa_close(spa, FTAG);
3847 if (error) {
3848 dsl_dataset_rele(ds, FTAG);
3849 return (error);
3850 }
3851 error = dmu_objset_from_ds(dsfrom, &fromsnap);
3852 if (error) {
3853 dsl_dataset_rele(dsfrom, FTAG);
3854 dsl_dataset_rele(ds, FTAG);
3855 return (error);
3856 }
3857 } else {
3858 spa_close(spa, FTAG);
3859 }
3860
3861 fp = getf(zc->zc_cookie);
3862 if (fp == NULL) {
3863 dsl_dataset_rele(ds, FTAG);
3864 if (dsfrom)
3865 dsl_dataset_rele(dsfrom, FTAG);
3866 return (EBADF);
3867 }
3868
3869 off = fp->f_offset;
3870 error = dmu_sendbackup(tosnap, fromsnap, zc->zc_obj, fp->f_vnode, &off);
3871
3872 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
3873 fp->f_offset = off;
3874 releasef(zc->zc_cookie);
3875 if (dsfrom)
3876 dsl_dataset_rele(dsfrom, FTAG);
3877 dsl_dataset_rele(ds, FTAG);
3878 return (error);
3879 }
3880
3881 static int
3882 zfs_ioc_inject_fault(zfs_cmd_t *zc)
3883 {
3884 int id, error;
3885
3886 error = zio_inject_fault(zc->zc_name, (int)zc->zc_guid, &id,
3887 &zc->zc_inject_record);
3888
3889 if (error == 0)
3890 zc->zc_guid = (uint64_t)id;
3891
3892 return (error);
3893 }
3894
3895 static int
3896 zfs_ioc_clear_fault(zfs_cmd_t *zc)
3897 {
3898 return (zio_clear_fault((int)zc->zc_guid));
3899 }
3900
3901 static int
3902 zfs_ioc_inject_list_next(zfs_cmd_t *zc)
3903 {
3904 int id = (int)zc->zc_guid;
3905 int error;
3906
3907 error = zio_inject_list_next(&id, zc->zc_name, sizeof (zc->zc_name),
3908 &zc->zc_inject_record);
3909
3910 zc->zc_guid = id;
3911
3912 return (error);
3913 }
3914
3915 static int
3916 zfs_ioc_error_log(zfs_cmd_t *zc)
3917 {
3918 spa_t *spa;
3919 int error;
3920 size_t count = (size_t)zc->zc_nvlist_dst_size;
3921
3922 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
3923 return (error);
3924
3925 error = spa_get_errlog(spa, (void *)(uintptr_t)zc->zc_nvlist_dst,
3926 &count);
3927 if (error == 0)
3928 zc->zc_nvlist_dst_size = count;
3929 else
3930 zc->zc_nvlist_dst_size = spa_get_errlog_size(spa);
3931
3932 spa_close(spa, FTAG);
3933
3934 return (error);
3935 }
3936
3937 static int
3938 zfs_ioc_clear(zfs_cmd_t *zc)
3939 {
3940 spa_t *spa;
3941 vdev_t *vd;
3942 int error;
3943
3944 /*
3945 * On zpool clear we also fix up missing slogs
3946 */
3947 mutex_enter(&spa_namespace_lock);
3948 spa = spa_lookup(zc->zc_name);
3949 if (spa == NULL) {
3950 mutex_exit(&spa_namespace_lock);
3951 return (EIO);
3952 }
3953 if (spa_get_log_state(spa) == SPA_LOG_MISSING) {
3954 /* we need to let spa_open/spa_load clear the chains */
3955 spa_set_log_state(spa, SPA_LOG_CLEAR);
3956 }
3957 spa->spa_last_open_failed = 0;
3958 mutex_exit(&spa_namespace_lock);
3959
3960 if (zc->zc_cookie & ZPOOL_NO_REWIND) {
3961 error = spa_open(zc->zc_name, &spa, FTAG);
3962 } else {
3963 nvlist_t *policy;
3964 nvlist_t *config = NULL;
3965
3966 if (zc->zc_nvlist_src == 0)
3967 return (EINVAL);
3968
3969 if ((error = get_nvlist(zc->zc_nvlist_src,
3970 zc->zc_nvlist_src_size, zc->zc_iflags, &policy)) == 0) {
3971 error = spa_open_rewind(zc->zc_name, &spa, FTAG,
3972 policy, &config);
3973 if (config != NULL) {
3974 int err;
3975
3976 if ((err = put_nvlist(zc, config)) != 0)
3977 error = err;
3978 nvlist_free(config);
3979 }
3980 nvlist_free(policy);
3981 }
3982 }
3983
3984 if (error)
3985 return (error);
3986
3987 spa_vdev_state_enter(spa, SCL_NONE);
3988
3989 if (zc->zc_guid == 0) {
3990 vd = NULL;
3991 } else {
3992 vd = spa_lookup_by_guid(spa, zc->zc_guid, B_TRUE);
3993 if (vd == NULL) {
3994 (void) spa_vdev_state_exit(spa, NULL, ENODEV);
3995 spa_close(spa, FTAG);
3996 return (ENODEV);
3997 }
3998 }
3999
4000 vdev_clear(spa, vd);
4001
4002 (void) spa_vdev_state_exit(spa, NULL, 0);
4003
4004 /*
4005 * Resume any suspended I/Os.
4006 */
4007 if (zio_resume(spa) != 0)
4008 error = EIO;
4009
4010 spa_close(spa, FTAG);
4011
4012 return (error);
4013 }
4014
4015 /*
4016 * inputs:
4017 * zc_name name of filesystem
4018 * zc_value name of origin snapshot
4019 *
4020 * outputs:
4021 * zc_string name of conflicting snapshot, if there is one
4022 */
4023 static int
4024 zfs_ioc_promote(zfs_cmd_t *zc)
4025 {
4026 char *cp;
4027
4028 /*
4029 * We don't need to unmount *all* the origin fs's snapshots, but
4030 * it's easier.
4031 */
4032 cp = strchr(zc->zc_value, '@');
4033 if (cp)
4034 *cp = '\0';
4035 (void) dmu_objset_find(zc->zc_value,
4036 zfs_unmount_snap, NULL, DS_FIND_SNAPSHOTS);
4037 return (dsl_dataset_promote(zc->zc_name, zc->zc_string));
4038 }
4039
4040 /*
4041 * Retrieve a single {user|group}{used|quota}@... property.
4042 *
4043 * inputs:
4044 * zc_name name of filesystem
4045 * zc_objset_type zfs_userquota_prop_t
4046 * zc_value domain name (eg. "S-1-234-567-89")
4047 * zc_guid RID/UID/GID
4048 *
4049 * outputs:
4050 * zc_cookie property value
4051 */
4052 static int
4053 zfs_ioc_userspace_one(zfs_cmd_t *zc)
4054 {
4055 zfs_sb_t *zsb;
4056 int error;
4057
4058 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
4059 return (EINVAL);
4060
4061 error = zfs_sb_hold(zc->zc_name, FTAG, &zsb, B_FALSE);
4062 if (error)
4063 return (error);
4064
4065 error = zfs_userspace_one(zsb,
4066 zc->zc_objset_type, zc->zc_value, zc->zc_guid, &zc->zc_cookie);
4067 zfs_sb_rele(zsb, FTAG);
4068
4069 return (error);
4070 }
4071
4072 /*
4073 * inputs:
4074 * zc_name name of filesystem
4075 * zc_cookie zap cursor
4076 * zc_objset_type zfs_userquota_prop_t
4077 * zc_nvlist_dst[_size] buffer to fill (not really an nvlist)
4078 *
4079 * outputs:
4080 * zc_nvlist_dst[_size] data buffer (array of zfs_useracct_t)
4081 * zc_cookie zap cursor
4082 */
4083 static int
4084 zfs_ioc_userspace_many(zfs_cmd_t *zc)
4085 {
4086 zfs_sb_t *zsb;
4087 int bufsize = zc->zc_nvlist_dst_size;
4088 int error;
4089 void *buf;
4090
4091 if (bufsize <= 0)
4092 return (ENOMEM);
4093
4094 error = zfs_sb_hold(zc->zc_name, FTAG, &zsb, B_FALSE);
4095 if (error)
4096 return (error);
4097
4098 buf = vmem_alloc(bufsize, KM_SLEEP);
4099
4100 error = zfs_userspace_many(zsb, zc->zc_objset_type, &zc->zc_cookie,
4101 buf, &zc->zc_nvlist_dst_size);
4102
4103 if (error == 0) {
4104 error = xcopyout(buf,
4105 (void *)(uintptr_t)zc->zc_nvlist_dst,
4106 zc->zc_nvlist_dst_size);
4107 }
4108 vmem_free(buf, bufsize);
4109 zfs_sb_rele(zsb, FTAG);
4110
4111 return (error);
4112 }
4113
4114 /*
4115 * inputs:
4116 * zc_name name of filesystem
4117 *
4118 * outputs:
4119 * none
4120 */
4121 static int
4122 zfs_ioc_userspace_upgrade(zfs_cmd_t *zc)
4123 {
4124 objset_t *os;
4125 int error = 0;
4126 zfs_sb_t *zsb;
4127
4128 if (get_zfs_sb(zc->zc_name, &zsb) == 0) {
4129 if (!dmu_objset_userused_enabled(zsb->z_os)) {
4130 /*
4131 * If userused is not enabled, it may be because the
4132 * objset needs to be closed & reopened (to grow the
4133 * objset_phys_t). Suspend/resume the fs will do that.
4134 */
4135 error = zfs_suspend_fs(zsb);
4136 if (error == 0)
4137 error = zfs_resume_fs(zsb, zc->zc_name);
4138 }
4139 if (error == 0)
4140 error = dmu_objset_userspace_upgrade(zsb->z_os);
4141 deactivate_super(zsb->z_sb);
4142 } else {
4143 /* XXX kind of reading contents without owning */
4144 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
4145 if (error)
4146 return (error);
4147
4148 error = dmu_objset_userspace_upgrade(os);
4149 dmu_objset_rele(os, FTAG);
4150 }
4151
4152 return (error);
4153 }
4154
4155 static int
4156 zfs_ioc_share(zfs_cmd_t *zc)
4157 {
4158 return (ENOSYS);
4159 }
4160
4161 ace_t full_access[] = {
4162 {(uid_t)-1, ACE_ALL_PERMS, ACE_EVERYONE, 0}
4163 };
4164
4165 /*
4166 * inputs:
4167 * zc_name name of containing filesystem
4168 * zc_obj object # beyond which we want next in-use object #
4169 *
4170 * outputs:
4171 * zc_obj next in-use object #
4172 */
4173 static int
4174 zfs_ioc_next_obj(zfs_cmd_t *zc)
4175 {
4176 objset_t *os = NULL;
4177 int error;
4178
4179 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
4180 if (error)
4181 return (error);
4182
4183 error = dmu_object_next(os, &zc->zc_obj, B_FALSE,
4184 os->os_dsl_dataset->ds_phys->ds_prev_snap_txg);
4185
4186 dmu_objset_rele(os, FTAG);
4187 return (error);
4188 }
4189
4190 /*
4191 * inputs:
4192 * zc_name name of filesystem
4193 * zc_value prefix name for snapshot
4194 * zc_cleanup_fd cleanup-on-exit file descriptor for calling process
4195 *
4196 * outputs:
4197 */
4198 static int
4199 zfs_ioc_tmp_snapshot(zfs_cmd_t *zc)
4200 {
4201 char *snap_name;
4202 int error;
4203
4204 snap_name = kmem_asprintf("%s-%016llx", zc->zc_value,
4205 (u_longlong_t)ddi_get_lbolt64());
4206
4207 if (strlen(snap_name) >= MAXNAMELEN) {
4208 strfree(snap_name);
4209 return (E2BIG);
4210 }
4211
4212 error = dmu_objset_snapshot(zc->zc_name, snap_name, snap_name,
4213 NULL, B_FALSE, B_TRUE, zc->zc_cleanup_fd);
4214 if (error != 0) {
4215 strfree(snap_name);
4216 return (error);
4217 }
4218
4219 (void) strcpy(zc->zc_value, snap_name);
4220 strfree(snap_name);
4221 return (0);
4222 }
4223
4224 /*
4225 * inputs:
4226 * zc_name name of "to" snapshot
4227 * zc_value name of "from" snapshot
4228 * zc_cookie file descriptor to write diff data on
4229 *
4230 * outputs:
4231 * dmu_diff_record_t's to the file descriptor
4232 */
4233 static int
4234 zfs_ioc_diff(zfs_cmd_t *zc)
4235 {
4236 objset_t *fromsnap;
4237 objset_t *tosnap;
4238 file_t *fp;
4239 offset_t off;
4240 int error;
4241
4242 error = dmu_objset_hold(zc->zc_name, FTAG, &tosnap);
4243 if (error)
4244 return (error);
4245
4246 error = dmu_objset_hold(zc->zc_value, FTAG, &fromsnap);
4247 if (error) {
4248 dmu_objset_rele(tosnap, FTAG);
4249 return (error);
4250 }
4251
4252 fp = getf(zc->zc_cookie);
4253 if (fp == NULL) {
4254 dmu_objset_rele(fromsnap, FTAG);
4255 dmu_objset_rele(tosnap, FTAG);
4256 return (EBADF);
4257 }
4258
4259 off = fp->f_offset;
4260
4261 error = dmu_diff(tosnap, fromsnap, fp->f_vnode, &off);
4262
4263 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
4264 fp->f_offset = off;
4265 releasef(zc->zc_cookie);
4266
4267 dmu_objset_rele(fromsnap, FTAG);
4268 dmu_objset_rele(tosnap, FTAG);
4269 return (error);
4270 }
4271
4272 /*
4273 * Remove all ACL files in shares dir
4274 */
4275 #ifdef HAVE_SMB_SHARE
4276 static int
4277 zfs_smb_acl_purge(znode_t *dzp)
4278 {
4279 zap_cursor_t zc;
4280 zap_attribute_t zap;
4281 zfs_sb_t *zsb = ZTOZSB(dzp);
4282 int error;
4283
4284 for (zap_cursor_init(&zc, zsb->z_os, dzp->z_id);
4285 (error = zap_cursor_retrieve(&zc, &zap)) == 0;
4286 zap_cursor_advance(&zc)) {
4287 if ((error = VOP_REMOVE(ZTOV(dzp), zap.za_name, kcred,
4288 NULL, 0)) != 0)
4289 break;
4290 }
4291 zap_cursor_fini(&zc);
4292 return (error);
4293 }
4294 #endif /* HAVE_SMB_SHARE */
4295
4296 static int
4297 zfs_ioc_smb_acl(zfs_cmd_t *zc)
4298 {
4299 #ifdef HAVE_SMB_SHARE
4300 vnode_t *vp;
4301 znode_t *dzp;
4302 vnode_t *resourcevp = NULL;
4303 znode_t *sharedir;
4304 zfs_sb_t *zsb;
4305 nvlist_t *nvlist;
4306 char *src, *target;
4307 vattr_t vattr;
4308 vsecattr_t vsec;
4309 int error = 0;
4310
4311 if ((error = lookupname(zc->zc_value, UIO_SYSSPACE,
4312 NO_FOLLOW, NULL, &vp)) != 0)
4313 return (error);
4314
4315 /* Now make sure mntpnt and dataset are ZFS */
4316
4317 if (vp->v_vfsp->vfs_fstype != zfsfstype ||
4318 (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource),
4319 zc->zc_name) != 0)) {
4320 VN_RELE(vp);
4321 return (EINVAL);
4322 }
4323
4324 dzp = VTOZ(vp);
4325 zsb = ZTOZSB(dzp);
4326 ZFS_ENTER(zsb);
4327
4328 /*
4329 * Create share dir if its missing.
4330 */
4331 mutex_enter(&zsb->z_lock);
4332 if (zsb->z_shares_dir == 0) {
4333 dmu_tx_t *tx;
4334
4335 tx = dmu_tx_create(zsb->z_os);
4336 dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, TRUE,
4337 ZFS_SHARES_DIR);
4338 dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL);
4339 error = dmu_tx_assign(tx, TXG_WAIT);
4340 if (error) {
4341 dmu_tx_abort(tx);
4342 } else {
4343 error = zfs_create_share_dir(zsb, tx);
4344 dmu_tx_commit(tx);
4345 }
4346 if (error) {
4347 mutex_exit(&zsb->z_lock);
4348 VN_RELE(vp);
4349 ZFS_EXIT(zsb);
4350 return (error);
4351 }
4352 }
4353 mutex_exit(&zsb->z_lock);
4354
4355 ASSERT(zsb->z_shares_dir);
4356 if ((error = zfs_zget(zsb, zsb->z_shares_dir, &sharedir)) != 0) {
4357 VN_RELE(vp);
4358 ZFS_EXIT(zsb);
4359 return (error);
4360 }
4361
4362 switch (zc->zc_cookie) {
4363 case ZFS_SMB_ACL_ADD:
4364 vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE;
4365 vattr.va_mode = S_IFREG|0777;
4366 vattr.va_uid = 0;
4367 vattr.va_gid = 0;
4368
4369 vsec.vsa_mask = VSA_ACE;
4370 vsec.vsa_aclentp = &full_access;
4371 vsec.vsa_aclentsz = sizeof (full_access);
4372 vsec.vsa_aclcnt = 1;
4373
4374 error = VOP_CREATE(ZTOV(sharedir), zc->zc_string,
4375 &vattr, EXCL, 0, &resourcevp, kcred, 0, NULL, &vsec);
4376 if (resourcevp)
4377 VN_RELE(resourcevp);
4378 break;
4379
4380 case ZFS_SMB_ACL_REMOVE:
4381 error = VOP_REMOVE(ZTOV(sharedir), zc->zc_string, kcred,
4382 NULL, 0);
4383 break;
4384
4385 case ZFS_SMB_ACL_RENAME:
4386 if ((error = get_nvlist(zc->zc_nvlist_src,
4387 zc->zc_nvlist_src_size, zc->zc_iflags, &nvlist)) != 0) {
4388 VN_RELE(vp);
4389 ZFS_EXIT(zsb);
4390 return (error);
4391 }
4392 if (nvlist_lookup_string(nvlist, ZFS_SMB_ACL_SRC, &src) ||
4393 nvlist_lookup_string(nvlist, ZFS_SMB_ACL_TARGET,
4394 &target)) {
4395 VN_RELE(vp);
4396 VN_RELE(ZTOV(sharedir));
4397 ZFS_EXIT(zsb);
4398 nvlist_free(nvlist);
4399 return (error);
4400 }
4401 error = VOP_RENAME(ZTOV(sharedir), src, ZTOV(sharedir), target,
4402 kcred, NULL, 0);
4403 nvlist_free(nvlist);
4404 break;
4405
4406 case ZFS_SMB_ACL_PURGE:
4407 error = zfs_smb_acl_purge(sharedir);
4408 break;
4409
4410 default:
4411 error = EINVAL;
4412 break;
4413 }
4414
4415 VN_RELE(vp);
4416 VN_RELE(ZTOV(sharedir));
4417
4418 ZFS_EXIT(zsb);
4419
4420 return (error);
4421 #else
4422 return (ENOTSUP);
4423 #endif /* HAVE_SMB_SHARE */
4424 }
4425
4426 /*
4427 * inputs:
4428 * zc_name name of filesystem
4429 * zc_value short name of snap
4430 * zc_string user-supplied tag for this hold
4431 * zc_cookie recursive flag
4432 * zc_temphold set if hold is temporary
4433 * zc_cleanup_fd cleanup-on-exit file descriptor for calling process
4434 * zc_sendobj if non-zero, the objid for zc_name@zc_value
4435 * zc_createtxg if zc_sendobj is non-zero, snap must have zc_createtxg
4436 *
4437 * outputs: none
4438 */
4439 static int
4440 zfs_ioc_hold(zfs_cmd_t *zc)
4441 {
4442 boolean_t recursive = zc->zc_cookie;
4443 spa_t *spa;
4444 dsl_pool_t *dp;
4445 dsl_dataset_t *ds;
4446 int error;
4447 minor_t minor = 0;
4448
4449 if (snapshot_namecheck(zc->zc_value, NULL, NULL) != 0)
4450 return (EINVAL);
4451
4452 if (zc->zc_sendobj == 0) {
4453 return (dsl_dataset_user_hold(zc->zc_name, zc->zc_value,
4454 zc->zc_string, recursive, zc->zc_temphold,
4455 zc->zc_cleanup_fd));
4456 }
4457
4458 if (recursive)
4459 return (EINVAL);
4460
4461 error = spa_open(zc->zc_name, &spa, FTAG);
4462 if (error)
4463 return (error);
4464
4465 dp = spa_get_dsl(spa);
4466 rw_enter(&dp->dp_config_rwlock, RW_READER);
4467 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &ds);
4468 rw_exit(&dp->dp_config_rwlock);
4469 spa_close(spa, FTAG);
4470 if (error)
4471 return (error);
4472
4473 /*
4474 * Until we have a hold on this snapshot, it's possible that
4475 * zc_sendobj could've been destroyed and reused as part
4476 * of a later txg. Make sure we're looking at the right object.
4477 */
4478 if (zc->zc_createtxg != ds->ds_phys->ds_creation_txg) {
4479 dsl_dataset_rele(ds, FTAG);
4480 return (ENOENT);
4481 }
4482
4483 if (zc->zc_cleanup_fd != -1 && zc->zc_temphold) {
4484 error = zfs_onexit_fd_hold(zc->zc_cleanup_fd, &minor);
4485 if (error) {
4486 dsl_dataset_rele(ds, FTAG);
4487 return (error);
4488 }
4489 }
4490
4491 error = dsl_dataset_user_hold_for_send(ds, zc->zc_string,
4492 zc->zc_temphold);
4493 if (minor != 0) {
4494 if (error == 0) {
4495 dsl_register_onexit_hold_cleanup(ds, zc->zc_string,
4496 minor);
4497 }
4498 zfs_onexit_fd_rele(zc->zc_cleanup_fd);
4499 }
4500 dsl_dataset_rele(ds, FTAG);
4501
4502 return (error);
4503 }
4504
4505 /*
4506 * inputs:
4507 * zc_name name of dataset from which we're releasing a user hold
4508 * zc_value short name of snap
4509 * zc_string user-supplied tag for this hold
4510 * zc_cookie recursive flag
4511 *
4512 * outputs: none
4513 */
4514 static int
4515 zfs_ioc_release(zfs_cmd_t *zc)
4516 {
4517 boolean_t recursive = zc->zc_cookie;
4518
4519 if (snapshot_namecheck(zc->zc_value, NULL, NULL) != 0)
4520 return (EINVAL);
4521
4522 return (dsl_dataset_user_release(zc->zc_name, zc->zc_value,
4523 zc->zc_string, recursive));
4524 }
4525
4526 /*
4527 * inputs:
4528 * zc_name name of filesystem
4529 *
4530 * outputs:
4531 * zc_nvlist_src{_size} nvlist of snapshot holds
4532 */
4533 static int
4534 zfs_ioc_get_holds(zfs_cmd_t *zc)
4535 {
4536 nvlist_t *nvp;
4537 int error;
4538
4539 if ((error = dsl_dataset_get_holds(zc->zc_name, &nvp)) == 0) {
4540 error = put_nvlist(zc, nvp);
4541 nvlist_free(nvp);
4542 }
4543
4544 return (error);
4545 }
4546
4547 /*
4548 * inputs:
4549 * zc_guid flags (ZEVENT_NONBLOCK)
4550 *
4551 * outputs:
4552 * zc_nvlist_dst next nvlist event
4553 * zc_cookie dropped events since last get
4554 * zc_cleanup_fd cleanup-on-exit file descriptor
4555 */
4556 static int
4557 zfs_ioc_events_next(zfs_cmd_t *zc)
4558 {
4559 zfs_zevent_t *ze;
4560 nvlist_t *event = NULL;
4561 minor_t minor;
4562 uint64_t dropped = 0;
4563 int error;
4564
4565 error = zfs_zevent_fd_hold(zc->zc_cleanup_fd, &minor, &ze);
4566 if (error != 0)
4567 return (error);
4568
4569 do {
4570 error = zfs_zevent_next(ze, &event,
4571 &zc->zc_nvlist_dst_size, &dropped);
4572 if (event != NULL) {
4573 zc->zc_cookie = dropped;
4574 error = put_nvlist(zc, event);
4575 nvlist_free(event);
4576 }
4577
4578 if (zc->zc_guid & ZEVENT_NONBLOCK)
4579 break;
4580
4581 if ((error == 0) || (error != ENOENT))
4582 break;
4583
4584 error = zfs_zevent_wait(ze);
4585 if (error)
4586 break;
4587 } while (1);
4588
4589 zfs_zevent_fd_rele(zc->zc_cleanup_fd);
4590
4591 return (error);
4592 }
4593
4594 /*
4595 * outputs:
4596 * zc_cookie cleared events count
4597 */
4598 static int
4599 zfs_ioc_events_clear(zfs_cmd_t *zc)
4600 {
4601 int count;
4602
4603 zfs_zevent_drain_all(&count);
4604 zc->zc_cookie = count;
4605
4606 return 0;
4607 }
4608
4609 /*
4610 * pool create, destroy, and export don't log the history as part of
4611 * zfsdev_ioctl, but rather zfs_ioc_pool_create, and zfs_ioc_pool_export
4612 * do the logging of those commands.
4613 */
4614 static zfs_ioc_vec_t zfs_ioc_vec[] = {
4615 { zfs_ioc_pool_create, zfs_secpolicy_config, POOL_NAME, B_FALSE,
4616 POOL_CHECK_NONE },
4617 { zfs_ioc_pool_destroy, zfs_secpolicy_config, POOL_NAME, B_FALSE,
4618 POOL_CHECK_NONE },
4619 { zfs_ioc_pool_import, zfs_secpolicy_config, POOL_NAME, B_TRUE,
4620 POOL_CHECK_NONE },
4621 { zfs_ioc_pool_export, zfs_secpolicy_config, POOL_NAME, B_FALSE,
4622 POOL_CHECK_NONE },
4623 { zfs_ioc_pool_configs, zfs_secpolicy_none, NO_NAME, B_FALSE,
4624 POOL_CHECK_NONE },
4625 { zfs_ioc_pool_stats, zfs_secpolicy_read, POOL_NAME, B_FALSE,
4626 POOL_CHECK_NONE },
4627 { zfs_ioc_pool_tryimport, zfs_secpolicy_config, NO_NAME, B_FALSE,
4628 POOL_CHECK_NONE },
4629 { zfs_ioc_pool_scan, zfs_secpolicy_config, POOL_NAME, B_TRUE,
4630 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4631 { zfs_ioc_pool_freeze, zfs_secpolicy_config, NO_NAME, B_FALSE,
4632 POOL_CHECK_READONLY },
4633 { zfs_ioc_pool_upgrade, zfs_secpolicy_config, POOL_NAME, B_TRUE,
4634 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4635 { zfs_ioc_pool_get_history, zfs_secpolicy_config, POOL_NAME, B_FALSE,
4636 POOL_CHECK_NONE },
4637 { zfs_ioc_vdev_add, zfs_secpolicy_config, POOL_NAME, B_TRUE,
4638 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4639 { zfs_ioc_vdev_remove, zfs_secpolicy_config, POOL_NAME, B_TRUE,
4640 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4641 { zfs_ioc_vdev_set_state, zfs_secpolicy_config, POOL_NAME, B_TRUE,
4642 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4643 { zfs_ioc_vdev_attach, zfs_secpolicy_config, POOL_NAME, B_TRUE,
4644 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4645 { zfs_ioc_vdev_detach, zfs_secpolicy_config, POOL_NAME, B_TRUE,
4646 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4647 { zfs_ioc_vdev_setpath, zfs_secpolicy_config, POOL_NAME, B_FALSE,
4648 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4649 { zfs_ioc_vdev_setfru, zfs_secpolicy_config, POOL_NAME, B_FALSE,
4650 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4651 { zfs_ioc_objset_stats, zfs_secpolicy_read, DATASET_NAME, B_FALSE,
4652 POOL_CHECK_SUSPENDED },
4653 { zfs_ioc_objset_zplprops, zfs_secpolicy_read, DATASET_NAME, B_FALSE,
4654 POOL_CHECK_NONE },
4655 { zfs_ioc_dataset_list_next, zfs_secpolicy_read, DATASET_NAME, B_FALSE,
4656 POOL_CHECK_SUSPENDED },
4657 { zfs_ioc_snapshot_list_next, zfs_secpolicy_read, DATASET_NAME, B_FALSE,
4658 POOL_CHECK_SUSPENDED },
4659 { zfs_ioc_set_prop, zfs_secpolicy_none, DATASET_NAME, B_TRUE,
4660 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4661 { zfs_ioc_create_minor, zfs_secpolicy_config, DATASET_NAME, B_FALSE,
4662 POOL_CHECK_NONE },
4663 { zfs_ioc_remove_minor, zfs_secpolicy_config, DATASET_NAME, B_FALSE,
4664 POOL_CHECK_NONE },
4665 { zfs_ioc_create, zfs_secpolicy_create, DATASET_NAME, B_TRUE,
4666 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4667 { zfs_ioc_destroy, zfs_secpolicy_destroy, DATASET_NAME, B_TRUE,
4668 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4669 { zfs_ioc_rollback, zfs_secpolicy_rollback, DATASET_NAME, B_TRUE,
4670 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4671 { zfs_ioc_rename, zfs_secpolicy_rename, DATASET_NAME, B_TRUE,
4672 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4673 { zfs_ioc_recv, zfs_secpolicy_receive, DATASET_NAME, B_TRUE,
4674 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4675 { zfs_ioc_send, zfs_secpolicy_send, DATASET_NAME, B_TRUE,
4676 POOL_CHECK_NONE },
4677 { zfs_ioc_inject_fault, zfs_secpolicy_inject, NO_NAME, B_FALSE,
4678 POOL_CHECK_NONE },
4679 { zfs_ioc_clear_fault, zfs_secpolicy_inject, NO_NAME, B_FALSE,
4680 POOL_CHECK_NONE },
4681 { zfs_ioc_inject_list_next, zfs_secpolicy_inject, NO_NAME, B_FALSE,
4682 POOL_CHECK_NONE },
4683 { zfs_ioc_error_log, zfs_secpolicy_inject, POOL_NAME, B_FALSE,
4684 POOL_CHECK_NONE },
4685 { zfs_ioc_clear, zfs_secpolicy_config, POOL_NAME, B_TRUE,
4686 POOL_CHECK_NONE },
4687 { zfs_ioc_promote, zfs_secpolicy_promote, DATASET_NAME, B_TRUE,
4688 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4689 { zfs_ioc_destroy_snaps, zfs_secpolicy_destroy_snaps, DATASET_NAME,
4690 B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4691 { zfs_ioc_snapshot, zfs_secpolicy_snapshot, DATASET_NAME, B_TRUE,
4692 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4693 { zfs_ioc_dsobj_to_dsname, zfs_secpolicy_diff, POOL_NAME, B_FALSE,
4694 POOL_CHECK_NONE },
4695 { zfs_ioc_obj_to_path, zfs_secpolicy_diff, DATASET_NAME, B_FALSE,
4696 POOL_CHECK_SUSPENDED },
4697 { zfs_ioc_pool_set_props, zfs_secpolicy_config, POOL_NAME, B_TRUE,
4698 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4699 { zfs_ioc_pool_get_props, zfs_secpolicy_read, POOL_NAME, B_FALSE,
4700 POOL_CHECK_NONE },
4701 { zfs_ioc_set_fsacl, zfs_secpolicy_fsacl, DATASET_NAME, B_TRUE,
4702 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4703 { zfs_ioc_get_fsacl, zfs_secpolicy_read, DATASET_NAME, B_FALSE,
4704 POOL_CHECK_NONE },
4705 { zfs_ioc_share, zfs_secpolicy_share, DATASET_NAME, B_FALSE,
4706 POOL_CHECK_NONE },
4707 { zfs_ioc_inherit_prop, zfs_secpolicy_inherit, DATASET_NAME, B_TRUE,
4708 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4709 { zfs_ioc_smb_acl, zfs_secpolicy_smb_acl, DATASET_NAME, B_FALSE,
4710 POOL_CHECK_NONE },
4711 { zfs_ioc_userspace_one, zfs_secpolicy_userspace_one, DATASET_NAME,
4712 B_FALSE, POOL_CHECK_NONE },
4713 { zfs_ioc_userspace_many, zfs_secpolicy_userspace_many, DATASET_NAME,
4714 B_FALSE, POOL_CHECK_NONE },
4715 { zfs_ioc_userspace_upgrade, zfs_secpolicy_userspace_upgrade,
4716 DATASET_NAME, B_FALSE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4717 { zfs_ioc_hold, zfs_secpolicy_hold, DATASET_NAME, B_TRUE,
4718 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4719 { zfs_ioc_release, zfs_secpolicy_release, DATASET_NAME, B_TRUE,
4720 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4721 { zfs_ioc_get_holds, zfs_secpolicy_read, DATASET_NAME, B_FALSE,
4722 POOL_CHECK_SUSPENDED },
4723 { zfs_ioc_objset_recvd_props, zfs_secpolicy_read, DATASET_NAME, B_FALSE,
4724 POOL_CHECK_NONE },
4725 { zfs_ioc_vdev_split, zfs_secpolicy_config, POOL_NAME, B_TRUE,
4726 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4727 { zfs_ioc_next_obj, zfs_secpolicy_read, DATASET_NAME, B_FALSE,
4728 POOL_CHECK_NONE },
4729 { zfs_ioc_diff, zfs_secpolicy_diff, DATASET_NAME, B_FALSE,
4730 POOL_CHECK_NONE },
4731 { zfs_ioc_tmp_snapshot, zfs_secpolicy_tmp_snapshot, DATASET_NAME,
4732 B_FALSE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY },
4733 { zfs_ioc_obj_to_stats, zfs_secpolicy_diff, DATASET_NAME, B_FALSE,
4734 POOL_CHECK_SUSPENDED },
4735 { zfs_ioc_events_next, zfs_secpolicy_config, NO_NAME, B_FALSE,
4736 POOL_CHECK_NONE },
4737 { zfs_ioc_events_clear, zfs_secpolicy_config, NO_NAME, B_FALSE,
4738 POOL_CHECK_NONE },
4739 };
4740
4741 int
4742 pool_status_check(const char *name, zfs_ioc_namecheck_t type,
4743 zfs_ioc_poolcheck_t check)
4744 {
4745 spa_t *spa;
4746 int error;
4747
4748 ASSERT(type == POOL_NAME || type == DATASET_NAME);
4749
4750 if (check & POOL_CHECK_NONE)
4751 return (0);
4752
4753 error = spa_open(name, &spa, FTAG);
4754 if (error == 0) {
4755 if ((check & POOL_CHECK_SUSPENDED) && spa_suspended(spa))
4756 error = EAGAIN;
4757 else if ((check & POOL_CHECK_READONLY) && !spa_writeable(spa))
4758 error = EROFS;
4759 spa_close(spa, FTAG);
4760 }
4761 return (error);
4762 }
4763
4764 static void *
4765 zfsdev_get_state_impl(minor_t minor, enum zfsdev_state_type which)
4766 {
4767 zfsdev_state_t *zs;
4768
4769 ASSERT(MUTEX_HELD(&zfsdev_state_lock));
4770
4771 for (zs = list_head(&zfsdev_state_list); zs != NULL;
4772 zs = list_next(&zfsdev_state_list, zs)) {
4773 if (zs->zs_minor == minor) {
4774 switch (which) {
4775 case ZST_ONEXIT: return (zs->zs_onexit);
4776 case ZST_ZEVENT: return (zs->zs_zevent);
4777 case ZST_ALL: return (zs);
4778 }
4779 }
4780 }
4781
4782 return NULL;
4783 }
4784
4785 void *
4786 zfsdev_get_state(minor_t minor, enum zfsdev_state_type which)
4787 {
4788 void *ptr;
4789
4790 mutex_enter(&zfsdev_state_lock);
4791 ptr = zfsdev_get_state_impl(minor, which);
4792 mutex_exit(&zfsdev_state_lock);
4793
4794 return ptr;
4795 }
4796
4797 minor_t
4798 zfsdev_getminor(struct file *filp)
4799 {
4800 ASSERT(filp != NULL);
4801 ASSERT(filp->private_data != NULL);
4802
4803 return (((zfsdev_state_t *)filp->private_data)->zs_minor);
4804 }
4805
4806 /*
4807 * Find a free minor number. The zfsdev_state_list is expected to
4808 * be short since it is only a list of currently open file handles.
4809 */
4810 minor_t
4811 zfsdev_minor_alloc(void)
4812 {
4813 static minor_t last_minor = 0;
4814 minor_t m;
4815
4816 ASSERT(MUTEX_HELD(&zfsdev_state_lock));
4817
4818 for (m = last_minor + 1; m != last_minor; m++) {
4819 if (m > ZFSDEV_MAX_MINOR)
4820 m = 1;
4821 if (zfsdev_get_state_impl(m, ZST_ALL) == NULL) {
4822 last_minor = m;
4823 return (m);
4824 }
4825 }
4826
4827 return (0);
4828 }
4829
4830 static int
4831 zfsdev_state_init(struct file *filp)
4832 {
4833 zfsdev_state_t *zs;
4834 minor_t minor;
4835
4836 ASSERT(MUTEX_HELD(&zfsdev_state_lock));
4837
4838 minor = zfsdev_minor_alloc();
4839 if (minor == 0)
4840 return (ENXIO);
4841
4842 zs = kmem_zalloc( sizeof(zfsdev_state_t), KM_SLEEP);
4843 if (zs == NULL)
4844 return (ENOMEM);
4845
4846 zs->zs_file = filp;
4847 zs->zs_minor = minor;
4848 filp->private_data = zs;
4849
4850 zfs_onexit_init((zfs_onexit_t **)&zs->zs_onexit);
4851 zfs_zevent_init((zfs_zevent_t **)&zs->zs_zevent);
4852
4853 list_insert_tail(&zfsdev_state_list, zs);
4854
4855 return (0);
4856 }
4857
4858 static int
4859 zfsdev_state_destroy(struct file *filp)
4860 {
4861 zfsdev_state_t *zs;
4862
4863 ASSERT(MUTEX_HELD(&zfsdev_state_lock));
4864 ASSERT(filp->private_data != NULL);
4865
4866 zs = filp->private_data;
4867 zfs_onexit_destroy(zs->zs_onexit);
4868 zfs_zevent_destroy(zs->zs_zevent);
4869
4870 list_remove(&zfsdev_state_list, zs);
4871 kmem_free(zs, sizeof(zfsdev_state_t));
4872
4873 return 0;
4874 }
4875
4876 static int
4877 zfsdev_open(struct inode *ino, struct file *filp)
4878 {
4879 int error;
4880
4881 mutex_enter(&zfsdev_state_lock);
4882 error = zfsdev_state_init(filp);
4883 mutex_exit(&zfsdev_state_lock);
4884
4885 return (-error);
4886 }
4887
4888 static int
4889 zfsdev_release(struct inode *ino, struct file *filp)
4890 {
4891 int error;
4892
4893 mutex_enter(&zfsdev_state_lock);
4894 error = zfsdev_state_destroy(filp);
4895 mutex_exit(&zfsdev_state_lock);
4896
4897 return (-error);
4898 }
4899
4900 static long
4901 zfsdev_ioctl(struct file *filp, unsigned cmd, unsigned long arg)
4902 {
4903 zfs_cmd_t *zc;
4904 uint_t vec;
4905 int error, rc, flag = 0;
4906
4907 vec = cmd - ZFS_IOC;
4908 if (vec >= sizeof (zfs_ioc_vec) / sizeof (zfs_ioc_vec[0]))
4909 return (-EINVAL);
4910
4911 zc = kmem_zalloc(sizeof (zfs_cmd_t), KM_SLEEP | KM_NODEBUG);
4912
4913 error = ddi_copyin((void *)arg, zc, sizeof (zfs_cmd_t), flag);
4914 if (error != 0)
4915 error = EFAULT;
4916
4917 if ((error == 0) && !(flag & FKIOCTL))
4918 error = zfs_ioc_vec[vec].zvec_secpolicy(zc, CRED());
4919
4920 /*
4921 * Ensure that all pool/dataset names are valid before we pass down to
4922 * the lower layers.
4923 */
4924 if (error == 0) {
4925 zc->zc_name[sizeof (zc->zc_name) - 1] = '\0';
4926 zc->zc_iflags = flag & FKIOCTL;
4927 switch (zfs_ioc_vec[vec].zvec_namecheck) {
4928 case POOL_NAME:
4929 if (pool_namecheck(zc->zc_name, NULL, NULL) != 0)
4930 error = EINVAL;
4931 error = pool_status_check(zc->zc_name,
4932 zfs_ioc_vec[vec].zvec_namecheck,
4933 zfs_ioc_vec[vec].zvec_pool_check);
4934 break;
4935
4936 case DATASET_NAME:
4937 if (dataset_namecheck(zc->zc_name, NULL, NULL) != 0)
4938 error = EINVAL;
4939 error = pool_status_check(zc->zc_name,
4940 zfs_ioc_vec[vec].zvec_namecheck,
4941 zfs_ioc_vec[vec].zvec_pool_check);
4942 break;
4943
4944 case NO_NAME:
4945 break;
4946 }
4947 }
4948
4949 if (error == 0)
4950 error = zfs_ioc_vec[vec].zvec_func(zc);
4951
4952 rc = ddi_copyout(zc, (void *)arg, sizeof (zfs_cmd_t), flag);
4953 if (error == 0) {
4954 if (rc != 0)
4955 error = EFAULT;
4956 if (zfs_ioc_vec[vec].zvec_his_log)
4957 zfs_log_history(zc);
4958 }
4959
4960 kmem_free(zc, sizeof (zfs_cmd_t));
4961 return (-error);
4962 }
4963
4964 #ifdef CONFIG_COMPAT
4965 static long
4966 zfsdev_compat_ioctl(struct file *filp, unsigned cmd, unsigned long arg)
4967 {
4968 return zfsdev_ioctl(filp, cmd, arg);
4969 }
4970 #else
4971 #define zfsdev_compat_ioctl NULL
4972 #endif
4973
4974 static const struct file_operations zfsdev_fops = {
4975 .open = zfsdev_open,
4976 .release = zfsdev_release,
4977 .unlocked_ioctl = zfsdev_ioctl,
4978 .compat_ioctl = zfsdev_compat_ioctl,
4979 .owner = THIS_MODULE,
4980 };
4981
4982 static struct miscdevice zfs_misc = {
4983 .minor = MISC_DYNAMIC_MINOR,
4984 .name = ZFS_DRIVER,
4985 .fops = &zfsdev_fops,
4986 };
4987
4988 static int
4989 zfs_attach(void)
4990 {
4991 int error;
4992
4993 mutex_init(&zfsdev_state_lock, NULL, MUTEX_DEFAULT, NULL);
4994 list_create(&zfsdev_state_list, sizeof (zfsdev_state_t),
4995 offsetof(zfsdev_state_t, zs_next));
4996
4997 error = misc_register(&zfs_misc);
4998 if (error) {
4999 printk(KERN_INFO "ZFS: misc_register() failed %d\n", error);
5000 return (error);
5001 }
5002
5003 return (0);
5004 }
5005
5006 static void
5007 zfs_detach(void)
5008 {
5009 int error;
5010
5011 error = misc_deregister(&zfs_misc);
5012 if (error)
5013 printk(KERN_INFO "ZFS: misc_deregister() failed %d\n", error);
5014
5015 mutex_destroy(&zfsdev_state_lock);
5016 list_destroy(&zfsdev_state_list);
5017 }
5018
5019 uint_t zfs_fsyncer_key;
5020 extern uint_t rrw_tsd_key;
5021
5022 #ifdef DEBUG
5023 #define ZFS_DEBUG_STR " (DEBUG mode)"
5024 #else
5025 #define ZFS_DEBUG_STR ""
5026 #endif
5027
5028 int
5029 _init(void)
5030 {
5031 int error;
5032
5033 spa_init(FREAD | FWRITE);
5034 zfs_init();
5035
5036 if ((error = zvol_init()) != 0)
5037 goto out1;
5038
5039 if ((error = zfs_attach()) != 0)
5040 goto out2;
5041
5042 tsd_create(&zfs_fsyncer_key, NULL);
5043 tsd_create(&rrw_tsd_key, NULL);
5044
5045 printk(KERN_NOTICE "ZFS: Loaded module v%s%s, "
5046 "ZFS pool version %s, ZFS filesystem version %s\n",
5047 ZFS_META_VERSION, ZFS_DEBUG_STR,
5048 SPA_VERSION_STRING, ZPL_VERSION_STRING);
5049
5050 return (0);
5051
5052 out2:
5053 (void) zvol_fini();
5054 out1:
5055 zfs_fini();
5056 spa_fini();
5057 printk(KERN_NOTICE "ZFS: Failed to Load ZFS Filesystem v%s%s"
5058 ", rc = %d\n", ZFS_META_VERSION, ZFS_DEBUG_STR, error);
5059
5060 return (error);
5061 }
5062
5063 int
5064 _fini(void)
5065 {
5066 zfs_detach();
5067 zvol_fini();
5068 zfs_fini();
5069 spa_fini();
5070
5071 tsd_destroy(&zfs_fsyncer_key);
5072 tsd_destroy(&rrw_tsd_key);
5073
5074 printk(KERN_NOTICE "ZFS: Unloaded module v%s%s\n",
5075 ZFS_META_VERSION, ZFS_DEBUG_STR);
5076
5077 return (0);
5078 }
5079
5080 #ifdef HAVE_SPL
5081 spl_module_init(_init);
5082 spl_module_exit(_fini);
5083
5084 MODULE_DESCRIPTION("ZFS");
5085 MODULE_AUTHOR(ZFS_META_AUTHOR);
5086 MODULE_LICENSE(ZFS_META_LICENSE);
5087 #endif /* HAVE_SPL */
ZFS On Linux mirror for PVE