1 /*
2 * CDDL HEADER START
3 *
4 * The contents of this file are subject to the terms of the
5 * Common Development and Distribution License (the "License").
6 * You may not use this file except in compliance with the License.
7 *
8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9 * or http://www.opensolaris.org/os/licensing.
10 * See the License for the specific language governing permissions
11 * and limitations under the License.
12 *
13 * When distributing Covered Code, include this CDDL HEADER in each
14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15 * If applicable, add the following below this CDDL HEADER, with the
16 * fields enclosed by brackets "[]" replaced with your own identifying
17 * information: Portions Copyright [yyyy] [name of copyright owner]
18 *
19 * CDDL HEADER END
20 */
21
22 /*
23 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
24 * Copyright (c) 2011-2012 Pawel Jakub Dawidek. All rights reserved.
25 * Portions Copyright 2011 Martin Matuska
26 * Copyright 2015, OmniTI Computer Consulting, Inc. All rights reserved.
27 * Copyright 2015 Nexenta Systems, Inc. All rights reserved.
28 * Copyright (c) 2014, 2016 Joyent, Inc. All rights reserved.
29 * Copyright (c) 2011, 2015 by Delphix. All rights reserved.
30 * Copyright (c) 2013 by Saso Kiselkov. All rights reserved.
31 * Copyright (c) 2013 Steven Hartland. All rights reserved.
32 * Copyright (c) 2014 Integros [integros.com]
33 */
34
35 /*
36 * ZFS ioctls.
37 *
38 * This file handles the ioctls to /dev/zfs, used for configuring ZFS storage
39 * pools and filesystems, e.g. with /sbin/zfs and /sbin/zpool.
40 *
41 * There are two ways that we handle ioctls: the legacy way where almost
42 * all of the logic is in the ioctl callback, and the new way where most
43 * of the marshalling is handled in the common entry point, zfsdev_ioctl().
44 *
45 * Non-legacy ioctls should be registered by calling
46 * zfs_ioctl_register() from zfs_ioctl_init(). The ioctl is invoked
47 * from userland by lzc_ioctl().
48 *
49 * The registration arguments are as follows:
50 *
51 * const char *name
52 * The name of the ioctl. This is used for history logging. If the
53 * ioctl returns successfully (the callback returns 0), and allow_log
54 * is true, then a history log entry will be recorded with the input &
55 * output nvlists. The log entry can be printed with "zpool history -i".
56 *
57 * zfs_ioc_t ioc
58 * The ioctl request number, which userland will pass to ioctl(2).
59 * The ioctl numbers can change from release to release, because
60 * the caller (libzfs) must be matched to the kernel.
61 *
62 * zfs_secpolicy_func_t *secpolicy
63 * This function will be called before the zfs_ioc_func_t, to
64 * determine if this operation is permitted. It should return EPERM
65 * on failure, and 0 on success. Checks include determining if the
66 * dataset is visible in this zone, and if the user has either all
67 * zfs privileges in the zone (SYS_MOUNT), or has been granted permission
68 * to do this operation on this dataset with "zfs allow".
69 *
70 * zfs_ioc_namecheck_t namecheck
71 * This specifies what to expect in the zfs_cmd_t:zc_name -- a pool
72 * name, a dataset name, or nothing. If the name is not well-formed,
73 * the ioctl will fail and the callback will not be called.
74 * Therefore, the callback can assume that the name is well-formed
75 * (e.g. is null-terminated, doesn't have more than one '@' character,
76 * doesn't have invalid characters).
77 *
78 * zfs_ioc_poolcheck_t pool_check
79 * This specifies requirements on the pool state. If the pool does
80 * not meet them (is suspended or is readonly), the ioctl will fail
81 * and the callback will not be called. If any checks are specified
82 * (i.e. it is not POOL_CHECK_NONE), namecheck must not be NO_NAME.
83 * Multiple checks can be or-ed together (e.g. POOL_CHECK_SUSPENDED |
84 * POOL_CHECK_READONLY).
85 *
86 * boolean_t smush_outnvlist
87 * If smush_outnvlist is true, then the output is presumed to be a
88 * list of errors, and it will be "smushed" down to fit into the
89 * caller's buffer, by removing some entries and replacing them with a
90 * single "N_MORE_ERRORS" entry indicating how many were removed. See
91 * nvlist_smush() for details. If smush_outnvlist is false, and the
92 * outnvlist does not fit into the userland-provided buffer, then the
93 * ioctl will fail with ENOMEM.
94 *
95 * zfs_ioc_func_t *func
96 * The callback function that will perform the operation.
97 *
98 * The callback should return 0 on success, or an error number on
99 * failure. If the function fails, the userland ioctl will return -1,
100 * and errno will be set to the callback's return value. The callback
101 * will be called with the following arguments:
102 *
103 * const char *name
104 * The name of the pool or dataset to operate on, from
105 * zfs_cmd_t:zc_name. The 'namecheck' argument specifies the
106 * expected type (pool, dataset, or none).
107 *
108 * nvlist_t *innvl
109 * The input nvlist, deserialized from zfs_cmd_t:zc_nvlist_src. Or
110 * NULL if no input nvlist was provided. Changes to this nvlist are
111 * ignored. If the input nvlist could not be deserialized, the
112 * ioctl will fail and the callback will not be called.
113 *
114 * nvlist_t *outnvl
115 * The output nvlist, initially empty. The callback can fill it in,
116 * and it will be returned to userland by serializing it into
117 * zfs_cmd_t:zc_nvlist_dst. If it is non-empty, and serialization
118 * fails (e.g. because the caller didn't supply a large enough
119 * buffer), then the overall ioctl will fail. See the
120 * 'smush_nvlist' argument above for additional behaviors.
121 *
122 * There are two typical uses of the output nvlist:
123 * - To return state, e.g. property values. In this case,
124 * smush_outnvlist should be false. If the buffer was not large
125 * enough, the caller will reallocate a larger buffer and try
126 * the ioctl again.
127 *
128 * - To return multiple errors from an ioctl which makes on-disk
129 * changes. In this case, smush_outnvlist should be true.
130 * Ioctls which make on-disk modifications should generally not
131 * use the outnvl if they succeed, because the caller can not
132 * distinguish between the operation failing, and
133 * deserialization failing.
134 */
135
136 #include <sys/types.h>
137 #include <sys/param.h>
138 #include <sys/errno.h>
139 #include <sys/uio.h>
140 #include <sys/buf.h>
141 #include <sys/modctl.h>
142 #include <sys/open.h>
143 #include <sys/file.h>
144 #include <sys/kmem.h>
145 #include <sys/conf.h>
146 #include <sys/cmn_err.h>
147 #include <sys/stat.h>
148 #include <sys/zfs_ioctl.h>
149 #include <sys/zfs_vfsops.h>
150 #include <sys/zfs_znode.h>
151 #include <sys/zap.h>
152 #include <sys/spa.h>
153 #include <sys/spa_impl.h>
154 #include <sys/vdev.h>
155 #include <sys/priv_impl.h>
156 #include <sys/dmu.h>
157 #include <sys/dsl_dir.h>
158 #include <sys/dsl_dataset.h>
159 #include <sys/dsl_prop.h>
160 #include <sys/dsl_deleg.h>
161 #include <sys/dmu_objset.h>
162 #include <sys/dmu_impl.h>
163 #include <sys/dmu_tx.h>
164 #include <sys/ddi.h>
165 #include <sys/sunddi.h>
166 #include <sys/sunldi.h>
167 #include <sys/policy.h>
168 #include <sys/zone.h>
169 #include <sys/nvpair.h>
170 #include <sys/pathname.h>
171 #include <sys/mount.h>
172 #include <sys/sdt.h>
173 #include <sys/fs/zfs.h>
174 #include <sys/zfs_ctldir.h>
175 #include <sys/zfs_dir.h>
176 #include <sys/zfs_onexit.h>
177 #include <sys/zvol.h>
178 #include <sys/dsl_scan.h>
179 #include <sharefs/share.h>
180 #include <sys/dmu_objset.h>
181 #include <sys/dmu_send.h>
182 #include <sys/dsl_destroy.h>
183 #include <sys/dsl_bookmark.h>
184 #include <sys/dsl_userhold.h>
185 #include <sys/zfeature.h>
186 #include <sys/zio_checksum.h>
187
188 #include "zfs_namecheck.h"
189 #include "zfs_prop.h"
190 #include "zfs_deleg.h"
191 #include "zfs_comutil.h"
192
193 extern struct modlfs zfs_modlfs;
194
195 extern void zfs_init(void);
196 extern void zfs_fini(void);
197
198 ldi_ident_t zfs_li = NULL;
199 dev_info_t *zfs_dip;
200
201 uint_t zfs_fsyncer_key;
202 extern uint_t rrw_tsd_key;
203 static uint_t zfs_allow_log_key;
204
205 typedef int zfs_ioc_legacy_func_t(zfs_cmd_t *);
206 typedef int zfs_ioc_func_t(const char *, nvlist_t *, nvlist_t *);
207 typedef int zfs_secpolicy_func_t(zfs_cmd_t *, nvlist_t *, cred_t *);
208
209 typedef enum {
210 NO_NAME,
211 POOL_NAME,
212 DATASET_NAME
213 } zfs_ioc_namecheck_t;
214
215 typedef enum {
216 POOL_CHECK_NONE = 1 << 0,
217 POOL_CHECK_SUSPENDED = 1 << 1,
218 POOL_CHECK_READONLY = 1 << 2,
219 } zfs_ioc_poolcheck_t;
220
221 typedef struct zfs_ioc_vec {
222 zfs_ioc_legacy_func_t *zvec_legacy_func;
223 zfs_ioc_func_t *zvec_func;
224 zfs_secpolicy_func_t *zvec_secpolicy;
225 zfs_ioc_namecheck_t zvec_namecheck;
226 boolean_t zvec_allow_log;
227 zfs_ioc_poolcheck_t zvec_pool_check;
228 boolean_t zvec_smush_outnvlist;
229 const char *zvec_name;
230 } zfs_ioc_vec_t;
231
232 /* This array is indexed by zfs_userquota_prop_t */
233 static const char *userquota_perms[] = {
234 ZFS_DELEG_PERM_USERUSED,
235 ZFS_DELEG_PERM_USERQUOTA,
236 ZFS_DELEG_PERM_GROUPUSED,
237 ZFS_DELEG_PERM_GROUPQUOTA,
238 };
239
240 static int zfs_ioc_userspace_upgrade(zfs_cmd_t *zc);
241 static int zfs_check_settable(const char *name, nvpair_t *property,
242 cred_t *cr);
243 static int zfs_check_clearable(char *dataset, nvlist_t *props,
244 nvlist_t **errors);
245 static int zfs_fill_zplprops_root(uint64_t, nvlist_t *, nvlist_t *,
246 boolean_t *);
247 int zfs_set_prop_nvlist(const char *, zprop_source_t, nvlist_t *, nvlist_t *);
248 static int get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp);
249
250 static int zfs_prop_activate_feature(spa_t *spa, spa_feature_t feature);
251
252 /* _NOTE(PRINTFLIKE(4)) - this is printf-like, but lint is too whiney */
253 void
254 __dprintf(const char *file, const char *func, int line, const char *fmt, ...)
255 {
256 const char *newfile;
257 char buf[512];
258 va_list adx;
259
260 /*
261 * Get rid of annoying "../common/" prefix to filename.
262 */
263 newfile = strrchr(file, '/');
264 if (newfile != NULL) {
265 newfile = newfile + 1; /* Get rid of leading / */
266 } else {
267 newfile = file;
268 }
269
270 va_start(adx, fmt);
271 (void) vsnprintf(buf, sizeof (buf), fmt, adx);
272 va_end(adx);
273
274 /*
275 * To get this data, use the zfs-dprintf probe as so:
276 * dtrace -q -n 'zfs-dprintf \
277 * /stringof(arg0) == "dbuf.c"/ \
278 * {printf("%s: %s", stringof(arg1), stringof(arg3))}'
279 * arg0 = file name
280 * arg1 = function name
281 * arg2 = line number
282 * arg3 = message
283 */
284 DTRACE_PROBE4(zfs__dprintf,
285 char *, newfile, char *, func, int, line, char *, buf);
286 }
287
288 static void
289 history_str_free(char *buf)
290 {
291 kmem_free(buf, HIS_MAX_RECORD_LEN);
292 }
293
294 static char *
295 history_str_get(zfs_cmd_t *zc)
296 {
297 char *buf;
298
299 if (zc->zc_history == NULL)
300 return (NULL);
301
302 buf = kmem_alloc(HIS_MAX_RECORD_LEN, KM_SLEEP);
303 if (copyinstr((void *)(uintptr_t)zc->zc_history,
304 buf, HIS_MAX_RECORD_LEN, NULL) != 0) {
305 history_str_free(buf);
306 return (NULL);
307 }
308
309 buf[HIS_MAX_RECORD_LEN -1] = '\0';
310
311 return (buf);
312 }
313
314 /*
315 * Check to see if the named dataset is currently defined as bootable
316 */
317 static boolean_t
318 zfs_is_bootfs(const char *name)
319 {
320 objset_t *os;
321
322 if (dmu_objset_hold(name, FTAG, &os) == 0) {
323 boolean_t ret;
324 ret = (dmu_objset_id(os) == spa_bootfs(dmu_objset_spa(os)));
325 dmu_objset_rele(os, FTAG);
326 return (ret);
327 }
328 return (B_FALSE);
329 }
330
331 /*
332 * Return non-zero if the spa version is less than requested version.
333 */
334 static int
335 zfs_earlier_version(const char *name, int version)
336 {
337 spa_t *spa;
338
339 if (spa_open(name, &spa, FTAG) == 0) {
340 if (spa_version(spa) < version) {
341 spa_close(spa, FTAG);
342 return (1);
343 }
344 spa_close(spa, FTAG);
345 }
346 return (0);
347 }
348
349 /*
350 * Return TRUE if the ZPL version is less than requested version.
351 */
352 static boolean_t
353 zpl_earlier_version(const char *name, int version)
354 {
355 objset_t *os;
356 boolean_t rc = B_TRUE;
357
358 if (dmu_objset_hold(name, FTAG, &os) == 0) {
359 uint64_t zplversion;
360
361 if (dmu_objset_type(os) != DMU_OST_ZFS) {
362 dmu_objset_rele(os, FTAG);
363 return (B_TRUE);
364 }
365 /* XXX reading from non-owned objset */
366 if (zfs_get_zplprop(os, ZFS_PROP_VERSION, &zplversion) == 0)
367 rc = zplversion < version;
368 dmu_objset_rele(os, FTAG);
369 }
370 return (rc);
371 }
372
373 static void
374 zfs_log_history(zfs_cmd_t *zc)
375 {
376 spa_t *spa;
377 char *buf;
378
379 if ((buf = history_str_get(zc)) == NULL)
380 return;
381
382 if (spa_open(zc->zc_name, &spa, FTAG) == 0) {
383 if (spa_version(spa) >= SPA_VERSION_ZPOOL_HISTORY)
384 (void) spa_history_log(spa, buf);
385 spa_close(spa, FTAG);
386 }
387 history_str_free(buf);
388 }
389
390 /*
391 * Policy for top-level read operations (list pools). Requires no privileges,
392 * and can be used in the local zone, as there is no associated dataset.
393 */
394 /* ARGSUSED */
395 static int
396 zfs_secpolicy_none(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
397 {
398 return (0);
399 }
400
401 /*
402 * Policy for dataset read operations (list children, get statistics). Requires
403 * no privileges, but must be visible in the local zone.
404 */
405 /* ARGSUSED */
406 static int
407 zfs_secpolicy_read(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
408 {
409 if (INGLOBALZONE(curproc) ||
410 zone_dataset_visible(zc->zc_name, NULL))
411 return (0);
412
413 return (SET_ERROR(ENOENT));
414 }
415
416 static int
417 zfs_dozonecheck_impl(const char *dataset, uint64_t zoned, cred_t *cr)
418 {
419 int writable = 1;
420
421 /*
422 * The dataset must be visible by this zone -- check this first
423 * so they don't see EPERM on something they shouldn't know about.
424 */
425 if (!INGLOBALZONE(curproc) &&
426 !zone_dataset_visible(dataset, &writable))
427 return (SET_ERROR(ENOENT));
428
429 if (INGLOBALZONE(curproc)) {
430 /*
431 * If the fs is zoned, only root can access it from the
432 * global zone.
433 */
434 if (secpolicy_zfs(cr) && zoned)
435 return (SET_ERROR(EPERM));
436 } else {
437 /*
438 * If we are in a local zone, the 'zoned' property must be set.
439 */
440 if (!zoned)
441 return (SET_ERROR(EPERM));
442
443 /* must be writable by this zone */
444 if (!writable)
445 return (SET_ERROR(EPERM));
446 }
447 return (0);
448 }
449
450 static int
451 zfs_dozonecheck(const char *dataset, cred_t *cr)
452 {
453 uint64_t zoned;
454
455 if (dsl_prop_get_integer(dataset, "zoned", &zoned, NULL))
456 return (SET_ERROR(ENOENT));
457
458 return (zfs_dozonecheck_impl(dataset, zoned, cr));
459 }
460
461 static int
462 zfs_dozonecheck_ds(const char *dataset, dsl_dataset_t *ds, cred_t *cr)
463 {
464 uint64_t zoned;
465
466 if (dsl_prop_get_int_ds(ds, "zoned", &zoned))
467 return (SET_ERROR(ENOENT));
468
469 return (zfs_dozonecheck_impl(dataset, zoned, cr));
470 }
471
472 static int
473 zfs_secpolicy_write_perms_ds(const char *name, dsl_dataset_t *ds,
474 const char *perm, cred_t *cr)
475 {
476 int error;
477
478 error = zfs_dozonecheck_ds(name, ds, cr);
479 if (error == 0) {
480 error = secpolicy_zfs(cr);
481 if (error != 0)
482 error = dsl_deleg_access_impl(ds, perm, cr);
483 }
484 return (error);
485 }
486
487 static int
488 zfs_secpolicy_write_perms(const char *name, const char *perm, cred_t *cr)
489 {
490 int error;
491 dsl_dataset_t *ds;
492 dsl_pool_t *dp;
493
494 error = dsl_pool_hold(name, FTAG, &dp);
495 if (error != 0)
496 return (error);
497
498 error = dsl_dataset_hold(dp, name, FTAG, &ds);
499 if (error != 0) {
500 dsl_pool_rele(dp, FTAG);
501 return (error);
502 }
503
504 error = zfs_secpolicy_write_perms_ds(name, ds, perm, cr);
505
506 dsl_dataset_rele(ds, FTAG);
507 dsl_pool_rele(dp, FTAG);
508 return (error);
509 }
510
511 /*
512 * Policy for setting the security label property.
513 *
514 * Returns 0 for success, non-zero for access and other errors.
515 */
516 static int
517 zfs_set_slabel_policy(const char *name, char *strval, cred_t *cr)
518 {
519 char ds_hexsl[MAXNAMELEN];
520 bslabel_t ds_sl, new_sl;
521 boolean_t new_default = FALSE;
522 uint64_t zoned;
523 int needed_priv = -1;
524 int error;
525
526 /* First get the existing dataset label. */
527 error = dsl_prop_get(name, zfs_prop_to_name(ZFS_PROP_MLSLABEL),
528 1, sizeof (ds_hexsl), &ds_hexsl, NULL);
529 if (error != 0)
530 return (SET_ERROR(EPERM));
531
532 if (strcasecmp(strval, ZFS_MLSLABEL_DEFAULT) == 0)
533 new_default = TRUE;
534
535 /* The label must be translatable */
536 if (!new_default && (hexstr_to_label(strval, &new_sl) != 0))
537 return (SET_ERROR(EINVAL));
538
539 /*
540 * In a non-global zone, disallow attempts to set a label that
541 * doesn't match that of the zone; otherwise no other checks
542 * are needed.
543 */
544 if (!INGLOBALZONE(curproc)) {
545 if (new_default || !blequal(&new_sl, CR_SL(CRED())))
546 return (SET_ERROR(EPERM));
547 return (0);
548 }
549
550 /*
551 * For global-zone datasets (i.e., those whose zoned property is
552 * "off", verify that the specified new label is valid for the
553 * global zone.
554 */
555 if (dsl_prop_get_integer(name,
556 zfs_prop_to_name(ZFS_PROP_ZONED), &zoned, NULL))
557 return (SET_ERROR(EPERM));
558 if (!zoned) {
559 if (zfs_check_global_label(name, strval) != 0)
560 return (SET_ERROR(EPERM));
561 }
562
563 /*
564 * If the existing dataset label is nondefault, check if the
565 * dataset is mounted (label cannot be changed while mounted).
566 * Get the zfsvfs; if there isn't one, then the dataset isn't
567 * mounted (or isn't a dataset, doesn't exist, ...).
568 */
569 if (strcasecmp(ds_hexsl, ZFS_MLSLABEL_DEFAULT) != 0) {
570 objset_t *os;
571 static char *setsl_tag = "setsl_tag";
572
573 /*
574 * Try to own the dataset; abort if there is any error,
575 * (e.g., already mounted, in use, or other error).
576 */
577 error = dmu_objset_own(name, DMU_OST_ZFS, B_TRUE,
578 setsl_tag, &os);
579 if (error != 0)
580 return (SET_ERROR(EPERM));
581
582 dmu_objset_disown(os, setsl_tag);
583
584 if (new_default) {
585 needed_priv = PRIV_FILE_DOWNGRADE_SL;
586 goto out_check;
587 }
588
589 if (hexstr_to_label(strval, &new_sl) != 0)
590 return (SET_ERROR(EPERM));
591
592 if (blstrictdom(&ds_sl, &new_sl))
593 needed_priv = PRIV_FILE_DOWNGRADE_SL;
594 else if (blstrictdom(&new_sl, &ds_sl))
595 needed_priv = PRIV_FILE_UPGRADE_SL;
596 } else {
597 /* dataset currently has a default label */
598 if (!new_default)
599 needed_priv = PRIV_FILE_UPGRADE_SL;
600 }
601
602 out_check:
603 if (needed_priv != -1)
604 return (PRIV_POLICY(cr, needed_priv, B_FALSE, EPERM, NULL));
605 return (0);
606 }
607
608 static int
609 zfs_secpolicy_setprop(const char *dsname, zfs_prop_t prop, nvpair_t *propval,
610 cred_t *cr)
611 {
612 char *strval;
613
614 /*
615 * Check permissions for special properties.
616 */
617 switch (prop) {
618 case ZFS_PROP_ZONED:
619 /*
620 * Disallow setting of 'zoned' from within a local zone.
621 */
622 if (!INGLOBALZONE(curproc))
623 return (SET_ERROR(EPERM));
624 break;
625
626 case ZFS_PROP_QUOTA:
627 case ZFS_PROP_FILESYSTEM_LIMIT:
628 case ZFS_PROP_SNAPSHOT_LIMIT:
629 if (!INGLOBALZONE(curproc)) {
630 uint64_t zoned;
631 char setpoint[ZFS_MAX_DATASET_NAME_LEN];
632 /*
633 * Unprivileged users are allowed to modify the
634 * limit on things *under* (ie. contained by)
635 * the thing they own.
636 */
637 if (dsl_prop_get_integer(dsname, "zoned", &zoned,
638 setpoint))
639 return (SET_ERROR(EPERM));
640 if (!zoned || strlen(dsname) <= strlen(setpoint))
641 return (SET_ERROR(EPERM));
642 }
643 break;
644
645 case ZFS_PROP_MLSLABEL:
646 if (!is_system_labeled())
647 return (SET_ERROR(EPERM));
648
649 if (nvpair_value_string(propval, &strval) == 0) {
650 int err;
651
652 err = zfs_set_slabel_policy(dsname, strval, CRED());
653 if (err != 0)
654 return (err);
655 }
656 break;
657 }
658
659 return (zfs_secpolicy_write_perms(dsname, zfs_prop_to_name(prop), cr));
660 }
661
662 /* ARGSUSED */
663 static int
664 zfs_secpolicy_set_fsacl(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
665 {
666 int error;
667
668 error = zfs_dozonecheck(zc->zc_name, cr);
669 if (error != 0)
670 return (error);
671
672 /*
673 * permission to set permissions will be evaluated later in
674 * dsl_deleg_can_allow()
675 */
676 return (0);
677 }
678
679 /* ARGSUSED */
680 static int
681 zfs_secpolicy_rollback(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
682 {
683 return (zfs_secpolicy_write_perms(zc->zc_name,
684 ZFS_DELEG_PERM_ROLLBACK, cr));
685 }
686
687 /* ARGSUSED */
688 static int
689 zfs_secpolicy_send(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
690 {
691 dsl_pool_t *dp;
692 dsl_dataset_t *ds;
693 char *cp;
694 int error;
695
696 /*
697 * Generate the current snapshot name from the given objsetid, then
698 * use that name for the secpolicy/zone checks.
699 */
700 cp = strchr(zc->zc_name, '@');
701 if (cp == NULL)
702 return (SET_ERROR(EINVAL));
703 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
704 if (error != 0)
705 return (error);
706
707 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &ds);
708 if (error != 0) {
709 dsl_pool_rele(dp, FTAG);
710 return (error);
711 }
712
713 dsl_dataset_name(ds, zc->zc_name);
714
715 error = zfs_secpolicy_write_perms_ds(zc->zc_name, ds,
716 ZFS_DELEG_PERM_SEND, cr);
717 dsl_dataset_rele(ds, FTAG);
718 dsl_pool_rele(dp, FTAG);
719
720 return (error);
721 }
722
723 /* ARGSUSED */
724 static int
725 zfs_secpolicy_send_new(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
726 {
727 return (zfs_secpolicy_write_perms(zc->zc_name,
728 ZFS_DELEG_PERM_SEND, cr));
729 }
730
731 /* ARGSUSED */
732 static int
733 zfs_secpolicy_deleg_share(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
734 {
735 vnode_t *vp;
736 int error;
737
738 if ((error = lookupname(zc->zc_value, UIO_SYSSPACE,
739 NO_FOLLOW, NULL, &vp)) != 0)
740 return (error);
741
742 /* Now make sure mntpnt and dataset are ZFS */
743
744 if (vp->v_vfsp->vfs_fstype != zfsfstype ||
745 (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource),
746 zc->zc_name) != 0)) {
747 VN_RELE(vp);
748 return (SET_ERROR(EPERM));
749 }
750
751 VN_RELE(vp);
752 return (dsl_deleg_access(zc->zc_name,
753 ZFS_DELEG_PERM_SHARE, cr));
754 }
755
756 int
757 zfs_secpolicy_share(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
758 {
759 if (!INGLOBALZONE(curproc))
760 return (SET_ERROR(EPERM));
761
762 if (secpolicy_nfs(cr) == 0) {
763 return (0);
764 } else {
765 return (zfs_secpolicy_deleg_share(zc, innvl, cr));
766 }
767 }
768
769 int
770 zfs_secpolicy_smb_acl(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
771 {
772 if (!INGLOBALZONE(curproc))
773 return (SET_ERROR(EPERM));
774
775 if (secpolicy_smb(cr) == 0) {
776 return (0);
777 } else {
778 return (zfs_secpolicy_deleg_share(zc, innvl, cr));
779 }
780 }
781
782 static int
783 zfs_get_parent(const char *datasetname, char *parent, int parentsize)
784 {
785 char *cp;
786
787 /*
788 * Remove the @bla or /bla from the end of the name to get the parent.
789 */
790 (void) strncpy(parent, datasetname, parentsize);
791 cp = strrchr(parent, '@');
792 if (cp != NULL) {
793 cp[0] = '\0';
794 } else {
795 cp = strrchr(parent, '/');
796 if (cp == NULL)
797 return (SET_ERROR(ENOENT));
798 cp[0] = '\0';
799 }
800
801 return (0);
802 }
803
804 int
805 zfs_secpolicy_destroy_perms(const char *name, cred_t *cr)
806 {
807 int error;
808
809 if ((error = zfs_secpolicy_write_perms(name,
810 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
811 return (error);
812
813 return (zfs_secpolicy_write_perms(name, ZFS_DELEG_PERM_DESTROY, cr));
814 }
815
816 /* ARGSUSED */
817 static int
818 zfs_secpolicy_destroy(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
819 {
820 return (zfs_secpolicy_destroy_perms(zc->zc_name, cr));
821 }
822
823 /*
824 * Destroying snapshots with delegated permissions requires
825 * descendant mount and destroy permissions.
826 */
827 /* ARGSUSED */
828 static int
829 zfs_secpolicy_destroy_snaps(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
830 {
831 nvlist_t *snaps;
832 nvpair_t *pair, *nextpair;
833 int error = 0;
834
835 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
836 return (SET_ERROR(EINVAL));
837 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
838 pair = nextpair) {
839 nextpair = nvlist_next_nvpair(snaps, pair);
840 error = zfs_secpolicy_destroy_perms(nvpair_name(pair), cr);
841 if (error == ENOENT) {
842 /*
843 * Ignore any snapshots that don't exist (we consider
844 * them "already destroyed"). Remove the name from the
845 * nvl here in case the snapshot is created between
846 * now and when we try to destroy it (in which case
847 * we don't want to destroy it since we haven't
848 * checked for permission).
849 */
850 fnvlist_remove_nvpair(snaps, pair);
851 error = 0;
852 }
853 if (error != 0)
854 break;
855 }
856
857 return (error);
858 }
859
860 int
861 zfs_secpolicy_rename_perms(const char *from, const char *to, cred_t *cr)
862 {
863 char parentname[ZFS_MAX_DATASET_NAME_LEN];
864 int error;
865
866 if ((error = zfs_secpolicy_write_perms(from,
867 ZFS_DELEG_PERM_RENAME, cr)) != 0)
868 return (error);
869
870 if ((error = zfs_secpolicy_write_perms(from,
871 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
872 return (error);
873
874 if ((error = zfs_get_parent(to, parentname,
875 sizeof (parentname))) != 0)
876 return (error);
877
878 if ((error = zfs_secpolicy_write_perms(parentname,
879 ZFS_DELEG_PERM_CREATE, cr)) != 0)
880 return (error);
881
882 if ((error = zfs_secpolicy_write_perms(parentname,
883 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
884 return (error);
885
886 return (error);
887 }
888
889 /* ARGSUSED */
890 static int
891 zfs_secpolicy_rename(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
892 {
893 return (zfs_secpolicy_rename_perms(zc->zc_name, zc->zc_value, cr));
894 }
895
896 /* ARGSUSED */
897 static int
898 zfs_secpolicy_promote(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
899 {
900 dsl_pool_t *dp;
901 dsl_dataset_t *clone;
902 int error;
903
904 error = zfs_secpolicy_write_perms(zc->zc_name,
905 ZFS_DELEG_PERM_PROMOTE, cr);
906 if (error != 0)
907 return (error);
908
909 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
910 if (error != 0)
911 return (error);
912
913 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &clone);
914
915 if (error == 0) {
916 char parentname[ZFS_MAX_DATASET_NAME_LEN];
917 dsl_dataset_t *origin = NULL;
918 dsl_dir_t *dd;
919 dd = clone->ds_dir;
920
921 error = dsl_dataset_hold_obj(dd->dd_pool,
922 dsl_dir_phys(dd)->dd_origin_obj, FTAG, &origin);
923 if (error != 0) {
924 dsl_dataset_rele(clone, FTAG);
925 dsl_pool_rele(dp, FTAG);
926 return (error);
927 }
928
929 error = zfs_secpolicy_write_perms_ds(zc->zc_name, clone,
930 ZFS_DELEG_PERM_MOUNT, cr);
931
932 dsl_dataset_name(origin, parentname);
933 if (error == 0) {
934 error = zfs_secpolicy_write_perms_ds(parentname, origin,
935 ZFS_DELEG_PERM_PROMOTE, cr);
936 }
937 dsl_dataset_rele(clone, FTAG);
938 dsl_dataset_rele(origin, FTAG);
939 }
940 dsl_pool_rele(dp, FTAG);
941 return (error);
942 }
943
944 /* ARGSUSED */
945 static int
946 zfs_secpolicy_recv(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
947 {
948 int error;
949
950 if ((error = zfs_secpolicy_write_perms(zc->zc_name,
951 ZFS_DELEG_PERM_RECEIVE, cr)) != 0)
952 return (error);
953
954 if ((error = zfs_secpolicy_write_perms(zc->zc_name,
955 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
956 return (error);
957
958 return (zfs_secpolicy_write_perms(zc->zc_name,
959 ZFS_DELEG_PERM_CREATE, cr));
960 }
961
962 int
963 zfs_secpolicy_snapshot_perms(const char *name, cred_t *cr)
964 {
965 return (zfs_secpolicy_write_perms(name,
966 ZFS_DELEG_PERM_SNAPSHOT, cr));
967 }
968
969 /*
970 * Check for permission to create each snapshot in the nvlist.
971 */
972 /* ARGSUSED */
973 static int
974 zfs_secpolicy_snapshot(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
975 {
976 nvlist_t *snaps;
977 int error = 0;
978 nvpair_t *pair;
979
980 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
981 return (SET_ERROR(EINVAL));
982 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
983 pair = nvlist_next_nvpair(snaps, pair)) {
984 char *name = nvpair_name(pair);
985 char *atp = strchr(name, '@');
986
987 if (atp == NULL) {
988 error = SET_ERROR(EINVAL);
989 break;
990 }
991 *atp = '\0';
992 error = zfs_secpolicy_snapshot_perms(name, cr);
993 *atp = '@';
994 if (error != 0)
995 break;
996 }
997 return (error);
998 }
999
1000 /*
1001 * Check for permission to create each snapshot in the nvlist.
1002 */
1003 /* ARGSUSED */
1004 static int
1005 zfs_secpolicy_bookmark(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1006 {
1007 int error = 0;
1008
1009 for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL);
1010 pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) {
1011 char *name = nvpair_name(pair);
1012 char *hashp = strchr(name, '#');
1013
1014 if (hashp == NULL) {
1015 error = SET_ERROR(EINVAL);
1016 break;
1017 }
1018 *hashp = '\0';
1019 error = zfs_secpolicy_write_perms(name,
1020 ZFS_DELEG_PERM_BOOKMARK, cr);
1021 *hashp = '#';
1022 if (error != 0)
1023 break;
1024 }
1025 return (error);
1026 }
1027
1028 /* ARGSUSED */
1029 static int
1030 zfs_secpolicy_destroy_bookmarks(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1031 {
1032 nvpair_t *pair, *nextpair;
1033 int error = 0;
1034
1035 for (pair = nvlist_next_nvpair(innvl, NULL); pair != NULL;
1036 pair = nextpair) {
1037 char *name = nvpair_name(pair);
1038 char *hashp = strchr(name, '#');
1039 nextpair = nvlist_next_nvpair(innvl, pair);
1040
1041 if (hashp == NULL) {
1042 error = SET_ERROR(EINVAL);
1043 break;
1044 }
1045
1046 *hashp = '\0';
1047 error = zfs_secpolicy_write_perms(name,
1048 ZFS_DELEG_PERM_DESTROY, cr);
1049 *hashp = '#';
1050 if (error == ENOENT) {
1051 /*
1052 * Ignore any filesystems that don't exist (we consider
1053 * their bookmarks "already destroyed"). Remove
1054 * the name from the nvl here in case the filesystem
1055 * is created between now and when we try to destroy
1056 * the bookmark (in which case we don't want to
1057 * destroy it since we haven't checked for permission).
1058 */
1059 fnvlist_remove_nvpair(innvl, pair);
1060 error = 0;
1061 }
1062 if (error != 0)
1063 break;
1064 }
1065
1066 return (error);
1067 }
1068
1069 /* ARGSUSED */
1070 static int
1071 zfs_secpolicy_log_history(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1072 {
1073 /*
1074 * Even root must have a proper TSD so that we know what pool
1075 * to log to.
1076 */
1077 if (tsd_get(zfs_allow_log_key) == NULL)
1078 return (SET_ERROR(EPERM));
1079 return (0);
1080 }
1081
1082 static int
1083 zfs_secpolicy_create_clone(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1084 {
1085 char parentname[ZFS_MAX_DATASET_NAME_LEN];
1086 int error;
1087 char *origin;
1088
1089 if ((error = zfs_get_parent(zc->zc_name, parentname,
1090 sizeof (parentname))) != 0)
1091 return (error);
1092
1093 if (nvlist_lookup_string(innvl, "origin", &origin) == 0 &&
1094 (error = zfs_secpolicy_write_perms(origin,
1095 ZFS_DELEG_PERM_CLONE, cr)) != 0)
1096 return (error);
1097
1098 if ((error = zfs_secpolicy_write_perms(parentname,
1099 ZFS_DELEG_PERM_CREATE, cr)) != 0)
1100 return (error);
1101
1102 return (zfs_secpolicy_write_perms(parentname,
1103 ZFS_DELEG_PERM_MOUNT, cr));
1104 }
1105
1106 /*
1107 * Policy for pool operations - create/destroy pools, add vdevs, etc. Requires
1108 * SYS_CONFIG privilege, which is not available in a local zone.
1109 */
1110 /* ARGSUSED */
1111 static int
1112 zfs_secpolicy_config(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1113 {
1114 if (secpolicy_sys_config(cr, B_FALSE) != 0)
1115 return (SET_ERROR(EPERM));
1116
1117 return (0);
1118 }
1119
1120 /*
1121 * Policy for object to name lookups.
1122 */
1123 /* ARGSUSED */
1124 static int
1125 zfs_secpolicy_diff(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1126 {
1127 int error;
1128
1129 if ((error = secpolicy_sys_config(cr, B_FALSE)) == 0)
1130 return (0);
1131
1132 error = zfs_secpolicy_write_perms(zc->zc_name, ZFS_DELEG_PERM_DIFF, cr);
1133 return (error);
1134 }
1135
1136 /*
1137 * Policy for fault injection. Requires all privileges.
1138 */
1139 /* ARGSUSED */
1140 static int
1141 zfs_secpolicy_inject(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1142 {
1143 return (secpolicy_zinject(cr));
1144 }
1145
1146 /* ARGSUSED */
1147 static int
1148 zfs_secpolicy_inherit_prop(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1149 {
1150 zfs_prop_t prop = zfs_name_to_prop(zc->zc_value);
1151
1152 if (prop == ZPROP_INVAL) {
1153 if (!zfs_prop_user(zc->zc_value))
1154 return (SET_ERROR(EINVAL));
1155 return (zfs_secpolicy_write_perms(zc->zc_name,
1156 ZFS_DELEG_PERM_USERPROP, cr));
1157 } else {
1158 return (zfs_secpolicy_setprop(zc->zc_name, prop,
1159 NULL, cr));
1160 }
1161 }
1162
1163 static int
1164 zfs_secpolicy_userspace_one(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1165 {
1166 int err = zfs_secpolicy_read(zc, innvl, cr);
1167 if (err)
1168 return (err);
1169
1170 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
1171 return (SET_ERROR(EINVAL));
1172
1173 if (zc->zc_value[0] == 0) {
1174 /*
1175 * They are asking about a posix uid/gid. If it's
1176 * themself, allow it.
1177 */
1178 if (zc->zc_objset_type == ZFS_PROP_USERUSED ||
1179 zc->zc_objset_type == ZFS_PROP_USERQUOTA) {
1180 if (zc->zc_guid == crgetuid(cr))
1181 return (0);
1182 } else {
1183 if (groupmember(zc->zc_guid, cr))
1184 return (0);
1185 }
1186 }
1187
1188 return (zfs_secpolicy_write_perms(zc->zc_name,
1189 userquota_perms[zc->zc_objset_type], cr));
1190 }
1191
1192 static int
1193 zfs_secpolicy_userspace_many(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1194 {
1195 int err = zfs_secpolicy_read(zc, innvl, cr);
1196 if (err)
1197 return (err);
1198
1199 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
1200 return (SET_ERROR(EINVAL));
1201
1202 return (zfs_secpolicy_write_perms(zc->zc_name,
1203 userquota_perms[zc->zc_objset_type], cr));
1204 }
1205
1206 /* ARGSUSED */
1207 static int
1208 zfs_secpolicy_userspace_upgrade(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1209 {
1210 return (zfs_secpolicy_setprop(zc->zc_name, ZFS_PROP_VERSION,
1211 NULL, cr));
1212 }
1213
1214 /* ARGSUSED */
1215 static int
1216 zfs_secpolicy_hold(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1217 {
1218 nvpair_t *pair;
1219 nvlist_t *holds;
1220 int error;
1221
1222 error = nvlist_lookup_nvlist(innvl, "holds", &holds);
1223 if (error != 0)
1224 return (SET_ERROR(EINVAL));
1225
1226 for (pair = nvlist_next_nvpair(holds, NULL); pair != NULL;
1227 pair = nvlist_next_nvpair(holds, pair)) {
1228 char fsname[ZFS_MAX_DATASET_NAME_LEN];
1229 error = dmu_fsname(nvpair_name(pair), fsname);
1230 if (error != 0)
1231 return (error);
1232 error = zfs_secpolicy_write_perms(fsname,
1233 ZFS_DELEG_PERM_HOLD, cr);
1234 if (error != 0)
1235 return (error);
1236 }
1237 return (0);
1238 }
1239
1240 /* ARGSUSED */
1241 static int
1242 zfs_secpolicy_release(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1243 {
1244 nvpair_t *pair;
1245 int error;
1246
1247 for (pair = nvlist_next_nvpair(innvl, NULL); pair != NULL;
1248 pair = nvlist_next_nvpair(innvl, pair)) {
1249 char fsname[ZFS_MAX_DATASET_NAME_LEN];
1250 error = dmu_fsname(nvpair_name(pair), fsname);
1251 if (error != 0)
1252 return (error);
1253 error = zfs_secpolicy_write_perms(fsname,
1254 ZFS_DELEG_PERM_RELEASE, cr);
1255 if (error != 0)
1256 return (error);
1257 }
1258 return (0);
1259 }
1260
1261 /*
1262 * Policy for allowing temporary snapshots to be taken or released
1263 */
1264 static int
1265 zfs_secpolicy_tmp_snapshot(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1266 {
1267 /*
1268 * A temporary snapshot is the same as a snapshot,
1269 * hold, destroy and release all rolled into one.
1270 * Delegated diff alone is sufficient that we allow this.
1271 */
1272 int error;
1273
1274 if ((error = zfs_secpolicy_write_perms(zc->zc_name,
1275 ZFS_DELEG_PERM_DIFF, cr)) == 0)
1276 return (0);
1277
1278 error = zfs_secpolicy_snapshot_perms(zc->zc_name, cr);
1279 if (error == 0)
1280 error = zfs_secpolicy_hold(zc, innvl, cr);
1281 if (error == 0)
1282 error = zfs_secpolicy_release(zc, innvl, cr);
1283 if (error == 0)
1284 error = zfs_secpolicy_destroy(zc, innvl, cr);
1285 return (error);
1286 }
1287
1288 /*
1289 * Returns the nvlist as specified by the user in the zfs_cmd_t.
1290 */
1291 static int
1292 get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp)
1293 {
1294 char *packed;
1295 int error;
1296 nvlist_t *list = NULL;
1297
1298 /*
1299 * Read in and unpack the user-supplied nvlist.
1300 */
1301 if (size == 0)
1302 return (SET_ERROR(EINVAL));
1303
1304 packed = kmem_alloc(size, KM_SLEEP);
1305
1306 if ((error = ddi_copyin((void *)(uintptr_t)nvl, packed, size,
1307 iflag)) != 0) {
1308 kmem_free(packed, size);
1309 return (SET_ERROR(EFAULT));
1310 }
1311
1312 if ((error = nvlist_unpack(packed, size, &list, 0)) != 0) {
1313 kmem_free(packed, size);
1314 return (error);
1315 }
1316
1317 kmem_free(packed, size);
1318
1319 *nvp = list;
1320 return (0);
1321 }
1322
1323 /*
1324 * Reduce the size of this nvlist until it can be serialized in 'max' bytes.
1325 * Entries will be removed from the end of the nvlist, and one int32 entry
1326 * named "N_MORE_ERRORS" will be added indicating how many entries were
1327 * removed.
1328 */
1329 static int
1330 nvlist_smush(nvlist_t *errors, size_t max)
1331 {
1332 size_t size;
1333
1334 size = fnvlist_size(errors);
1335
1336 if (size > max) {
1337 nvpair_t *more_errors;
1338 int n = 0;
1339
1340 if (max < 1024)
1341 return (SET_ERROR(ENOMEM));
1342
1343 fnvlist_add_int32(errors, ZPROP_N_MORE_ERRORS, 0);
1344 more_errors = nvlist_prev_nvpair(errors, NULL);
1345
1346 do {
1347 nvpair_t *pair = nvlist_prev_nvpair(errors,
1348 more_errors);
1349 fnvlist_remove_nvpair(errors, pair);
1350 n++;
1351 size = fnvlist_size(errors);
1352 } while (size > max);
1353
1354 fnvlist_remove_nvpair(errors, more_errors);
1355 fnvlist_add_int32(errors, ZPROP_N_MORE_ERRORS, n);
1356 ASSERT3U(fnvlist_size(errors), <=, max);
1357 }
1358
1359 return (0);
1360 }
1361
1362 static int
1363 put_nvlist(zfs_cmd_t *zc, nvlist_t *nvl)
1364 {
1365 char *packed = NULL;
1366 int error = 0;
1367 size_t size;
1368
1369 size = fnvlist_size(nvl);
1370
1371 if (size > zc->zc_nvlist_dst_size) {
1372 error = SET_ERROR(ENOMEM);
1373 } else {
1374 packed = fnvlist_pack(nvl, &size);
1375 if (ddi_copyout(packed, (void *)(uintptr_t)zc->zc_nvlist_dst,
1376 size, zc->zc_iflags) != 0)
1377 error = SET_ERROR(EFAULT);
1378 fnvlist_pack_free(packed, size);
1379 }
1380
1381 zc->zc_nvlist_dst_size = size;
1382 zc->zc_nvlist_dst_filled = B_TRUE;
1383 return (error);
1384 }
1385
1386 static int
1387 getzfsvfs(const char *dsname, zfsvfs_t **zfvp)
1388 {
1389 objset_t *os;
1390 int error;
1391
1392 error = dmu_objset_hold(dsname, FTAG, &os);
1393 if (error != 0)
1394 return (error);
1395 if (dmu_objset_type(os) != DMU_OST_ZFS) {
1396 dmu_objset_rele(os, FTAG);
1397 return (SET_ERROR(EINVAL));
1398 }
1399
1400 mutex_enter(&os->os_user_ptr_lock);
1401 *zfvp = dmu_objset_get_user(os);
1402 if (*zfvp) {
1403 VFS_HOLD((*zfvp)->z_vfs);
1404 } else {
1405 error = SET_ERROR(ESRCH);
1406 }
1407 mutex_exit(&os->os_user_ptr_lock);
1408 dmu_objset_rele(os, FTAG);
1409 return (error);
1410 }
1411
1412 /*
1413 * Find a zfsvfs_t for a mounted filesystem, or create our own, in which
1414 * case its z_vfs will be NULL, and it will be opened as the owner.
1415 * If 'writer' is set, the z_teardown_lock will be held for RW_WRITER,
1416 * which prevents all vnode ops from running.
1417 */
1418 static int
1419 zfsvfs_hold(const char *name, void *tag, zfsvfs_t **zfvp, boolean_t writer)
1420 {
1421 int error = 0;
1422
1423 if (getzfsvfs(name, zfvp) != 0)
1424 error = zfsvfs_create(name, zfvp);
1425 if (error == 0) {
1426 rrm_enter(&(*zfvp)->z_teardown_lock, (writer) ? RW_WRITER :
1427 RW_READER, tag);
1428 if ((*zfvp)->z_unmounted) {
1429 /*
1430 * XXX we could probably try again, since the unmounting
1431 * thread should be just about to disassociate the
1432 * objset from the zfsvfs.
1433 */
1434 rrm_exit(&(*zfvp)->z_teardown_lock, tag);
1435 return (SET_ERROR(EBUSY));
1436 }
1437 }
1438 return (error);
1439 }
1440
1441 static void
1442 zfsvfs_rele(zfsvfs_t *zfsvfs, void *tag)
1443 {
1444 rrm_exit(&zfsvfs->z_teardown_lock, tag);
1445
1446 if (zfsvfs->z_vfs) {
1447 VFS_RELE(zfsvfs->z_vfs);
1448 } else {
1449 dmu_objset_disown(zfsvfs->z_os, zfsvfs);
1450 zfsvfs_free(zfsvfs);
1451 }
1452 }
1453
1454 static int
1455 zfs_ioc_pool_create(zfs_cmd_t *zc)
1456 {
1457 int error;
1458 nvlist_t *config, *props = NULL;
1459 nvlist_t *rootprops = NULL;
1460 nvlist_t *zplprops = NULL;
1461
1462 if (error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1463 zc->zc_iflags, &config))
1464 return (error);
1465
1466 if (zc->zc_nvlist_src_size != 0 && (error =
1467 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1468 zc->zc_iflags, &props))) {
1469 nvlist_free(config);
1470 return (error);
1471 }
1472
1473 if (props) {
1474 nvlist_t *nvl = NULL;
1475 uint64_t version = SPA_VERSION;
1476
1477 (void) nvlist_lookup_uint64(props,
1478 zpool_prop_to_name(ZPOOL_PROP_VERSION), &version);
1479 if (!SPA_VERSION_IS_SUPPORTED(version)) {
1480 error = SET_ERROR(EINVAL);
1481 goto pool_props_bad;
1482 }
1483 (void) nvlist_lookup_nvlist(props, ZPOOL_ROOTFS_PROPS, &nvl);
1484 if (nvl) {
1485 error = nvlist_dup(nvl, &rootprops, KM_SLEEP);
1486 if (error != 0) {
1487 nvlist_free(config);
1488 nvlist_free(props);
1489 return (error);
1490 }
1491 (void) nvlist_remove_all(props, ZPOOL_ROOTFS_PROPS);
1492 }
1493 VERIFY(nvlist_alloc(&zplprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
1494 error = zfs_fill_zplprops_root(version, rootprops,
1495 zplprops, NULL);
1496 if (error != 0)
1497 goto pool_props_bad;
1498 }
1499
1500 error = spa_create(zc->zc_name, config, props, zplprops);
1501
1502 /*
1503 * Set the remaining root properties
1504 */
1505 if (!error && (error = zfs_set_prop_nvlist(zc->zc_name,
1506 ZPROP_SRC_LOCAL, rootprops, NULL)) != 0)
1507 (void) spa_destroy(zc->zc_name);
1508
1509 pool_props_bad:
1510 nvlist_free(rootprops);
1511 nvlist_free(zplprops);
1512 nvlist_free(config);
1513 nvlist_free(props);
1514
1515 return (error);
1516 }
1517
1518 static int
1519 zfs_ioc_pool_destroy(zfs_cmd_t *zc)
1520 {
1521 int error;
1522 zfs_log_history(zc);
1523 error = spa_destroy(zc->zc_name);
1524 if (error == 0)
1525 zvol_remove_minors(zc->zc_name);
1526 return (error);
1527 }
1528
1529 static int
1530 zfs_ioc_pool_import(zfs_cmd_t *zc)
1531 {
1532 nvlist_t *config, *props = NULL;
1533 uint64_t guid;
1534 int error;
1535
1536 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1537 zc->zc_iflags, &config)) != 0)
1538 return (error);
1539
1540 if (zc->zc_nvlist_src_size != 0 && (error =
1541 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1542 zc->zc_iflags, &props))) {
1543 nvlist_free(config);
1544 return (error);
1545 }
1546
1547 if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, &guid) != 0 ||
1548 guid != zc->zc_guid)
1549 error = SET_ERROR(EINVAL);
1550 else
1551 error = spa_import(zc->zc_name, config, props, zc->zc_cookie);
1552
1553 if (zc->zc_nvlist_dst != 0) {
1554 int err;
1555
1556 if ((err = put_nvlist(zc, config)) != 0)
1557 error = err;
1558 }
1559
1560 nvlist_free(config);
1561
1562 nvlist_free(props);
1563
1564 return (error);
1565 }
1566
1567 static int
1568 zfs_ioc_pool_export(zfs_cmd_t *zc)
1569 {
1570 int error;
1571 boolean_t force = (boolean_t)zc->zc_cookie;
1572 boolean_t hardforce = (boolean_t)zc->zc_guid;
1573
1574 zfs_log_history(zc);
1575 error = spa_export(zc->zc_name, NULL, force, hardforce);
1576 if (error == 0)
1577 zvol_remove_minors(zc->zc_name);
1578 return (error);
1579 }
1580
1581 static int
1582 zfs_ioc_pool_configs(zfs_cmd_t *zc)
1583 {
1584 nvlist_t *configs;
1585 int error;
1586
1587 if ((configs = spa_all_configs(&zc->zc_cookie)) == NULL)
1588 return (SET_ERROR(EEXIST));
1589
1590 error = put_nvlist(zc, configs);
1591
1592 nvlist_free(configs);
1593
1594 return (error);
1595 }
1596
1597 /*
1598 * inputs:
1599 * zc_name name of the pool
1600 *
1601 * outputs:
1602 * zc_cookie real errno
1603 * zc_nvlist_dst config nvlist
1604 * zc_nvlist_dst_size size of config nvlist
1605 */
1606 static int
1607 zfs_ioc_pool_stats(zfs_cmd_t *zc)
1608 {
1609 nvlist_t *config;
1610 int error;
1611 int ret = 0;
1612
1613 error = spa_get_stats(zc->zc_name, &config, zc->zc_value,
1614 sizeof (zc->zc_value));
1615
1616 if (config != NULL) {
1617 ret = put_nvlist(zc, config);
1618 nvlist_free(config);
1619
1620 /*
1621 * The config may be present even if 'error' is non-zero.
1622 * In this case we return success, and preserve the real errno
1623 * in 'zc_cookie'.
1624 */
1625 zc->zc_cookie = error;
1626 } else {
1627 ret = error;
1628 }
1629
1630 return (ret);
1631 }
1632
1633 /*
1634 * Try to import the given pool, returning pool stats as appropriate so that
1635 * user land knows which devices are available and overall pool health.
1636 */
1637 static int
1638 zfs_ioc_pool_tryimport(zfs_cmd_t *zc)
1639 {
1640 nvlist_t *tryconfig, *config;
1641 int error;
1642
1643 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1644 zc->zc_iflags, &tryconfig)) != 0)
1645 return (error);
1646
1647 config = spa_tryimport(tryconfig);
1648
1649 nvlist_free(tryconfig);
1650
1651 if (config == NULL)
1652 return (SET_ERROR(EINVAL));
1653
1654 error = put_nvlist(zc, config);
1655 nvlist_free(config);
1656
1657 return (error);
1658 }
1659
1660 /*
1661 * inputs:
1662 * zc_name name of the pool
1663 * zc_cookie scan func (pool_scan_func_t)
1664 */
1665 static int
1666 zfs_ioc_pool_scan(zfs_cmd_t *zc)
1667 {
1668 spa_t *spa;
1669 int error;
1670
1671 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1672 return (error);
1673
1674 if (zc->zc_cookie == POOL_SCAN_NONE)
1675 error = spa_scan_stop(spa);
1676 else
1677 error = spa_scan(spa, zc->zc_cookie);
1678
1679 spa_close(spa, FTAG);
1680
1681 return (error);
1682 }
1683
1684 static int
1685 zfs_ioc_pool_freeze(zfs_cmd_t *zc)
1686 {
1687 spa_t *spa;
1688 int error;
1689
1690 error = spa_open(zc->zc_name, &spa, FTAG);
1691 if (error == 0) {
1692 spa_freeze(spa);
1693 spa_close(spa, FTAG);
1694 }
1695 return (error);
1696 }
1697
1698 static int
1699 zfs_ioc_pool_upgrade(zfs_cmd_t *zc)
1700 {
1701 spa_t *spa;
1702 int error;
1703
1704 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1705 return (error);
1706
1707 if (zc->zc_cookie < spa_version(spa) ||
1708 !SPA_VERSION_IS_SUPPORTED(zc->zc_cookie)) {
1709 spa_close(spa, FTAG);
1710 return (SET_ERROR(EINVAL));
1711 }
1712
1713 spa_upgrade(spa, zc->zc_cookie);
1714 spa_close(spa, FTAG);
1715
1716 return (error);
1717 }
1718
1719 static int
1720 zfs_ioc_pool_get_history(zfs_cmd_t *zc)
1721 {
1722 spa_t *spa;
1723 char *hist_buf;
1724 uint64_t size;
1725 int error;
1726
1727 if ((size = zc->zc_history_len) == 0)
1728 return (SET_ERROR(EINVAL));
1729
1730 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1731 return (error);
1732
1733 if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) {
1734 spa_close(spa, FTAG);
1735 return (SET_ERROR(ENOTSUP));
1736 }
1737
1738 hist_buf = kmem_alloc(size, KM_SLEEP);
1739 if ((error = spa_history_get(spa, &zc->zc_history_offset,
1740 &zc->zc_history_len, hist_buf)) == 0) {
1741 error = ddi_copyout(hist_buf,
1742 (void *)(uintptr_t)zc->zc_history,
1743 zc->zc_history_len, zc->zc_iflags);
1744 }
1745
1746 spa_close(spa, FTAG);
1747 kmem_free(hist_buf, size);
1748 return (error);
1749 }
1750
1751 static int
1752 zfs_ioc_pool_reguid(zfs_cmd_t *zc)
1753 {
1754 spa_t *spa;
1755 int error;
1756
1757 error = spa_open(zc->zc_name, &spa, FTAG);
1758 if (error == 0) {
1759 error = spa_change_guid(spa);
1760 spa_close(spa, FTAG);
1761 }
1762 return (error);
1763 }
1764
1765 static int
1766 zfs_ioc_dsobj_to_dsname(zfs_cmd_t *zc)
1767 {
1768 return (dsl_dsobj_to_dsname(zc->zc_name, zc->zc_obj, zc->zc_value));
1769 }
1770
1771 /*
1772 * inputs:
1773 * zc_name name of filesystem
1774 * zc_obj object to find
1775 *
1776 * outputs:
1777 * zc_value name of object
1778 */
1779 static int
1780 zfs_ioc_obj_to_path(zfs_cmd_t *zc)
1781 {
1782 objset_t *os;
1783 int error;
1784
1785 /* XXX reading from objset not owned */
1786 if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os)) != 0)
1787 return (error);
1788 if (dmu_objset_type(os) != DMU_OST_ZFS) {
1789 dmu_objset_rele(os, FTAG);
1790 return (SET_ERROR(EINVAL));
1791 }
1792 error = zfs_obj_to_path(os, zc->zc_obj, zc->zc_value,
1793 sizeof (zc->zc_value));
1794 dmu_objset_rele(os, FTAG);
1795
1796 return (error);
1797 }
1798
1799 /*
1800 * inputs:
1801 * zc_name name of filesystem
1802 * zc_obj object to find
1803 *
1804 * outputs:
1805 * zc_stat stats on object
1806 * zc_value path to object
1807 */
1808 static int
1809 zfs_ioc_obj_to_stats(zfs_cmd_t *zc)
1810 {
1811 objset_t *os;
1812 int error;
1813
1814 /* XXX reading from objset not owned */
1815 if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os)) != 0)
1816 return (error);
1817 if (dmu_objset_type(os) != DMU_OST_ZFS) {
1818 dmu_objset_rele(os, FTAG);
1819 return (SET_ERROR(EINVAL));
1820 }
1821 error = zfs_obj_to_stats(os, zc->zc_obj, &zc->zc_stat, zc->zc_value,
1822 sizeof (zc->zc_value));
1823 dmu_objset_rele(os, FTAG);
1824
1825 return (error);
1826 }
1827
1828 static int
1829 zfs_ioc_vdev_add(zfs_cmd_t *zc)
1830 {
1831 spa_t *spa;
1832 int error;
1833 nvlist_t *config, **l2cache, **spares;
1834 uint_t nl2cache = 0, nspares = 0;
1835
1836 error = spa_open(zc->zc_name, &spa, FTAG);
1837 if (error != 0)
1838 return (error);
1839
1840 error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1841 zc->zc_iflags, &config);
1842 (void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_L2CACHE,
1843 &l2cache, &nl2cache);
1844
1845 (void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_SPARES,
1846 &spares, &nspares);
1847
1848 /*
1849 * A root pool with concatenated devices is not supported.
1850 * Thus, can not add a device to a root pool.
1851 *
1852 * Intent log device can not be added to a rootpool because
1853 * during mountroot, zil is replayed, a seperated log device
1854 * can not be accessed during the mountroot time.
1855 *
1856 * l2cache and spare devices are ok to be added to a rootpool.
1857 */
1858 if (spa_bootfs(spa) != 0 && nl2cache == 0 && nspares == 0) {
1859 nvlist_free(config);
1860 spa_close(spa, FTAG);
1861 return (SET_ERROR(EDOM));
1862 }
1863
1864 if (error == 0) {
1865 error = spa_vdev_add(spa, config);
1866 nvlist_free(config);
1867 }
1868 spa_close(spa, FTAG);
1869 return (error);
1870 }
1871
1872 /*
1873 * inputs:
1874 * zc_name name of the pool
1875 * zc_nvlist_conf nvlist of devices to remove
1876 * zc_cookie to stop the remove?
1877 */
1878 static int
1879 zfs_ioc_vdev_remove(zfs_cmd_t *zc)
1880 {
1881 spa_t *spa;
1882 int error;
1883
1884 error = spa_open(zc->zc_name, &spa, FTAG);
1885 if (error != 0)
1886 return (error);
1887 error = spa_vdev_remove(spa, zc->zc_guid, B_FALSE);
1888 spa_close(spa, FTAG);
1889 return (error);
1890 }
1891
1892 static int
1893 zfs_ioc_vdev_set_state(zfs_cmd_t *zc)
1894 {
1895 spa_t *spa;
1896 int error;
1897 vdev_state_t newstate = VDEV_STATE_UNKNOWN;
1898
1899 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1900 return (error);
1901 switch (zc->zc_cookie) {
1902 case VDEV_STATE_ONLINE:
1903 error = vdev_online(spa, zc->zc_guid, zc->zc_obj, &newstate);
1904 break;
1905
1906 case VDEV_STATE_OFFLINE:
1907 error = vdev_offline(spa, zc->zc_guid, zc->zc_obj);
1908 break;
1909
1910 case VDEV_STATE_FAULTED:
1911 if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED &&
1912 zc->zc_obj != VDEV_AUX_EXTERNAL)
1913 zc->zc_obj = VDEV_AUX_ERR_EXCEEDED;
1914
1915 error = vdev_fault(spa, zc->zc_guid, zc->zc_obj);
1916 break;
1917
1918 case VDEV_STATE_DEGRADED:
1919 if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED &&
1920 zc->zc_obj != VDEV_AUX_EXTERNAL)
1921 zc->zc_obj = VDEV_AUX_ERR_EXCEEDED;
1922
1923 error = vdev_degrade(spa, zc->zc_guid, zc->zc_obj);
1924 break;
1925
1926 default:
1927 error = SET_ERROR(EINVAL);
1928 }
1929 zc->zc_cookie = newstate;
1930 spa_close(spa, FTAG);
1931 return (error);
1932 }
1933
1934 static int
1935 zfs_ioc_vdev_attach(zfs_cmd_t *zc)
1936 {
1937 spa_t *spa;
1938 int replacing = zc->zc_cookie;
1939 nvlist_t *config;
1940 int error;
1941
1942 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1943 return (error);
1944
1945 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1946 zc->zc_iflags, &config)) == 0) {
1947 error = spa_vdev_attach(spa, zc->zc_guid, config, replacing);
1948 nvlist_free(config);
1949 }
1950
1951 spa_close(spa, FTAG);
1952 return (error);
1953 }
1954
1955 static int
1956 zfs_ioc_vdev_detach(zfs_cmd_t *zc)
1957 {
1958 spa_t *spa;
1959 int error;
1960
1961 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1962 return (error);
1963
1964 error = spa_vdev_detach(spa, zc->zc_guid, 0, B_FALSE);
1965
1966 spa_close(spa, FTAG);
1967 return (error);
1968 }
1969
1970 static int
1971 zfs_ioc_vdev_split(zfs_cmd_t *zc)
1972 {
1973 spa_t *spa;
1974 nvlist_t *config, *props = NULL;
1975 int error;
1976 boolean_t exp = !!(zc->zc_cookie & ZPOOL_EXPORT_AFTER_SPLIT);
1977
1978 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1979 return (error);
1980
1981 if (error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1982 zc->zc_iflags, &config)) {
1983 spa_close(spa, FTAG);
1984 return (error);
1985 }
1986
1987 if (zc->zc_nvlist_src_size != 0 && (error =
1988 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1989 zc->zc_iflags, &props))) {
1990 spa_close(spa, FTAG);
1991 nvlist_free(config);
1992 return (error);
1993 }
1994
1995 error = spa_vdev_split_mirror(spa, zc->zc_string, config, props, exp);
1996
1997 spa_close(spa, FTAG);
1998
1999 nvlist_free(config);
2000 nvlist_free(props);
2001
2002 return (error);
2003 }
2004
2005 static int
2006 zfs_ioc_vdev_setpath(zfs_cmd_t *zc)
2007 {
2008 spa_t *spa;
2009 char *path = zc->zc_value;
2010 uint64_t guid = zc->zc_guid;
2011 int error;
2012
2013 error = spa_open(zc->zc_name, &spa, FTAG);
2014 if (error != 0)
2015 return (error);
2016
2017 error = spa_vdev_setpath(spa, guid, path);
2018 spa_close(spa, FTAG);
2019 return (error);
2020 }
2021
2022 static int
2023 zfs_ioc_vdev_setfru(zfs_cmd_t *zc)
2024 {
2025 spa_t *spa;
2026 char *fru = zc->zc_value;
2027 uint64_t guid = zc->zc_guid;
2028 int error;
2029
2030 error = spa_open(zc->zc_name, &spa, FTAG);
2031 if (error != 0)
2032 return (error);
2033
2034 error = spa_vdev_setfru(spa, guid, fru);
2035 spa_close(spa, FTAG);
2036 return (error);
2037 }
2038
2039 static int
2040 zfs_ioc_objset_stats_impl(zfs_cmd_t *zc, objset_t *os)
2041 {
2042 int error = 0;
2043 nvlist_t *nv;
2044
2045 dmu_objset_fast_stat(os, &zc->zc_objset_stats);
2046
2047 if (zc->zc_nvlist_dst != 0 &&
2048 (error = dsl_prop_get_all(os, &nv)) == 0) {
2049 dmu_objset_stats(os, nv);
2050 /*
2051 * NB: zvol_get_stats() will read the objset contents,
2052 * which we aren't supposed to do with a
2053 * DS_MODE_USER hold, because it could be
2054 * inconsistent. So this is a bit of a workaround...
2055 * XXX reading with out owning
2056 */
2057 if (!zc->zc_objset_stats.dds_inconsistent &&
2058 dmu_objset_type(os) == DMU_OST_ZVOL) {
2059 error = zvol_get_stats(os, nv);
2060 if (error == EIO)
2061 return (error);
2062 VERIFY0(error);
2063 }
2064 error = put_nvlist(zc, nv);
2065 nvlist_free(nv);
2066 }
2067
2068 return (error);
2069 }
2070
2071 /*
2072 * inputs:
2073 * zc_name name of filesystem
2074 * zc_nvlist_dst_size size of buffer for property nvlist
2075 *
2076 * outputs:
2077 * zc_objset_stats stats
2078 * zc_nvlist_dst property nvlist
2079 * zc_nvlist_dst_size size of property nvlist
2080 */
2081 static int
2082 zfs_ioc_objset_stats(zfs_cmd_t *zc)
2083 {
2084 objset_t *os;
2085 int error;
2086
2087 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
2088 if (error == 0) {
2089 error = zfs_ioc_objset_stats_impl(zc, os);
2090 dmu_objset_rele(os, FTAG);
2091 }
2092
2093 return (error);
2094 }
2095
2096 /*
2097 * inputs:
2098 * zc_name name of filesystem
2099 * zc_nvlist_dst_size size of buffer for property nvlist
2100 *
2101 * outputs:
2102 * zc_nvlist_dst received property nvlist
2103 * zc_nvlist_dst_size size of received property nvlist
2104 *
2105 * Gets received properties (distinct from local properties on or after
2106 * SPA_VERSION_RECVD_PROPS) for callers who want to differentiate received from
2107 * local property values.
2108 */
2109 static int
2110 zfs_ioc_objset_recvd_props(zfs_cmd_t *zc)
2111 {
2112 int error = 0;
2113 nvlist_t *nv;
2114
2115 /*
2116 * Without this check, we would return local property values if the
2117 * caller has not already received properties on or after
2118 * SPA_VERSION_RECVD_PROPS.
2119 */
2120 if (!dsl_prop_get_hasrecvd(zc->zc_name))
2121 return (SET_ERROR(ENOTSUP));
2122
2123 if (zc->zc_nvlist_dst != 0 &&
2124 (error = dsl_prop_get_received(zc->zc_name, &nv)) == 0) {
2125 error = put_nvlist(zc, nv);
2126 nvlist_free(nv);
2127 }
2128
2129 return (error);
2130 }
2131
2132 static int
2133 nvl_add_zplprop(objset_t *os, nvlist_t *props, zfs_prop_t prop)
2134 {
2135 uint64_t value;
2136 int error;
2137
2138 /*
2139 * zfs_get_zplprop() will either find a value or give us
2140 * the default value (if there is one).
2141 */
2142 if ((error = zfs_get_zplprop(os, prop, &value)) != 0)
2143 return (error);
2144 VERIFY(nvlist_add_uint64(props, zfs_prop_to_name(prop), value) == 0);
2145 return (0);
2146 }
2147
2148 /*
2149 * inputs:
2150 * zc_name name of filesystem
2151 * zc_nvlist_dst_size size of buffer for zpl property nvlist
2152 *
2153 * outputs:
2154 * zc_nvlist_dst zpl property nvlist
2155 * zc_nvlist_dst_size size of zpl property nvlist
2156 */
2157 static int
2158 zfs_ioc_objset_zplprops(zfs_cmd_t *zc)
2159 {
2160 objset_t *os;
2161 int err;
2162
2163 /* XXX reading without owning */
2164 if (err = dmu_objset_hold(zc->zc_name, FTAG, &os))
2165 return (err);
2166
2167 dmu_objset_fast_stat(os, &zc->zc_objset_stats);
2168
2169 /*
2170 * NB: nvl_add_zplprop() will read the objset contents,
2171 * which we aren't supposed to do with a DS_MODE_USER
2172 * hold, because it could be inconsistent.
2173 */
2174 if (zc->zc_nvlist_dst != NULL &&
2175 !zc->zc_objset_stats.dds_inconsistent &&
2176 dmu_objset_type(os) == DMU_OST_ZFS) {
2177 nvlist_t *nv;
2178
2179 VERIFY(nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2180 if ((err = nvl_add_zplprop(os, nv, ZFS_PROP_VERSION)) == 0 &&
2181 (err = nvl_add_zplprop(os, nv, ZFS_PROP_NORMALIZE)) == 0 &&
2182 (err = nvl_add_zplprop(os, nv, ZFS_PROP_UTF8ONLY)) == 0 &&
2183 (err = nvl_add_zplprop(os, nv, ZFS_PROP_CASE)) == 0)
2184 err = put_nvlist(zc, nv);
2185 nvlist_free(nv);
2186 } else {
2187 err = SET_ERROR(ENOENT);
2188 }
2189 dmu_objset_rele(os, FTAG);
2190 return (err);
2191 }
2192
2193 static boolean_t
2194 dataset_name_hidden(const char *name)
2195 {
2196 /*
2197 * Skip over datasets that are not visible in this zone,
2198 * internal datasets (which have a $ in their name), and
2199 * temporary datasets (which have a % in their name).
2200 */
2201 if (strchr(name, '$') != NULL)
2202 return (B_TRUE);
2203 if (strchr(name, '%') != NULL)
2204 return (B_TRUE);
2205 if (!INGLOBALZONE(curproc) && !zone_dataset_visible(name, NULL))
2206 return (B_TRUE);
2207 return (B_FALSE);
2208 }
2209
2210 /*
2211 * inputs:
2212 * zc_name name of filesystem
2213 * zc_cookie zap cursor
2214 * zc_nvlist_dst_size size of buffer for property nvlist
2215 *
2216 * outputs:
2217 * zc_name name of next filesystem
2218 * zc_cookie zap cursor
2219 * zc_objset_stats stats
2220 * zc_nvlist_dst property nvlist
2221 * zc_nvlist_dst_size size of property nvlist
2222 */
2223 static int
2224 zfs_ioc_dataset_list_next(zfs_cmd_t *zc)
2225 {
2226 objset_t *os;
2227 int error;
2228 char *p;
2229 size_t orig_len = strlen(zc->zc_name);
2230
2231 top:
2232 if (error = dmu_objset_hold(zc->zc_name, FTAG, &os)) {
2233 if (error == ENOENT)
2234 error = SET_ERROR(ESRCH);
2235 return (error);
2236 }
2237
2238 p = strrchr(zc->zc_name, '/');
2239 if (p == NULL || p[1] != '\0')
2240 (void) strlcat(zc->zc_name, "/", sizeof (zc->zc_name));
2241 p = zc->zc_name + strlen(zc->zc_name);
2242
2243 do {
2244 error = dmu_dir_list_next(os,
2245 sizeof (zc->zc_name) - (p - zc->zc_name), p,
2246 NULL, &zc->zc_cookie);
2247 if (error == ENOENT)
2248 error = SET_ERROR(ESRCH);
2249 } while (error == 0 && dataset_name_hidden(zc->zc_name));
2250 dmu_objset_rele(os, FTAG);
2251
2252 /*
2253 * If it's an internal dataset (ie. with a '$' in its name),
2254 * don't try to get stats for it, otherwise we'll return ENOENT.
2255 */
2256 if (error == 0 && strchr(zc->zc_name, '$') == NULL) {
2257 error = zfs_ioc_objset_stats(zc); /* fill in the stats */
2258 if (error == ENOENT) {
2259 /* We lost a race with destroy, get the next one. */
2260 zc->zc_name[orig_len] = '\0';
2261 goto top;
2262 }
2263 }
2264 return (error);
2265 }
2266
2267 /*
2268 * inputs:
2269 * zc_name name of filesystem
2270 * zc_cookie zap cursor
2271 * zc_nvlist_dst_size size of buffer for property nvlist
2272 * zc_simple when set, only name is requested
2273 *
2274 * outputs:
2275 * zc_name name of next snapshot
2276 * zc_objset_stats stats
2277 * zc_nvlist_dst property nvlist
2278 * zc_nvlist_dst_size size of property nvlist
2279 */
2280 static int
2281 zfs_ioc_snapshot_list_next(zfs_cmd_t *zc)
2282 {
2283 objset_t *os;
2284 int error;
2285
2286 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
2287 if (error != 0) {
2288 return (error == ENOENT ? ESRCH : error);
2289 }
2290
2291 /*
2292 * A dataset name of maximum length cannot have any snapshots,
2293 * so exit immediately.
2294 */
2295 if (strlcat(zc->zc_name, "@", sizeof (zc->zc_name)) >=
2296 ZFS_MAX_DATASET_NAME_LEN) {
2297 dmu_objset_rele(os, FTAG);
2298 return (SET_ERROR(ESRCH));
2299 }
2300
2301 error = dmu_snapshot_list_next(os,
2302 sizeof (zc->zc_name) - strlen(zc->zc_name),
2303 zc->zc_name + strlen(zc->zc_name), &zc->zc_obj, &zc->zc_cookie,
2304 NULL);
2305
2306 if (error == 0 && !zc->zc_simple) {
2307 dsl_dataset_t *ds;
2308 dsl_pool_t *dp = os->os_dsl_dataset->ds_dir->dd_pool;
2309
2310 error = dsl_dataset_hold_obj(dp, zc->zc_obj, FTAG, &ds);
2311 if (error == 0) {
2312 objset_t *ossnap;
2313
2314 error = dmu_objset_from_ds(ds, &ossnap);
2315 if (error == 0)
2316 error = zfs_ioc_objset_stats_impl(zc, ossnap);
2317 dsl_dataset_rele(ds, FTAG);
2318 }
2319 } else if (error == ENOENT) {
2320 error = SET_ERROR(ESRCH);
2321 }
2322
2323 dmu_objset_rele(os, FTAG);
2324 /* if we failed, undo the @ that we tacked on to zc_name */
2325 if (error != 0)
2326 *strchr(zc->zc_name, '@') = '\0';
2327 return (error);
2328 }
2329
2330 static int
2331 zfs_prop_set_userquota(const char *dsname, nvpair_t *pair)
2332 {
2333 const char *propname = nvpair_name(pair);
2334 uint64_t *valary;
2335 unsigned int vallen;
2336 const char *domain;
2337 char *dash;
2338 zfs_userquota_prop_t type;
2339 uint64_t rid;
2340 uint64_t quota;
2341 zfsvfs_t *zfsvfs;
2342 int err;
2343
2344 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2345 nvlist_t *attrs;
2346 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
2347 if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2348 &pair) != 0)
2349 return (SET_ERROR(EINVAL));
2350 }
2351
2352 /*
2353 * A correctly constructed propname is encoded as
2354 * userquota@<rid>-<domain>.
2355 */
2356 if ((dash = strchr(propname, '-')) == NULL ||
2357 nvpair_value_uint64_array(pair, &valary, &vallen) != 0 ||
2358 vallen != 3)
2359 return (SET_ERROR(EINVAL));
2360
2361 domain = dash + 1;
2362 type = valary[0];
2363 rid = valary[1];
2364 quota = valary[2];
2365
2366 err = zfsvfs_hold(dsname, FTAG, &zfsvfs, B_FALSE);
2367 if (err == 0) {
2368 err = zfs_set_userquota(zfsvfs, type, domain, rid, quota);
2369 zfsvfs_rele(zfsvfs, FTAG);
2370 }
2371
2372 return (err);
2373 }
2374
2375 /*
2376 * If the named property is one that has a special function to set its value,
2377 * return 0 on success and a positive error code on failure; otherwise if it is
2378 * not one of the special properties handled by this function, return -1.
2379 *
2380 * XXX: It would be better for callers of the property interface if we handled
2381 * these special cases in dsl_prop.c (in the dsl layer).
2382 */
2383 static int
2384 zfs_prop_set_special(const char *dsname, zprop_source_t source,
2385 nvpair_t *pair)
2386 {
2387 const char *propname = nvpair_name(pair);
2388 zfs_prop_t prop = zfs_name_to_prop(propname);
2389 uint64_t intval;
2390 int err = -1;
2391
2392 if (prop == ZPROP_INVAL) {
2393 if (zfs_prop_userquota(propname))
2394 return (zfs_prop_set_userquota(dsname, pair));
2395 return (-1);
2396 }
2397
2398 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2399 nvlist_t *attrs;
2400 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
2401 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2402 &pair) == 0);
2403 }
2404
2405 if (zfs_prop_get_type(prop) == PROP_TYPE_STRING)
2406 return (-1);
2407
2408 VERIFY(0 == nvpair_value_uint64(pair, &intval));
2409
2410 switch (prop) {
2411 case ZFS_PROP_QUOTA:
2412 err = dsl_dir_set_quota(dsname, source, intval);
2413 break;
2414 case ZFS_PROP_REFQUOTA:
2415 err = dsl_dataset_set_refquota(dsname, source, intval);
2416 break;
2417 case ZFS_PROP_FILESYSTEM_LIMIT:
2418 case ZFS_PROP_SNAPSHOT_LIMIT:
2419 if (intval == UINT64_MAX) {
2420 /* clearing the limit, just do it */
2421 err = 0;
2422 } else {
2423 err = dsl_dir_activate_fs_ss_limit(dsname);
2424 }
2425 /*
2426 * Set err to -1 to force the zfs_set_prop_nvlist code down the
2427 * default path to set the value in the nvlist.
2428 */
2429 if (err == 0)
2430 err = -1;
2431 break;
2432 case ZFS_PROP_RESERVATION:
2433 err = dsl_dir_set_reservation(dsname, source, intval);
2434 break;
2435 case ZFS_PROP_REFRESERVATION:
2436 err = dsl_dataset_set_refreservation(dsname, source, intval);
2437 break;
2438 case ZFS_PROP_VOLSIZE:
2439 err = zvol_set_volsize(dsname, intval);
2440 break;
2441 case ZFS_PROP_VERSION:
2442 {
2443 zfsvfs_t *zfsvfs;
2444
2445 if ((err = zfsvfs_hold(dsname, FTAG, &zfsvfs, B_TRUE)) != 0)
2446 break;
2447
2448 err = zfs_set_version(zfsvfs, intval);
2449 zfsvfs_rele(zfsvfs, FTAG);
2450
2451 if (err == 0 && intval >= ZPL_VERSION_USERSPACE) {
2452 zfs_cmd_t *zc;
2453
2454 zc = kmem_zalloc(sizeof (zfs_cmd_t), KM_SLEEP);
2455 (void) strcpy(zc->zc_name, dsname);
2456 (void) zfs_ioc_userspace_upgrade(zc);
2457 kmem_free(zc, sizeof (zfs_cmd_t));
2458 }
2459 break;
2460 }
2461 default:
2462 err = -1;
2463 }
2464
2465 return (err);
2466 }
2467
2468 /*
2469 * This function is best effort. If it fails to set any of the given properties,
2470 * it continues to set as many as it can and returns the last error
2471 * encountered. If the caller provides a non-NULL errlist, it will be filled in
2472 * with the list of names of all the properties that failed along with the
2473 * corresponding error numbers.
2474 *
2475 * If every property is set successfully, zero is returned and errlist is not
2476 * modified.
2477 */
2478 int
2479 zfs_set_prop_nvlist(const char *dsname, zprop_source_t source, nvlist_t *nvl,
2480 nvlist_t *errlist)
2481 {
2482 nvpair_t *pair;
2483 nvpair_t *propval;
2484 int rv = 0;
2485 uint64_t intval;
2486 char *strval;
2487 nvlist_t *genericnvl = fnvlist_alloc();
2488 nvlist_t *retrynvl = fnvlist_alloc();
2489
2490 retry:
2491 pair = NULL;
2492 while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) {
2493 const char *propname = nvpair_name(pair);
2494 zfs_prop_t prop = zfs_name_to_prop(propname);
2495 int err = 0;
2496
2497 /* decode the property value */
2498 propval = pair;
2499 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2500 nvlist_t *attrs;
2501 attrs = fnvpair_value_nvlist(pair);
2502 if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2503 &propval) != 0)
2504 err = SET_ERROR(EINVAL);
2505 }
2506
2507 /* Validate value type */
2508 if (err == 0 && prop == ZPROP_INVAL) {
2509 if (zfs_prop_user(propname)) {
2510 if (nvpair_type(propval) != DATA_TYPE_STRING)
2511 err = SET_ERROR(EINVAL);
2512 } else if (zfs_prop_userquota(propname)) {
2513 if (nvpair_type(propval) !=
2514 DATA_TYPE_UINT64_ARRAY)
2515 err = SET_ERROR(EINVAL);
2516 } else {
2517 err = SET_ERROR(EINVAL);
2518 }
2519 } else if (err == 0) {
2520 if (nvpair_type(propval) == DATA_TYPE_STRING) {
2521 if (zfs_prop_get_type(prop) != PROP_TYPE_STRING)
2522 err = SET_ERROR(EINVAL);
2523 } else if (nvpair_type(propval) == DATA_TYPE_UINT64) {
2524 const char *unused;
2525
2526 intval = fnvpair_value_uint64(propval);
2527
2528 switch (zfs_prop_get_type(prop)) {
2529 case PROP_TYPE_NUMBER:
2530 break;
2531 case PROP_TYPE_STRING:
2532 err = SET_ERROR(EINVAL);
2533 break;
2534 case PROP_TYPE_INDEX:
2535 if (zfs_prop_index_to_string(prop,
2536 intval, &unused) != 0)
2537 err = SET_ERROR(EINVAL);
2538 break;
2539 default:
2540 cmn_err(CE_PANIC,
2541 "unknown property type");
2542 }
2543 } else {
2544 err = SET_ERROR(EINVAL);
2545 }
2546 }
2547
2548 /* Validate permissions */
2549 if (err == 0)
2550 err = zfs_check_settable(dsname, pair, CRED());
2551
2552 if (err == 0) {
2553 err = zfs_prop_set_special(dsname, source, pair);
2554 if (err == -1) {
2555 /*
2556 * For better performance we build up a list of
2557 * properties to set in a single transaction.
2558 */
2559 err = nvlist_add_nvpair(genericnvl, pair);
2560 } else if (err != 0 && nvl != retrynvl) {
2561 /*
2562 * This may be a spurious error caused by
2563 * receiving quota and reservation out of order.
2564 * Try again in a second pass.
2565 */
2566 err = nvlist_add_nvpair(retrynvl, pair);
2567 }
2568 }
2569
2570 if (err != 0) {
2571 if (errlist != NULL)
2572 fnvlist_add_int32(errlist, propname, err);
2573 rv = err;
2574 }
2575 }
2576
2577 if (nvl != retrynvl && !nvlist_empty(retrynvl)) {
2578 nvl = retrynvl;
2579 goto retry;
2580 }
2581
2582 if (!nvlist_empty(genericnvl) &&
2583 dsl_props_set(dsname, source, genericnvl) != 0) {
2584 /*
2585 * If this fails, we still want to set as many properties as we
2586 * can, so try setting them individually.
2587 */
2588 pair = NULL;
2589 while ((pair = nvlist_next_nvpair(genericnvl, pair)) != NULL) {
2590 const char *propname = nvpair_name(pair);
2591 int err = 0;
2592
2593 propval = pair;
2594 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2595 nvlist_t *attrs;
2596 attrs = fnvpair_value_nvlist(pair);
2597 propval = fnvlist_lookup_nvpair(attrs,
2598 ZPROP_VALUE);
2599 }
2600
2601 if (nvpair_type(propval) == DATA_TYPE_STRING) {
2602 strval = fnvpair_value_string(propval);
2603 err = dsl_prop_set_string(dsname, propname,
2604 source, strval);
2605 } else {
2606 intval = fnvpair_value_uint64(propval);
2607 err = dsl_prop_set_int(dsname, propname, source,
2608 intval);
2609 }
2610
2611 if (err != 0) {
2612 if (errlist != NULL) {
2613 fnvlist_add_int32(errlist, propname,
2614 err);
2615 }
2616 rv = err;
2617 }
2618 }
2619 }
2620 nvlist_free(genericnvl);
2621 nvlist_free(retrynvl);
2622
2623 return (rv);
2624 }
2625
2626 /*
2627 * Check that all the properties are valid user properties.
2628 */
2629 static int
2630 zfs_check_userprops(const char *fsname, nvlist_t *nvl)
2631 {
2632 nvpair_t *pair = NULL;
2633 int error = 0;
2634
2635 while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) {
2636 const char *propname = nvpair_name(pair);
2637
2638 if (!zfs_prop_user(propname) ||
2639 nvpair_type(pair) != DATA_TYPE_STRING)
2640 return (SET_ERROR(EINVAL));
2641
2642 if (error = zfs_secpolicy_write_perms(fsname,
2643 ZFS_DELEG_PERM_USERPROP, CRED()))
2644 return (error);
2645
2646 if (strlen(propname) >= ZAP_MAXNAMELEN)
2647 return (SET_ERROR(ENAMETOOLONG));
2648
2649 if (strlen(fnvpair_value_string(pair)) >= ZAP_MAXVALUELEN)
2650 return (E2BIG);
2651 }
2652 return (0);
2653 }
2654
2655 static void
2656 props_skip(nvlist_t *props, nvlist_t *skipped, nvlist_t **newprops)
2657 {
2658 nvpair_t *pair;
2659
2660 VERIFY(nvlist_alloc(newprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2661
2662 pair = NULL;
2663 while ((pair = nvlist_next_nvpair(props, pair)) != NULL) {
2664 if (nvlist_exists(skipped, nvpair_name(pair)))
2665 continue;
2666
2667 VERIFY(nvlist_add_nvpair(*newprops, pair) == 0);
2668 }
2669 }
2670
2671 static int
2672 clear_received_props(const char *dsname, nvlist_t *props,
2673 nvlist_t *skipped)
2674 {
2675 int err = 0;
2676 nvlist_t *cleared_props = NULL;
2677 props_skip(props, skipped, &cleared_props);
2678 if (!nvlist_empty(cleared_props)) {
2679 /*
2680 * Acts on local properties until the dataset has received
2681 * properties at least once on or after SPA_VERSION_RECVD_PROPS.
2682 */
2683 zprop_source_t flags = (ZPROP_SRC_NONE |
2684 (dsl_prop_get_hasrecvd(dsname) ? ZPROP_SRC_RECEIVED : 0));
2685 err = zfs_set_prop_nvlist(dsname, flags, cleared_props, NULL);
2686 }
2687 nvlist_free(cleared_props);
2688 return (err);
2689 }
2690
2691 /*
2692 * inputs:
2693 * zc_name name of filesystem
2694 * zc_value name of property to set
2695 * zc_nvlist_src{_size} nvlist of properties to apply
2696 * zc_cookie received properties flag
2697 *
2698 * outputs:
2699 * zc_nvlist_dst{_size} error for each unapplied received property
2700 */
2701 static int
2702 zfs_ioc_set_prop(zfs_cmd_t *zc)
2703 {
2704 nvlist_t *nvl;
2705 boolean_t received = zc->zc_cookie;
2706 zprop_source_t source = (received ? ZPROP_SRC_RECEIVED :
2707 ZPROP_SRC_LOCAL);
2708 nvlist_t *errors;
2709 int error;
2710
2711 if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2712 zc->zc_iflags, &nvl)) != 0)
2713 return (error);
2714
2715 if (received) {
2716 nvlist_t *origprops;
2717
2718 if (dsl_prop_get_received(zc->zc_name, &origprops) == 0) {
2719 (void) clear_received_props(zc->zc_name,
2720 origprops, nvl);
2721 nvlist_free(origprops);
2722 }
2723
2724 error = dsl_prop_set_hasrecvd(zc->zc_name);
2725 }
2726
2727 errors = fnvlist_alloc();
2728 if (error == 0)
2729 error = zfs_set_prop_nvlist(zc->zc_name, source, nvl, errors);
2730
2731 if (zc->zc_nvlist_dst != NULL && errors != NULL) {
2732 (void) put_nvlist(zc, errors);
2733 }
2734
2735 nvlist_free(errors);
2736 nvlist_free(nvl);
2737 return (error);
2738 }
2739
2740 /*
2741 * inputs:
2742 * zc_name name of filesystem
2743 * zc_value name of property to inherit
2744 * zc_cookie revert to received value if TRUE
2745 *
2746 * outputs: none
2747 */
2748 static int
2749 zfs_ioc_inherit_prop(zfs_cmd_t *zc)
2750 {
2751 const char *propname = zc->zc_value;
2752 zfs_prop_t prop = zfs_name_to_prop(propname);
2753 boolean_t received = zc->zc_cookie;
2754 zprop_source_t source = (received
2755 ? ZPROP_SRC_NONE /* revert to received value, if any */
2756 : ZPROP_SRC_INHERITED); /* explicitly inherit */
2757
2758 if (received) {
2759 nvlist_t *dummy;
2760 nvpair_t *pair;
2761 zprop_type_t type;
2762 int err;
2763
2764 /*
2765 * zfs_prop_set_special() expects properties in the form of an
2766 * nvpair with type info.
2767 */
2768 if (prop == ZPROP_INVAL) {
2769 if (!zfs_prop_user(propname))
2770 return (SET_ERROR(EINVAL));
2771
2772 type = PROP_TYPE_STRING;
2773 } else if (prop == ZFS_PROP_VOLSIZE ||
2774 prop == ZFS_PROP_VERSION) {
2775 return (SET_ERROR(EINVAL));
2776 } else {
2777 type = zfs_prop_get_type(prop);
2778 }
2779
2780 VERIFY(nvlist_alloc(&dummy, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2781
2782 switch (type) {
2783 case PROP_TYPE_STRING:
2784 VERIFY(0 == nvlist_add_string(dummy, propname, ""));
2785 break;
2786 case PROP_TYPE_NUMBER:
2787 case PROP_TYPE_INDEX:
2788 VERIFY(0 == nvlist_add_uint64(dummy, propname, 0));
2789 break;
2790 default:
2791 nvlist_free(dummy);
2792 return (SET_ERROR(EINVAL));
2793 }
2794
2795 pair = nvlist_next_nvpair(dummy, NULL);
2796 err = zfs_prop_set_special(zc->zc_name, source, pair);
2797 nvlist_free(dummy);
2798 if (err != -1)
2799 return (err); /* special property already handled */
2800 } else {
2801 /*
2802 * Only check this in the non-received case. We want to allow
2803 * 'inherit -S' to revert non-inheritable properties like quota
2804 * and reservation to the received or default values even though
2805 * they are not considered inheritable.
2806 */
2807 if (prop != ZPROP_INVAL && !zfs_prop_inheritable(prop))
2808 return (SET_ERROR(EINVAL));
2809 }
2810
2811 /* property name has been validated by zfs_secpolicy_inherit_prop() */
2812 return (dsl_prop_inherit(zc->zc_name, zc->zc_value, source));
2813 }
2814
2815 static int
2816 zfs_ioc_pool_set_props(zfs_cmd_t *zc)
2817 {
2818 nvlist_t *props;
2819 spa_t *spa;
2820 int error;
2821 nvpair_t *pair;
2822
2823 if (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2824 zc->zc_iflags, &props))
2825 return (error);
2826
2827 /*
2828 * If the only property is the configfile, then just do a spa_lookup()
2829 * to handle the faulted case.
2830 */
2831 pair = nvlist_next_nvpair(props, NULL);
2832 if (pair != NULL && strcmp(nvpair_name(pair),
2833 zpool_prop_to_name(ZPOOL_PROP_CACHEFILE)) == 0 &&
2834 nvlist_next_nvpair(props, pair) == NULL) {
2835 mutex_enter(&spa_namespace_lock);
2836 if ((spa = spa_lookup(zc->zc_name)) != NULL) {
2837 spa_configfile_set(spa, props, B_FALSE);
2838 spa_config_sync(spa, B_FALSE, B_TRUE);
2839 }
2840 mutex_exit(&spa_namespace_lock);
2841 if (spa != NULL) {
2842 nvlist_free(props);
2843 return (0);
2844 }
2845 }
2846
2847 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) {
2848 nvlist_free(props);
2849 return (error);
2850 }
2851
2852 error = spa_prop_set(spa, props);
2853
2854 nvlist_free(props);
2855 spa_close(spa, FTAG);
2856
2857 return (error);
2858 }
2859
2860 static int
2861 zfs_ioc_pool_get_props(zfs_cmd_t *zc)
2862 {
2863 spa_t *spa;
2864 int error;
2865 nvlist_t *nvp = NULL;
2866
2867 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) {
2868 /*
2869 * If the pool is faulted, there may be properties we can still
2870 * get (such as altroot and cachefile), so attempt to get them
2871 * anyway.
2872 */
2873 mutex_enter(&spa_namespace_lock);
2874 if ((spa = spa_lookup(zc->zc_name)) != NULL)
2875 error = spa_prop_get(spa, &nvp);
2876 mutex_exit(&spa_namespace_lock);
2877 } else {
2878 error = spa_prop_get(spa, &nvp);
2879 spa_close(spa, FTAG);
2880 }
2881
2882 if (error == 0 && zc->zc_nvlist_dst != NULL)
2883 error = put_nvlist(zc, nvp);
2884 else
2885 error = SET_ERROR(EFAULT);
2886
2887 nvlist_free(nvp);
2888 return (error);
2889 }
2890
2891 /*
2892 * inputs:
2893 * zc_name name of filesystem
2894 * zc_nvlist_src{_size} nvlist of delegated permissions
2895 * zc_perm_action allow/unallow flag
2896 *
2897 * outputs: none
2898 */
2899 static int
2900 zfs_ioc_set_fsacl(zfs_cmd_t *zc)
2901 {
2902 int error;
2903 nvlist_t *fsaclnv = NULL;
2904
2905 if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2906 zc->zc_iflags, &fsaclnv)) != 0)
2907 return (error);
2908
2909 /*
2910 * Verify nvlist is constructed correctly
2911 */
2912 if ((error = zfs_deleg_verify_nvlist(fsaclnv)) != 0) {
2913 nvlist_free(fsaclnv);
2914 return (SET_ERROR(EINVAL));
2915 }
2916
2917 /*
2918 * If we don't have PRIV_SYS_MOUNT, then validate
2919 * that user is allowed to hand out each permission in
2920 * the nvlist(s)
2921 */
2922
2923 error = secpolicy_zfs(CRED());
2924 if (error != 0) {
2925 if (zc->zc_perm_action == B_FALSE) {
2926 error = dsl_deleg_can_allow(zc->zc_name,
2927 fsaclnv, CRED());
2928 } else {
2929 error = dsl_deleg_can_unallow(zc->zc_name,
2930 fsaclnv, CRED());
2931 }
2932 }
2933
2934 if (error == 0)
2935 error = dsl_deleg_set(zc->zc_name, fsaclnv, zc->zc_perm_action);
2936
2937 nvlist_free(fsaclnv);
2938 return (error);
2939 }
2940
2941 /*
2942 * inputs:
2943 * zc_name name of filesystem
2944 *
2945 * outputs:
2946 * zc_nvlist_src{_size} nvlist of delegated permissions
2947 */
2948 static int
2949 zfs_ioc_get_fsacl(zfs_cmd_t *zc)
2950 {
2951 nvlist_t *nvp;
2952 int error;
2953
2954 if ((error = dsl_deleg_get(zc->zc_name, &nvp)) == 0) {
2955 error = put_nvlist(zc, nvp);
2956 nvlist_free(nvp);
2957 }
2958
2959 return (error);
2960 }
2961
2962 /*
2963 * Search the vfs list for a specified resource. Returns a pointer to it
2964 * or NULL if no suitable entry is found. The caller of this routine
2965 * is responsible for releasing the returned vfs pointer.
2966 */
2967 static vfs_t *
2968 zfs_get_vfs(const char *resource)
2969 {
2970 struct vfs *vfsp;
2971 struct vfs *vfs_found = NULL;
2972
2973 vfs_list_read_lock();
2974 vfsp = rootvfs;
2975 do {
2976 if (strcmp(refstr_value(vfsp->vfs_resource), resource) == 0) {
2977 VFS_HOLD(vfsp);
2978 vfs_found = vfsp;
2979 break;
2980 }
2981 vfsp = vfsp->vfs_next;
2982 } while (vfsp != rootvfs);
2983 vfs_list_unlock();
2984 return (vfs_found);
2985 }
2986
2987 /* ARGSUSED */
2988 static void
2989 zfs_create_cb(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx)
2990 {
2991 zfs_creat_t *zct = arg;
2992
2993 zfs_create_fs(os, cr, zct->zct_zplprops, tx);
2994 }
2995
2996 #define ZFS_PROP_UNDEFINED ((uint64_t)-1)
2997
2998 /*
2999 * inputs:
3000 * os parent objset pointer (NULL if root fs)
3001 * fuids_ok fuids allowed in this version of the spa?
3002 * sa_ok SAs allowed in this version of the spa?
3003 * createprops list of properties requested by creator
3004 *
3005 * outputs:
3006 * zplprops values for the zplprops we attach to the master node object
3007 * is_ci true if requested file system will be purely case-insensitive
3008 *
3009 * Determine the settings for utf8only, normalization and
3010 * casesensitivity. Specific values may have been requested by the
3011 * creator and/or we can inherit values from the parent dataset. If
3012 * the file system is of too early a vintage, a creator can not
3013 * request settings for these properties, even if the requested
3014 * setting is the default value. We don't actually want to create dsl
3015 * properties for these, so remove them from the source nvlist after
3016 * processing.
3017 */
3018 static int
3019 zfs_fill_zplprops_impl(objset_t *os, uint64_t zplver,
3020 boolean_t fuids_ok, boolean_t sa_ok, nvlist_t *createprops,
3021 nvlist_t *zplprops, boolean_t *is_ci)
3022 {
3023 uint64_t sense = ZFS_PROP_UNDEFINED;
3024 uint64_t norm = ZFS_PROP_UNDEFINED;
3025 uint64_t u8 = ZFS_PROP_UNDEFINED;
3026
3027 ASSERT(zplprops != NULL);
3028
3029 /*
3030 * Pull out creator prop choices, if any.
3031 */
3032 if (createprops) {
3033 (void) nvlist_lookup_uint64(createprops,
3034 zfs_prop_to_name(ZFS_PROP_VERSION), &zplver);
3035 (void) nvlist_lookup_uint64(createprops,
3036 zfs_prop_to_name(ZFS_PROP_NORMALIZE), &norm);
3037 (void) nvlist_remove_all(createprops,
3038 zfs_prop_to_name(ZFS_PROP_NORMALIZE));
3039 (void) nvlist_lookup_uint64(createprops,
3040 zfs_prop_to_name(ZFS_PROP_UTF8ONLY), &u8);
3041 (void) nvlist_remove_all(createprops,
3042 zfs_prop_to_name(ZFS_PROP_UTF8ONLY));
3043 (void) nvlist_lookup_uint64(createprops,
3044 zfs_prop_to_name(ZFS_PROP_CASE), &sense);
3045 (void) nvlist_remove_all(createprops,
3046 zfs_prop_to_name(ZFS_PROP_CASE));
3047 }
3048
3049 /*
3050 * If the zpl version requested is whacky or the file system
3051 * or pool is version is too "young" to support normalization
3052 * and the creator tried to set a value for one of the props,
3053 * error out.
3054 */
3055 if ((zplver < ZPL_VERSION_INITIAL || zplver > ZPL_VERSION) ||
3056 (zplver >= ZPL_VERSION_FUID && !fuids_ok) ||
3057 (zplver >= ZPL_VERSION_SA && !sa_ok) ||
3058 (zplver < ZPL_VERSION_NORMALIZATION &&
3059 (norm != ZFS_PROP_UNDEFINED || u8 != ZFS_PROP_UNDEFINED ||
3060 sense != ZFS_PROP_UNDEFINED)))
3061 return (SET_ERROR(ENOTSUP));
3062
3063 /*
3064 * Put the version in the zplprops
3065 */
3066 VERIFY(nvlist_add_uint64(zplprops,
3067 zfs_prop_to_name(ZFS_PROP_VERSION), zplver) == 0);
3068
3069 if (norm == ZFS_PROP_UNDEFINED)
3070 VERIFY(zfs_get_zplprop(os, ZFS_PROP_NORMALIZE, &norm) == 0);
3071 VERIFY(nvlist_add_uint64(zplprops,
3072 zfs_prop_to_name(ZFS_PROP_NORMALIZE), norm) == 0);
3073
3074 /*
3075 * If we're normalizing, names must always be valid UTF-8 strings.
3076 */
3077 if (norm)
3078 u8 = 1;
3079 if (u8 == ZFS_PROP_UNDEFINED)
3080 VERIFY(zfs_get_zplprop(os, ZFS_PROP_UTF8ONLY, &u8) == 0);
3081 VERIFY(nvlist_add_uint64(zplprops,
3082 zfs_prop_to_name(ZFS_PROP_UTF8ONLY), u8) == 0);
3083
3084 if (sense == ZFS_PROP_UNDEFINED)
3085 VERIFY(zfs_get_zplprop(os, ZFS_PROP_CASE, &sense) == 0);
3086 VERIFY(nvlist_add_uint64(zplprops,
3087 zfs_prop_to_name(ZFS_PROP_CASE), sense) == 0);
3088
3089 if (is_ci)
3090 *is_ci = (sense == ZFS_CASE_INSENSITIVE);
3091
3092 return (0);
3093 }
3094
3095 static int
3096 zfs_fill_zplprops(const char *dataset, nvlist_t *createprops,
3097 nvlist_t *zplprops, boolean_t *is_ci)
3098 {
3099 boolean_t fuids_ok, sa_ok;
3100 uint64_t zplver = ZPL_VERSION;
3101 objset_t *os = NULL;
3102 char parentname[ZFS_MAX_DATASET_NAME_LEN];
3103 char *cp;
3104 spa_t *spa;
3105 uint64_t spa_vers;
3106 int error;
3107
3108 (void) strlcpy(parentname, dataset, sizeof (parentname));
3109 cp = strrchr(parentname, '/');
3110 ASSERT(cp != NULL);
3111 cp[0] = '\0';
3112
3113 if ((error = spa_open(dataset, &spa, FTAG)) != 0)
3114 return (error);
3115
3116 spa_vers = spa_version(spa);
3117 spa_close(spa, FTAG);
3118
3119 zplver = zfs_zpl_version_map(spa_vers);
3120 fuids_ok = (zplver >= ZPL_VERSION_FUID);
3121 sa_ok = (zplver >= ZPL_VERSION_SA);
3122
3123 /*
3124 * Open parent object set so we can inherit zplprop values.
3125 */
3126 if ((error = dmu_objset_hold(parentname, FTAG, &os)) != 0)
3127 return (error);
3128
3129 error = zfs_fill_zplprops_impl(os, zplver, fuids_ok, sa_ok, createprops,
3130 zplprops, is_ci);
3131 dmu_objset_rele(os, FTAG);
3132 return (error);
3133 }
3134
3135 static int
3136 zfs_fill_zplprops_root(uint64_t spa_vers, nvlist_t *createprops,
3137 nvlist_t *zplprops, boolean_t *is_ci)
3138 {
3139 boolean_t fuids_ok;
3140 boolean_t sa_ok;
3141 uint64_t zplver = ZPL_VERSION;
3142 int error;
3143
3144 zplver = zfs_zpl_version_map(spa_vers);
3145 fuids_ok = (zplver >= ZPL_VERSION_FUID);
3146 sa_ok = (zplver >= ZPL_VERSION_SA);
3147
3148 error = zfs_fill_zplprops_impl(NULL, zplver, fuids_ok, sa_ok,
3149 createprops, zplprops, is_ci);
3150 return (error);
3151 }
3152
3153 /*
3154 * innvl: {
3155 * "type" -> dmu_objset_type_t (int32)
3156 * (optional) "props" -> { prop -> value }
3157 * }
3158 *
3159 * outnvl: propname -> error code (int32)
3160 */
3161 static int
3162 zfs_ioc_create(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3163 {
3164 int error = 0;
3165 zfs_creat_t zct = { 0 };
3166 nvlist_t *nvprops = NULL;
3167 void (*cbfunc)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx);
3168 int32_t type32;
3169 dmu_objset_type_t type;
3170 boolean_t is_insensitive = B_FALSE;
3171
3172 if (nvlist_lookup_int32(innvl, "type", &type32) != 0)
3173 return (SET_ERROR(EINVAL));
3174 type = type32;
3175 (void) nvlist_lookup_nvlist(innvl, "props", &nvprops);
3176
3177 switch (type) {
3178 case DMU_OST_ZFS:
3179 cbfunc = zfs_create_cb;
3180 break;
3181
3182 case DMU_OST_ZVOL:
3183 cbfunc = zvol_create_cb;
3184 break;
3185
3186 default:
3187 cbfunc = NULL;
3188 break;
3189 }
3190 if (strchr(fsname, '@') ||
3191 strchr(fsname, '%'))
3192 return (SET_ERROR(EINVAL));
3193
3194 zct.zct_props = nvprops;
3195
3196 if (cbfunc == NULL)
3197 return (SET_ERROR(EINVAL));
3198
3199 if (type == DMU_OST_ZVOL) {
3200 uint64_t volsize, volblocksize;
3201
3202 if (nvprops == NULL)
3203 return (SET_ERROR(EINVAL));
3204 if (nvlist_lookup_uint64(nvprops,
3205 zfs_prop_to_name(ZFS_PROP_VOLSIZE), &volsize) != 0)
3206 return (SET_ERROR(EINVAL));
3207
3208 if ((error = nvlist_lookup_uint64(nvprops,
3209 zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE),
3210 &volblocksize)) != 0 && error != ENOENT)
3211 return (SET_ERROR(EINVAL));
3212
3213 if (error != 0)
3214 volblocksize = zfs_prop_default_numeric(
3215 ZFS_PROP_VOLBLOCKSIZE);
3216
3217 if ((error = zvol_check_volblocksize(
3218 volblocksize)) != 0 ||
3219 (error = zvol_check_volsize(volsize,
3220 volblocksize)) != 0)
3221 return (error);
3222 } else if (type == DMU_OST_ZFS) {
3223 int error;
3224
3225 /*
3226 * We have to have normalization and
3227 * case-folding flags correct when we do the
3228 * file system creation, so go figure them out
3229 * now.
3230 */
3231 VERIFY(nvlist_alloc(&zct.zct_zplprops,
3232 NV_UNIQUE_NAME, KM_SLEEP) == 0);
3233 error = zfs_fill_zplprops(fsname, nvprops,
3234 zct.zct_zplprops, &is_insensitive);
3235 if (error != 0) {
3236 nvlist_free(zct.zct_zplprops);
3237 return (error);
3238 }
3239 }
3240
3241 error = dmu_objset_create(fsname, type,
3242 is_insensitive ? DS_FLAG_CI_DATASET : 0, cbfunc, &zct);
3243 nvlist_free(zct.zct_zplprops);
3244
3245 /*
3246 * It would be nice to do this atomically.
3247 */
3248 if (error == 0) {
3249 error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL,
3250 nvprops, outnvl);
3251 if (error != 0)
3252 (void) dsl_destroy_head(fsname);
3253 }
3254 return (error);
3255 }
3256
3257 /*
3258 * innvl: {
3259 * "origin" -> name of origin snapshot
3260 * (optional) "props" -> { prop -> value }
3261 * }
3262 *
3263 * outnvl: propname -> error code (int32)
3264 */
3265 static int
3266 zfs_ioc_clone(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3267 {
3268 int error = 0;
3269 nvlist_t *nvprops = NULL;
3270 char *origin_name;
3271
3272 if (nvlist_lookup_string(innvl, "origin", &origin_name) != 0)
3273 return (SET_ERROR(EINVAL));
3274 (void) nvlist_lookup_nvlist(innvl, "props", &nvprops);
3275
3276 if (strchr(fsname, '@') ||
3277 strchr(fsname, '%'))
3278 return (SET_ERROR(EINVAL));
3279
3280 if (dataset_namecheck(origin_name, NULL, NULL) != 0)
3281 return (SET_ERROR(EINVAL));
3282 error = dmu_objset_clone(fsname, origin_name);
3283 if (error != 0)
3284 return (error);
3285
3286 /*
3287 * It would be nice to do this atomically.
3288 */
3289 if (error == 0) {
3290 error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL,
3291 nvprops, outnvl);
3292 if (error != 0)
3293 (void) dsl_destroy_head(fsname);
3294 }
3295 return (error);
3296 }
3297
3298 /*
3299 * innvl: {
3300 * "snaps" -> { snapshot1, snapshot2 }
3301 * (optional) "props" -> { prop -> value (string) }
3302 * }
3303 *
3304 * outnvl: snapshot -> error code (int32)
3305 */
3306 static int
3307 zfs_ioc_snapshot(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3308 {
3309 nvlist_t *snaps;
3310 nvlist_t *props = NULL;
3311 int error, poollen;
3312 nvpair_t *pair;
3313
3314 (void) nvlist_lookup_nvlist(innvl, "props", &props);
3315 if ((error = zfs_check_userprops(poolname, props)) != 0)
3316 return (error);
3317
3318 if (!nvlist_empty(props) &&
3319 zfs_earlier_version(poolname, SPA_VERSION_SNAP_PROPS))
3320 return (SET_ERROR(ENOTSUP));
3321
3322 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
3323 return (SET_ERROR(EINVAL));
3324 poollen = strlen(poolname);
3325 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
3326 pair = nvlist_next_nvpair(snaps, pair)) {
3327 const char *name = nvpair_name(pair);
3328 const char *cp = strchr(name, '@');
3329
3330 /*
3331 * The snap name must contain an @, and the part after it must
3332 * contain only valid characters.
3333 */
3334 if (cp == NULL ||
3335 zfs_component_namecheck(cp + 1, NULL, NULL) != 0)
3336 return (SET_ERROR(EINVAL));
3337
3338 /*
3339 * The snap must be in the specified pool.
3340 */
3341 if (strncmp(name, poolname, poollen) != 0 ||
3342 (name[poollen] != '/' && name[poollen] != '@'))
3343 return (SET_ERROR(EXDEV));
3344
3345 /* This must be the only snap of this fs. */
3346 for (nvpair_t *pair2 = nvlist_next_nvpair(snaps, pair);
3347 pair2 != NULL; pair2 = nvlist_next_nvpair(snaps, pair2)) {
3348 if (strncmp(name, nvpair_name(pair2), cp - name + 1)
3349 == 0) {
3350 return (SET_ERROR(EXDEV));
3351 }
3352 }
3353 }
3354
3355 error = dsl_dataset_snapshot(snaps, props, outnvl);
3356 return (error);
3357 }
3358
3359 /*
3360 * innvl: "message" -> string
3361 */
3362 /* ARGSUSED */
3363 static int
3364 zfs_ioc_log_history(const char *unused, nvlist_t *innvl, nvlist_t *outnvl)
3365 {
3366 char *message;
3367 spa_t *spa;
3368 int error;
3369 char *poolname;
3370
3371 /*
3372 * The poolname in the ioctl is not set, we get it from the TSD,
3373 * which was set at the end of the last successful ioctl that allows
3374 * logging. The secpolicy func already checked that it is set.
3375 * Only one log ioctl is allowed after each successful ioctl, so
3376 * we clear the TSD here.
3377 */
3378 poolname = tsd_get(zfs_allow_log_key);
3379 (void) tsd_set(zfs_allow_log_key, NULL);
3380 error = spa_open(poolname, &spa, FTAG);
3381 strfree(poolname);
3382 if (error != 0)
3383 return (error);
3384
3385 if (nvlist_lookup_string(innvl, "message", &message) != 0) {
3386 spa_close(spa, FTAG);
3387 return (SET_ERROR(EINVAL));
3388 }
3389
3390 if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) {
3391 spa_close(spa, FTAG);
3392 return (SET_ERROR(ENOTSUP));
3393 }
3394
3395 error = spa_history_log(spa, message);
3396 spa_close(spa, FTAG);
3397 return (error);
3398 }
3399
3400 /*
3401 * The dp_config_rwlock must not be held when calling this, because the
3402 * unmount may need to write out data.
3403 *
3404 * This function is best-effort. Callers must deal gracefully if it
3405 * remains mounted (or is remounted after this call).
3406 *
3407 * Returns 0 if the argument is not a snapshot, or it is not currently a
3408 * filesystem, or we were able to unmount it. Returns error code otherwise.
3409 */
3410 int
3411 zfs_unmount_snap(const char *snapname)
3412 {
3413 vfs_t *vfsp;
3414 zfsvfs_t *zfsvfs;
3415 int err;
3416
3417 if (strchr(snapname, '@') == NULL)
3418 return (0);
3419
3420 vfsp = zfs_get_vfs(snapname);
3421 if (vfsp == NULL)
3422 return (0);
3423
3424 zfsvfs = vfsp->vfs_data;
3425 ASSERT(!dsl_pool_config_held(dmu_objset_pool(zfsvfs->z_os)));
3426
3427 err = vn_vfswlock(vfsp->vfs_vnodecovered);
3428 VFS_RELE(vfsp);
3429 if (err != 0)
3430 return (SET_ERROR(err));
3431
3432 /*
3433 * Always force the unmount for snapshots.
3434 */
3435 (void) dounmount(vfsp, MS_FORCE, kcred);
3436 return (0);
3437 }
3438
3439 /* ARGSUSED */
3440 static int
3441 zfs_unmount_snap_cb(const char *snapname, void *arg)
3442 {
3443 return (zfs_unmount_snap(snapname));
3444 }
3445
3446 /*
3447 * When a clone is destroyed, its origin may also need to be destroyed,
3448 * in which case it must be unmounted. This routine will do that unmount
3449 * if necessary.
3450 */
3451 void
3452 zfs_destroy_unmount_origin(const char *fsname)
3453 {
3454 int error;
3455 objset_t *os;
3456 dsl_dataset_t *ds;
3457
3458 error = dmu_objset_hold(fsname, FTAG, &os);
3459 if (error != 0)
3460 return;
3461 ds = dmu_objset_ds(os);
3462 if (dsl_dir_is_clone(ds->ds_dir) && DS_IS_DEFER_DESTROY(ds->ds_prev)) {
3463 char originname[ZFS_MAX_DATASET_NAME_LEN];
3464 dsl_dataset_name(ds->ds_prev, originname);
3465 dmu_objset_rele(os, FTAG);
3466 (void) zfs_unmount_snap(originname);
3467 } else {
3468 dmu_objset_rele(os, FTAG);
3469 }
3470 }
3471
3472 /*
3473 * innvl: {
3474 * "snaps" -> { snapshot1, snapshot2 }
3475 * (optional boolean) "defer"
3476 * }
3477 *
3478 * outnvl: snapshot -> error code (int32)
3479 *
3480 */
3481 /* ARGSUSED */
3482 static int
3483 zfs_ioc_destroy_snaps(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3484 {
3485 nvlist_t *snaps;
3486 nvpair_t *pair;
3487 boolean_t defer;
3488
3489 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
3490 return (SET_ERROR(EINVAL));
3491 defer = nvlist_exists(innvl, "defer");
3492
3493 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
3494 pair = nvlist_next_nvpair(snaps, pair)) {
3495 (void) zfs_unmount_snap(nvpair_name(pair));
3496 }
3497
3498 return (dsl_destroy_snapshots_nvl(snaps, defer, outnvl));
3499 }
3500
3501 /*
3502 * Create bookmarks. Bookmark names are of the form <fs>#<bmark>.
3503 * All bookmarks must be in the same pool.
3504 *
3505 * innvl: {
3506 * bookmark1 -> snapshot1, bookmark2 -> snapshot2
3507 * }
3508 *
3509 * outnvl: bookmark -> error code (int32)
3510 *
3511 */
3512 /* ARGSUSED */
3513 static int
3514 zfs_ioc_bookmark(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3515 {
3516 for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL);
3517 pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) {
3518 char *snap_name;
3519
3520 /*
3521 * Verify the snapshot argument.
3522 */
3523 if (nvpair_value_string(pair, &snap_name) != 0)
3524 return (SET_ERROR(EINVAL));
3525
3526
3527 /* Verify that the keys (bookmarks) are unique */
3528 for (nvpair_t *pair2 = nvlist_next_nvpair(innvl, pair);
3529 pair2 != NULL; pair2 = nvlist_next_nvpair(innvl, pair2)) {
3530 if (strcmp(nvpair_name(pair), nvpair_name(pair2)) == 0)
3531 return (SET_ERROR(EINVAL));
3532 }
3533 }
3534
3535 return (dsl_bookmark_create(innvl, outnvl));
3536 }
3537
3538 /*
3539 * innvl: {
3540 * property 1, property 2, ...
3541 * }
3542 *
3543 * outnvl: {
3544 * bookmark name 1 -> { property 1, property 2, ... },
3545 * bookmark name 2 -> { property 1, property 2, ... }
3546 * }
3547 *
3548 */
3549 static int
3550 zfs_ioc_get_bookmarks(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3551 {
3552 return (dsl_get_bookmarks(fsname, innvl, outnvl));
3553 }
3554
3555 /*
3556 * innvl: {
3557 * bookmark name 1, bookmark name 2
3558 * }
3559 *
3560 * outnvl: bookmark -> error code (int32)
3561 *
3562 */
3563 static int
3564 zfs_ioc_destroy_bookmarks(const char *poolname, nvlist_t *innvl,
3565 nvlist_t *outnvl)
3566 {
3567 int error, poollen;
3568
3569 poollen = strlen(poolname);
3570 for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL);
3571 pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) {
3572 const char *name = nvpair_name(pair);
3573 const char *cp = strchr(name, '#');
3574
3575 /*
3576 * The bookmark name must contain an #, and the part after it
3577 * must contain only valid characters.
3578 */
3579 if (cp == NULL ||
3580 zfs_component_namecheck(cp + 1, NULL, NULL) != 0)
3581 return (SET_ERROR(EINVAL));
3582
3583 /*
3584 * The bookmark must be in the specified pool.
3585 */
3586 if (strncmp(name, poolname, poollen) != 0 ||
3587 (name[poollen] != '/' && name[poollen] != '#'))
3588 return (SET_ERROR(EXDEV));
3589 }
3590
3591 error = dsl_bookmark_destroy(innvl, outnvl);
3592 return (error);
3593 }
3594
3595 /*
3596 * inputs:
3597 * zc_name name of dataset to destroy
3598 * zc_objset_type type of objset
3599 * zc_defer_destroy mark for deferred destroy
3600 *
3601 * outputs: none
3602 */
3603 static int
3604 zfs_ioc_destroy(zfs_cmd_t *zc)
3605 {
3606 int err;
3607
3608 if (zc->zc_objset_type == DMU_OST_ZFS) {
3609 err = zfs_unmount_snap(zc->zc_name);
3610 if (err != 0)
3611 return (err);
3612 }
3613
3614 if (strchr(zc->zc_name, '@'))
3615 err = dsl_destroy_snapshot(zc->zc_name, zc->zc_defer_destroy);
3616 else
3617 err = dsl_destroy_head(zc->zc_name);
3618 if (zc->zc_objset_type == DMU_OST_ZVOL && err == 0)
3619 (void) zvol_remove_minor(zc->zc_name);
3620 return (err);
3621 }
3622
3623 /*
3624 * fsname is name of dataset to rollback (to most recent snapshot)
3625 *
3626 * innvl is not used.
3627 *
3628 * outnvl: "target" -> name of most recent snapshot
3629 * }
3630 */
3631 /* ARGSUSED */
3632 static int
3633 zfs_ioc_rollback(const char *fsname, nvlist_t *args, nvlist_t *outnvl)
3634 {
3635 zfsvfs_t *zfsvfs;
3636 int error;
3637
3638 if (getzfsvfs(fsname, &zfsvfs) == 0) {
3639 error = zfs_suspend_fs(zfsvfs);
3640 if (error == 0) {
3641 int resume_err;
3642
3643 error = dsl_dataset_rollback(fsname, zfsvfs, outnvl);
3644 resume_err = zfs_resume_fs(zfsvfs, fsname);
3645 error = error ? error : resume_err;
3646 }
3647 VFS_RELE(zfsvfs->z_vfs);
3648 } else {
3649 error = dsl_dataset_rollback(fsname, NULL, outnvl);
3650 }
3651 return (error);
3652 }
3653
3654 static int
3655 recursive_unmount(const char *fsname, void *arg)
3656 {
3657 const char *snapname = arg;
3658 char fullname[ZFS_MAX_DATASET_NAME_LEN];
3659
3660 (void) snprintf(fullname, sizeof (fullname), "%s@%s", fsname, snapname);
3661 return (zfs_unmount_snap(fullname));
3662 }
3663
3664 /*
3665 * inputs:
3666 * zc_name old name of dataset
3667 * zc_value new name of dataset
3668 * zc_cookie recursive flag (only valid for snapshots)
3669 *
3670 * outputs: none
3671 */
3672 static int
3673 zfs_ioc_rename(zfs_cmd_t *zc)
3674 {
3675 boolean_t recursive = zc->zc_cookie & 1;
3676 char *at;
3677
3678 zc->zc_value[sizeof (zc->zc_value) - 1] = '\0';
3679 if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
3680 strchr(zc->zc_value, '%'))
3681 return (SET_ERROR(EINVAL));
3682
3683 at = strchr(zc->zc_name, '@');
3684 if (at != NULL) {
3685 /* snaps must be in same fs */
3686 int error;
3687
3688 if (strncmp(zc->zc_name, zc->zc_value, at - zc->zc_name + 1))
3689 return (SET_ERROR(EXDEV));
3690 *at = '\0';
3691 if (zc->zc_objset_type == DMU_OST_ZFS) {
3692 error = dmu_objset_find(zc->zc_name,
3693 recursive_unmount, at + 1,
3694 recursive ? DS_FIND_CHILDREN : 0);
3695 if (error != 0) {
3696 *at = '@';
3697 return (error);
3698 }
3699 }
3700 error = dsl_dataset_rename_snapshot(zc->zc_name,
3701 at + 1, strchr(zc->zc_value, '@') + 1, recursive);
3702 *at = '@';
3703
3704 return (error);
3705 } else {
3706 if (zc->zc_objset_type == DMU_OST_ZVOL)
3707 (void) zvol_remove_minor(zc->zc_name);
3708 return (dsl_dir_rename(zc->zc_name, zc->zc_value));
3709 }
3710 }
3711
3712 static int
3713 zfs_check_settable(const char *dsname, nvpair_t *pair, cred_t *cr)
3714 {
3715 const char *propname = nvpair_name(pair);
3716 boolean_t issnap = (strchr(dsname, '@') != NULL);
3717 zfs_prop_t prop = zfs_name_to_prop(propname);
3718 uint64_t intval;
3719 int err;
3720
3721 if (prop == ZPROP_INVAL) {
3722 if (zfs_prop_user(propname)) {
3723 if (err = zfs_secpolicy_write_perms(dsname,
3724 ZFS_DELEG_PERM_USERPROP, cr))
3725 return (err);
3726 return (0);
3727 }
3728
3729 if (!issnap && zfs_prop_userquota(propname)) {
3730 const char *perm = NULL;
3731 const char *uq_prefix =
3732 zfs_userquota_prop_prefixes[ZFS_PROP_USERQUOTA];
3733 const char *gq_prefix =
3734 zfs_userquota_prop_prefixes[ZFS_PROP_GROUPQUOTA];
3735
3736 if (strncmp(propname, uq_prefix,
3737 strlen(uq_prefix)) == 0) {
3738 perm = ZFS_DELEG_PERM_USERQUOTA;
3739 } else if (strncmp(propname, gq_prefix,
3740 strlen(gq_prefix)) == 0) {
3741 perm = ZFS_DELEG_PERM_GROUPQUOTA;
3742 } else {
3743 /* USERUSED and GROUPUSED are read-only */
3744 return (SET_ERROR(EINVAL));
3745 }
3746
3747 if (err = zfs_secpolicy_write_perms(dsname, perm, cr))
3748 return (err);
3749 return (0);
3750 }
3751
3752 return (SET_ERROR(EINVAL));
3753 }
3754
3755 if (issnap)
3756 return (SET_ERROR(EINVAL));
3757
3758 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
3759 /*
3760 * dsl_prop_get_all_impl() returns properties in this
3761 * format.
3762 */
3763 nvlist_t *attrs;
3764 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
3765 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
3766 &pair) == 0);
3767 }
3768
3769 /*
3770 * Check that this value is valid for this pool version
3771 */
3772 switch (prop) {
3773 case ZFS_PROP_COMPRESSION:
3774 /*
3775 * If the user specified gzip compression, make sure
3776 * the SPA supports it. We ignore any errors here since
3777 * we'll catch them later.
3778 */
3779 if (nvpair_value_uint64(pair, &intval) == 0) {
3780 if (intval >= ZIO_COMPRESS_GZIP_1 &&
3781 intval <= ZIO_COMPRESS_GZIP_9 &&
3782 zfs_earlier_version(dsname,
3783 SPA_VERSION_GZIP_COMPRESSION)) {
3784 return (SET_ERROR(ENOTSUP));
3785 }
3786
3787 if (intval == ZIO_COMPRESS_ZLE &&
3788 zfs_earlier_version(dsname,
3789 SPA_VERSION_ZLE_COMPRESSION))
3790 return (SET_ERROR(ENOTSUP));
3791
3792 if (intval == ZIO_COMPRESS_LZ4) {
3793 spa_t *spa;
3794
3795 if ((err = spa_open(dsname, &spa, FTAG)) != 0)
3796 return (err);
3797
3798 if (!spa_feature_is_enabled(spa,
3799 SPA_FEATURE_LZ4_COMPRESS)) {
3800 spa_close(spa, FTAG);
3801 return (SET_ERROR(ENOTSUP));
3802 }
3803 spa_close(spa, FTAG);
3804 }
3805
3806 /*
3807 * If this is a bootable dataset then
3808 * verify that the compression algorithm
3809 * is supported for booting. We must return
3810 * something other than ENOTSUP since it
3811 * implies a downrev pool version.
3812 */
3813 if (zfs_is_bootfs(dsname) &&
3814 !BOOTFS_COMPRESS_VALID(intval)) {
3815 return (SET_ERROR(ERANGE));
3816 }
3817 }
3818 break;
3819
3820 case ZFS_PROP_COPIES:
3821 if (zfs_earlier_version(dsname, SPA_VERSION_DITTO_BLOCKS))
3822 return (SET_ERROR(ENOTSUP));
3823 break;
3824
3825 case ZFS_PROP_RECORDSIZE:
3826 /* Record sizes above 128k need the feature to be enabled */
3827 if (nvpair_value_uint64(pair, &intval) == 0 &&
3828 intval > SPA_OLD_MAXBLOCKSIZE) {
3829 spa_t *spa;
3830
3831 /*
3832 * If this is a bootable dataset then
3833 * the we don't allow large (>128K) blocks,
3834 * because GRUB doesn't support them.
3835 */
3836 if (zfs_is_bootfs(dsname) &&
3837 intval > SPA_OLD_MAXBLOCKSIZE) {
3838 return (SET_ERROR(ERANGE));
3839 }
3840
3841 /*
3842 * We don't allow setting the property above 1MB,
3843 * unless the tunable has been changed.
3844 */
3845 if (intval > zfs_max_recordsize ||
3846 intval > SPA_MAXBLOCKSIZE)
3847 return (SET_ERROR(ERANGE));
3848
3849 if ((err = spa_open(dsname, &spa, FTAG)) != 0)
3850 return (err);
3851
3852 if (!spa_feature_is_enabled(spa,
3853 SPA_FEATURE_LARGE_BLOCKS)) {
3854 spa_close(spa, FTAG);
3855 return (SET_ERROR(ENOTSUP));
3856 }
3857 spa_close(spa, FTAG);
3858 }
3859 break;
3860
3861 case ZFS_PROP_SHARESMB:
3862 if (zpl_earlier_version(dsname, ZPL_VERSION_FUID))
3863 return (SET_ERROR(ENOTSUP));
3864 break;
3865
3866 case ZFS_PROP_ACLINHERIT:
3867 if (nvpair_type(pair) == DATA_TYPE_UINT64 &&
3868 nvpair_value_uint64(pair, &intval) == 0) {
3869 if (intval == ZFS_ACL_PASSTHROUGH_X &&
3870 zfs_earlier_version(dsname,
3871 SPA_VERSION_PASSTHROUGH_X))
3872 return (SET_ERROR(ENOTSUP));
3873 }
3874 break;
3875
3876 case ZFS_PROP_CHECKSUM:
3877 case ZFS_PROP_DEDUP:
3878 {
3879 spa_feature_t feature;
3880 spa_t *spa;
3881
3882 /* dedup feature version checks */
3883 if (prop == ZFS_PROP_DEDUP &&
3884 zfs_earlier_version(dsname, SPA_VERSION_DEDUP))
3885 return (SET_ERROR(ENOTSUP));
3886
3887 if (nvpair_value_uint64(pair, &intval) != 0)
3888 return (SET_ERROR(EINVAL));
3889
3890 /* check prop value is enabled in features */
3891 feature = zio_checksum_to_feature(intval & ZIO_CHECKSUM_MASK);
3892 if (feature == SPA_FEATURE_NONE)
3893 break;
3894
3895 if ((err = spa_open(dsname, &spa, FTAG)) != 0)
3896 return (err);
3897 /*
3898 * Salted checksums are not supported on root pools.
3899 */
3900 if (spa_bootfs(spa) != 0 &&
3901 intval < ZIO_CHECKSUM_FUNCTIONS &&
3902 (zio_checksum_table[intval].ci_flags &
3903 ZCHECKSUM_FLAG_SALTED)) {
3904 spa_close(spa, FTAG);
3905 return (SET_ERROR(ERANGE));
3906 }
3907 if (!spa_feature_is_enabled(spa, feature)) {
3908 spa_close(spa, FTAG);
3909 return (SET_ERROR(ENOTSUP));
3910 }
3911 spa_close(spa, FTAG);
3912 break;
3913 }
3914 }
3915
3916 return (zfs_secpolicy_setprop(dsname, prop, pair, CRED()));
3917 }
3918
3919 /*
3920 * Checks for a race condition to make sure we don't increment a feature flag
3921 * multiple times.
3922 */
3923 static int
3924 zfs_prop_activate_feature_check(void *arg, dmu_tx_t *tx)
3925 {
3926 spa_t *spa = dmu_tx_pool(tx)->dp_spa;
3927 spa_feature_t *featurep = arg;
3928
3929 if (!spa_feature_is_active(spa, *featurep))
3930 return (0);
3931 else
3932 return (SET_ERROR(EBUSY));
3933 }
3934
3935 /*
3936 * The callback invoked on feature activation in the sync task caused by
3937 * zfs_prop_activate_feature.
3938 */
3939 static void
3940 zfs_prop_activate_feature_sync(void *arg, dmu_tx_t *tx)
3941 {
3942 spa_t *spa = dmu_tx_pool(tx)->dp_spa;
3943 spa_feature_t *featurep = arg;
3944
3945 spa_feature_incr(spa, *featurep, tx);
3946 }
3947
3948 /*
3949 * Activates a feature on a pool in response to a property setting. This
3950 * creates a new sync task which modifies the pool to reflect the feature
3951 * as being active.
3952 */
3953 static int
3954 zfs_prop_activate_feature(spa_t *spa, spa_feature_t feature)
3955 {
3956 int err;
3957
3958 /* EBUSY here indicates that the feature is already active */
3959 err = dsl_sync_task(spa_name(spa),
3960 zfs_prop_activate_feature_check, zfs_prop_activate_feature_sync,
3961 &feature, 2, ZFS_SPACE_CHECK_RESERVED);
3962
3963 if (err != 0 && err != EBUSY)
3964 return (err);
3965 else
3966 return (0);
3967 }
3968
3969 /*
3970 * Removes properties from the given props list that fail permission checks
3971 * needed to clear them and to restore them in case of a receive error. For each
3972 * property, make sure we have both set and inherit permissions.
3973 *
3974 * Returns the first error encountered if any permission checks fail. If the
3975 * caller provides a non-NULL errlist, it also gives the complete list of names
3976 * of all the properties that failed a permission check along with the
3977 * corresponding error numbers. The caller is responsible for freeing the
3978 * returned errlist.
3979 *
3980 * If every property checks out successfully, zero is returned and the list
3981 * pointed at by errlist is NULL.
3982 */
3983 static int
3984 zfs_check_clearable(char *dataset, nvlist_t *props, nvlist_t **errlist)
3985 {
3986 zfs_cmd_t *zc;
3987 nvpair_t *pair, *next_pair;
3988 nvlist_t *errors;
3989 int err, rv = 0;
3990
3991 if (props == NULL)
3992 return (0);
3993
3994 VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0);
3995
3996 zc = kmem_alloc(sizeof (zfs_cmd_t), KM_SLEEP);
3997 (void) strcpy(zc->zc_name, dataset);
3998 pair = nvlist_next_nvpair(props, NULL);
3999 while (pair != NULL) {
4000 next_pair = nvlist_next_nvpair(props, pair);
4001
4002 (void) strcpy(zc->zc_value, nvpair_name(pair));
4003 if ((err = zfs_check_settable(dataset, pair, CRED())) != 0 ||
4004 (err = zfs_secpolicy_inherit_prop(zc, NULL, CRED())) != 0) {
4005 VERIFY(nvlist_remove_nvpair(props, pair) == 0);
4006 VERIFY(nvlist_add_int32(errors,
4007 zc->zc_value, err) == 0);
4008 }
4009 pair = next_pair;
4010 }
4011 kmem_free(zc, sizeof (zfs_cmd_t));
4012
4013 if ((pair = nvlist_next_nvpair(errors, NULL)) == NULL) {
4014 nvlist_free(errors);
4015 errors = NULL;
4016 } else {
4017 VERIFY(nvpair_value_int32(pair, &rv) == 0);
4018 }
4019
4020 if (errlist == NULL)
4021 nvlist_free(errors);
4022 else
4023 *errlist = errors;
4024
4025 return (rv);
4026 }
4027
4028 static boolean_t
4029 propval_equals(nvpair_t *p1, nvpair_t *p2)
4030 {
4031 if (nvpair_type(p1) == DATA_TYPE_NVLIST) {
4032 /* dsl_prop_get_all_impl() format */
4033 nvlist_t *attrs;
4034 VERIFY(nvpair_value_nvlist(p1, &attrs) == 0);
4035 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
4036 &p1) == 0);
4037 }
4038
4039 if (nvpair_type(p2) == DATA_TYPE_NVLIST) {
4040 nvlist_t *attrs;
4041 VERIFY(nvpair_value_nvlist(p2, &attrs) == 0);
4042 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
4043 &p2) == 0);
4044 }
4045
4046 if (nvpair_type(p1) != nvpair_type(p2))
4047 return (B_FALSE);
4048
4049 if (nvpair_type(p1) == DATA_TYPE_STRING) {
4050 char *valstr1, *valstr2;
4051
4052 VERIFY(nvpair_value_string(p1, (char **)&valstr1) == 0);
4053 VERIFY(nvpair_value_string(p2, (char **)&valstr2) == 0);
4054 return (strcmp(valstr1, valstr2) == 0);
4055 } else {
4056 uint64_t intval1, intval2;
4057
4058 VERIFY(nvpair_value_uint64(p1, &intval1) == 0);
4059 VERIFY(nvpair_value_uint64(p2, &intval2) == 0);
4060 return (intval1 == intval2);
4061 }
4062 }
4063
4064 /*
4065 * Remove properties from props if they are not going to change (as determined
4066 * by comparison with origprops). Remove them from origprops as well, since we
4067 * do not need to clear or restore properties that won't change.
4068 */
4069 static void
4070 props_reduce(nvlist_t *props, nvlist_t *origprops)
4071 {
4072 nvpair_t *pair, *next_pair;
4073
4074 if (origprops == NULL)
4075 return; /* all props need to be received */
4076
4077 pair = nvlist_next_nvpair(props, NULL);
4078 while (pair != NULL) {
4079 const char *propname = nvpair_name(pair);
4080 nvpair_t *match;
4081
4082 next_pair = nvlist_next_nvpair(props, pair);
4083
4084 if ((nvlist_lookup_nvpair(origprops, propname,
4085 &match) != 0) || !propval_equals(pair, match))
4086 goto next; /* need to set received value */
4087
4088 /* don't clear the existing received value */
4089 (void) nvlist_remove_nvpair(origprops, match);
4090 /* don't bother receiving the property */
4091 (void) nvlist_remove_nvpair(props, pair);
4092 next:
4093 pair = next_pair;
4094 }
4095 }
4096
4097 /*
4098 * Extract properties that cannot be set PRIOR to the receipt of a dataset.
4099 * For example, refquota cannot be set until after the receipt of a dataset,
4100 * because in replication streams, an older/earlier snapshot may exceed the
4101 * refquota. We want to receive the older/earlier snapshot, but setting
4102 * refquota pre-receipt will set the dsl's ACTUAL quota, which will prevent
4103 * the older/earlier snapshot from being received (with EDQUOT).
4104 *
4105 * The ZFS test "zfs_receive_011_pos" demonstrates such a scenario.
4106 *
4107 * libzfs will need to be judicious handling errors encountered by props
4108 * extracted by this function.
4109 */
4110 static nvlist_t *
4111 extract_delay_props(nvlist_t *props)
4112 {
4113 nvlist_t *delayprops;
4114 nvpair_t *nvp, *tmp;
4115 static const zfs_prop_t delayable[] = { ZFS_PROP_REFQUOTA, 0 };
4116 int i;
4117
4118 VERIFY(nvlist_alloc(&delayprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
4119
4120 for (nvp = nvlist_next_nvpair(props, NULL); nvp != NULL;
4121 nvp = nvlist_next_nvpair(props, nvp)) {
4122 /*
4123 * strcmp() is safe because zfs_prop_to_name() always returns
4124 * a bounded string.
4125 */
4126 for (i = 0; delayable[i] != 0; i++) {
4127 if (strcmp(zfs_prop_to_name(delayable[i]),
4128 nvpair_name(nvp)) == 0) {
4129 break;
4130 }
4131 }
4132 if (delayable[i] != 0) {
4133 tmp = nvlist_prev_nvpair(props, nvp);
4134 VERIFY(nvlist_add_nvpair(delayprops, nvp) == 0);
4135 VERIFY(nvlist_remove_nvpair(props, nvp) == 0);
4136 nvp = tmp;
4137 }
4138 }
4139
4140 if (nvlist_empty(delayprops)) {
4141 nvlist_free(delayprops);
4142 delayprops = NULL;
4143 }
4144 return (delayprops);
4145 }
4146
4147 #ifdef DEBUG
4148 static boolean_t zfs_ioc_recv_inject_err;
4149 #endif
4150
4151 /*
4152 * inputs:
4153 * zc_name name of containing filesystem
4154 * zc_nvlist_src{_size} nvlist of properties to apply
4155 * zc_value name of snapshot to create
4156 * zc_string name of clone origin (if DRR_FLAG_CLONE)
4157 * zc_cookie file descriptor to recv from
4158 * zc_begin_record the BEGIN record of the stream (not byteswapped)
4159 * zc_guid force flag
4160 * zc_cleanup_fd cleanup-on-exit file descriptor
4161 * zc_action_handle handle for this guid/ds mapping (or zero on first call)
4162 * zc_resumable if data is incomplete assume sender will resume
4163 *
4164 * outputs:
4165 * zc_cookie number of bytes read
4166 * zc_nvlist_dst{_size} error for each unapplied received property
4167 * zc_obj zprop_errflags_t
4168 * zc_action_handle handle for this guid/ds mapping
4169 */
4170 static int
4171 zfs_ioc_recv(zfs_cmd_t *zc)
4172 {
4173 file_t *fp;
4174 dmu_recv_cookie_t drc;
4175 boolean_t force = (boolean_t)zc->zc_guid;
4176 int fd;
4177 int error = 0;
4178 int props_error = 0;
4179 nvlist_t *errors;
4180 offset_t off;
4181 nvlist_t *props = NULL; /* sent properties */
4182 nvlist_t *origprops = NULL; /* existing properties */
4183 nvlist_t *delayprops = NULL; /* sent properties applied post-receive */
4184 char *origin = NULL;
4185 char *tosnap;
4186 char tofs[ZFS_MAX_DATASET_NAME_LEN];
4187 boolean_t first_recvd_props = B_FALSE;
4188
4189 if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
4190 strchr(zc->zc_value, '@') == NULL ||
4191 strchr(zc->zc_value, '%'))
4192 return (SET_ERROR(EINVAL));
4193
4194 (void) strcpy(tofs, zc->zc_value);
4195 tosnap = strchr(tofs, '@');
4196 *tosnap++ = '\0';
4197
4198 if (zc->zc_nvlist_src != NULL &&
4199 (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
4200 zc->zc_iflags, &props)) != 0)
4201 return (error);
4202
4203 fd = zc->zc_cookie;
4204 fp = getf(fd);
4205 if (fp == NULL) {
4206 nvlist_free(props);
4207 return (SET_ERROR(EBADF));
4208 }
4209
4210 errors = fnvlist_alloc();
4211
4212 if (zc->zc_string[0])
4213 origin = zc->zc_string;
4214
4215 error = dmu_recv_begin(tofs, tosnap,
4216 &zc->zc_begin_record, force, zc->zc_resumable, origin, &drc);
4217 if (error != 0)
4218 goto out;
4219
4220 /*
4221 * Set properties before we receive the stream so that they are applied
4222 * to the new data. Note that we must call dmu_recv_stream() if
4223 * dmu_recv_begin() succeeds.
4224 */
4225 if (props != NULL && !drc.drc_newfs) {
4226 if (spa_version(dsl_dataset_get_spa(drc.drc_ds)) >=
4227 SPA_VERSION_RECVD_PROPS &&
4228 !dsl_prop_get_hasrecvd(tofs))
4229 first_recvd_props = B_TRUE;
4230
4231 /*
4232 * If new received properties are supplied, they are to
4233 * completely replace the existing received properties, so stash
4234 * away the existing ones.
4235 */
4236 if (dsl_prop_get_received(tofs, &origprops) == 0) {
4237 nvlist_t *errlist = NULL;
4238 /*
4239 * Don't bother writing a property if its value won't
4240 * change (and avoid the unnecessary security checks).
4241 *
4242 * The first receive after SPA_VERSION_RECVD_PROPS is a
4243 * special case where we blow away all local properties
4244 * regardless.
4245 */
4246 if (!first_recvd_props)
4247 props_reduce(props, origprops);
4248 if (zfs_check_clearable(tofs, origprops, &errlist) != 0)
4249 (void) nvlist_merge(errors, errlist, 0);
4250 nvlist_free(errlist);
4251
4252 if (clear_received_props(tofs, origprops,
4253 first_recvd_props ? NULL : props) != 0)
4254 zc->zc_obj |= ZPROP_ERR_NOCLEAR;
4255 } else {
4256 zc->zc_obj |= ZPROP_ERR_NOCLEAR;
4257 }
4258 }
4259
4260 if (props != NULL) {
4261 props_error = dsl_prop_set_hasrecvd(tofs);
4262
4263 if (props_error == 0) {
4264 delayprops = extract_delay_props(props);
4265 (void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_RECEIVED,
4266 props, errors);
4267 }
4268 }
4269
4270 off = fp->f_offset;
4271 error = dmu_recv_stream(&drc, fp->f_vnode, &off, zc->zc_cleanup_fd,
4272 &zc->zc_action_handle);
4273
4274 if (error == 0) {
4275 zfsvfs_t *zfsvfs = NULL;
4276
4277 if (getzfsvfs(tofs, &zfsvfs) == 0) {
4278 /* online recv */
4279 int end_err;
4280
4281 error = zfs_suspend_fs(zfsvfs);
4282 /*
4283 * If the suspend fails, then the recv_end will
4284 * likely also fail, and clean up after itself.
4285 */
4286 end_err = dmu_recv_end(&drc, zfsvfs);
4287 if (error == 0)
4288 error = zfs_resume_fs(zfsvfs, tofs);
4289 error = error ? error : end_err;
4290 VFS_RELE(zfsvfs->z_vfs);
4291 } else {
4292 error = dmu_recv_end(&drc, NULL);
4293 }
4294
4295 /* Set delayed properties now, after we're done receiving. */
4296 if (delayprops != NULL && error == 0) {
4297 (void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_RECEIVED,
4298 delayprops, errors);
4299 }
4300 }
4301
4302 if (delayprops != NULL) {
4303 /*
4304 * Merge delayed props back in with initial props, in case
4305 * we're DEBUG and zfs_ioc_recv_inject_err is set (which means
4306 * we have to make sure clear_received_props() includes
4307 * the delayed properties).
4308 *
4309 * Since zfs_ioc_recv_inject_err is only in DEBUG kernels,
4310 * using ASSERT() will be just like a VERIFY.
4311 */
4312 ASSERT(nvlist_merge(props, delayprops, 0) == 0);
4313 nvlist_free(delayprops);
4314 }
4315
4316 /*
4317 * Now that all props, initial and delayed, are set, report the prop
4318 * errors to the caller.
4319 */
4320 if (zc->zc_nvlist_dst_size != 0 &&
4321 (nvlist_smush(errors, zc->zc_nvlist_dst_size) != 0 ||
4322 put_nvlist(zc, errors) != 0)) {
4323 /*
4324 * Caller made zc->zc_nvlist_dst less than the minimum expected
4325 * size or supplied an invalid address.
4326 */
4327 props_error = SET_ERROR(EINVAL);
4328 }
4329
4330 zc->zc_cookie = off - fp->f_offset;
4331 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
4332 fp->f_offset = off;
4333
4334 #ifdef DEBUG
4335 if (zfs_ioc_recv_inject_err) {
4336 zfs_ioc_recv_inject_err = B_FALSE;
4337 error = 1;
4338 }
4339 #endif
4340 /*
4341 * On error, restore the original props.
4342 */
4343 if (error != 0 && props != NULL && !drc.drc_newfs) {
4344 if (clear_received_props(tofs, props, NULL) != 0) {
4345 /*
4346 * We failed to clear the received properties.
4347 * Since we may have left a $recvd value on the
4348 * system, we can't clear the $hasrecvd flag.
4349 */
4350 zc->zc_obj |= ZPROP_ERR_NORESTORE;
4351 } else if (first_recvd_props) {
4352 dsl_prop_unset_hasrecvd(tofs);
4353 }
4354
4355 if (origprops == NULL && !drc.drc_newfs) {
4356 /* We failed to stash the original properties. */
4357 zc->zc_obj |= ZPROP_ERR_NORESTORE;
4358 }
4359
4360 /*
4361 * dsl_props_set() will not convert RECEIVED to LOCAL on or
4362 * after SPA_VERSION_RECVD_PROPS, so we need to specify LOCAL
4363 * explictly if we're restoring local properties cleared in the
4364 * first new-style receive.
4365 */
4366 if (origprops != NULL &&
4367 zfs_set_prop_nvlist(tofs, (first_recvd_props ?
4368 ZPROP_SRC_LOCAL : ZPROP_SRC_RECEIVED),
4369 origprops, NULL) != 0) {
4370 /*
4371 * We stashed the original properties but failed to
4372 * restore them.
4373 */
4374 zc->zc_obj |= ZPROP_ERR_NORESTORE;
4375 }
4376 }
4377 out:
4378 nvlist_free(props);
4379 nvlist_free(origprops);
4380 nvlist_free(errors);
4381 releasef(fd);
4382
4383 if (error == 0)
4384 error = props_error;
4385
4386 return (error);
4387 }
4388
4389 /*
4390 * inputs:
4391 * zc_name name of snapshot to send
4392 * zc_cookie file descriptor to send stream to
4393 * zc_obj fromorigin flag (mutually exclusive with zc_fromobj)
4394 * zc_sendobj objsetid of snapshot to send
4395 * zc_fromobj objsetid of incremental fromsnap (may be zero)
4396 * zc_guid if set, estimate size of stream only. zc_cookie is ignored.
4397 * output size in zc_objset_type.
4398 * zc_flags lzc_send_flags
4399 *
4400 * outputs:
4401 * zc_objset_type estimated size, if zc_guid is set
4402 */
4403 static int
4404 zfs_ioc_send(zfs_cmd_t *zc)
4405 {
4406 int error;
4407 offset_t off;
4408 boolean_t estimate = (zc->zc_guid != 0);
4409 boolean_t embedok = (zc->zc_flags & 0x1);
4410 boolean_t large_block_ok = (zc->zc_flags & 0x2);
4411
4412 if (zc->zc_obj != 0) {
4413 dsl_pool_t *dp;
4414 dsl_dataset_t *tosnap;
4415
4416 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
4417 if (error != 0)
4418 return (error);
4419
4420 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &tosnap);
4421 if (error != 0) {
4422 dsl_pool_rele(dp, FTAG);
4423 return (error);
4424 }
4425
4426 if (dsl_dir_is_clone(tosnap->ds_dir))
4427 zc->zc_fromobj =
4428 dsl_dir_phys(tosnap->ds_dir)->dd_origin_obj;
4429 dsl_dataset_rele(tosnap, FTAG);
4430 dsl_pool_rele(dp, FTAG);
4431 }
4432
4433 if (estimate) {
4434 dsl_pool_t *dp;
4435 dsl_dataset_t *tosnap;
4436 dsl_dataset_t *fromsnap = NULL;
4437
4438 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
4439 if (error != 0)
4440 return (error);
4441
4442 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &tosnap);
4443 if (error != 0) {
4444 dsl_pool_rele(dp, FTAG);
4445 return (error);
4446 }
4447
4448 if (zc->zc_fromobj != 0) {
4449 error = dsl_dataset_hold_obj(dp, zc->zc_fromobj,
4450 FTAG, &fromsnap);
4451 if (error != 0) {
4452 dsl_dataset_rele(tosnap, FTAG);
4453 dsl_pool_rele(dp, FTAG);
4454 return (error);
4455 }
4456 }
4457
4458 error = dmu_send_estimate(tosnap, fromsnap,
4459 &zc->zc_objset_type);
4460
4461 if (fromsnap != NULL)
4462 dsl_dataset_rele(fromsnap, FTAG);
4463 dsl_dataset_rele(tosnap, FTAG);
4464 dsl_pool_rele(dp, FTAG);
4465 } else {
4466 file_t *fp = getf(zc->zc_cookie);
4467 if (fp == NULL)
4468 return (SET_ERROR(EBADF));
4469
4470 off = fp->f_offset;
4471 error = dmu_send_obj(zc->zc_name, zc->zc_sendobj,
4472 zc->zc_fromobj, embedok, large_block_ok,
4473 zc->zc_cookie, fp->f_vnode, &off);
4474
4475 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
4476 fp->f_offset = off;
4477 releasef(zc->zc_cookie);
4478 }
4479 return (error);
4480 }
4481
4482 /*
4483 * inputs:
4484 * zc_name name of snapshot on which to report progress
4485 * zc_cookie file descriptor of send stream
4486 *
4487 * outputs:
4488 * zc_cookie number of bytes written in send stream thus far
4489 */
4490 static int
4491 zfs_ioc_send_progress(zfs_cmd_t *zc)
4492 {
4493 dsl_pool_t *dp;
4494 dsl_dataset_t *ds;
4495 dmu_sendarg_t *dsp = NULL;
4496 int error;
4497
4498 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
4499 if (error != 0)
4500 return (error);
4501
4502 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &ds);
4503 if (error != 0) {
4504 dsl_pool_rele(dp, FTAG);
4505 return (error);
4506 }
4507
4508 mutex_enter(&ds->ds_sendstream_lock);
4509
4510 /*
4511 * Iterate over all the send streams currently active on this dataset.
4512 * If there's one which matches the specified file descriptor _and_ the
4513 * stream was started by the current process, return the progress of
4514 * that stream.
4515 */
4516 for (dsp = list_head(&ds->ds_sendstreams); dsp != NULL;
4517 dsp = list_next(&ds->ds_sendstreams, dsp)) {
4518 if (dsp->dsa_outfd == zc->zc_cookie &&
4519 dsp->dsa_proc == curproc)
4520 break;
4521 }
4522
4523 if (dsp != NULL)
4524 zc->zc_cookie = *(dsp->dsa_off);
4525 else
4526 error = SET_ERROR(ENOENT);
4527
4528 mutex_exit(&ds->ds_sendstream_lock);
4529 dsl_dataset_rele(ds, FTAG);
4530 dsl_pool_rele(dp, FTAG);
4531 return (error);
4532 }
4533
4534 static int
4535 zfs_ioc_inject_fault(zfs_cmd_t *zc)
4536 {
4537 int id, error;
4538
4539 error = zio_inject_fault(zc->zc_name, (int)zc->zc_guid, &id,
4540 &zc->zc_inject_record);
4541
4542 if (error == 0)
4543 zc->zc_guid = (uint64_t)id;
4544
4545 return (error);
4546 }
4547
4548 static int
4549 zfs_ioc_clear_fault(zfs_cmd_t *zc)
4550 {
4551 return (zio_clear_fault((int)zc->zc_guid));
4552 }
4553
4554 static int
4555 zfs_ioc_inject_list_next(zfs_cmd_t *zc)
4556 {
4557 int id = (int)zc->zc_guid;
4558 int error;
4559
4560 error = zio_inject_list_next(&id, zc->zc_name, sizeof (zc->zc_name),
4561 &zc->zc_inject_record);
4562
4563 zc->zc_guid = id;
4564
4565 return (error);
4566 }
4567
4568 static int
4569 zfs_ioc_error_log(zfs_cmd_t *zc)
4570 {
4571 spa_t *spa;
4572 int error;
4573 size_t count = (size_t)zc->zc_nvlist_dst_size;
4574
4575 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
4576 return (error);
4577
4578 error = spa_get_errlog(spa, (void *)(uintptr_t)zc->zc_nvlist_dst,
4579 &count);
4580 if (error == 0)
4581 zc->zc_nvlist_dst_size = count;
4582 else
4583 zc->zc_nvlist_dst_size = spa_get_errlog_size(spa);
4584
4585 spa_close(spa, FTAG);
4586
4587 return (error);
4588 }
4589
4590 static int
4591 zfs_ioc_clear(zfs_cmd_t *zc)
4592 {
4593 spa_t *spa;
4594 vdev_t *vd;
4595 int error;
4596
4597 /*
4598 * On zpool clear we also fix up missing slogs
4599 */
4600 mutex_enter(&spa_namespace_lock);
4601 spa = spa_lookup(zc->zc_name);
4602 if (spa == NULL) {
4603 mutex_exit(&spa_namespace_lock);
4604 return (SET_ERROR(EIO));
4605 }
4606 if (spa_get_log_state(spa) == SPA_LOG_MISSING) {
4607 /* we need to let spa_open/spa_load clear the chains */
4608 spa_set_log_state(spa, SPA_LOG_CLEAR);
4609 }
4610 spa->spa_last_open_failed = 0;
4611 mutex_exit(&spa_namespace_lock);
4612
4613 if (zc->zc_cookie & ZPOOL_NO_REWIND) {
4614 error = spa_open(zc->zc_name, &spa, FTAG);
4615 } else {
4616 nvlist_t *policy;
4617 nvlist_t *config = NULL;
4618
4619 if (zc->zc_nvlist_src == NULL)
4620 return (SET_ERROR(EINVAL));
4621
4622 if ((error = get_nvlist(zc->zc_nvlist_src,
4623 zc->zc_nvlist_src_size, zc->zc_iflags, &policy)) == 0) {
4624 error = spa_open_rewind(zc->zc_name, &spa, FTAG,
4625 policy, &config);
4626 if (config != NULL) {
4627 int err;
4628
4629 if ((err = put_nvlist(zc, config)) != 0)
4630 error = err;
4631 nvlist_free(config);
4632 }
4633 nvlist_free(policy);
4634 }
4635 }
4636
4637 if (error != 0)
4638 return (error);
4639
4640 spa_vdev_state_enter(spa, SCL_NONE);
4641
4642 if (zc->zc_guid == 0) {
4643 vd = NULL;
4644 } else {
4645 vd = spa_lookup_by_guid(spa, zc->zc_guid, B_TRUE);
4646 if (vd == NULL) {
4647 (void) spa_vdev_state_exit(spa, NULL, ENODEV);
4648 spa_close(spa, FTAG);
4649 return (SET_ERROR(ENODEV));
4650 }
4651 }
4652
4653 vdev_clear(spa, vd);
4654
4655 (void) spa_vdev_state_exit(spa, NULL, 0);
4656
4657 /*
4658 * Resume any suspended I/Os.
4659 */
4660 if (zio_resume(spa) != 0)
4661 error = SET_ERROR(EIO);
4662
4663 spa_close(spa, FTAG);
4664
4665 return (error);
4666 }
4667
4668 static int
4669 zfs_ioc_pool_reopen(zfs_cmd_t *zc)
4670 {
4671 spa_t *spa;
4672 int error;
4673
4674 error = spa_open(zc->zc_name, &spa, FTAG);
4675 if (error != 0)
4676 return (error);
4677
4678 spa_vdev_state_enter(spa, SCL_NONE);
4679
4680 /*
4681 * If a resilver is already in progress then set the
4682 * spa_scrub_reopen flag to B_TRUE so that we don't restart
4683 * the scan as a side effect of the reopen. Otherwise, let
4684 * vdev_open() decided if a resilver is required.
4685 */
4686 spa->spa_scrub_reopen = dsl_scan_resilvering(spa->spa_dsl_pool);
4687 vdev_reopen(spa->spa_root_vdev);
4688 spa->spa_scrub_reopen = B_FALSE;
4689
4690 (void) spa_vdev_state_exit(spa, NULL, 0);
4691 spa_close(spa, FTAG);
4692 return (0);
4693 }
4694 /*
4695 * inputs:
4696 * zc_name name of filesystem
4697 * zc_value name of origin snapshot
4698 *
4699 * outputs:
4700 * zc_string name of conflicting snapshot, if there is one
4701 */
4702 static int
4703 zfs_ioc_promote(zfs_cmd_t *zc)
4704 {
4705 char *cp;
4706
4707 /*
4708 * We don't need to unmount *all* the origin fs's snapshots, but
4709 * it's easier.
4710 */
4711 cp = strchr(zc->zc_value, '@');
4712 if (cp)
4713 *cp = '\0';
4714 (void) dmu_objset_find(zc->zc_value,
4715 zfs_unmount_snap_cb, NULL, DS_FIND_SNAPSHOTS);
4716 return (dsl_dataset_promote(zc->zc_name, zc->zc_string));
4717 }
4718
4719 /*
4720 * Retrieve a single {user|group}{used|quota}@... property.
4721 *
4722 * inputs:
4723 * zc_name name of filesystem
4724 * zc_objset_type zfs_userquota_prop_t
4725 * zc_value domain name (eg. "S-1-234-567-89")
4726 * zc_guid RID/UID/GID
4727 *
4728 * outputs:
4729 * zc_cookie property value
4730 */
4731 static int
4732 zfs_ioc_userspace_one(zfs_cmd_t *zc)
4733 {
4734 zfsvfs_t *zfsvfs;
4735 int error;
4736
4737 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
4738 return (SET_ERROR(EINVAL));
4739
4740 error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE);
4741 if (error != 0)
4742 return (error);
4743
4744 error = zfs_userspace_one(zfsvfs,
4745 zc->zc_objset_type, zc->zc_value, zc->zc_guid, &zc->zc_cookie);
4746 zfsvfs_rele(zfsvfs, FTAG);
4747
4748 return (error);
4749 }
4750
4751 /*
4752 * inputs:
4753 * zc_name name of filesystem
4754 * zc_cookie zap cursor
4755 * zc_objset_type zfs_userquota_prop_t
4756 * zc_nvlist_dst[_size] buffer to fill (not really an nvlist)
4757 *
4758 * outputs:
4759 * zc_nvlist_dst[_size] data buffer (array of zfs_useracct_t)
4760 * zc_cookie zap cursor
4761 */
4762 static int
4763 zfs_ioc_userspace_many(zfs_cmd_t *zc)
4764 {
4765 zfsvfs_t *zfsvfs;
4766 int bufsize = zc->zc_nvlist_dst_size;
4767
4768 if (bufsize <= 0)
4769 return (SET_ERROR(ENOMEM));
4770
4771 int error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE);
4772 if (error != 0)
4773 return (error);
4774
4775 void *buf = kmem_alloc(bufsize, KM_SLEEP);
4776
4777 error = zfs_userspace_many(zfsvfs, zc->zc_objset_type, &zc->zc_cookie,
4778 buf, &zc->zc_nvlist_dst_size);
4779
4780 if (error == 0) {
4781 error = xcopyout(buf,
4782 (void *)(uintptr_t)zc->zc_nvlist_dst,
4783 zc->zc_nvlist_dst_size);
4784 }
4785 kmem_free(buf, bufsize);
4786 zfsvfs_rele(zfsvfs, FTAG);
4787
4788 return (error);
4789 }
4790
4791 /*
4792 * inputs:
4793 * zc_name name of filesystem
4794 *
4795 * outputs:
4796 * none
4797 */
4798 static int
4799 zfs_ioc_userspace_upgrade(zfs_cmd_t *zc)
4800 {
4801 objset_t *os;
4802 int error = 0;
4803 zfsvfs_t *zfsvfs;
4804
4805 if (getzfsvfs(zc->zc_name, &zfsvfs) == 0) {
4806 if (!dmu_objset_userused_enabled(zfsvfs->z_os)) {
4807 /*
4808 * If userused is not enabled, it may be because the
4809 * objset needs to be closed & reopened (to grow the
4810 * objset_phys_t). Suspend/resume the fs will do that.
4811 */
4812 error = zfs_suspend_fs(zfsvfs);
4813 if (error == 0) {
4814 dmu_objset_refresh_ownership(zfsvfs->z_os,
4815 zfsvfs);
4816 error = zfs_resume_fs(zfsvfs, zc->zc_name);
4817 }
4818 }
4819 if (error == 0)
4820 error = dmu_objset_userspace_upgrade(zfsvfs->z_os);
4821 VFS_RELE(zfsvfs->z_vfs);
4822 } else {
4823 /* XXX kind of reading contents without owning */
4824 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
4825 if (error != 0)
4826 return (error);
4827
4828 error = dmu_objset_userspace_upgrade(os);
4829 dmu_objset_rele(os, FTAG);
4830 }
4831
4832 return (error);
4833 }
4834
4835 /*
4836 * We don't want to have a hard dependency
4837 * against some special symbols in sharefs
4838 * nfs, and smbsrv. Determine them if needed when
4839 * the first file system is shared.
4840 * Neither sharefs, nfs or smbsrv are unloadable modules.
4841 */
4842 int (*znfsexport_fs)(void *arg);
4843 int (*zshare_fs)(enum sharefs_sys_op, share_t *, uint32_t);
4844 int (*zsmbexport_fs)(void *arg, boolean_t add_share);
4845
4846 int zfs_nfsshare_inited;
4847 int zfs_smbshare_inited;
4848
4849 ddi_modhandle_t nfs_mod;
4850 ddi_modhandle_t sharefs_mod;
4851 ddi_modhandle_t smbsrv_mod;
4852 kmutex_t zfs_share_lock;
4853
4854 static int
4855 zfs_init_sharefs()
4856 {
4857 int error;
4858
4859 ASSERT(MUTEX_HELD(&zfs_share_lock));
4860 /* Both NFS and SMB shares also require sharetab support. */
4861 if (sharefs_mod == NULL && ((sharefs_mod =
4862 ddi_modopen("fs/sharefs",
4863 KRTLD_MODE_FIRST, &error)) == NULL)) {
4864 return (SET_ERROR(ENOSYS));
4865 }
4866 if (zshare_fs == NULL && ((zshare_fs =
4867 (int (*)(enum sharefs_sys_op, share_t *, uint32_t))
4868 ddi_modsym(sharefs_mod, "sharefs_impl", &error)) == NULL)) {
4869 return (SET_ERROR(ENOSYS));
4870 }
4871 return (0);
4872 }
4873
4874 static int
4875 zfs_ioc_share(zfs_cmd_t *zc)
4876 {
4877 int error;
4878 int opcode;
4879
4880 switch (zc->zc_share.z_sharetype) {
4881 case ZFS_SHARE_NFS:
4882 case ZFS_UNSHARE_NFS:
4883 if (zfs_nfsshare_inited == 0) {
4884 mutex_enter(&zfs_share_lock);
4885 if (nfs_mod == NULL && ((nfs_mod = ddi_modopen("fs/nfs",
4886 KRTLD_MODE_FIRST, &error)) == NULL)) {
4887 mutex_exit(&zfs_share_lock);
4888 return (SET_ERROR(ENOSYS));
4889 }
4890 if (znfsexport_fs == NULL &&
4891 ((znfsexport_fs = (int (*)(void *))
4892 ddi_modsym(nfs_mod,
4893 "nfs_export", &error)) == NULL)) {
4894 mutex_exit(&zfs_share_lock);
4895 return (SET_ERROR(ENOSYS));
4896 }
4897 error = zfs_init_sharefs();
4898 if (error != 0) {
4899 mutex_exit(&zfs_share_lock);
4900 return (SET_ERROR(ENOSYS));
4901 }
4902 zfs_nfsshare_inited = 1;
4903 mutex_exit(&zfs_share_lock);
4904 }
4905 break;
4906 case ZFS_SHARE_SMB:
4907 case ZFS_UNSHARE_SMB:
4908 if (zfs_smbshare_inited == 0) {
4909 mutex_enter(&zfs_share_lock);
4910 if (smbsrv_mod == NULL && ((smbsrv_mod =
4911 ddi_modopen("drv/smbsrv",
4912 KRTLD_MODE_FIRST, &error)) == NULL)) {
4913 mutex_exit(&zfs_share_lock);
4914 return (SET_ERROR(ENOSYS));
4915 }
4916 if (zsmbexport_fs == NULL && ((zsmbexport_fs =
4917 (int (*)(void *, boolean_t))ddi_modsym(smbsrv_mod,
4918 "smb_server_share", &error)) == NULL)) {
4919 mutex_exit(&zfs_share_lock);
4920 return (SET_ERROR(ENOSYS));
4921 }
4922 error = zfs_init_sharefs();
4923 if (error != 0) {
4924 mutex_exit(&zfs_share_lock);
4925 return (SET_ERROR(ENOSYS));
4926 }
4927 zfs_smbshare_inited = 1;
4928 mutex_exit(&zfs_share_lock);
4929 }
4930 break;
4931 default:
4932 return (SET_ERROR(EINVAL));
4933 }
4934
4935 switch (zc->zc_share.z_sharetype) {
4936 case ZFS_SHARE_NFS:
4937 case ZFS_UNSHARE_NFS:
4938 if (error =
4939 znfsexport_fs((void *)
4940 (uintptr_t)zc->zc_share.z_exportdata))
4941 return (error);
4942 break;
4943 case ZFS_SHARE_SMB:
4944 case ZFS_UNSHARE_SMB:
4945 if (error = zsmbexport_fs((void *)
4946 (uintptr_t)zc->zc_share.z_exportdata,
4947 zc->zc_share.z_sharetype == ZFS_SHARE_SMB ?
4948 B_TRUE: B_FALSE)) {
4949 return (error);
4950 }
4951 break;
4952 }
4953
4954 opcode = (zc->zc_share.z_sharetype == ZFS_SHARE_NFS ||
4955 zc->zc_share.z_sharetype == ZFS_SHARE_SMB) ?
4956 SHAREFS_ADD : SHAREFS_REMOVE;
4957
4958 /*
4959 * Add or remove share from sharetab
4960 */
4961 error = zshare_fs(opcode,
4962 (void *)(uintptr_t)zc->zc_share.z_sharedata,
4963 zc->zc_share.z_sharemax);
4964
4965 return (error);
4966
4967 }
4968
4969 ace_t full_access[] = {
4970 {(uid_t)-1, ACE_ALL_PERMS, ACE_EVERYONE, 0}
4971 };
4972
4973 /*
4974 * inputs:
4975 * zc_name name of containing filesystem
4976 * zc_obj object # beyond which we want next in-use object #
4977 *
4978 * outputs:
4979 * zc_obj next in-use object #
4980 */
4981 static int
4982 zfs_ioc_next_obj(zfs_cmd_t *zc)
4983 {
4984 objset_t *os = NULL;
4985 int error;
4986
4987 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
4988 if (error != 0)
4989 return (error);
4990
4991 error = dmu_object_next(os, &zc->zc_obj, B_FALSE,
4992 dsl_dataset_phys(os->os_dsl_dataset)->ds_prev_snap_txg);
4993
4994 dmu_objset_rele(os, FTAG);
4995 return (error);
4996 }
4997
4998 /*
4999 * inputs:
5000 * zc_name name of filesystem
5001 * zc_value prefix name for snapshot
5002 * zc_cleanup_fd cleanup-on-exit file descriptor for calling process
5003 *
5004 * outputs:
5005 * zc_value short name of new snapshot
5006 */
5007 static int
5008 zfs_ioc_tmp_snapshot(zfs_cmd_t *zc)
5009 {
5010 char *snap_name;
5011 char *hold_name;
5012 int error;
5013 minor_t minor;
5014
5015 error = zfs_onexit_fd_hold(zc->zc_cleanup_fd, &minor);
5016 if (error != 0)
5017 return (error);
5018
5019 snap_name = kmem_asprintf("%s-%016llx", zc->zc_value,
5020 (u_longlong_t)ddi_get_lbolt64());
5021 hold_name = kmem_asprintf("%%%s", zc->zc_value);
5022
5023 error = dsl_dataset_snapshot_tmp(zc->zc_name, snap_name, minor,
5024 hold_name);
5025 if (error == 0)
5026 (void) strcpy(zc->zc_value, snap_name);
5027 strfree(snap_name);
5028 strfree(hold_name);
5029 zfs_onexit_fd_rele(zc->zc_cleanup_fd);
5030 return (error);
5031 }
5032
5033 /*
5034 * inputs:
5035 * zc_name name of "to" snapshot
5036 * zc_value name of "from" snapshot
5037 * zc_cookie file descriptor to write diff data on
5038 *
5039 * outputs:
5040 * dmu_diff_record_t's to the file descriptor
5041 */
5042 static int
5043 zfs_ioc_diff(zfs_cmd_t *zc)
5044 {
5045 file_t *fp;
5046 offset_t off;
5047 int error;
5048
5049 fp = getf(zc->zc_cookie);
5050 if (fp == NULL)
5051 return (SET_ERROR(EBADF));
5052
5053 off = fp->f_offset;
5054
5055 error = dmu_diff(zc->zc_name, zc->zc_value, fp->f_vnode, &off);
5056
5057 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
5058 fp->f_offset = off;
5059 releasef(zc->zc_cookie);
5060
5061 return (error);
5062 }
5063
5064 /*
5065 * Remove all ACL files in shares dir
5066 */
5067 static int
5068 zfs_smb_acl_purge(znode_t *dzp)
5069 {
5070 zap_cursor_t zc;
5071 zap_attribute_t zap;
5072 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
5073 int error;
5074
5075 for (zap_cursor_init(&zc, zfsvfs->z_os, dzp->z_id);
5076 (error = zap_cursor_retrieve(&zc, &zap)) == 0;
5077 zap_cursor_advance(&zc)) {
5078 if ((error = VOP_REMOVE(ZTOV(dzp), zap.za_name, kcred,
5079 NULL, 0)) != 0)
5080 break;
5081 }
5082 zap_cursor_fini(&zc);
5083 return (error);
5084 }
5085
5086 static int
5087 zfs_ioc_smb_acl(zfs_cmd_t *zc)
5088 {
5089 vnode_t *vp;
5090 znode_t *dzp;
5091 vnode_t *resourcevp = NULL;
5092 znode_t *sharedir;
5093 zfsvfs_t *zfsvfs;
5094 nvlist_t *nvlist;
5095 char *src, *target;
5096 vattr_t vattr;
5097 vsecattr_t vsec;
5098 int error = 0;
5099
5100 if ((error = lookupname(zc->zc_value, UIO_SYSSPACE,
5101 NO_FOLLOW, NULL, &vp)) != 0)
5102 return (error);
5103
5104 /* Now make sure mntpnt and dataset are ZFS */
5105
5106 if (vp->v_vfsp->vfs_fstype != zfsfstype ||
5107 (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource),
5108 zc->zc_name) != 0)) {
5109 VN_RELE(vp);
5110 return (SET_ERROR(EINVAL));
5111 }
5112
5113 dzp = VTOZ(vp);
5114 zfsvfs = dzp->z_zfsvfs;
5115 ZFS_ENTER(zfsvfs);
5116
5117 /*
5118 * Create share dir if its missing.
5119 */
5120 mutex_enter(&zfsvfs->z_lock);
5121 if (zfsvfs->z_shares_dir == 0) {
5122 dmu_tx_t *tx;
5123
5124 tx = dmu_tx_create(zfsvfs->z_os);
5125 dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, TRUE,
5126 ZFS_SHARES_DIR);
5127 dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL);
5128 error = dmu_tx_assign(tx, TXG_WAIT);
5129 if (error != 0) {
5130 dmu_tx_abort(tx);
5131 } else {
5132 error = zfs_create_share_dir(zfsvfs, tx);
5133 dmu_tx_commit(tx);
5134 }
5135 if (error != 0) {
5136 mutex_exit(&zfsvfs->z_lock);
5137 VN_RELE(vp);
5138 ZFS_EXIT(zfsvfs);
5139 return (error);
5140 }
5141 }
5142 mutex_exit(&zfsvfs->z_lock);
5143
5144 ASSERT(zfsvfs->z_shares_dir);
5145 if ((error = zfs_zget(zfsvfs, zfsvfs->z_shares_dir, &sharedir)) != 0) {
5146 VN_RELE(vp);
5147 ZFS_EXIT(zfsvfs);
5148 return (error);
5149 }
5150
5151 switch (zc->zc_cookie) {
5152 case ZFS_SMB_ACL_ADD:
5153 vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE;
5154 vattr.va_type = VREG;
5155 vattr.va_mode = S_IFREG|0777;
5156 vattr.va_uid = 0;
5157 vattr.va_gid = 0;
5158
5159 vsec.vsa_mask = VSA_ACE;
5160 vsec.vsa_aclentp = &full_access;
5161 vsec.vsa_aclentsz = sizeof (full_access);
5162 vsec.vsa_aclcnt = 1;
5163
5164 error = VOP_CREATE(ZTOV(sharedir), zc->zc_string,
5165 &vattr, EXCL, 0, &resourcevp, kcred, 0, NULL, &vsec);
5166 if (resourcevp)
5167 VN_RELE(resourcevp);
5168 break;
5169
5170 case ZFS_SMB_ACL_REMOVE:
5171 error = VOP_REMOVE(ZTOV(sharedir), zc->zc_string, kcred,
5172 NULL, 0);
5173 break;
5174
5175 case ZFS_SMB_ACL_RENAME:
5176 if ((error = get_nvlist(zc->zc_nvlist_src,
5177 zc->zc_nvlist_src_size, zc->zc_iflags, &nvlist)) != 0) {
5178 VN_RELE(vp);
5179 VN_RELE(ZTOV(sharedir));
5180 ZFS_EXIT(zfsvfs);
5181 return (error);
5182 }
5183 if (nvlist_lookup_string(nvlist, ZFS_SMB_ACL_SRC, &src) ||
5184 nvlist_lookup_string(nvlist, ZFS_SMB_ACL_TARGET,
5185 &target)) {
5186 VN_RELE(vp);
5187 VN_RELE(ZTOV(sharedir));
5188 ZFS_EXIT(zfsvfs);
5189 nvlist_free(nvlist);
5190 return (error);
5191 }
5192 error = VOP_RENAME(ZTOV(sharedir), src, ZTOV(sharedir), target,
5193 kcred, NULL, 0);
5194 nvlist_free(nvlist);
5195 break;
5196
5197 case ZFS_SMB_ACL_PURGE:
5198 error = zfs_smb_acl_purge(sharedir);
5199 break;
5200
5201 default:
5202 error = SET_ERROR(EINVAL);
5203 break;
5204 }
5205
5206 VN_RELE(vp);
5207 VN_RELE(ZTOV(sharedir));
5208
5209 ZFS_EXIT(zfsvfs);
5210
5211 return (error);
5212 }
5213
5214 /*
5215 * innvl: {
5216 * "holds" -> { snapname -> holdname (string), ... }
5217 * (optional) "cleanup_fd" -> fd (int32)
5218 * }
5219 *
5220 * outnvl: {
5221 * snapname -> error value (int32)
5222 * ...
5223 * }
5224 */
5225 /* ARGSUSED */
5226 static int
5227 zfs_ioc_hold(const char *pool, nvlist_t *args, nvlist_t *errlist)
5228 {
5229 nvpair_t *pair;
5230 nvlist_t *holds;
5231 int cleanup_fd = -1;
5232 int error;
5233 minor_t minor = 0;
5234
5235 error = nvlist_lookup_nvlist(args, "holds", &holds);
5236 if (error != 0)
5237 return (SET_ERROR(EINVAL));
5238
5239 /* make sure the user didn't pass us any invalid (empty) tags */
5240 for (pair = nvlist_next_nvpair(holds, NULL); pair != NULL;
5241 pair = nvlist_next_nvpair(holds, pair)) {
5242 char *htag;
5243
5244 error = nvpair_value_string(pair, &htag);
5245 if (error != 0)
5246 return (SET_ERROR(error));
5247
5248 if (strlen(htag) == 0)
5249 return (SET_ERROR(EINVAL));
5250 }
5251
5252 if (nvlist_lookup_int32(args, "cleanup_fd", &cleanup_fd) == 0) {
5253 error = zfs_onexit_fd_hold(cleanup_fd, &minor);
5254 if (error != 0)
5255 return (error);
5256 }
5257
5258 error = dsl_dataset_user_hold(holds, minor, errlist);
5259 if (minor != 0)
5260 zfs_onexit_fd_rele(cleanup_fd);
5261 return (error);
5262 }
5263
5264 /*
5265 * innvl is not used.
5266 *
5267 * outnvl: {
5268 * holdname -> time added (uint64 seconds since epoch)
5269 * ...
5270 * }
5271 */
5272 /* ARGSUSED */
5273 static int
5274 zfs_ioc_get_holds(const char *snapname, nvlist_t *args, nvlist_t *outnvl)
5275 {
5276 return (dsl_dataset_get_holds(snapname, outnvl));
5277 }
5278
5279 /*
5280 * innvl: {
5281 * snapname -> { holdname, ... }
5282 * ...
5283 * }
5284 *
5285 * outnvl: {
5286 * snapname -> error value (int32)
5287 * ...
5288 * }
5289 */
5290 /* ARGSUSED */
5291 static int
5292 zfs_ioc_release(const char *pool, nvlist_t *holds, nvlist_t *errlist)
5293 {
5294 return (dsl_dataset_user_release(holds, errlist));
5295 }
5296
5297 /*
5298 * inputs:
5299 * zc_name name of new filesystem or snapshot
5300 * zc_value full name of old snapshot
5301 *
5302 * outputs:
5303 * zc_cookie space in bytes
5304 * zc_objset_type compressed space in bytes
5305 * zc_perm_action uncompressed space in bytes
5306 */
5307 static int
5308 zfs_ioc_space_written(zfs_cmd_t *zc)
5309 {
5310 int error;
5311 dsl_pool_t *dp;
5312 dsl_dataset_t *new, *old;
5313
5314 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
5315 if (error != 0)
5316 return (error);
5317 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &new);
5318 if (error != 0) {
5319 dsl_pool_rele(dp, FTAG);
5320 return (error);
5321 }
5322 error = dsl_dataset_hold(dp, zc->zc_value, FTAG, &old);
5323 if (error != 0) {
5324 dsl_dataset_rele(new, FTAG);
5325 dsl_pool_rele(dp, FTAG);
5326 return (error);
5327 }
5328
5329 error = dsl_dataset_space_written(old, new, &zc->zc_cookie,
5330 &zc->zc_objset_type, &zc->zc_perm_action);
5331 dsl_dataset_rele(old, FTAG);
5332 dsl_dataset_rele(new, FTAG);
5333 dsl_pool_rele(dp, FTAG);
5334 return (error);
5335 }
5336
5337 /*
5338 * innvl: {
5339 * "firstsnap" -> snapshot name
5340 * }
5341 *
5342 * outnvl: {
5343 * "used" -> space in bytes
5344 * "compressed" -> compressed space in bytes
5345 * "uncompressed" -> uncompressed space in bytes
5346 * }
5347 */
5348 static int
5349 zfs_ioc_space_snaps(const char *lastsnap, nvlist_t *innvl, nvlist_t *outnvl)
5350 {
5351 int error;
5352 dsl_pool_t *dp;
5353 dsl_dataset_t *new, *old;
5354 char *firstsnap;
5355 uint64_t used, comp, uncomp;
5356
5357 if (nvlist_lookup_string(innvl, "firstsnap", &firstsnap) != 0)
5358 return (SET_ERROR(EINVAL));
5359
5360 error = dsl_pool_hold(lastsnap, FTAG, &dp);
5361 if (error != 0)
5362 return (error);
5363
5364 error = dsl_dataset_hold(dp, lastsnap, FTAG, &new);
5365 if (error == 0 && !new->ds_is_snapshot) {
5366 dsl_dataset_rele(new, FTAG);
5367 error = SET_ERROR(EINVAL);
5368 }
5369 if (error != 0) {
5370 dsl_pool_rele(dp, FTAG);
5371 return (error);
5372 }
5373 error = dsl_dataset_hold(dp, firstsnap, FTAG, &old);
5374 if (error == 0 && !old->ds_is_snapshot) {
5375 dsl_dataset_rele(old, FTAG);
5376 error = SET_ERROR(EINVAL);
5377 }
5378 if (error != 0) {
5379 dsl_dataset_rele(new, FTAG);
5380 dsl_pool_rele(dp, FTAG);
5381 return (error);
5382 }
5383
5384 error = dsl_dataset_space_wouldfree(old, new, &used, &comp, &uncomp);
5385 dsl_dataset_rele(old, FTAG);
5386 dsl_dataset_rele(new, FTAG);
5387 dsl_pool_rele(dp, FTAG);
5388 fnvlist_add_uint64(outnvl, "used", used);
5389 fnvlist_add_uint64(outnvl, "compressed", comp);
5390 fnvlist_add_uint64(outnvl, "uncompressed", uncomp);
5391 return (error);
5392 }
5393
5394 /*
5395 * innvl: {
5396 * "fd" -> file descriptor to write stream to (int32)
5397 * (optional) "fromsnap" -> full snap name to send an incremental from
5398 * (optional) "largeblockok" -> (value ignored)
5399 * indicates that blocks > 128KB are permitted
5400 * (optional) "embedok" -> (value ignored)
5401 * presence indicates DRR_WRITE_EMBEDDED records are permitted
5402 * (optional) "resume_object" and "resume_offset" -> (uint64)
5403 * if present, resume send stream from specified object and offset.
5404 * }
5405 *
5406 * outnvl is unused
5407 */
5408 /* ARGSUSED */
5409 static int
5410 zfs_ioc_send_new(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl)
5411 {
5412 int error;
5413 offset_t off;
5414 char *fromname = NULL;
5415 int fd;
5416 boolean_t largeblockok;
5417 boolean_t embedok;
5418 uint64_t resumeobj = 0;
5419 uint64_t resumeoff = 0;
5420
5421 error = nvlist_lookup_int32(innvl, "fd", &fd);
5422 if (error != 0)
5423 return (SET_ERROR(EINVAL));
5424
5425 (void) nvlist_lookup_string(innvl, "fromsnap", &fromname);
5426
5427 largeblockok = nvlist_exists(innvl, "largeblockok");
5428 embedok = nvlist_exists(innvl, "embedok");
5429
5430 (void) nvlist_lookup_uint64(innvl, "resume_object", &resumeobj);
5431 (void) nvlist_lookup_uint64(innvl, "resume_offset", &resumeoff);
5432
5433 file_t *fp = getf(fd);
5434 if (fp == NULL)
5435 return (SET_ERROR(EBADF));
5436
5437 off = fp->f_offset;
5438 error = dmu_send(snapname, fromname, embedok, largeblockok, fd,
5439 resumeobj, resumeoff, fp->f_vnode, &off);
5440
5441 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
5442 fp->f_offset = off;
5443 releasef(fd);
5444 return (error);
5445 }
5446
5447 /*
5448 * Determine approximately how large a zfs send stream will be -- the number
5449 * of bytes that will be written to the fd supplied to zfs_ioc_send_new().
5450 *
5451 * innvl: {
5452 * (optional) "from" -> full snap or bookmark name to send an incremental
5453 * from
5454 * }
5455 *
5456 * outnvl: {
5457 * "space" -> bytes of space (uint64)
5458 * }
5459 */
5460 static int
5461 zfs_ioc_send_space(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl)
5462 {
5463 dsl_pool_t *dp;
5464 dsl_dataset_t *tosnap;
5465 int error;
5466 char *fromname;
5467 uint64_t space;
5468
5469 error = dsl_pool_hold(snapname, FTAG, &dp);
5470 if (error != 0)
5471 return (error);
5472
5473 error = dsl_dataset_hold(dp, snapname, FTAG, &tosnap);
5474 if (error != 0) {
5475 dsl_pool_rele(dp, FTAG);
5476 return (error);
5477 }
5478
5479 error = nvlist_lookup_string(innvl, "from", &fromname);
5480 if (error == 0) {
5481 if (strchr(fromname, '@') != NULL) {
5482 /*
5483 * If from is a snapshot, hold it and use the more
5484 * efficient dmu_send_estimate to estimate send space
5485 * size using deadlists.
5486 */
5487 dsl_dataset_t *fromsnap;
5488 error = dsl_dataset_hold(dp, fromname, FTAG, &fromsnap);
5489 if (error != 0)
5490 goto out;
5491 error = dmu_send_estimate(tosnap, fromsnap, &space);
5492 dsl_dataset_rele(fromsnap, FTAG);
5493 } else if (strchr(fromname, '#') != NULL) {
5494 /*
5495 * If from is a bookmark, fetch the creation TXG of the
5496 * snapshot it was created from and use that to find
5497 * blocks that were born after it.
5498 */
5499 zfs_bookmark_phys_t frombm;
5500
5501 error = dsl_bookmark_lookup(dp, fromname, tosnap,
5502 &frombm);
5503 if (error != 0)
5504 goto out;
5505 error = dmu_send_estimate_from_txg(tosnap,
5506 frombm.zbm_creation_txg, &space);
5507 } else {
5508 /*
5509 * from is not properly formatted as a snapshot or
5510 * bookmark
5511 */
5512 error = SET_ERROR(EINVAL);
5513 goto out;
5514 }
5515 } else {
5516 // If estimating the size of a full send, use dmu_send_estimate
5517 error = dmu_send_estimate(tosnap, NULL, &space);
5518 }
5519
5520 fnvlist_add_uint64(outnvl, "space", space);
5521
5522 out:
5523 dsl_dataset_rele(tosnap, FTAG);
5524 dsl_pool_rele(dp, FTAG);
5525 return (error);
5526 }
5527
5528 static zfs_ioc_vec_t zfs_ioc_vec[ZFS_IOC_LAST - ZFS_IOC_FIRST];
5529
5530 static void
5531 zfs_ioctl_register_legacy(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5532 zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck,
5533 boolean_t log_history, zfs_ioc_poolcheck_t pool_check)
5534 {
5535 zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST];
5536
5537 ASSERT3U(ioc, >=, ZFS_IOC_FIRST);
5538 ASSERT3U(ioc, <, ZFS_IOC_LAST);
5539 ASSERT3P(vec->zvec_legacy_func, ==, NULL);
5540 ASSERT3P(vec->zvec_func, ==, NULL);
5541
5542 vec->zvec_legacy_func = func;
5543 vec->zvec_secpolicy = secpolicy;
5544 vec->zvec_namecheck = namecheck;
5545 vec->zvec_allow_log = log_history;
5546 vec->zvec_pool_check = pool_check;
5547 }
5548
5549 /*
5550 * See the block comment at the beginning of this file for details on
5551 * each argument to this function.
5552 */
5553 static void
5554 zfs_ioctl_register(const char *name, zfs_ioc_t ioc, zfs_ioc_func_t *func,
5555 zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck,
5556 zfs_ioc_poolcheck_t pool_check, boolean_t smush_outnvlist,
5557 boolean_t allow_log)
5558 {
5559 zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST];
5560
5561 ASSERT3U(ioc, >=, ZFS_IOC_FIRST);
5562 ASSERT3U(ioc, <, ZFS_IOC_LAST);
5563 ASSERT3P(vec->zvec_legacy_func, ==, NULL);
5564 ASSERT3P(vec->zvec_func, ==, NULL);
5565
5566 /* if we are logging, the name must be valid */
5567 ASSERT(!allow_log || namecheck != NO_NAME);
5568
5569 vec->zvec_name = name;
5570 vec->zvec_func = func;
5571 vec->zvec_secpolicy = secpolicy;
5572 vec->zvec_namecheck = namecheck;
5573 vec->zvec_pool_check = pool_check;
5574 vec->zvec_smush_outnvlist = smush_outnvlist;
5575 vec->zvec_allow_log = allow_log;
5576 }
5577
5578 static void
5579 zfs_ioctl_register_pool(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5580 zfs_secpolicy_func_t *secpolicy, boolean_t log_history,
5581 zfs_ioc_poolcheck_t pool_check)
5582 {
5583 zfs_ioctl_register_legacy(ioc, func, secpolicy,
5584 POOL_NAME, log_history, pool_check);
5585 }
5586
5587 static void
5588 zfs_ioctl_register_dataset_nolog(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5589 zfs_secpolicy_func_t *secpolicy, zfs_ioc_poolcheck_t pool_check)
5590 {
5591 zfs_ioctl_register_legacy(ioc, func, secpolicy,
5592 DATASET_NAME, B_FALSE, pool_check);
5593 }
5594
5595 static void
5596 zfs_ioctl_register_pool_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func)
5597 {
5598 zfs_ioctl_register_legacy(ioc, func, zfs_secpolicy_config,
5599 POOL_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5600 }
5601
5602 static void
5603 zfs_ioctl_register_pool_meta(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5604 zfs_secpolicy_func_t *secpolicy)
5605 {
5606 zfs_ioctl_register_legacy(ioc, func, secpolicy,
5607 NO_NAME, B_FALSE, POOL_CHECK_NONE);
5608 }
5609
5610 static void
5611 zfs_ioctl_register_dataset_read_secpolicy(zfs_ioc_t ioc,
5612 zfs_ioc_legacy_func_t *func, zfs_secpolicy_func_t *secpolicy)
5613 {
5614 zfs_ioctl_register_legacy(ioc, func, secpolicy,
5615 DATASET_NAME, B_FALSE, POOL_CHECK_SUSPENDED);
5616 }
5617
5618 static void
5619 zfs_ioctl_register_dataset_read(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func)
5620 {
5621 zfs_ioctl_register_dataset_read_secpolicy(ioc, func,
5622 zfs_secpolicy_read);
5623 }
5624
5625 static void
5626 zfs_ioctl_register_dataset_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5627 zfs_secpolicy_func_t *secpolicy)
5628 {
5629 zfs_ioctl_register_legacy(ioc, func, secpolicy,
5630 DATASET_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5631 }
5632
5633 static void
5634 zfs_ioctl_init(void)
5635 {
5636 zfs_ioctl_register("snapshot", ZFS_IOC_SNAPSHOT,
5637 zfs_ioc_snapshot, zfs_secpolicy_snapshot, POOL_NAME,
5638 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5639
5640 zfs_ioctl_register("log_history", ZFS_IOC_LOG_HISTORY,
5641 zfs_ioc_log_history, zfs_secpolicy_log_history, NO_NAME,
5642 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_FALSE);
5643
5644 zfs_ioctl_register("space_snaps", ZFS_IOC_SPACE_SNAPS,
5645 zfs_ioc_space_snaps, zfs_secpolicy_read, DATASET_NAME,
5646 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5647
5648 zfs_ioctl_register("send", ZFS_IOC_SEND_NEW,
5649 zfs_ioc_send_new, zfs_secpolicy_send_new, DATASET_NAME,
5650 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5651
5652 zfs_ioctl_register("send_space", ZFS_IOC_SEND_SPACE,
5653 zfs_ioc_send_space, zfs_secpolicy_read, DATASET_NAME,
5654 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5655
5656 zfs_ioctl_register("create", ZFS_IOC_CREATE,
5657 zfs_ioc_create, zfs_secpolicy_create_clone, DATASET_NAME,
5658 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5659
5660 zfs_ioctl_register("clone", ZFS_IOC_CLONE,
5661 zfs_ioc_clone, zfs_secpolicy_create_clone, DATASET_NAME,
5662 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5663
5664 zfs_ioctl_register("destroy_snaps", ZFS_IOC_DESTROY_SNAPS,
5665 zfs_ioc_destroy_snaps, zfs_secpolicy_destroy_snaps, POOL_NAME,
5666 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5667
5668 zfs_ioctl_register("hold", ZFS_IOC_HOLD,
5669 zfs_ioc_hold, zfs_secpolicy_hold, POOL_NAME,
5670 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5671 zfs_ioctl_register("release", ZFS_IOC_RELEASE,
5672 zfs_ioc_release, zfs_secpolicy_release, POOL_NAME,
5673 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5674
5675 zfs_ioctl_register("get_holds", ZFS_IOC_GET_HOLDS,
5676 zfs_ioc_get_holds, zfs_secpolicy_read, DATASET_NAME,
5677 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5678
5679 zfs_ioctl_register("rollback", ZFS_IOC_ROLLBACK,
5680 zfs_ioc_rollback, zfs_secpolicy_rollback, DATASET_NAME,
5681 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_TRUE);
5682
5683 zfs_ioctl_register("bookmark", ZFS_IOC_BOOKMARK,
5684 zfs_ioc_bookmark, zfs_secpolicy_bookmark, POOL_NAME,
5685 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5686
5687 zfs_ioctl_register("get_bookmarks", ZFS_IOC_GET_BOOKMARKS,
5688 zfs_ioc_get_bookmarks, zfs_secpolicy_read, DATASET_NAME,
5689 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5690
5691 zfs_ioctl_register("destroy_bookmarks", ZFS_IOC_DESTROY_BOOKMARKS,
5692 zfs_ioc_destroy_bookmarks, zfs_secpolicy_destroy_bookmarks,
5693 POOL_NAME,
5694 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5695
5696 /* IOCTLS that use the legacy function signature */
5697
5698 zfs_ioctl_register_legacy(ZFS_IOC_POOL_FREEZE, zfs_ioc_pool_freeze,
5699 zfs_secpolicy_config, NO_NAME, B_FALSE, POOL_CHECK_READONLY);
5700
5701 zfs_ioctl_register_pool(ZFS_IOC_POOL_CREATE, zfs_ioc_pool_create,
5702 zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE);
5703 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SCAN,
5704 zfs_ioc_pool_scan);
5705 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_UPGRADE,
5706 zfs_ioc_pool_upgrade);
5707 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ADD,
5708 zfs_ioc_vdev_add);
5709 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_REMOVE,
5710 zfs_ioc_vdev_remove);
5711 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SET_STATE,
5712 zfs_ioc_vdev_set_state);
5713 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ATTACH,
5714 zfs_ioc_vdev_attach);
5715 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_DETACH,
5716 zfs_ioc_vdev_detach);
5717 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETPATH,
5718 zfs_ioc_vdev_setpath);
5719 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETFRU,
5720 zfs_ioc_vdev_setfru);
5721 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SET_PROPS,
5722 zfs_ioc_pool_set_props);
5723 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SPLIT,
5724 zfs_ioc_vdev_split);
5725 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_REGUID,
5726 zfs_ioc_pool_reguid);
5727
5728 zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_CONFIGS,
5729 zfs_ioc_pool_configs, zfs_secpolicy_none);
5730 zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_TRYIMPORT,
5731 zfs_ioc_pool_tryimport, zfs_secpolicy_config);
5732 zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_FAULT,
5733 zfs_ioc_inject_fault, zfs_secpolicy_inject);
5734 zfs_ioctl_register_pool_meta(ZFS_IOC_CLEAR_FAULT,
5735 zfs_ioc_clear_fault, zfs_secpolicy_inject);
5736 zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_LIST_NEXT,
5737 zfs_ioc_inject_list_next, zfs_secpolicy_inject);
5738
5739 /*
5740 * pool destroy, and export don't log the history as part of
5741 * zfsdev_ioctl, but rather zfs_ioc_pool_export
5742 * does the logging of those commands.
5743 */
5744 zfs_ioctl_register_pool(ZFS_IOC_POOL_DESTROY, zfs_ioc_pool_destroy,
5745 zfs_secpolicy_config, B_FALSE, POOL_CHECK_NONE);
5746 zfs_ioctl_register_pool(ZFS_IOC_POOL_EXPORT, zfs_ioc_pool_export,
5747 zfs_secpolicy_config, B_FALSE, POOL_CHECK_NONE);
5748
5749 zfs_ioctl_register_pool(ZFS_IOC_POOL_STATS, zfs_ioc_pool_stats,
5750 zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE);
5751 zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_PROPS, zfs_ioc_pool_get_props,
5752 zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE);
5753
5754 zfs_ioctl_register_pool(ZFS_IOC_ERROR_LOG, zfs_ioc_error_log,
5755 zfs_secpolicy_inject, B_FALSE, POOL_CHECK_SUSPENDED);
5756 zfs_ioctl_register_pool(ZFS_IOC_DSOBJ_TO_DSNAME,
5757 zfs_ioc_dsobj_to_dsname,
5758 zfs_secpolicy_diff, B_FALSE, POOL_CHECK_SUSPENDED);
5759 zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_HISTORY,
5760 zfs_ioc_pool_get_history,
5761 zfs_secpolicy_config, B_FALSE, POOL_CHECK_SUSPENDED);
5762
5763 zfs_ioctl_register_pool(ZFS_IOC_POOL_IMPORT, zfs_ioc_pool_import,
5764 zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE);
5765
5766 zfs_ioctl_register_pool(ZFS_IOC_CLEAR, zfs_ioc_clear,
5767 zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE);
5768 zfs_ioctl_register_pool(ZFS_IOC_POOL_REOPEN, zfs_ioc_pool_reopen,
5769 zfs_secpolicy_config, B_TRUE, POOL_CHECK_SUSPENDED);
5770
5771 zfs_ioctl_register_dataset_read(ZFS_IOC_SPACE_WRITTEN,
5772 zfs_ioc_space_written);
5773 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_RECVD_PROPS,
5774 zfs_ioc_objset_recvd_props);
5775 zfs_ioctl_register_dataset_read(ZFS_IOC_NEXT_OBJ,
5776 zfs_ioc_next_obj);
5777 zfs_ioctl_register_dataset_read(ZFS_IOC_GET_FSACL,
5778 zfs_ioc_get_fsacl);
5779 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_STATS,
5780 zfs_ioc_objset_stats);
5781 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_ZPLPROPS,
5782 zfs_ioc_objset_zplprops);
5783 zfs_ioctl_register_dataset_read(ZFS_IOC_DATASET_LIST_NEXT,
5784 zfs_ioc_dataset_list_next);
5785 zfs_ioctl_register_dataset_read(ZFS_IOC_SNAPSHOT_LIST_NEXT,
5786 zfs_ioc_snapshot_list_next);
5787 zfs_ioctl_register_dataset_read(ZFS_IOC_SEND_PROGRESS,
5788 zfs_ioc_send_progress);
5789
5790 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_DIFF,
5791 zfs_ioc_diff, zfs_secpolicy_diff);
5792 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_STATS,
5793 zfs_ioc_obj_to_stats, zfs_secpolicy_diff);
5794 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_PATH,
5795 zfs_ioc_obj_to_path, zfs_secpolicy_diff);
5796 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_ONE,
5797 zfs_ioc_userspace_one, zfs_secpolicy_userspace_one);
5798 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_MANY,
5799 zfs_ioc_userspace_many, zfs_secpolicy_userspace_many);
5800 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_SEND,
5801 zfs_ioc_send, zfs_secpolicy_send);
5802
5803 zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_PROP, zfs_ioc_set_prop,
5804 zfs_secpolicy_none);
5805 zfs_ioctl_register_dataset_modify(ZFS_IOC_DESTROY, zfs_ioc_destroy,
5806 zfs_secpolicy_destroy);
5807 zfs_ioctl_register_dataset_modify(ZFS_IOC_RENAME, zfs_ioc_rename,
5808 zfs_secpolicy_rename);
5809 zfs_ioctl_register_dataset_modify(ZFS_IOC_RECV, zfs_ioc_recv,
5810 zfs_secpolicy_recv);
5811 zfs_ioctl_register_dataset_modify(ZFS_IOC_PROMOTE, zfs_ioc_promote,
5812 zfs_secpolicy_promote);
5813 zfs_ioctl_register_dataset_modify(ZFS_IOC_INHERIT_PROP,
5814 zfs_ioc_inherit_prop, zfs_secpolicy_inherit_prop);
5815 zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_FSACL, zfs_ioc_set_fsacl,
5816 zfs_secpolicy_set_fsacl);
5817
5818 zfs_ioctl_register_dataset_nolog(ZFS_IOC_SHARE, zfs_ioc_share,
5819 zfs_secpolicy_share, POOL_CHECK_NONE);
5820 zfs_ioctl_register_dataset_nolog(ZFS_IOC_SMB_ACL, zfs_ioc_smb_acl,
5821 zfs_secpolicy_smb_acl, POOL_CHECK_NONE);
5822 zfs_ioctl_register_dataset_nolog(ZFS_IOC_USERSPACE_UPGRADE,
5823 zfs_ioc_userspace_upgrade, zfs_secpolicy_userspace_upgrade,
5824 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5825 zfs_ioctl_register_dataset_nolog(ZFS_IOC_TMP_SNAPSHOT,
5826 zfs_ioc_tmp_snapshot, zfs_secpolicy_tmp_snapshot,
5827 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5828 }
5829
5830 int
5831 pool_status_check(const char *name, zfs_ioc_namecheck_t type,
5832 zfs_ioc_poolcheck_t check)
5833 {
5834 spa_t *spa;
5835 int error;
5836
5837 ASSERT(type == POOL_NAME || type == DATASET_NAME);
5838
5839 if (check & POOL_CHECK_NONE)
5840 return (0);
5841
5842 error = spa_open(name, &spa, FTAG);
5843 if (error == 0) {
5844 if ((check & POOL_CHECK_SUSPENDED) && spa_suspended(spa))
5845 error = SET_ERROR(EAGAIN);
5846 else if ((check & POOL_CHECK_READONLY) && !spa_writeable(spa))
5847 error = SET_ERROR(EROFS);
5848 spa_close(spa, FTAG);
5849 }
5850 return (error);
5851 }
5852
5853 /*
5854 * Find a free minor number.
5855 */
5856 minor_t
5857 zfsdev_minor_alloc(void)
5858 {
5859 static minor_t last_minor;
5860 minor_t m;
5861
5862 ASSERT(MUTEX_HELD(&zfsdev_state_lock));
5863
5864 for (m = last_minor + 1; m != last_minor; m++) {
5865 if (m > ZFSDEV_MAX_MINOR)
5866 m = 1;
5867 if (ddi_get_soft_state(zfsdev_state, m) == NULL) {
5868 last_minor = m;
5869 return (m);
5870 }
5871 }
5872
5873 return (0);
5874 }
5875
5876 static int
5877 zfs_ctldev_init(dev_t *devp)
5878 {
5879 minor_t minor;
5880 zfs_soft_state_t *zs;
5881
5882 ASSERT(MUTEX_HELD(&zfsdev_state_lock));
5883 ASSERT(getminor(*devp) == 0);
5884
5885 minor = zfsdev_minor_alloc();
5886 if (minor == 0)
5887 return (SET_ERROR(ENXIO));
5888
5889 if (ddi_soft_state_zalloc(zfsdev_state, minor) != DDI_SUCCESS)
5890 return (SET_ERROR(EAGAIN));
5891
5892 *devp = makedevice(getemajor(*devp), minor);
5893
5894 zs = ddi_get_soft_state(zfsdev_state, minor);
5895 zs->zss_type = ZSST_CTLDEV;
5896 zfs_onexit_init((zfs_onexit_t **)&zs->zss_data);
5897
5898 return (0);
5899 }
5900
5901 static void
5902 zfs_ctldev_destroy(zfs_onexit_t *zo, minor_t minor)
5903 {
5904 ASSERT(MUTEX_HELD(&zfsdev_state_lock));
5905
5906 zfs_onexit_destroy(zo);
5907 ddi_soft_state_free(zfsdev_state, minor);
5908 }
5909
5910 void *
5911 zfsdev_get_soft_state(minor_t minor, enum zfs_soft_state_type which)
5912 {
5913 zfs_soft_state_t *zp;
5914
5915 zp = ddi_get_soft_state(zfsdev_state, minor);
5916 if (zp == NULL || zp->zss_type != which)
5917 return (NULL);
5918
5919 return (zp->zss_data);
5920 }
5921
5922 static int
5923 zfsdev_open(dev_t *devp, int flag, int otyp, cred_t *cr)
5924 {
5925 int error = 0;
5926
5927 if (getminor(*devp) != 0)
5928 return (zvol_open(devp, flag, otyp, cr));
5929
5930 /* This is the control device. Allocate a new minor if requested. */
5931 if (flag & FEXCL) {
5932 mutex_enter(&zfsdev_state_lock);
5933 error = zfs_ctldev_init(devp);
5934 mutex_exit(&zfsdev_state_lock);
5935 }
5936
5937 return (error);
5938 }
5939
5940 static int
5941 zfsdev_close(dev_t dev, int flag, int otyp, cred_t *cr)
5942 {
5943 zfs_onexit_t *zo;
5944 minor_t minor = getminor(dev);
5945
5946 if (minor == 0)
5947 return (0);
5948
5949 mutex_enter(&zfsdev_state_lock);
5950 zo = zfsdev_get_soft_state(minor, ZSST_CTLDEV);
5951 if (zo == NULL) {
5952 mutex_exit(&zfsdev_state_lock);
5953 return (zvol_close(dev, flag, otyp, cr));
5954 }
5955 zfs_ctldev_destroy(zo, minor);
5956 mutex_exit(&zfsdev_state_lock);
5957
5958 return (0);
5959 }
5960
5961 static int
5962 zfsdev_ioctl(dev_t dev, int cmd, intptr_t arg, int flag, cred_t *cr, int *rvalp)
5963 {
5964 zfs_cmd_t *zc;
5965 uint_t vecnum;
5966 int error, rc, len;
5967 minor_t minor = getminor(dev);
5968 const zfs_ioc_vec_t *vec;
5969 char *saved_poolname = NULL;
5970 nvlist_t *innvl = NULL;
5971
5972 if (minor != 0 &&
5973 zfsdev_get_soft_state(minor, ZSST_CTLDEV) == NULL)
5974 return (zvol_ioctl(dev, cmd, arg, flag, cr, rvalp));
5975
5976 vecnum = cmd - ZFS_IOC_FIRST;
5977 ASSERT3U(getmajor(dev), ==, ddi_driver_major(zfs_dip));
5978
5979 if (vecnum >= sizeof (zfs_ioc_vec) / sizeof (zfs_ioc_vec[0]))
5980 return (SET_ERROR(EINVAL));
5981 vec = &zfs_ioc_vec[vecnum];
5982
5983 zc = kmem_zalloc(sizeof (zfs_cmd_t), KM_SLEEP);
5984
5985 error = ddi_copyin((void *)arg, zc, sizeof (zfs_cmd_t), flag);
5986 if (error != 0) {
5987 error = SET_ERROR(EFAULT);
5988 goto out;
5989 }
5990
5991 zc->zc_iflags = flag & FKIOCTL;
5992 if (zc->zc_nvlist_src_size != 0) {
5993 error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
5994 zc->zc_iflags, &innvl);
5995 if (error != 0)
5996 goto out;
5997 }
5998
5999 /*
6000 * Ensure that all pool/dataset names are valid before we pass down to
6001 * the lower layers.
6002 */
6003 zc->zc_name[sizeof (zc->zc_name) - 1] = '\0';
6004 switch (vec->zvec_namecheck) {
6005 case POOL_NAME:
6006 if (pool_namecheck(zc->zc_name, NULL, NULL) != 0)
6007 error = SET_ERROR(EINVAL);
6008 else
6009 error = pool_status_check(zc->zc_name,
6010 vec->zvec_namecheck, vec->zvec_pool_check);
6011 break;
6012
6013 case DATASET_NAME:
6014 if (dataset_namecheck(zc->zc_name, NULL, NULL) != 0)
6015 error = SET_ERROR(EINVAL);
6016 else
6017 error = pool_status_check(zc->zc_name,
6018 vec->zvec_namecheck, vec->zvec_pool_check);
6019 break;
6020
6021 case NO_NAME:
6022 break;
6023 }
6024
6025
6026 if (error == 0)
6027 error = vec->zvec_secpolicy(zc, innvl, cr);
6028
6029 if (error != 0)
6030 goto out;
6031
6032 /* legacy ioctls can modify zc_name */
6033 len = strcspn(zc->zc_name, "/@#") + 1;
6034 saved_poolname = kmem_alloc(len, KM_SLEEP);
6035 (void) strlcpy(saved_poolname, zc->zc_name, len);
6036
6037 if (vec->zvec_func != NULL) {
6038 nvlist_t *outnvl;
6039 int puterror = 0;
6040 spa_t *spa;
6041 nvlist_t *lognv = NULL;
6042
6043 ASSERT(vec->zvec_legacy_func == NULL);
6044
6045 /*
6046 * Add the innvl to the lognv before calling the func,
6047 * in case the func changes the innvl.
6048 */
6049 if (vec->zvec_allow_log) {
6050 lognv = fnvlist_alloc();
6051 fnvlist_add_string(lognv, ZPOOL_HIST_IOCTL,
6052 vec->zvec_name);
6053 if (!nvlist_empty(innvl)) {
6054 fnvlist_add_nvlist(lognv, ZPOOL_HIST_INPUT_NVL,
6055 innvl);
6056 }
6057 }
6058
6059 outnvl = fnvlist_alloc();
6060 error = vec->zvec_func(zc->zc_name, innvl, outnvl);
6061
6062 if (error == 0 && vec->zvec_allow_log &&
6063 spa_open(zc->zc_name, &spa, FTAG) == 0) {
6064 if (!nvlist_empty(outnvl)) {
6065 fnvlist_add_nvlist(lognv, ZPOOL_HIST_OUTPUT_NVL,
6066 outnvl);
6067 }
6068 (void) spa_history_log_nvl(spa, lognv);
6069 spa_close(spa, FTAG);
6070 }
6071 fnvlist_free(lognv);
6072
6073 if (!nvlist_empty(outnvl) || zc->zc_nvlist_dst_size != 0) {
6074 int smusherror = 0;
6075 if (vec->zvec_smush_outnvlist) {
6076 smusherror = nvlist_smush(outnvl,
6077 zc->zc_nvlist_dst_size);
6078 }
6079 if (smusherror == 0)
6080 puterror = put_nvlist(zc, outnvl);
6081 }
6082
6083 if (puterror != 0)
6084 error = puterror;
6085
6086 nvlist_free(outnvl);
6087 } else {
6088 error = vec->zvec_legacy_func(zc);
6089 }
6090
6091 out:
6092 nvlist_free(innvl);
6093 rc = ddi_copyout(zc, (void *)arg, sizeof (zfs_cmd_t), flag);
6094 if (error == 0 && rc != 0)
6095 error = SET_ERROR(EFAULT);
6096 if (error == 0 && vec->zvec_allow_log) {
6097 char *s = tsd_get(zfs_allow_log_key);
6098 if (s != NULL)
6099 strfree(s);
6100 (void) tsd_set(zfs_allow_log_key, saved_poolname);
6101 } else {
6102 if (saved_poolname != NULL)
6103 strfree(saved_poolname);
6104 }
6105
6106 kmem_free(zc, sizeof (zfs_cmd_t));
6107 return (error);
6108 }
6109
6110 static int
6111 zfs_attach(dev_info_t *dip, ddi_attach_cmd_t cmd)
6112 {
6113 if (cmd != DDI_ATTACH)
6114 return (DDI_FAILURE);
6115
6116 if (ddi_create_minor_node(dip, "zfs", S_IFCHR, 0,
6117 DDI_PSEUDO, 0) == DDI_FAILURE)
6118 return (DDI_FAILURE);
6119
6120 zfs_dip = dip;
6121
6122 ddi_report_dev(dip);
6123
6124 return (DDI_SUCCESS);
6125 }
6126
6127 static int
6128 zfs_detach(dev_info_t *dip, ddi_detach_cmd_t cmd)
6129 {
6130 if (spa_busy() || zfs_busy() || zvol_busy())
6131 return (DDI_FAILURE);
6132
6133 if (cmd != DDI_DETACH)
6134 return (DDI_FAILURE);
6135
6136 zfs_dip = NULL;
6137
6138 ddi_prop_remove_all(dip);
6139 ddi_remove_minor_node(dip, NULL);
6140
6141 return (DDI_SUCCESS);
6142 }
6143
6144 /*ARGSUSED*/
6145 static int
6146 zfs_info(dev_info_t *dip, ddi_info_cmd_t infocmd, void *arg, void **result)
6147 {
6148 switch (infocmd) {
6149 case DDI_INFO_DEVT2DEVINFO:
6150 *result = zfs_dip;
6151 return (DDI_SUCCESS);
6152
6153 case DDI_INFO_DEVT2INSTANCE:
6154 *result = (void *)0;
6155 return (DDI_SUCCESS);
6156 }
6157
6158 return (DDI_FAILURE);
6159 }
6160
6161 /*
6162 * OK, so this is a little weird.
6163 *
6164 * /dev/zfs is the control node, i.e. minor 0.
6165 * /dev/zvol/[r]dsk/pool/dataset are the zvols, minor > 0.
6166 *
6167 * /dev/zfs has basically nothing to do except serve up ioctls,
6168 * so most of the standard driver entry points are in zvol.c.
6169 */
6170 static struct cb_ops zfs_cb_ops = {
6171 zfsdev_open, /* open */
6172 zfsdev_close, /* close */
6173 zvol_strategy, /* strategy */
6174 nodev, /* print */
6175 zvol_dump, /* dump */
6176 zvol_read, /* read */
6177 zvol_write, /* write */
6178 zfsdev_ioctl, /* ioctl */
6179 nodev, /* devmap */
6180 nodev, /* mmap */
6181 nodev, /* segmap */
6182 nochpoll, /* poll */
6183 ddi_prop_op, /* prop_op */
6184 NULL, /* streamtab */
6185 D_NEW | D_MP | D_64BIT, /* Driver compatibility flag */
6186 CB_REV, /* version */
6187 nodev, /* async read */
6188 nodev, /* async write */
6189 };
6190
6191 static struct dev_ops zfs_dev_ops = {
6192 DEVO_REV, /* version */
6193 0, /* refcnt */
6194 zfs_info, /* info */
6195 nulldev, /* identify */
6196 nulldev, /* probe */
6197 zfs_attach, /* attach */
6198 zfs_detach, /* detach */
6199 nodev, /* reset */
6200 &zfs_cb_ops, /* driver operations */
6201 NULL, /* no bus operations */
6202 NULL, /* power */
6203 ddi_quiesce_not_needed, /* quiesce */
6204 };
6205
6206 static struct modldrv zfs_modldrv = {
6207 &mod_driverops,
6208 "ZFS storage pool",
6209 &zfs_dev_ops
6210 };
6211
6212 static struct modlinkage modlinkage = {
6213 MODREV_1,
6214 (void *)&zfs_modlfs,
6215 (void *)&zfs_modldrv,
6216 NULL
6217 };
6218
6219 static void
6220 zfs_allow_log_destroy(void *arg)
6221 {
6222 char *poolname = arg;
6223 strfree(poolname);
6224 }
6225
6226 int
6227 _init(void)
6228 {
6229 int error;
6230
6231 spa_init(FREAD | FWRITE);
6232 zfs_init();
6233 zvol_init();
6234 zfs_ioctl_init();
6235
6236 if ((error = mod_install(&modlinkage)) != 0) {
6237 zvol_fini();
6238 zfs_fini();
6239 spa_fini();
6240 return (error);
6241 }
6242
6243 tsd_create(&zfs_fsyncer_key, NULL);
6244 tsd_create(&rrw_tsd_key, rrw_tsd_destroy);
6245 tsd_create(&zfs_allow_log_key, zfs_allow_log_destroy);
6246
6247 error = ldi_ident_from_mod(&modlinkage, &zfs_li);
6248 ASSERT(error == 0);
6249 mutex_init(&zfs_share_lock, NULL, MUTEX_DEFAULT, NULL);
6250
6251 return (0);
6252 }
6253
6254 int
6255 _fini(void)
6256 {
6257 int error;
6258
6259 if (spa_busy() || zfs_busy() || zvol_busy() || zio_injection_enabled)
6260 return (SET_ERROR(EBUSY));
6261
6262 if ((error = mod_remove(&modlinkage)) != 0)
6263 return (error);
6264
6265 zvol_fini();
6266 zfs_fini();
6267 spa_fini();
6268 if (zfs_nfsshare_inited)
6269 (void) ddi_modclose(nfs_mod);
6270 if (zfs_smbshare_inited)
6271 (void) ddi_modclose(smbsrv_mod);
6272 if (zfs_nfsshare_inited || zfs_smbshare_inited)
6273 (void) ddi_modclose(sharefs_mod);
6274
6275 tsd_destroy(&zfs_fsyncer_key);
6276 ldi_ident_release(zfs_li);
6277 zfs_li = NULL;
6278 mutex_destroy(&zfs_share_lock);
6279
6280 return (error);
6281 }
6282
6283 int
6284 _info(struct modinfo *modinfop)
6285 {
6286 return (mod_info(&modlinkage, modinfop));
6287 }