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 * Copyright 2016 Toomas Soome <tsoome@me.com>
34 */
35
36 /*
37 * ZFS ioctls.
38 *
39 * This file handles the ioctls to /dev/zfs, used for configuring ZFS storage
40 * pools and filesystems, e.g. with /sbin/zfs and /sbin/zpool.
41 *
42 * There are two ways that we handle ioctls: the legacy way where almost
43 * all of the logic is in the ioctl callback, and the new way where most
44 * of the marshalling is handled in the common entry point, zfsdev_ioctl().
45 *
46 * Non-legacy ioctls should be registered by calling
47 * zfs_ioctl_register() from zfs_ioctl_init(). The ioctl is invoked
48 * from userland by lzc_ioctl().
49 *
50 * The registration arguments are as follows:
51 *
52 * const char *name
53 * The name of the ioctl. This is used for history logging. If the
54 * ioctl returns successfully (the callback returns 0), and allow_log
55 * is true, then a history log entry will be recorded with the input &
56 * output nvlists. The log entry can be printed with "zpool history -i".
57 *
58 * zfs_ioc_t ioc
59 * The ioctl request number, which userland will pass to ioctl(2).
60 * The ioctl numbers can change from release to release, because
61 * the caller (libzfs) must be matched to the kernel.
62 *
63 * zfs_secpolicy_func_t *secpolicy
64 * This function will be called before the zfs_ioc_func_t, to
65 * determine if this operation is permitted. It should return EPERM
66 * on failure, and 0 on success. Checks include determining if the
67 * dataset is visible in this zone, and if the user has either all
68 * zfs privileges in the zone (SYS_MOUNT), or has been granted permission
69 * to do this operation on this dataset with "zfs allow".
70 *
71 * zfs_ioc_namecheck_t namecheck
72 * This specifies what to expect in the zfs_cmd_t:zc_name -- a pool
73 * name, a dataset name, or nothing. If the name is not well-formed,
74 * the ioctl will fail and the callback will not be called.
75 * Therefore, the callback can assume that the name is well-formed
76 * (e.g. is null-terminated, doesn't have more than one '@' character,
77 * doesn't have invalid characters).
78 *
79 * zfs_ioc_poolcheck_t pool_check
80 * This specifies requirements on the pool state. If the pool does
81 * not meet them (is suspended or is readonly), the ioctl will fail
82 * and the callback will not be called. If any checks are specified
83 * (i.e. it is not POOL_CHECK_NONE), namecheck must not be NO_NAME.
84 * Multiple checks can be or-ed together (e.g. POOL_CHECK_SUSPENDED |
85 * POOL_CHECK_READONLY).
86 *
87 * boolean_t smush_outnvlist
88 * If smush_outnvlist is true, then the output is presumed to be a
89 * list of errors, and it will be "smushed" down to fit into the
90 * caller's buffer, by removing some entries and replacing them with a
91 * single "N_MORE_ERRORS" entry indicating how many were removed. See
92 * nvlist_smush() for details. If smush_outnvlist is false, and the
93 * outnvlist does not fit into the userland-provided buffer, then the
94 * ioctl will fail with ENOMEM.
95 *
96 * zfs_ioc_func_t *func
97 * The callback function that will perform the operation.
98 *
99 * The callback should return 0 on success, or an error number on
100 * failure. If the function fails, the userland ioctl will return -1,
101 * and errno will be set to the callback's return value. The callback
102 * will be called with the following arguments:
103 *
104 * const char *name
105 * The name of the pool or dataset to operate on, from
106 * zfs_cmd_t:zc_name. The 'namecheck' argument specifies the
107 * expected type (pool, dataset, or none).
108 *
109 * nvlist_t *innvl
110 * The input nvlist, deserialized from zfs_cmd_t:zc_nvlist_src. Or
111 * NULL if no input nvlist was provided. Changes to this nvlist are
112 * ignored. If the input nvlist could not be deserialized, the
113 * ioctl will fail and the callback will not be called.
114 *
115 * nvlist_t *outnvl
116 * The output nvlist, initially empty. The callback can fill it in,
117 * and it will be returned to userland by serializing it into
118 * zfs_cmd_t:zc_nvlist_dst. If it is non-empty, and serialization
119 * fails (e.g. because the caller didn't supply a large enough
120 * buffer), then the overall ioctl will fail. See the
121 * 'smush_nvlist' argument above for additional behaviors.
122 *
123 * There are two typical uses of the output nvlist:
124 * - To return state, e.g. property values. In this case,
125 * smush_outnvlist should be false. If the buffer was not large
126 * enough, the caller will reallocate a larger buffer and try
127 * the ioctl again.
128 *
129 * - To return multiple errors from an ioctl which makes on-disk
130 * changes. In this case, smush_outnvlist should be true.
131 * Ioctls which make on-disk modifications should generally not
132 * use the outnvl if they succeed, because the caller can not
133 * distinguish between the operation failing, and
134 * deserialization failing.
135 */
136
137 #include <sys/types.h>
138 #include <sys/param.h>
139 #include <sys/errno.h>
140 #include <sys/uio.h>
141 #include <sys/buf.h>
142 #include <sys/modctl.h>
143 #include <sys/open.h>
144 #include <sys/file.h>
145 #include <sys/kmem.h>
146 #include <sys/conf.h>
147 #include <sys/cmn_err.h>
148 #include <sys/stat.h>
149 #include <sys/zfs_ioctl.h>
150 #include <sys/zfs_vfsops.h>
151 #include <sys/zfs_znode.h>
152 #include <sys/zap.h>
153 #include <sys/spa.h>
154 #include <sys/spa_impl.h>
155 #include <sys/vdev.h>
156 #include <sys/priv_impl.h>
157 #include <sys/dmu.h>
158 #include <sys/dsl_dir.h>
159 #include <sys/dsl_dataset.h>
160 #include <sys/dsl_prop.h>
161 #include <sys/dsl_deleg.h>
162 #include <sys/dmu_objset.h>
163 #include <sys/dmu_impl.h>
164 #include <sys/dmu_tx.h>
165 #include <sys/ddi.h>
166 #include <sys/sunddi.h>
167 #include <sys/sunldi.h>
168 #include <sys/policy.h>
169 #include <sys/zone.h>
170 #include <sys/nvpair.h>
171 #include <sys/pathname.h>
172 #include <sys/mount.h>
173 #include <sys/sdt.h>
174 #include <sys/fs/zfs.h>
175 #include <sys/zfs_ctldir.h>
176 #include <sys/zfs_dir.h>
177 #include <sys/zfs_onexit.h>
178 #include <sys/zvol.h>
179 #include <sys/dsl_scan.h>
180 #include <sharefs/share.h>
181 #include <sys/dmu_objset.h>
182 #include <sys/dmu_send.h>
183 #include <sys/dsl_destroy.h>
184 #include <sys/dsl_bookmark.h>
185 #include <sys/dsl_userhold.h>
186 #include <sys/zfeature.h>
187 #include <sys/zio_checksum.h>
188
189 #include "zfs_namecheck.h"
190 #include "zfs_prop.h"
191 #include "zfs_deleg.h"
192 #include "zfs_comutil.h"
193
194 extern struct modlfs zfs_modlfs;
195
196 extern void zfs_init(void);
197 extern void zfs_fini(void);
198
199 ldi_ident_t zfs_li = NULL;
200 dev_info_t *zfs_dip;
201
202 uint_t zfs_fsyncer_key;
203 extern uint_t rrw_tsd_key;
204 static uint_t zfs_allow_log_key;
205
206 typedef int zfs_ioc_legacy_func_t(zfs_cmd_t *);
207 typedef int zfs_ioc_func_t(const char *, nvlist_t *, nvlist_t *);
208 typedef int zfs_secpolicy_func_t(zfs_cmd_t *, nvlist_t *, cred_t *);
209
210 typedef enum {
211 NO_NAME,
212 POOL_NAME,
213 DATASET_NAME
214 } zfs_ioc_namecheck_t;
215
216 typedef enum {
217 POOL_CHECK_NONE = 1 << 0,
218 POOL_CHECK_SUSPENDED = 1 << 1,
219 POOL_CHECK_READONLY = 1 << 2,
220 } zfs_ioc_poolcheck_t;
221
222 typedef struct zfs_ioc_vec {
223 zfs_ioc_legacy_func_t *zvec_legacy_func;
224 zfs_ioc_func_t *zvec_func;
225 zfs_secpolicy_func_t *zvec_secpolicy;
226 zfs_ioc_namecheck_t zvec_namecheck;
227 boolean_t zvec_allow_log;
228 zfs_ioc_poolcheck_t zvec_pool_check;
229 boolean_t zvec_smush_outnvlist;
230 const char *zvec_name;
231 } zfs_ioc_vec_t;
232
233 /* This array is indexed by zfs_userquota_prop_t */
234 static const char *userquota_perms[] = {
235 ZFS_DELEG_PERM_USERUSED,
236 ZFS_DELEG_PERM_USERQUOTA,
237 ZFS_DELEG_PERM_GROUPUSED,
238 ZFS_DELEG_PERM_GROUPQUOTA,
239 };
240
241 static int zfs_ioc_userspace_upgrade(zfs_cmd_t *zc);
242 static int zfs_check_settable(const char *name, nvpair_t *property,
243 cred_t *cr);
244 static int zfs_check_clearable(char *dataset, nvlist_t *props,
245 nvlist_t **errors);
246 static int zfs_fill_zplprops_root(uint64_t, nvlist_t *, nvlist_t *,
247 boolean_t *);
248 int zfs_set_prop_nvlist(const char *, zprop_source_t, nvlist_t *, nvlist_t *);
249 static int get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp);
250
251 static int zfs_prop_activate_feature(spa_t *spa, spa_feature_t feature);
252
253 /* _NOTE(PRINTFLIKE(4)) - this is printf-like, but lint is too whiney */
254 void
255 __dprintf(const char *file, const char *func, int line, const char *fmt, ...)
256 {
257 const char *newfile;
258 char buf[512];
259 va_list adx;
260
261 /*
262 * Get rid of annoying "../common/" prefix to filename.
263 */
264 newfile = strrchr(file, '/');
265 if (newfile != NULL) {
266 newfile = newfile + 1; /* Get rid of leading / */
267 } else {
268 newfile = file;
269 }
270
271 va_start(adx, fmt);
272 (void) vsnprintf(buf, sizeof (buf), fmt, adx);
273 va_end(adx);
274
275 /*
276 * To get this data, use the zfs-dprintf probe as so:
277 * dtrace -q -n 'zfs-dprintf \
278 * /stringof(arg0) == "dbuf.c"/ \
279 * {printf("%s: %s", stringof(arg1), stringof(arg3))}'
280 * arg0 = file name
281 * arg1 = function name
282 * arg2 = line number
283 * arg3 = message
284 */
285 DTRACE_PROBE4(zfs__dprintf,
286 char *, newfile, char *, func, int, line, char *, buf);
287 }
288
289 static void
290 history_str_free(char *buf)
291 {
292 kmem_free(buf, HIS_MAX_RECORD_LEN);
293 }
294
295 static char *
296 history_str_get(zfs_cmd_t *zc)
297 {
298 char *buf;
299
300 if (zc->zc_history == NULL)
301 return (NULL);
302
303 buf = kmem_alloc(HIS_MAX_RECORD_LEN, KM_SLEEP);
304 if (copyinstr((void *)(uintptr_t)zc->zc_history,
305 buf, HIS_MAX_RECORD_LEN, NULL) != 0) {
306 history_str_free(buf);
307 return (NULL);
308 }
309
310 buf[HIS_MAX_RECORD_LEN -1] = '\0';
311
312 return (buf);
313 }
314
315 /*
316 * Check to see if the named dataset is currently defined as bootable
317 */
318 static boolean_t
319 zfs_is_bootfs(const char *name)
320 {
321 objset_t *os;
322
323 if (dmu_objset_hold(name, FTAG, &os) == 0) {
324 boolean_t ret;
325 ret = (dmu_objset_id(os) == spa_bootfs(dmu_objset_spa(os)));
326 dmu_objset_rele(os, FTAG);
327 return (ret);
328 }
329 return (B_FALSE);
330 }
331
332 /*
333 * Return non-zero if the spa version is less than requested version.
334 */
335 static int
336 zfs_earlier_version(const char *name, int version)
337 {
338 spa_t *spa;
339
340 if (spa_open(name, &spa, FTAG) == 0) {
341 if (spa_version(spa) < version) {
342 spa_close(spa, FTAG);
343 return (1);
344 }
345 spa_close(spa, FTAG);
346 }
347 return (0);
348 }
349
350 /*
351 * Return TRUE if the ZPL version is less than requested version.
352 */
353 static boolean_t
354 zpl_earlier_version(const char *name, int version)
355 {
356 objset_t *os;
357 boolean_t rc = B_TRUE;
358
359 if (dmu_objset_hold(name, FTAG, &os) == 0) {
360 uint64_t zplversion;
361
362 if (dmu_objset_type(os) != DMU_OST_ZFS) {
363 dmu_objset_rele(os, FTAG);
364 return (B_TRUE);
365 }
366 /* XXX reading from non-owned objset */
367 if (zfs_get_zplprop(os, ZFS_PROP_VERSION, &zplversion) == 0)
368 rc = zplversion < version;
369 dmu_objset_rele(os, FTAG);
370 }
371 return (rc);
372 }
373
374 static void
375 zfs_log_history(zfs_cmd_t *zc)
376 {
377 spa_t *spa;
378 char *buf;
379
380 if ((buf = history_str_get(zc)) == NULL)
381 return;
382
383 if (spa_open(zc->zc_name, &spa, FTAG) == 0) {
384 if (spa_version(spa) >= SPA_VERSION_ZPOOL_HISTORY)
385 (void) spa_history_log(spa, buf);
386 spa_close(spa, FTAG);
387 }
388 history_str_free(buf);
389 }
390
391 /*
392 * Policy for top-level read operations (list pools). Requires no privileges,
393 * and can be used in the local zone, as there is no associated dataset.
394 */
395 /* ARGSUSED */
396 static int
397 zfs_secpolicy_none(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
398 {
399 return (0);
400 }
401
402 /*
403 * Policy for dataset read operations (list children, get statistics). Requires
404 * no privileges, but must be visible in the local zone.
405 */
406 /* ARGSUSED */
407 static int
408 zfs_secpolicy_read(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
409 {
410 if (INGLOBALZONE(curproc) ||
411 zone_dataset_visible(zc->zc_name, NULL))
412 return (0);
413
414 return (SET_ERROR(ENOENT));
415 }
416
417 static int
418 zfs_dozonecheck_impl(const char *dataset, uint64_t zoned, cred_t *cr)
419 {
420 int writable = 1;
421
422 /*
423 * The dataset must be visible by this zone -- check this first
424 * so they don't see EPERM on something they shouldn't know about.
425 */
426 if (!INGLOBALZONE(curproc) &&
427 !zone_dataset_visible(dataset, &writable))
428 return (SET_ERROR(ENOENT));
429
430 if (INGLOBALZONE(curproc)) {
431 /*
432 * If the fs is zoned, only root can access it from the
433 * global zone.
434 */
435 if (secpolicy_zfs(cr) && zoned)
436 return (SET_ERROR(EPERM));
437 } else {
438 /*
439 * If we are in a local zone, the 'zoned' property must be set.
440 */
441 if (!zoned)
442 return (SET_ERROR(EPERM));
443
444 /* must be writable by this zone */
445 if (!writable)
446 return (SET_ERROR(EPERM));
447 }
448 return (0);
449 }
450
451 static int
452 zfs_dozonecheck(const char *dataset, cred_t *cr)
453 {
454 uint64_t zoned;
455
456 if (dsl_prop_get_integer(dataset, "zoned", &zoned, NULL))
457 return (SET_ERROR(ENOENT));
458
459 return (zfs_dozonecheck_impl(dataset, zoned, cr));
460 }
461
462 static int
463 zfs_dozonecheck_ds(const char *dataset, dsl_dataset_t *ds, cred_t *cr)
464 {
465 uint64_t zoned;
466
467 if (dsl_prop_get_int_ds(ds, "zoned", &zoned))
468 return (SET_ERROR(ENOENT));
469
470 return (zfs_dozonecheck_impl(dataset, zoned, cr));
471 }
472
473 static int
474 zfs_secpolicy_write_perms_ds(const char *name, dsl_dataset_t *ds,
475 const char *perm, cred_t *cr)
476 {
477 int error;
478
479 error = zfs_dozonecheck_ds(name, ds, cr);
480 if (error == 0) {
481 error = secpolicy_zfs(cr);
482 if (error != 0)
483 error = dsl_deleg_access_impl(ds, perm, cr);
484 }
485 return (error);
486 }
487
488 static int
489 zfs_secpolicy_write_perms(const char *name, const char *perm, cred_t *cr)
490 {
491 int error;
492 dsl_dataset_t *ds;
493 dsl_pool_t *dp;
494
495 /*
496 * First do a quick check for root in the global zone, which
497 * is allowed to do all write_perms. This ensures that zfs_ioc_*
498 * will get to handle nonexistent datasets.
499 */
500 if (INGLOBALZONE(curproc) && secpolicy_zfs(cr) == 0)
501 return (0);
502
503 error = dsl_pool_hold(name, FTAG, &dp);
504 if (error != 0)
505 return (error);
506
507 error = dsl_dataset_hold(dp, name, FTAG, &ds);
508 if (error != 0) {
509 dsl_pool_rele(dp, FTAG);
510 return (error);
511 }
512
513 error = zfs_secpolicy_write_perms_ds(name, ds, perm, cr);
514
515 dsl_dataset_rele(ds, FTAG);
516 dsl_pool_rele(dp, FTAG);
517 return (error);
518 }
519
520 /*
521 * Policy for setting the security label property.
522 *
523 * Returns 0 for success, non-zero for access and other errors.
524 */
525 static int
526 zfs_set_slabel_policy(const char *name, char *strval, cred_t *cr)
527 {
528 char ds_hexsl[MAXNAMELEN];
529 bslabel_t ds_sl, new_sl;
530 boolean_t new_default = FALSE;
531 uint64_t zoned;
532 int needed_priv = -1;
533 int error;
534
535 /* First get the existing dataset label. */
536 error = dsl_prop_get(name, zfs_prop_to_name(ZFS_PROP_MLSLABEL),
537 1, sizeof (ds_hexsl), &ds_hexsl, NULL);
538 if (error != 0)
539 return (SET_ERROR(EPERM));
540
541 if (strcasecmp(strval, ZFS_MLSLABEL_DEFAULT) == 0)
542 new_default = TRUE;
543
544 /* The label must be translatable */
545 if (!new_default && (hexstr_to_label(strval, &new_sl) != 0))
546 return (SET_ERROR(EINVAL));
547
548 /*
549 * In a non-global zone, disallow attempts to set a label that
550 * doesn't match that of the zone; otherwise no other checks
551 * are needed.
552 */
553 if (!INGLOBALZONE(curproc)) {
554 if (new_default || !blequal(&new_sl, CR_SL(CRED())))
555 return (SET_ERROR(EPERM));
556 return (0);
557 }
558
559 /*
560 * For global-zone datasets (i.e., those whose zoned property is
561 * "off", verify that the specified new label is valid for the
562 * global zone.
563 */
564 if (dsl_prop_get_integer(name,
565 zfs_prop_to_name(ZFS_PROP_ZONED), &zoned, NULL))
566 return (SET_ERROR(EPERM));
567 if (!zoned) {
568 if (zfs_check_global_label(name, strval) != 0)
569 return (SET_ERROR(EPERM));
570 }
571
572 /*
573 * If the existing dataset label is nondefault, check if the
574 * dataset is mounted (label cannot be changed while mounted).
575 * Get the zfsvfs; if there isn't one, then the dataset isn't
576 * mounted (or isn't a dataset, doesn't exist, ...).
577 */
578 if (strcasecmp(ds_hexsl, ZFS_MLSLABEL_DEFAULT) != 0) {
579 objset_t *os;
580 static char *setsl_tag = "setsl_tag";
581
582 /*
583 * Try to own the dataset; abort if there is any error,
584 * (e.g., already mounted, in use, or other error).
585 */
586 error = dmu_objset_own(name, DMU_OST_ZFS, B_TRUE,
587 setsl_tag, &os);
588 if (error != 0)
589 return (SET_ERROR(EPERM));
590
591 dmu_objset_disown(os, setsl_tag);
592
593 if (new_default) {
594 needed_priv = PRIV_FILE_DOWNGRADE_SL;
595 goto out_check;
596 }
597
598 if (hexstr_to_label(strval, &new_sl) != 0)
599 return (SET_ERROR(EPERM));
600
601 if (blstrictdom(&ds_sl, &new_sl))
602 needed_priv = PRIV_FILE_DOWNGRADE_SL;
603 else if (blstrictdom(&new_sl, &ds_sl))
604 needed_priv = PRIV_FILE_UPGRADE_SL;
605 } else {
606 /* dataset currently has a default label */
607 if (!new_default)
608 needed_priv = PRIV_FILE_UPGRADE_SL;
609 }
610
611 out_check:
612 if (needed_priv != -1)
613 return (PRIV_POLICY(cr, needed_priv, B_FALSE, EPERM, NULL));
614 return (0);
615 }
616
617 static int
618 zfs_secpolicy_setprop(const char *dsname, zfs_prop_t prop, nvpair_t *propval,
619 cred_t *cr)
620 {
621 char *strval;
622
623 /*
624 * Check permissions for special properties.
625 */
626 switch (prop) {
627 case ZFS_PROP_ZONED:
628 /*
629 * Disallow setting of 'zoned' from within a local zone.
630 */
631 if (!INGLOBALZONE(curproc))
632 return (SET_ERROR(EPERM));
633 break;
634
635 case ZFS_PROP_QUOTA:
636 case ZFS_PROP_FILESYSTEM_LIMIT:
637 case ZFS_PROP_SNAPSHOT_LIMIT:
638 if (!INGLOBALZONE(curproc)) {
639 uint64_t zoned;
640 char setpoint[ZFS_MAX_DATASET_NAME_LEN];
641 /*
642 * Unprivileged users are allowed to modify the
643 * limit on things *under* (ie. contained by)
644 * the thing they own.
645 */
646 if (dsl_prop_get_integer(dsname, "zoned", &zoned,
647 setpoint))
648 return (SET_ERROR(EPERM));
649 if (!zoned || strlen(dsname) <= strlen(setpoint))
650 return (SET_ERROR(EPERM));
651 }
652 break;
653
654 case ZFS_PROP_MLSLABEL:
655 if (!is_system_labeled())
656 return (SET_ERROR(EPERM));
657
658 if (nvpair_value_string(propval, &strval) == 0) {
659 int err;
660
661 err = zfs_set_slabel_policy(dsname, strval, CRED());
662 if (err != 0)
663 return (err);
664 }
665 break;
666 }
667
668 return (zfs_secpolicy_write_perms(dsname, zfs_prop_to_name(prop), cr));
669 }
670
671 /* ARGSUSED */
672 static int
673 zfs_secpolicy_set_fsacl(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
674 {
675 int error;
676
677 error = zfs_dozonecheck(zc->zc_name, cr);
678 if (error != 0)
679 return (error);
680
681 /*
682 * permission to set permissions will be evaluated later in
683 * dsl_deleg_can_allow()
684 */
685 return (0);
686 }
687
688 /* ARGSUSED */
689 static int
690 zfs_secpolicy_rollback(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
691 {
692 return (zfs_secpolicy_write_perms(zc->zc_name,
693 ZFS_DELEG_PERM_ROLLBACK, cr));
694 }
695
696 /* ARGSUSED */
697 static int
698 zfs_secpolicy_send(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
699 {
700 dsl_pool_t *dp;
701 dsl_dataset_t *ds;
702 char *cp;
703 int error;
704
705 /*
706 * Generate the current snapshot name from the given objsetid, then
707 * use that name for the secpolicy/zone checks.
708 */
709 cp = strchr(zc->zc_name, '@');
710 if (cp == NULL)
711 return (SET_ERROR(EINVAL));
712 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
713 if (error != 0)
714 return (error);
715
716 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &ds);
717 if (error != 0) {
718 dsl_pool_rele(dp, FTAG);
719 return (error);
720 }
721
722 dsl_dataset_name(ds, zc->zc_name);
723
724 error = zfs_secpolicy_write_perms_ds(zc->zc_name, ds,
725 ZFS_DELEG_PERM_SEND, cr);
726 dsl_dataset_rele(ds, FTAG);
727 dsl_pool_rele(dp, FTAG);
728
729 return (error);
730 }
731
732 /* ARGSUSED */
733 static int
734 zfs_secpolicy_send_new(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
735 {
736 return (zfs_secpolicy_write_perms(zc->zc_name,
737 ZFS_DELEG_PERM_SEND, cr));
738 }
739
740 /* ARGSUSED */
741 static int
742 zfs_secpolicy_deleg_share(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
743 {
744 vnode_t *vp;
745 int error;
746
747 if ((error = lookupname(zc->zc_value, UIO_SYSSPACE,
748 NO_FOLLOW, NULL, &vp)) != 0)
749 return (error);
750
751 /* Now make sure mntpnt and dataset are ZFS */
752
753 if (vp->v_vfsp->vfs_fstype != zfsfstype ||
754 (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource),
755 zc->zc_name) != 0)) {
756 VN_RELE(vp);
757 return (SET_ERROR(EPERM));
758 }
759
760 VN_RELE(vp);
761 return (dsl_deleg_access(zc->zc_name,
762 ZFS_DELEG_PERM_SHARE, cr));
763 }
764
765 int
766 zfs_secpolicy_share(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
767 {
768 if (!INGLOBALZONE(curproc))
769 return (SET_ERROR(EPERM));
770
771 if (secpolicy_nfs(cr) == 0) {
772 return (0);
773 } else {
774 return (zfs_secpolicy_deleg_share(zc, innvl, cr));
775 }
776 }
777
778 int
779 zfs_secpolicy_smb_acl(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
780 {
781 if (!INGLOBALZONE(curproc))
782 return (SET_ERROR(EPERM));
783
784 if (secpolicy_smb(cr) == 0) {
785 return (0);
786 } else {
787 return (zfs_secpolicy_deleg_share(zc, innvl, cr));
788 }
789 }
790
791 static int
792 zfs_get_parent(const char *datasetname, char *parent, int parentsize)
793 {
794 char *cp;
795
796 /*
797 * Remove the @bla or /bla from the end of the name to get the parent.
798 */
799 (void) strncpy(parent, datasetname, parentsize);
800 cp = strrchr(parent, '@');
801 if (cp != NULL) {
802 cp[0] = '\0';
803 } else {
804 cp = strrchr(parent, '/');
805 if (cp == NULL)
806 return (SET_ERROR(ENOENT));
807 cp[0] = '\0';
808 }
809
810 return (0);
811 }
812
813 int
814 zfs_secpolicy_destroy_perms(const char *name, cred_t *cr)
815 {
816 int error;
817
818 if ((error = zfs_secpolicy_write_perms(name,
819 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
820 return (error);
821
822 return (zfs_secpolicy_write_perms(name, ZFS_DELEG_PERM_DESTROY, cr));
823 }
824
825 /* ARGSUSED */
826 static int
827 zfs_secpolicy_destroy(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
828 {
829 return (zfs_secpolicy_destroy_perms(zc->zc_name, cr));
830 }
831
832 /*
833 * Destroying snapshots with delegated permissions requires
834 * descendant mount and destroy permissions.
835 */
836 /* ARGSUSED */
837 static int
838 zfs_secpolicy_destroy_snaps(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
839 {
840 nvlist_t *snaps;
841 nvpair_t *pair, *nextpair;
842 int error = 0;
843
844 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
845 return (SET_ERROR(EINVAL));
846 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
847 pair = nextpair) {
848 nextpair = nvlist_next_nvpair(snaps, pair);
849 error = zfs_secpolicy_destroy_perms(nvpair_name(pair), cr);
850 if (error == ENOENT) {
851 /*
852 * Ignore any snapshots that don't exist (we consider
853 * them "already destroyed"). Remove the name from the
854 * nvl here in case the snapshot is created between
855 * now and when we try to destroy it (in which case
856 * we don't want to destroy it since we haven't
857 * checked for permission).
858 */
859 fnvlist_remove_nvpair(snaps, pair);
860 error = 0;
861 }
862 if (error != 0)
863 break;
864 }
865
866 return (error);
867 }
868
869 int
870 zfs_secpolicy_rename_perms(const char *from, const char *to, cred_t *cr)
871 {
872 char parentname[ZFS_MAX_DATASET_NAME_LEN];
873 int error;
874
875 if ((error = zfs_secpolicy_write_perms(from,
876 ZFS_DELEG_PERM_RENAME, cr)) != 0)
877 return (error);
878
879 if ((error = zfs_secpolicy_write_perms(from,
880 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
881 return (error);
882
883 if ((error = zfs_get_parent(to, parentname,
884 sizeof (parentname))) != 0)
885 return (error);
886
887 if ((error = zfs_secpolicy_write_perms(parentname,
888 ZFS_DELEG_PERM_CREATE, cr)) != 0)
889 return (error);
890
891 if ((error = zfs_secpolicy_write_perms(parentname,
892 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
893 return (error);
894
895 return (error);
896 }
897
898 /* ARGSUSED */
899 static int
900 zfs_secpolicy_rename(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
901 {
902 return (zfs_secpolicy_rename_perms(zc->zc_name, zc->zc_value, cr));
903 }
904
905 /* ARGSUSED */
906 static int
907 zfs_secpolicy_promote(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
908 {
909 dsl_pool_t *dp;
910 dsl_dataset_t *clone;
911 int error;
912
913 error = zfs_secpolicy_write_perms(zc->zc_name,
914 ZFS_DELEG_PERM_PROMOTE, cr);
915 if (error != 0)
916 return (error);
917
918 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
919 if (error != 0)
920 return (error);
921
922 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &clone);
923
924 if (error == 0) {
925 char parentname[ZFS_MAX_DATASET_NAME_LEN];
926 dsl_dataset_t *origin = NULL;
927 dsl_dir_t *dd;
928 dd = clone->ds_dir;
929
930 error = dsl_dataset_hold_obj(dd->dd_pool,
931 dsl_dir_phys(dd)->dd_origin_obj, FTAG, &origin);
932 if (error != 0) {
933 dsl_dataset_rele(clone, FTAG);
934 dsl_pool_rele(dp, FTAG);
935 return (error);
936 }
937
938 error = zfs_secpolicy_write_perms_ds(zc->zc_name, clone,
939 ZFS_DELEG_PERM_MOUNT, cr);
940
941 dsl_dataset_name(origin, parentname);
942 if (error == 0) {
943 error = zfs_secpolicy_write_perms_ds(parentname, origin,
944 ZFS_DELEG_PERM_PROMOTE, cr);
945 }
946 dsl_dataset_rele(clone, FTAG);
947 dsl_dataset_rele(origin, FTAG);
948 }
949 dsl_pool_rele(dp, FTAG);
950 return (error);
951 }
952
953 /* ARGSUSED */
954 static int
955 zfs_secpolicy_recv(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
956 {
957 int error;
958
959 if ((error = zfs_secpolicy_write_perms(zc->zc_name,
960 ZFS_DELEG_PERM_RECEIVE, cr)) != 0)
961 return (error);
962
963 if ((error = zfs_secpolicy_write_perms(zc->zc_name,
964 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
965 return (error);
966
967 return (zfs_secpolicy_write_perms(zc->zc_name,
968 ZFS_DELEG_PERM_CREATE, cr));
969 }
970
971 int
972 zfs_secpolicy_snapshot_perms(const char *name, cred_t *cr)
973 {
974 return (zfs_secpolicy_write_perms(name,
975 ZFS_DELEG_PERM_SNAPSHOT, cr));
976 }
977
978 /*
979 * Check for permission to create each snapshot in the nvlist.
980 */
981 /* ARGSUSED */
982 static int
983 zfs_secpolicy_snapshot(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
984 {
985 nvlist_t *snaps;
986 int error = 0;
987 nvpair_t *pair;
988
989 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
990 return (SET_ERROR(EINVAL));
991 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
992 pair = nvlist_next_nvpair(snaps, pair)) {
993 char *name = nvpair_name(pair);
994 char *atp = strchr(name, '@');
995
996 if (atp == NULL) {
997 error = SET_ERROR(EINVAL);
998 break;
999 }
1000 *atp = '\0';
1001 error = zfs_secpolicy_snapshot_perms(name, cr);
1002 *atp = '@';
1003 if (error != 0)
1004 break;
1005 }
1006 return (error);
1007 }
1008
1009 /*
1010 * Check for permission to create each snapshot in the nvlist.
1011 */
1012 /* ARGSUSED */
1013 static int
1014 zfs_secpolicy_bookmark(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1015 {
1016 int error = 0;
1017
1018 for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL);
1019 pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) {
1020 char *name = nvpair_name(pair);
1021 char *hashp = strchr(name, '#');
1022
1023 if (hashp == NULL) {
1024 error = SET_ERROR(EINVAL);
1025 break;
1026 }
1027 *hashp = '\0';
1028 error = zfs_secpolicy_write_perms(name,
1029 ZFS_DELEG_PERM_BOOKMARK, cr);
1030 *hashp = '#';
1031 if (error != 0)
1032 break;
1033 }
1034 return (error);
1035 }
1036
1037 /* ARGSUSED */
1038 static int
1039 zfs_secpolicy_destroy_bookmarks(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1040 {
1041 nvpair_t *pair, *nextpair;
1042 int error = 0;
1043
1044 for (pair = nvlist_next_nvpair(innvl, NULL); pair != NULL;
1045 pair = nextpair) {
1046 char *name = nvpair_name(pair);
1047 char *hashp = strchr(name, '#');
1048 nextpair = nvlist_next_nvpair(innvl, pair);
1049
1050 if (hashp == NULL) {
1051 error = SET_ERROR(EINVAL);
1052 break;
1053 }
1054
1055 *hashp = '\0';
1056 error = zfs_secpolicy_write_perms(name,
1057 ZFS_DELEG_PERM_DESTROY, cr);
1058 *hashp = '#';
1059 if (error == ENOENT) {
1060 /*
1061 * Ignore any filesystems that don't exist (we consider
1062 * their bookmarks "already destroyed"). Remove
1063 * the name from the nvl here in case the filesystem
1064 * is created between now and when we try to destroy
1065 * the bookmark (in which case we don't want to
1066 * destroy it since we haven't checked for permission).
1067 */
1068 fnvlist_remove_nvpair(innvl, pair);
1069 error = 0;
1070 }
1071 if (error != 0)
1072 break;
1073 }
1074
1075 return (error);
1076 }
1077
1078 /* ARGSUSED */
1079 static int
1080 zfs_secpolicy_log_history(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1081 {
1082 /*
1083 * Even root must have a proper TSD so that we know what pool
1084 * to log to.
1085 */
1086 if (tsd_get(zfs_allow_log_key) == NULL)
1087 return (SET_ERROR(EPERM));
1088 return (0);
1089 }
1090
1091 static int
1092 zfs_secpolicy_create_clone(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1093 {
1094 char parentname[ZFS_MAX_DATASET_NAME_LEN];
1095 int error;
1096 char *origin;
1097
1098 if ((error = zfs_get_parent(zc->zc_name, parentname,
1099 sizeof (parentname))) != 0)
1100 return (error);
1101
1102 if (nvlist_lookup_string(innvl, "origin", &origin) == 0 &&
1103 (error = zfs_secpolicy_write_perms(origin,
1104 ZFS_DELEG_PERM_CLONE, cr)) != 0)
1105 return (error);
1106
1107 if ((error = zfs_secpolicy_write_perms(parentname,
1108 ZFS_DELEG_PERM_CREATE, cr)) != 0)
1109 return (error);
1110
1111 return (zfs_secpolicy_write_perms(parentname,
1112 ZFS_DELEG_PERM_MOUNT, cr));
1113 }
1114
1115 /*
1116 * Policy for pool operations - create/destroy pools, add vdevs, etc. Requires
1117 * SYS_CONFIG privilege, which is not available in a local zone.
1118 */
1119 /* ARGSUSED */
1120 static int
1121 zfs_secpolicy_config(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1122 {
1123 if (secpolicy_sys_config(cr, B_FALSE) != 0)
1124 return (SET_ERROR(EPERM));
1125
1126 return (0);
1127 }
1128
1129 /*
1130 * Policy for object to name lookups.
1131 */
1132 /* ARGSUSED */
1133 static int
1134 zfs_secpolicy_diff(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1135 {
1136 int error;
1137
1138 if ((error = secpolicy_sys_config(cr, B_FALSE)) == 0)
1139 return (0);
1140
1141 error = zfs_secpolicy_write_perms(zc->zc_name, ZFS_DELEG_PERM_DIFF, cr);
1142 return (error);
1143 }
1144
1145 /*
1146 * Policy for fault injection. Requires all privileges.
1147 */
1148 /* ARGSUSED */
1149 static int
1150 zfs_secpolicy_inject(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1151 {
1152 return (secpolicy_zinject(cr));
1153 }
1154
1155 /* ARGSUSED */
1156 static int
1157 zfs_secpolicy_inherit_prop(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1158 {
1159 zfs_prop_t prop = zfs_name_to_prop(zc->zc_value);
1160
1161 if (prop == ZPROP_INVAL) {
1162 if (!zfs_prop_user(zc->zc_value))
1163 return (SET_ERROR(EINVAL));
1164 return (zfs_secpolicy_write_perms(zc->zc_name,
1165 ZFS_DELEG_PERM_USERPROP, cr));
1166 } else {
1167 return (zfs_secpolicy_setprop(zc->zc_name, prop,
1168 NULL, cr));
1169 }
1170 }
1171
1172 static int
1173 zfs_secpolicy_userspace_one(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1174 {
1175 int err = zfs_secpolicy_read(zc, innvl, cr);
1176 if (err)
1177 return (err);
1178
1179 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
1180 return (SET_ERROR(EINVAL));
1181
1182 if (zc->zc_value[0] == 0) {
1183 /*
1184 * They are asking about a posix uid/gid. If it's
1185 * themself, allow it.
1186 */
1187 if (zc->zc_objset_type == ZFS_PROP_USERUSED ||
1188 zc->zc_objset_type == ZFS_PROP_USERQUOTA) {
1189 if (zc->zc_guid == crgetuid(cr))
1190 return (0);
1191 } else {
1192 if (groupmember(zc->zc_guid, cr))
1193 return (0);
1194 }
1195 }
1196
1197 return (zfs_secpolicy_write_perms(zc->zc_name,
1198 userquota_perms[zc->zc_objset_type], cr));
1199 }
1200
1201 static int
1202 zfs_secpolicy_userspace_many(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1203 {
1204 int err = zfs_secpolicy_read(zc, innvl, cr);
1205 if (err)
1206 return (err);
1207
1208 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
1209 return (SET_ERROR(EINVAL));
1210
1211 return (zfs_secpolicy_write_perms(zc->zc_name,
1212 userquota_perms[zc->zc_objset_type], cr));
1213 }
1214
1215 /* ARGSUSED */
1216 static int
1217 zfs_secpolicy_userspace_upgrade(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1218 {
1219 return (zfs_secpolicy_setprop(zc->zc_name, ZFS_PROP_VERSION,
1220 NULL, cr));
1221 }
1222
1223 /* ARGSUSED */
1224 static int
1225 zfs_secpolicy_hold(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1226 {
1227 nvpair_t *pair;
1228 nvlist_t *holds;
1229 int error;
1230
1231 error = nvlist_lookup_nvlist(innvl, "holds", &holds);
1232 if (error != 0)
1233 return (SET_ERROR(EINVAL));
1234
1235 for (pair = nvlist_next_nvpair(holds, NULL); pair != NULL;
1236 pair = nvlist_next_nvpair(holds, pair)) {
1237 char fsname[ZFS_MAX_DATASET_NAME_LEN];
1238 error = dmu_fsname(nvpair_name(pair), fsname);
1239 if (error != 0)
1240 return (error);
1241 error = zfs_secpolicy_write_perms(fsname,
1242 ZFS_DELEG_PERM_HOLD, cr);
1243 if (error != 0)
1244 return (error);
1245 }
1246 return (0);
1247 }
1248
1249 /* ARGSUSED */
1250 static int
1251 zfs_secpolicy_release(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1252 {
1253 nvpair_t *pair;
1254 int error;
1255
1256 for (pair = nvlist_next_nvpair(innvl, NULL); pair != NULL;
1257 pair = nvlist_next_nvpair(innvl, pair)) {
1258 char fsname[ZFS_MAX_DATASET_NAME_LEN];
1259 error = dmu_fsname(nvpair_name(pair), fsname);
1260 if (error != 0)
1261 return (error);
1262 error = zfs_secpolicy_write_perms(fsname,
1263 ZFS_DELEG_PERM_RELEASE, cr);
1264 if (error != 0)
1265 return (error);
1266 }
1267 return (0);
1268 }
1269
1270 /*
1271 * Policy for allowing temporary snapshots to be taken or released
1272 */
1273 static int
1274 zfs_secpolicy_tmp_snapshot(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1275 {
1276 /*
1277 * A temporary snapshot is the same as a snapshot,
1278 * hold, destroy and release all rolled into one.
1279 * Delegated diff alone is sufficient that we allow this.
1280 */
1281 int error;
1282
1283 if ((error = zfs_secpolicy_write_perms(zc->zc_name,
1284 ZFS_DELEG_PERM_DIFF, cr)) == 0)
1285 return (0);
1286
1287 error = zfs_secpolicy_snapshot_perms(zc->zc_name, cr);
1288 if (error == 0)
1289 error = zfs_secpolicy_hold(zc, innvl, cr);
1290 if (error == 0)
1291 error = zfs_secpolicy_release(zc, innvl, cr);
1292 if (error == 0)
1293 error = zfs_secpolicy_destroy(zc, innvl, cr);
1294 return (error);
1295 }
1296
1297 /*
1298 * Returns the nvlist as specified by the user in the zfs_cmd_t.
1299 */
1300 static int
1301 get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp)
1302 {
1303 char *packed;
1304 int error;
1305 nvlist_t *list = NULL;
1306
1307 /*
1308 * Read in and unpack the user-supplied nvlist.
1309 */
1310 if (size == 0)
1311 return (SET_ERROR(EINVAL));
1312
1313 packed = kmem_alloc(size, KM_SLEEP);
1314
1315 if ((error = ddi_copyin((void *)(uintptr_t)nvl, packed, size,
1316 iflag)) != 0) {
1317 kmem_free(packed, size);
1318 return (SET_ERROR(EFAULT));
1319 }
1320
1321 if ((error = nvlist_unpack(packed, size, &list, 0)) != 0) {
1322 kmem_free(packed, size);
1323 return (error);
1324 }
1325
1326 kmem_free(packed, size);
1327
1328 *nvp = list;
1329 return (0);
1330 }
1331
1332 /*
1333 * Reduce the size of this nvlist until it can be serialized in 'max' bytes.
1334 * Entries will be removed from the end of the nvlist, and one int32 entry
1335 * named "N_MORE_ERRORS" will be added indicating how many entries were
1336 * removed.
1337 */
1338 static int
1339 nvlist_smush(nvlist_t *errors, size_t max)
1340 {
1341 size_t size;
1342
1343 size = fnvlist_size(errors);
1344
1345 if (size > max) {
1346 nvpair_t *more_errors;
1347 int n = 0;
1348
1349 if (max < 1024)
1350 return (SET_ERROR(ENOMEM));
1351
1352 fnvlist_add_int32(errors, ZPROP_N_MORE_ERRORS, 0);
1353 more_errors = nvlist_prev_nvpair(errors, NULL);
1354
1355 do {
1356 nvpair_t *pair = nvlist_prev_nvpair(errors,
1357 more_errors);
1358 fnvlist_remove_nvpair(errors, pair);
1359 n++;
1360 size = fnvlist_size(errors);
1361 } while (size > max);
1362
1363 fnvlist_remove_nvpair(errors, more_errors);
1364 fnvlist_add_int32(errors, ZPROP_N_MORE_ERRORS, n);
1365 ASSERT3U(fnvlist_size(errors), <=, max);
1366 }
1367
1368 return (0);
1369 }
1370
1371 static int
1372 put_nvlist(zfs_cmd_t *zc, nvlist_t *nvl)
1373 {
1374 char *packed = NULL;
1375 int error = 0;
1376 size_t size;
1377
1378 size = fnvlist_size(nvl);
1379
1380 if (size > zc->zc_nvlist_dst_size) {
1381 error = SET_ERROR(ENOMEM);
1382 } else {
1383 packed = fnvlist_pack(nvl, &size);
1384 if (ddi_copyout(packed, (void *)(uintptr_t)zc->zc_nvlist_dst,
1385 size, zc->zc_iflags) != 0)
1386 error = SET_ERROR(EFAULT);
1387 fnvlist_pack_free(packed, size);
1388 }
1389
1390 zc->zc_nvlist_dst_size = size;
1391 zc->zc_nvlist_dst_filled = B_TRUE;
1392 return (error);
1393 }
1394
1395 static int
1396 getzfsvfs(const char *dsname, zfsvfs_t **zfvp)
1397 {
1398 objset_t *os;
1399 int error;
1400
1401 error = dmu_objset_hold(dsname, FTAG, &os);
1402 if (error != 0)
1403 return (error);
1404 if (dmu_objset_type(os) != DMU_OST_ZFS) {
1405 dmu_objset_rele(os, FTAG);
1406 return (SET_ERROR(EINVAL));
1407 }
1408
1409 mutex_enter(&os->os_user_ptr_lock);
1410 *zfvp = dmu_objset_get_user(os);
1411 if (*zfvp) {
1412 VFS_HOLD((*zfvp)->z_vfs);
1413 } else {
1414 error = SET_ERROR(ESRCH);
1415 }
1416 mutex_exit(&os->os_user_ptr_lock);
1417 dmu_objset_rele(os, FTAG);
1418 return (error);
1419 }
1420
1421 /*
1422 * Find a zfsvfs_t for a mounted filesystem, or create our own, in which
1423 * case its z_vfs will be NULL, and it will be opened as the owner.
1424 * If 'writer' is set, the z_teardown_lock will be held for RW_WRITER,
1425 * which prevents all vnode ops from running.
1426 */
1427 static int
1428 zfsvfs_hold(const char *name, void *tag, zfsvfs_t **zfvp, boolean_t writer)
1429 {
1430 int error = 0;
1431
1432 if (getzfsvfs(name, zfvp) != 0)
1433 error = zfsvfs_create(name, zfvp);
1434 if (error == 0) {
1435 rrm_enter(&(*zfvp)->z_teardown_lock, (writer) ? RW_WRITER :
1436 RW_READER, tag);
1437 if ((*zfvp)->z_unmounted) {
1438 /*
1439 * XXX we could probably try again, since the unmounting
1440 * thread should be just about to disassociate the
1441 * objset from the zfsvfs.
1442 */
1443 rrm_exit(&(*zfvp)->z_teardown_lock, tag);
1444 return (SET_ERROR(EBUSY));
1445 }
1446 }
1447 return (error);
1448 }
1449
1450 static void
1451 zfsvfs_rele(zfsvfs_t *zfsvfs, void *tag)
1452 {
1453 rrm_exit(&zfsvfs->z_teardown_lock, tag);
1454
1455 if (zfsvfs->z_vfs) {
1456 VFS_RELE(zfsvfs->z_vfs);
1457 } else {
1458 dmu_objset_disown(zfsvfs->z_os, zfsvfs);
1459 zfsvfs_free(zfsvfs);
1460 }
1461 }
1462
1463 static int
1464 zfs_ioc_pool_create(zfs_cmd_t *zc)
1465 {
1466 int error;
1467 nvlist_t *config, *props = NULL;
1468 nvlist_t *rootprops = NULL;
1469 nvlist_t *zplprops = NULL;
1470
1471 if (error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1472 zc->zc_iflags, &config))
1473 return (error);
1474
1475 if (zc->zc_nvlist_src_size != 0 && (error =
1476 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1477 zc->zc_iflags, &props))) {
1478 nvlist_free(config);
1479 return (error);
1480 }
1481
1482 if (props) {
1483 nvlist_t *nvl = NULL;
1484 uint64_t version = SPA_VERSION;
1485
1486 (void) nvlist_lookup_uint64(props,
1487 zpool_prop_to_name(ZPOOL_PROP_VERSION), &version);
1488 if (!SPA_VERSION_IS_SUPPORTED(version)) {
1489 error = SET_ERROR(EINVAL);
1490 goto pool_props_bad;
1491 }
1492 (void) nvlist_lookup_nvlist(props, ZPOOL_ROOTFS_PROPS, &nvl);
1493 if (nvl) {
1494 error = nvlist_dup(nvl, &rootprops, KM_SLEEP);
1495 if (error != 0) {
1496 nvlist_free(config);
1497 nvlist_free(props);
1498 return (error);
1499 }
1500 (void) nvlist_remove_all(props, ZPOOL_ROOTFS_PROPS);
1501 }
1502 VERIFY(nvlist_alloc(&zplprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
1503 error = zfs_fill_zplprops_root(version, rootprops,
1504 zplprops, NULL);
1505 if (error != 0)
1506 goto pool_props_bad;
1507 }
1508
1509 error = spa_create(zc->zc_name, config, props, zplprops);
1510
1511 /*
1512 * Set the remaining root properties
1513 */
1514 if (!error && (error = zfs_set_prop_nvlist(zc->zc_name,
1515 ZPROP_SRC_LOCAL, rootprops, NULL)) != 0)
1516 (void) spa_destroy(zc->zc_name);
1517
1518 pool_props_bad:
1519 nvlist_free(rootprops);
1520 nvlist_free(zplprops);
1521 nvlist_free(config);
1522 nvlist_free(props);
1523
1524 return (error);
1525 }
1526
1527 static int
1528 zfs_ioc_pool_destroy(zfs_cmd_t *zc)
1529 {
1530 int error;
1531 zfs_log_history(zc);
1532 error = spa_destroy(zc->zc_name);
1533 if (error == 0)
1534 zvol_remove_minors(zc->zc_name);
1535 return (error);
1536 }
1537
1538 static int
1539 zfs_ioc_pool_import(zfs_cmd_t *zc)
1540 {
1541 nvlist_t *config, *props = NULL;
1542 uint64_t guid;
1543 int error;
1544
1545 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1546 zc->zc_iflags, &config)) != 0)
1547 return (error);
1548
1549 if (zc->zc_nvlist_src_size != 0 && (error =
1550 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1551 zc->zc_iflags, &props))) {
1552 nvlist_free(config);
1553 return (error);
1554 }
1555
1556 if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, &guid) != 0 ||
1557 guid != zc->zc_guid)
1558 error = SET_ERROR(EINVAL);
1559 else
1560 error = spa_import(zc->zc_name, config, props, zc->zc_cookie);
1561
1562 if (zc->zc_nvlist_dst != 0) {
1563 int err;
1564
1565 if ((err = put_nvlist(zc, config)) != 0)
1566 error = err;
1567 }
1568
1569 nvlist_free(config);
1570
1571 nvlist_free(props);
1572
1573 return (error);
1574 }
1575
1576 static int
1577 zfs_ioc_pool_export(zfs_cmd_t *zc)
1578 {
1579 int error;
1580 boolean_t force = (boolean_t)zc->zc_cookie;
1581 boolean_t hardforce = (boolean_t)zc->zc_guid;
1582
1583 zfs_log_history(zc);
1584 error = spa_export(zc->zc_name, NULL, force, hardforce);
1585 if (error == 0)
1586 zvol_remove_minors(zc->zc_name);
1587 return (error);
1588 }
1589
1590 static int
1591 zfs_ioc_pool_configs(zfs_cmd_t *zc)
1592 {
1593 nvlist_t *configs;
1594 int error;
1595
1596 if ((configs = spa_all_configs(&zc->zc_cookie)) == NULL)
1597 return (SET_ERROR(EEXIST));
1598
1599 error = put_nvlist(zc, configs);
1600
1601 nvlist_free(configs);
1602
1603 return (error);
1604 }
1605
1606 /*
1607 * inputs:
1608 * zc_name name of the pool
1609 *
1610 * outputs:
1611 * zc_cookie real errno
1612 * zc_nvlist_dst config nvlist
1613 * zc_nvlist_dst_size size of config nvlist
1614 */
1615 static int
1616 zfs_ioc_pool_stats(zfs_cmd_t *zc)
1617 {
1618 nvlist_t *config;
1619 int error;
1620 int ret = 0;
1621
1622 error = spa_get_stats(zc->zc_name, &config, zc->zc_value,
1623 sizeof (zc->zc_value));
1624
1625 if (config != NULL) {
1626 ret = put_nvlist(zc, config);
1627 nvlist_free(config);
1628
1629 /*
1630 * The config may be present even if 'error' is non-zero.
1631 * In this case we return success, and preserve the real errno
1632 * in 'zc_cookie'.
1633 */
1634 zc->zc_cookie = error;
1635 } else {
1636 ret = error;
1637 }
1638
1639 return (ret);
1640 }
1641
1642 /*
1643 * Try to import the given pool, returning pool stats as appropriate so that
1644 * user land knows which devices are available and overall pool health.
1645 */
1646 static int
1647 zfs_ioc_pool_tryimport(zfs_cmd_t *zc)
1648 {
1649 nvlist_t *tryconfig, *config;
1650 int error;
1651
1652 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1653 zc->zc_iflags, &tryconfig)) != 0)
1654 return (error);
1655
1656 config = spa_tryimport(tryconfig);
1657
1658 nvlist_free(tryconfig);
1659
1660 if (config == NULL)
1661 return (SET_ERROR(EINVAL));
1662
1663 error = put_nvlist(zc, config);
1664 nvlist_free(config);
1665
1666 return (error);
1667 }
1668
1669 /*
1670 * inputs:
1671 * zc_name name of the pool
1672 * zc_cookie scan func (pool_scan_func_t)
1673 */
1674 static int
1675 zfs_ioc_pool_scan(zfs_cmd_t *zc)
1676 {
1677 spa_t *spa;
1678 int error;
1679
1680 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1681 return (error);
1682
1683 if (zc->zc_cookie == POOL_SCAN_NONE)
1684 error = spa_scan_stop(spa);
1685 else
1686 error = spa_scan(spa, zc->zc_cookie);
1687
1688 spa_close(spa, FTAG);
1689
1690 return (error);
1691 }
1692
1693 static int
1694 zfs_ioc_pool_freeze(zfs_cmd_t *zc)
1695 {
1696 spa_t *spa;
1697 int error;
1698
1699 error = spa_open(zc->zc_name, &spa, FTAG);
1700 if (error == 0) {
1701 spa_freeze(spa);
1702 spa_close(spa, FTAG);
1703 }
1704 return (error);
1705 }
1706
1707 static int
1708 zfs_ioc_pool_upgrade(zfs_cmd_t *zc)
1709 {
1710 spa_t *spa;
1711 int error;
1712
1713 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1714 return (error);
1715
1716 if (zc->zc_cookie < spa_version(spa) ||
1717 !SPA_VERSION_IS_SUPPORTED(zc->zc_cookie)) {
1718 spa_close(spa, FTAG);
1719 return (SET_ERROR(EINVAL));
1720 }
1721
1722 spa_upgrade(spa, zc->zc_cookie);
1723 spa_close(spa, FTAG);
1724
1725 return (error);
1726 }
1727
1728 static int
1729 zfs_ioc_pool_get_history(zfs_cmd_t *zc)
1730 {
1731 spa_t *spa;
1732 char *hist_buf;
1733 uint64_t size;
1734 int error;
1735
1736 if ((size = zc->zc_history_len) == 0)
1737 return (SET_ERROR(EINVAL));
1738
1739 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1740 return (error);
1741
1742 if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) {
1743 spa_close(spa, FTAG);
1744 return (SET_ERROR(ENOTSUP));
1745 }
1746
1747 hist_buf = kmem_alloc(size, KM_SLEEP);
1748 if ((error = spa_history_get(spa, &zc->zc_history_offset,
1749 &zc->zc_history_len, hist_buf)) == 0) {
1750 error = ddi_copyout(hist_buf,
1751 (void *)(uintptr_t)zc->zc_history,
1752 zc->zc_history_len, zc->zc_iflags);
1753 }
1754
1755 spa_close(spa, FTAG);
1756 kmem_free(hist_buf, size);
1757 return (error);
1758 }
1759
1760 static int
1761 zfs_ioc_pool_reguid(zfs_cmd_t *zc)
1762 {
1763 spa_t *spa;
1764 int error;
1765
1766 error = spa_open(zc->zc_name, &spa, FTAG);
1767 if (error == 0) {
1768 error = spa_change_guid(spa);
1769 spa_close(spa, FTAG);
1770 }
1771 return (error);
1772 }
1773
1774 static int
1775 zfs_ioc_dsobj_to_dsname(zfs_cmd_t *zc)
1776 {
1777 return (dsl_dsobj_to_dsname(zc->zc_name, zc->zc_obj, zc->zc_value));
1778 }
1779
1780 /*
1781 * inputs:
1782 * zc_name name of filesystem
1783 * zc_obj object to find
1784 *
1785 * outputs:
1786 * zc_value name of object
1787 */
1788 static int
1789 zfs_ioc_obj_to_path(zfs_cmd_t *zc)
1790 {
1791 objset_t *os;
1792 int error;
1793
1794 /* XXX reading from objset not owned */
1795 if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os)) != 0)
1796 return (error);
1797 if (dmu_objset_type(os) != DMU_OST_ZFS) {
1798 dmu_objset_rele(os, FTAG);
1799 return (SET_ERROR(EINVAL));
1800 }
1801 error = zfs_obj_to_path(os, zc->zc_obj, zc->zc_value,
1802 sizeof (zc->zc_value));
1803 dmu_objset_rele(os, FTAG);
1804
1805 return (error);
1806 }
1807
1808 /*
1809 * inputs:
1810 * zc_name name of filesystem
1811 * zc_obj object to find
1812 *
1813 * outputs:
1814 * zc_stat stats on object
1815 * zc_value path to object
1816 */
1817 static int
1818 zfs_ioc_obj_to_stats(zfs_cmd_t *zc)
1819 {
1820 objset_t *os;
1821 int error;
1822
1823 /* XXX reading from objset not owned */
1824 if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os)) != 0)
1825 return (error);
1826 if (dmu_objset_type(os) != DMU_OST_ZFS) {
1827 dmu_objset_rele(os, FTAG);
1828 return (SET_ERROR(EINVAL));
1829 }
1830 error = zfs_obj_to_stats(os, zc->zc_obj, &zc->zc_stat, zc->zc_value,
1831 sizeof (zc->zc_value));
1832 dmu_objset_rele(os, FTAG);
1833
1834 return (error);
1835 }
1836
1837 static int
1838 zfs_ioc_vdev_add(zfs_cmd_t *zc)
1839 {
1840 spa_t *spa;
1841 int error;
1842 nvlist_t *config, **l2cache, **spares;
1843 uint_t nl2cache = 0, nspares = 0;
1844
1845 error = spa_open(zc->zc_name, &spa, FTAG);
1846 if (error != 0)
1847 return (error);
1848
1849 error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1850 zc->zc_iflags, &config);
1851 (void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_L2CACHE,
1852 &l2cache, &nl2cache);
1853
1854 (void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_SPARES,
1855 &spares, &nspares);
1856
1857 /*
1858 * A root pool with concatenated devices is not supported.
1859 * Thus, can not add a device to a root pool.
1860 *
1861 * Intent log device can not be added to a rootpool because
1862 * during mountroot, zil is replayed, a seperated log device
1863 * can not be accessed during the mountroot time.
1864 *
1865 * l2cache and spare devices are ok to be added to a rootpool.
1866 */
1867 if (spa_bootfs(spa) != 0 && nl2cache == 0 && nspares == 0) {
1868 nvlist_free(config);
1869 spa_close(spa, FTAG);
1870 return (SET_ERROR(EDOM));
1871 }
1872
1873 if (error == 0) {
1874 error = spa_vdev_add(spa, config);
1875 nvlist_free(config);
1876 }
1877 spa_close(spa, FTAG);
1878 return (error);
1879 }
1880
1881 /*
1882 * inputs:
1883 * zc_name name of the pool
1884 * zc_nvlist_conf nvlist of devices to remove
1885 * zc_cookie to stop the remove?
1886 */
1887 static int
1888 zfs_ioc_vdev_remove(zfs_cmd_t *zc)
1889 {
1890 spa_t *spa;
1891 int error;
1892
1893 error = spa_open(zc->zc_name, &spa, FTAG);
1894 if (error != 0)
1895 return (error);
1896 error = spa_vdev_remove(spa, zc->zc_guid, B_FALSE);
1897 spa_close(spa, FTAG);
1898 return (error);
1899 }
1900
1901 static int
1902 zfs_ioc_vdev_set_state(zfs_cmd_t *zc)
1903 {
1904 spa_t *spa;
1905 int error;
1906 vdev_state_t newstate = VDEV_STATE_UNKNOWN;
1907
1908 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1909 return (error);
1910 switch (zc->zc_cookie) {
1911 case VDEV_STATE_ONLINE:
1912 error = vdev_online(spa, zc->zc_guid, zc->zc_obj, &newstate);
1913 break;
1914
1915 case VDEV_STATE_OFFLINE:
1916 error = vdev_offline(spa, zc->zc_guid, zc->zc_obj);
1917 break;
1918
1919 case VDEV_STATE_FAULTED:
1920 if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED &&
1921 zc->zc_obj != VDEV_AUX_EXTERNAL)
1922 zc->zc_obj = VDEV_AUX_ERR_EXCEEDED;
1923
1924 error = vdev_fault(spa, zc->zc_guid, zc->zc_obj);
1925 break;
1926
1927 case VDEV_STATE_DEGRADED:
1928 if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED &&
1929 zc->zc_obj != VDEV_AUX_EXTERNAL)
1930 zc->zc_obj = VDEV_AUX_ERR_EXCEEDED;
1931
1932 error = vdev_degrade(spa, zc->zc_guid, zc->zc_obj);
1933 break;
1934
1935 default:
1936 error = SET_ERROR(EINVAL);
1937 }
1938 zc->zc_cookie = newstate;
1939 spa_close(spa, FTAG);
1940 return (error);
1941 }
1942
1943 static int
1944 zfs_ioc_vdev_attach(zfs_cmd_t *zc)
1945 {
1946 spa_t *spa;
1947 int replacing = zc->zc_cookie;
1948 nvlist_t *config;
1949 int error;
1950
1951 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1952 return (error);
1953
1954 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1955 zc->zc_iflags, &config)) == 0) {
1956 error = spa_vdev_attach(spa, zc->zc_guid, config, replacing);
1957 nvlist_free(config);
1958 }
1959
1960 spa_close(spa, FTAG);
1961 return (error);
1962 }
1963
1964 static int
1965 zfs_ioc_vdev_detach(zfs_cmd_t *zc)
1966 {
1967 spa_t *spa;
1968 int error;
1969
1970 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1971 return (error);
1972
1973 error = spa_vdev_detach(spa, zc->zc_guid, 0, B_FALSE);
1974
1975 spa_close(spa, FTAG);
1976 return (error);
1977 }
1978
1979 static int
1980 zfs_ioc_vdev_split(zfs_cmd_t *zc)
1981 {
1982 spa_t *spa;
1983 nvlist_t *config, *props = NULL;
1984 int error;
1985 boolean_t exp = !!(zc->zc_cookie & ZPOOL_EXPORT_AFTER_SPLIT);
1986
1987 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1988 return (error);
1989
1990 if (error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1991 zc->zc_iflags, &config)) {
1992 spa_close(spa, FTAG);
1993 return (error);
1994 }
1995
1996 if (zc->zc_nvlist_src_size != 0 && (error =
1997 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1998 zc->zc_iflags, &props))) {
1999 spa_close(spa, FTAG);
2000 nvlist_free(config);
2001 return (error);
2002 }
2003
2004 error = spa_vdev_split_mirror(spa, zc->zc_string, config, props, exp);
2005
2006 spa_close(spa, FTAG);
2007
2008 nvlist_free(config);
2009 nvlist_free(props);
2010
2011 return (error);
2012 }
2013
2014 static int
2015 zfs_ioc_vdev_setpath(zfs_cmd_t *zc)
2016 {
2017 spa_t *spa;
2018 char *path = zc->zc_value;
2019 uint64_t guid = zc->zc_guid;
2020 int error;
2021
2022 error = spa_open(zc->zc_name, &spa, FTAG);
2023 if (error != 0)
2024 return (error);
2025
2026 error = spa_vdev_setpath(spa, guid, path);
2027 spa_close(spa, FTAG);
2028 return (error);
2029 }
2030
2031 static int
2032 zfs_ioc_vdev_setfru(zfs_cmd_t *zc)
2033 {
2034 spa_t *spa;
2035 char *fru = zc->zc_value;
2036 uint64_t guid = zc->zc_guid;
2037 int error;
2038
2039 error = spa_open(zc->zc_name, &spa, FTAG);
2040 if (error != 0)
2041 return (error);
2042
2043 error = spa_vdev_setfru(spa, guid, fru);
2044 spa_close(spa, FTAG);
2045 return (error);
2046 }
2047
2048 static int
2049 zfs_ioc_objset_stats_impl(zfs_cmd_t *zc, objset_t *os)
2050 {
2051 int error = 0;
2052 nvlist_t *nv;
2053
2054 dmu_objset_fast_stat(os, &zc->zc_objset_stats);
2055
2056 if (zc->zc_nvlist_dst != 0 &&
2057 (error = dsl_prop_get_all(os, &nv)) == 0) {
2058 dmu_objset_stats(os, nv);
2059 /*
2060 * NB: zvol_get_stats() will read the objset contents,
2061 * which we aren't supposed to do with a
2062 * DS_MODE_USER hold, because it could be
2063 * inconsistent. So this is a bit of a workaround...
2064 * XXX reading with out owning
2065 */
2066 if (!zc->zc_objset_stats.dds_inconsistent &&
2067 dmu_objset_type(os) == DMU_OST_ZVOL) {
2068 error = zvol_get_stats(os, nv);
2069 if (error == EIO)
2070 return (error);
2071 VERIFY0(error);
2072 }
2073 error = put_nvlist(zc, nv);
2074 nvlist_free(nv);
2075 }
2076
2077 return (error);
2078 }
2079
2080 /*
2081 * inputs:
2082 * zc_name name of filesystem
2083 * zc_nvlist_dst_size size of buffer for property nvlist
2084 *
2085 * outputs:
2086 * zc_objset_stats stats
2087 * zc_nvlist_dst property nvlist
2088 * zc_nvlist_dst_size size of property nvlist
2089 */
2090 static int
2091 zfs_ioc_objset_stats(zfs_cmd_t *zc)
2092 {
2093 objset_t *os;
2094 int error;
2095
2096 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
2097 if (error == 0) {
2098 error = zfs_ioc_objset_stats_impl(zc, os);
2099 dmu_objset_rele(os, FTAG);
2100 }
2101
2102 return (error);
2103 }
2104
2105 /*
2106 * inputs:
2107 * zc_name name of filesystem
2108 * zc_nvlist_dst_size size of buffer for property nvlist
2109 *
2110 * outputs:
2111 * zc_nvlist_dst received property nvlist
2112 * zc_nvlist_dst_size size of received property nvlist
2113 *
2114 * Gets received properties (distinct from local properties on or after
2115 * SPA_VERSION_RECVD_PROPS) for callers who want to differentiate received from
2116 * local property values.
2117 */
2118 static int
2119 zfs_ioc_objset_recvd_props(zfs_cmd_t *zc)
2120 {
2121 int error = 0;
2122 nvlist_t *nv;
2123
2124 /*
2125 * Without this check, we would return local property values if the
2126 * caller has not already received properties on or after
2127 * SPA_VERSION_RECVD_PROPS.
2128 */
2129 if (!dsl_prop_get_hasrecvd(zc->zc_name))
2130 return (SET_ERROR(ENOTSUP));
2131
2132 if (zc->zc_nvlist_dst != 0 &&
2133 (error = dsl_prop_get_received(zc->zc_name, &nv)) == 0) {
2134 error = put_nvlist(zc, nv);
2135 nvlist_free(nv);
2136 }
2137
2138 return (error);
2139 }
2140
2141 static int
2142 nvl_add_zplprop(objset_t *os, nvlist_t *props, zfs_prop_t prop)
2143 {
2144 uint64_t value;
2145 int error;
2146
2147 /*
2148 * zfs_get_zplprop() will either find a value or give us
2149 * the default value (if there is one).
2150 */
2151 if ((error = zfs_get_zplprop(os, prop, &value)) != 0)
2152 return (error);
2153 VERIFY(nvlist_add_uint64(props, zfs_prop_to_name(prop), value) == 0);
2154 return (0);
2155 }
2156
2157 /*
2158 * inputs:
2159 * zc_name name of filesystem
2160 * zc_nvlist_dst_size size of buffer for zpl property nvlist
2161 *
2162 * outputs:
2163 * zc_nvlist_dst zpl property nvlist
2164 * zc_nvlist_dst_size size of zpl property nvlist
2165 */
2166 static int
2167 zfs_ioc_objset_zplprops(zfs_cmd_t *zc)
2168 {
2169 objset_t *os;
2170 int err;
2171
2172 /* XXX reading without owning */
2173 if (err = dmu_objset_hold(zc->zc_name, FTAG, &os))
2174 return (err);
2175
2176 dmu_objset_fast_stat(os, &zc->zc_objset_stats);
2177
2178 /*
2179 * NB: nvl_add_zplprop() will read the objset contents,
2180 * which we aren't supposed to do with a DS_MODE_USER
2181 * hold, because it could be inconsistent.
2182 */
2183 if (zc->zc_nvlist_dst != NULL &&
2184 !zc->zc_objset_stats.dds_inconsistent &&
2185 dmu_objset_type(os) == DMU_OST_ZFS) {
2186 nvlist_t *nv;
2187
2188 VERIFY(nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2189 if ((err = nvl_add_zplprop(os, nv, ZFS_PROP_VERSION)) == 0 &&
2190 (err = nvl_add_zplprop(os, nv, ZFS_PROP_NORMALIZE)) == 0 &&
2191 (err = nvl_add_zplprop(os, nv, ZFS_PROP_UTF8ONLY)) == 0 &&
2192 (err = nvl_add_zplprop(os, nv, ZFS_PROP_CASE)) == 0)
2193 err = put_nvlist(zc, nv);
2194 nvlist_free(nv);
2195 } else {
2196 err = SET_ERROR(ENOENT);
2197 }
2198 dmu_objset_rele(os, FTAG);
2199 return (err);
2200 }
2201
2202 static boolean_t
2203 dataset_name_hidden(const char *name)
2204 {
2205 /*
2206 * Skip over datasets that are not visible in this zone,
2207 * internal datasets (which have a $ in their name), and
2208 * temporary datasets (which have a % in their name).
2209 */
2210 if (strchr(name, '$') != NULL)
2211 return (B_TRUE);
2212 if (strchr(name, '%') != NULL)
2213 return (B_TRUE);
2214 if (!INGLOBALZONE(curproc) && !zone_dataset_visible(name, NULL))
2215 return (B_TRUE);
2216 return (B_FALSE);
2217 }
2218
2219 /*
2220 * inputs:
2221 * zc_name name of filesystem
2222 * zc_cookie zap cursor
2223 * zc_nvlist_dst_size size of buffer for property nvlist
2224 *
2225 * outputs:
2226 * zc_name name of next filesystem
2227 * zc_cookie zap cursor
2228 * zc_objset_stats stats
2229 * zc_nvlist_dst property nvlist
2230 * zc_nvlist_dst_size size of property nvlist
2231 */
2232 static int
2233 zfs_ioc_dataset_list_next(zfs_cmd_t *zc)
2234 {
2235 objset_t *os;
2236 int error;
2237 char *p;
2238 size_t orig_len = strlen(zc->zc_name);
2239
2240 top:
2241 if (error = dmu_objset_hold(zc->zc_name, FTAG, &os)) {
2242 if (error == ENOENT)
2243 error = SET_ERROR(ESRCH);
2244 return (error);
2245 }
2246
2247 p = strrchr(zc->zc_name, '/');
2248 if (p == NULL || p[1] != '\0')
2249 (void) strlcat(zc->zc_name, "/", sizeof (zc->zc_name));
2250 p = zc->zc_name + strlen(zc->zc_name);
2251
2252 do {
2253 error = dmu_dir_list_next(os,
2254 sizeof (zc->zc_name) - (p - zc->zc_name), p,
2255 NULL, &zc->zc_cookie);
2256 if (error == ENOENT)
2257 error = SET_ERROR(ESRCH);
2258 } while (error == 0 && dataset_name_hidden(zc->zc_name));
2259 dmu_objset_rele(os, FTAG);
2260
2261 /*
2262 * If it's an internal dataset (ie. with a '$' in its name),
2263 * don't try to get stats for it, otherwise we'll return ENOENT.
2264 */
2265 if (error == 0 && strchr(zc->zc_name, '$') == NULL) {
2266 error = zfs_ioc_objset_stats(zc); /* fill in the stats */
2267 if (error == ENOENT) {
2268 /* We lost a race with destroy, get the next one. */
2269 zc->zc_name[orig_len] = '\0';
2270 goto top;
2271 }
2272 }
2273 return (error);
2274 }
2275
2276 /*
2277 * inputs:
2278 * zc_name name of filesystem
2279 * zc_cookie zap cursor
2280 * zc_nvlist_dst_size size of buffer for property nvlist
2281 * zc_simple when set, only name is requested
2282 *
2283 * outputs:
2284 * zc_name name of next snapshot
2285 * zc_objset_stats stats
2286 * zc_nvlist_dst property nvlist
2287 * zc_nvlist_dst_size size of property nvlist
2288 */
2289 static int
2290 zfs_ioc_snapshot_list_next(zfs_cmd_t *zc)
2291 {
2292 objset_t *os;
2293 int error;
2294
2295 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
2296 if (error != 0) {
2297 return (error == ENOENT ? ESRCH : error);
2298 }
2299
2300 /*
2301 * A dataset name of maximum length cannot have any snapshots,
2302 * so exit immediately.
2303 */
2304 if (strlcat(zc->zc_name, "@", sizeof (zc->zc_name)) >=
2305 ZFS_MAX_DATASET_NAME_LEN) {
2306 dmu_objset_rele(os, FTAG);
2307 return (SET_ERROR(ESRCH));
2308 }
2309
2310 error = dmu_snapshot_list_next(os,
2311 sizeof (zc->zc_name) - strlen(zc->zc_name),
2312 zc->zc_name + strlen(zc->zc_name), &zc->zc_obj, &zc->zc_cookie,
2313 NULL);
2314
2315 if (error == 0 && !zc->zc_simple) {
2316 dsl_dataset_t *ds;
2317 dsl_pool_t *dp = os->os_dsl_dataset->ds_dir->dd_pool;
2318
2319 error = dsl_dataset_hold_obj(dp, zc->zc_obj, FTAG, &ds);
2320 if (error == 0) {
2321 objset_t *ossnap;
2322
2323 error = dmu_objset_from_ds(ds, &ossnap);
2324 if (error == 0)
2325 error = zfs_ioc_objset_stats_impl(zc, ossnap);
2326 dsl_dataset_rele(ds, FTAG);
2327 }
2328 } else if (error == ENOENT) {
2329 error = SET_ERROR(ESRCH);
2330 }
2331
2332 dmu_objset_rele(os, FTAG);
2333 /* if we failed, undo the @ that we tacked on to zc_name */
2334 if (error != 0)
2335 *strchr(zc->zc_name, '@') = '\0';
2336 return (error);
2337 }
2338
2339 static int
2340 zfs_prop_set_userquota(const char *dsname, nvpair_t *pair)
2341 {
2342 const char *propname = nvpair_name(pair);
2343 uint64_t *valary;
2344 unsigned int vallen;
2345 const char *domain;
2346 char *dash;
2347 zfs_userquota_prop_t type;
2348 uint64_t rid;
2349 uint64_t quota;
2350 zfsvfs_t *zfsvfs;
2351 int err;
2352
2353 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2354 nvlist_t *attrs;
2355 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
2356 if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2357 &pair) != 0)
2358 return (SET_ERROR(EINVAL));
2359 }
2360
2361 /*
2362 * A correctly constructed propname is encoded as
2363 * userquota@<rid>-<domain>.
2364 */
2365 if ((dash = strchr(propname, '-')) == NULL ||
2366 nvpair_value_uint64_array(pair, &valary, &vallen) != 0 ||
2367 vallen != 3)
2368 return (SET_ERROR(EINVAL));
2369
2370 domain = dash + 1;
2371 type = valary[0];
2372 rid = valary[1];
2373 quota = valary[2];
2374
2375 err = zfsvfs_hold(dsname, FTAG, &zfsvfs, B_FALSE);
2376 if (err == 0) {
2377 err = zfs_set_userquota(zfsvfs, type, domain, rid, quota);
2378 zfsvfs_rele(zfsvfs, FTAG);
2379 }
2380
2381 return (err);
2382 }
2383
2384 /*
2385 * If the named property is one that has a special function to set its value,
2386 * return 0 on success and a positive error code on failure; otherwise if it is
2387 * not one of the special properties handled by this function, return -1.
2388 *
2389 * XXX: It would be better for callers of the property interface if we handled
2390 * these special cases in dsl_prop.c (in the dsl layer).
2391 */
2392 static int
2393 zfs_prop_set_special(const char *dsname, zprop_source_t source,
2394 nvpair_t *pair)
2395 {
2396 const char *propname = nvpair_name(pair);
2397 zfs_prop_t prop = zfs_name_to_prop(propname);
2398 uint64_t intval;
2399 int err = -1;
2400
2401 if (prop == ZPROP_INVAL) {
2402 if (zfs_prop_userquota(propname))
2403 return (zfs_prop_set_userquota(dsname, pair));
2404 return (-1);
2405 }
2406
2407 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2408 nvlist_t *attrs;
2409 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
2410 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2411 &pair) == 0);
2412 }
2413
2414 if (zfs_prop_get_type(prop) == PROP_TYPE_STRING)
2415 return (-1);
2416
2417 VERIFY(0 == nvpair_value_uint64(pair, &intval));
2418
2419 switch (prop) {
2420 case ZFS_PROP_QUOTA:
2421 err = dsl_dir_set_quota(dsname, source, intval);
2422 break;
2423 case ZFS_PROP_REFQUOTA:
2424 err = dsl_dataset_set_refquota(dsname, source, intval);
2425 break;
2426 case ZFS_PROP_FILESYSTEM_LIMIT:
2427 case ZFS_PROP_SNAPSHOT_LIMIT:
2428 if (intval == UINT64_MAX) {
2429 /* clearing the limit, just do it */
2430 err = 0;
2431 } else {
2432 err = dsl_dir_activate_fs_ss_limit(dsname);
2433 }
2434 /*
2435 * Set err to -1 to force the zfs_set_prop_nvlist code down the
2436 * default path to set the value in the nvlist.
2437 */
2438 if (err == 0)
2439 err = -1;
2440 break;
2441 case ZFS_PROP_RESERVATION:
2442 err = dsl_dir_set_reservation(dsname, source, intval);
2443 break;
2444 case ZFS_PROP_REFRESERVATION:
2445 err = dsl_dataset_set_refreservation(dsname, source, intval);
2446 break;
2447 case ZFS_PROP_VOLSIZE:
2448 err = zvol_set_volsize(dsname, intval);
2449 break;
2450 case ZFS_PROP_VERSION:
2451 {
2452 zfsvfs_t *zfsvfs;
2453
2454 if ((err = zfsvfs_hold(dsname, FTAG, &zfsvfs, B_TRUE)) != 0)
2455 break;
2456
2457 err = zfs_set_version(zfsvfs, intval);
2458 zfsvfs_rele(zfsvfs, FTAG);
2459
2460 if (err == 0 && intval >= ZPL_VERSION_USERSPACE) {
2461 zfs_cmd_t *zc;
2462
2463 zc = kmem_zalloc(sizeof (zfs_cmd_t), KM_SLEEP);
2464 (void) strcpy(zc->zc_name, dsname);
2465 (void) zfs_ioc_userspace_upgrade(zc);
2466 kmem_free(zc, sizeof (zfs_cmd_t));
2467 }
2468 break;
2469 }
2470 default:
2471 err = -1;
2472 }
2473
2474 return (err);
2475 }
2476
2477 /*
2478 * This function is best effort. If it fails to set any of the given properties,
2479 * it continues to set as many as it can and returns the last error
2480 * encountered. If the caller provides a non-NULL errlist, it will be filled in
2481 * with the list of names of all the properties that failed along with the
2482 * corresponding error numbers.
2483 *
2484 * If every property is set successfully, zero is returned and errlist is not
2485 * modified.
2486 */
2487 int
2488 zfs_set_prop_nvlist(const char *dsname, zprop_source_t source, nvlist_t *nvl,
2489 nvlist_t *errlist)
2490 {
2491 nvpair_t *pair;
2492 nvpair_t *propval;
2493 int rv = 0;
2494 uint64_t intval;
2495 char *strval;
2496 nvlist_t *genericnvl = fnvlist_alloc();
2497 nvlist_t *retrynvl = fnvlist_alloc();
2498
2499 retry:
2500 pair = NULL;
2501 while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) {
2502 const char *propname = nvpair_name(pair);
2503 zfs_prop_t prop = zfs_name_to_prop(propname);
2504 int err = 0;
2505
2506 /* decode the property value */
2507 propval = pair;
2508 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2509 nvlist_t *attrs;
2510 attrs = fnvpair_value_nvlist(pair);
2511 if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2512 &propval) != 0)
2513 err = SET_ERROR(EINVAL);
2514 }
2515
2516 /* Validate value type */
2517 if (err == 0 && prop == ZPROP_INVAL) {
2518 if (zfs_prop_user(propname)) {
2519 if (nvpair_type(propval) != DATA_TYPE_STRING)
2520 err = SET_ERROR(EINVAL);
2521 } else if (zfs_prop_userquota(propname)) {
2522 if (nvpair_type(propval) !=
2523 DATA_TYPE_UINT64_ARRAY)
2524 err = SET_ERROR(EINVAL);
2525 } else {
2526 err = SET_ERROR(EINVAL);
2527 }
2528 } else if (err == 0) {
2529 if (nvpair_type(propval) == DATA_TYPE_STRING) {
2530 if (zfs_prop_get_type(prop) != PROP_TYPE_STRING)
2531 err = SET_ERROR(EINVAL);
2532 } else if (nvpair_type(propval) == DATA_TYPE_UINT64) {
2533 const char *unused;
2534
2535 intval = fnvpair_value_uint64(propval);
2536
2537 switch (zfs_prop_get_type(prop)) {
2538 case PROP_TYPE_NUMBER:
2539 break;
2540 case PROP_TYPE_STRING:
2541 err = SET_ERROR(EINVAL);
2542 break;
2543 case PROP_TYPE_INDEX:
2544 if (zfs_prop_index_to_string(prop,
2545 intval, &unused) != 0)
2546 err = SET_ERROR(EINVAL);
2547 break;
2548 default:
2549 cmn_err(CE_PANIC,
2550 "unknown property type");
2551 }
2552 } else {
2553 err = SET_ERROR(EINVAL);
2554 }
2555 }
2556
2557 /* Validate permissions */
2558 if (err == 0)
2559 err = zfs_check_settable(dsname, pair, CRED());
2560
2561 if (err == 0) {
2562 err = zfs_prop_set_special(dsname, source, pair);
2563 if (err == -1) {
2564 /*
2565 * For better performance we build up a list of
2566 * properties to set in a single transaction.
2567 */
2568 err = nvlist_add_nvpair(genericnvl, pair);
2569 } else if (err != 0 && nvl != retrynvl) {
2570 /*
2571 * This may be a spurious error caused by
2572 * receiving quota and reservation out of order.
2573 * Try again in a second pass.
2574 */
2575 err = nvlist_add_nvpair(retrynvl, pair);
2576 }
2577 }
2578
2579 if (err != 0) {
2580 if (errlist != NULL)
2581 fnvlist_add_int32(errlist, propname, err);
2582 rv = err;
2583 }
2584 }
2585
2586 if (nvl != retrynvl && !nvlist_empty(retrynvl)) {
2587 nvl = retrynvl;
2588 goto retry;
2589 }
2590
2591 if (!nvlist_empty(genericnvl) &&
2592 dsl_props_set(dsname, source, genericnvl) != 0) {
2593 /*
2594 * If this fails, we still want to set as many properties as we
2595 * can, so try setting them individually.
2596 */
2597 pair = NULL;
2598 while ((pair = nvlist_next_nvpair(genericnvl, pair)) != NULL) {
2599 const char *propname = nvpair_name(pair);
2600 int err = 0;
2601
2602 propval = pair;
2603 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2604 nvlist_t *attrs;
2605 attrs = fnvpair_value_nvlist(pair);
2606 propval = fnvlist_lookup_nvpair(attrs,
2607 ZPROP_VALUE);
2608 }
2609
2610 if (nvpair_type(propval) == DATA_TYPE_STRING) {
2611 strval = fnvpair_value_string(propval);
2612 err = dsl_prop_set_string(dsname, propname,
2613 source, strval);
2614 } else {
2615 intval = fnvpair_value_uint64(propval);
2616 err = dsl_prop_set_int(dsname, propname, source,
2617 intval);
2618 }
2619
2620 if (err != 0) {
2621 if (errlist != NULL) {
2622 fnvlist_add_int32(errlist, propname,
2623 err);
2624 }
2625 rv = err;
2626 }
2627 }
2628 }
2629 nvlist_free(genericnvl);
2630 nvlist_free(retrynvl);
2631
2632 return (rv);
2633 }
2634
2635 /*
2636 * Check that all the properties are valid user properties.
2637 */
2638 static int
2639 zfs_check_userprops(const char *fsname, nvlist_t *nvl)
2640 {
2641 nvpair_t *pair = NULL;
2642 int error = 0;
2643
2644 while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) {
2645 const char *propname = nvpair_name(pair);
2646
2647 if (!zfs_prop_user(propname) ||
2648 nvpair_type(pair) != DATA_TYPE_STRING)
2649 return (SET_ERROR(EINVAL));
2650
2651 if (error = zfs_secpolicy_write_perms(fsname,
2652 ZFS_DELEG_PERM_USERPROP, CRED()))
2653 return (error);
2654
2655 if (strlen(propname) >= ZAP_MAXNAMELEN)
2656 return (SET_ERROR(ENAMETOOLONG));
2657
2658 if (strlen(fnvpair_value_string(pair)) >= ZAP_MAXVALUELEN)
2659 return (E2BIG);
2660 }
2661 return (0);
2662 }
2663
2664 static void
2665 props_skip(nvlist_t *props, nvlist_t *skipped, nvlist_t **newprops)
2666 {
2667 nvpair_t *pair;
2668
2669 VERIFY(nvlist_alloc(newprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2670
2671 pair = NULL;
2672 while ((pair = nvlist_next_nvpair(props, pair)) != NULL) {
2673 if (nvlist_exists(skipped, nvpair_name(pair)))
2674 continue;
2675
2676 VERIFY(nvlist_add_nvpair(*newprops, pair) == 0);
2677 }
2678 }
2679
2680 static int
2681 clear_received_props(const char *dsname, nvlist_t *props,
2682 nvlist_t *skipped)
2683 {
2684 int err = 0;
2685 nvlist_t *cleared_props = NULL;
2686 props_skip(props, skipped, &cleared_props);
2687 if (!nvlist_empty(cleared_props)) {
2688 /*
2689 * Acts on local properties until the dataset has received
2690 * properties at least once on or after SPA_VERSION_RECVD_PROPS.
2691 */
2692 zprop_source_t flags = (ZPROP_SRC_NONE |
2693 (dsl_prop_get_hasrecvd(dsname) ? ZPROP_SRC_RECEIVED : 0));
2694 err = zfs_set_prop_nvlist(dsname, flags, cleared_props, NULL);
2695 }
2696 nvlist_free(cleared_props);
2697 return (err);
2698 }
2699
2700 /*
2701 * inputs:
2702 * zc_name name of filesystem
2703 * zc_value name of property to set
2704 * zc_nvlist_src{_size} nvlist of properties to apply
2705 * zc_cookie received properties flag
2706 *
2707 * outputs:
2708 * zc_nvlist_dst{_size} error for each unapplied received property
2709 */
2710 static int
2711 zfs_ioc_set_prop(zfs_cmd_t *zc)
2712 {
2713 nvlist_t *nvl;
2714 boolean_t received = zc->zc_cookie;
2715 zprop_source_t source = (received ? ZPROP_SRC_RECEIVED :
2716 ZPROP_SRC_LOCAL);
2717 nvlist_t *errors;
2718 int error;
2719
2720 if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2721 zc->zc_iflags, &nvl)) != 0)
2722 return (error);
2723
2724 if (received) {
2725 nvlist_t *origprops;
2726
2727 if (dsl_prop_get_received(zc->zc_name, &origprops) == 0) {
2728 (void) clear_received_props(zc->zc_name,
2729 origprops, nvl);
2730 nvlist_free(origprops);
2731 }
2732
2733 error = dsl_prop_set_hasrecvd(zc->zc_name);
2734 }
2735
2736 errors = fnvlist_alloc();
2737 if (error == 0)
2738 error = zfs_set_prop_nvlist(zc->zc_name, source, nvl, errors);
2739
2740 if (zc->zc_nvlist_dst != NULL && errors != NULL) {
2741 (void) put_nvlist(zc, errors);
2742 }
2743
2744 nvlist_free(errors);
2745 nvlist_free(nvl);
2746 return (error);
2747 }
2748
2749 /*
2750 * inputs:
2751 * zc_name name of filesystem
2752 * zc_value name of property to inherit
2753 * zc_cookie revert to received value if TRUE
2754 *
2755 * outputs: none
2756 */
2757 static int
2758 zfs_ioc_inherit_prop(zfs_cmd_t *zc)
2759 {
2760 const char *propname = zc->zc_value;
2761 zfs_prop_t prop = zfs_name_to_prop(propname);
2762 boolean_t received = zc->zc_cookie;
2763 zprop_source_t source = (received
2764 ? ZPROP_SRC_NONE /* revert to received value, if any */
2765 : ZPROP_SRC_INHERITED); /* explicitly inherit */
2766
2767 if (received) {
2768 nvlist_t *dummy;
2769 nvpair_t *pair;
2770 zprop_type_t type;
2771 int err;
2772
2773 /*
2774 * zfs_prop_set_special() expects properties in the form of an
2775 * nvpair with type info.
2776 */
2777 if (prop == ZPROP_INVAL) {
2778 if (!zfs_prop_user(propname))
2779 return (SET_ERROR(EINVAL));
2780
2781 type = PROP_TYPE_STRING;
2782 } else if (prop == ZFS_PROP_VOLSIZE ||
2783 prop == ZFS_PROP_VERSION) {
2784 return (SET_ERROR(EINVAL));
2785 } else {
2786 type = zfs_prop_get_type(prop);
2787 }
2788
2789 VERIFY(nvlist_alloc(&dummy, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2790
2791 switch (type) {
2792 case PROP_TYPE_STRING:
2793 VERIFY(0 == nvlist_add_string(dummy, propname, ""));
2794 break;
2795 case PROP_TYPE_NUMBER:
2796 case PROP_TYPE_INDEX:
2797 VERIFY(0 == nvlist_add_uint64(dummy, propname, 0));
2798 break;
2799 default:
2800 nvlist_free(dummy);
2801 return (SET_ERROR(EINVAL));
2802 }
2803
2804 pair = nvlist_next_nvpair(dummy, NULL);
2805 err = zfs_prop_set_special(zc->zc_name, source, pair);
2806 nvlist_free(dummy);
2807 if (err != -1)
2808 return (err); /* special property already handled */
2809 } else {
2810 /*
2811 * Only check this in the non-received case. We want to allow
2812 * 'inherit -S' to revert non-inheritable properties like quota
2813 * and reservation to the received or default values even though
2814 * they are not considered inheritable.
2815 */
2816 if (prop != ZPROP_INVAL && !zfs_prop_inheritable(prop))
2817 return (SET_ERROR(EINVAL));
2818 }
2819
2820 /* property name has been validated by zfs_secpolicy_inherit_prop() */
2821 return (dsl_prop_inherit(zc->zc_name, zc->zc_value, source));
2822 }
2823
2824 static int
2825 zfs_ioc_pool_set_props(zfs_cmd_t *zc)
2826 {
2827 nvlist_t *props;
2828 spa_t *spa;
2829 int error;
2830 nvpair_t *pair;
2831
2832 if (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2833 zc->zc_iflags, &props))
2834 return (error);
2835
2836 /*
2837 * If the only property is the configfile, then just do a spa_lookup()
2838 * to handle the faulted case.
2839 */
2840 pair = nvlist_next_nvpair(props, NULL);
2841 if (pair != NULL && strcmp(nvpair_name(pair),
2842 zpool_prop_to_name(ZPOOL_PROP_CACHEFILE)) == 0 &&
2843 nvlist_next_nvpair(props, pair) == NULL) {
2844 mutex_enter(&spa_namespace_lock);
2845 if ((spa = spa_lookup(zc->zc_name)) != NULL) {
2846 spa_configfile_set(spa, props, B_FALSE);
2847 spa_config_sync(spa, B_FALSE, B_TRUE);
2848 }
2849 mutex_exit(&spa_namespace_lock);
2850 if (spa != NULL) {
2851 nvlist_free(props);
2852 return (0);
2853 }
2854 }
2855
2856 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) {
2857 nvlist_free(props);
2858 return (error);
2859 }
2860
2861 error = spa_prop_set(spa, props);
2862
2863 nvlist_free(props);
2864 spa_close(spa, FTAG);
2865
2866 return (error);
2867 }
2868
2869 static int
2870 zfs_ioc_pool_get_props(zfs_cmd_t *zc)
2871 {
2872 spa_t *spa;
2873 int error;
2874 nvlist_t *nvp = NULL;
2875
2876 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) {
2877 /*
2878 * If the pool is faulted, there may be properties we can still
2879 * get (such as altroot and cachefile), so attempt to get them
2880 * anyway.
2881 */
2882 mutex_enter(&spa_namespace_lock);
2883 if ((spa = spa_lookup(zc->zc_name)) != NULL)
2884 error = spa_prop_get(spa, &nvp);
2885 mutex_exit(&spa_namespace_lock);
2886 } else {
2887 error = spa_prop_get(spa, &nvp);
2888 spa_close(spa, FTAG);
2889 }
2890
2891 if (error == 0 && zc->zc_nvlist_dst != NULL)
2892 error = put_nvlist(zc, nvp);
2893 else
2894 error = SET_ERROR(EFAULT);
2895
2896 nvlist_free(nvp);
2897 return (error);
2898 }
2899
2900 /*
2901 * inputs:
2902 * zc_name name of filesystem
2903 * zc_nvlist_src{_size} nvlist of delegated permissions
2904 * zc_perm_action allow/unallow flag
2905 *
2906 * outputs: none
2907 */
2908 static int
2909 zfs_ioc_set_fsacl(zfs_cmd_t *zc)
2910 {
2911 int error;
2912 nvlist_t *fsaclnv = NULL;
2913
2914 if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2915 zc->zc_iflags, &fsaclnv)) != 0)
2916 return (error);
2917
2918 /*
2919 * Verify nvlist is constructed correctly
2920 */
2921 if ((error = zfs_deleg_verify_nvlist(fsaclnv)) != 0) {
2922 nvlist_free(fsaclnv);
2923 return (SET_ERROR(EINVAL));
2924 }
2925
2926 /*
2927 * If we don't have PRIV_SYS_MOUNT, then validate
2928 * that user is allowed to hand out each permission in
2929 * the nvlist(s)
2930 */
2931
2932 error = secpolicy_zfs(CRED());
2933 if (error != 0) {
2934 if (zc->zc_perm_action == B_FALSE) {
2935 error = dsl_deleg_can_allow(zc->zc_name,
2936 fsaclnv, CRED());
2937 } else {
2938 error = dsl_deleg_can_unallow(zc->zc_name,
2939 fsaclnv, CRED());
2940 }
2941 }
2942
2943 if (error == 0)
2944 error = dsl_deleg_set(zc->zc_name, fsaclnv, zc->zc_perm_action);
2945
2946 nvlist_free(fsaclnv);
2947 return (error);
2948 }
2949
2950 /*
2951 * inputs:
2952 * zc_name name of filesystem
2953 *
2954 * outputs:
2955 * zc_nvlist_src{_size} nvlist of delegated permissions
2956 */
2957 static int
2958 zfs_ioc_get_fsacl(zfs_cmd_t *zc)
2959 {
2960 nvlist_t *nvp;
2961 int error;
2962
2963 if ((error = dsl_deleg_get(zc->zc_name, &nvp)) == 0) {
2964 error = put_nvlist(zc, nvp);
2965 nvlist_free(nvp);
2966 }
2967
2968 return (error);
2969 }
2970
2971 /*
2972 * Search the vfs list for a specified resource. Returns a pointer to it
2973 * or NULL if no suitable entry is found. The caller of this routine
2974 * is responsible for releasing the returned vfs pointer.
2975 */
2976 static vfs_t *
2977 zfs_get_vfs(const char *resource)
2978 {
2979 struct vfs *vfsp;
2980 struct vfs *vfs_found = NULL;
2981
2982 vfs_list_read_lock();
2983 vfsp = rootvfs;
2984 do {
2985 if (strcmp(refstr_value(vfsp->vfs_resource), resource) == 0) {
2986 VFS_HOLD(vfsp);
2987 vfs_found = vfsp;
2988 break;
2989 }
2990 vfsp = vfsp->vfs_next;
2991 } while (vfsp != rootvfs);
2992 vfs_list_unlock();
2993 return (vfs_found);
2994 }
2995
2996 /* ARGSUSED */
2997 static void
2998 zfs_create_cb(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx)
2999 {
3000 zfs_creat_t *zct = arg;
3001
3002 zfs_create_fs(os, cr, zct->zct_zplprops, tx);
3003 }
3004
3005 #define ZFS_PROP_UNDEFINED ((uint64_t)-1)
3006
3007 /*
3008 * inputs:
3009 * os parent objset pointer (NULL if root fs)
3010 * fuids_ok fuids allowed in this version of the spa?
3011 * sa_ok SAs allowed in this version of the spa?
3012 * createprops list of properties requested by creator
3013 *
3014 * outputs:
3015 * zplprops values for the zplprops we attach to the master node object
3016 * is_ci true if requested file system will be purely case-insensitive
3017 *
3018 * Determine the settings for utf8only, normalization and
3019 * casesensitivity. Specific values may have been requested by the
3020 * creator and/or we can inherit values from the parent dataset. If
3021 * the file system is of too early a vintage, a creator can not
3022 * request settings for these properties, even if the requested
3023 * setting is the default value. We don't actually want to create dsl
3024 * properties for these, so remove them from the source nvlist after
3025 * processing.
3026 */
3027 static int
3028 zfs_fill_zplprops_impl(objset_t *os, uint64_t zplver,
3029 boolean_t fuids_ok, boolean_t sa_ok, nvlist_t *createprops,
3030 nvlist_t *zplprops, boolean_t *is_ci)
3031 {
3032 uint64_t sense = ZFS_PROP_UNDEFINED;
3033 uint64_t norm = ZFS_PROP_UNDEFINED;
3034 uint64_t u8 = ZFS_PROP_UNDEFINED;
3035
3036 ASSERT(zplprops != NULL);
3037
3038 if (os != NULL && os->os_phys->os_type != DMU_OST_ZFS)
3039 return (SET_ERROR(EINVAL));
3040
3041 /*
3042 * Pull out creator prop choices, if any.
3043 */
3044 if (createprops) {
3045 (void) nvlist_lookup_uint64(createprops,
3046 zfs_prop_to_name(ZFS_PROP_VERSION), &zplver);
3047 (void) nvlist_lookup_uint64(createprops,
3048 zfs_prop_to_name(ZFS_PROP_NORMALIZE), &norm);
3049 (void) nvlist_remove_all(createprops,
3050 zfs_prop_to_name(ZFS_PROP_NORMALIZE));
3051 (void) nvlist_lookup_uint64(createprops,
3052 zfs_prop_to_name(ZFS_PROP_UTF8ONLY), &u8);
3053 (void) nvlist_remove_all(createprops,
3054 zfs_prop_to_name(ZFS_PROP_UTF8ONLY));
3055 (void) nvlist_lookup_uint64(createprops,
3056 zfs_prop_to_name(ZFS_PROP_CASE), &sense);
3057 (void) nvlist_remove_all(createprops,
3058 zfs_prop_to_name(ZFS_PROP_CASE));
3059 }
3060
3061 /*
3062 * If the zpl version requested is whacky or the file system
3063 * or pool is version is too "young" to support normalization
3064 * and the creator tried to set a value for one of the props,
3065 * error out.
3066 */
3067 if ((zplver < ZPL_VERSION_INITIAL || zplver > ZPL_VERSION) ||
3068 (zplver >= ZPL_VERSION_FUID && !fuids_ok) ||
3069 (zplver >= ZPL_VERSION_SA && !sa_ok) ||
3070 (zplver < ZPL_VERSION_NORMALIZATION &&
3071 (norm != ZFS_PROP_UNDEFINED || u8 != ZFS_PROP_UNDEFINED ||
3072 sense != ZFS_PROP_UNDEFINED)))
3073 return (SET_ERROR(ENOTSUP));
3074
3075 /*
3076 * Put the version in the zplprops
3077 */
3078 VERIFY(nvlist_add_uint64(zplprops,
3079 zfs_prop_to_name(ZFS_PROP_VERSION), zplver) == 0);
3080
3081 if (norm == ZFS_PROP_UNDEFINED)
3082 VERIFY(zfs_get_zplprop(os, ZFS_PROP_NORMALIZE, &norm) == 0);
3083 VERIFY(nvlist_add_uint64(zplprops,
3084 zfs_prop_to_name(ZFS_PROP_NORMALIZE), norm) == 0);
3085
3086 /*
3087 * If we're normalizing, names must always be valid UTF-8 strings.
3088 */
3089 if (norm)
3090 u8 = 1;
3091 if (u8 == ZFS_PROP_UNDEFINED)
3092 VERIFY(zfs_get_zplprop(os, ZFS_PROP_UTF8ONLY, &u8) == 0);
3093 VERIFY(nvlist_add_uint64(zplprops,
3094 zfs_prop_to_name(ZFS_PROP_UTF8ONLY), u8) == 0);
3095
3096 if (sense == ZFS_PROP_UNDEFINED)
3097 VERIFY(zfs_get_zplprop(os, ZFS_PROP_CASE, &sense) == 0);
3098 VERIFY(nvlist_add_uint64(zplprops,
3099 zfs_prop_to_name(ZFS_PROP_CASE), sense) == 0);
3100
3101 if (is_ci)
3102 *is_ci = (sense == ZFS_CASE_INSENSITIVE);
3103
3104 return (0);
3105 }
3106
3107 static int
3108 zfs_fill_zplprops(const char *dataset, nvlist_t *createprops,
3109 nvlist_t *zplprops, boolean_t *is_ci)
3110 {
3111 boolean_t fuids_ok, sa_ok;
3112 uint64_t zplver = ZPL_VERSION;
3113 objset_t *os = NULL;
3114 char parentname[ZFS_MAX_DATASET_NAME_LEN];
3115 char *cp;
3116 spa_t *spa;
3117 uint64_t spa_vers;
3118 int error;
3119
3120 (void) strlcpy(parentname, dataset, sizeof (parentname));
3121 cp = strrchr(parentname, '/');
3122 ASSERT(cp != NULL);
3123 cp[0] = '\0';
3124
3125 if ((error = spa_open(dataset, &spa, FTAG)) != 0)
3126 return (error);
3127
3128 spa_vers = spa_version(spa);
3129 spa_close(spa, FTAG);
3130
3131 zplver = zfs_zpl_version_map(spa_vers);
3132 fuids_ok = (zplver >= ZPL_VERSION_FUID);
3133 sa_ok = (zplver >= ZPL_VERSION_SA);
3134
3135 /*
3136 * Open parent object set so we can inherit zplprop values.
3137 */
3138 if ((error = dmu_objset_hold(parentname, FTAG, &os)) != 0)
3139 return (error);
3140
3141 error = zfs_fill_zplprops_impl(os, zplver, fuids_ok, sa_ok, createprops,
3142 zplprops, is_ci);
3143 dmu_objset_rele(os, FTAG);
3144 return (error);
3145 }
3146
3147 static int
3148 zfs_fill_zplprops_root(uint64_t spa_vers, nvlist_t *createprops,
3149 nvlist_t *zplprops, boolean_t *is_ci)
3150 {
3151 boolean_t fuids_ok;
3152 boolean_t sa_ok;
3153 uint64_t zplver = ZPL_VERSION;
3154 int error;
3155
3156 zplver = zfs_zpl_version_map(spa_vers);
3157 fuids_ok = (zplver >= ZPL_VERSION_FUID);
3158 sa_ok = (zplver >= ZPL_VERSION_SA);
3159
3160 error = zfs_fill_zplprops_impl(NULL, zplver, fuids_ok, sa_ok,
3161 createprops, zplprops, is_ci);
3162 return (error);
3163 }
3164
3165 /*
3166 * innvl: {
3167 * "type" -> dmu_objset_type_t (int32)
3168 * (optional) "props" -> { prop -> value }
3169 * }
3170 *
3171 * outnvl: propname -> error code (int32)
3172 */
3173 static int
3174 zfs_ioc_create(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3175 {
3176 int error = 0;
3177 zfs_creat_t zct = { 0 };
3178 nvlist_t *nvprops = NULL;
3179 void (*cbfunc)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx);
3180 int32_t type32;
3181 dmu_objset_type_t type;
3182 boolean_t is_insensitive = B_FALSE;
3183
3184 if (nvlist_lookup_int32(innvl, "type", &type32) != 0)
3185 return (SET_ERROR(EINVAL));
3186 type = type32;
3187 (void) nvlist_lookup_nvlist(innvl, "props", &nvprops);
3188
3189 switch (type) {
3190 case DMU_OST_ZFS:
3191 cbfunc = zfs_create_cb;
3192 break;
3193
3194 case DMU_OST_ZVOL:
3195 cbfunc = zvol_create_cb;
3196 break;
3197
3198 default:
3199 cbfunc = NULL;
3200 break;
3201 }
3202 if (strchr(fsname, '@') ||
3203 strchr(fsname, '%'))
3204 return (SET_ERROR(EINVAL));
3205
3206 zct.zct_props = nvprops;
3207
3208 if (cbfunc == NULL)
3209 return (SET_ERROR(EINVAL));
3210
3211 if (type == DMU_OST_ZVOL) {
3212 uint64_t volsize, volblocksize;
3213
3214 if (nvprops == NULL)
3215 return (SET_ERROR(EINVAL));
3216 if (nvlist_lookup_uint64(nvprops,
3217 zfs_prop_to_name(ZFS_PROP_VOLSIZE), &volsize) != 0)
3218 return (SET_ERROR(EINVAL));
3219
3220 if ((error = nvlist_lookup_uint64(nvprops,
3221 zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE),
3222 &volblocksize)) != 0 && error != ENOENT)
3223 return (SET_ERROR(EINVAL));
3224
3225 if (error != 0)
3226 volblocksize = zfs_prop_default_numeric(
3227 ZFS_PROP_VOLBLOCKSIZE);
3228
3229 if ((error = zvol_check_volblocksize(
3230 volblocksize)) != 0 ||
3231 (error = zvol_check_volsize(volsize,
3232 volblocksize)) != 0)
3233 return (error);
3234 } else if (type == DMU_OST_ZFS) {
3235 int error;
3236
3237 /*
3238 * We have to have normalization and
3239 * case-folding flags correct when we do the
3240 * file system creation, so go figure them out
3241 * now.
3242 */
3243 VERIFY(nvlist_alloc(&zct.zct_zplprops,
3244 NV_UNIQUE_NAME, KM_SLEEP) == 0);
3245 error = zfs_fill_zplprops(fsname, nvprops,
3246 zct.zct_zplprops, &is_insensitive);
3247 if (error != 0) {
3248 nvlist_free(zct.zct_zplprops);
3249 return (error);
3250 }
3251 }
3252
3253 error = dmu_objset_create(fsname, type,
3254 is_insensitive ? DS_FLAG_CI_DATASET : 0, cbfunc, &zct);
3255 nvlist_free(zct.zct_zplprops);
3256
3257 /*
3258 * It would be nice to do this atomically.
3259 */
3260 if (error == 0) {
3261 error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL,
3262 nvprops, outnvl);
3263 if (error != 0)
3264 (void) dsl_destroy_head(fsname);
3265 }
3266 return (error);
3267 }
3268
3269 /*
3270 * innvl: {
3271 * "origin" -> name of origin snapshot
3272 * (optional) "props" -> { prop -> value }
3273 * }
3274 *
3275 * outnvl: propname -> error code (int32)
3276 */
3277 static int
3278 zfs_ioc_clone(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3279 {
3280 int error = 0;
3281 nvlist_t *nvprops = NULL;
3282 char *origin_name;
3283
3284 if (nvlist_lookup_string(innvl, "origin", &origin_name) != 0)
3285 return (SET_ERROR(EINVAL));
3286 (void) nvlist_lookup_nvlist(innvl, "props", &nvprops);
3287
3288 if (strchr(fsname, '@') ||
3289 strchr(fsname, '%'))
3290 return (SET_ERROR(EINVAL));
3291
3292 if (dataset_namecheck(origin_name, NULL, NULL) != 0)
3293 return (SET_ERROR(EINVAL));
3294 error = dmu_objset_clone(fsname, origin_name);
3295 if (error != 0)
3296 return (error);
3297
3298 /*
3299 * It would be nice to do this atomically.
3300 */
3301 if (error == 0) {
3302 error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL,
3303 nvprops, outnvl);
3304 if (error != 0)
3305 (void) dsl_destroy_head(fsname);
3306 }
3307 return (error);
3308 }
3309
3310 /*
3311 * innvl: {
3312 * "snaps" -> { snapshot1, snapshot2 }
3313 * (optional) "props" -> { prop -> value (string) }
3314 * }
3315 *
3316 * outnvl: snapshot -> error code (int32)
3317 */
3318 static int
3319 zfs_ioc_snapshot(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3320 {
3321 nvlist_t *snaps;
3322 nvlist_t *props = NULL;
3323 int error, poollen;
3324 nvpair_t *pair;
3325
3326 (void) nvlist_lookup_nvlist(innvl, "props", &props);
3327 if ((error = zfs_check_userprops(poolname, props)) != 0)
3328 return (error);
3329
3330 if (!nvlist_empty(props) &&
3331 zfs_earlier_version(poolname, SPA_VERSION_SNAP_PROPS))
3332 return (SET_ERROR(ENOTSUP));
3333
3334 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
3335 return (SET_ERROR(EINVAL));
3336 poollen = strlen(poolname);
3337 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
3338 pair = nvlist_next_nvpair(snaps, pair)) {
3339 const char *name = nvpair_name(pair);
3340 const char *cp = strchr(name, '@');
3341
3342 /*
3343 * The snap name must contain an @, and the part after it must
3344 * contain only valid characters.
3345 */
3346 if (cp == NULL ||
3347 zfs_component_namecheck(cp + 1, NULL, NULL) != 0)
3348 return (SET_ERROR(EINVAL));
3349
3350 /*
3351 * The snap must be in the specified pool.
3352 */
3353 if (strncmp(name, poolname, poollen) != 0 ||
3354 (name[poollen] != '/' && name[poollen] != '@'))
3355 return (SET_ERROR(EXDEV));
3356
3357 /* This must be the only snap of this fs. */
3358 for (nvpair_t *pair2 = nvlist_next_nvpair(snaps, pair);
3359 pair2 != NULL; pair2 = nvlist_next_nvpair(snaps, pair2)) {
3360 if (strncmp(name, nvpair_name(pair2), cp - name + 1)
3361 == 0) {
3362 return (SET_ERROR(EXDEV));
3363 }
3364 }
3365 }
3366
3367 error = dsl_dataset_snapshot(snaps, props, outnvl);
3368 return (error);
3369 }
3370
3371 /*
3372 * innvl: "message" -> string
3373 */
3374 /* ARGSUSED */
3375 static int
3376 zfs_ioc_log_history(const char *unused, nvlist_t *innvl, nvlist_t *outnvl)
3377 {
3378 char *message;
3379 spa_t *spa;
3380 int error;
3381 char *poolname;
3382
3383 /*
3384 * The poolname in the ioctl is not set, we get it from the TSD,
3385 * which was set at the end of the last successful ioctl that allows
3386 * logging. The secpolicy func already checked that it is set.
3387 * Only one log ioctl is allowed after each successful ioctl, so
3388 * we clear the TSD here.
3389 */
3390 poolname = tsd_get(zfs_allow_log_key);
3391 (void) tsd_set(zfs_allow_log_key, NULL);
3392 error = spa_open(poolname, &spa, FTAG);
3393 strfree(poolname);
3394 if (error != 0)
3395 return (error);
3396
3397 if (nvlist_lookup_string(innvl, "message", &message) != 0) {
3398 spa_close(spa, FTAG);
3399 return (SET_ERROR(EINVAL));
3400 }
3401
3402 if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) {
3403 spa_close(spa, FTAG);
3404 return (SET_ERROR(ENOTSUP));
3405 }
3406
3407 error = spa_history_log(spa, message);
3408 spa_close(spa, FTAG);
3409 return (error);
3410 }
3411
3412 /*
3413 * The dp_config_rwlock must not be held when calling this, because the
3414 * unmount may need to write out data.
3415 *
3416 * This function is best-effort. Callers must deal gracefully if it
3417 * remains mounted (or is remounted after this call).
3418 *
3419 * Returns 0 if the argument is not a snapshot, or it is not currently a
3420 * filesystem, or we were able to unmount it. Returns error code otherwise.
3421 */
3422 int
3423 zfs_unmount_snap(const char *snapname)
3424 {
3425 vfs_t *vfsp;
3426 zfsvfs_t *zfsvfs;
3427 int err;
3428
3429 if (strchr(snapname, '@') == NULL)
3430 return (0);
3431
3432 vfsp = zfs_get_vfs(snapname);
3433 if (vfsp == NULL)
3434 return (0);
3435
3436 zfsvfs = vfsp->vfs_data;
3437 ASSERT(!dsl_pool_config_held(dmu_objset_pool(zfsvfs->z_os)));
3438
3439 err = vn_vfswlock(vfsp->vfs_vnodecovered);
3440 VFS_RELE(vfsp);
3441 if (err != 0)
3442 return (SET_ERROR(err));
3443
3444 /*
3445 * Always force the unmount for snapshots.
3446 */
3447 (void) dounmount(vfsp, MS_FORCE, kcred);
3448 return (0);
3449 }
3450
3451 /* ARGSUSED */
3452 static int
3453 zfs_unmount_snap_cb(const char *snapname, void *arg)
3454 {
3455 return (zfs_unmount_snap(snapname));
3456 }
3457
3458 /*
3459 * When a clone is destroyed, its origin may also need to be destroyed,
3460 * in which case it must be unmounted. This routine will do that unmount
3461 * if necessary.
3462 */
3463 void
3464 zfs_destroy_unmount_origin(const char *fsname)
3465 {
3466 int error;
3467 objset_t *os;
3468 dsl_dataset_t *ds;
3469
3470 error = dmu_objset_hold(fsname, FTAG, &os);
3471 if (error != 0)
3472 return;
3473 ds = dmu_objset_ds(os);
3474 if (dsl_dir_is_clone(ds->ds_dir) && DS_IS_DEFER_DESTROY(ds->ds_prev)) {
3475 char originname[ZFS_MAX_DATASET_NAME_LEN];
3476 dsl_dataset_name(ds->ds_prev, originname);
3477 dmu_objset_rele(os, FTAG);
3478 (void) zfs_unmount_snap(originname);
3479 } else {
3480 dmu_objset_rele(os, FTAG);
3481 }
3482 }
3483
3484 /*
3485 * innvl: {
3486 * "snaps" -> { snapshot1, snapshot2 }
3487 * (optional boolean) "defer"
3488 * }
3489 *
3490 * outnvl: snapshot -> error code (int32)
3491 *
3492 */
3493 /* ARGSUSED */
3494 static int
3495 zfs_ioc_destroy_snaps(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3496 {
3497 nvlist_t *snaps;
3498 nvpair_t *pair;
3499 boolean_t defer;
3500
3501 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
3502 return (SET_ERROR(EINVAL));
3503 defer = nvlist_exists(innvl, "defer");
3504
3505 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
3506 pair = nvlist_next_nvpair(snaps, pair)) {
3507 (void) zfs_unmount_snap(nvpair_name(pair));
3508 }
3509
3510 return (dsl_destroy_snapshots_nvl(snaps, defer, outnvl));
3511 }
3512
3513 /*
3514 * Create bookmarks. Bookmark names are of the form <fs>#<bmark>.
3515 * All bookmarks must be in the same pool.
3516 *
3517 * innvl: {
3518 * bookmark1 -> snapshot1, bookmark2 -> snapshot2
3519 * }
3520 *
3521 * outnvl: bookmark -> error code (int32)
3522 *
3523 */
3524 /* ARGSUSED */
3525 static int
3526 zfs_ioc_bookmark(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3527 {
3528 for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL);
3529 pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) {
3530 char *snap_name;
3531
3532 /*
3533 * Verify the snapshot argument.
3534 */
3535 if (nvpair_value_string(pair, &snap_name) != 0)
3536 return (SET_ERROR(EINVAL));
3537
3538
3539 /* Verify that the keys (bookmarks) are unique */
3540 for (nvpair_t *pair2 = nvlist_next_nvpair(innvl, pair);
3541 pair2 != NULL; pair2 = nvlist_next_nvpair(innvl, pair2)) {
3542 if (strcmp(nvpair_name(pair), nvpair_name(pair2)) == 0)
3543 return (SET_ERROR(EINVAL));
3544 }
3545 }
3546
3547 return (dsl_bookmark_create(innvl, outnvl));
3548 }
3549
3550 /*
3551 * innvl: {
3552 * property 1, property 2, ...
3553 * }
3554 *
3555 * outnvl: {
3556 * bookmark name 1 -> { property 1, property 2, ... },
3557 * bookmark name 2 -> { property 1, property 2, ... }
3558 * }
3559 *
3560 */
3561 static int
3562 zfs_ioc_get_bookmarks(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3563 {
3564 return (dsl_get_bookmarks(fsname, innvl, outnvl));
3565 }
3566
3567 /*
3568 * innvl: {
3569 * bookmark name 1, bookmark name 2
3570 * }
3571 *
3572 * outnvl: bookmark -> error code (int32)
3573 *
3574 */
3575 static int
3576 zfs_ioc_destroy_bookmarks(const char *poolname, nvlist_t *innvl,
3577 nvlist_t *outnvl)
3578 {
3579 int error, poollen;
3580
3581 poollen = strlen(poolname);
3582 for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL);
3583 pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) {
3584 const char *name = nvpair_name(pair);
3585 const char *cp = strchr(name, '#');
3586
3587 /*
3588 * The bookmark name must contain an #, and the part after it
3589 * must contain only valid characters.
3590 */
3591 if (cp == NULL ||
3592 zfs_component_namecheck(cp + 1, NULL, NULL) != 0)
3593 return (SET_ERROR(EINVAL));
3594
3595 /*
3596 * The bookmark must be in the specified pool.
3597 */
3598 if (strncmp(name, poolname, poollen) != 0 ||
3599 (name[poollen] != '/' && name[poollen] != '#'))
3600 return (SET_ERROR(EXDEV));
3601 }
3602
3603 error = dsl_bookmark_destroy(innvl, outnvl);
3604 return (error);
3605 }
3606
3607 /*
3608 * inputs:
3609 * zc_name name of dataset to destroy
3610 * zc_objset_type type of objset
3611 * zc_defer_destroy mark for deferred destroy
3612 *
3613 * outputs: none
3614 */
3615 static int
3616 zfs_ioc_destroy(zfs_cmd_t *zc)
3617 {
3618 int err;
3619
3620 if (zc->zc_objset_type == DMU_OST_ZFS) {
3621 err = zfs_unmount_snap(zc->zc_name);
3622 if (err != 0)
3623 return (err);
3624 }
3625
3626 if (strchr(zc->zc_name, '@'))
3627 err = dsl_destroy_snapshot(zc->zc_name, zc->zc_defer_destroy);
3628 else
3629 err = dsl_destroy_head(zc->zc_name);
3630 if (zc->zc_objset_type == DMU_OST_ZVOL && err == 0)
3631 (void) zvol_remove_minor(zc->zc_name);
3632 return (err);
3633 }
3634
3635 /*
3636 * fsname is name of dataset to rollback (to most recent snapshot)
3637 *
3638 * innvl is not used.
3639 *
3640 * outnvl: "target" -> name of most recent snapshot
3641 * }
3642 */
3643 /* ARGSUSED */
3644 static int
3645 zfs_ioc_rollback(const char *fsname, nvlist_t *args, nvlist_t *outnvl)
3646 {
3647 zfsvfs_t *zfsvfs;
3648 int error;
3649
3650 if (getzfsvfs(fsname, &zfsvfs) == 0) {
3651 dsl_dataset_t *ds;
3652
3653 ds = dmu_objset_ds(zfsvfs->z_os);
3654 error = zfs_suspend_fs(zfsvfs);
3655 if (error == 0) {
3656 int resume_err;
3657
3658 error = dsl_dataset_rollback(fsname, zfsvfs, outnvl);
3659 resume_err = zfs_resume_fs(zfsvfs, ds);
3660 error = error ? error : resume_err;
3661 }
3662 VFS_RELE(zfsvfs->z_vfs);
3663 } else {
3664 error = dsl_dataset_rollback(fsname, NULL, outnvl);
3665 }
3666 return (error);
3667 }
3668
3669 static int
3670 recursive_unmount(const char *fsname, void *arg)
3671 {
3672 const char *snapname = arg;
3673 char fullname[ZFS_MAX_DATASET_NAME_LEN];
3674
3675 (void) snprintf(fullname, sizeof (fullname), "%s@%s", fsname, snapname);
3676 return (zfs_unmount_snap(fullname));
3677 }
3678
3679 /*
3680 * inputs:
3681 * zc_name old name of dataset
3682 * zc_value new name of dataset
3683 * zc_cookie recursive flag (only valid for snapshots)
3684 *
3685 * outputs: none
3686 */
3687 static int
3688 zfs_ioc_rename(zfs_cmd_t *zc)
3689 {
3690 boolean_t recursive = zc->zc_cookie & 1;
3691 char *at;
3692
3693 zc->zc_value[sizeof (zc->zc_value) - 1] = '\0';
3694 if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
3695 strchr(zc->zc_value, '%'))
3696 return (SET_ERROR(EINVAL));
3697
3698 at = strchr(zc->zc_name, '@');
3699 if (at != NULL) {
3700 /* snaps must be in same fs */
3701 int error;
3702
3703 if (strncmp(zc->zc_name, zc->zc_value, at - zc->zc_name + 1))
3704 return (SET_ERROR(EXDEV));
3705 *at = '\0';
3706 if (zc->zc_objset_type == DMU_OST_ZFS) {
3707 error = dmu_objset_find(zc->zc_name,
3708 recursive_unmount, at + 1,
3709 recursive ? DS_FIND_CHILDREN : 0);
3710 if (error != 0) {
3711 *at = '@';
3712 return (error);
3713 }
3714 }
3715 error = dsl_dataset_rename_snapshot(zc->zc_name,
3716 at + 1, strchr(zc->zc_value, '@') + 1, recursive);
3717 *at = '@';
3718
3719 return (error);
3720 } else {
3721 if (zc->zc_objset_type == DMU_OST_ZVOL)
3722 (void) zvol_remove_minor(zc->zc_name);
3723 return (dsl_dir_rename(zc->zc_name, zc->zc_value));
3724 }
3725 }
3726
3727 static int
3728 zfs_check_settable(const char *dsname, nvpair_t *pair, cred_t *cr)
3729 {
3730 const char *propname = nvpair_name(pair);
3731 boolean_t issnap = (strchr(dsname, '@') != NULL);
3732 zfs_prop_t prop = zfs_name_to_prop(propname);
3733 uint64_t intval;
3734 int err;
3735
3736 if (prop == ZPROP_INVAL) {
3737 if (zfs_prop_user(propname)) {
3738 if (err = zfs_secpolicy_write_perms(dsname,
3739 ZFS_DELEG_PERM_USERPROP, cr))
3740 return (err);
3741 return (0);
3742 }
3743
3744 if (!issnap && zfs_prop_userquota(propname)) {
3745 const char *perm = NULL;
3746 const char *uq_prefix =
3747 zfs_userquota_prop_prefixes[ZFS_PROP_USERQUOTA];
3748 const char *gq_prefix =
3749 zfs_userquota_prop_prefixes[ZFS_PROP_GROUPQUOTA];
3750
3751 if (strncmp(propname, uq_prefix,
3752 strlen(uq_prefix)) == 0) {
3753 perm = ZFS_DELEG_PERM_USERQUOTA;
3754 } else if (strncmp(propname, gq_prefix,
3755 strlen(gq_prefix)) == 0) {
3756 perm = ZFS_DELEG_PERM_GROUPQUOTA;
3757 } else {
3758 /* USERUSED and GROUPUSED are read-only */
3759 return (SET_ERROR(EINVAL));
3760 }
3761
3762 if (err = zfs_secpolicy_write_perms(dsname, perm, cr))
3763 return (err);
3764 return (0);
3765 }
3766
3767 return (SET_ERROR(EINVAL));
3768 }
3769
3770 if (issnap)
3771 return (SET_ERROR(EINVAL));
3772
3773 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
3774 /*
3775 * dsl_prop_get_all_impl() returns properties in this
3776 * format.
3777 */
3778 nvlist_t *attrs;
3779 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
3780 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
3781 &pair) == 0);
3782 }
3783
3784 /*
3785 * Check that this value is valid for this pool version
3786 */
3787 switch (prop) {
3788 case ZFS_PROP_COMPRESSION:
3789 /*
3790 * If the user specified gzip compression, make sure
3791 * the SPA supports it. We ignore any errors here since
3792 * we'll catch them later.
3793 */
3794 if (nvpair_value_uint64(pair, &intval) == 0) {
3795 if (intval >= ZIO_COMPRESS_GZIP_1 &&
3796 intval <= ZIO_COMPRESS_GZIP_9 &&
3797 zfs_earlier_version(dsname,
3798 SPA_VERSION_GZIP_COMPRESSION)) {
3799 return (SET_ERROR(ENOTSUP));
3800 }
3801
3802 if (intval == ZIO_COMPRESS_ZLE &&
3803 zfs_earlier_version(dsname,
3804 SPA_VERSION_ZLE_COMPRESSION))
3805 return (SET_ERROR(ENOTSUP));
3806
3807 if (intval == ZIO_COMPRESS_LZ4) {
3808 spa_t *spa;
3809
3810 if ((err = spa_open(dsname, &spa, FTAG)) != 0)
3811 return (err);
3812
3813 if (!spa_feature_is_enabled(spa,
3814 SPA_FEATURE_LZ4_COMPRESS)) {
3815 spa_close(spa, FTAG);
3816 return (SET_ERROR(ENOTSUP));
3817 }
3818 spa_close(spa, FTAG);
3819 }
3820
3821 /*
3822 * If this is a bootable dataset then
3823 * verify that the compression algorithm
3824 * is supported for booting. We must return
3825 * something other than ENOTSUP since it
3826 * implies a downrev pool version.
3827 */
3828 if (zfs_is_bootfs(dsname) &&
3829 !BOOTFS_COMPRESS_VALID(intval)) {
3830 return (SET_ERROR(ERANGE));
3831 }
3832 }
3833 break;
3834
3835 case ZFS_PROP_COPIES:
3836 if (zfs_earlier_version(dsname, SPA_VERSION_DITTO_BLOCKS))
3837 return (SET_ERROR(ENOTSUP));
3838 break;
3839
3840 case ZFS_PROP_RECORDSIZE:
3841 /* Record sizes above 128k need the feature to be enabled */
3842 if (nvpair_value_uint64(pair, &intval) == 0 &&
3843 intval > SPA_OLD_MAXBLOCKSIZE) {
3844 spa_t *spa;
3845
3846 /*
3847 * We don't allow setting the property above 1MB,
3848 * unless the tunable has been changed.
3849 */
3850 if (intval > zfs_max_recordsize ||
3851 intval > SPA_MAXBLOCKSIZE)
3852 return (SET_ERROR(ERANGE));
3853
3854 if ((err = spa_open(dsname, &spa, FTAG)) != 0)
3855 return (err);
3856
3857 if (!spa_feature_is_enabled(spa,
3858 SPA_FEATURE_LARGE_BLOCKS)) {
3859 spa_close(spa, FTAG);
3860 return (SET_ERROR(ENOTSUP));
3861 }
3862 spa_close(spa, FTAG);
3863 }
3864 break;
3865
3866 case ZFS_PROP_SHARESMB:
3867 if (zpl_earlier_version(dsname, ZPL_VERSION_FUID))
3868 return (SET_ERROR(ENOTSUP));
3869 break;
3870
3871 case ZFS_PROP_ACLINHERIT:
3872 if (nvpair_type(pair) == DATA_TYPE_UINT64 &&
3873 nvpair_value_uint64(pair, &intval) == 0) {
3874 if (intval == ZFS_ACL_PASSTHROUGH_X &&
3875 zfs_earlier_version(dsname,
3876 SPA_VERSION_PASSTHROUGH_X))
3877 return (SET_ERROR(ENOTSUP));
3878 }
3879 break;
3880
3881 case ZFS_PROP_CHECKSUM:
3882 case ZFS_PROP_DEDUP:
3883 {
3884 spa_feature_t feature;
3885 spa_t *spa;
3886
3887 /* dedup feature version checks */
3888 if (prop == ZFS_PROP_DEDUP &&
3889 zfs_earlier_version(dsname, SPA_VERSION_DEDUP))
3890 return (SET_ERROR(ENOTSUP));
3891
3892 if (nvpair_value_uint64(pair, &intval) != 0)
3893 return (SET_ERROR(EINVAL));
3894
3895 /* check prop value is enabled in features */
3896 feature = zio_checksum_to_feature(intval & ZIO_CHECKSUM_MASK);
3897 if (feature == SPA_FEATURE_NONE)
3898 break;
3899
3900 if ((err = spa_open(dsname, &spa, FTAG)) != 0)
3901 return (err);
3902 /*
3903 * Salted checksums are not supported on root pools.
3904 */
3905 if (spa_bootfs(spa) != 0 &&
3906 intval < ZIO_CHECKSUM_FUNCTIONS &&
3907 (zio_checksum_table[intval].ci_flags &
3908 ZCHECKSUM_FLAG_SALTED)) {
3909 spa_close(spa, FTAG);
3910 return (SET_ERROR(ERANGE));
3911 }
3912 if (!spa_feature_is_enabled(spa, feature)) {
3913 spa_close(spa, FTAG);
3914 return (SET_ERROR(ENOTSUP));
3915 }
3916 spa_close(spa, FTAG);
3917 break;
3918 }
3919 }
3920
3921 return (zfs_secpolicy_setprop(dsname, prop, pair, CRED()));
3922 }
3923
3924 /*
3925 * Checks for a race condition to make sure we don't increment a feature flag
3926 * multiple times.
3927 */
3928 static int
3929 zfs_prop_activate_feature_check(void *arg, dmu_tx_t *tx)
3930 {
3931 spa_t *spa = dmu_tx_pool(tx)->dp_spa;
3932 spa_feature_t *featurep = arg;
3933
3934 if (!spa_feature_is_active(spa, *featurep))
3935 return (0);
3936 else
3937 return (SET_ERROR(EBUSY));
3938 }
3939
3940 /*
3941 * The callback invoked on feature activation in the sync task caused by
3942 * zfs_prop_activate_feature.
3943 */
3944 static void
3945 zfs_prop_activate_feature_sync(void *arg, dmu_tx_t *tx)
3946 {
3947 spa_t *spa = dmu_tx_pool(tx)->dp_spa;
3948 spa_feature_t *featurep = arg;
3949
3950 spa_feature_incr(spa, *featurep, tx);
3951 }
3952
3953 /*
3954 * Activates a feature on a pool in response to a property setting. This
3955 * creates a new sync task which modifies the pool to reflect the feature
3956 * as being active.
3957 */
3958 static int
3959 zfs_prop_activate_feature(spa_t *spa, spa_feature_t feature)
3960 {
3961 int err;
3962
3963 /* EBUSY here indicates that the feature is already active */
3964 err = dsl_sync_task(spa_name(spa),
3965 zfs_prop_activate_feature_check, zfs_prop_activate_feature_sync,
3966 &feature, 2, ZFS_SPACE_CHECK_RESERVED);
3967
3968 if (err != 0 && err != EBUSY)
3969 return (err);
3970 else
3971 return (0);
3972 }
3973
3974 /*
3975 * Removes properties from the given props list that fail permission checks
3976 * needed to clear them and to restore them in case of a receive error. For each
3977 * property, make sure we have both set and inherit permissions.
3978 *
3979 * Returns the first error encountered if any permission checks fail. If the
3980 * caller provides a non-NULL errlist, it also gives the complete list of names
3981 * of all the properties that failed a permission check along with the
3982 * corresponding error numbers. The caller is responsible for freeing the
3983 * returned errlist.
3984 *
3985 * If every property checks out successfully, zero is returned and the list
3986 * pointed at by errlist is NULL.
3987 */
3988 static int
3989 zfs_check_clearable(char *dataset, nvlist_t *props, nvlist_t **errlist)
3990 {
3991 zfs_cmd_t *zc;
3992 nvpair_t *pair, *next_pair;
3993 nvlist_t *errors;
3994 int err, rv = 0;
3995
3996 if (props == NULL)
3997 return (0);
3998
3999 VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0);
4000
4001 zc = kmem_alloc(sizeof (zfs_cmd_t), KM_SLEEP);
4002 (void) strcpy(zc->zc_name, dataset);
4003 pair = nvlist_next_nvpair(props, NULL);
4004 while (pair != NULL) {
4005 next_pair = nvlist_next_nvpair(props, pair);
4006
4007 (void) strcpy(zc->zc_value, nvpair_name(pair));
4008 if ((err = zfs_check_settable(dataset, pair, CRED())) != 0 ||
4009 (err = zfs_secpolicy_inherit_prop(zc, NULL, CRED())) != 0) {
4010 VERIFY(nvlist_remove_nvpair(props, pair) == 0);
4011 VERIFY(nvlist_add_int32(errors,
4012 zc->zc_value, err) == 0);
4013 }
4014 pair = next_pair;
4015 }
4016 kmem_free(zc, sizeof (zfs_cmd_t));
4017
4018 if ((pair = nvlist_next_nvpair(errors, NULL)) == NULL) {
4019 nvlist_free(errors);
4020 errors = NULL;
4021 } else {
4022 VERIFY(nvpair_value_int32(pair, &rv) == 0);
4023 }
4024
4025 if (errlist == NULL)
4026 nvlist_free(errors);
4027 else
4028 *errlist = errors;
4029
4030 return (rv);
4031 }
4032
4033 static boolean_t
4034 propval_equals(nvpair_t *p1, nvpair_t *p2)
4035 {
4036 if (nvpair_type(p1) == DATA_TYPE_NVLIST) {
4037 /* dsl_prop_get_all_impl() format */
4038 nvlist_t *attrs;
4039 VERIFY(nvpair_value_nvlist(p1, &attrs) == 0);
4040 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
4041 &p1) == 0);
4042 }
4043
4044 if (nvpair_type(p2) == DATA_TYPE_NVLIST) {
4045 nvlist_t *attrs;
4046 VERIFY(nvpair_value_nvlist(p2, &attrs) == 0);
4047 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
4048 &p2) == 0);
4049 }
4050
4051 if (nvpair_type(p1) != nvpair_type(p2))
4052 return (B_FALSE);
4053
4054 if (nvpair_type(p1) == DATA_TYPE_STRING) {
4055 char *valstr1, *valstr2;
4056
4057 VERIFY(nvpair_value_string(p1, (char **)&valstr1) == 0);
4058 VERIFY(nvpair_value_string(p2, (char **)&valstr2) == 0);
4059 return (strcmp(valstr1, valstr2) == 0);
4060 } else {
4061 uint64_t intval1, intval2;
4062
4063 VERIFY(nvpair_value_uint64(p1, &intval1) == 0);
4064 VERIFY(nvpair_value_uint64(p2, &intval2) == 0);
4065 return (intval1 == intval2);
4066 }
4067 }
4068
4069 /*
4070 * Remove properties from props if they are not going to change (as determined
4071 * by comparison with origprops). Remove them from origprops as well, since we
4072 * do not need to clear or restore properties that won't change.
4073 */
4074 static void
4075 props_reduce(nvlist_t *props, nvlist_t *origprops)
4076 {
4077 nvpair_t *pair, *next_pair;
4078
4079 if (origprops == NULL)
4080 return; /* all props need to be received */
4081
4082 pair = nvlist_next_nvpair(props, NULL);
4083 while (pair != NULL) {
4084 const char *propname = nvpair_name(pair);
4085 nvpair_t *match;
4086
4087 next_pair = nvlist_next_nvpair(props, pair);
4088
4089 if ((nvlist_lookup_nvpair(origprops, propname,
4090 &match) != 0) || !propval_equals(pair, match))
4091 goto next; /* need to set received value */
4092
4093 /* don't clear the existing received value */
4094 (void) nvlist_remove_nvpair(origprops, match);
4095 /* don't bother receiving the property */
4096 (void) nvlist_remove_nvpair(props, pair);
4097 next:
4098 pair = next_pair;
4099 }
4100 }
4101
4102 /*
4103 * Extract properties that cannot be set PRIOR to the receipt of a dataset.
4104 * For example, refquota cannot be set until after the receipt of a dataset,
4105 * because in replication streams, an older/earlier snapshot may exceed the
4106 * refquota. We want to receive the older/earlier snapshot, but setting
4107 * refquota pre-receipt will set the dsl's ACTUAL quota, which will prevent
4108 * the older/earlier snapshot from being received (with EDQUOT).
4109 *
4110 * The ZFS test "zfs_receive_011_pos" demonstrates such a scenario.
4111 *
4112 * libzfs will need to be judicious handling errors encountered by props
4113 * extracted by this function.
4114 */
4115 static nvlist_t *
4116 extract_delay_props(nvlist_t *props)
4117 {
4118 nvlist_t *delayprops;
4119 nvpair_t *nvp, *tmp;
4120 static const zfs_prop_t delayable[] = { ZFS_PROP_REFQUOTA, 0 };
4121 int i;
4122
4123 VERIFY(nvlist_alloc(&delayprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
4124
4125 for (nvp = nvlist_next_nvpair(props, NULL); nvp != NULL;
4126 nvp = nvlist_next_nvpair(props, nvp)) {
4127 /*
4128 * strcmp() is safe because zfs_prop_to_name() always returns
4129 * a bounded string.
4130 */
4131 for (i = 0; delayable[i] != 0; i++) {
4132 if (strcmp(zfs_prop_to_name(delayable[i]),
4133 nvpair_name(nvp)) == 0) {
4134 break;
4135 }
4136 }
4137 if (delayable[i] != 0) {
4138 tmp = nvlist_prev_nvpair(props, nvp);
4139 VERIFY(nvlist_add_nvpair(delayprops, nvp) == 0);
4140 VERIFY(nvlist_remove_nvpair(props, nvp) == 0);
4141 nvp = tmp;
4142 }
4143 }
4144
4145 if (nvlist_empty(delayprops)) {
4146 nvlist_free(delayprops);
4147 delayprops = NULL;
4148 }
4149 return (delayprops);
4150 }
4151
4152 #ifdef DEBUG
4153 static boolean_t zfs_ioc_recv_inject_err;
4154 #endif
4155
4156 /*
4157 * inputs:
4158 * zc_name name of containing filesystem
4159 * zc_nvlist_src{_size} nvlist of properties to apply
4160 * zc_value name of snapshot to create
4161 * zc_string name of clone origin (if DRR_FLAG_CLONE)
4162 * zc_cookie file descriptor to recv from
4163 * zc_begin_record the BEGIN record of the stream (not byteswapped)
4164 * zc_guid force flag
4165 * zc_cleanup_fd cleanup-on-exit file descriptor
4166 * zc_action_handle handle for this guid/ds mapping (or zero on first call)
4167 * zc_resumable if data is incomplete assume sender will resume
4168 *
4169 * outputs:
4170 * zc_cookie number of bytes read
4171 * zc_nvlist_dst{_size} error for each unapplied received property
4172 * zc_obj zprop_errflags_t
4173 * zc_action_handle handle for this guid/ds mapping
4174 */
4175 static int
4176 zfs_ioc_recv(zfs_cmd_t *zc)
4177 {
4178 file_t *fp;
4179 dmu_recv_cookie_t drc;
4180 boolean_t force = (boolean_t)zc->zc_guid;
4181 int fd;
4182 int error = 0;
4183 int props_error = 0;
4184 nvlist_t *errors;
4185 offset_t off;
4186 nvlist_t *props = NULL; /* sent properties */
4187 nvlist_t *origprops = NULL; /* existing properties */
4188 nvlist_t *delayprops = NULL; /* sent properties applied post-receive */
4189 char *origin = NULL;
4190 char *tosnap;
4191 char tofs[ZFS_MAX_DATASET_NAME_LEN];
4192 boolean_t first_recvd_props = B_FALSE;
4193
4194 if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
4195 strchr(zc->zc_value, '@') == NULL ||
4196 strchr(zc->zc_value, '%'))
4197 return (SET_ERROR(EINVAL));
4198
4199 (void) strcpy(tofs, zc->zc_value);
4200 tosnap = strchr(tofs, '@');
4201 *tosnap++ = '\0';
4202
4203 if (zc->zc_nvlist_src != NULL &&
4204 (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
4205 zc->zc_iflags, &props)) != 0)
4206 return (error);
4207
4208 fd = zc->zc_cookie;
4209 fp = getf(fd);
4210 if (fp == NULL) {
4211 nvlist_free(props);
4212 return (SET_ERROR(EBADF));
4213 }
4214
4215 errors = fnvlist_alloc();
4216
4217 if (zc->zc_string[0])
4218 origin = zc->zc_string;
4219
4220 error = dmu_recv_begin(tofs, tosnap,
4221 &zc->zc_begin_record, force, zc->zc_resumable, origin, &drc);
4222 if (error != 0)
4223 goto out;
4224
4225 /*
4226 * Set properties before we receive the stream so that they are applied
4227 * to the new data. Note that we must call dmu_recv_stream() if
4228 * dmu_recv_begin() succeeds.
4229 */
4230 if (props != NULL && !drc.drc_newfs) {
4231 if (spa_version(dsl_dataset_get_spa(drc.drc_ds)) >=
4232 SPA_VERSION_RECVD_PROPS &&
4233 !dsl_prop_get_hasrecvd(tofs))
4234 first_recvd_props = B_TRUE;
4235
4236 /*
4237 * If new received properties are supplied, they are to
4238 * completely replace the existing received properties, so stash
4239 * away the existing ones.
4240 */
4241 if (dsl_prop_get_received(tofs, &origprops) == 0) {
4242 nvlist_t *errlist = NULL;
4243 /*
4244 * Don't bother writing a property if its value won't
4245 * change (and avoid the unnecessary security checks).
4246 *
4247 * The first receive after SPA_VERSION_RECVD_PROPS is a
4248 * special case where we blow away all local properties
4249 * regardless.
4250 */
4251 if (!first_recvd_props)
4252 props_reduce(props, origprops);
4253 if (zfs_check_clearable(tofs, origprops, &errlist) != 0)
4254 (void) nvlist_merge(errors, errlist, 0);
4255 nvlist_free(errlist);
4256
4257 if (clear_received_props(tofs, origprops,
4258 first_recvd_props ? NULL : props) != 0)
4259 zc->zc_obj |= ZPROP_ERR_NOCLEAR;
4260 } else {
4261 zc->zc_obj |= ZPROP_ERR_NOCLEAR;
4262 }
4263 }
4264
4265 if (props != NULL) {
4266 props_error = dsl_prop_set_hasrecvd(tofs);
4267
4268 if (props_error == 0) {
4269 delayprops = extract_delay_props(props);
4270 (void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_RECEIVED,
4271 props, errors);
4272 }
4273 }
4274
4275 off = fp->f_offset;
4276 error = dmu_recv_stream(&drc, fp->f_vnode, &off, zc->zc_cleanup_fd,
4277 &zc->zc_action_handle);
4278
4279 if (error == 0) {
4280 zfsvfs_t *zfsvfs = NULL;
4281
4282 if (getzfsvfs(tofs, &zfsvfs) == 0) {
4283 /* online recv */
4284 dsl_dataset_t *ds;
4285 int end_err;
4286
4287 ds = dmu_objset_ds(zfsvfs->z_os);
4288 error = zfs_suspend_fs(zfsvfs);
4289 /*
4290 * If the suspend fails, then the recv_end will
4291 * likely also fail, and clean up after itself.
4292 */
4293 end_err = dmu_recv_end(&drc, zfsvfs);
4294 if (error == 0)
4295 error = zfs_resume_fs(zfsvfs, ds);
4296 error = error ? error : end_err;
4297 VFS_RELE(zfsvfs->z_vfs);
4298 } else {
4299 error = dmu_recv_end(&drc, NULL);
4300 }
4301
4302 /* Set delayed properties now, after we're done receiving. */
4303 if (delayprops != NULL && error == 0) {
4304 (void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_RECEIVED,
4305 delayprops, errors);
4306 }
4307 }
4308
4309 if (delayprops != NULL) {
4310 /*
4311 * Merge delayed props back in with initial props, in case
4312 * we're DEBUG and zfs_ioc_recv_inject_err is set (which means
4313 * we have to make sure clear_received_props() includes
4314 * the delayed properties).
4315 *
4316 * Since zfs_ioc_recv_inject_err is only in DEBUG kernels,
4317 * using ASSERT() will be just like a VERIFY.
4318 */
4319 ASSERT(nvlist_merge(props, delayprops, 0) == 0);
4320 nvlist_free(delayprops);
4321 }
4322
4323 /*
4324 * Now that all props, initial and delayed, are set, report the prop
4325 * errors to the caller.
4326 */
4327 if (zc->zc_nvlist_dst_size != 0 &&
4328 (nvlist_smush(errors, zc->zc_nvlist_dst_size) != 0 ||
4329 put_nvlist(zc, errors) != 0)) {
4330 /*
4331 * Caller made zc->zc_nvlist_dst less than the minimum expected
4332 * size or supplied an invalid address.
4333 */
4334 props_error = SET_ERROR(EINVAL);
4335 }
4336
4337 zc->zc_cookie = off - fp->f_offset;
4338 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
4339 fp->f_offset = off;
4340
4341 #ifdef DEBUG
4342 if (zfs_ioc_recv_inject_err) {
4343 zfs_ioc_recv_inject_err = B_FALSE;
4344 error = 1;
4345 }
4346 #endif
4347 /*
4348 * On error, restore the original props.
4349 */
4350 if (error != 0 && props != NULL && !drc.drc_newfs) {
4351 if (clear_received_props(tofs, props, NULL) != 0) {
4352 /*
4353 * We failed to clear the received properties.
4354 * Since we may have left a $recvd value on the
4355 * system, we can't clear the $hasrecvd flag.
4356 */
4357 zc->zc_obj |= ZPROP_ERR_NORESTORE;
4358 } else if (first_recvd_props) {
4359 dsl_prop_unset_hasrecvd(tofs);
4360 }
4361
4362 if (origprops == NULL && !drc.drc_newfs) {
4363 /* We failed to stash the original properties. */
4364 zc->zc_obj |= ZPROP_ERR_NORESTORE;
4365 }
4366
4367 /*
4368 * dsl_props_set() will not convert RECEIVED to LOCAL on or
4369 * after SPA_VERSION_RECVD_PROPS, so we need to specify LOCAL
4370 * explictly if we're restoring local properties cleared in the
4371 * first new-style receive.
4372 */
4373 if (origprops != NULL &&
4374 zfs_set_prop_nvlist(tofs, (first_recvd_props ?
4375 ZPROP_SRC_LOCAL : ZPROP_SRC_RECEIVED),
4376 origprops, NULL) != 0) {
4377 /*
4378 * We stashed the original properties but failed to
4379 * restore them.
4380 */
4381 zc->zc_obj |= ZPROP_ERR_NORESTORE;
4382 }
4383 }
4384 out:
4385 nvlist_free(props);
4386 nvlist_free(origprops);
4387 nvlist_free(errors);
4388 releasef(fd);
4389
4390 if (error == 0)
4391 error = props_error;
4392
4393 return (error);
4394 }
4395
4396 /*
4397 * inputs:
4398 * zc_name name of snapshot to send
4399 * zc_cookie file descriptor to send stream to
4400 * zc_obj fromorigin flag (mutually exclusive with zc_fromobj)
4401 * zc_sendobj objsetid of snapshot to send
4402 * zc_fromobj objsetid of incremental fromsnap (may be zero)
4403 * zc_guid if set, estimate size of stream only. zc_cookie is ignored.
4404 * output size in zc_objset_type.
4405 * zc_flags lzc_send_flags
4406 *
4407 * outputs:
4408 * zc_objset_type estimated size, if zc_guid is set
4409 */
4410 static int
4411 zfs_ioc_send(zfs_cmd_t *zc)
4412 {
4413 int error;
4414 offset_t off;
4415 boolean_t estimate = (zc->zc_guid != 0);
4416 boolean_t embedok = (zc->zc_flags & 0x1);
4417 boolean_t large_block_ok = (zc->zc_flags & 0x2);
4418 boolean_t compressok = (zc->zc_flags & 0x4);
4419
4420 if (zc->zc_obj != 0) {
4421 dsl_pool_t *dp;
4422 dsl_dataset_t *tosnap;
4423
4424 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
4425 if (error != 0)
4426 return (error);
4427
4428 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &tosnap);
4429 if (error != 0) {
4430 dsl_pool_rele(dp, FTAG);
4431 return (error);
4432 }
4433
4434 if (dsl_dir_is_clone(tosnap->ds_dir))
4435 zc->zc_fromobj =
4436 dsl_dir_phys(tosnap->ds_dir)->dd_origin_obj;
4437 dsl_dataset_rele(tosnap, FTAG);
4438 dsl_pool_rele(dp, FTAG);
4439 }
4440
4441 if (estimate) {
4442 dsl_pool_t *dp;
4443 dsl_dataset_t *tosnap;
4444 dsl_dataset_t *fromsnap = NULL;
4445
4446 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
4447 if (error != 0)
4448 return (error);
4449
4450 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &tosnap);
4451 if (error != 0) {
4452 dsl_pool_rele(dp, FTAG);
4453 return (error);
4454 }
4455
4456 if (zc->zc_fromobj != 0) {
4457 error = dsl_dataset_hold_obj(dp, zc->zc_fromobj,
4458 FTAG, &fromsnap);
4459 if (error != 0) {
4460 dsl_dataset_rele(tosnap, FTAG);
4461 dsl_pool_rele(dp, FTAG);
4462 return (error);
4463 }
4464 }
4465
4466 error = dmu_send_estimate(tosnap, fromsnap, compressok,
4467 &zc->zc_objset_type);
4468
4469 if (fromsnap != NULL)
4470 dsl_dataset_rele(fromsnap, FTAG);
4471 dsl_dataset_rele(tosnap, FTAG);
4472 dsl_pool_rele(dp, FTAG);
4473 } else {
4474 file_t *fp = getf(zc->zc_cookie);
4475 if (fp == NULL)
4476 return (SET_ERROR(EBADF));
4477
4478 off = fp->f_offset;
4479 error = dmu_send_obj(zc->zc_name, zc->zc_sendobj,
4480 zc->zc_fromobj, embedok, large_block_ok, compressok,
4481 zc->zc_cookie, fp->f_vnode, &off);
4482
4483 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
4484 fp->f_offset = off;
4485 releasef(zc->zc_cookie);
4486 }
4487 return (error);
4488 }
4489
4490 /*
4491 * inputs:
4492 * zc_name name of snapshot on which to report progress
4493 * zc_cookie file descriptor of send stream
4494 *
4495 * outputs:
4496 * zc_cookie number of bytes written in send stream thus far
4497 */
4498 static int
4499 zfs_ioc_send_progress(zfs_cmd_t *zc)
4500 {
4501 dsl_pool_t *dp;
4502 dsl_dataset_t *ds;
4503 dmu_sendarg_t *dsp = NULL;
4504 int error;
4505
4506 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
4507 if (error != 0)
4508 return (error);
4509
4510 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &ds);
4511 if (error != 0) {
4512 dsl_pool_rele(dp, FTAG);
4513 return (error);
4514 }
4515
4516 mutex_enter(&ds->ds_sendstream_lock);
4517
4518 /*
4519 * Iterate over all the send streams currently active on this dataset.
4520 * If there's one which matches the specified file descriptor _and_ the
4521 * stream was started by the current process, return the progress of
4522 * that stream.
4523 */
4524 for (dsp = list_head(&ds->ds_sendstreams); dsp != NULL;
4525 dsp = list_next(&ds->ds_sendstreams, dsp)) {
4526 if (dsp->dsa_outfd == zc->zc_cookie &&
4527 dsp->dsa_proc == curproc)
4528 break;
4529 }
4530
4531 if (dsp != NULL)
4532 zc->zc_cookie = *(dsp->dsa_off);
4533 else
4534 error = SET_ERROR(ENOENT);
4535
4536 mutex_exit(&ds->ds_sendstream_lock);
4537 dsl_dataset_rele(ds, FTAG);
4538 dsl_pool_rele(dp, FTAG);
4539 return (error);
4540 }
4541
4542 static int
4543 zfs_ioc_inject_fault(zfs_cmd_t *zc)
4544 {
4545 int id, error;
4546
4547 error = zio_inject_fault(zc->zc_name, (int)zc->zc_guid, &id,
4548 &zc->zc_inject_record);
4549
4550 if (error == 0)
4551 zc->zc_guid = (uint64_t)id;
4552
4553 return (error);
4554 }
4555
4556 static int
4557 zfs_ioc_clear_fault(zfs_cmd_t *zc)
4558 {
4559 return (zio_clear_fault((int)zc->zc_guid));
4560 }
4561
4562 static int
4563 zfs_ioc_inject_list_next(zfs_cmd_t *zc)
4564 {
4565 int id = (int)zc->zc_guid;
4566 int error;
4567
4568 error = zio_inject_list_next(&id, zc->zc_name, sizeof (zc->zc_name),
4569 &zc->zc_inject_record);
4570
4571 zc->zc_guid = id;
4572
4573 return (error);
4574 }
4575
4576 static int
4577 zfs_ioc_error_log(zfs_cmd_t *zc)
4578 {
4579 spa_t *spa;
4580 int error;
4581 size_t count = (size_t)zc->zc_nvlist_dst_size;
4582
4583 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
4584 return (error);
4585
4586 error = spa_get_errlog(spa, (void *)(uintptr_t)zc->zc_nvlist_dst,
4587 &count);
4588 if (error == 0)
4589 zc->zc_nvlist_dst_size = count;
4590 else
4591 zc->zc_nvlist_dst_size = spa_get_errlog_size(spa);
4592
4593 spa_close(spa, FTAG);
4594
4595 return (error);
4596 }
4597
4598 static int
4599 zfs_ioc_clear(zfs_cmd_t *zc)
4600 {
4601 spa_t *spa;
4602 vdev_t *vd;
4603 int error;
4604
4605 /*
4606 * On zpool clear we also fix up missing slogs
4607 */
4608 mutex_enter(&spa_namespace_lock);
4609 spa = spa_lookup(zc->zc_name);
4610 if (spa == NULL) {
4611 mutex_exit(&spa_namespace_lock);
4612 return (SET_ERROR(EIO));
4613 }
4614 if (spa_get_log_state(spa) == SPA_LOG_MISSING) {
4615 /* we need to let spa_open/spa_load clear the chains */
4616 spa_set_log_state(spa, SPA_LOG_CLEAR);
4617 }
4618 spa->spa_last_open_failed = 0;
4619 mutex_exit(&spa_namespace_lock);
4620
4621 if (zc->zc_cookie & ZPOOL_NO_REWIND) {
4622 error = spa_open(zc->zc_name, &spa, FTAG);
4623 } else {
4624 nvlist_t *policy;
4625 nvlist_t *config = NULL;
4626
4627 if (zc->zc_nvlist_src == NULL)
4628 return (SET_ERROR(EINVAL));
4629
4630 if ((error = get_nvlist(zc->zc_nvlist_src,
4631 zc->zc_nvlist_src_size, zc->zc_iflags, &policy)) == 0) {
4632 error = spa_open_rewind(zc->zc_name, &spa, FTAG,
4633 policy, &config);
4634 if (config != NULL) {
4635 int err;
4636
4637 if ((err = put_nvlist(zc, config)) != 0)
4638 error = err;
4639 nvlist_free(config);
4640 }
4641 nvlist_free(policy);
4642 }
4643 }
4644
4645 if (error != 0)
4646 return (error);
4647
4648 spa_vdev_state_enter(spa, SCL_NONE);
4649
4650 if (zc->zc_guid == 0) {
4651 vd = NULL;
4652 } else {
4653 vd = spa_lookup_by_guid(spa, zc->zc_guid, B_TRUE);
4654 if (vd == NULL) {
4655 (void) spa_vdev_state_exit(spa, NULL, ENODEV);
4656 spa_close(spa, FTAG);
4657 return (SET_ERROR(ENODEV));
4658 }
4659 }
4660
4661 vdev_clear(spa, vd);
4662
4663 (void) spa_vdev_state_exit(spa, NULL, 0);
4664
4665 /*
4666 * Resume any suspended I/Os.
4667 */
4668 if (zio_resume(spa) != 0)
4669 error = SET_ERROR(EIO);
4670
4671 spa_close(spa, FTAG);
4672
4673 return (error);
4674 }
4675
4676 static int
4677 zfs_ioc_pool_reopen(zfs_cmd_t *zc)
4678 {
4679 spa_t *spa;
4680 int error;
4681
4682 error = spa_open(zc->zc_name, &spa, FTAG);
4683 if (error != 0)
4684 return (error);
4685
4686 spa_vdev_state_enter(spa, SCL_NONE);
4687
4688 /*
4689 * If a resilver is already in progress then set the
4690 * spa_scrub_reopen flag to B_TRUE so that we don't restart
4691 * the scan as a side effect of the reopen. Otherwise, let
4692 * vdev_open() decided if a resilver is required.
4693 */
4694 spa->spa_scrub_reopen = dsl_scan_resilvering(spa->spa_dsl_pool);
4695 vdev_reopen(spa->spa_root_vdev);
4696 spa->spa_scrub_reopen = B_FALSE;
4697
4698 (void) spa_vdev_state_exit(spa, NULL, 0);
4699 spa_close(spa, FTAG);
4700 return (0);
4701 }
4702 /*
4703 * inputs:
4704 * zc_name name of filesystem
4705 * zc_value name of origin snapshot
4706 *
4707 * outputs:
4708 * zc_string name of conflicting snapshot, if there is one
4709 */
4710 static int
4711 zfs_ioc_promote(zfs_cmd_t *zc)
4712 {
4713 char *cp;
4714
4715 /*
4716 * We don't need to unmount *all* the origin fs's snapshots, but
4717 * it's easier.
4718 */
4719 cp = strchr(zc->zc_value, '@');
4720 if (cp)
4721 *cp = '\0';
4722 (void) dmu_objset_find(zc->zc_value,
4723 zfs_unmount_snap_cb, NULL, DS_FIND_SNAPSHOTS);
4724 return (dsl_dataset_promote(zc->zc_name, zc->zc_string));
4725 }
4726
4727 /*
4728 * Retrieve a single {user|group}{used|quota}@... property.
4729 *
4730 * inputs:
4731 * zc_name name of filesystem
4732 * zc_objset_type zfs_userquota_prop_t
4733 * zc_value domain name (eg. "S-1-234-567-89")
4734 * zc_guid RID/UID/GID
4735 *
4736 * outputs:
4737 * zc_cookie property value
4738 */
4739 static int
4740 zfs_ioc_userspace_one(zfs_cmd_t *zc)
4741 {
4742 zfsvfs_t *zfsvfs;
4743 int error;
4744
4745 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
4746 return (SET_ERROR(EINVAL));
4747
4748 error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE);
4749 if (error != 0)
4750 return (error);
4751
4752 error = zfs_userspace_one(zfsvfs,
4753 zc->zc_objset_type, zc->zc_value, zc->zc_guid, &zc->zc_cookie);
4754 zfsvfs_rele(zfsvfs, FTAG);
4755
4756 return (error);
4757 }
4758
4759 /*
4760 * inputs:
4761 * zc_name name of filesystem
4762 * zc_cookie zap cursor
4763 * zc_objset_type zfs_userquota_prop_t
4764 * zc_nvlist_dst[_size] buffer to fill (not really an nvlist)
4765 *
4766 * outputs:
4767 * zc_nvlist_dst[_size] data buffer (array of zfs_useracct_t)
4768 * zc_cookie zap cursor
4769 */
4770 static int
4771 zfs_ioc_userspace_many(zfs_cmd_t *zc)
4772 {
4773 zfsvfs_t *zfsvfs;
4774 int bufsize = zc->zc_nvlist_dst_size;
4775
4776 if (bufsize <= 0)
4777 return (SET_ERROR(ENOMEM));
4778
4779 int error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE);
4780 if (error != 0)
4781 return (error);
4782
4783 void *buf = kmem_alloc(bufsize, KM_SLEEP);
4784
4785 error = zfs_userspace_many(zfsvfs, zc->zc_objset_type, &zc->zc_cookie,
4786 buf, &zc->zc_nvlist_dst_size);
4787
4788 if (error == 0) {
4789 error = xcopyout(buf,
4790 (void *)(uintptr_t)zc->zc_nvlist_dst,
4791 zc->zc_nvlist_dst_size);
4792 }
4793 kmem_free(buf, bufsize);
4794 zfsvfs_rele(zfsvfs, FTAG);
4795
4796 return (error);
4797 }
4798
4799 /*
4800 * inputs:
4801 * zc_name name of filesystem
4802 *
4803 * outputs:
4804 * none
4805 */
4806 static int
4807 zfs_ioc_userspace_upgrade(zfs_cmd_t *zc)
4808 {
4809 objset_t *os;
4810 int error = 0;
4811 zfsvfs_t *zfsvfs;
4812
4813 if (getzfsvfs(zc->zc_name, &zfsvfs) == 0) {
4814 if (!dmu_objset_userused_enabled(zfsvfs->z_os)) {
4815 /*
4816 * If userused is not enabled, it may be because the
4817 * objset needs to be closed & reopened (to grow the
4818 * objset_phys_t). Suspend/resume the fs will do that.
4819 */
4820 dsl_dataset_t *ds;
4821
4822 ds = dmu_objset_ds(zfsvfs->z_os);
4823 error = zfs_suspend_fs(zfsvfs);
4824 if (error == 0) {
4825 dmu_objset_refresh_ownership(zfsvfs->z_os,
4826 zfsvfs);
4827 error = zfs_resume_fs(zfsvfs, ds);
4828 }
4829 }
4830 if (error == 0)
4831 error = dmu_objset_userspace_upgrade(zfsvfs->z_os);
4832 VFS_RELE(zfsvfs->z_vfs);
4833 } else {
4834 /* XXX kind of reading contents without owning */
4835 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
4836 if (error != 0)
4837 return (error);
4838
4839 error = dmu_objset_userspace_upgrade(os);
4840 dmu_objset_rele(os, FTAG);
4841 }
4842
4843 return (error);
4844 }
4845
4846 /*
4847 * We don't want to have a hard dependency
4848 * against some special symbols in sharefs
4849 * nfs, and smbsrv. Determine them if needed when
4850 * the first file system is shared.
4851 * Neither sharefs, nfs or smbsrv are unloadable modules.
4852 */
4853 int (*znfsexport_fs)(void *arg);
4854 int (*zshare_fs)(enum sharefs_sys_op, share_t *, uint32_t);
4855 int (*zsmbexport_fs)(void *arg, boolean_t add_share);
4856
4857 int zfs_nfsshare_inited;
4858 int zfs_smbshare_inited;
4859
4860 ddi_modhandle_t nfs_mod;
4861 ddi_modhandle_t sharefs_mod;
4862 ddi_modhandle_t smbsrv_mod;
4863 kmutex_t zfs_share_lock;
4864
4865 static int
4866 zfs_init_sharefs()
4867 {
4868 int error;
4869
4870 ASSERT(MUTEX_HELD(&zfs_share_lock));
4871 /* Both NFS and SMB shares also require sharetab support. */
4872 if (sharefs_mod == NULL && ((sharefs_mod =
4873 ddi_modopen("fs/sharefs",
4874 KRTLD_MODE_FIRST, &error)) == NULL)) {
4875 return (SET_ERROR(ENOSYS));
4876 }
4877 if (zshare_fs == NULL && ((zshare_fs =
4878 (int (*)(enum sharefs_sys_op, share_t *, uint32_t))
4879 ddi_modsym(sharefs_mod, "sharefs_impl", &error)) == NULL)) {
4880 return (SET_ERROR(ENOSYS));
4881 }
4882 return (0);
4883 }
4884
4885 static int
4886 zfs_ioc_share(zfs_cmd_t *zc)
4887 {
4888 int error;
4889 int opcode;
4890
4891 switch (zc->zc_share.z_sharetype) {
4892 case ZFS_SHARE_NFS:
4893 case ZFS_UNSHARE_NFS:
4894 if (zfs_nfsshare_inited == 0) {
4895 mutex_enter(&zfs_share_lock);
4896 if (nfs_mod == NULL && ((nfs_mod = ddi_modopen("fs/nfs",
4897 KRTLD_MODE_FIRST, &error)) == NULL)) {
4898 mutex_exit(&zfs_share_lock);
4899 return (SET_ERROR(ENOSYS));
4900 }
4901 if (znfsexport_fs == NULL &&
4902 ((znfsexport_fs = (int (*)(void *))
4903 ddi_modsym(nfs_mod,
4904 "nfs_export", &error)) == NULL)) {
4905 mutex_exit(&zfs_share_lock);
4906 return (SET_ERROR(ENOSYS));
4907 }
4908 error = zfs_init_sharefs();
4909 if (error != 0) {
4910 mutex_exit(&zfs_share_lock);
4911 return (SET_ERROR(ENOSYS));
4912 }
4913 zfs_nfsshare_inited = 1;
4914 mutex_exit(&zfs_share_lock);
4915 }
4916 break;
4917 case ZFS_SHARE_SMB:
4918 case ZFS_UNSHARE_SMB:
4919 if (zfs_smbshare_inited == 0) {
4920 mutex_enter(&zfs_share_lock);
4921 if (smbsrv_mod == NULL && ((smbsrv_mod =
4922 ddi_modopen("drv/smbsrv",
4923 KRTLD_MODE_FIRST, &error)) == NULL)) {
4924 mutex_exit(&zfs_share_lock);
4925 return (SET_ERROR(ENOSYS));
4926 }
4927 if (zsmbexport_fs == NULL && ((zsmbexport_fs =
4928 (int (*)(void *, boolean_t))ddi_modsym(smbsrv_mod,
4929 "smb_server_share", &error)) == NULL)) {
4930 mutex_exit(&zfs_share_lock);
4931 return (SET_ERROR(ENOSYS));
4932 }
4933 error = zfs_init_sharefs();
4934 if (error != 0) {
4935 mutex_exit(&zfs_share_lock);
4936 return (SET_ERROR(ENOSYS));
4937 }
4938 zfs_smbshare_inited = 1;
4939 mutex_exit(&zfs_share_lock);
4940 }
4941 break;
4942 default:
4943 return (SET_ERROR(EINVAL));
4944 }
4945
4946 switch (zc->zc_share.z_sharetype) {
4947 case ZFS_SHARE_NFS:
4948 case ZFS_UNSHARE_NFS:
4949 if (error =
4950 znfsexport_fs((void *)
4951 (uintptr_t)zc->zc_share.z_exportdata))
4952 return (error);
4953 break;
4954 case ZFS_SHARE_SMB:
4955 case ZFS_UNSHARE_SMB:
4956 if (error = zsmbexport_fs((void *)
4957 (uintptr_t)zc->zc_share.z_exportdata,
4958 zc->zc_share.z_sharetype == ZFS_SHARE_SMB ?
4959 B_TRUE: B_FALSE)) {
4960 return (error);
4961 }
4962 break;
4963 }
4964
4965 opcode = (zc->zc_share.z_sharetype == ZFS_SHARE_NFS ||
4966 zc->zc_share.z_sharetype == ZFS_SHARE_SMB) ?
4967 SHAREFS_ADD : SHAREFS_REMOVE;
4968
4969 /*
4970 * Add or remove share from sharetab
4971 */
4972 error = zshare_fs(opcode,
4973 (void *)(uintptr_t)zc->zc_share.z_sharedata,
4974 zc->zc_share.z_sharemax);
4975
4976 return (error);
4977
4978 }
4979
4980 ace_t full_access[] = {
4981 {(uid_t)-1, ACE_ALL_PERMS, ACE_EVERYONE, 0}
4982 };
4983
4984 /*
4985 * inputs:
4986 * zc_name name of containing filesystem
4987 * zc_obj object # beyond which we want next in-use object #
4988 *
4989 * outputs:
4990 * zc_obj next in-use object #
4991 */
4992 static int
4993 zfs_ioc_next_obj(zfs_cmd_t *zc)
4994 {
4995 objset_t *os = NULL;
4996 int error;
4997
4998 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
4999 if (error != 0)
5000 return (error);
5001
5002 error = dmu_object_next(os, &zc->zc_obj, B_FALSE,
5003 dsl_dataset_phys(os->os_dsl_dataset)->ds_prev_snap_txg);
5004
5005 dmu_objset_rele(os, FTAG);
5006 return (error);
5007 }
5008
5009 /*
5010 * inputs:
5011 * zc_name name of filesystem
5012 * zc_value prefix name for snapshot
5013 * zc_cleanup_fd cleanup-on-exit file descriptor for calling process
5014 *
5015 * outputs:
5016 * zc_value short name of new snapshot
5017 */
5018 static int
5019 zfs_ioc_tmp_snapshot(zfs_cmd_t *zc)
5020 {
5021 char *snap_name;
5022 char *hold_name;
5023 int error;
5024 minor_t minor;
5025
5026 error = zfs_onexit_fd_hold(zc->zc_cleanup_fd, &minor);
5027 if (error != 0)
5028 return (error);
5029
5030 snap_name = kmem_asprintf("%s-%016llx", zc->zc_value,
5031 (u_longlong_t)ddi_get_lbolt64());
5032 hold_name = kmem_asprintf("%%%s", zc->zc_value);
5033
5034 error = dsl_dataset_snapshot_tmp(zc->zc_name, snap_name, minor,
5035 hold_name);
5036 if (error == 0)
5037 (void) strcpy(zc->zc_value, snap_name);
5038 strfree(snap_name);
5039 strfree(hold_name);
5040 zfs_onexit_fd_rele(zc->zc_cleanup_fd);
5041 return (error);
5042 }
5043
5044 /*
5045 * inputs:
5046 * zc_name name of "to" snapshot
5047 * zc_value name of "from" snapshot
5048 * zc_cookie file descriptor to write diff data on
5049 *
5050 * outputs:
5051 * dmu_diff_record_t's to the file descriptor
5052 */
5053 static int
5054 zfs_ioc_diff(zfs_cmd_t *zc)
5055 {
5056 file_t *fp;
5057 offset_t off;
5058 int error;
5059
5060 fp = getf(zc->zc_cookie);
5061 if (fp == NULL)
5062 return (SET_ERROR(EBADF));
5063
5064 off = fp->f_offset;
5065
5066 error = dmu_diff(zc->zc_name, zc->zc_value, fp->f_vnode, &off);
5067
5068 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
5069 fp->f_offset = off;
5070 releasef(zc->zc_cookie);
5071
5072 return (error);
5073 }
5074
5075 /*
5076 * Remove all ACL files in shares dir
5077 */
5078 static int
5079 zfs_smb_acl_purge(znode_t *dzp)
5080 {
5081 zap_cursor_t zc;
5082 zap_attribute_t zap;
5083 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
5084 int error;
5085
5086 for (zap_cursor_init(&zc, zfsvfs->z_os, dzp->z_id);
5087 (error = zap_cursor_retrieve(&zc, &zap)) == 0;
5088 zap_cursor_advance(&zc)) {
5089 if ((error = VOP_REMOVE(ZTOV(dzp), zap.za_name, kcred,
5090 NULL, 0)) != 0)
5091 break;
5092 }
5093 zap_cursor_fini(&zc);
5094 return (error);
5095 }
5096
5097 static int
5098 zfs_ioc_smb_acl(zfs_cmd_t *zc)
5099 {
5100 vnode_t *vp;
5101 znode_t *dzp;
5102 vnode_t *resourcevp = NULL;
5103 znode_t *sharedir;
5104 zfsvfs_t *zfsvfs;
5105 nvlist_t *nvlist;
5106 char *src, *target;
5107 vattr_t vattr;
5108 vsecattr_t vsec;
5109 int error = 0;
5110
5111 if ((error = lookupname(zc->zc_value, UIO_SYSSPACE,
5112 NO_FOLLOW, NULL, &vp)) != 0)
5113 return (error);
5114
5115 /* Now make sure mntpnt and dataset are ZFS */
5116
5117 if (vp->v_vfsp->vfs_fstype != zfsfstype ||
5118 (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource),
5119 zc->zc_name) != 0)) {
5120 VN_RELE(vp);
5121 return (SET_ERROR(EINVAL));
5122 }
5123
5124 dzp = VTOZ(vp);
5125 zfsvfs = dzp->z_zfsvfs;
5126 ZFS_ENTER(zfsvfs);
5127
5128 /*
5129 * Create share dir if its missing.
5130 */
5131 mutex_enter(&zfsvfs->z_lock);
5132 if (zfsvfs->z_shares_dir == 0) {
5133 dmu_tx_t *tx;
5134
5135 tx = dmu_tx_create(zfsvfs->z_os);
5136 dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, TRUE,
5137 ZFS_SHARES_DIR);
5138 dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL);
5139 error = dmu_tx_assign(tx, TXG_WAIT);
5140 if (error != 0) {
5141 dmu_tx_abort(tx);
5142 } else {
5143 error = zfs_create_share_dir(zfsvfs, tx);
5144 dmu_tx_commit(tx);
5145 }
5146 if (error != 0) {
5147 mutex_exit(&zfsvfs->z_lock);
5148 VN_RELE(vp);
5149 ZFS_EXIT(zfsvfs);
5150 return (error);
5151 }
5152 }
5153 mutex_exit(&zfsvfs->z_lock);
5154
5155 ASSERT(zfsvfs->z_shares_dir);
5156 if ((error = zfs_zget(zfsvfs, zfsvfs->z_shares_dir, &sharedir)) != 0) {
5157 VN_RELE(vp);
5158 ZFS_EXIT(zfsvfs);
5159 return (error);
5160 }
5161
5162 switch (zc->zc_cookie) {
5163 case ZFS_SMB_ACL_ADD:
5164 vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE;
5165 vattr.va_type = VREG;
5166 vattr.va_mode = S_IFREG|0777;
5167 vattr.va_uid = 0;
5168 vattr.va_gid = 0;
5169
5170 vsec.vsa_mask = VSA_ACE;
5171 vsec.vsa_aclentp = &full_access;
5172 vsec.vsa_aclentsz = sizeof (full_access);
5173 vsec.vsa_aclcnt = 1;
5174
5175 error = VOP_CREATE(ZTOV(sharedir), zc->zc_string,
5176 &vattr, EXCL, 0, &resourcevp, kcred, 0, NULL, &vsec);
5177 if (resourcevp)
5178 VN_RELE(resourcevp);
5179 break;
5180
5181 case ZFS_SMB_ACL_REMOVE:
5182 error = VOP_REMOVE(ZTOV(sharedir), zc->zc_string, kcred,
5183 NULL, 0);
5184 break;
5185
5186 case ZFS_SMB_ACL_RENAME:
5187 if ((error = get_nvlist(zc->zc_nvlist_src,
5188 zc->zc_nvlist_src_size, zc->zc_iflags, &nvlist)) != 0) {
5189 VN_RELE(vp);
5190 VN_RELE(ZTOV(sharedir));
5191 ZFS_EXIT(zfsvfs);
5192 return (error);
5193 }
5194 if (nvlist_lookup_string(nvlist, ZFS_SMB_ACL_SRC, &src) ||
5195 nvlist_lookup_string(nvlist, ZFS_SMB_ACL_TARGET,
5196 &target)) {
5197 VN_RELE(vp);
5198 VN_RELE(ZTOV(sharedir));
5199 ZFS_EXIT(zfsvfs);
5200 nvlist_free(nvlist);
5201 return (error);
5202 }
5203 error = VOP_RENAME(ZTOV(sharedir), src, ZTOV(sharedir), target,
5204 kcred, NULL, 0);
5205 nvlist_free(nvlist);
5206 break;
5207
5208 case ZFS_SMB_ACL_PURGE:
5209 error = zfs_smb_acl_purge(sharedir);
5210 break;
5211
5212 default:
5213 error = SET_ERROR(EINVAL);
5214 break;
5215 }
5216
5217 VN_RELE(vp);
5218 VN_RELE(ZTOV(sharedir));
5219
5220 ZFS_EXIT(zfsvfs);
5221
5222 return (error);
5223 }
5224
5225 /*
5226 * innvl: {
5227 * "holds" -> { snapname -> holdname (string), ... }
5228 * (optional) "cleanup_fd" -> fd (int32)
5229 * }
5230 *
5231 * outnvl: {
5232 * snapname -> error value (int32)
5233 * ...
5234 * }
5235 */
5236 /* ARGSUSED */
5237 static int
5238 zfs_ioc_hold(const char *pool, nvlist_t *args, nvlist_t *errlist)
5239 {
5240 nvpair_t *pair;
5241 nvlist_t *holds;
5242 int cleanup_fd = -1;
5243 int error;
5244 minor_t minor = 0;
5245
5246 error = nvlist_lookup_nvlist(args, "holds", &holds);
5247 if (error != 0)
5248 return (SET_ERROR(EINVAL));
5249
5250 /* make sure the user didn't pass us any invalid (empty) tags */
5251 for (pair = nvlist_next_nvpair(holds, NULL); pair != NULL;
5252 pair = nvlist_next_nvpair(holds, pair)) {
5253 char *htag;
5254
5255 error = nvpair_value_string(pair, &htag);
5256 if (error != 0)
5257 return (SET_ERROR(error));
5258
5259 if (strlen(htag) == 0)
5260 return (SET_ERROR(EINVAL));
5261 }
5262
5263 if (nvlist_lookup_int32(args, "cleanup_fd", &cleanup_fd) == 0) {
5264 error = zfs_onexit_fd_hold(cleanup_fd, &minor);
5265 if (error != 0)
5266 return (error);
5267 }
5268
5269 error = dsl_dataset_user_hold(holds, minor, errlist);
5270 if (minor != 0)
5271 zfs_onexit_fd_rele(cleanup_fd);
5272 return (error);
5273 }
5274
5275 /*
5276 * innvl is not used.
5277 *
5278 * outnvl: {
5279 * holdname -> time added (uint64 seconds since epoch)
5280 * ...
5281 * }
5282 */
5283 /* ARGSUSED */
5284 static int
5285 zfs_ioc_get_holds(const char *snapname, nvlist_t *args, nvlist_t *outnvl)
5286 {
5287 return (dsl_dataset_get_holds(snapname, outnvl));
5288 }
5289
5290 /*
5291 * innvl: {
5292 * snapname -> { holdname, ... }
5293 * ...
5294 * }
5295 *
5296 * outnvl: {
5297 * snapname -> error value (int32)
5298 * ...
5299 * }
5300 */
5301 /* ARGSUSED */
5302 static int
5303 zfs_ioc_release(const char *pool, nvlist_t *holds, nvlist_t *errlist)
5304 {
5305 return (dsl_dataset_user_release(holds, errlist));
5306 }
5307
5308 /*
5309 * inputs:
5310 * zc_name name of new filesystem or snapshot
5311 * zc_value full name of old snapshot
5312 *
5313 * outputs:
5314 * zc_cookie space in bytes
5315 * zc_objset_type compressed space in bytes
5316 * zc_perm_action uncompressed space in bytes
5317 */
5318 static int
5319 zfs_ioc_space_written(zfs_cmd_t *zc)
5320 {
5321 int error;
5322 dsl_pool_t *dp;
5323 dsl_dataset_t *new, *old;
5324
5325 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
5326 if (error != 0)
5327 return (error);
5328 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &new);
5329 if (error != 0) {
5330 dsl_pool_rele(dp, FTAG);
5331 return (error);
5332 }
5333 error = dsl_dataset_hold(dp, zc->zc_value, FTAG, &old);
5334 if (error != 0) {
5335 dsl_dataset_rele(new, FTAG);
5336 dsl_pool_rele(dp, FTAG);
5337 return (error);
5338 }
5339
5340 error = dsl_dataset_space_written(old, new, &zc->zc_cookie,
5341 &zc->zc_objset_type, &zc->zc_perm_action);
5342 dsl_dataset_rele(old, FTAG);
5343 dsl_dataset_rele(new, FTAG);
5344 dsl_pool_rele(dp, FTAG);
5345 return (error);
5346 }
5347
5348 /*
5349 * innvl: {
5350 * "firstsnap" -> snapshot name
5351 * }
5352 *
5353 * outnvl: {
5354 * "used" -> space in bytes
5355 * "compressed" -> compressed space in bytes
5356 * "uncompressed" -> uncompressed space in bytes
5357 * }
5358 */
5359 static int
5360 zfs_ioc_space_snaps(const char *lastsnap, nvlist_t *innvl, nvlist_t *outnvl)
5361 {
5362 int error;
5363 dsl_pool_t *dp;
5364 dsl_dataset_t *new, *old;
5365 char *firstsnap;
5366 uint64_t used, comp, uncomp;
5367
5368 if (nvlist_lookup_string(innvl, "firstsnap", &firstsnap) != 0)
5369 return (SET_ERROR(EINVAL));
5370
5371 error = dsl_pool_hold(lastsnap, FTAG, &dp);
5372 if (error != 0)
5373 return (error);
5374
5375 error = dsl_dataset_hold(dp, lastsnap, FTAG, &new);
5376 if (error == 0 && !new->ds_is_snapshot) {
5377 dsl_dataset_rele(new, FTAG);
5378 error = SET_ERROR(EINVAL);
5379 }
5380 if (error != 0) {
5381 dsl_pool_rele(dp, FTAG);
5382 return (error);
5383 }
5384 error = dsl_dataset_hold(dp, firstsnap, FTAG, &old);
5385 if (error == 0 && !old->ds_is_snapshot) {
5386 dsl_dataset_rele(old, FTAG);
5387 error = SET_ERROR(EINVAL);
5388 }
5389 if (error != 0) {
5390 dsl_dataset_rele(new, FTAG);
5391 dsl_pool_rele(dp, FTAG);
5392 return (error);
5393 }
5394
5395 error = dsl_dataset_space_wouldfree(old, new, &used, &comp, &uncomp);
5396 dsl_dataset_rele(old, FTAG);
5397 dsl_dataset_rele(new, FTAG);
5398 dsl_pool_rele(dp, FTAG);
5399 fnvlist_add_uint64(outnvl, "used", used);
5400 fnvlist_add_uint64(outnvl, "compressed", comp);
5401 fnvlist_add_uint64(outnvl, "uncompressed", uncomp);
5402 return (error);
5403 }
5404
5405 /*
5406 * innvl: {
5407 * "fd" -> file descriptor to write stream to (int32)
5408 * (optional) "fromsnap" -> full snap name to send an incremental from
5409 * (optional) "largeblockok" -> (value ignored)
5410 * indicates that blocks > 128KB are permitted
5411 * (optional) "embedok" -> (value ignored)
5412 * presence indicates DRR_WRITE_EMBEDDED records are permitted
5413 * (optional) "compressok" -> (value ignored)
5414 * presence indicates compressed DRR_WRITE records are permitted
5415 * (optional) "resume_object" and "resume_offset" -> (uint64)
5416 * if present, resume send stream from specified object and offset.
5417 * }
5418 *
5419 * outnvl is unused
5420 */
5421 /* ARGSUSED */
5422 static int
5423 zfs_ioc_send_new(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl)
5424 {
5425 int error;
5426 offset_t off;
5427 char *fromname = NULL;
5428 int fd;
5429 boolean_t largeblockok;
5430 boolean_t embedok;
5431 boolean_t compressok;
5432 uint64_t resumeobj = 0;
5433 uint64_t resumeoff = 0;
5434
5435 error = nvlist_lookup_int32(innvl, "fd", &fd);
5436 if (error != 0)
5437 return (SET_ERROR(EINVAL));
5438
5439 (void) nvlist_lookup_string(innvl, "fromsnap", &fromname);
5440
5441 largeblockok = nvlist_exists(innvl, "largeblockok");
5442 embedok = nvlist_exists(innvl, "embedok");
5443 compressok = nvlist_exists(innvl, "compressok");
5444
5445 (void) nvlist_lookup_uint64(innvl, "resume_object", &resumeobj);
5446 (void) nvlist_lookup_uint64(innvl, "resume_offset", &resumeoff);
5447
5448 file_t *fp = getf(fd);
5449 if (fp == NULL)
5450 return (SET_ERROR(EBADF));
5451
5452 off = fp->f_offset;
5453 error = dmu_send(snapname, fromname, embedok, largeblockok, compressok,
5454 fd, resumeobj, resumeoff, fp->f_vnode, &off);
5455
5456 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
5457 fp->f_offset = off;
5458 releasef(fd);
5459 return (error);
5460 }
5461
5462 /*
5463 * Determine approximately how large a zfs send stream will be -- the number
5464 * of bytes that will be written to the fd supplied to zfs_ioc_send_new().
5465 *
5466 * innvl: {
5467 * (optional) "from" -> full snap or bookmark name to send an incremental
5468 * from
5469 * (optional) "largeblockok" -> (value ignored)
5470 * indicates that blocks > 128KB are permitted
5471 * (optional) "embedok" -> (value ignored)
5472 * presence indicates DRR_WRITE_EMBEDDED records are permitted
5473 * (optional) "compressok" -> (value ignored)
5474 * presence indicates compressed DRR_WRITE records are permitted
5475 * }
5476 *
5477 * outnvl: {
5478 * "space" -> bytes of space (uint64)
5479 * }
5480 */
5481 static int
5482 zfs_ioc_send_space(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl)
5483 {
5484 dsl_pool_t *dp;
5485 dsl_dataset_t *tosnap;
5486 int error;
5487 char *fromname;
5488 /* LINTED E_FUNC_SET_NOT_USED */
5489 boolean_t largeblockok;
5490 /* LINTED E_FUNC_SET_NOT_USED */
5491 boolean_t embedok;
5492 boolean_t compressok;
5493 uint64_t space;
5494
5495 error = dsl_pool_hold(snapname, FTAG, &dp);
5496 if (error != 0)
5497 return (error);
5498
5499 error = dsl_dataset_hold(dp, snapname, FTAG, &tosnap);
5500 if (error != 0) {
5501 dsl_pool_rele(dp, FTAG);
5502 return (error);
5503 }
5504
5505 largeblockok = nvlist_exists(innvl, "largeblockok");
5506 embedok = nvlist_exists(innvl, "embedok");
5507 compressok = nvlist_exists(innvl, "compressok");
5508
5509 error = nvlist_lookup_string(innvl, "from", &fromname);
5510 if (error == 0) {
5511 if (strchr(fromname, '@') != NULL) {
5512 /*
5513 * If from is a snapshot, hold it and use the more
5514 * efficient dmu_send_estimate to estimate send space
5515 * size using deadlists.
5516 */
5517 dsl_dataset_t *fromsnap;
5518 error = dsl_dataset_hold(dp, fromname, FTAG, &fromsnap);
5519 if (error != 0)
5520 goto out;
5521 error = dmu_send_estimate(tosnap, fromsnap, compressok,
5522 &space);
5523 dsl_dataset_rele(fromsnap, FTAG);
5524 } else if (strchr(fromname, '#') != NULL) {
5525 /*
5526 * If from is a bookmark, fetch the creation TXG of the
5527 * snapshot it was created from and use that to find
5528 * blocks that were born after it.
5529 */
5530 zfs_bookmark_phys_t frombm;
5531
5532 error = dsl_bookmark_lookup(dp, fromname, tosnap,
5533 &frombm);
5534 if (error != 0)
5535 goto out;
5536 error = dmu_send_estimate_from_txg(tosnap,
5537 frombm.zbm_creation_txg, compressok, &space);
5538 } else {
5539 /*
5540 * from is not properly formatted as a snapshot or
5541 * bookmark
5542 */
5543 error = SET_ERROR(EINVAL);
5544 goto out;
5545 }
5546 } else {
5547 // If estimating the size of a full send, use dmu_send_estimate
5548 error = dmu_send_estimate(tosnap, NULL, compressok, &space);
5549 }
5550
5551 fnvlist_add_uint64(outnvl, "space", space);
5552
5553 out:
5554 dsl_dataset_rele(tosnap, FTAG);
5555 dsl_pool_rele(dp, FTAG);
5556 return (error);
5557 }
5558
5559 static zfs_ioc_vec_t zfs_ioc_vec[ZFS_IOC_LAST - ZFS_IOC_FIRST];
5560
5561 static void
5562 zfs_ioctl_register_legacy(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5563 zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck,
5564 boolean_t log_history, zfs_ioc_poolcheck_t pool_check)
5565 {
5566 zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST];
5567
5568 ASSERT3U(ioc, >=, ZFS_IOC_FIRST);
5569 ASSERT3U(ioc, <, ZFS_IOC_LAST);
5570 ASSERT3P(vec->zvec_legacy_func, ==, NULL);
5571 ASSERT3P(vec->zvec_func, ==, NULL);
5572
5573 vec->zvec_legacy_func = func;
5574 vec->zvec_secpolicy = secpolicy;
5575 vec->zvec_namecheck = namecheck;
5576 vec->zvec_allow_log = log_history;
5577 vec->zvec_pool_check = pool_check;
5578 }
5579
5580 /*
5581 * See the block comment at the beginning of this file for details on
5582 * each argument to this function.
5583 */
5584 static void
5585 zfs_ioctl_register(const char *name, zfs_ioc_t ioc, zfs_ioc_func_t *func,
5586 zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck,
5587 zfs_ioc_poolcheck_t pool_check, boolean_t smush_outnvlist,
5588 boolean_t allow_log)
5589 {
5590 zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST];
5591
5592 ASSERT3U(ioc, >=, ZFS_IOC_FIRST);
5593 ASSERT3U(ioc, <, ZFS_IOC_LAST);
5594 ASSERT3P(vec->zvec_legacy_func, ==, NULL);
5595 ASSERT3P(vec->zvec_func, ==, NULL);
5596
5597 /* if we are logging, the name must be valid */
5598 ASSERT(!allow_log || namecheck != NO_NAME);
5599
5600 vec->zvec_name = name;
5601 vec->zvec_func = func;
5602 vec->zvec_secpolicy = secpolicy;
5603 vec->zvec_namecheck = namecheck;
5604 vec->zvec_pool_check = pool_check;
5605 vec->zvec_smush_outnvlist = smush_outnvlist;
5606 vec->zvec_allow_log = allow_log;
5607 }
5608
5609 static void
5610 zfs_ioctl_register_pool(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5611 zfs_secpolicy_func_t *secpolicy, boolean_t log_history,
5612 zfs_ioc_poolcheck_t pool_check)
5613 {
5614 zfs_ioctl_register_legacy(ioc, func, secpolicy,
5615 POOL_NAME, log_history, pool_check);
5616 }
5617
5618 static void
5619 zfs_ioctl_register_dataset_nolog(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5620 zfs_secpolicy_func_t *secpolicy, zfs_ioc_poolcheck_t pool_check)
5621 {
5622 zfs_ioctl_register_legacy(ioc, func, secpolicy,
5623 DATASET_NAME, B_FALSE, pool_check);
5624 }
5625
5626 static void
5627 zfs_ioctl_register_pool_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func)
5628 {
5629 zfs_ioctl_register_legacy(ioc, func, zfs_secpolicy_config,
5630 POOL_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5631 }
5632
5633 static void
5634 zfs_ioctl_register_pool_meta(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5635 zfs_secpolicy_func_t *secpolicy)
5636 {
5637 zfs_ioctl_register_legacy(ioc, func, secpolicy,
5638 NO_NAME, B_FALSE, POOL_CHECK_NONE);
5639 }
5640
5641 static void
5642 zfs_ioctl_register_dataset_read_secpolicy(zfs_ioc_t ioc,
5643 zfs_ioc_legacy_func_t *func, zfs_secpolicy_func_t *secpolicy)
5644 {
5645 zfs_ioctl_register_legacy(ioc, func, secpolicy,
5646 DATASET_NAME, B_FALSE, POOL_CHECK_SUSPENDED);
5647 }
5648
5649 static void
5650 zfs_ioctl_register_dataset_read(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func)
5651 {
5652 zfs_ioctl_register_dataset_read_secpolicy(ioc, func,
5653 zfs_secpolicy_read);
5654 }
5655
5656 static void
5657 zfs_ioctl_register_dataset_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5658 zfs_secpolicy_func_t *secpolicy)
5659 {
5660 zfs_ioctl_register_legacy(ioc, func, secpolicy,
5661 DATASET_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5662 }
5663
5664 static void
5665 zfs_ioctl_init(void)
5666 {
5667 zfs_ioctl_register("snapshot", ZFS_IOC_SNAPSHOT,
5668 zfs_ioc_snapshot, zfs_secpolicy_snapshot, POOL_NAME,
5669 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5670
5671 zfs_ioctl_register("log_history", ZFS_IOC_LOG_HISTORY,
5672 zfs_ioc_log_history, zfs_secpolicy_log_history, NO_NAME,
5673 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_FALSE);
5674
5675 zfs_ioctl_register("space_snaps", ZFS_IOC_SPACE_SNAPS,
5676 zfs_ioc_space_snaps, zfs_secpolicy_read, DATASET_NAME,
5677 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5678
5679 zfs_ioctl_register("send", ZFS_IOC_SEND_NEW,
5680 zfs_ioc_send_new, zfs_secpolicy_send_new, DATASET_NAME,
5681 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5682
5683 zfs_ioctl_register("send_space", ZFS_IOC_SEND_SPACE,
5684 zfs_ioc_send_space, zfs_secpolicy_read, DATASET_NAME,
5685 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5686
5687 zfs_ioctl_register("create", ZFS_IOC_CREATE,
5688 zfs_ioc_create, zfs_secpolicy_create_clone, DATASET_NAME,
5689 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5690
5691 zfs_ioctl_register("clone", ZFS_IOC_CLONE,
5692 zfs_ioc_clone, zfs_secpolicy_create_clone, DATASET_NAME,
5693 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5694
5695 zfs_ioctl_register("destroy_snaps", ZFS_IOC_DESTROY_SNAPS,
5696 zfs_ioc_destroy_snaps, zfs_secpolicy_destroy_snaps, POOL_NAME,
5697 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5698
5699 zfs_ioctl_register("hold", ZFS_IOC_HOLD,
5700 zfs_ioc_hold, zfs_secpolicy_hold, POOL_NAME,
5701 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5702 zfs_ioctl_register("release", ZFS_IOC_RELEASE,
5703 zfs_ioc_release, zfs_secpolicy_release, POOL_NAME,
5704 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5705
5706 zfs_ioctl_register("get_holds", ZFS_IOC_GET_HOLDS,
5707 zfs_ioc_get_holds, zfs_secpolicy_read, DATASET_NAME,
5708 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5709
5710 zfs_ioctl_register("rollback", ZFS_IOC_ROLLBACK,
5711 zfs_ioc_rollback, zfs_secpolicy_rollback, DATASET_NAME,
5712 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_TRUE);
5713
5714 zfs_ioctl_register("bookmark", ZFS_IOC_BOOKMARK,
5715 zfs_ioc_bookmark, zfs_secpolicy_bookmark, POOL_NAME,
5716 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5717
5718 zfs_ioctl_register("get_bookmarks", ZFS_IOC_GET_BOOKMARKS,
5719 zfs_ioc_get_bookmarks, zfs_secpolicy_read, DATASET_NAME,
5720 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5721
5722 zfs_ioctl_register("destroy_bookmarks", ZFS_IOC_DESTROY_BOOKMARKS,
5723 zfs_ioc_destroy_bookmarks, zfs_secpolicy_destroy_bookmarks,
5724 POOL_NAME,
5725 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5726
5727 /* IOCTLS that use the legacy function signature */
5728
5729 zfs_ioctl_register_legacy(ZFS_IOC_POOL_FREEZE, zfs_ioc_pool_freeze,
5730 zfs_secpolicy_config, NO_NAME, B_FALSE, POOL_CHECK_READONLY);
5731
5732 zfs_ioctl_register_pool(ZFS_IOC_POOL_CREATE, zfs_ioc_pool_create,
5733 zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE);
5734 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SCAN,
5735 zfs_ioc_pool_scan);
5736 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_UPGRADE,
5737 zfs_ioc_pool_upgrade);
5738 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ADD,
5739 zfs_ioc_vdev_add);
5740 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_REMOVE,
5741 zfs_ioc_vdev_remove);
5742 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SET_STATE,
5743 zfs_ioc_vdev_set_state);
5744 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ATTACH,
5745 zfs_ioc_vdev_attach);
5746 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_DETACH,
5747 zfs_ioc_vdev_detach);
5748 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETPATH,
5749 zfs_ioc_vdev_setpath);
5750 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETFRU,
5751 zfs_ioc_vdev_setfru);
5752 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SET_PROPS,
5753 zfs_ioc_pool_set_props);
5754 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SPLIT,
5755 zfs_ioc_vdev_split);
5756 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_REGUID,
5757 zfs_ioc_pool_reguid);
5758
5759 zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_CONFIGS,
5760 zfs_ioc_pool_configs, zfs_secpolicy_none);
5761 zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_TRYIMPORT,
5762 zfs_ioc_pool_tryimport, zfs_secpolicy_config);
5763 zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_FAULT,
5764 zfs_ioc_inject_fault, zfs_secpolicy_inject);
5765 zfs_ioctl_register_pool_meta(ZFS_IOC_CLEAR_FAULT,
5766 zfs_ioc_clear_fault, zfs_secpolicy_inject);
5767 zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_LIST_NEXT,
5768 zfs_ioc_inject_list_next, zfs_secpolicy_inject);
5769
5770 /*
5771 * pool destroy, and export don't log the history as part of
5772 * zfsdev_ioctl, but rather zfs_ioc_pool_export
5773 * does the logging of those commands.
5774 */
5775 zfs_ioctl_register_pool(ZFS_IOC_POOL_DESTROY, zfs_ioc_pool_destroy,
5776 zfs_secpolicy_config, B_FALSE, POOL_CHECK_NONE);
5777 zfs_ioctl_register_pool(ZFS_IOC_POOL_EXPORT, zfs_ioc_pool_export,
5778 zfs_secpolicy_config, B_FALSE, POOL_CHECK_NONE);
5779
5780 zfs_ioctl_register_pool(ZFS_IOC_POOL_STATS, zfs_ioc_pool_stats,
5781 zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE);
5782 zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_PROPS, zfs_ioc_pool_get_props,
5783 zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE);
5784
5785 zfs_ioctl_register_pool(ZFS_IOC_ERROR_LOG, zfs_ioc_error_log,
5786 zfs_secpolicy_inject, B_FALSE, POOL_CHECK_SUSPENDED);
5787 zfs_ioctl_register_pool(ZFS_IOC_DSOBJ_TO_DSNAME,
5788 zfs_ioc_dsobj_to_dsname,
5789 zfs_secpolicy_diff, B_FALSE, POOL_CHECK_SUSPENDED);
5790 zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_HISTORY,
5791 zfs_ioc_pool_get_history,
5792 zfs_secpolicy_config, B_FALSE, POOL_CHECK_SUSPENDED);
5793
5794 zfs_ioctl_register_pool(ZFS_IOC_POOL_IMPORT, zfs_ioc_pool_import,
5795 zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE);
5796
5797 zfs_ioctl_register_pool(ZFS_IOC_CLEAR, zfs_ioc_clear,
5798 zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE);
5799 zfs_ioctl_register_pool(ZFS_IOC_POOL_REOPEN, zfs_ioc_pool_reopen,
5800 zfs_secpolicy_config, B_TRUE, POOL_CHECK_SUSPENDED);
5801
5802 zfs_ioctl_register_dataset_read(ZFS_IOC_SPACE_WRITTEN,
5803 zfs_ioc_space_written);
5804 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_RECVD_PROPS,
5805 zfs_ioc_objset_recvd_props);
5806 zfs_ioctl_register_dataset_read(ZFS_IOC_NEXT_OBJ,
5807 zfs_ioc_next_obj);
5808 zfs_ioctl_register_dataset_read(ZFS_IOC_GET_FSACL,
5809 zfs_ioc_get_fsacl);
5810 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_STATS,
5811 zfs_ioc_objset_stats);
5812 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_ZPLPROPS,
5813 zfs_ioc_objset_zplprops);
5814 zfs_ioctl_register_dataset_read(ZFS_IOC_DATASET_LIST_NEXT,
5815 zfs_ioc_dataset_list_next);
5816 zfs_ioctl_register_dataset_read(ZFS_IOC_SNAPSHOT_LIST_NEXT,
5817 zfs_ioc_snapshot_list_next);
5818 zfs_ioctl_register_dataset_read(ZFS_IOC_SEND_PROGRESS,
5819 zfs_ioc_send_progress);
5820
5821 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_DIFF,
5822 zfs_ioc_diff, zfs_secpolicy_diff);
5823 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_STATS,
5824 zfs_ioc_obj_to_stats, zfs_secpolicy_diff);
5825 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_PATH,
5826 zfs_ioc_obj_to_path, zfs_secpolicy_diff);
5827 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_ONE,
5828 zfs_ioc_userspace_one, zfs_secpolicy_userspace_one);
5829 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_MANY,
5830 zfs_ioc_userspace_many, zfs_secpolicy_userspace_many);
5831 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_SEND,
5832 zfs_ioc_send, zfs_secpolicy_send);
5833
5834 zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_PROP, zfs_ioc_set_prop,
5835 zfs_secpolicy_none);
5836 zfs_ioctl_register_dataset_modify(ZFS_IOC_DESTROY, zfs_ioc_destroy,
5837 zfs_secpolicy_destroy);
5838 zfs_ioctl_register_dataset_modify(ZFS_IOC_RENAME, zfs_ioc_rename,
5839 zfs_secpolicy_rename);
5840 zfs_ioctl_register_dataset_modify(ZFS_IOC_RECV, zfs_ioc_recv,
5841 zfs_secpolicy_recv);
5842 zfs_ioctl_register_dataset_modify(ZFS_IOC_PROMOTE, zfs_ioc_promote,
5843 zfs_secpolicy_promote);
5844 zfs_ioctl_register_dataset_modify(ZFS_IOC_INHERIT_PROP,
5845 zfs_ioc_inherit_prop, zfs_secpolicy_inherit_prop);
5846 zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_FSACL, zfs_ioc_set_fsacl,
5847 zfs_secpolicy_set_fsacl);
5848
5849 zfs_ioctl_register_dataset_nolog(ZFS_IOC_SHARE, zfs_ioc_share,
5850 zfs_secpolicy_share, POOL_CHECK_NONE);
5851 zfs_ioctl_register_dataset_nolog(ZFS_IOC_SMB_ACL, zfs_ioc_smb_acl,
5852 zfs_secpolicy_smb_acl, POOL_CHECK_NONE);
5853 zfs_ioctl_register_dataset_nolog(ZFS_IOC_USERSPACE_UPGRADE,
5854 zfs_ioc_userspace_upgrade, zfs_secpolicy_userspace_upgrade,
5855 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5856 zfs_ioctl_register_dataset_nolog(ZFS_IOC_TMP_SNAPSHOT,
5857 zfs_ioc_tmp_snapshot, zfs_secpolicy_tmp_snapshot,
5858 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5859 }
5860
5861 int
5862 pool_status_check(const char *name, zfs_ioc_namecheck_t type,
5863 zfs_ioc_poolcheck_t check)
5864 {
5865 spa_t *spa;
5866 int error;
5867
5868 ASSERT(type == POOL_NAME || type == DATASET_NAME);
5869
5870 if (check & POOL_CHECK_NONE)
5871 return (0);
5872
5873 error = spa_open(name, &spa, FTAG);
5874 if (error == 0) {
5875 if ((check & POOL_CHECK_SUSPENDED) && spa_suspended(spa))
5876 error = SET_ERROR(EAGAIN);
5877 else if ((check & POOL_CHECK_READONLY) && !spa_writeable(spa))
5878 error = SET_ERROR(EROFS);
5879 spa_close(spa, FTAG);
5880 }
5881 return (error);
5882 }
5883
5884 /*
5885 * Find a free minor number.
5886 */
5887 minor_t
5888 zfsdev_minor_alloc(void)
5889 {
5890 static minor_t last_minor;
5891 minor_t m;
5892
5893 ASSERT(MUTEX_HELD(&zfsdev_state_lock));
5894
5895 for (m = last_minor + 1; m != last_minor; m++) {
5896 if (m > ZFSDEV_MAX_MINOR)
5897 m = 1;
5898 if (ddi_get_soft_state(zfsdev_state, m) == NULL) {
5899 last_minor = m;
5900 return (m);
5901 }
5902 }
5903
5904 return (0);
5905 }
5906
5907 static int
5908 zfs_ctldev_init(dev_t *devp)
5909 {
5910 minor_t minor;
5911 zfs_soft_state_t *zs;
5912
5913 ASSERT(MUTEX_HELD(&zfsdev_state_lock));
5914 ASSERT(getminor(*devp) == 0);
5915
5916 minor = zfsdev_minor_alloc();
5917 if (minor == 0)
5918 return (SET_ERROR(ENXIO));
5919
5920 if (ddi_soft_state_zalloc(zfsdev_state, minor) != DDI_SUCCESS)
5921 return (SET_ERROR(EAGAIN));
5922
5923 *devp = makedevice(getemajor(*devp), minor);
5924
5925 zs = ddi_get_soft_state(zfsdev_state, minor);
5926 zs->zss_type = ZSST_CTLDEV;
5927 zfs_onexit_init((zfs_onexit_t **)&zs->zss_data);
5928
5929 return (0);
5930 }
5931
5932 static void
5933 zfs_ctldev_destroy(zfs_onexit_t *zo, minor_t minor)
5934 {
5935 ASSERT(MUTEX_HELD(&zfsdev_state_lock));
5936
5937 zfs_onexit_destroy(zo);
5938 ddi_soft_state_free(zfsdev_state, minor);
5939 }
5940
5941 void *
5942 zfsdev_get_soft_state(minor_t minor, enum zfs_soft_state_type which)
5943 {
5944 zfs_soft_state_t *zp;
5945
5946 zp = ddi_get_soft_state(zfsdev_state, minor);
5947 if (zp == NULL || zp->zss_type != which)
5948 return (NULL);
5949
5950 return (zp->zss_data);
5951 }
5952
5953 static int
5954 zfsdev_open(dev_t *devp, int flag, int otyp, cred_t *cr)
5955 {
5956 int error = 0;
5957
5958 if (getminor(*devp) != 0)
5959 return (zvol_open(devp, flag, otyp, cr));
5960
5961 /* This is the control device. Allocate a new minor if requested. */
5962 if (flag & FEXCL) {
5963 mutex_enter(&zfsdev_state_lock);
5964 error = zfs_ctldev_init(devp);
5965 mutex_exit(&zfsdev_state_lock);
5966 }
5967
5968 return (error);
5969 }
5970
5971 static int
5972 zfsdev_close(dev_t dev, int flag, int otyp, cred_t *cr)
5973 {
5974 zfs_onexit_t *zo;
5975 minor_t minor = getminor(dev);
5976
5977 if (minor == 0)
5978 return (0);
5979
5980 mutex_enter(&zfsdev_state_lock);
5981 zo = zfsdev_get_soft_state(minor, ZSST_CTLDEV);
5982 if (zo == NULL) {
5983 mutex_exit(&zfsdev_state_lock);
5984 return (zvol_close(dev, flag, otyp, cr));
5985 }
5986 zfs_ctldev_destroy(zo, minor);
5987 mutex_exit(&zfsdev_state_lock);
5988
5989 return (0);
5990 }
5991
5992 static int
5993 zfsdev_ioctl(dev_t dev, int cmd, intptr_t arg, int flag, cred_t *cr, int *rvalp)
5994 {
5995 zfs_cmd_t *zc;
5996 uint_t vecnum;
5997 int error, rc, len;
5998 minor_t minor = getminor(dev);
5999 const zfs_ioc_vec_t *vec;
6000 char *saved_poolname = NULL;
6001 nvlist_t *innvl = NULL;
6002
6003 if (minor != 0 &&
6004 zfsdev_get_soft_state(minor, ZSST_CTLDEV) == NULL)
6005 return (zvol_ioctl(dev, cmd, arg, flag, cr, rvalp));
6006
6007 vecnum = cmd - ZFS_IOC_FIRST;
6008 ASSERT3U(getmajor(dev), ==, ddi_driver_major(zfs_dip));
6009
6010 if (vecnum >= sizeof (zfs_ioc_vec) / sizeof (zfs_ioc_vec[0]))
6011 return (SET_ERROR(EINVAL));
6012 vec = &zfs_ioc_vec[vecnum];
6013
6014 zc = kmem_zalloc(sizeof (zfs_cmd_t), KM_SLEEP);
6015
6016 error = ddi_copyin((void *)arg, zc, sizeof (zfs_cmd_t), flag);
6017 if (error != 0) {
6018 error = SET_ERROR(EFAULT);
6019 goto out;
6020 }
6021
6022 zc->zc_iflags = flag & FKIOCTL;
6023 if (zc->zc_nvlist_src_size != 0) {
6024 error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
6025 zc->zc_iflags, &innvl);
6026 if (error != 0)
6027 goto out;
6028 }
6029
6030 /*
6031 * Ensure that all pool/dataset names are valid before we pass down to
6032 * the lower layers.
6033 */
6034 zc->zc_name[sizeof (zc->zc_name) - 1] = '\0';
6035 switch (vec->zvec_namecheck) {
6036 case POOL_NAME:
6037 if (pool_namecheck(zc->zc_name, NULL, NULL) != 0)
6038 error = SET_ERROR(EINVAL);
6039 else
6040 error = pool_status_check(zc->zc_name,
6041 vec->zvec_namecheck, vec->zvec_pool_check);
6042 break;
6043
6044 case DATASET_NAME:
6045 if (dataset_namecheck(zc->zc_name, NULL, NULL) != 0)
6046 error = SET_ERROR(EINVAL);
6047 else
6048 error = pool_status_check(zc->zc_name,
6049 vec->zvec_namecheck, vec->zvec_pool_check);
6050 break;
6051
6052 case NO_NAME:
6053 break;
6054 }
6055
6056
6057 if (error == 0)
6058 error = vec->zvec_secpolicy(zc, innvl, cr);
6059
6060 if (error != 0)
6061 goto out;
6062
6063 /* legacy ioctls can modify zc_name */
6064 len = strcspn(zc->zc_name, "/@#") + 1;
6065 saved_poolname = kmem_alloc(len, KM_SLEEP);
6066 (void) strlcpy(saved_poolname, zc->zc_name, len);
6067
6068 if (vec->zvec_func != NULL) {
6069 nvlist_t *outnvl;
6070 int puterror = 0;
6071 spa_t *spa;
6072 nvlist_t *lognv = NULL;
6073
6074 ASSERT(vec->zvec_legacy_func == NULL);
6075
6076 /*
6077 * Add the innvl to the lognv before calling the func,
6078 * in case the func changes the innvl.
6079 */
6080 if (vec->zvec_allow_log) {
6081 lognv = fnvlist_alloc();
6082 fnvlist_add_string(lognv, ZPOOL_HIST_IOCTL,
6083 vec->zvec_name);
6084 if (!nvlist_empty(innvl)) {
6085 fnvlist_add_nvlist(lognv, ZPOOL_HIST_INPUT_NVL,
6086 innvl);
6087 }
6088 }
6089
6090 outnvl = fnvlist_alloc();
6091 error = vec->zvec_func(zc->zc_name, innvl, outnvl);
6092
6093 if (error == 0 && vec->zvec_allow_log &&
6094 spa_open(zc->zc_name, &spa, FTAG) == 0) {
6095 if (!nvlist_empty(outnvl)) {
6096 fnvlist_add_nvlist(lognv, ZPOOL_HIST_OUTPUT_NVL,
6097 outnvl);
6098 }
6099 (void) spa_history_log_nvl(spa, lognv);
6100 spa_close(spa, FTAG);
6101 }
6102 fnvlist_free(lognv);
6103
6104 if (!nvlist_empty(outnvl) || zc->zc_nvlist_dst_size != 0) {
6105 int smusherror = 0;
6106 if (vec->zvec_smush_outnvlist) {
6107 smusherror = nvlist_smush(outnvl,
6108 zc->zc_nvlist_dst_size);
6109 }
6110 if (smusherror == 0)
6111 puterror = put_nvlist(zc, outnvl);
6112 }
6113
6114 if (puterror != 0)
6115 error = puterror;
6116
6117 nvlist_free(outnvl);
6118 } else {
6119 error = vec->zvec_legacy_func(zc);
6120 }
6121
6122 out:
6123 nvlist_free(innvl);
6124 rc = ddi_copyout(zc, (void *)arg, sizeof (zfs_cmd_t), flag);
6125 if (error == 0 && rc != 0)
6126 error = SET_ERROR(EFAULT);
6127 if (error == 0 && vec->zvec_allow_log) {
6128 char *s = tsd_get(zfs_allow_log_key);
6129 if (s != NULL)
6130 strfree(s);
6131 (void) tsd_set(zfs_allow_log_key, saved_poolname);
6132 } else {
6133 if (saved_poolname != NULL)
6134 strfree(saved_poolname);
6135 }
6136
6137 kmem_free(zc, sizeof (zfs_cmd_t));
6138 return (error);
6139 }
6140
6141 static int
6142 zfs_attach(dev_info_t *dip, ddi_attach_cmd_t cmd)
6143 {
6144 if (cmd != DDI_ATTACH)
6145 return (DDI_FAILURE);
6146
6147 if (ddi_create_minor_node(dip, "zfs", S_IFCHR, 0,
6148 DDI_PSEUDO, 0) == DDI_FAILURE)
6149 return (DDI_FAILURE);
6150
6151 zfs_dip = dip;
6152
6153 ddi_report_dev(dip);
6154
6155 return (DDI_SUCCESS);
6156 }
6157
6158 static int
6159 zfs_detach(dev_info_t *dip, ddi_detach_cmd_t cmd)
6160 {
6161 if (spa_busy() || zfs_busy() || zvol_busy())
6162 return (DDI_FAILURE);
6163
6164 if (cmd != DDI_DETACH)
6165 return (DDI_FAILURE);
6166
6167 zfs_dip = NULL;
6168
6169 ddi_prop_remove_all(dip);
6170 ddi_remove_minor_node(dip, NULL);
6171
6172 return (DDI_SUCCESS);
6173 }
6174
6175 /*ARGSUSED*/
6176 static int
6177 zfs_info(dev_info_t *dip, ddi_info_cmd_t infocmd, void *arg, void **result)
6178 {
6179 switch (infocmd) {
6180 case DDI_INFO_DEVT2DEVINFO:
6181 *result = zfs_dip;
6182 return (DDI_SUCCESS);
6183
6184 case DDI_INFO_DEVT2INSTANCE:
6185 *result = (void *)0;
6186 return (DDI_SUCCESS);
6187 }
6188
6189 return (DDI_FAILURE);
6190 }
6191
6192 /*
6193 * OK, so this is a little weird.
6194 *
6195 * /dev/zfs is the control node, i.e. minor 0.
6196 * /dev/zvol/[r]dsk/pool/dataset are the zvols, minor > 0.
6197 *
6198 * /dev/zfs has basically nothing to do except serve up ioctls,
6199 * so most of the standard driver entry points are in zvol.c.
6200 */
6201 static struct cb_ops zfs_cb_ops = {
6202 zfsdev_open, /* open */
6203 zfsdev_close, /* close */
6204 zvol_strategy, /* strategy */
6205 nodev, /* print */
6206 zvol_dump, /* dump */
6207 zvol_read, /* read */
6208 zvol_write, /* write */
6209 zfsdev_ioctl, /* ioctl */
6210 nodev, /* devmap */
6211 nodev, /* mmap */
6212 nodev, /* segmap */
6213 nochpoll, /* poll */
6214 ddi_prop_op, /* prop_op */
6215 NULL, /* streamtab */
6216 D_NEW | D_MP | D_64BIT, /* Driver compatibility flag */
6217 CB_REV, /* version */
6218 nodev, /* async read */
6219 nodev, /* async write */
6220 };
6221
6222 static struct dev_ops zfs_dev_ops = {
6223 DEVO_REV, /* version */
6224 0, /* refcnt */
6225 zfs_info, /* info */
6226 nulldev, /* identify */
6227 nulldev, /* probe */
6228 zfs_attach, /* attach */
6229 zfs_detach, /* detach */
6230 nodev, /* reset */
6231 &zfs_cb_ops, /* driver operations */
6232 NULL, /* no bus operations */
6233 NULL, /* power */
6234 ddi_quiesce_not_needed, /* quiesce */
6235 };
6236
6237 static struct modldrv zfs_modldrv = {
6238 &mod_driverops,
6239 "ZFS storage pool",
6240 &zfs_dev_ops
6241 };
6242
6243 static struct modlinkage modlinkage = {
6244 MODREV_1,
6245 (void *)&zfs_modlfs,
6246 (void *)&zfs_modldrv,
6247 NULL
6248 };
6249
6250 static void
6251 zfs_allow_log_destroy(void *arg)
6252 {
6253 char *poolname = arg;
6254 strfree(poolname);
6255 }
6256
6257 int
6258 _init(void)
6259 {
6260 int error;
6261
6262 spa_init(FREAD | FWRITE);
6263 zfs_init();
6264 zvol_init();
6265 zfs_ioctl_init();
6266
6267 if ((error = mod_install(&modlinkage)) != 0) {
6268 zvol_fini();
6269 zfs_fini();
6270 spa_fini();
6271 return (error);
6272 }
6273
6274 tsd_create(&zfs_fsyncer_key, NULL);
6275 tsd_create(&rrw_tsd_key, rrw_tsd_destroy);
6276 tsd_create(&zfs_allow_log_key, zfs_allow_log_destroy);
6277
6278 error = ldi_ident_from_mod(&modlinkage, &zfs_li);
6279 ASSERT(error == 0);
6280 mutex_init(&zfs_share_lock, NULL, MUTEX_DEFAULT, NULL);
6281
6282 return (0);
6283 }
6284
6285 int
6286 _fini(void)
6287 {
6288 int error;
6289
6290 if (spa_busy() || zfs_busy() || zvol_busy() || zio_injection_enabled)
6291 return (SET_ERROR(EBUSY));
6292
6293 if ((error = mod_remove(&modlinkage)) != 0)
6294 return (error);
6295
6296 zvol_fini();
6297 zfs_fini();
6298 spa_fini();
6299 if (zfs_nfsshare_inited)
6300 (void) ddi_modclose(nfs_mod);
6301 if (zfs_smbshare_inited)
6302 (void) ddi_modclose(smbsrv_mod);
6303 if (zfs_nfsshare_inited || zfs_smbshare_inited)
6304 (void) ddi_modclose(sharefs_mod);
6305
6306 tsd_destroy(&zfs_fsyncer_key);
6307 ldi_ident_release(zfs_li);
6308 zfs_li = NULL;
6309 mutex_destroy(&zfs_share_lock);
6310
6311 return (error);
6312 }
6313
6314 int
6315 _info(struct modinfo *modinfop)
6316 {
6317 return (mod_info(&modlinkage, modinfop));
6318 }