Print this page
5882 Temporary pool names
Reviewed by: Matt Ahrens <matt@delphix.com>
Reviewed by: Igor Kozhukhov <igor@dilos.org>
Reviewed by: John Kennedy <john.kennedy@delphix.com>
Approved by: Dan McDonald <danmcd@joyent.com>
| Split |
Close |
| Expand all |
| Collapse all |
--- old/usr/src/uts/common/fs/zfs/zfs_ioctl.c
+++ new/usr/src/uts/common/fs/zfs/zfs_ioctl.c
1 1 /*
2 2 * CDDL HEADER START
3 3 *
4 4 * The contents of this file are subject to the terms of the
5 5 * Common Development and Distribution License (the "License").
6 6 * You may not use this file except in compliance with the License.
7 7 *
8 8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9 9 * or http://www.opensolaris.org/os/licensing.
10 10 * See the License for the specific language governing permissions
11 11 * and limitations under the License.
12 12 *
13 13 * When distributing Covered Code, include this CDDL HEADER in each
14 14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15 15 * If applicable, add the following below this CDDL HEADER, with the
16 16 * fields enclosed by brackets "[]" replaced with your own identifying
17 17 * information: Portions Copyright [yyyy] [name of copyright owner]
18 18 *
19 19 * CDDL HEADER END
20 20 */
21 21
22 22 /*
23 23 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
24 24 * Copyright (c) 2011-2012 Pawel Jakub Dawidek. All rights reserved.
25 25 * Portions Copyright 2011 Martin Matuska
26 26 * Copyright 2015, OmniTI Computer Consulting, Inc. All rights reserved.
27 27 * Copyright 2015 Nexenta Systems, Inc. All rights reserved.
28 28 * Copyright (c) 2014, 2016 Joyent, Inc. All rights reserved.
29 29 * Copyright (c) 2011, 2017 by Delphix. All rights reserved.
30 30 * Copyright (c) 2013 by Saso Kiselkov. All rights reserved.
31 31 * Copyright (c) 2013 Steven Hartland. All rights reserved.
32 32 * Copyright (c) 2014 Integros [integros.com]
33 33 * Copyright 2016 Toomas Soome <tsoome@me.com>
34 34 * Copyright 2017 RackTop Systems.
35 35 * Copyright (c) 2017 Datto Inc.
36 36 */
37 37
38 38 /*
39 39 * ZFS ioctls.
40 40 *
41 41 * This file handles the ioctls to /dev/zfs, used for configuring ZFS storage
42 42 * pools and filesystems, e.g. with /sbin/zfs and /sbin/zpool.
43 43 *
44 44 * There are two ways that we handle ioctls: the legacy way where almost
45 45 * all of the logic is in the ioctl callback, and the new way where most
46 46 * of the marshalling is handled in the common entry point, zfsdev_ioctl().
47 47 *
48 48 * Non-legacy ioctls should be registered by calling
49 49 * zfs_ioctl_register() from zfs_ioctl_init(). The ioctl is invoked
50 50 * from userland by lzc_ioctl().
51 51 *
52 52 * The registration arguments are as follows:
53 53 *
54 54 * const char *name
55 55 * The name of the ioctl. This is used for history logging. If the
56 56 * ioctl returns successfully (the callback returns 0), and allow_log
57 57 * is true, then a history log entry will be recorded with the input &
58 58 * output nvlists. The log entry can be printed with "zpool history -i".
59 59 *
60 60 * zfs_ioc_t ioc
61 61 * The ioctl request number, which userland will pass to ioctl(2).
62 62 * The ioctl numbers can change from release to release, because
63 63 * the caller (libzfs) must be matched to the kernel.
64 64 *
65 65 * zfs_secpolicy_func_t *secpolicy
66 66 * This function will be called before the zfs_ioc_func_t, to
67 67 * determine if this operation is permitted. It should return EPERM
68 68 * on failure, and 0 on success. Checks include determining if the
69 69 * dataset is visible in this zone, and if the user has either all
70 70 * zfs privileges in the zone (SYS_MOUNT), or has been granted permission
71 71 * to do this operation on this dataset with "zfs allow".
72 72 *
73 73 * zfs_ioc_namecheck_t namecheck
74 74 * This specifies what to expect in the zfs_cmd_t:zc_name -- a pool
75 75 * name, a dataset name, or nothing. If the name is not well-formed,
76 76 * the ioctl will fail and the callback will not be called.
77 77 * Therefore, the callback can assume that the name is well-formed
78 78 * (e.g. is null-terminated, doesn't have more than one '@' character,
79 79 * doesn't have invalid characters).
80 80 *
81 81 * zfs_ioc_poolcheck_t pool_check
82 82 * This specifies requirements on the pool state. If the pool does
83 83 * not meet them (is suspended or is readonly), the ioctl will fail
84 84 * and the callback will not be called. If any checks are specified
85 85 * (i.e. it is not POOL_CHECK_NONE), namecheck must not be NO_NAME.
86 86 * Multiple checks can be or-ed together (e.g. POOL_CHECK_SUSPENDED |
87 87 * POOL_CHECK_READONLY).
88 88 *
89 89 * boolean_t smush_outnvlist
90 90 * If smush_outnvlist is true, then the output is presumed to be a
91 91 * list of errors, and it will be "smushed" down to fit into the
92 92 * caller's buffer, by removing some entries and replacing them with a
93 93 * single "N_MORE_ERRORS" entry indicating how many were removed. See
94 94 * nvlist_smush() for details. If smush_outnvlist is false, and the
95 95 * outnvlist does not fit into the userland-provided buffer, then the
96 96 * ioctl will fail with ENOMEM.
97 97 *
98 98 * zfs_ioc_func_t *func
99 99 * The callback function that will perform the operation.
100 100 *
101 101 * The callback should return 0 on success, or an error number on
102 102 * failure. If the function fails, the userland ioctl will return -1,
103 103 * and errno will be set to the callback's return value. The callback
104 104 * will be called with the following arguments:
105 105 *
106 106 * const char *name
107 107 * The name of the pool or dataset to operate on, from
108 108 * zfs_cmd_t:zc_name. The 'namecheck' argument specifies the
109 109 * expected type (pool, dataset, or none).
110 110 *
111 111 * nvlist_t *innvl
112 112 * The input nvlist, deserialized from zfs_cmd_t:zc_nvlist_src. Or
113 113 * NULL if no input nvlist was provided. Changes to this nvlist are
114 114 * ignored. If the input nvlist could not be deserialized, the
115 115 * ioctl will fail and the callback will not be called.
116 116 *
117 117 * nvlist_t *outnvl
118 118 * The output nvlist, initially empty. The callback can fill it in,
119 119 * and it will be returned to userland by serializing it into
120 120 * zfs_cmd_t:zc_nvlist_dst. If it is non-empty, and serialization
121 121 * fails (e.g. because the caller didn't supply a large enough
122 122 * buffer), then the overall ioctl will fail. See the
123 123 * 'smush_nvlist' argument above for additional behaviors.
124 124 *
125 125 * There are two typical uses of the output nvlist:
126 126 * - To return state, e.g. property values. In this case,
127 127 * smush_outnvlist should be false. If the buffer was not large
128 128 * enough, the caller will reallocate a larger buffer and try
129 129 * the ioctl again.
130 130 *
131 131 * - To return multiple errors from an ioctl which makes on-disk
132 132 * changes. In this case, smush_outnvlist should be true.
133 133 * Ioctls which make on-disk modifications should generally not
134 134 * use the outnvl if they succeed, because the caller can not
135 135 * distinguish between the operation failing, and
136 136 * deserialization failing.
137 137 */
138 138
139 139 #include <sys/types.h>
140 140 #include <sys/param.h>
141 141 #include <sys/errno.h>
142 142 #include <sys/uio.h>
143 143 #include <sys/buf.h>
144 144 #include <sys/modctl.h>
145 145 #include <sys/open.h>
146 146 #include <sys/file.h>
147 147 #include <sys/kmem.h>
148 148 #include <sys/conf.h>
149 149 #include <sys/cmn_err.h>
150 150 #include <sys/stat.h>
151 151 #include <sys/zfs_ioctl.h>
152 152 #include <sys/zfs_vfsops.h>
153 153 #include <sys/zfs_znode.h>
154 154 #include <sys/zap.h>
155 155 #include <sys/spa.h>
156 156 #include <sys/spa_impl.h>
157 157 #include <sys/vdev.h>
158 158 #include <sys/priv_impl.h>
159 159 #include <sys/dmu.h>
160 160 #include <sys/dsl_dir.h>
161 161 #include <sys/dsl_dataset.h>
162 162 #include <sys/dsl_prop.h>
163 163 #include <sys/dsl_deleg.h>
164 164 #include <sys/dmu_objset.h>
165 165 #include <sys/dmu_impl.h>
166 166 #include <sys/dmu_tx.h>
167 167 #include <sys/ddi.h>
168 168 #include <sys/sunddi.h>
169 169 #include <sys/sunldi.h>
170 170 #include <sys/policy.h>
171 171 #include <sys/zone.h>
172 172 #include <sys/nvpair.h>
173 173 #include <sys/pathname.h>
174 174 #include <sys/mount.h>
175 175 #include <sys/sdt.h>
176 176 #include <sys/fs/zfs.h>
177 177 #include <sys/zfs_ctldir.h>
178 178 #include <sys/zfs_dir.h>
179 179 #include <sys/zfs_onexit.h>
180 180 #include <sys/zvol.h>
181 181 #include <sys/dsl_scan.h>
182 182 #include <sharefs/share.h>
183 183 #include <sys/dmu_objset.h>
184 184 #include <sys/dmu_send.h>
185 185 #include <sys/dsl_destroy.h>
186 186 #include <sys/dsl_bookmark.h>
187 187 #include <sys/dsl_userhold.h>
188 188 #include <sys/zfeature.h>
189 189 #include <sys/zcp.h>
190 190 #include <sys/zio_checksum.h>
191 191 #include <sys/vdev_removal.h>
192 192 #include <sys/vdev_impl.h>
193 193 #include <sys/vdev_initialize.h>
194 194
195 195 #include "zfs_namecheck.h"
196 196 #include "zfs_prop.h"
197 197 #include "zfs_deleg.h"
198 198 #include "zfs_comutil.h"
199 199
200 200 #include "lua.h"
201 201 #include "lauxlib.h"
202 202
203 203 extern struct modlfs zfs_modlfs;
204 204
205 205 extern void zfs_init(void);
206 206 extern void zfs_fini(void);
207 207
208 208 ldi_ident_t zfs_li = NULL;
209 209 dev_info_t *zfs_dip;
210 210
211 211 uint_t zfs_fsyncer_key;
212 212 extern uint_t rrw_tsd_key;
213 213 static uint_t zfs_allow_log_key;
214 214
215 215 typedef int zfs_ioc_legacy_func_t(zfs_cmd_t *);
216 216 typedef int zfs_ioc_func_t(const char *, nvlist_t *, nvlist_t *);
217 217 typedef int zfs_secpolicy_func_t(zfs_cmd_t *, nvlist_t *, cred_t *);
218 218
219 219 typedef enum {
220 220 NO_NAME,
221 221 POOL_NAME,
222 222 DATASET_NAME
223 223 } zfs_ioc_namecheck_t;
224 224
225 225 typedef enum {
226 226 POOL_CHECK_NONE = 1 << 0,
227 227 POOL_CHECK_SUSPENDED = 1 << 1,
228 228 POOL_CHECK_READONLY = 1 << 2,
229 229 } zfs_ioc_poolcheck_t;
230 230
231 231 typedef struct zfs_ioc_vec {
232 232 zfs_ioc_legacy_func_t *zvec_legacy_func;
233 233 zfs_ioc_func_t *zvec_func;
234 234 zfs_secpolicy_func_t *zvec_secpolicy;
235 235 zfs_ioc_namecheck_t zvec_namecheck;
236 236 boolean_t zvec_allow_log;
237 237 zfs_ioc_poolcheck_t zvec_pool_check;
238 238 boolean_t zvec_smush_outnvlist;
239 239 const char *zvec_name;
240 240 } zfs_ioc_vec_t;
241 241
242 242 /* This array is indexed by zfs_userquota_prop_t */
243 243 static const char *userquota_perms[] = {
244 244 ZFS_DELEG_PERM_USERUSED,
245 245 ZFS_DELEG_PERM_USERQUOTA,
246 246 ZFS_DELEG_PERM_GROUPUSED,
247 247 ZFS_DELEG_PERM_GROUPQUOTA,
248 248 };
249 249
250 250 static int zfs_ioc_userspace_upgrade(zfs_cmd_t *zc);
251 251 static int zfs_check_settable(const char *name, nvpair_t *property,
252 252 cred_t *cr);
253 253 static int zfs_check_clearable(char *dataset, nvlist_t *props,
254 254 nvlist_t **errors);
255 255 static int zfs_fill_zplprops_root(uint64_t, nvlist_t *, nvlist_t *,
256 256 boolean_t *);
257 257 int zfs_set_prop_nvlist(const char *, zprop_source_t, nvlist_t *, nvlist_t *);
258 258 static int get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp);
259 259
260 260 static int zfs_prop_activate_feature(spa_t *spa, spa_feature_t feature);
261 261
262 262 /* _NOTE(PRINTFLIKE(4)) - this is printf-like, but lint is too whiney */
263 263 void
264 264 __dprintf(const char *file, const char *func, int line, const char *fmt, ...)
265 265 {
266 266 const char *newfile;
267 267 char buf[512];
268 268 va_list adx;
269 269
270 270 /*
271 271 * Get rid of annoying "../common/" prefix to filename.
272 272 */
273 273 newfile = strrchr(file, '/');
274 274 if (newfile != NULL) {
275 275 newfile = newfile + 1; /* Get rid of leading / */
276 276 } else {
277 277 newfile = file;
278 278 }
279 279
280 280 va_start(adx, fmt);
281 281 (void) vsnprintf(buf, sizeof (buf), fmt, adx);
282 282 va_end(adx);
283 283
284 284 /*
285 285 * To get this data, use the zfs-dprintf probe as so:
286 286 * dtrace -q -n 'zfs-dprintf \
287 287 * /stringof(arg0) == "dbuf.c"/ \
288 288 * {printf("%s: %s", stringof(arg1), stringof(arg3))}'
289 289 * arg0 = file name
290 290 * arg1 = function name
291 291 * arg2 = line number
292 292 * arg3 = message
293 293 */
294 294 DTRACE_PROBE4(zfs__dprintf,
295 295 char *, newfile, char *, func, int, line, char *, buf);
296 296 }
297 297
298 298 static void
299 299 history_str_free(char *buf)
300 300 {
301 301 kmem_free(buf, HIS_MAX_RECORD_LEN);
302 302 }
303 303
304 304 static char *
305 305 history_str_get(zfs_cmd_t *zc)
306 306 {
307 307 char *buf;
308 308
309 309 if (zc->zc_history == NULL)
310 310 return (NULL);
311 311
312 312 buf = kmem_alloc(HIS_MAX_RECORD_LEN, KM_SLEEP);
313 313 if (copyinstr((void *)(uintptr_t)zc->zc_history,
314 314 buf, HIS_MAX_RECORD_LEN, NULL) != 0) {
315 315 history_str_free(buf);
316 316 return (NULL);
317 317 }
318 318
319 319 buf[HIS_MAX_RECORD_LEN -1] = '\0';
320 320
321 321 return (buf);
322 322 }
323 323
324 324 /*
325 325 * Check to see if the named dataset is currently defined as bootable
326 326 */
327 327 static boolean_t
328 328 zfs_is_bootfs(const char *name)
329 329 {
330 330 objset_t *os;
331 331
332 332 if (dmu_objset_hold(name, FTAG, &os) == 0) {
333 333 boolean_t ret;
334 334 ret = (dmu_objset_id(os) == spa_bootfs(dmu_objset_spa(os)));
335 335 dmu_objset_rele(os, FTAG);
336 336 return (ret);
337 337 }
338 338 return (B_FALSE);
339 339 }
340 340
341 341 /*
342 342 * Return non-zero if the spa version is less than requested version.
343 343 */
344 344 static int
345 345 zfs_earlier_version(const char *name, int version)
346 346 {
347 347 spa_t *spa;
348 348
349 349 if (spa_open(name, &spa, FTAG) == 0) {
350 350 if (spa_version(spa) < version) {
351 351 spa_close(spa, FTAG);
352 352 return (1);
353 353 }
354 354 spa_close(spa, FTAG);
355 355 }
356 356 return (0);
357 357 }
358 358
359 359 /*
360 360 * Return TRUE if the ZPL version is less than requested version.
361 361 */
362 362 static boolean_t
363 363 zpl_earlier_version(const char *name, int version)
364 364 {
365 365 objset_t *os;
366 366 boolean_t rc = B_TRUE;
367 367
368 368 if (dmu_objset_hold(name, FTAG, &os) == 0) {
369 369 uint64_t zplversion;
370 370
371 371 if (dmu_objset_type(os) != DMU_OST_ZFS) {
372 372 dmu_objset_rele(os, FTAG);
373 373 return (B_TRUE);
374 374 }
375 375 /* XXX reading from non-owned objset */
376 376 if (zfs_get_zplprop(os, ZFS_PROP_VERSION, &zplversion) == 0)
377 377 rc = zplversion < version;
378 378 dmu_objset_rele(os, FTAG);
379 379 }
380 380 return (rc);
381 381 }
382 382
383 383 static void
384 384 zfs_log_history(zfs_cmd_t *zc)
385 385 {
386 386 spa_t *spa;
387 387 char *buf;
388 388
389 389 if ((buf = history_str_get(zc)) == NULL)
390 390 return;
391 391
392 392 if (spa_open(zc->zc_name, &spa, FTAG) == 0) {
393 393 if (spa_version(spa) >= SPA_VERSION_ZPOOL_HISTORY)
394 394 (void) spa_history_log(spa, buf);
395 395 spa_close(spa, FTAG);
396 396 }
397 397 history_str_free(buf);
398 398 }
399 399
400 400 /*
401 401 * Policy for top-level read operations (list pools). Requires no privileges,
402 402 * and can be used in the local zone, as there is no associated dataset.
403 403 */
404 404 /* ARGSUSED */
405 405 static int
406 406 zfs_secpolicy_none(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
407 407 {
408 408 return (0);
409 409 }
410 410
411 411 /*
412 412 * Policy for dataset read operations (list children, get statistics). Requires
413 413 * no privileges, but must be visible in the local zone.
414 414 */
415 415 /* ARGSUSED */
416 416 static int
417 417 zfs_secpolicy_read(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
418 418 {
419 419 if (INGLOBALZONE(curproc) ||
420 420 zone_dataset_visible(zc->zc_name, NULL))
421 421 return (0);
422 422
423 423 return (SET_ERROR(ENOENT));
424 424 }
425 425
426 426 static int
427 427 zfs_dozonecheck_impl(const char *dataset, uint64_t zoned, cred_t *cr)
428 428 {
429 429 int writable = 1;
430 430
431 431 /*
432 432 * The dataset must be visible by this zone -- check this first
433 433 * so they don't see EPERM on something they shouldn't know about.
434 434 */
435 435 if (!INGLOBALZONE(curproc) &&
436 436 !zone_dataset_visible(dataset, &writable))
437 437 return (SET_ERROR(ENOENT));
438 438
439 439 if (INGLOBALZONE(curproc)) {
440 440 /*
441 441 * If the fs is zoned, only root can access it from the
442 442 * global zone.
443 443 */
444 444 if (secpolicy_zfs(cr) && zoned)
445 445 return (SET_ERROR(EPERM));
446 446 } else {
447 447 /*
448 448 * If we are in a local zone, the 'zoned' property must be set.
449 449 */
450 450 if (!zoned)
451 451 return (SET_ERROR(EPERM));
452 452
453 453 /* must be writable by this zone */
454 454 if (!writable)
455 455 return (SET_ERROR(EPERM));
456 456 }
457 457 return (0);
458 458 }
459 459
460 460 static int
461 461 zfs_dozonecheck(const char *dataset, cred_t *cr)
462 462 {
463 463 uint64_t zoned;
464 464
465 465 if (dsl_prop_get_integer(dataset, "zoned", &zoned, NULL))
466 466 return (SET_ERROR(ENOENT));
467 467
468 468 return (zfs_dozonecheck_impl(dataset, zoned, cr));
469 469 }
470 470
471 471 static int
472 472 zfs_dozonecheck_ds(const char *dataset, dsl_dataset_t *ds, cred_t *cr)
473 473 {
474 474 uint64_t zoned;
475 475
476 476 if (dsl_prop_get_int_ds(ds, "zoned", &zoned))
477 477 return (SET_ERROR(ENOENT));
478 478
479 479 return (zfs_dozonecheck_impl(dataset, zoned, cr));
480 480 }
481 481
482 482 static int
483 483 zfs_secpolicy_write_perms_ds(const char *name, dsl_dataset_t *ds,
484 484 const char *perm, cred_t *cr)
485 485 {
486 486 int error;
487 487
488 488 error = zfs_dozonecheck_ds(name, ds, cr);
489 489 if (error == 0) {
490 490 error = secpolicy_zfs(cr);
491 491 if (error != 0)
492 492 error = dsl_deleg_access_impl(ds, perm, cr);
493 493 }
494 494 return (error);
495 495 }
496 496
497 497 static int
498 498 zfs_secpolicy_write_perms(const char *name, const char *perm, cred_t *cr)
499 499 {
500 500 int error;
501 501 dsl_dataset_t *ds;
502 502 dsl_pool_t *dp;
503 503
504 504 /*
505 505 * First do a quick check for root in the global zone, which
506 506 * is allowed to do all write_perms. This ensures that zfs_ioc_*
507 507 * will get to handle nonexistent datasets.
508 508 */
509 509 if (INGLOBALZONE(curproc) && secpolicy_zfs(cr) == 0)
510 510 return (0);
511 511
512 512 error = dsl_pool_hold(name, FTAG, &dp);
513 513 if (error != 0)
514 514 return (error);
515 515
516 516 error = dsl_dataset_hold(dp, name, FTAG, &ds);
517 517 if (error != 0) {
518 518 dsl_pool_rele(dp, FTAG);
519 519 return (error);
520 520 }
521 521
522 522 error = zfs_secpolicy_write_perms_ds(name, ds, perm, cr);
523 523
524 524 dsl_dataset_rele(ds, FTAG);
525 525 dsl_pool_rele(dp, FTAG);
526 526 return (error);
527 527 }
528 528
529 529 /*
530 530 * Policy for setting the security label property.
531 531 *
532 532 * Returns 0 for success, non-zero for access and other errors.
533 533 */
534 534 static int
535 535 zfs_set_slabel_policy(const char *name, char *strval, cred_t *cr)
536 536 {
537 537 char ds_hexsl[MAXNAMELEN];
538 538 bslabel_t ds_sl, new_sl;
539 539 boolean_t new_default = FALSE;
540 540 uint64_t zoned;
541 541 int needed_priv = -1;
542 542 int error;
543 543
544 544 /* First get the existing dataset label. */
545 545 error = dsl_prop_get(name, zfs_prop_to_name(ZFS_PROP_MLSLABEL),
546 546 1, sizeof (ds_hexsl), &ds_hexsl, NULL);
547 547 if (error != 0)
548 548 return (SET_ERROR(EPERM));
549 549
550 550 if (strcasecmp(strval, ZFS_MLSLABEL_DEFAULT) == 0)
551 551 new_default = TRUE;
552 552
553 553 /* The label must be translatable */
554 554 if (!new_default && (hexstr_to_label(strval, &new_sl) != 0))
555 555 return (SET_ERROR(EINVAL));
556 556
557 557 /*
558 558 * In a non-global zone, disallow attempts to set a label that
559 559 * doesn't match that of the zone; otherwise no other checks
560 560 * are needed.
561 561 */
562 562 if (!INGLOBALZONE(curproc)) {
563 563 if (new_default || !blequal(&new_sl, CR_SL(CRED())))
564 564 return (SET_ERROR(EPERM));
565 565 return (0);
566 566 }
567 567
568 568 /*
569 569 * For global-zone datasets (i.e., those whose zoned property is
570 570 * "off", verify that the specified new label is valid for the
571 571 * global zone.
572 572 */
573 573 if (dsl_prop_get_integer(name,
574 574 zfs_prop_to_name(ZFS_PROP_ZONED), &zoned, NULL))
575 575 return (SET_ERROR(EPERM));
576 576 if (!zoned) {
577 577 if (zfs_check_global_label(name, strval) != 0)
578 578 return (SET_ERROR(EPERM));
579 579 }
580 580
581 581 /*
582 582 * If the existing dataset label is nondefault, check if the
583 583 * dataset is mounted (label cannot be changed while mounted).
584 584 * Get the zfsvfs; if there isn't one, then the dataset isn't
585 585 * mounted (or isn't a dataset, doesn't exist, ...).
586 586 */
587 587 if (strcasecmp(ds_hexsl, ZFS_MLSLABEL_DEFAULT) != 0) {
588 588 objset_t *os;
589 589 static char *setsl_tag = "setsl_tag";
590 590
591 591 /*
592 592 * Try to own the dataset; abort if there is any error,
593 593 * (e.g., already mounted, in use, or other error).
594 594 */
595 595 error = dmu_objset_own(name, DMU_OST_ZFS, B_TRUE,
596 596 setsl_tag, &os);
597 597 if (error != 0)
598 598 return (SET_ERROR(EPERM));
599 599
600 600 dmu_objset_disown(os, setsl_tag);
601 601
602 602 if (new_default) {
603 603 needed_priv = PRIV_FILE_DOWNGRADE_SL;
604 604 goto out_check;
605 605 }
606 606
607 607 if (hexstr_to_label(strval, &new_sl) != 0)
608 608 return (SET_ERROR(EPERM));
609 609
610 610 if (blstrictdom(&ds_sl, &new_sl))
611 611 needed_priv = PRIV_FILE_DOWNGRADE_SL;
612 612 else if (blstrictdom(&new_sl, &ds_sl))
613 613 needed_priv = PRIV_FILE_UPGRADE_SL;
614 614 } else {
615 615 /* dataset currently has a default label */
616 616 if (!new_default)
617 617 needed_priv = PRIV_FILE_UPGRADE_SL;
618 618 }
619 619
620 620 out_check:
621 621 if (needed_priv != -1)
622 622 return (PRIV_POLICY(cr, needed_priv, B_FALSE, EPERM, NULL));
623 623 return (0);
624 624 }
625 625
626 626 static int
627 627 zfs_secpolicy_setprop(const char *dsname, zfs_prop_t prop, nvpair_t *propval,
628 628 cred_t *cr)
629 629 {
630 630 char *strval;
631 631
632 632 /*
633 633 * Check permissions for special properties.
634 634 */
635 635 switch (prop) {
636 636 case ZFS_PROP_ZONED:
637 637 /*
638 638 * Disallow setting of 'zoned' from within a local zone.
639 639 */
640 640 if (!INGLOBALZONE(curproc))
641 641 return (SET_ERROR(EPERM));
642 642 break;
643 643
644 644 case ZFS_PROP_QUOTA:
645 645 case ZFS_PROP_FILESYSTEM_LIMIT:
646 646 case ZFS_PROP_SNAPSHOT_LIMIT:
647 647 if (!INGLOBALZONE(curproc)) {
648 648 uint64_t zoned;
649 649 char setpoint[ZFS_MAX_DATASET_NAME_LEN];
650 650 /*
651 651 * Unprivileged users are allowed to modify the
652 652 * limit on things *under* (ie. contained by)
653 653 * the thing they own.
654 654 */
655 655 if (dsl_prop_get_integer(dsname, "zoned", &zoned,
656 656 setpoint))
657 657 return (SET_ERROR(EPERM));
658 658 if (!zoned || strlen(dsname) <= strlen(setpoint))
659 659 return (SET_ERROR(EPERM));
660 660 }
661 661 break;
662 662
663 663 case ZFS_PROP_MLSLABEL:
664 664 if (!is_system_labeled())
665 665 return (SET_ERROR(EPERM));
666 666
667 667 if (nvpair_value_string(propval, &strval) == 0) {
668 668 int err;
669 669
670 670 err = zfs_set_slabel_policy(dsname, strval, CRED());
671 671 if (err != 0)
672 672 return (err);
673 673 }
674 674 break;
675 675 }
676 676
677 677 return (zfs_secpolicy_write_perms(dsname, zfs_prop_to_name(prop), cr));
678 678 }
679 679
680 680 /* ARGSUSED */
681 681 static int
682 682 zfs_secpolicy_set_fsacl(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
683 683 {
684 684 int error;
685 685
686 686 error = zfs_dozonecheck(zc->zc_name, cr);
687 687 if (error != 0)
688 688 return (error);
689 689
690 690 /*
691 691 * permission to set permissions will be evaluated later in
692 692 * dsl_deleg_can_allow()
693 693 */
694 694 return (0);
695 695 }
696 696
697 697 /* ARGSUSED */
698 698 static int
699 699 zfs_secpolicy_rollback(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
700 700 {
701 701 return (zfs_secpolicy_write_perms(zc->zc_name,
702 702 ZFS_DELEG_PERM_ROLLBACK, cr));
703 703 }
704 704
705 705 /* ARGSUSED */
706 706 static int
707 707 zfs_secpolicy_send(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
708 708 {
709 709 dsl_pool_t *dp;
710 710 dsl_dataset_t *ds;
711 711 char *cp;
712 712 int error;
713 713
714 714 /*
715 715 * Generate the current snapshot name from the given objsetid, then
716 716 * use that name for the secpolicy/zone checks.
717 717 */
718 718 cp = strchr(zc->zc_name, '@');
719 719 if (cp == NULL)
720 720 return (SET_ERROR(EINVAL));
721 721 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
722 722 if (error != 0)
723 723 return (error);
724 724
725 725 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &ds);
726 726 if (error != 0) {
727 727 dsl_pool_rele(dp, FTAG);
728 728 return (error);
729 729 }
730 730
731 731 dsl_dataset_name(ds, zc->zc_name);
732 732
733 733 error = zfs_secpolicy_write_perms_ds(zc->zc_name, ds,
734 734 ZFS_DELEG_PERM_SEND, cr);
735 735 dsl_dataset_rele(ds, FTAG);
736 736 dsl_pool_rele(dp, FTAG);
737 737
738 738 return (error);
739 739 }
740 740
741 741 /* ARGSUSED */
742 742 static int
743 743 zfs_secpolicy_send_new(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
744 744 {
745 745 return (zfs_secpolicy_write_perms(zc->zc_name,
746 746 ZFS_DELEG_PERM_SEND, cr));
747 747 }
748 748
749 749 /* ARGSUSED */
750 750 static int
751 751 zfs_secpolicy_deleg_share(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
752 752 {
753 753 vnode_t *vp;
754 754 int error;
755 755
756 756 if ((error = lookupname(zc->zc_value, UIO_SYSSPACE,
757 757 NO_FOLLOW, NULL, &vp)) != 0)
758 758 return (error);
759 759
760 760 /* Now make sure mntpnt and dataset are ZFS */
761 761
762 762 if (vp->v_vfsp->vfs_fstype != zfsfstype ||
763 763 (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource),
764 764 zc->zc_name) != 0)) {
765 765 VN_RELE(vp);
766 766 return (SET_ERROR(EPERM));
767 767 }
768 768
769 769 VN_RELE(vp);
770 770 return (dsl_deleg_access(zc->zc_name,
771 771 ZFS_DELEG_PERM_SHARE, cr));
772 772 }
773 773
774 774 int
775 775 zfs_secpolicy_share(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
776 776 {
777 777 if (!INGLOBALZONE(curproc))
778 778 return (SET_ERROR(EPERM));
779 779
780 780 if (secpolicy_nfs(cr) == 0) {
781 781 return (0);
782 782 } else {
783 783 return (zfs_secpolicy_deleg_share(zc, innvl, cr));
784 784 }
785 785 }
786 786
787 787 int
788 788 zfs_secpolicy_smb_acl(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
789 789 {
790 790 if (!INGLOBALZONE(curproc))
791 791 return (SET_ERROR(EPERM));
792 792
793 793 if (secpolicy_smb(cr) == 0) {
794 794 return (0);
795 795 } else {
796 796 return (zfs_secpolicy_deleg_share(zc, innvl, cr));
797 797 }
798 798 }
799 799
800 800 static int
801 801 zfs_get_parent(const char *datasetname, char *parent, int parentsize)
802 802 {
803 803 char *cp;
804 804
805 805 /*
806 806 * Remove the @bla or /bla from the end of the name to get the parent.
807 807 */
808 808 (void) strncpy(parent, datasetname, parentsize);
809 809 cp = strrchr(parent, '@');
810 810 if (cp != NULL) {
811 811 cp[0] = '\0';
812 812 } else {
813 813 cp = strrchr(parent, '/');
814 814 if (cp == NULL)
815 815 return (SET_ERROR(ENOENT));
816 816 cp[0] = '\0';
817 817 }
818 818
819 819 return (0);
820 820 }
821 821
822 822 int
823 823 zfs_secpolicy_destroy_perms(const char *name, cred_t *cr)
824 824 {
825 825 int error;
826 826
827 827 if ((error = zfs_secpolicy_write_perms(name,
828 828 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
829 829 return (error);
830 830
831 831 return (zfs_secpolicy_write_perms(name, ZFS_DELEG_PERM_DESTROY, cr));
832 832 }
833 833
834 834 /* ARGSUSED */
835 835 static int
836 836 zfs_secpolicy_destroy(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
837 837 {
838 838 return (zfs_secpolicy_destroy_perms(zc->zc_name, cr));
839 839 }
840 840
841 841 /*
842 842 * Destroying snapshots with delegated permissions requires
843 843 * descendant mount and destroy permissions.
844 844 */
845 845 /* ARGSUSED */
846 846 static int
847 847 zfs_secpolicy_destroy_snaps(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
848 848 {
849 849 nvlist_t *snaps;
850 850 nvpair_t *pair, *nextpair;
851 851 int error = 0;
852 852
853 853 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
854 854 return (SET_ERROR(EINVAL));
855 855 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
856 856 pair = nextpair) {
857 857 nextpair = nvlist_next_nvpair(snaps, pair);
858 858 error = zfs_secpolicy_destroy_perms(nvpair_name(pair), cr);
859 859 if (error == ENOENT) {
860 860 /*
861 861 * Ignore any snapshots that don't exist (we consider
862 862 * them "already destroyed"). Remove the name from the
863 863 * nvl here in case the snapshot is created between
864 864 * now and when we try to destroy it (in which case
865 865 * we don't want to destroy it since we haven't
866 866 * checked for permission).
867 867 */
868 868 fnvlist_remove_nvpair(snaps, pair);
869 869 error = 0;
870 870 }
871 871 if (error != 0)
872 872 break;
873 873 }
874 874
875 875 return (error);
876 876 }
877 877
878 878 int
879 879 zfs_secpolicy_rename_perms(const char *from, const char *to, cred_t *cr)
880 880 {
881 881 char parentname[ZFS_MAX_DATASET_NAME_LEN];
882 882 int error;
883 883
884 884 if ((error = zfs_secpolicy_write_perms(from,
885 885 ZFS_DELEG_PERM_RENAME, cr)) != 0)
886 886 return (error);
887 887
888 888 if ((error = zfs_secpolicy_write_perms(from,
889 889 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
890 890 return (error);
891 891
892 892 if ((error = zfs_get_parent(to, parentname,
893 893 sizeof (parentname))) != 0)
894 894 return (error);
895 895
896 896 if ((error = zfs_secpolicy_write_perms(parentname,
897 897 ZFS_DELEG_PERM_CREATE, cr)) != 0)
898 898 return (error);
899 899
900 900 if ((error = zfs_secpolicy_write_perms(parentname,
901 901 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
902 902 return (error);
903 903
904 904 return (error);
905 905 }
906 906
907 907 /* ARGSUSED */
908 908 static int
909 909 zfs_secpolicy_rename(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
910 910 {
911 911 return (zfs_secpolicy_rename_perms(zc->zc_name, zc->zc_value, cr));
912 912 }
913 913
914 914 /* ARGSUSED */
915 915 static int
916 916 zfs_secpolicy_promote(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
917 917 {
918 918 dsl_pool_t *dp;
919 919 dsl_dataset_t *clone;
920 920 int error;
921 921
922 922 error = zfs_secpolicy_write_perms(zc->zc_name,
923 923 ZFS_DELEG_PERM_PROMOTE, cr);
924 924 if (error != 0)
925 925 return (error);
926 926
927 927 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
928 928 if (error != 0)
929 929 return (error);
930 930
931 931 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &clone);
932 932
933 933 if (error == 0) {
934 934 char parentname[ZFS_MAX_DATASET_NAME_LEN];
935 935 dsl_dataset_t *origin = NULL;
936 936 dsl_dir_t *dd;
937 937 dd = clone->ds_dir;
938 938
939 939 error = dsl_dataset_hold_obj(dd->dd_pool,
940 940 dsl_dir_phys(dd)->dd_origin_obj, FTAG, &origin);
941 941 if (error != 0) {
942 942 dsl_dataset_rele(clone, FTAG);
943 943 dsl_pool_rele(dp, FTAG);
944 944 return (error);
945 945 }
946 946
947 947 error = zfs_secpolicy_write_perms_ds(zc->zc_name, clone,
948 948 ZFS_DELEG_PERM_MOUNT, cr);
949 949
950 950 dsl_dataset_name(origin, parentname);
951 951 if (error == 0) {
952 952 error = zfs_secpolicy_write_perms_ds(parentname, origin,
953 953 ZFS_DELEG_PERM_PROMOTE, cr);
954 954 }
955 955 dsl_dataset_rele(clone, FTAG);
956 956 dsl_dataset_rele(origin, FTAG);
957 957 }
958 958 dsl_pool_rele(dp, FTAG);
959 959 return (error);
960 960 }
961 961
962 962 /* ARGSUSED */
963 963 static int
964 964 zfs_secpolicy_recv(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
965 965 {
966 966 int error;
967 967
968 968 if ((error = zfs_secpolicy_write_perms(zc->zc_name,
969 969 ZFS_DELEG_PERM_RECEIVE, cr)) != 0)
970 970 return (error);
971 971
972 972 if ((error = zfs_secpolicy_write_perms(zc->zc_name,
973 973 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
974 974 return (error);
975 975
976 976 return (zfs_secpolicy_write_perms(zc->zc_name,
977 977 ZFS_DELEG_PERM_CREATE, cr));
978 978 }
979 979
980 980 int
981 981 zfs_secpolicy_snapshot_perms(const char *name, cred_t *cr)
982 982 {
983 983 return (zfs_secpolicy_write_perms(name,
984 984 ZFS_DELEG_PERM_SNAPSHOT, cr));
985 985 }
986 986
987 987 /*
988 988 * Check for permission to create each snapshot in the nvlist.
989 989 */
990 990 /* ARGSUSED */
991 991 static int
992 992 zfs_secpolicy_snapshot(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
993 993 {
994 994 nvlist_t *snaps;
995 995 int error = 0;
996 996 nvpair_t *pair;
997 997
998 998 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
999 999 return (SET_ERROR(EINVAL));
1000 1000 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
1001 1001 pair = nvlist_next_nvpair(snaps, pair)) {
1002 1002 char *name = nvpair_name(pair);
1003 1003 char *atp = strchr(name, '@');
1004 1004
1005 1005 if (atp == NULL) {
1006 1006 error = SET_ERROR(EINVAL);
1007 1007 break;
1008 1008 }
1009 1009 *atp = '\0';
1010 1010 error = zfs_secpolicy_snapshot_perms(name, cr);
1011 1011 *atp = '@';
1012 1012 if (error != 0)
1013 1013 break;
1014 1014 }
1015 1015 return (error);
1016 1016 }
1017 1017
1018 1018 /*
1019 1019 * Check for permission to create each snapshot in the nvlist.
1020 1020 */
1021 1021 /* ARGSUSED */
1022 1022 static int
1023 1023 zfs_secpolicy_bookmark(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1024 1024 {
1025 1025 int error = 0;
1026 1026
1027 1027 for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL);
1028 1028 pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) {
1029 1029 char *name = nvpair_name(pair);
1030 1030 char *hashp = strchr(name, '#');
1031 1031
1032 1032 if (hashp == NULL) {
1033 1033 error = SET_ERROR(EINVAL);
1034 1034 break;
1035 1035 }
1036 1036 *hashp = '\0';
1037 1037 error = zfs_secpolicy_write_perms(name,
1038 1038 ZFS_DELEG_PERM_BOOKMARK, cr);
1039 1039 *hashp = '#';
1040 1040 if (error != 0)
1041 1041 break;
1042 1042 }
1043 1043 return (error);
1044 1044 }
1045 1045
1046 1046 /* ARGSUSED */
1047 1047 static int
1048 1048 zfs_secpolicy_remap(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1049 1049 {
1050 1050 return (zfs_secpolicy_write_perms(zc->zc_name,
1051 1051 ZFS_DELEG_PERM_REMAP, cr));
1052 1052 }
1053 1053
1054 1054 /* ARGSUSED */
1055 1055 static int
1056 1056 zfs_secpolicy_destroy_bookmarks(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1057 1057 {
1058 1058 nvpair_t *pair, *nextpair;
1059 1059 int error = 0;
1060 1060
1061 1061 for (pair = nvlist_next_nvpair(innvl, NULL); pair != NULL;
1062 1062 pair = nextpair) {
1063 1063 char *name = nvpair_name(pair);
1064 1064 char *hashp = strchr(name, '#');
1065 1065 nextpair = nvlist_next_nvpair(innvl, pair);
1066 1066
1067 1067 if (hashp == NULL) {
1068 1068 error = SET_ERROR(EINVAL);
1069 1069 break;
1070 1070 }
1071 1071
1072 1072 *hashp = '\0';
1073 1073 error = zfs_secpolicy_write_perms(name,
1074 1074 ZFS_DELEG_PERM_DESTROY, cr);
1075 1075 *hashp = '#';
1076 1076 if (error == ENOENT) {
1077 1077 /*
1078 1078 * Ignore any filesystems that don't exist (we consider
1079 1079 * their bookmarks "already destroyed"). Remove
1080 1080 * the name from the nvl here in case the filesystem
1081 1081 * is created between now and when we try to destroy
1082 1082 * the bookmark (in which case we don't want to
1083 1083 * destroy it since we haven't checked for permission).
1084 1084 */
1085 1085 fnvlist_remove_nvpair(innvl, pair);
1086 1086 error = 0;
1087 1087 }
1088 1088 if (error != 0)
1089 1089 break;
1090 1090 }
1091 1091
1092 1092 return (error);
1093 1093 }
1094 1094
1095 1095 /* ARGSUSED */
1096 1096 static int
1097 1097 zfs_secpolicy_log_history(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1098 1098 {
1099 1099 /*
1100 1100 * Even root must have a proper TSD so that we know what pool
1101 1101 * to log to.
1102 1102 */
1103 1103 if (tsd_get(zfs_allow_log_key) == NULL)
1104 1104 return (SET_ERROR(EPERM));
1105 1105 return (0);
1106 1106 }
1107 1107
1108 1108 static int
1109 1109 zfs_secpolicy_create_clone(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1110 1110 {
1111 1111 char parentname[ZFS_MAX_DATASET_NAME_LEN];
1112 1112 int error;
1113 1113 char *origin;
1114 1114
1115 1115 if ((error = zfs_get_parent(zc->zc_name, parentname,
1116 1116 sizeof (parentname))) != 0)
1117 1117 return (error);
1118 1118
1119 1119 if (nvlist_lookup_string(innvl, "origin", &origin) == 0 &&
1120 1120 (error = zfs_secpolicy_write_perms(origin,
1121 1121 ZFS_DELEG_PERM_CLONE, cr)) != 0)
1122 1122 return (error);
1123 1123
1124 1124 if ((error = zfs_secpolicy_write_perms(parentname,
1125 1125 ZFS_DELEG_PERM_CREATE, cr)) != 0)
1126 1126 return (error);
1127 1127
1128 1128 return (zfs_secpolicy_write_perms(parentname,
1129 1129 ZFS_DELEG_PERM_MOUNT, cr));
1130 1130 }
1131 1131
1132 1132 /*
1133 1133 * Policy for pool operations - create/destroy pools, add vdevs, etc. Requires
1134 1134 * SYS_CONFIG privilege, which is not available in a local zone.
1135 1135 */
1136 1136 /* ARGSUSED */
1137 1137 static int
1138 1138 zfs_secpolicy_config(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1139 1139 {
1140 1140 if (secpolicy_sys_config(cr, B_FALSE) != 0)
1141 1141 return (SET_ERROR(EPERM));
1142 1142
1143 1143 return (0);
1144 1144 }
1145 1145
1146 1146 /*
1147 1147 * Policy for object to name lookups.
1148 1148 */
1149 1149 /* ARGSUSED */
1150 1150 static int
1151 1151 zfs_secpolicy_diff(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1152 1152 {
1153 1153 int error;
1154 1154
1155 1155 if ((error = secpolicy_sys_config(cr, B_FALSE)) == 0)
1156 1156 return (0);
1157 1157
1158 1158 error = zfs_secpolicy_write_perms(zc->zc_name, ZFS_DELEG_PERM_DIFF, cr);
1159 1159 return (error);
1160 1160 }
1161 1161
1162 1162 /*
1163 1163 * Policy for fault injection. Requires all privileges.
1164 1164 */
1165 1165 /* ARGSUSED */
1166 1166 static int
1167 1167 zfs_secpolicy_inject(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1168 1168 {
1169 1169 return (secpolicy_zinject(cr));
1170 1170 }
1171 1171
1172 1172 /* ARGSUSED */
1173 1173 static int
1174 1174 zfs_secpolicy_inherit_prop(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1175 1175 {
1176 1176 zfs_prop_t prop = zfs_name_to_prop(zc->zc_value);
1177 1177
1178 1178 if (prop == ZPROP_INVAL) {
1179 1179 if (!zfs_prop_user(zc->zc_value))
1180 1180 return (SET_ERROR(EINVAL));
1181 1181 return (zfs_secpolicy_write_perms(zc->zc_name,
1182 1182 ZFS_DELEG_PERM_USERPROP, cr));
1183 1183 } else {
1184 1184 return (zfs_secpolicy_setprop(zc->zc_name, prop,
1185 1185 NULL, cr));
1186 1186 }
1187 1187 }
1188 1188
1189 1189 static int
1190 1190 zfs_secpolicy_userspace_one(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1191 1191 {
1192 1192 int err = zfs_secpolicy_read(zc, innvl, cr);
1193 1193 if (err)
1194 1194 return (err);
1195 1195
1196 1196 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
1197 1197 return (SET_ERROR(EINVAL));
1198 1198
1199 1199 if (zc->zc_value[0] == 0) {
1200 1200 /*
1201 1201 * They are asking about a posix uid/gid. If it's
1202 1202 * themself, allow it.
1203 1203 */
1204 1204 if (zc->zc_objset_type == ZFS_PROP_USERUSED ||
1205 1205 zc->zc_objset_type == ZFS_PROP_USERQUOTA) {
1206 1206 if (zc->zc_guid == crgetuid(cr))
1207 1207 return (0);
1208 1208 } else {
1209 1209 if (groupmember(zc->zc_guid, cr))
1210 1210 return (0);
1211 1211 }
1212 1212 }
1213 1213
1214 1214 return (zfs_secpolicy_write_perms(zc->zc_name,
1215 1215 userquota_perms[zc->zc_objset_type], cr));
1216 1216 }
1217 1217
1218 1218 static int
1219 1219 zfs_secpolicy_userspace_many(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1220 1220 {
1221 1221 int err = zfs_secpolicy_read(zc, innvl, cr);
1222 1222 if (err)
1223 1223 return (err);
1224 1224
1225 1225 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
1226 1226 return (SET_ERROR(EINVAL));
1227 1227
1228 1228 return (zfs_secpolicy_write_perms(zc->zc_name,
1229 1229 userquota_perms[zc->zc_objset_type], cr));
1230 1230 }
1231 1231
1232 1232 /* ARGSUSED */
1233 1233 static int
1234 1234 zfs_secpolicy_userspace_upgrade(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1235 1235 {
1236 1236 return (zfs_secpolicy_setprop(zc->zc_name, ZFS_PROP_VERSION,
1237 1237 NULL, cr));
1238 1238 }
1239 1239
1240 1240 /* ARGSUSED */
1241 1241 static int
1242 1242 zfs_secpolicy_hold(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1243 1243 {
1244 1244 nvpair_t *pair;
1245 1245 nvlist_t *holds;
1246 1246 int error;
1247 1247
1248 1248 error = nvlist_lookup_nvlist(innvl, "holds", &holds);
1249 1249 if (error != 0)
1250 1250 return (SET_ERROR(EINVAL));
1251 1251
1252 1252 for (pair = nvlist_next_nvpair(holds, NULL); pair != NULL;
1253 1253 pair = nvlist_next_nvpair(holds, pair)) {
1254 1254 char fsname[ZFS_MAX_DATASET_NAME_LEN];
1255 1255 error = dmu_fsname(nvpair_name(pair), fsname);
1256 1256 if (error != 0)
1257 1257 return (error);
1258 1258 error = zfs_secpolicy_write_perms(fsname,
1259 1259 ZFS_DELEG_PERM_HOLD, cr);
1260 1260 if (error != 0)
1261 1261 return (error);
1262 1262 }
1263 1263 return (0);
1264 1264 }
1265 1265
1266 1266 /* ARGSUSED */
1267 1267 static int
1268 1268 zfs_secpolicy_release(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1269 1269 {
1270 1270 nvpair_t *pair;
1271 1271 int error;
1272 1272
1273 1273 for (pair = nvlist_next_nvpair(innvl, NULL); pair != NULL;
1274 1274 pair = nvlist_next_nvpair(innvl, pair)) {
1275 1275 char fsname[ZFS_MAX_DATASET_NAME_LEN];
1276 1276 error = dmu_fsname(nvpair_name(pair), fsname);
1277 1277 if (error != 0)
1278 1278 return (error);
1279 1279 error = zfs_secpolicy_write_perms(fsname,
1280 1280 ZFS_DELEG_PERM_RELEASE, cr);
1281 1281 if (error != 0)
1282 1282 return (error);
1283 1283 }
1284 1284 return (0);
1285 1285 }
1286 1286
1287 1287 /*
1288 1288 * Policy for allowing temporary snapshots to be taken or released
1289 1289 */
1290 1290 static int
1291 1291 zfs_secpolicy_tmp_snapshot(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1292 1292 {
1293 1293 /*
1294 1294 * A temporary snapshot is the same as a snapshot,
1295 1295 * hold, destroy and release all rolled into one.
1296 1296 * Delegated diff alone is sufficient that we allow this.
1297 1297 */
1298 1298 int error;
1299 1299
1300 1300 if ((error = zfs_secpolicy_write_perms(zc->zc_name,
1301 1301 ZFS_DELEG_PERM_DIFF, cr)) == 0)
1302 1302 return (0);
1303 1303
1304 1304 error = zfs_secpolicy_snapshot_perms(zc->zc_name, cr);
1305 1305 if (error == 0)
1306 1306 error = zfs_secpolicy_hold(zc, innvl, cr);
1307 1307 if (error == 0)
1308 1308 error = zfs_secpolicy_release(zc, innvl, cr);
1309 1309 if (error == 0)
1310 1310 error = zfs_secpolicy_destroy(zc, innvl, cr);
1311 1311 return (error);
1312 1312 }
1313 1313
1314 1314 /*
1315 1315 * Returns the nvlist as specified by the user in the zfs_cmd_t.
1316 1316 */
1317 1317 static int
1318 1318 get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp)
1319 1319 {
1320 1320 char *packed;
1321 1321 int error;
1322 1322 nvlist_t *list = NULL;
1323 1323
1324 1324 /*
1325 1325 * Read in and unpack the user-supplied nvlist.
1326 1326 */
1327 1327 if (size == 0)
1328 1328 return (SET_ERROR(EINVAL));
1329 1329
1330 1330 packed = kmem_alloc(size, KM_SLEEP);
1331 1331
1332 1332 if ((error = ddi_copyin((void *)(uintptr_t)nvl, packed, size,
1333 1333 iflag)) != 0) {
1334 1334 kmem_free(packed, size);
1335 1335 return (SET_ERROR(EFAULT));
1336 1336 }
1337 1337
1338 1338 if ((error = nvlist_unpack(packed, size, &list, 0)) != 0) {
1339 1339 kmem_free(packed, size);
1340 1340 return (error);
1341 1341 }
1342 1342
1343 1343 kmem_free(packed, size);
1344 1344
1345 1345 *nvp = list;
1346 1346 return (0);
1347 1347 }
1348 1348
1349 1349 /*
1350 1350 * Reduce the size of this nvlist until it can be serialized in 'max' bytes.
1351 1351 * Entries will be removed from the end of the nvlist, and one int32 entry
1352 1352 * named "N_MORE_ERRORS" will be added indicating how many entries were
1353 1353 * removed.
1354 1354 */
1355 1355 static int
1356 1356 nvlist_smush(nvlist_t *errors, size_t max)
1357 1357 {
1358 1358 size_t size;
1359 1359
1360 1360 size = fnvlist_size(errors);
1361 1361
1362 1362 if (size > max) {
1363 1363 nvpair_t *more_errors;
1364 1364 int n = 0;
1365 1365
1366 1366 if (max < 1024)
1367 1367 return (SET_ERROR(ENOMEM));
1368 1368
1369 1369 fnvlist_add_int32(errors, ZPROP_N_MORE_ERRORS, 0);
1370 1370 more_errors = nvlist_prev_nvpair(errors, NULL);
1371 1371
1372 1372 do {
1373 1373 nvpair_t *pair = nvlist_prev_nvpair(errors,
1374 1374 more_errors);
1375 1375 fnvlist_remove_nvpair(errors, pair);
1376 1376 n++;
1377 1377 size = fnvlist_size(errors);
1378 1378 } while (size > max);
1379 1379
1380 1380 fnvlist_remove_nvpair(errors, more_errors);
1381 1381 fnvlist_add_int32(errors, ZPROP_N_MORE_ERRORS, n);
1382 1382 ASSERT3U(fnvlist_size(errors), <=, max);
1383 1383 }
1384 1384
1385 1385 return (0);
1386 1386 }
1387 1387
1388 1388 static int
1389 1389 put_nvlist(zfs_cmd_t *zc, nvlist_t *nvl)
1390 1390 {
1391 1391 char *packed = NULL;
1392 1392 int error = 0;
1393 1393 size_t size;
1394 1394
1395 1395 size = fnvlist_size(nvl);
1396 1396
1397 1397 if (size > zc->zc_nvlist_dst_size) {
1398 1398 error = SET_ERROR(ENOMEM);
1399 1399 } else {
1400 1400 packed = fnvlist_pack(nvl, &size);
1401 1401 if (ddi_copyout(packed, (void *)(uintptr_t)zc->zc_nvlist_dst,
1402 1402 size, zc->zc_iflags) != 0)
1403 1403 error = SET_ERROR(EFAULT);
1404 1404 fnvlist_pack_free(packed, size);
1405 1405 }
1406 1406
1407 1407 zc->zc_nvlist_dst_size = size;
1408 1408 zc->zc_nvlist_dst_filled = B_TRUE;
1409 1409 return (error);
1410 1410 }
1411 1411
1412 1412 int
1413 1413 getzfsvfs_impl(objset_t *os, zfsvfs_t **zfvp)
1414 1414 {
1415 1415 int error = 0;
1416 1416 if (dmu_objset_type(os) != DMU_OST_ZFS) {
1417 1417 return (SET_ERROR(EINVAL));
1418 1418 }
1419 1419
1420 1420 mutex_enter(&os->os_user_ptr_lock);
1421 1421 *zfvp = dmu_objset_get_user(os);
1422 1422 if (*zfvp) {
1423 1423 VFS_HOLD((*zfvp)->z_vfs);
1424 1424 } else {
1425 1425 error = SET_ERROR(ESRCH);
1426 1426 }
1427 1427 mutex_exit(&os->os_user_ptr_lock);
1428 1428 return (error);
1429 1429 }
1430 1430
1431 1431 int
1432 1432 getzfsvfs(const char *dsname, zfsvfs_t **zfvp)
1433 1433 {
1434 1434 objset_t *os;
1435 1435 int error;
1436 1436
1437 1437 error = dmu_objset_hold(dsname, FTAG, &os);
1438 1438 if (error != 0)
1439 1439 return (error);
1440 1440
1441 1441 error = getzfsvfs_impl(os, zfvp);
1442 1442 dmu_objset_rele(os, FTAG);
1443 1443 return (error);
1444 1444 }
1445 1445
1446 1446 /*
1447 1447 * Find a zfsvfs_t for a mounted filesystem, or create our own, in which
1448 1448 * case its z_vfs will be NULL, and it will be opened as the owner.
1449 1449 * If 'writer' is set, the z_teardown_lock will be held for RW_WRITER,
1450 1450 * which prevents all vnode ops from running.
1451 1451 */
1452 1452 static int
1453 1453 zfsvfs_hold(const char *name, void *tag, zfsvfs_t **zfvp, boolean_t writer)
1454 1454 {
1455 1455 int error = 0;
1456 1456
1457 1457 if (getzfsvfs(name, zfvp) != 0)
1458 1458 error = zfsvfs_create(name, zfvp);
1459 1459 if (error == 0) {
1460 1460 rrm_enter(&(*zfvp)->z_teardown_lock, (writer) ? RW_WRITER :
1461 1461 RW_READER, tag);
1462 1462 if ((*zfvp)->z_unmounted) {
1463 1463 /*
1464 1464 * XXX we could probably try again, since the unmounting
1465 1465 * thread should be just about to disassociate the
1466 1466 * objset from the zfsvfs.
1467 1467 */
1468 1468 rrm_exit(&(*zfvp)->z_teardown_lock, tag);
1469 1469 return (SET_ERROR(EBUSY));
1470 1470 }
1471 1471 }
1472 1472 return (error);
1473 1473 }
1474 1474
1475 1475 static void
1476 1476 zfsvfs_rele(zfsvfs_t *zfsvfs, void *tag)
1477 1477 {
1478 1478 rrm_exit(&zfsvfs->z_teardown_lock, tag);
1479 1479
1480 1480 if (zfsvfs->z_vfs) {
1481 1481 VFS_RELE(zfsvfs->z_vfs);
1482 1482 } else {
1483 1483 dmu_objset_disown(zfsvfs->z_os, zfsvfs);
1484 1484 zfsvfs_free(zfsvfs);
|
↓ open down ↓ |
1484 lines elided |
↑ open up ↑ |
1485 1485 }
1486 1486 }
1487 1487
1488 1488 static int
1489 1489 zfs_ioc_pool_create(zfs_cmd_t *zc)
1490 1490 {
1491 1491 int error;
1492 1492 nvlist_t *config, *props = NULL;
1493 1493 nvlist_t *rootprops = NULL;
1494 1494 nvlist_t *zplprops = NULL;
1495 + char *spa_name = zc->zc_name;
1495 1496
1496 1497 if (error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1497 1498 zc->zc_iflags, &config))
1498 1499 return (error);
1499 1500
1500 1501 if (zc->zc_nvlist_src_size != 0 && (error =
1501 1502 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1502 1503 zc->zc_iflags, &props))) {
1503 1504 nvlist_free(config);
1504 1505 return (error);
1505 1506 }
1506 1507
1507 1508 if (props) {
1508 1509 nvlist_t *nvl = NULL;
1509 1510 uint64_t version = SPA_VERSION;
1511 + char *tname;
1510 1512
1511 1513 (void) nvlist_lookup_uint64(props,
1512 1514 zpool_prop_to_name(ZPOOL_PROP_VERSION), &version);
1513 1515 if (!SPA_VERSION_IS_SUPPORTED(version)) {
1514 1516 error = SET_ERROR(EINVAL);
1515 1517 goto pool_props_bad;
1516 1518 }
1517 1519 (void) nvlist_lookup_nvlist(props, ZPOOL_ROOTFS_PROPS, &nvl);
1518 1520 if (nvl) {
1519 1521 error = nvlist_dup(nvl, &rootprops, KM_SLEEP);
1520 1522 if (error != 0) {
1521 1523 nvlist_free(config);
|
↓ open down ↓ |
2 lines elided |
↑ open up ↑ |
1522 1524 nvlist_free(props);
1523 1525 return (error);
1524 1526 }
1525 1527 (void) nvlist_remove_all(props, ZPOOL_ROOTFS_PROPS);
1526 1528 }
1527 1529 VERIFY(nvlist_alloc(&zplprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
1528 1530 error = zfs_fill_zplprops_root(version, rootprops,
1529 1531 zplprops, NULL);
1530 1532 if (error != 0)
1531 1533 goto pool_props_bad;
1534 +
1535 + if (nvlist_lookup_string(props,
1536 + zpool_prop_to_name(ZPOOL_PROP_TNAME), &tname) == 0)
1537 + spa_name = tname;
1532 1538 }
1533 1539
1534 1540 error = spa_create(zc->zc_name, config, props, zplprops);
1535 1541
1536 1542 /*
1537 1543 * Set the remaining root properties
1538 1544 */
1539 - if (!error && (error = zfs_set_prop_nvlist(zc->zc_name,
1545 + if (!error && (error = zfs_set_prop_nvlist(spa_name,
1540 1546 ZPROP_SRC_LOCAL, rootprops, NULL)) != 0)
1541 - (void) spa_destroy(zc->zc_name);
1547 + (void) spa_destroy(spa_name);
1542 1548
1543 1549 pool_props_bad:
1544 1550 nvlist_free(rootprops);
1545 1551 nvlist_free(zplprops);
1546 1552 nvlist_free(config);
1547 1553 nvlist_free(props);
1548 1554
1549 1555 return (error);
1550 1556 }
1551 1557
1552 1558 static int
1553 1559 zfs_ioc_pool_destroy(zfs_cmd_t *zc)
1554 1560 {
1555 1561 int error;
1556 1562 zfs_log_history(zc);
1557 1563 error = spa_destroy(zc->zc_name);
1558 1564 if (error == 0)
1559 1565 zvol_remove_minors(zc->zc_name);
1560 1566 return (error);
1561 1567 }
1562 1568
1563 1569 static int
1564 1570 zfs_ioc_pool_import(zfs_cmd_t *zc)
1565 1571 {
1566 1572 nvlist_t *config, *props = NULL;
1567 1573 uint64_t guid;
1568 1574 int error;
1569 1575
1570 1576 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1571 1577 zc->zc_iflags, &config)) != 0)
1572 1578 return (error);
1573 1579
1574 1580 if (zc->zc_nvlist_src_size != 0 && (error =
1575 1581 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1576 1582 zc->zc_iflags, &props))) {
1577 1583 nvlist_free(config);
1578 1584 return (error);
1579 1585 }
1580 1586
1581 1587 if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, &guid) != 0 ||
1582 1588 guid != zc->zc_guid)
1583 1589 error = SET_ERROR(EINVAL);
1584 1590 else
1585 1591 error = spa_import(zc->zc_name, config, props, zc->zc_cookie);
1586 1592
1587 1593 if (zc->zc_nvlist_dst != 0) {
1588 1594 int err;
1589 1595
1590 1596 if ((err = put_nvlist(zc, config)) != 0)
1591 1597 error = err;
1592 1598 }
1593 1599
1594 1600 nvlist_free(config);
1595 1601
1596 1602 nvlist_free(props);
1597 1603
1598 1604 return (error);
1599 1605 }
1600 1606
1601 1607 static int
1602 1608 zfs_ioc_pool_export(zfs_cmd_t *zc)
1603 1609 {
1604 1610 int error;
1605 1611 boolean_t force = (boolean_t)zc->zc_cookie;
1606 1612 boolean_t hardforce = (boolean_t)zc->zc_guid;
1607 1613
1608 1614 zfs_log_history(zc);
1609 1615 error = spa_export(zc->zc_name, NULL, force, hardforce);
1610 1616 if (error == 0)
1611 1617 zvol_remove_minors(zc->zc_name);
1612 1618 return (error);
1613 1619 }
1614 1620
1615 1621 static int
1616 1622 zfs_ioc_pool_configs(zfs_cmd_t *zc)
1617 1623 {
1618 1624 nvlist_t *configs;
1619 1625 int error;
1620 1626
1621 1627 if ((configs = spa_all_configs(&zc->zc_cookie)) == NULL)
1622 1628 return (SET_ERROR(EEXIST));
1623 1629
1624 1630 error = put_nvlist(zc, configs);
1625 1631
1626 1632 nvlist_free(configs);
1627 1633
1628 1634 return (error);
1629 1635 }
1630 1636
1631 1637 /*
1632 1638 * inputs:
1633 1639 * zc_name name of the pool
1634 1640 *
1635 1641 * outputs:
1636 1642 * zc_cookie real errno
1637 1643 * zc_nvlist_dst config nvlist
1638 1644 * zc_nvlist_dst_size size of config nvlist
1639 1645 */
1640 1646 static int
1641 1647 zfs_ioc_pool_stats(zfs_cmd_t *zc)
1642 1648 {
1643 1649 nvlist_t *config;
1644 1650 int error;
1645 1651 int ret = 0;
1646 1652
1647 1653 error = spa_get_stats(zc->zc_name, &config, zc->zc_value,
1648 1654 sizeof (zc->zc_value));
1649 1655
1650 1656 if (config != NULL) {
1651 1657 ret = put_nvlist(zc, config);
1652 1658 nvlist_free(config);
1653 1659
1654 1660 /*
1655 1661 * The config may be present even if 'error' is non-zero.
1656 1662 * In this case we return success, and preserve the real errno
1657 1663 * in 'zc_cookie'.
1658 1664 */
1659 1665 zc->zc_cookie = error;
1660 1666 } else {
1661 1667 ret = error;
1662 1668 }
1663 1669
1664 1670 return (ret);
1665 1671 }
1666 1672
1667 1673 /*
1668 1674 * Try to import the given pool, returning pool stats as appropriate so that
1669 1675 * user land knows which devices are available and overall pool health.
1670 1676 */
1671 1677 static int
1672 1678 zfs_ioc_pool_tryimport(zfs_cmd_t *zc)
1673 1679 {
1674 1680 nvlist_t *tryconfig, *config;
1675 1681 int error;
1676 1682
1677 1683 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1678 1684 zc->zc_iflags, &tryconfig)) != 0)
1679 1685 return (error);
1680 1686
1681 1687 config = spa_tryimport(tryconfig);
1682 1688
1683 1689 nvlist_free(tryconfig);
1684 1690
1685 1691 if (config == NULL)
1686 1692 return (SET_ERROR(EINVAL));
1687 1693
1688 1694 error = put_nvlist(zc, config);
1689 1695 nvlist_free(config);
1690 1696
1691 1697 return (error);
1692 1698 }
1693 1699
1694 1700 /*
1695 1701 * inputs:
1696 1702 * zc_name name of the pool
1697 1703 * zc_cookie scan func (pool_scan_func_t)
1698 1704 * zc_flags scrub pause/resume flag (pool_scrub_cmd_t)
1699 1705 */
1700 1706 static int
1701 1707 zfs_ioc_pool_scan(zfs_cmd_t *zc)
1702 1708 {
1703 1709 spa_t *spa;
1704 1710 int error;
1705 1711
1706 1712 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1707 1713 return (error);
1708 1714
1709 1715 if (zc->zc_flags >= POOL_SCRUB_FLAGS_END)
1710 1716 return (SET_ERROR(EINVAL));
1711 1717
1712 1718 if (zc->zc_flags == POOL_SCRUB_PAUSE)
1713 1719 error = spa_scrub_pause_resume(spa, POOL_SCRUB_PAUSE);
1714 1720 else if (zc->zc_cookie == POOL_SCAN_NONE)
1715 1721 error = spa_scan_stop(spa);
1716 1722 else
1717 1723 error = spa_scan(spa, zc->zc_cookie);
1718 1724
1719 1725 spa_close(spa, FTAG);
1720 1726
1721 1727 return (error);
1722 1728 }
1723 1729
1724 1730 static int
1725 1731 zfs_ioc_pool_freeze(zfs_cmd_t *zc)
1726 1732 {
1727 1733 spa_t *spa;
1728 1734 int error;
1729 1735
1730 1736 error = spa_open(zc->zc_name, &spa, FTAG);
1731 1737 if (error == 0) {
1732 1738 spa_freeze(spa);
1733 1739 spa_close(spa, FTAG);
1734 1740 }
1735 1741 return (error);
1736 1742 }
1737 1743
1738 1744 static int
1739 1745 zfs_ioc_pool_upgrade(zfs_cmd_t *zc)
1740 1746 {
1741 1747 spa_t *spa;
1742 1748 int error;
1743 1749
1744 1750 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1745 1751 return (error);
1746 1752
1747 1753 if (zc->zc_cookie < spa_version(spa) ||
1748 1754 !SPA_VERSION_IS_SUPPORTED(zc->zc_cookie)) {
1749 1755 spa_close(spa, FTAG);
1750 1756 return (SET_ERROR(EINVAL));
1751 1757 }
1752 1758
1753 1759 spa_upgrade(spa, zc->zc_cookie);
1754 1760 spa_close(spa, FTAG);
1755 1761
1756 1762 return (error);
1757 1763 }
1758 1764
1759 1765 static int
1760 1766 zfs_ioc_pool_get_history(zfs_cmd_t *zc)
1761 1767 {
1762 1768 spa_t *spa;
1763 1769 char *hist_buf;
1764 1770 uint64_t size;
1765 1771 int error;
1766 1772
1767 1773 if ((size = zc->zc_history_len) == 0)
1768 1774 return (SET_ERROR(EINVAL));
1769 1775
1770 1776 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1771 1777 return (error);
1772 1778
1773 1779 if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) {
1774 1780 spa_close(spa, FTAG);
1775 1781 return (SET_ERROR(ENOTSUP));
1776 1782 }
1777 1783
1778 1784 hist_buf = kmem_alloc(size, KM_SLEEP);
1779 1785 if ((error = spa_history_get(spa, &zc->zc_history_offset,
1780 1786 &zc->zc_history_len, hist_buf)) == 0) {
1781 1787 error = ddi_copyout(hist_buf,
1782 1788 (void *)(uintptr_t)zc->zc_history,
1783 1789 zc->zc_history_len, zc->zc_iflags);
1784 1790 }
1785 1791
1786 1792 spa_close(spa, FTAG);
1787 1793 kmem_free(hist_buf, size);
1788 1794 return (error);
1789 1795 }
1790 1796
1791 1797 static int
1792 1798 zfs_ioc_pool_reguid(zfs_cmd_t *zc)
1793 1799 {
1794 1800 spa_t *spa;
1795 1801 int error;
1796 1802
1797 1803 error = spa_open(zc->zc_name, &spa, FTAG);
1798 1804 if (error == 0) {
1799 1805 error = spa_change_guid(spa);
1800 1806 spa_close(spa, FTAG);
1801 1807 }
1802 1808 return (error);
1803 1809 }
1804 1810
1805 1811 static int
1806 1812 zfs_ioc_dsobj_to_dsname(zfs_cmd_t *zc)
1807 1813 {
1808 1814 return (dsl_dsobj_to_dsname(zc->zc_name, zc->zc_obj, zc->zc_value));
1809 1815 }
1810 1816
1811 1817 /*
1812 1818 * inputs:
1813 1819 * zc_name name of filesystem
1814 1820 * zc_obj object to find
1815 1821 *
1816 1822 * outputs:
1817 1823 * zc_value name of object
1818 1824 */
1819 1825 static int
1820 1826 zfs_ioc_obj_to_path(zfs_cmd_t *zc)
1821 1827 {
1822 1828 objset_t *os;
1823 1829 int error;
1824 1830
1825 1831 /* XXX reading from objset not owned */
1826 1832 if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os)) != 0)
1827 1833 return (error);
1828 1834 if (dmu_objset_type(os) != DMU_OST_ZFS) {
1829 1835 dmu_objset_rele(os, FTAG);
1830 1836 return (SET_ERROR(EINVAL));
1831 1837 }
1832 1838 error = zfs_obj_to_path(os, zc->zc_obj, zc->zc_value,
1833 1839 sizeof (zc->zc_value));
1834 1840 dmu_objset_rele(os, FTAG);
1835 1841
1836 1842 return (error);
1837 1843 }
1838 1844
1839 1845 /*
1840 1846 * inputs:
1841 1847 * zc_name name of filesystem
1842 1848 * zc_obj object to find
1843 1849 *
1844 1850 * outputs:
1845 1851 * zc_stat stats on object
1846 1852 * zc_value path to object
1847 1853 */
1848 1854 static int
1849 1855 zfs_ioc_obj_to_stats(zfs_cmd_t *zc)
1850 1856 {
1851 1857 objset_t *os;
1852 1858 int error;
1853 1859
1854 1860 /* XXX reading from objset not owned */
1855 1861 if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os)) != 0)
1856 1862 return (error);
1857 1863 if (dmu_objset_type(os) != DMU_OST_ZFS) {
1858 1864 dmu_objset_rele(os, FTAG);
1859 1865 return (SET_ERROR(EINVAL));
1860 1866 }
1861 1867 error = zfs_obj_to_stats(os, zc->zc_obj, &zc->zc_stat, zc->zc_value,
1862 1868 sizeof (zc->zc_value));
1863 1869 dmu_objset_rele(os, FTAG);
1864 1870
1865 1871 return (error);
1866 1872 }
1867 1873
1868 1874 static int
1869 1875 zfs_ioc_vdev_add(zfs_cmd_t *zc)
1870 1876 {
1871 1877 spa_t *spa;
1872 1878 int error;
1873 1879 nvlist_t *config, **l2cache, **spares;
1874 1880 uint_t nl2cache = 0, nspares = 0;
1875 1881
1876 1882 error = spa_open(zc->zc_name, &spa, FTAG);
1877 1883 if (error != 0)
1878 1884 return (error);
1879 1885
1880 1886 error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1881 1887 zc->zc_iflags, &config);
1882 1888 (void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_L2CACHE,
1883 1889 &l2cache, &nl2cache);
1884 1890
1885 1891 (void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_SPARES,
1886 1892 &spares, &nspares);
1887 1893
1888 1894 /*
1889 1895 * A root pool with concatenated devices is not supported.
1890 1896 * Thus, can not add a device to a root pool.
1891 1897 *
1892 1898 * Intent log device can not be added to a rootpool because
1893 1899 * during mountroot, zil is replayed, a seperated log device
1894 1900 * can not be accessed during the mountroot time.
1895 1901 *
1896 1902 * l2cache and spare devices are ok to be added to a rootpool.
1897 1903 */
1898 1904 if (spa_bootfs(spa) != 0 && nl2cache == 0 && nspares == 0) {
1899 1905 nvlist_free(config);
1900 1906 spa_close(spa, FTAG);
1901 1907 return (SET_ERROR(EDOM));
1902 1908 }
1903 1909
1904 1910 if (error == 0) {
1905 1911 error = spa_vdev_add(spa, config);
1906 1912 nvlist_free(config);
1907 1913 }
1908 1914 spa_close(spa, FTAG);
1909 1915 return (error);
1910 1916 }
1911 1917
1912 1918 /*
1913 1919 * inputs:
1914 1920 * zc_name name of the pool
1915 1921 * zc_guid guid of vdev to remove
1916 1922 * zc_cookie cancel removal
1917 1923 */
1918 1924 static int
1919 1925 zfs_ioc_vdev_remove(zfs_cmd_t *zc)
1920 1926 {
1921 1927 spa_t *spa;
1922 1928 int error;
1923 1929
1924 1930 error = spa_open(zc->zc_name, &spa, FTAG);
1925 1931 if (error != 0)
1926 1932 return (error);
1927 1933 if (zc->zc_cookie != 0) {
1928 1934 error = spa_vdev_remove_cancel(spa);
1929 1935 } else {
1930 1936 error = spa_vdev_remove(spa, zc->zc_guid, B_FALSE);
1931 1937 }
1932 1938 spa_close(spa, FTAG);
1933 1939 return (error);
1934 1940 }
1935 1941
1936 1942 static int
1937 1943 zfs_ioc_vdev_set_state(zfs_cmd_t *zc)
1938 1944 {
1939 1945 spa_t *spa;
1940 1946 int error;
1941 1947 vdev_state_t newstate = VDEV_STATE_UNKNOWN;
1942 1948
1943 1949 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1944 1950 return (error);
1945 1951 switch (zc->zc_cookie) {
1946 1952 case VDEV_STATE_ONLINE:
1947 1953 error = vdev_online(spa, zc->zc_guid, zc->zc_obj, &newstate);
1948 1954 break;
1949 1955
1950 1956 case VDEV_STATE_OFFLINE:
1951 1957 error = vdev_offline(spa, zc->zc_guid, zc->zc_obj);
1952 1958 break;
1953 1959
1954 1960 case VDEV_STATE_FAULTED:
1955 1961 if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED &&
1956 1962 zc->zc_obj != VDEV_AUX_EXTERNAL)
1957 1963 zc->zc_obj = VDEV_AUX_ERR_EXCEEDED;
1958 1964
1959 1965 error = vdev_fault(spa, zc->zc_guid, zc->zc_obj);
1960 1966 break;
1961 1967
1962 1968 case VDEV_STATE_DEGRADED:
1963 1969 if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED &&
1964 1970 zc->zc_obj != VDEV_AUX_EXTERNAL)
1965 1971 zc->zc_obj = VDEV_AUX_ERR_EXCEEDED;
1966 1972
1967 1973 error = vdev_degrade(spa, zc->zc_guid, zc->zc_obj);
1968 1974 break;
1969 1975
1970 1976 default:
1971 1977 error = SET_ERROR(EINVAL);
1972 1978 }
1973 1979 zc->zc_cookie = newstate;
1974 1980 spa_close(spa, FTAG);
1975 1981 return (error);
1976 1982 }
1977 1983
1978 1984 static int
1979 1985 zfs_ioc_vdev_attach(zfs_cmd_t *zc)
1980 1986 {
1981 1987 spa_t *spa;
1982 1988 int replacing = zc->zc_cookie;
1983 1989 nvlist_t *config;
1984 1990 int error;
1985 1991
1986 1992 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1987 1993 return (error);
1988 1994
1989 1995 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1990 1996 zc->zc_iflags, &config)) == 0) {
1991 1997 error = spa_vdev_attach(spa, zc->zc_guid, config, replacing);
1992 1998 nvlist_free(config);
1993 1999 }
1994 2000
1995 2001 spa_close(spa, FTAG);
1996 2002 return (error);
1997 2003 }
1998 2004
1999 2005 static int
2000 2006 zfs_ioc_vdev_detach(zfs_cmd_t *zc)
2001 2007 {
2002 2008 spa_t *spa;
2003 2009 int error;
2004 2010
2005 2011 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
2006 2012 return (error);
2007 2013
2008 2014 error = spa_vdev_detach(spa, zc->zc_guid, 0, B_FALSE);
2009 2015
2010 2016 spa_close(spa, FTAG);
2011 2017 return (error);
2012 2018 }
2013 2019
2014 2020 static int
2015 2021 zfs_ioc_vdev_split(zfs_cmd_t *zc)
2016 2022 {
2017 2023 spa_t *spa;
2018 2024 nvlist_t *config, *props = NULL;
2019 2025 int error;
2020 2026 boolean_t exp = !!(zc->zc_cookie & ZPOOL_EXPORT_AFTER_SPLIT);
2021 2027
2022 2028 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
2023 2029 return (error);
2024 2030
2025 2031 if (error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
2026 2032 zc->zc_iflags, &config)) {
2027 2033 spa_close(spa, FTAG);
2028 2034 return (error);
2029 2035 }
2030 2036
2031 2037 if (zc->zc_nvlist_src_size != 0 && (error =
2032 2038 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2033 2039 zc->zc_iflags, &props))) {
2034 2040 spa_close(spa, FTAG);
2035 2041 nvlist_free(config);
2036 2042 return (error);
2037 2043 }
2038 2044
2039 2045 error = spa_vdev_split_mirror(spa, zc->zc_string, config, props, exp);
2040 2046
2041 2047 spa_close(spa, FTAG);
2042 2048
2043 2049 nvlist_free(config);
2044 2050 nvlist_free(props);
2045 2051
2046 2052 return (error);
2047 2053 }
2048 2054
2049 2055 static int
2050 2056 zfs_ioc_vdev_setpath(zfs_cmd_t *zc)
2051 2057 {
2052 2058 spa_t *spa;
2053 2059 char *path = zc->zc_value;
2054 2060 uint64_t guid = zc->zc_guid;
2055 2061 int error;
2056 2062
2057 2063 error = spa_open(zc->zc_name, &spa, FTAG);
2058 2064 if (error != 0)
2059 2065 return (error);
2060 2066
2061 2067 error = spa_vdev_setpath(spa, guid, path);
2062 2068 spa_close(spa, FTAG);
2063 2069 return (error);
2064 2070 }
2065 2071
2066 2072 static int
2067 2073 zfs_ioc_vdev_setfru(zfs_cmd_t *zc)
2068 2074 {
2069 2075 spa_t *spa;
2070 2076 char *fru = zc->zc_value;
2071 2077 uint64_t guid = zc->zc_guid;
2072 2078 int error;
2073 2079
2074 2080 error = spa_open(zc->zc_name, &spa, FTAG);
2075 2081 if (error != 0)
2076 2082 return (error);
2077 2083
2078 2084 error = spa_vdev_setfru(spa, guid, fru);
2079 2085 spa_close(spa, FTAG);
2080 2086 return (error);
2081 2087 }
2082 2088
2083 2089 static int
2084 2090 zfs_ioc_objset_stats_impl(zfs_cmd_t *zc, objset_t *os)
2085 2091 {
2086 2092 int error = 0;
2087 2093 nvlist_t *nv;
2088 2094
2089 2095 dmu_objset_fast_stat(os, &zc->zc_objset_stats);
2090 2096
2091 2097 if (zc->zc_nvlist_dst != 0 &&
2092 2098 (error = dsl_prop_get_all(os, &nv)) == 0) {
2093 2099 dmu_objset_stats(os, nv);
2094 2100 /*
2095 2101 * NB: zvol_get_stats() will read the objset contents,
2096 2102 * which we aren't supposed to do with a
2097 2103 * DS_MODE_USER hold, because it could be
2098 2104 * inconsistent. So this is a bit of a workaround...
2099 2105 * XXX reading with out owning
2100 2106 */
2101 2107 if (!zc->zc_objset_stats.dds_inconsistent &&
2102 2108 dmu_objset_type(os) == DMU_OST_ZVOL) {
2103 2109 error = zvol_get_stats(os, nv);
2104 2110 if (error == EIO)
2105 2111 return (error);
2106 2112 VERIFY0(error);
2107 2113 }
2108 2114 error = put_nvlist(zc, nv);
2109 2115 nvlist_free(nv);
2110 2116 }
2111 2117
2112 2118 return (error);
2113 2119 }
2114 2120
2115 2121 /*
2116 2122 * inputs:
2117 2123 * zc_name name of filesystem
2118 2124 * zc_nvlist_dst_size size of buffer for property nvlist
2119 2125 *
2120 2126 * outputs:
2121 2127 * zc_objset_stats stats
2122 2128 * zc_nvlist_dst property nvlist
2123 2129 * zc_nvlist_dst_size size of property nvlist
2124 2130 */
2125 2131 static int
2126 2132 zfs_ioc_objset_stats(zfs_cmd_t *zc)
2127 2133 {
2128 2134 objset_t *os;
2129 2135 int error;
2130 2136
2131 2137 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
2132 2138 if (error == 0) {
2133 2139 error = zfs_ioc_objset_stats_impl(zc, os);
2134 2140 dmu_objset_rele(os, FTAG);
2135 2141 }
2136 2142
2137 2143 return (error);
2138 2144 }
2139 2145
2140 2146 /*
2141 2147 * inputs:
2142 2148 * zc_name name of filesystem
2143 2149 * zc_nvlist_dst_size size of buffer for property nvlist
2144 2150 *
2145 2151 * outputs:
2146 2152 * zc_nvlist_dst received property nvlist
2147 2153 * zc_nvlist_dst_size size of received property nvlist
2148 2154 *
2149 2155 * Gets received properties (distinct from local properties on or after
2150 2156 * SPA_VERSION_RECVD_PROPS) for callers who want to differentiate received from
2151 2157 * local property values.
2152 2158 */
2153 2159 static int
2154 2160 zfs_ioc_objset_recvd_props(zfs_cmd_t *zc)
2155 2161 {
2156 2162 int error = 0;
2157 2163 nvlist_t *nv;
2158 2164
2159 2165 /*
2160 2166 * Without this check, we would return local property values if the
2161 2167 * caller has not already received properties on or after
2162 2168 * SPA_VERSION_RECVD_PROPS.
2163 2169 */
2164 2170 if (!dsl_prop_get_hasrecvd(zc->zc_name))
2165 2171 return (SET_ERROR(ENOTSUP));
2166 2172
2167 2173 if (zc->zc_nvlist_dst != 0 &&
2168 2174 (error = dsl_prop_get_received(zc->zc_name, &nv)) == 0) {
2169 2175 error = put_nvlist(zc, nv);
2170 2176 nvlist_free(nv);
2171 2177 }
2172 2178
2173 2179 return (error);
2174 2180 }
2175 2181
2176 2182 static int
2177 2183 nvl_add_zplprop(objset_t *os, nvlist_t *props, zfs_prop_t prop)
2178 2184 {
2179 2185 uint64_t value;
2180 2186 int error;
2181 2187
2182 2188 /*
2183 2189 * zfs_get_zplprop() will either find a value or give us
2184 2190 * the default value (if there is one).
2185 2191 */
2186 2192 if ((error = zfs_get_zplprop(os, prop, &value)) != 0)
2187 2193 return (error);
2188 2194 VERIFY(nvlist_add_uint64(props, zfs_prop_to_name(prop), value) == 0);
2189 2195 return (0);
2190 2196 }
2191 2197
2192 2198 /*
2193 2199 * inputs:
2194 2200 * zc_name name of filesystem
2195 2201 * zc_nvlist_dst_size size of buffer for zpl property nvlist
2196 2202 *
2197 2203 * outputs:
2198 2204 * zc_nvlist_dst zpl property nvlist
2199 2205 * zc_nvlist_dst_size size of zpl property nvlist
2200 2206 */
2201 2207 static int
2202 2208 zfs_ioc_objset_zplprops(zfs_cmd_t *zc)
2203 2209 {
2204 2210 objset_t *os;
2205 2211 int err;
2206 2212
2207 2213 /* XXX reading without owning */
2208 2214 if (err = dmu_objset_hold(zc->zc_name, FTAG, &os))
2209 2215 return (err);
2210 2216
2211 2217 dmu_objset_fast_stat(os, &zc->zc_objset_stats);
2212 2218
2213 2219 /*
2214 2220 * NB: nvl_add_zplprop() will read the objset contents,
2215 2221 * which we aren't supposed to do with a DS_MODE_USER
2216 2222 * hold, because it could be inconsistent.
2217 2223 */
2218 2224 if (zc->zc_nvlist_dst != NULL &&
2219 2225 !zc->zc_objset_stats.dds_inconsistent &&
2220 2226 dmu_objset_type(os) == DMU_OST_ZFS) {
2221 2227 nvlist_t *nv;
2222 2228
2223 2229 VERIFY(nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2224 2230 if ((err = nvl_add_zplprop(os, nv, ZFS_PROP_VERSION)) == 0 &&
2225 2231 (err = nvl_add_zplprop(os, nv, ZFS_PROP_NORMALIZE)) == 0 &&
2226 2232 (err = nvl_add_zplprop(os, nv, ZFS_PROP_UTF8ONLY)) == 0 &&
2227 2233 (err = nvl_add_zplprop(os, nv, ZFS_PROP_CASE)) == 0)
2228 2234 err = put_nvlist(zc, nv);
2229 2235 nvlist_free(nv);
2230 2236 } else {
2231 2237 err = SET_ERROR(ENOENT);
2232 2238 }
2233 2239 dmu_objset_rele(os, FTAG);
2234 2240 return (err);
2235 2241 }
2236 2242
2237 2243 static boolean_t
2238 2244 dataset_name_hidden(const char *name)
2239 2245 {
2240 2246 /*
2241 2247 * Skip over datasets that are not visible in this zone,
2242 2248 * internal datasets (which have a $ in their name), and
2243 2249 * temporary datasets (which have a % in their name).
2244 2250 */
2245 2251 if (strchr(name, '$') != NULL)
2246 2252 return (B_TRUE);
2247 2253 if (strchr(name, '%') != NULL)
2248 2254 return (B_TRUE);
2249 2255 if (!INGLOBALZONE(curproc) && !zone_dataset_visible(name, NULL))
2250 2256 return (B_TRUE);
2251 2257 return (B_FALSE);
2252 2258 }
2253 2259
2254 2260 /*
2255 2261 * inputs:
2256 2262 * zc_name name of filesystem
2257 2263 * zc_cookie zap cursor
2258 2264 * zc_nvlist_dst_size size of buffer for property nvlist
2259 2265 *
2260 2266 * outputs:
2261 2267 * zc_name name of next filesystem
2262 2268 * zc_cookie zap cursor
2263 2269 * zc_objset_stats stats
2264 2270 * zc_nvlist_dst property nvlist
2265 2271 * zc_nvlist_dst_size size of property nvlist
2266 2272 */
2267 2273 static int
2268 2274 zfs_ioc_dataset_list_next(zfs_cmd_t *zc)
2269 2275 {
2270 2276 objset_t *os;
2271 2277 int error;
2272 2278 char *p;
2273 2279 size_t orig_len = strlen(zc->zc_name);
2274 2280
2275 2281 top:
2276 2282 if (error = dmu_objset_hold(zc->zc_name, FTAG, &os)) {
2277 2283 if (error == ENOENT)
2278 2284 error = SET_ERROR(ESRCH);
2279 2285 return (error);
2280 2286 }
2281 2287
2282 2288 p = strrchr(zc->zc_name, '/');
2283 2289 if (p == NULL || p[1] != '\0')
2284 2290 (void) strlcat(zc->zc_name, "/", sizeof (zc->zc_name));
2285 2291 p = zc->zc_name + strlen(zc->zc_name);
2286 2292
2287 2293 do {
2288 2294 error = dmu_dir_list_next(os,
2289 2295 sizeof (zc->zc_name) - (p - zc->zc_name), p,
2290 2296 NULL, &zc->zc_cookie);
2291 2297 if (error == ENOENT)
2292 2298 error = SET_ERROR(ESRCH);
2293 2299 } while (error == 0 && dataset_name_hidden(zc->zc_name));
2294 2300 dmu_objset_rele(os, FTAG);
2295 2301
2296 2302 /*
2297 2303 * If it's an internal dataset (ie. with a '$' in its name),
2298 2304 * don't try to get stats for it, otherwise we'll return ENOENT.
2299 2305 */
2300 2306 if (error == 0 && strchr(zc->zc_name, '$') == NULL) {
2301 2307 error = zfs_ioc_objset_stats(zc); /* fill in the stats */
2302 2308 if (error == ENOENT) {
2303 2309 /* We lost a race with destroy, get the next one. */
2304 2310 zc->zc_name[orig_len] = '\0';
2305 2311 goto top;
2306 2312 }
2307 2313 }
2308 2314 return (error);
2309 2315 }
2310 2316
2311 2317 /*
2312 2318 * inputs:
2313 2319 * zc_name name of filesystem
2314 2320 * zc_cookie zap cursor
2315 2321 * zc_nvlist_dst_size size of buffer for property nvlist
2316 2322 * zc_simple when set, only name is requested
2317 2323 *
2318 2324 * outputs:
2319 2325 * zc_name name of next snapshot
2320 2326 * zc_objset_stats stats
2321 2327 * zc_nvlist_dst property nvlist
2322 2328 * zc_nvlist_dst_size size of property nvlist
2323 2329 */
2324 2330 static int
2325 2331 zfs_ioc_snapshot_list_next(zfs_cmd_t *zc)
2326 2332 {
2327 2333 objset_t *os;
2328 2334 int error;
2329 2335
2330 2336 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
2331 2337 if (error != 0) {
2332 2338 return (error == ENOENT ? ESRCH : error);
2333 2339 }
2334 2340
2335 2341 /*
2336 2342 * A dataset name of maximum length cannot have any snapshots,
2337 2343 * so exit immediately.
2338 2344 */
2339 2345 if (strlcat(zc->zc_name, "@", sizeof (zc->zc_name)) >=
2340 2346 ZFS_MAX_DATASET_NAME_LEN) {
2341 2347 dmu_objset_rele(os, FTAG);
2342 2348 return (SET_ERROR(ESRCH));
2343 2349 }
2344 2350
2345 2351 error = dmu_snapshot_list_next(os,
2346 2352 sizeof (zc->zc_name) - strlen(zc->zc_name),
2347 2353 zc->zc_name + strlen(zc->zc_name), &zc->zc_obj, &zc->zc_cookie,
2348 2354 NULL);
2349 2355
2350 2356 if (error == 0 && !zc->zc_simple) {
2351 2357 dsl_dataset_t *ds;
2352 2358 dsl_pool_t *dp = os->os_dsl_dataset->ds_dir->dd_pool;
2353 2359
2354 2360 error = dsl_dataset_hold_obj(dp, zc->zc_obj, FTAG, &ds);
2355 2361 if (error == 0) {
2356 2362 objset_t *ossnap;
2357 2363
2358 2364 error = dmu_objset_from_ds(ds, &ossnap);
2359 2365 if (error == 0)
2360 2366 error = zfs_ioc_objset_stats_impl(zc, ossnap);
2361 2367 dsl_dataset_rele(ds, FTAG);
2362 2368 }
2363 2369 } else if (error == ENOENT) {
2364 2370 error = SET_ERROR(ESRCH);
2365 2371 }
2366 2372
2367 2373 dmu_objset_rele(os, FTAG);
2368 2374 /* if we failed, undo the @ that we tacked on to zc_name */
2369 2375 if (error != 0)
2370 2376 *strchr(zc->zc_name, '@') = '\0';
2371 2377 return (error);
2372 2378 }
2373 2379
2374 2380 static int
2375 2381 zfs_prop_set_userquota(const char *dsname, nvpair_t *pair)
2376 2382 {
2377 2383 const char *propname = nvpair_name(pair);
2378 2384 uint64_t *valary;
2379 2385 unsigned int vallen;
2380 2386 const char *domain;
2381 2387 char *dash;
2382 2388 zfs_userquota_prop_t type;
2383 2389 uint64_t rid;
2384 2390 uint64_t quota;
2385 2391 zfsvfs_t *zfsvfs;
2386 2392 int err;
2387 2393
2388 2394 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2389 2395 nvlist_t *attrs;
2390 2396 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
2391 2397 if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2392 2398 &pair) != 0)
2393 2399 return (SET_ERROR(EINVAL));
2394 2400 }
2395 2401
2396 2402 /*
2397 2403 * A correctly constructed propname is encoded as
2398 2404 * userquota@<rid>-<domain>.
2399 2405 */
2400 2406 if ((dash = strchr(propname, '-')) == NULL ||
2401 2407 nvpair_value_uint64_array(pair, &valary, &vallen) != 0 ||
2402 2408 vallen != 3)
2403 2409 return (SET_ERROR(EINVAL));
2404 2410
2405 2411 domain = dash + 1;
2406 2412 type = valary[0];
2407 2413 rid = valary[1];
2408 2414 quota = valary[2];
2409 2415
2410 2416 err = zfsvfs_hold(dsname, FTAG, &zfsvfs, B_FALSE);
2411 2417 if (err == 0) {
2412 2418 err = zfs_set_userquota(zfsvfs, type, domain, rid, quota);
2413 2419 zfsvfs_rele(zfsvfs, FTAG);
2414 2420 }
2415 2421
2416 2422 return (err);
2417 2423 }
2418 2424
2419 2425 /*
2420 2426 * If the named property is one that has a special function to set its value,
2421 2427 * return 0 on success and a positive error code on failure; otherwise if it is
2422 2428 * not one of the special properties handled by this function, return -1.
2423 2429 *
2424 2430 * XXX: It would be better for callers of the property interface if we handled
2425 2431 * these special cases in dsl_prop.c (in the dsl layer).
2426 2432 */
2427 2433 static int
2428 2434 zfs_prop_set_special(const char *dsname, zprop_source_t source,
2429 2435 nvpair_t *pair)
2430 2436 {
2431 2437 const char *propname = nvpair_name(pair);
2432 2438 zfs_prop_t prop = zfs_name_to_prop(propname);
2433 2439 uint64_t intval;
2434 2440 int err = -1;
2435 2441
2436 2442 if (prop == ZPROP_INVAL) {
2437 2443 if (zfs_prop_userquota(propname))
2438 2444 return (zfs_prop_set_userquota(dsname, pair));
2439 2445 return (-1);
2440 2446 }
2441 2447
2442 2448 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2443 2449 nvlist_t *attrs;
2444 2450 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
2445 2451 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2446 2452 &pair) == 0);
2447 2453 }
2448 2454
2449 2455 if (zfs_prop_get_type(prop) == PROP_TYPE_STRING)
2450 2456 return (-1);
2451 2457
2452 2458 VERIFY(0 == nvpair_value_uint64(pair, &intval));
2453 2459
2454 2460 switch (prop) {
2455 2461 case ZFS_PROP_QUOTA:
2456 2462 err = dsl_dir_set_quota(dsname, source, intval);
2457 2463 break;
2458 2464 case ZFS_PROP_REFQUOTA:
2459 2465 err = dsl_dataset_set_refquota(dsname, source, intval);
2460 2466 break;
2461 2467 case ZFS_PROP_FILESYSTEM_LIMIT:
2462 2468 case ZFS_PROP_SNAPSHOT_LIMIT:
2463 2469 if (intval == UINT64_MAX) {
2464 2470 /* clearing the limit, just do it */
2465 2471 err = 0;
2466 2472 } else {
2467 2473 err = dsl_dir_activate_fs_ss_limit(dsname);
2468 2474 }
2469 2475 /*
2470 2476 * Set err to -1 to force the zfs_set_prop_nvlist code down the
2471 2477 * default path to set the value in the nvlist.
2472 2478 */
2473 2479 if (err == 0)
2474 2480 err = -1;
2475 2481 break;
2476 2482 case ZFS_PROP_RESERVATION:
2477 2483 err = dsl_dir_set_reservation(dsname, source, intval);
2478 2484 break;
2479 2485 case ZFS_PROP_REFRESERVATION:
2480 2486 err = dsl_dataset_set_refreservation(dsname, source, intval);
2481 2487 break;
2482 2488 case ZFS_PROP_VOLSIZE:
2483 2489 err = zvol_set_volsize(dsname, intval);
2484 2490 break;
2485 2491 case ZFS_PROP_VERSION:
2486 2492 {
2487 2493 zfsvfs_t *zfsvfs;
2488 2494
2489 2495 if ((err = zfsvfs_hold(dsname, FTAG, &zfsvfs, B_TRUE)) != 0)
2490 2496 break;
2491 2497
2492 2498 err = zfs_set_version(zfsvfs, intval);
2493 2499 zfsvfs_rele(zfsvfs, FTAG);
2494 2500
2495 2501 if (err == 0 && intval >= ZPL_VERSION_USERSPACE) {
2496 2502 zfs_cmd_t *zc;
2497 2503
2498 2504 zc = kmem_zalloc(sizeof (zfs_cmd_t), KM_SLEEP);
2499 2505 (void) strcpy(zc->zc_name, dsname);
2500 2506 (void) zfs_ioc_userspace_upgrade(zc);
2501 2507 kmem_free(zc, sizeof (zfs_cmd_t));
2502 2508 }
2503 2509 break;
2504 2510 }
2505 2511 default:
2506 2512 err = -1;
2507 2513 }
2508 2514
2509 2515 return (err);
2510 2516 }
2511 2517
2512 2518 /*
2513 2519 * This function is best effort. If it fails to set any of the given properties,
2514 2520 * it continues to set as many as it can and returns the last error
2515 2521 * encountered. If the caller provides a non-NULL errlist, it will be filled in
2516 2522 * with the list of names of all the properties that failed along with the
2517 2523 * corresponding error numbers.
2518 2524 *
2519 2525 * If every property is set successfully, zero is returned and errlist is not
2520 2526 * modified.
2521 2527 */
2522 2528 int
2523 2529 zfs_set_prop_nvlist(const char *dsname, zprop_source_t source, nvlist_t *nvl,
2524 2530 nvlist_t *errlist)
2525 2531 {
2526 2532 nvpair_t *pair;
2527 2533 nvpair_t *propval;
2528 2534 int rv = 0;
2529 2535 uint64_t intval;
2530 2536 char *strval;
2531 2537 nvlist_t *genericnvl = fnvlist_alloc();
2532 2538 nvlist_t *retrynvl = fnvlist_alloc();
2533 2539
2534 2540 retry:
2535 2541 pair = NULL;
2536 2542 while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) {
2537 2543 const char *propname = nvpair_name(pair);
2538 2544 zfs_prop_t prop = zfs_name_to_prop(propname);
2539 2545 int err = 0;
2540 2546
2541 2547 /* decode the property value */
2542 2548 propval = pair;
2543 2549 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2544 2550 nvlist_t *attrs;
2545 2551 attrs = fnvpair_value_nvlist(pair);
2546 2552 if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2547 2553 &propval) != 0)
2548 2554 err = SET_ERROR(EINVAL);
2549 2555 }
2550 2556
2551 2557 /* Validate value type */
2552 2558 if (err == 0 && prop == ZPROP_INVAL) {
2553 2559 if (zfs_prop_user(propname)) {
2554 2560 if (nvpair_type(propval) != DATA_TYPE_STRING)
2555 2561 err = SET_ERROR(EINVAL);
2556 2562 } else if (zfs_prop_userquota(propname)) {
2557 2563 if (nvpair_type(propval) !=
2558 2564 DATA_TYPE_UINT64_ARRAY)
2559 2565 err = SET_ERROR(EINVAL);
2560 2566 } else {
2561 2567 err = SET_ERROR(EINVAL);
2562 2568 }
2563 2569 } else if (err == 0) {
2564 2570 if (nvpair_type(propval) == DATA_TYPE_STRING) {
2565 2571 if (zfs_prop_get_type(prop) != PROP_TYPE_STRING)
2566 2572 err = SET_ERROR(EINVAL);
2567 2573 } else if (nvpair_type(propval) == DATA_TYPE_UINT64) {
2568 2574 const char *unused;
2569 2575
2570 2576 intval = fnvpair_value_uint64(propval);
2571 2577
2572 2578 switch (zfs_prop_get_type(prop)) {
2573 2579 case PROP_TYPE_NUMBER:
2574 2580 break;
2575 2581 case PROP_TYPE_STRING:
2576 2582 err = SET_ERROR(EINVAL);
2577 2583 break;
2578 2584 case PROP_TYPE_INDEX:
2579 2585 if (zfs_prop_index_to_string(prop,
2580 2586 intval, &unused) != 0)
2581 2587 err = SET_ERROR(EINVAL);
2582 2588 break;
2583 2589 default:
2584 2590 cmn_err(CE_PANIC,
2585 2591 "unknown property type");
2586 2592 }
2587 2593 } else {
2588 2594 err = SET_ERROR(EINVAL);
2589 2595 }
2590 2596 }
2591 2597
2592 2598 /* Validate permissions */
2593 2599 if (err == 0)
2594 2600 err = zfs_check_settable(dsname, pair, CRED());
2595 2601
2596 2602 if (err == 0) {
2597 2603 err = zfs_prop_set_special(dsname, source, pair);
2598 2604 if (err == -1) {
2599 2605 /*
2600 2606 * For better performance we build up a list of
2601 2607 * properties to set in a single transaction.
2602 2608 */
2603 2609 err = nvlist_add_nvpair(genericnvl, pair);
2604 2610 } else if (err != 0 && nvl != retrynvl) {
2605 2611 /*
2606 2612 * This may be a spurious error caused by
2607 2613 * receiving quota and reservation out of order.
2608 2614 * Try again in a second pass.
2609 2615 */
2610 2616 err = nvlist_add_nvpair(retrynvl, pair);
2611 2617 }
2612 2618 }
2613 2619
2614 2620 if (err != 0) {
2615 2621 if (errlist != NULL)
2616 2622 fnvlist_add_int32(errlist, propname, err);
2617 2623 rv = err;
2618 2624 }
2619 2625 }
2620 2626
2621 2627 if (nvl != retrynvl && !nvlist_empty(retrynvl)) {
2622 2628 nvl = retrynvl;
2623 2629 goto retry;
2624 2630 }
2625 2631
2626 2632 if (!nvlist_empty(genericnvl) &&
2627 2633 dsl_props_set(dsname, source, genericnvl) != 0) {
2628 2634 /*
2629 2635 * If this fails, we still want to set as many properties as we
2630 2636 * can, so try setting them individually.
2631 2637 */
2632 2638 pair = NULL;
2633 2639 while ((pair = nvlist_next_nvpair(genericnvl, pair)) != NULL) {
2634 2640 const char *propname = nvpair_name(pair);
2635 2641 int err = 0;
2636 2642
2637 2643 propval = pair;
2638 2644 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2639 2645 nvlist_t *attrs;
2640 2646 attrs = fnvpair_value_nvlist(pair);
2641 2647 propval = fnvlist_lookup_nvpair(attrs,
2642 2648 ZPROP_VALUE);
2643 2649 }
2644 2650
2645 2651 if (nvpair_type(propval) == DATA_TYPE_STRING) {
2646 2652 strval = fnvpair_value_string(propval);
2647 2653 err = dsl_prop_set_string(dsname, propname,
2648 2654 source, strval);
2649 2655 } else {
2650 2656 intval = fnvpair_value_uint64(propval);
2651 2657 err = dsl_prop_set_int(dsname, propname, source,
2652 2658 intval);
2653 2659 }
2654 2660
2655 2661 if (err != 0) {
2656 2662 if (errlist != NULL) {
2657 2663 fnvlist_add_int32(errlist, propname,
2658 2664 err);
2659 2665 }
2660 2666 rv = err;
2661 2667 }
2662 2668 }
2663 2669 }
2664 2670 nvlist_free(genericnvl);
2665 2671 nvlist_free(retrynvl);
2666 2672
2667 2673 return (rv);
2668 2674 }
2669 2675
2670 2676 /*
2671 2677 * Check that all the properties are valid user properties.
2672 2678 */
2673 2679 static int
2674 2680 zfs_check_userprops(const char *fsname, nvlist_t *nvl)
2675 2681 {
2676 2682 nvpair_t *pair = NULL;
2677 2683 int error = 0;
2678 2684
2679 2685 while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) {
2680 2686 const char *propname = nvpair_name(pair);
2681 2687
2682 2688 if (!zfs_prop_user(propname) ||
2683 2689 nvpair_type(pair) != DATA_TYPE_STRING)
2684 2690 return (SET_ERROR(EINVAL));
2685 2691
2686 2692 if (error = zfs_secpolicy_write_perms(fsname,
2687 2693 ZFS_DELEG_PERM_USERPROP, CRED()))
2688 2694 return (error);
2689 2695
2690 2696 if (strlen(propname) >= ZAP_MAXNAMELEN)
2691 2697 return (SET_ERROR(ENAMETOOLONG));
2692 2698
2693 2699 if (strlen(fnvpair_value_string(pair)) >= ZAP_MAXVALUELEN)
2694 2700 return (E2BIG);
2695 2701 }
2696 2702 return (0);
2697 2703 }
2698 2704
2699 2705 static void
2700 2706 props_skip(nvlist_t *props, nvlist_t *skipped, nvlist_t **newprops)
2701 2707 {
2702 2708 nvpair_t *pair;
2703 2709
2704 2710 VERIFY(nvlist_alloc(newprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2705 2711
2706 2712 pair = NULL;
2707 2713 while ((pair = nvlist_next_nvpair(props, pair)) != NULL) {
2708 2714 if (nvlist_exists(skipped, nvpair_name(pair)))
2709 2715 continue;
2710 2716
2711 2717 VERIFY(nvlist_add_nvpair(*newprops, pair) == 0);
2712 2718 }
2713 2719 }
2714 2720
2715 2721 static int
2716 2722 clear_received_props(const char *dsname, nvlist_t *props,
2717 2723 nvlist_t *skipped)
2718 2724 {
2719 2725 int err = 0;
2720 2726 nvlist_t *cleared_props = NULL;
2721 2727 props_skip(props, skipped, &cleared_props);
2722 2728 if (!nvlist_empty(cleared_props)) {
2723 2729 /*
2724 2730 * Acts on local properties until the dataset has received
2725 2731 * properties at least once on or after SPA_VERSION_RECVD_PROPS.
2726 2732 */
2727 2733 zprop_source_t flags = (ZPROP_SRC_NONE |
2728 2734 (dsl_prop_get_hasrecvd(dsname) ? ZPROP_SRC_RECEIVED : 0));
2729 2735 err = zfs_set_prop_nvlist(dsname, flags, cleared_props, NULL);
2730 2736 }
2731 2737 nvlist_free(cleared_props);
2732 2738 return (err);
2733 2739 }
2734 2740
2735 2741 /*
2736 2742 * inputs:
2737 2743 * zc_name name of filesystem
2738 2744 * zc_value name of property to set
2739 2745 * zc_nvlist_src{_size} nvlist of properties to apply
2740 2746 * zc_cookie received properties flag
2741 2747 *
2742 2748 * outputs:
2743 2749 * zc_nvlist_dst{_size} error for each unapplied received property
2744 2750 */
2745 2751 static int
2746 2752 zfs_ioc_set_prop(zfs_cmd_t *zc)
2747 2753 {
2748 2754 nvlist_t *nvl;
2749 2755 boolean_t received = zc->zc_cookie;
2750 2756 zprop_source_t source = (received ? ZPROP_SRC_RECEIVED :
2751 2757 ZPROP_SRC_LOCAL);
2752 2758 nvlist_t *errors;
2753 2759 int error;
2754 2760
2755 2761 if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2756 2762 zc->zc_iflags, &nvl)) != 0)
2757 2763 return (error);
2758 2764
2759 2765 if (received) {
2760 2766 nvlist_t *origprops;
2761 2767
2762 2768 if (dsl_prop_get_received(zc->zc_name, &origprops) == 0) {
2763 2769 (void) clear_received_props(zc->zc_name,
2764 2770 origprops, nvl);
2765 2771 nvlist_free(origprops);
2766 2772 }
2767 2773
2768 2774 error = dsl_prop_set_hasrecvd(zc->zc_name);
2769 2775 }
2770 2776
2771 2777 errors = fnvlist_alloc();
2772 2778 if (error == 0)
2773 2779 error = zfs_set_prop_nvlist(zc->zc_name, source, nvl, errors);
2774 2780
2775 2781 if (zc->zc_nvlist_dst != NULL && errors != NULL) {
2776 2782 (void) put_nvlist(zc, errors);
2777 2783 }
2778 2784
2779 2785 nvlist_free(errors);
2780 2786 nvlist_free(nvl);
2781 2787 return (error);
2782 2788 }
2783 2789
2784 2790 /*
2785 2791 * inputs:
2786 2792 * zc_name name of filesystem
2787 2793 * zc_value name of property to inherit
2788 2794 * zc_cookie revert to received value if TRUE
2789 2795 *
2790 2796 * outputs: none
2791 2797 */
2792 2798 static int
2793 2799 zfs_ioc_inherit_prop(zfs_cmd_t *zc)
2794 2800 {
2795 2801 const char *propname = zc->zc_value;
2796 2802 zfs_prop_t prop = zfs_name_to_prop(propname);
2797 2803 boolean_t received = zc->zc_cookie;
2798 2804 zprop_source_t source = (received
2799 2805 ? ZPROP_SRC_NONE /* revert to received value, if any */
2800 2806 : ZPROP_SRC_INHERITED); /* explicitly inherit */
2801 2807
2802 2808 if (received) {
2803 2809 nvlist_t *dummy;
2804 2810 nvpair_t *pair;
2805 2811 zprop_type_t type;
2806 2812 int err;
2807 2813
2808 2814 /*
2809 2815 * zfs_prop_set_special() expects properties in the form of an
2810 2816 * nvpair with type info.
2811 2817 */
2812 2818 if (prop == ZPROP_INVAL) {
2813 2819 if (!zfs_prop_user(propname))
2814 2820 return (SET_ERROR(EINVAL));
2815 2821
2816 2822 type = PROP_TYPE_STRING;
2817 2823 } else if (prop == ZFS_PROP_VOLSIZE ||
2818 2824 prop == ZFS_PROP_VERSION) {
2819 2825 return (SET_ERROR(EINVAL));
2820 2826 } else {
2821 2827 type = zfs_prop_get_type(prop);
2822 2828 }
2823 2829
2824 2830 VERIFY(nvlist_alloc(&dummy, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2825 2831
2826 2832 switch (type) {
2827 2833 case PROP_TYPE_STRING:
2828 2834 VERIFY(0 == nvlist_add_string(dummy, propname, ""));
2829 2835 break;
2830 2836 case PROP_TYPE_NUMBER:
2831 2837 case PROP_TYPE_INDEX:
2832 2838 VERIFY(0 == nvlist_add_uint64(dummy, propname, 0));
2833 2839 break;
2834 2840 default:
2835 2841 nvlist_free(dummy);
2836 2842 return (SET_ERROR(EINVAL));
2837 2843 }
2838 2844
2839 2845 pair = nvlist_next_nvpair(dummy, NULL);
2840 2846 err = zfs_prop_set_special(zc->zc_name, source, pair);
2841 2847 nvlist_free(dummy);
2842 2848 if (err != -1)
2843 2849 return (err); /* special property already handled */
2844 2850 } else {
2845 2851 /*
2846 2852 * Only check this in the non-received case. We want to allow
2847 2853 * 'inherit -S' to revert non-inheritable properties like quota
2848 2854 * and reservation to the received or default values even though
2849 2855 * they are not considered inheritable.
2850 2856 */
2851 2857 if (prop != ZPROP_INVAL && !zfs_prop_inheritable(prop))
2852 2858 return (SET_ERROR(EINVAL));
2853 2859 }
2854 2860
2855 2861 /* property name has been validated by zfs_secpolicy_inherit_prop() */
2856 2862 return (dsl_prop_inherit(zc->zc_name, zc->zc_value, source));
2857 2863 }
2858 2864
2859 2865 static int
2860 2866 zfs_ioc_pool_set_props(zfs_cmd_t *zc)
2861 2867 {
2862 2868 nvlist_t *props;
2863 2869 spa_t *spa;
2864 2870 int error;
2865 2871 nvpair_t *pair;
2866 2872
2867 2873 if (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2868 2874 zc->zc_iflags, &props))
2869 2875 return (error);
2870 2876
2871 2877 /*
2872 2878 * If the only property is the configfile, then just do a spa_lookup()
2873 2879 * to handle the faulted case.
2874 2880 */
2875 2881 pair = nvlist_next_nvpair(props, NULL);
2876 2882 if (pair != NULL && strcmp(nvpair_name(pair),
2877 2883 zpool_prop_to_name(ZPOOL_PROP_CACHEFILE)) == 0 &&
2878 2884 nvlist_next_nvpair(props, pair) == NULL) {
2879 2885 mutex_enter(&spa_namespace_lock);
2880 2886 if ((spa = spa_lookup(zc->zc_name)) != NULL) {
2881 2887 spa_configfile_set(spa, props, B_FALSE);
2882 2888 spa_write_cachefile(spa, B_FALSE, B_TRUE);
2883 2889 }
2884 2890 mutex_exit(&spa_namespace_lock);
2885 2891 if (spa != NULL) {
2886 2892 nvlist_free(props);
2887 2893 return (0);
2888 2894 }
2889 2895 }
2890 2896
2891 2897 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) {
2892 2898 nvlist_free(props);
2893 2899 return (error);
2894 2900 }
2895 2901
2896 2902 error = spa_prop_set(spa, props);
2897 2903
2898 2904 nvlist_free(props);
2899 2905 spa_close(spa, FTAG);
2900 2906
2901 2907 return (error);
2902 2908 }
2903 2909
2904 2910 static int
2905 2911 zfs_ioc_pool_get_props(zfs_cmd_t *zc)
2906 2912 {
2907 2913 spa_t *spa;
2908 2914 int error;
2909 2915 nvlist_t *nvp = NULL;
2910 2916
2911 2917 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) {
2912 2918 /*
2913 2919 * If the pool is faulted, there may be properties we can still
2914 2920 * get (such as altroot and cachefile), so attempt to get them
2915 2921 * anyway.
2916 2922 */
2917 2923 mutex_enter(&spa_namespace_lock);
2918 2924 if ((spa = spa_lookup(zc->zc_name)) != NULL)
2919 2925 error = spa_prop_get(spa, &nvp);
2920 2926 mutex_exit(&spa_namespace_lock);
2921 2927 } else {
2922 2928 error = spa_prop_get(spa, &nvp);
2923 2929 spa_close(spa, FTAG);
2924 2930 }
2925 2931
2926 2932 if (error == 0 && zc->zc_nvlist_dst != NULL)
2927 2933 error = put_nvlist(zc, nvp);
2928 2934 else
2929 2935 error = SET_ERROR(EFAULT);
2930 2936
2931 2937 nvlist_free(nvp);
2932 2938 return (error);
2933 2939 }
2934 2940
2935 2941 /*
2936 2942 * inputs:
2937 2943 * zc_name name of filesystem
2938 2944 * zc_nvlist_src{_size} nvlist of delegated permissions
2939 2945 * zc_perm_action allow/unallow flag
2940 2946 *
2941 2947 * outputs: none
2942 2948 */
2943 2949 static int
2944 2950 zfs_ioc_set_fsacl(zfs_cmd_t *zc)
2945 2951 {
2946 2952 int error;
2947 2953 nvlist_t *fsaclnv = NULL;
2948 2954
2949 2955 if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2950 2956 zc->zc_iflags, &fsaclnv)) != 0)
2951 2957 return (error);
2952 2958
2953 2959 /*
2954 2960 * Verify nvlist is constructed correctly
2955 2961 */
2956 2962 if ((error = zfs_deleg_verify_nvlist(fsaclnv)) != 0) {
2957 2963 nvlist_free(fsaclnv);
2958 2964 return (SET_ERROR(EINVAL));
2959 2965 }
2960 2966
2961 2967 /*
2962 2968 * If we don't have PRIV_SYS_MOUNT, then validate
2963 2969 * that user is allowed to hand out each permission in
2964 2970 * the nvlist(s)
2965 2971 */
2966 2972
2967 2973 error = secpolicy_zfs(CRED());
2968 2974 if (error != 0) {
2969 2975 if (zc->zc_perm_action == B_FALSE) {
2970 2976 error = dsl_deleg_can_allow(zc->zc_name,
2971 2977 fsaclnv, CRED());
2972 2978 } else {
2973 2979 error = dsl_deleg_can_unallow(zc->zc_name,
2974 2980 fsaclnv, CRED());
2975 2981 }
2976 2982 }
2977 2983
2978 2984 if (error == 0)
2979 2985 error = dsl_deleg_set(zc->zc_name, fsaclnv, zc->zc_perm_action);
2980 2986
2981 2987 nvlist_free(fsaclnv);
2982 2988 return (error);
2983 2989 }
2984 2990
2985 2991 /*
2986 2992 * inputs:
2987 2993 * zc_name name of filesystem
2988 2994 *
2989 2995 * outputs:
2990 2996 * zc_nvlist_src{_size} nvlist of delegated permissions
2991 2997 */
2992 2998 static int
2993 2999 zfs_ioc_get_fsacl(zfs_cmd_t *zc)
2994 3000 {
2995 3001 nvlist_t *nvp;
2996 3002 int error;
2997 3003
2998 3004 if ((error = dsl_deleg_get(zc->zc_name, &nvp)) == 0) {
2999 3005 error = put_nvlist(zc, nvp);
3000 3006 nvlist_free(nvp);
3001 3007 }
3002 3008
3003 3009 return (error);
3004 3010 }
3005 3011
3006 3012 /* ARGSUSED */
3007 3013 static void
3008 3014 zfs_create_cb(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx)
3009 3015 {
3010 3016 zfs_creat_t *zct = arg;
3011 3017
3012 3018 zfs_create_fs(os, cr, zct->zct_zplprops, tx);
3013 3019 }
3014 3020
3015 3021 #define ZFS_PROP_UNDEFINED ((uint64_t)-1)
3016 3022
3017 3023 /*
3018 3024 * inputs:
3019 3025 * os parent objset pointer (NULL if root fs)
3020 3026 * fuids_ok fuids allowed in this version of the spa?
3021 3027 * sa_ok SAs allowed in this version of the spa?
3022 3028 * createprops list of properties requested by creator
3023 3029 *
3024 3030 * outputs:
3025 3031 * zplprops values for the zplprops we attach to the master node object
3026 3032 * is_ci true if requested file system will be purely case-insensitive
3027 3033 *
3028 3034 * Determine the settings for utf8only, normalization and
3029 3035 * casesensitivity. Specific values may have been requested by the
3030 3036 * creator and/or we can inherit values from the parent dataset. If
3031 3037 * the file system is of too early a vintage, a creator can not
3032 3038 * request settings for these properties, even if the requested
3033 3039 * setting is the default value. We don't actually want to create dsl
3034 3040 * properties for these, so remove them from the source nvlist after
3035 3041 * processing.
3036 3042 */
3037 3043 static int
3038 3044 zfs_fill_zplprops_impl(objset_t *os, uint64_t zplver,
3039 3045 boolean_t fuids_ok, boolean_t sa_ok, nvlist_t *createprops,
3040 3046 nvlist_t *zplprops, boolean_t *is_ci)
3041 3047 {
3042 3048 uint64_t sense = ZFS_PROP_UNDEFINED;
3043 3049 uint64_t norm = ZFS_PROP_UNDEFINED;
3044 3050 uint64_t u8 = ZFS_PROP_UNDEFINED;
3045 3051
3046 3052 ASSERT(zplprops != NULL);
3047 3053
3048 3054 if (os != NULL && os->os_phys->os_type != DMU_OST_ZFS)
3049 3055 return (SET_ERROR(EINVAL));
3050 3056
3051 3057 /*
3052 3058 * Pull out creator prop choices, if any.
3053 3059 */
3054 3060 if (createprops) {
3055 3061 (void) nvlist_lookup_uint64(createprops,
3056 3062 zfs_prop_to_name(ZFS_PROP_VERSION), &zplver);
3057 3063 (void) nvlist_lookup_uint64(createprops,
3058 3064 zfs_prop_to_name(ZFS_PROP_NORMALIZE), &norm);
3059 3065 (void) nvlist_remove_all(createprops,
3060 3066 zfs_prop_to_name(ZFS_PROP_NORMALIZE));
3061 3067 (void) nvlist_lookup_uint64(createprops,
3062 3068 zfs_prop_to_name(ZFS_PROP_UTF8ONLY), &u8);
3063 3069 (void) nvlist_remove_all(createprops,
3064 3070 zfs_prop_to_name(ZFS_PROP_UTF8ONLY));
3065 3071 (void) nvlist_lookup_uint64(createprops,
3066 3072 zfs_prop_to_name(ZFS_PROP_CASE), &sense);
3067 3073 (void) nvlist_remove_all(createprops,
3068 3074 zfs_prop_to_name(ZFS_PROP_CASE));
3069 3075 }
3070 3076
3071 3077 /*
3072 3078 * If the zpl version requested is whacky or the file system
3073 3079 * or pool is version is too "young" to support normalization
3074 3080 * and the creator tried to set a value for one of the props,
3075 3081 * error out.
3076 3082 */
3077 3083 if ((zplver < ZPL_VERSION_INITIAL || zplver > ZPL_VERSION) ||
3078 3084 (zplver >= ZPL_VERSION_FUID && !fuids_ok) ||
3079 3085 (zplver >= ZPL_VERSION_SA && !sa_ok) ||
3080 3086 (zplver < ZPL_VERSION_NORMALIZATION &&
3081 3087 (norm != ZFS_PROP_UNDEFINED || u8 != ZFS_PROP_UNDEFINED ||
3082 3088 sense != ZFS_PROP_UNDEFINED)))
3083 3089 return (SET_ERROR(ENOTSUP));
3084 3090
3085 3091 /*
3086 3092 * Put the version in the zplprops
3087 3093 */
3088 3094 VERIFY(nvlist_add_uint64(zplprops,
3089 3095 zfs_prop_to_name(ZFS_PROP_VERSION), zplver) == 0);
3090 3096
3091 3097 if (norm == ZFS_PROP_UNDEFINED)
3092 3098 VERIFY(zfs_get_zplprop(os, ZFS_PROP_NORMALIZE, &norm) == 0);
3093 3099 VERIFY(nvlist_add_uint64(zplprops,
3094 3100 zfs_prop_to_name(ZFS_PROP_NORMALIZE), norm) == 0);
3095 3101
3096 3102 /*
3097 3103 * If we're normalizing, names must always be valid UTF-8 strings.
3098 3104 */
3099 3105 if (norm)
3100 3106 u8 = 1;
3101 3107 if (u8 == ZFS_PROP_UNDEFINED)
3102 3108 VERIFY(zfs_get_zplprop(os, ZFS_PROP_UTF8ONLY, &u8) == 0);
3103 3109 VERIFY(nvlist_add_uint64(zplprops,
3104 3110 zfs_prop_to_name(ZFS_PROP_UTF8ONLY), u8) == 0);
3105 3111
3106 3112 if (sense == ZFS_PROP_UNDEFINED)
3107 3113 VERIFY(zfs_get_zplprop(os, ZFS_PROP_CASE, &sense) == 0);
3108 3114 VERIFY(nvlist_add_uint64(zplprops,
3109 3115 zfs_prop_to_name(ZFS_PROP_CASE), sense) == 0);
3110 3116
3111 3117 if (is_ci)
3112 3118 *is_ci = (sense == ZFS_CASE_INSENSITIVE);
3113 3119
3114 3120 return (0);
3115 3121 }
3116 3122
3117 3123 static int
3118 3124 zfs_fill_zplprops(const char *dataset, nvlist_t *createprops,
3119 3125 nvlist_t *zplprops, boolean_t *is_ci)
3120 3126 {
3121 3127 boolean_t fuids_ok, sa_ok;
3122 3128 uint64_t zplver = ZPL_VERSION;
3123 3129 objset_t *os = NULL;
3124 3130 char parentname[ZFS_MAX_DATASET_NAME_LEN];
3125 3131 char *cp;
3126 3132 spa_t *spa;
3127 3133 uint64_t spa_vers;
3128 3134 int error;
3129 3135
3130 3136 (void) strlcpy(parentname, dataset, sizeof (parentname));
3131 3137 cp = strrchr(parentname, '/');
3132 3138 ASSERT(cp != NULL);
3133 3139 cp[0] = '\0';
3134 3140
3135 3141 if ((error = spa_open(dataset, &spa, FTAG)) != 0)
3136 3142 return (error);
3137 3143
3138 3144 spa_vers = spa_version(spa);
3139 3145 spa_close(spa, FTAG);
3140 3146
3141 3147 zplver = zfs_zpl_version_map(spa_vers);
3142 3148 fuids_ok = (zplver >= ZPL_VERSION_FUID);
3143 3149 sa_ok = (zplver >= ZPL_VERSION_SA);
3144 3150
3145 3151 /*
3146 3152 * Open parent object set so we can inherit zplprop values.
3147 3153 */
3148 3154 if ((error = dmu_objset_hold(parentname, FTAG, &os)) != 0)
3149 3155 return (error);
3150 3156
3151 3157 error = zfs_fill_zplprops_impl(os, zplver, fuids_ok, sa_ok, createprops,
3152 3158 zplprops, is_ci);
3153 3159 dmu_objset_rele(os, FTAG);
3154 3160 return (error);
3155 3161 }
3156 3162
3157 3163 static int
3158 3164 zfs_fill_zplprops_root(uint64_t spa_vers, nvlist_t *createprops,
3159 3165 nvlist_t *zplprops, boolean_t *is_ci)
3160 3166 {
3161 3167 boolean_t fuids_ok;
3162 3168 boolean_t sa_ok;
3163 3169 uint64_t zplver = ZPL_VERSION;
3164 3170 int error;
3165 3171
3166 3172 zplver = zfs_zpl_version_map(spa_vers);
3167 3173 fuids_ok = (zplver >= ZPL_VERSION_FUID);
3168 3174 sa_ok = (zplver >= ZPL_VERSION_SA);
3169 3175
3170 3176 error = zfs_fill_zplprops_impl(NULL, zplver, fuids_ok, sa_ok,
3171 3177 createprops, zplprops, is_ci);
3172 3178 return (error);
3173 3179 }
3174 3180
3175 3181 /*
3176 3182 * innvl: {
3177 3183 * "type" -> dmu_objset_type_t (int32)
3178 3184 * (optional) "props" -> { prop -> value }
3179 3185 * }
3180 3186 *
3181 3187 * outnvl: propname -> error code (int32)
3182 3188 */
3183 3189 static int
3184 3190 zfs_ioc_create(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3185 3191 {
3186 3192 int error = 0;
3187 3193 zfs_creat_t zct = { 0 };
3188 3194 nvlist_t *nvprops = NULL;
3189 3195 void (*cbfunc)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx);
3190 3196 int32_t type32;
3191 3197 dmu_objset_type_t type;
3192 3198 boolean_t is_insensitive = B_FALSE;
3193 3199
3194 3200 if (nvlist_lookup_int32(innvl, "type", &type32) != 0)
3195 3201 return (SET_ERROR(EINVAL));
3196 3202 type = type32;
3197 3203 (void) nvlist_lookup_nvlist(innvl, "props", &nvprops);
3198 3204
3199 3205 switch (type) {
3200 3206 case DMU_OST_ZFS:
3201 3207 cbfunc = zfs_create_cb;
3202 3208 break;
3203 3209
3204 3210 case DMU_OST_ZVOL:
3205 3211 cbfunc = zvol_create_cb;
3206 3212 break;
3207 3213
3208 3214 default:
3209 3215 cbfunc = NULL;
3210 3216 break;
3211 3217 }
3212 3218 if (strchr(fsname, '@') ||
3213 3219 strchr(fsname, '%'))
3214 3220 return (SET_ERROR(EINVAL));
3215 3221
3216 3222 zct.zct_props = nvprops;
3217 3223
3218 3224 if (cbfunc == NULL)
3219 3225 return (SET_ERROR(EINVAL));
3220 3226
3221 3227 if (type == DMU_OST_ZVOL) {
3222 3228 uint64_t volsize, volblocksize;
3223 3229
3224 3230 if (nvprops == NULL)
3225 3231 return (SET_ERROR(EINVAL));
3226 3232 if (nvlist_lookup_uint64(nvprops,
3227 3233 zfs_prop_to_name(ZFS_PROP_VOLSIZE), &volsize) != 0)
3228 3234 return (SET_ERROR(EINVAL));
3229 3235
3230 3236 if ((error = nvlist_lookup_uint64(nvprops,
3231 3237 zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE),
3232 3238 &volblocksize)) != 0 && error != ENOENT)
3233 3239 return (SET_ERROR(EINVAL));
3234 3240
3235 3241 if (error != 0)
3236 3242 volblocksize = zfs_prop_default_numeric(
3237 3243 ZFS_PROP_VOLBLOCKSIZE);
3238 3244
3239 3245 if ((error = zvol_check_volblocksize(
3240 3246 volblocksize)) != 0 ||
3241 3247 (error = zvol_check_volsize(volsize,
3242 3248 volblocksize)) != 0)
3243 3249 return (error);
3244 3250 } else if (type == DMU_OST_ZFS) {
3245 3251 int error;
3246 3252
3247 3253 /*
3248 3254 * We have to have normalization and
3249 3255 * case-folding flags correct when we do the
3250 3256 * file system creation, so go figure them out
3251 3257 * now.
3252 3258 */
3253 3259 VERIFY(nvlist_alloc(&zct.zct_zplprops,
3254 3260 NV_UNIQUE_NAME, KM_SLEEP) == 0);
3255 3261 error = zfs_fill_zplprops(fsname, nvprops,
3256 3262 zct.zct_zplprops, &is_insensitive);
3257 3263 if (error != 0) {
3258 3264 nvlist_free(zct.zct_zplprops);
3259 3265 return (error);
3260 3266 }
3261 3267 }
3262 3268
3263 3269 error = dmu_objset_create(fsname, type,
3264 3270 is_insensitive ? DS_FLAG_CI_DATASET : 0, cbfunc, &zct);
3265 3271 nvlist_free(zct.zct_zplprops);
3266 3272
3267 3273 /*
3268 3274 * It would be nice to do this atomically.
3269 3275 */
3270 3276 if (error == 0) {
3271 3277 error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL,
3272 3278 nvprops, outnvl);
3273 3279 if (error != 0)
3274 3280 (void) dsl_destroy_head(fsname);
3275 3281 }
3276 3282 return (error);
3277 3283 }
3278 3284
3279 3285 /*
3280 3286 * innvl: {
3281 3287 * "origin" -> name of origin snapshot
3282 3288 * (optional) "props" -> { prop -> value }
3283 3289 * }
3284 3290 *
3285 3291 * outnvl: propname -> error code (int32)
3286 3292 */
3287 3293 static int
3288 3294 zfs_ioc_clone(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3289 3295 {
3290 3296 int error = 0;
3291 3297 nvlist_t *nvprops = NULL;
3292 3298 char *origin_name;
3293 3299
3294 3300 if (nvlist_lookup_string(innvl, "origin", &origin_name) != 0)
3295 3301 return (SET_ERROR(EINVAL));
3296 3302 (void) nvlist_lookup_nvlist(innvl, "props", &nvprops);
3297 3303
3298 3304 if (strchr(fsname, '@') ||
3299 3305 strchr(fsname, '%'))
3300 3306 return (SET_ERROR(EINVAL));
3301 3307
3302 3308 if (dataset_namecheck(origin_name, NULL, NULL) != 0)
3303 3309 return (SET_ERROR(EINVAL));
3304 3310 error = dmu_objset_clone(fsname, origin_name);
3305 3311 if (error != 0)
3306 3312 return (error);
3307 3313
3308 3314 /*
3309 3315 * It would be nice to do this atomically.
3310 3316 */
3311 3317 if (error == 0) {
3312 3318 error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL,
3313 3319 nvprops, outnvl);
3314 3320 if (error != 0)
3315 3321 (void) dsl_destroy_head(fsname);
3316 3322 }
3317 3323 return (error);
3318 3324 }
3319 3325
3320 3326 /* ARGSUSED */
3321 3327 static int
3322 3328 zfs_ioc_remap(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3323 3329 {
3324 3330 if (strchr(fsname, '@') ||
3325 3331 strchr(fsname, '%'))
3326 3332 return (SET_ERROR(EINVAL));
3327 3333
3328 3334 return (dmu_objset_remap_indirects(fsname));
3329 3335 }
3330 3336
3331 3337 /*
3332 3338 * innvl: {
3333 3339 * "snaps" -> { snapshot1, snapshot2 }
3334 3340 * (optional) "props" -> { prop -> value (string) }
3335 3341 * }
3336 3342 *
3337 3343 * outnvl: snapshot -> error code (int32)
3338 3344 */
3339 3345 static int
3340 3346 zfs_ioc_snapshot(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3341 3347 {
3342 3348 nvlist_t *snaps;
3343 3349 nvlist_t *props = NULL;
3344 3350 int error, poollen;
3345 3351 nvpair_t *pair;
3346 3352
3347 3353 (void) nvlist_lookup_nvlist(innvl, "props", &props);
3348 3354 if ((error = zfs_check_userprops(poolname, props)) != 0)
3349 3355 return (error);
3350 3356
3351 3357 if (!nvlist_empty(props) &&
3352 3358 zfs_earlier_version(poolname, SPA_VERSION_SNAP_PROPS))
3353 3359 return (SET_ERROR(ENOTSUP));
3354 3360
3355 3361 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
3356 3362 return (SET_ERROR(EINVAL));
3357 3363 poollen = strlen(poolname);
3358 3364 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
3359 3365 pair = nvlist_next_nvpair(snaps, pair)) {
3360 3366 const char *name = nvpair_name(pair);
3361 3367 const char *cp = strchr(name, '@');
3362 3368
3363 3369 /*
3364 3370 * The snap name must contain an @, and the part after it must
3365 3371 * contain only valid characters.
3366 3372 */
3367 3373 if (cp == NULL ||
3368 3374 zfs_component_namecheck(cp + 1, NULL, NULL) != 0)
3369 3375 return (SET_ERROR(EINVAL));
3370 3376
3371 3377 /*
3372 3378 * The snap must be in the specified pool.
3373 3379 */
3374 3380 if (strncmp(name, poolname, poollen) != 0 ||
3375 3381 (name[poollen] != '/' && name[poollen] != '@'))
3376 3382 return (SET_ERROR(EXDEV));
3377 3383
3378 3384 /* This must be the only snap of this fs. */
3379 3385 for (nvpair_t *pair2 = nvlist_next_nvpair(snaps, pair);
3380 3386 pair2 != NULL; pair2 = nvlist_next_nvpair(snaps, pair2)) {
3381 3387 if (strncmp(name, nvpair_name(pair2), cp - name + 1)
3382 3388 == 0) {
3383 3389 return (SET_ERROR(EXDEV));
3384 3390 }
3385 3391 }
3386 3392 }
3387 3393
3388 3394 error = dsl_dataset_snapshot(snaps, props, outnvl);
3389 3395 return (error);
3390 3396 }
3391 3397
3392 3398 /*
3393 3399 * innvl: "message" -> string
3394 3400 */
3395 3401 /* ARGSUSED */
3396 3402 static int
3397 3403 zfs_ioc_log_history(const char *unused, nvlist_t *innvl, nvlist_t *outnvl)
3398 3404 {
3399 3405 char *message;
3400 3406 spa_t *spa;
3401 3407 int error;
3402 3408 char *poolname;
3403 3409
3404 3410 /*
3405 3411 * The poolname in the ioctl is not set, we get it from the TSD,
3406 3412 * which was set at the end of the last successful ioctl that allows
3407 3413 * logging. The secpolicy func already checked that it is set.
3408 3414 * Only one log ioctl is allowed after each successful ioctl, so
3409 3415 * we clear the TSD here.
3410 3416 */
3411 3417 poolname = tsd_get(zfs_allow_log_key);
3412 3418 (void) tsd_set(zfs_allow_log_key, NULL);
3413 3419 error = spa_open(poolname, &spa, FTAG);
3414 3420 strfree(poolname);
3415 3421 if (error != 0)
3416 3422 return (error);
3417 3423
3418 3424 if (nvlist_lookup_string(innvl, "message", &message) != 0) {
3419 3425 spa_close(spa, FTAG);
3420 3426 return (SET_ERROR(EINVAL));
3421 3427 }
3422 3428
3423 3429 if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) {
3424 3430 spa_close(spa, FTAG);
3425 3431 return (SET_ERROR(ENOTSUP));
3426 3432 }
3427 3433
3428 3434 error = spa_history_log(spa, message);
3429 3435 spa_close(spa, FTAG);
3430 3436 return (error);
3431 3437 }
3432 3438
3433 3439 /*
3434 3440 * The dp_config_rwlock must not be held when calling this, because the
3435 3441 * unmount may need to write out data.
3436 3442 *
3437 3443 * This function is best-effort. Callers must deal gracefully if it
3438 3444 * remains mounted (or is remounted after this call).
3439 3445 *
3440 3446 * Returns 0 if the argument is not a snapshot, or it is not currently a
3441 3447 * filesystem, or we were able to unmount it. Returns error code otherwise.
3442 3448 */
3443 3449 void
3444 3450 zfs_unmount_snap(const char *snapname)
3445 3451 {
3446 3452 vfs_t *vfsp = NULL;
3447 3453 zfsvfs_t *zfsvfs = NULL;
3448 3454
3449 3455 if (strchr(snapname, '@') == NULL)
3450 3456 return;
3451 3457
3452 3458 int err = getzfsvfs(snapname, &zfsvfs);
3453 3459 if (err != 0) {
3454 3460 ASSERT3P(zfsvfs, ==, NULL);
3455 3461 return;
3456 3462 }
3457 3463 vfsp = zfsvfs->z_vfs;
3458 3464
3459 3465 ASSERT(!dsl_pool_config_held(dmu_objset_pool(zfsvfs->z_os)));
3460 3466
3461 3467 err = vn_vfswlock(vfsp->vfs_vnodecovered);
3462 3468 VFS_RELE(vfsp);
3463 3469 if (err != 0)
3464 3470 return;
3465 3471
3466 3472 /*
3467 3473 * Always force the unmount for snapshots.
3468 3474 */
3469 3475 (void) dounmount(vfsp, MS_FORCE, kcred);
3470 3476 }
3471 3477
3472 3478 /* ARGSUSED */
3473 3479 static int
3474 3480 zfs_unmount_snap_cb(const char *snapname, void *arg)
3475 3481 {
3476 3482 zfs_unmount_snap(snapname);
3477 3483 return (0);
3478 3484 }
3479 3485
3480 3486 /*
3481 3487 * When a clone is destroyed, its origin may also need to be destroyed,
3482 3488 * in which case it must be unmounted. This routine will do that unmount
3483 3489 * if necessary.
3484 3490 */
3485 3491 void
3486 3492 zfs_destroy_unmount_origin(const char *fsname)
3487 3493 {
3488 3494 int error;
3489 3495 objset_t *os;
3490 3496 dsl_dataset_t *ds;
3491 3497
3492 3498 error = dmu_objset_hold(fsname, FTAG, &os);
3493 3499 if (error != 0)
3494 3500 return;
3495 3501 ds = dmu_objset_ds(os);
3496 3502 if (dsl_dir_is_clone(ds->ds_dir) && DS_IS_DEFER_DESTROY(ds->ds_prev)) {
3497 3503 char originname[ZFS_MAX_DATASET_NAME_LEN];
3498 3504 dsl_dataset_name(ds->ds_prev, originname);
3499 3505 dmu_objset_rele(os, FTAG);
3500 3506 zfs_unmount_snap(originname);
3501 3507 } else {
3502 3508 dmu_objset_rele(os, FTAG);
3503 3509 }
3504 3510 }
3505 3511
3506 3512 /*
3507 3513 * innvl: {
3508 3514 * "snaps" -> { snapshot1, snapshot2 }
3509 3515 * (optional boolean) "defer"
3510 3516 * }
3511 3517 *
3512 3518 * outnvl: snapshot -> error code (int32)
3513 3519 *
3514 3520 */
3515 3521 /* ARGSUSED */
3516 3522 static int
3517 3523 zfs_ioc_destroy_snaps(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3518 3524 {
3519 3525 nvlist_t *snaps;
3520 3526 nvpair_t *pair;
3521 3527 boolean_t defer;
3522 3528
3523 3529 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
3524 3530 return (SET_ERROR(EINVAL));
3525 3531 defer = nvlist_exists(innvl, "defer");
3526 3532
3527 3533 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
3528 3534 pair = nvlist_next_nvpair(snaps, pair)) {
3529 3535 zfs_unmount_snap(nvpair_name(pair));
3530 3536 }
3531 3537
3532 3538 return (dsl_destroy_snapshots_nvl(snaps, defer, outnvl));
3533 3539 }
3534 3540
3535 3541 /*
3536 3542 * Create bookmarks. Bookmark names are of the form <fs>#<bmark>.
3537 3543 * All bookmarks must be in the same pool.
3538 3544 *
3539 3545 * innvl: {
3540 3546 * bookmark1 -> snapshot1, bookmark2 -> snapshot2
3541 3547 * }
3542 3548 *
3543 3549 * outnvl: bookmark -> error code (int32)
3544 3550 *
3545 3551 */
3546 3552 /* ARGSUSED */
3547 3553 static int
3548 3554 zfs_ioc_bookmark(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3549 3555 {
3550 3556 for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL);
3551 3557 pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) {
3552 3558 char *snap_name;
3553 3559
3554 3560 /*
3555 3561 * Verify the snapshot argument.
3556 3562 */
3557 3563 if (nvpair_value_string(pair, &snap_name) != 0)
3558 3564 return (SET_ERROR(EINVAL));
3559 3565
3560 3566
3561 3567 /* Verify that the keys (bookmarks) are unique */
3562 3568 for (nvpair_t *pair2 = nvlist_next_nvpair(innvl, pair);
3563 3569 pair2 != NULL; pair2 = nvlist_next_nvpair(innvl, pair2)) {
3564 3570 if (strcmp(nvpair_name(pair), nvpair_name(pair2)) == 0)
3565 3571 return (SET_ERROR(EINVAL));
3566 3572 }
3567 3573 }
3568 3574
3569 3575 return (dsl_bookmark_create(innvl, outnvl));
3570 3576 }
3571 3577
3572 3578 /*
3573 3579 * innvl: {
3574 3580 * property 1, property 2, ...
3575 3581 * }
3576 3582 *
3577 3583 * outnvl: {
3578 3584 * bookmark name 1 -> { property 1, property 2, ... },
3579 3585 * bookmark name 2 -> { property 1, property 2, ... }
3580 3586 * }
3581 3587 *
3582 3588 */
3583 3589 static int
3584 3590 zfs_ioc_get_bookmarks(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3585 3591 {
3586 3592 return (dsl_get_bookmarks(fsname, innvl, outnvl));
3587 3593 }
3588 3594
3589 3595 /*
3590 3596 * innvl: {
3591 3597 * bookmark name 1, bookmark name 2
3592 3598 * }
3593 3599 *
3594 3600 * outnvl: bookmark -> error code (int32)
3595 3601 *
3596 3602 */
3597 3603 static int
3598 3604 zfs_ioc_destroy_bookmarks(const char *poolname, nvlist_t *innvl,
3599 3605 nvlist_t *outnvl)
3600 3606 {
3601 3607 int error, poollen;
3602 3608
3603 3609 poollen = strlen(poolname);
3604 3610 for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL);
3605 3611 pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) {
3606 3612 const char *name = nvpair_name(pair);
3607 3613 const char *cp = strchr(name, '#');
3608 3614
3609 3615 /*
3610 3616 * The bookmark name must contain an #, and the part after it
3611 3617 * must contain only valid characters.
3612 3618 */
3613 3619 if (cp == NULL ||
3614 3620 zfs_component_namecheck(cp + 1, NULL, NULL) != 0)
3615 3621 return (SET_ERROR(EINVAL));
3616 3622
3617 3623 /*
3618 3624 * The bookmark must be in the specified pool.
3619 3625 */
3620 3626 if (strncmp(name, poolname, poollen) != 0 ||
3621 3627 (name[poollen] != '/' && name[poollen] != '#'))
3622 3628 return (SET_ERROR(EXDEV));
3623 3629 }
3624 3630
3625 3631 error = dsl_bookmark_destroy(innvl, outnvl);
3626 3632 return (error);
3627 3633 }
3628 3634
3629 3635 static int
3630 3636 zfs_ioc_channel_program(const char *poolname, nvlist_t *innvl,
3631 3637 nvlist_t *outnvl)
3632 3638 {
3633 3639 char *program;
3634 3640 uint64_t instrlimit, memlimit;
3635 3641 boolean_t sync_flag;
3636 3642 nvpair_t *nvarg = NULL;
3637 3643
3638 3644 if (0 != nvlist_lookup_string(innvl, ZCP_ARG_PROGRAM, &program)) {
3639 3645 return (EINVAL);
3640 3646 }
3641 3647 if (0 != nvlist_lookup_boolean_value(innvl, ZCP_ARG_SYNC, &sync_flag)) {
3642 3648 sync_flag = B_TRUE;
3643 3649 }
3644 3650 if (0 != nvlist_lookup_uint64(innvl, ZCP_ARG_INSTRLIMIT, &instrlimit)) {
3645 3651 instrlimit = ZCP_DEFAULT_INSTRLIMIT;
3646 3652 }
3647 3653 if (0 != nvlist_lookup_uint64(innvl, ZCP_ARG_MEMLIMIT, &memlimit)) {
3648 3654 memlimit = ZCP_DEFAULT_MEMLIMIT;
3649 3655 }
3650 3656 if (0 != nvlist_lookup_nvpair(innvl, ZCP_ARG_ARGLIST, &nvarg)) {
3651 3657 return (EINVAL);
3652 3658 }
3653 3659
3654 3660 if (instrlimit == 0 || instrlimit > zfs_lua_max_instrlimit)
3655 3661 return (EINVAL);
3656 3662 if (memlimit == 0 || memlimit > zfs_lua_max_memlimit)
3657 3663 return (EINVAL);
3658 3664
3659 3665 return (zcp_eval(poolname, program, sync_flag, instrlimit, memlimit,
3660 3666 nvarg, outnvl));
3661 3667 }
3662 3668
3663 3669 /*
3664 3670 * innvl: unused
3665 3671 * outnvl: empty
3666 3672 */
3667 3673 /* ARGSUSED */
3668 3674 static int
3669 3675 zfs_ioc_pool_checkpoint(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3670 3676 {
3671 3677 return (spa_checkpoint(poolname));
3672 3678 }
3673 3679
3674 3680 /*
3675 3681 * innvl: unused
3676 3682 * outnvl: empty
3677 3683 */
3678 3684 /* ARGSUSED */
3679 3685 static int
3680 3686 zfs_ioc_pool_discard_checkpoint(const char *poolname, nvlist_t *innvl,
3681 3687 nvlist_t *outnvl)
3682 3688 {
3683 3689 return (spa_checkpoint_discard(poolname));
3684 3690 }
3685 3691
3686 3692 /*
3687 3693 * inputs:
3688 3694 * zc_name name of dataset to destroy
3689 3695 * zc_defer_destroy mark for deferred destroy
3690 3696 *
3691 3697 * outputs: none
3692 3698 */
3693 3699 static int
3694 3700 zfs_ioc_destroy(zfs_cmd_t *zc)
3695 3701 {
3696 3702 objset_t *os;
3697 3703 dmu_objset_type_t ost;
3698 3704 int err;
3699 3705
3700 3706 err = dmu_objset_hold(zc->zc_name, FTAG, &os);
3701 3707 if (err != 0)
3702 3708 return (err);
3703 3709 ost = dmu_objset_type(os);
3704 3710 dmu_objset_rele(os, FTAG);
3705 3711
3706 3712 if (ost == DMU_OST_ZFS)
3707 3713 zfs_unmount_snap(zc->zc_name);
3708 3714
3709 3715 if (strchr(zc->zc_name, '@'))
3710 3716 err = dsl_destroy_snapshot(zc->zc_name, zc->zc_defer_destroy);
3711 3717 else
3712 3718 err = dsl_destroy_head(zc->zc_name);
3713 3719 if (ost == DMU_OST_ZVOL && err == 0)
3714 3720 (void) zvol_remove_minor(zc->zc_name);
3715 3721 return (err);
3716 3722 }
3717 3723
3718 3724 /*
3719 3725 * innvl: {
3720 3726 * vdevs: {
3721 3727 * guid 1, guid 2, ...
3722 3728 * },
3723 3729 * func: POOL_INITIALIZE_{CANCEL|DO|SUSPEND}
3724 3730 * }
3725 3731 *
3726 3732 * outnvl: {
3727 3733 * [func: EINVAL (if provided command type didn't make sense)],
3728 3734 * [vdevs: {
3729 3735 * guid1: errno, (see function body for possible errnos)
3730 3736 * ...
3731 3737 * }]
3732 3738 * }
3733 3739 *
3734 3740 */
3735 3741 static int
3736 3742 zfs_ioc_pool_initialize(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3737 3743 {
3738 3744 spa_t *spa;
3739 3745 int error;
3740 3746
3741 3747 error = spa_open(poolname, &spa, FTAG);
3742 3748 if (error != 0)
3743 3749 return (error);
3744 3750
3745 3751 uint64_t cmd_type;
3746 3752 if (nvlist_lookup_uint64(innvl, ZPOOL_INITIALIZE_COMMAND,
3747 3753 &cmd_type) != 0) {
3748 3754 spa_close(spa, FTAG);
3749 3755 return (SET_ERROR(EINVAL));
3750 3756 }
3751 3757 if (!(cmd_type == POOL_INITIALIZE_CANCEL ||
3752 3758 cmd_type == POOL_INITIALIZE_DO ||
3753 3759 cmd_type == POOL_INITIALIZE_SUSPEND)) {
3754 3760 spa_close(spa, FTAG);
3755 3761 return (SET_ERROR(EINVAL));
3756 3762 }
3757 3763
3758 3764 nvlist_t *vdev_guids;
3759 3765 if (nvlist_lookup_nvlist(innvl, ZPOOL_INITIALIZE_VDEVS,
3760 3766 &vdev_guids) != 0) {
3761 3767 spa_close(spa, FTAG);
3762 3768 return (SET_ERROR(EINVAL));
3763 3769 }
3764 3770
3765 3771 nvlist_t *vdev_errlist = fnvlist_alloc();
3766 3772 int total_errors = 0;
3767 3773
3768 3774 for (nvpair_t *pair = nvlist_next_nvpair(vdev_guids, NULL);
3769 3775 pair != NULL; pair = nvlist_next_nvpair(vdev_guids, pair)) {
3770 3776 uint64_t vdev_guid = fnvpair_value_uint64(pair);
3771 3777
3772 3778 error = spa_vdev_initialize(spa, vdev_guid, cmd_type);
3773 3779 if (error != 0) {
3774 3780 char guid_as_str[MAXNAMELEN];
3775 3781
3776 3782 (void) snprintf(guid_as_str, sizeof (guid_as_str),
3777 3783 "%llu", (unsigned long long)vdev_guid);
3778 3784 fnvlist_add_int64(vdev_errlist, guid_as_str, error);
3779 3785 total_errors++;
3780 3786 }
3781 3787 }
3782 3788 if (fnvlist_size(vdev_errlist) > 0) {
3783 3789 fnvlist_add_nvlist(outnvl, ZPOOL_INITIALIZE_VDEVS,
3784 3790 vdev_errlist);
3785 3791 }
3786 3792 fnvlist_free(vdev_errlist);
3787 3793
3788 3794 spa_close(spa, FTAG);
3789 3795 return (total_errors > 0 ? EINVAL : 0);
3790 3796 }
3791 3797
3792 3798 /*
3793 3799 * fsname is name of dataset to rollback (to most recent snapshot)
3794 3800 *
3795 3801 * innvl may contain name of expected target snapshot
3796 3802 *
3797 3803 * outnvl: "target" -> name of most recent snapshot
3798 3804 * }
3799 3805 */
3800 3806 /* ARGSUSED */
3801 3807 static int
3802 3808 zfs_ioc_rollback(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3803 3809 {
3804 3810 zfsvfs_t *zfsvfs;
3805 3811 char *target = NULL;
3806 3812 int error;
3807 3813
3808 3814 (void) nvlist_lookup_string(innvl, "target", &target);
3809 3815 if (target != NULL) {
3810 3816 const char *cp = strchr(target, '@');
3811 3817
3812 3818 /*
3813 3819 * The snap name must contain an @, and the part after it must
3814 3820 * contain only valid characters.
3815 3821 */
3816 3822 if (cp == NULL ||
3817 3823 zfs_component_namecheck(cp + 1, NULL, NULL) != 0)
3818 3824 return (SET_ERROR(EINVAL));
3819 3825 }
3820 3826
3821 3827 if (getzfsvfs(fsname, &zfsvfs) == 0) {
3822 3828 dsl_dataset_t *ds;
3823 3829
3824 3830 ds = dmu_objset_ds(zfsvfs->z_os);
3825 3831 error = zfs_suspend_fs(zfsvfs);
3826 3832 if (error == 0) {
3827 3833 int resume_err;
3828 3834
3829 3835 error = dsl_dataset_rollback(fsname, target, zfsvfs,
3830 3836 outnvl);
3831 3837 resume_err = zfs_resume_fs(zfsvfs, ds);
3832 3838 error = error ? error : resume_err;
3833 3839 }
3834 3840 VFS_RELE(zfsvfs->z_vfs);
3835 3841 } else {
3836 3842 error = dsl_dataset_rollback(fsname, target, NULL, outnvl);
3837 3843 }
3838 3844 return (error);
3839 3845 }
3840 3846
3841 3847 static int
3842 3848 recursive_unmount(const char *fsname, void *arg)
3843 3849 {
3844 3850 const char *snapname = arg;
3845 3851 char fullname[ZFS_MAX_DATASET_NAME_LEN];
3846 3852
3847 3853 (void) snprintf(fullname, sizeof (fullname), "%s@%s", fsname, snapname);
3848 3854 zfs_unmount_snap(fullname);
3849 3855
3850 3856 return (0);
3851 3857 }
3852 3858
3853 3859 /*
3854 3860 * inputs:
3855 3861 * zc_name old name of dataset
3856 3862 * zc_value new name of dataset
3857 3863 * zc_cookie recursive flag (only valid for snapshots)
3858 3864 *
3859 3865 * outputs: none
3860 3866 */
3861 3867 static int
3862 3868 zfs_ioc_rename(zfs_cmd_t *zc)
3863 3869 {
3864 3870 objset_t *os;
3865 3871 dmu_objset_type_t ost;
3866 3872 boolean_t recursive = zc->zc_cookie & 1;
3867 3873 char *at;
3868 3874 int err;
3869 3875
3870 3876 /* "zfs rename" from and to ...%recv datasets should both fail */
3871 3877 zc->zc_name[sizeof (zc->zc_name) - 1] = '\0';
3872 3878 zc->zc_value[sizeof (zc->zc_value) - 1] = '\0';
3873 3879 if (dataset_namecheck(zc->zc_name, NULL, NULL) != 0 ||
3874 3880 dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
3875 3881 strchr(zc->zc_name, '%') || strchr(zc->zc_value, '%'))
3876 3882 return (SET_ERROR(EINVAL));
3877 3883
3878 3884 err = dmu_objset_hold(zc->zc_name, FTAG, &os);
3879 3885 if (err != 0)
3880 3886 return (err);
3881 3887 ost = dmu_objset_type(os);
3882 3888 dmu_objset_rele(os, FTAG);
3883 3889
3884 3890 at = strchr(zc->zc_name, '@');
3885 3891 if (at != NULL) {
3886 3892 /* snaps must be in same fs */
3887 3893 int error;
3888 3894
3889 3895 if (strncmp(zc->zc_name, zc->zc_value, at - zc->zc_name + 1))
3890 3896 return (SET_ERROR(EXDEV));
3891 3897 *at = '\0';
3892 3898 if (ost == DMU_OST_ZFS) {
3893 3899 error = dmu_objset_find(zc->zc_name,
3894 3900 recursive_unmount, at + 1,
3895 3901 recursive ? DS_FIND_CHILDREN : 0);
3896 3902 if (error != 0) {
3897 3903 *at = '@';
3898 3904 return (error);
3899 3905 }
3900 3906 }
3901 3907 error = dsl_dataset_rename_snapshot(zc->zc_name,
3902 3908 at + 1, strchr(zc->zc_value, '@') + 1, recursive);
3903 3909 *at = '@';
3904 3910
3905 3911 return (error);
3906 3912 } else {
3907 3913 if (ost == DMU_OST_ZVOL)
3908 3914 (void) zvol_remove_minor(zc->zc_name);
3909 3915 return (dsl_dir_rename(zc->zc_name, zc->zc_value));
3910 3916 }
3911 3917 }
3912 3918
3913 3919 static int
3914 3920 zfs_check_settable(const char *dsname, nvpair_t *pair, cred_t *cr)
3915 3921 {
3916 3922 const char *propname = nvpair_name(pair);
3917 3923 boolean_t issnap = (strchr(dsname, '@') != NULL);
3918 3924 zfs_prop_t prop = zfs_name_to_prop(propname);
3919 3925 uint64_t intval;
3920 3926 int err;
3921 3927
3922 3928 if (prop == ZPROP_INVAL) {
3923 3929 if (zfs_prop_user(propname)) {
3924 3930 if (err = zfs_secpolicy_write_perms(dsname,
3925 3931 ZFS_DELEG_PERM_USERPROP, cr))
3926 3932 return (err);
3927 3933 return (0);
3928 3934 }
3929 3935
3930 3936 if (!issnap && zfs_prop_userquota(propname)) {
3931 3937 const char *perm = NULL;
3932 3938 const char *uq_prefix =
3933 3939 zfs_userquota_prop_prefixes[ZFS_PROP_USERQUOTA];
3934 3940 const char *gq_prefix =
3935 3941 zfs_userquota_prop_prefixes[ZFS_PROP_GROUPQUOTA];
3936 3942
3937 3943 if (strncmp(propname, uq_prefix,
3938 3944 strlen(uq_prefix)) == 0) {
3939 3945 perm = ZFS_DELEG_PERM_USERQUOTA;
3940 3946 } else if (strncmp(propname, gq_prefix,
3941 3947 strlen(gq_prefix)) == 0) {
3942 3948 perm = ZFS_DELEG_PERM_GROUPQUOTA;
3943 3949 } else {
3944 3950 /* USERUSED and GROUPUSED are read-only */
3945 3951 return (SET_ERROR(EINVAL));
3946 3952 }
3947 3953
3948 3954 if (err = zfs_secpolicy_write_perms(dsname, perm, cr))
3949 3955 return (err);
3950 3956 return (0);
3951 3957 }
3952 3958
3953 3959 return (SET_ERROR(EINVAL));
3954 3960 }
3955 3961
3956 3962 if (issnap)
3957 3963 return (SET_ERROR(EINVAL));
3958 3964
3959 3965 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
3960 3966 /*
3961 3967 * dsl_prop_get_all_impl() returns properties in this
3962 3968 * format.
3963 3969 */
3964 3970 nvlist_t *attrs;
3965 3971 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
3966 3972 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
3967 3973 &pair) == 0);
3968 3974 }
3969 3975
3970 3976 /*
3971 3977 * Check that this value is valid for this pool version
3972 3978 */
3973 3979 switch (prop) {
3974 3980 case ZFS_PROP_COMPRESSION:
3975 3981 /*
3976 3982 * If the user specified gzip compression, make sure
3977 3983 * the SPA supports it. We ignore any errors here since
3978 3984 * we'll catch them later.
3979 3985 */
3980 3986 if (nvpair_value_uint64(pair, &intval) == 0) {
3981 3987 if (intval >= ZIO_COMPRESS_GZIP_1 &&
3982 3988 intval <= ZIO_COMPRESS_GZIP_9 &&
3983 3989 zfs_earlier_version(dsname,
3984 3990 SPA_VERSION_GZIP_COMPRESSION)) {
3985 3991 return (SET_ERROR(ENOTSUP));
3986 3992 }
3987 3993
3988 3994 if (intval == ZIO_COMPRESS_ZLE &&
3989 3995 zfs_earlier_version(dsname,
3990 3996 SPA_VERSION_ZLE_COMPRESSION))
3991 3997 return (SET_ERROR(ENOTSUP));
3992 3998
3993 3999 if (intval == ZIO_COMPRESS_LZ4) {
3994 4000 spa_t *spa;
3995 4001
3996 4002 if ((err = spa_open(dsname, &spa, FTAG)) != 0)
3997 4003 return (err);
3998 4004
3999 4005 if (!spa_feature_is_enabled(spa,
4000 4006 SPA_FEATURE_LZ4_COMPRESS)) {
4001 4007 spa_close(spa, FTAG);
4002 4008 return (SET_ERROR(ENOTSUP));
4003 4009 }
4004 4010 spa_close(spa, FTAG);
4005 4011 }
4006 4012
4007 4013 /*
4008 4014 * If this is a bootable dataset then
4009 4015 * verify that the compression algorithm
4010 4016 * is supported for booting. We must return
4011 4017 * something other than ENOTSUP since it
4012 4018 * implies a downrev pool version.
4013 4019 */
4014 4020 if (zfs_is_bootfs(dsname) &&
4015 4021 !BOOTFS_COMPRESS_VALID(intval)) {
4016 4022 return (SET_ERROR(ERANGE));
4017 4023 }
4018 4024 }
4019 4025 break;
4020 4026
4021 4027 case ZFS_PROP_COPIES:
4022 4028 if (zfs_earlier_version(dsname, SPA_VERSION_DITTO_BLOCKS))
4023 4029 return (SET_ERROR(ENOTSUP));
4024 4030 break;
4025 4031
4026 4032 case ZFS_PROP_RECORDSIZE:
4027 4033 /* Record sizes above 128k need the feature to be enabled */
4028 4034 if (nvpair_value_uint64(pair, &intval) == 0 &&
4029 4035 intval > SPA_OLD_MAXBLOCKSIZE) {
4030 4036 spa_t *spa;
4031 4037
4032 4038 /*
4033 4039 * We don't allow setting the property above 1MB,
4034 4040 * unless the tunable has been changed.
4035 4041 */
4036 4042 if (intval > zfs_max_recordsize ||
4037 4043 intval > SPA_MAXBLOCKSIZE)
4038 4044 return (SET_ERROR(ERANGE));
4039 4045
4040 4046 if ((err = spa_open(dsname, &spa, FTAG)) != 0)
4041 4047 return (err);
4042 4048
4043 4049 if (!spa_feature_is_enabled(spa,
4044 4050 SPA_FEATURE_LARGE_BLOCKS)) {
4045 4051 spa_close(spa, FTAG);
4046 4052 return (SET_ERROR(ENOTSUP));
4047 4053 }
4048 4054 spa_close(spa, FTAG);
4049 4055 }
4050 4056 break;
4051 4057
4052 4058 case ZFS_PROP_SHARESMB:
4053 4059 if (zpl_earlier_version(dsname, ZPL_VERSION_FUID))
4054 4060 return (SET_ERROR(ENOTSUP));
4055 4061 break;
4056 4062
4057 4063 case ZFS_PROP_ACLINHERIT:
4058 4064 if (nvpair_type(pair) == DATA_TYPE_UINT64 &&
4059 4065 nvpair_value_uint64(pair, &intval) == 0) {
4060 4066 if (intval == ZFS_ACL_PASSTHROUGH_X &&
4061 4067 zfs_earlier_version(dsname,
4062 4068 SPA_VERSION_PASSTHROUGH_X))
4063 4069 return (SET_ERROR(ENOTSUP));
4064 4070 }
4065 4071 break;
4066 4072
4067 4073 case ZFS_PROP_CHECKSUM:
4068 4074 case ZFS_PROP_DEDUP:
4069 4075 {
4070 4076 spa_feature_t feature;
4071 4077 spa_t *spa;
4072 4078
4073 4079 /* dedup feature version checks */
4074 4080 if (prop == ZFS_PROP_DEDUP &&
4075 4081 zfs_earlier_version(dsname, SPA_VERSION_DEDUP))
4076 4082 return (SET_ERROR(ENOTSUP));
4077 4083
4078 4084 if (nvpair_value_uint64(pair, &intval) != 0)
4079 4085 return (SET_ERROR(EINVAL));
4080 4086
4081 4087 /* check prop value is enabled in features */
4082 4088 feature = zio_checksum_to_feature(intval & ZIO_CHECKSUM_MASK);
4083 4089 if (feature == SPA_FEATURE_NONE)
4084 4090 break;
4085 4091
4086 4092 if ((err = spa_open(dsname, &spa, FTAG)) != 0)
4087 4093 return (err);
4088 4094
4089 4095 if (!spa_feature_is_enabled(spa, feature)) {
4090 4096 spa_close(spa, FTAG);
4091 4097 return (SET_ERROR(ENOTSUP));
4092 4098 }
4093 4099 spa_close(spa, FTAG);
4094 4100 break;
4095 4101 }
4096 4102 }
4097 4103
4098 4104 return (zfs_secpolicy_setprop(dsname, prop, pair, CRED()));
4099 4105 }
4100 4106
4101 4107 /*
4102 4108 * Checks for a race condition to make sure we don't increment a feature flag
4103 4109 * multiple times.
4104 4110 */
4105 4111 static int
4106 4112 zfs_prop_activate_feature_check(void *arg, dmu_tx_t *tx)
4107 4113 {
4108 4114 spa_t *spa = dmu_tx_pool(tx)->dp_spa;
4109 4115 spa_feature_t *featurep = arg;
4110 4116
4111 4117 if (!spa_feature_is_active(spa, *featurep))
4112 4118 return (0);
4113 4119 else
4114 4120 return (SET_ERROR(EBUSY));
4115 4121 }
4116 4122
4117 4123 /*
4118 4124 * The callback invoked on feature activation in the sync task caused by
4119 4125 * zfs_prop_activate_feature.
4120 4126 */
4121 4127 static void
4122 4128 zfs_prop_activate_feature_sync(void *arg, dmu_tx_t *tx)
4123 4129 {
4124 4130 spa_t *spa = dmu_tx_pool(tx)->dp_spa;
4125 4131 spa_feature_t *featurep = arg;
4126 4132
4127 4133 spa_feature_incr(spa, *featurep, tx);
4128 4134 }
4129 4135
4130 4136 /*
4131 4137 * Activates a feature on a pool in response to a property setting. This
4132 4138 * creates a new sync task which modifies the pool to reflect the feature
4133 4139 * as being active.
4134 4140 */
4135 4141 static int
4136 4142 zfs_prop_activate_feature(spa_t *spa, spa_feature_t feature)
4137 4143 {
4138 4144 int err;
4139 4145
4140 4146 /* EBUSY here indicates that the feature is already active */
4141 4147 err = dsl_sync_task(spa_name(spa),
4142 4148 zfs_prop_activate_feature_check, zfs_prop_activate_feature_sync,
4143 4149 &feature, 2, ZFS_SPACE_CHECK_RESERVED);
4144 4150
4145 4151 if (err != 0 && err != EBUSY)
4146 4152 return (err);
4147 4153 else
4148 4154 return (0);
4149 4155 }
4150 4156
4151 4157 /*
4152 4158 * Removes properties from the given props list that fail permission checks
4153 4159 * needed to clear them and to restore them in case of a receive error. For each
4154 4160 * property, make sure we have both set and inherit permissions.
4155 4161 *
4156 4162 * Returns the first error encountered if any permission checks fail. If the
4157 4163 * caller provides a non-NULL errlist, it also gives the complete list of names
4158 4164 * of all the properties that failed a permission check along with the
4159 4165 * corresponding error numbers. The caller is responsible for freeing the
4160 4166 * returned errlist.
4161 4167 *
4162 4168 * If every property checks out successfully, zero is returned and the list
4163 4169 * pointed at by errlist is NULL.
4164 4170 */
4165 4171 static int
4166 4172 zfs_check_clearable(char *dataset, nvlist_t *props, nvlist_t **errlist)
4167 4173 {
4168 4174 zfs_cmd_t *zc;
4169 4175 nvpair_t *pair, *next_pair;
4170 4176 nvlist_t *errors;
4171 4177 int err, rv = 0;
4172 4178
4173 4179 if (props == NULL)
4174 4180 return (0);
4175 4181
4176 4182 VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0);
4177 4183
4178 4184 zc = kmem_alloc(sizeof (zfs_cmd_t), KM_SLEEP);
4179 4185 (void) strcpy(zc->zc_name, dataset);
4180 4186 pair = nvlist_next_nvpair(props, NULL);
4181 4187 while (pair != NULL) {
4182 4188 next_pair = nvlist_next_nvpair(props, pair);
4183 4189
4184 4190 (void) strcpy(zc->zc_value, nvpair_name(pair));
4185 4191 if ((err = zfs_check_settable(dataset, pair, CRED())) != 0 ||
4186 4192 (err = zfs_secpolicy_inherit_prop(zc, NULL, CRED())) != 0) {
4187 4193 VERIFY(nvlist_remove_nvpair(props, pair) == 0);
4188 4194 VERIFY(nvlist_add_int32(errors,
4189 4195 zc->zc_value, err) == 0);
4190 4196 }
4191 4197 pair = next_pair;
4192 4198 }
4193 4199 kmem_free(zc, sizeof (zfs_cmd_t));
4194 4200
4195 4201 if ((pair = nvlist_next_nvpair(errors, NULL)) == NULL) {
4196 4202 nvlist_free(errors);
4197 4203 errors = NULL;
4198 4204 } else {
4199 4205 VERIFY(nvpair_value_int32(pair, &rv) == 0);
4200 4206 }
4201 4207
4202 4208 if (errlist == NULL)
4203 4209 nvlist_free(errors);
4204 4210 else
4205 4211 *errlist = errors;
4206 4212
4207 4213 return (rv);
4208 4214 }
4209 4215
4210 4216 static boolean_t
4211 4217 propval_equals(nvpair_t *p1, nvpair_t *p2)
4212 4218 {
4213 4219 if (nvpair_type(p1) == DATA_TYPE_NVLIST) {
4214 4220 /* dsl_prop_get_all_impl() format */
4215 4221 nvlist_t *attrs;
4216 4222 VERIFY(nvpair_value_nvlist(p1, &attrs) == 0);
4217 4223 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
4218 4224 &p1) == 0);
4219 4225 }
4220 4226
4221 4227 if (nvpair_type(p2) == DATA_TYPE_NVLIST) {
4222 4228 nvlist_t *attrs;
4223 4229 VERIFY(nvpair_value_nvlist(p2, &attrs) == 0);
4224 4230 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
4225 4231 &p2) == 0);
4226 4232 }
4227 4233
4228 4234 if (nvpair_type(p1) != nvpair_type(p2))
4229 4235 return (B_FALSE);
4230 4236
4231 4237 if (nvpair_type(p1) == DATA_TYPE_STRING) {
4232 4238 char *valstr1, *valstr2;
4233 4239
4234 4240 VERIFY(nvpair_value_string(p1, (char **)&valstr1) == 0);
4235 4241 VERIFY(nvpair_value_string(p2, (char **)&valstr2) == 0);
4236 4242 return (strcmp(valstr1, valstr2) == 0);
4237 4243 } else {
4238 4244 uint64_t intval1, intval2;
4239 4245
4240 4246 VERIFY(nvpair_value_uint64(p1, &intval1) == 0);
4241 4247 VERIFY(nvpair_value_uint64(p2, &intval2) == 0);
4242 4248 return (intval1 == intval2);
4243 4249 }
4244 4250 }
4245 4251
4246 4252 /*
4247 4253 * Remove properties from props if they are not going to change (as determined
4248 4254 * by comparison with origprops). Remove them from origprops as well, since we
4249 4255 * do not need to clear or restore properties that won't change.
4250 4256 */
4251 4257 static void
4252 4258 props_reduce(nvlist_t *props, nvlist_t *origprops)
4253 4259 {
4254 4260 nvpair_t *pair, *next_pair;
4255 4261
4256 4262 if (origprops == NULL)
4257 4263 return; /* all props need to be received */
4258 4264
4259 4265 pair = nvlist_next_nvpair(props, NULL);
4260 4266 while (pair != NULL) {
4261 4267 const char *propname = nvpair_name(pair);
4262 4268 nvpair_t *match;
4263 4269
4264 4270 next_pair = nvlist_next_nvpair(props, pair);
4265 4271
4266 4272 if ((nvlist_lookup_nvpair(origprops, propname,
4267 4273 &match) != 0) || !propval_equals(pair, match))
4268 4274 goto next; /* need to set received value */
4269 4275
4270 4276 /* don't clear the existing received value */
4271 4277 (void) nvlist_remove_nvpair(origprops, match);
4272 4278 /* don't bother receiving the property */
4273 4279 (void) nvlist_remove_nvpair(props, pair);
4274 4280 next:
4275 4281 pair = next_pair;
4276 4282 }
4277 4283 }
4278 4284
4279 4285 /*
4280 4286 * Extract properties that cannot be set PRIOR to the receipt of a dataset.
4281 4287 * For example, refquota cannot be set until after the receipt of a dataset,
4282 4288 * because in replication streams, an older/earlier snapshot may exceed the
4283 4289 * refquota. We want to receive the older/earlier snapshot, but setting
4284 4290 * refquota pre-receipt will set the dsl's ACTUAL quota, which will prevent
4285 4291 * the older/earlier snapshot from being received (with EDQUOT).
4286 4292 *
4287 4293 * The ZFS test "zfs_receive_011_pos" demonstrates such a scenario.
4288 4294 *
4289 4295 * libzfs will need to be judicious handling errors encountered by props
4290 4296 * extracted by this function.
4291 4297 */
4292 4298 static nvlist_t *
4293 4299 extract_delay_props(nvlist_t *props)
4294 4300 {
4295 4301 nvlist_t *delayprops;
4296 4302 nvpair_t *nvp, *tmp;
4297 4303 static const zfs_prop_t delayable[] = { ZFS_PROP_REFQUOTA, 0 };
4298 4304 int i;
4299 4305
4300 4306 VERIFY(nvlist_alloc(&delayprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
4301 4307
4302 4308 for (nvp = nvlist_next_nvpair(props, NULL); nvp != NULL;
4303 4309 nvp = nvlist_next_nvpair(props, nvp)) {
4304 4310 /*
4305 4311 * strcmp() is safe because zfs_prop_to_name() always returns
4306 4312 * a bounded string.
4307 4313 */
4308 4314 for (i = 0; delayable[i] != 0; i++) {
4309 4315 if (strcmp(zfs_prop_to_name(delayable[i]),
4310 4316 nvpair_name(nvp)) == 0) {
4311 4317 break;
4312 4318 }
4313 4319 }
4314 4320 if (delayable[i] != 0) {
4315 4321 tmp = nvlist_prev_nvpair(props, nvp);
4316 4322 VERIFY(nvlist_add_nvpair(delayprops, nvp) == 0);
4317 4323 VERIFY(nvlist_remove_nvpair(props, nvp) == 0);
4318 4324 nvp = tmp;
4319 4325 }
4320 4326 }
4321 4327
4322 4328 if (nvlist_empty(delayprops)) {
4323 4329 nvlist_free(delayprops);
4324 4330 delayprops = NULL;
4325 4331 }
4326 4332 return (delayprops);
4327 4333 }
4328 4334
4329 4335 #ifdef DEBUG
4330 4336 static boolean_t zfs_ioc_recv_inject_err;
4331 4337 #endif
4332 4338
4333 4339 /*
4334 4340 * inputs:
4335 4341 * zc_name name of containing filesystem
4336 4342 * zc_nvlist_src{_size} nvlist of properties to apply
4337 4343 * zc_value name of snapshot to create
4338 4344 * zc_string name of clone origin (if DRR_FLAG_CLONE)
4339 4345 * zc_cookie file descriptor to recv from
4340 4346 * zc_begin_record the BEGIN record of the stream (not byteswapped)
4341 4347 * zc_guid force flag
4342 4348 * zc_cleanup_fd cleanup-on-exit file descriptor
4343 4349 * zc_action_handle handle for this guid/ds mapping (or zero on first call)
4344 4350 * zc_resumable if data is incomplete assume sender will resume
4345 4351 *
4346 4352 * outputs:
4347 4353 * zc_cookie number of bytes read
4348 4354 * zc_nvlist_dst{_size} error for each unapplied received property
4349 4355 * zc_obj zprop_errflags_t
4350 4356 * zc_action_handle handle for this guid/ds mapping
4351 4357 */
4352 4358 static int
4353 4359 zfs_ioc_recv(zfs_cmd_t *zc)
4354 4360 {
4355 4361 file_t *fp;
4356 4362 dmu_recv_cookie_t drc;
4357 4363 boolean_t force = (boolean_t)zc->zc_guid;
4358 4364 int fd;
4359 4365 int error = 0;
4360 4366 int props_error = 0;
4361 4367 nvlist_t *errors;
4362 4368 offset_t off;
4363 4369 nvlist_t *props = NULL; /* sent properties */
4364 4370 nvlist_t *origprops = NULL; /* existing properties */
4365 4371 nvlist_t *delayprops = NULL; /* sent properties applied post-receive */
4366 4372 char *origin = NULL;
4367 4373 char *tosnap;
4368 4374 char tofs[ZFS_MAX_DATASET_NAME_LEN];
4369 4375 boolean_t first_recvd_props = B_FALSE;
4370 4376
4371 4377 if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
4372 4378 strchr(zc->zc_value, '@') == NULL ||
4373 4379 strchr(zc->zc_value, '%'))
4374 4380 return (SET_ERROR(EINVAL));
4375 4381
4376 4382 (void) strcpy(tofs, zc->zc_value);
4377 4383 tosnap = strchr(tofs, '@');
4378 4384 *tosnap++ = '\0';
4379 4385
4380 4386 if (zc->zc_nvlist_src != NULL &&
4381 4387 (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
4382 4388 zc->zc_iflags, &props)) != 0)
4383 4389 return (error);
4384 4390
4385 4391 fd = zc->zc_cookie;
4386 4392 fp = getf(fd);
4387 4393 if (fp == NULL) {
4388 4394 nvlist_free(props);
4389 4395 return (SET_ERROR(EBADF));
4390 4396 }
4391 4397
4392 4398 errors = fnvlist_alloc();
4393 4399
4394 4400 if (zc->zc_string[0])
4395 4401 origin = zc->zc_string;
4396 4402
4397 4403 error = dmu_recv_begin(tofs, tosnap,
4398 4404 &zc->zc_begin_record, force, zc->zc_resumable, origin, &drc);
4399 4405 if (error != 0)
4400 4406 goto out;
4401 4407
4402 4408 /*
4403 4409 * Set properties before we receive the stream so that they are applied
4404 4410 * to the new data. Note that we must call dmu_recv_stream() if
4405 4411 * dmu_recv_begin() succeeds.
4406 4412 */
4407 4413 if (props != NULL && !drc.drc_newfs) {
4408 4414 if (spa_version(dsl_dataset_get_spa(drc.drc_ds)) >=
4409 4415 SPA_VERSION_RECVD_PROPS &&
4410 4416 !dsl_prop_get_hasrecvd(tofs))
4411 4417 first_recvd_props = B_TRUE;
4412 4418
4413 4419 /*
4414 4420 * If new received properties are supplied, they are to
4415 4421 * completely replace the existing received properties, so stash
4416 4422 * away the existing ones.
4417 4423 */
4418 4424 if (dsl_prop_get_received(tofs, &origprops) == 0) {
4419 4425 nvlist_t *errlist = NULL;
4420 4426 /*
4421 4427 * Don't bother writing a property if its value won't
4422 4428 * change (and avoid the unnecessary security checks).
4423 4429 *
4424 4430 * The first receive after SPA_VERSION_RECVD_PROPS is a
4425 4431 * special case where we blow away all local properties
4426 4432 * regardless.
4427 4433 */
4428 4434 if (!first_recvd_props)
4429 4435 props_reduce(props, origprops);
4430 4436 if (zfs_check_clearable(tofs, origprops, &errlist) != 0)
4431 4437 (void) nvlist_merge(errors, errlist, 0);
4432 4438 nvlist_free(errlist);
4433 4439
4434 4440 if (clear_received_props(tofs, origprops,
4435 4441 first_recvd_props ? NULL : props) != 0)
4436 4442 zc->zc_obj |= ZPROP_ERR_NOCLEAR;
4437 4443 } else {
4438 4444 zc->zc_obj |= ZPROP_ERR_NOCLEAR;
4439 4445 }
4440 4446 }
4441 4447
4442 4448 if (props != NULL) {
4443 4449 props_error = dsl_prop_set_hasrecvd(tofs);
4444 4450
4445 4451 if (props_error == 0) {
4446 4452 delayprops = extract_delay_props(props);
4447 4453 (void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_RECEIVED,
4448 4454 props, errors);
4449 4455 }
4450 4456 }
4451 4457
4452 4458 off = fp->f_offset;
4453 4459 error = dmu_recv_stream(&drc, fp->f_vnode, &off, zc->zc_cleanup_fd,
4454 4460 &zc->zc_action_handle);
4455 4461
4456 4462 if (error == 0) {
4457 4463 zfsvfs_t *zfsvfs = NULL;
4458 4464
4459 4465 if (getzfsvfs(tofs, &zfsvfs) == 0) {
4460 4466 /* online recv */
4461 4467 dsl_dataset_t *ds;
4462 4468 int end_err;
4463 4469
4464 4470 ds = dmu_objset_ds(zfsvfs->z_os);
4465 4471 error = zfs_suspend_fs(zfsvfs);
4466 4472 /*
4467 4473 * If the suspend fails, then the recv_end will
4468 4474 * likely also fail, and clean up after itself.
4469 4475 */
4470 4476 end_err = dmu_recv_end(&drc, zfsvfs);
4471 4477 if (error == 0)
4472 4478 error = zfs_resume_fs(zfsvfs, ds);
4473 4479 error = error ? error : end_err;
4474 4480 VFS_RELE(zfsvfs->z_vfs);
4475 4481 } else {
4476 4482 error = dmu_recv_end(&drc, NULL);
4477 4483 }
4478 4484
4479 4485 /* Set delayed properties now, after we're done receiving. */
4480 4486 if (delayprops != NULL && error == 0) {
4481 4487 (void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_RECEIVED,
4482 4488 delayprops, errors);
4483 4489 }
4484 4490 }
4485 4491
4486 4492 if (delayprops != NULL) {
4487 4493 /*
4488 4494 * Merge delayed props back in with initial props, in case
4489 4495 * we're DEBUG and zfs_ioc_recv_inject_err is set (which means
4490 4496 * we have to make sure clear_received_props() includes
4491 4497 * the delayed properties).
4492 4498 *
4493 4499 * Since zfs_ioc_recv_inject_err is only in DEBUG kernels,
4494 4500 * using ASSERT() will be just like a VERIFY.
4495 4501 */
4496 4502 ASSERT(nvlist_merge(props, delayprops, 0) == 0);
4497 4503 nvlist_free(delayprops);
4498 4504 }
4499 4505
4500 4506 /*
4501 4507 * Now that all props, initial and delayed, are set, report the prop
4502 4508 * errors to the caller.
4503 4509 */
4504 4510 if (zc->zc_nvlist_dst_size != 0 &&
4505 4511 (nvlist_smush(errors, zc->zc_nvlist_dst_size) != 0 ||
4506 4512 put_nvlist(zc, errors) != 0)) {
4507 4513 /*
4508 4514 * Caller made zc->zc_nvlist_dst less than the minimum expected
4509 4515 * size or supplied an invalid address.
4510 4516 */
4511 4517 props_error = SET_ERROR(EINVAL);
4512 4518 }
4513 4519
4514 4520 zc->zc_cookie = off - fp->f_offset;
4515 4521 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
4516 4522 fp->f_offset = off;
4517 4523
4518 4524 #ifdef DEBUG
4519 4525 if (zfs_ioc_recv_inject_err) {
4520 4526 zfs_ioc_recv_inject_err = B_FALSE;
4521 4527 error = 1;
4522 4528 }
4523 4529 #endif
4524 4530 /*
4525 4531 * On error, restore the original props.
4526 4532 */
4527 4533 if (error != 0 && props != NULL && !drc.drc_newfs) {
4528 4534 if (clear_received_props(tofs, props, NULL) != 0) {
4529 4535 /*
4530 4536 * We failed to clear the received properties.
4531 4537 * Since we may have left a $recvd value on the
4532 4538 * system, we can't clear the $hasrecvd flag.
4533 4539 */
4534 4540 zc->zc_obj |= ZPROP_ERR_NORESTORE;
4535 4541 } else if (first_recvd_props) {
4536 4542 dsl_prop_unset_hasrecvd(tofs);
4537 4543 }
4538 4544
4539 4545 if (origprops == NULL && !drc.drc_newfs) {
4540 4546 /* We failed to stash the original properties. */
4541 4547 zc->zc_obj |= ZPROP_ERR_NORESTORE;
4542 4548 }
4543 4549
4544 4550 /*
4545 4551 * dsl_props_set() will not convert RECEIVED to LOCAL on or
4546 4552 * after SPA_VERSION_RECVD_PROPS, so we need to specify LOCAL
4547 4553 * explictly if we're restoring local properties cleared in the
4548 4554 * first new-style receive.
4549 4555 */
4550 4556 if (origprops != NULL &&
4551 4557 zfs_set_prop_nvlist(tofs, (first_recvd_props ?
4552 4558 ZPROP_SRC_LOCAL : ZPROP_SRC_RECEIVED),
4553 4559 origprops, NULL) != 0) {
4554 4560 /*
4555 4561 * We stashed the original properties but failed to
4556 4562 * restore them.
4557 4563 */
4558 4564 zc->zc_obj |= ZPROP_ERR_NORESTORE;
4559 4565 }
4560 4566 }
4561 4567 out:
4562 4568 nvlist_free(props);
4563 4569 nvlist_free(origprops);
4564 4570 nvlist_free(errors);
4565 4571 releasef(fd);
4566 4572
4567 4573 if (error == 0)
4568 4574 error = props_error;
4569 4575
4570 4576 return (error);
4571 4577 }
4572 4578
4573 4579 /*
4574 4580 * inputs:
4575 4581 * zc_name name of snapshot to send
4576 4582 * zc_cookie file descriptor to send stream to
4577 4583 * zc_obj fromorigin flag (mutually exclusive with zc_fromobj)
4578 4584 * zc_sendobj objsetid of snapshot to send
4579 4585 * zc_fromobj objsetid of incremental fromsnap (may be zero)
4580 4586 * zc_guid if set, estimate size of stream only. zc_cookie is ignored.
4581 4587 * output size in zc_objset_type.
4582 4588 * zc_flags lzc_send_flags
4583 4589 *
4584 4590 * outputs:
4585 4591 * zc_objset_type estimated size, if zc_guid is set
4586 4592 */
4587 4593 static int
4588 4594 zfs_ioc_send(zfs_cmd_t *zc)
4589 4595 {
4590 4596 int error;
4591 4597 offset_t off;
4592 4598 boolean_t estimate = (zc->zc_guid != 0);
4593 4599 boolean_t embedok = (zc->zc_flags & 0x1);
4594 4600 boolean_t large_block_ok = (zc->zc_flags & 0x2);
4595 4601 boolean_t compressok = (zc->zc_flags & 0x4);
4596 4602
4597 4603 if (zc->zc_obj != 0) {
4598 4604 dsl_pool_t *dp;
4599 4605 dsl_dataset_t *tosnap;
4600 4606
4601 4607 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
4602 4608 if (error != 0)
4603 4609 return (error);
4604 4610
4605 4611 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &tosnap);
4606 4612 if (error != 0) {
4607 4613 dsl_pool_rele(dp, FTAG);
4608 4614 return (error);
4609 4615 }
4610 4616
4611 4617 if (dsl_dir_is_clone(tosnap->ds_dir))
4612 4618 zc->zc_fromobj =
4613 4619 dsl_dir_phys(tosnap->ds_dir)->dd_origin_obj;
4614 4620 dsl_dataset_rele(tosnap, FTAG);
4615 4621 dsl_pool_rele(dp, FTAG);
4616 4622 }
4617 4623
4618 4624 if (estimate) {
4619 4625 dsl_pool_t *dp;
4620 4626 dsl_dataset_t *tosnap;
4621 4627 dsl_dataset_t *fromsnap = NULL;
4622 4628
4623 4629 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
4624 4630 if (error != 0)
4625 4631 return (error);
4626 4632
4627 4633 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &tosnap);
4628 4634 if (error != 0) {
4629 4635 dsl_pool_rele(dp, FTAG);
4630 4636 return (error);
4631 4637 }
4632 4638
4633 4639 if (zc->zc_fromobj != 0) {
4634 4640 error = dsl_dataset_hold_obj(dp, zc->zc_fromobj,
4635 4641 FTAG, &fromsnap);
4636 4642 if (error != 0) {
4637 4643 dsl_dataset_rele(tosnap, FTAG);
4638 4644 dsl_pool_rele(dp, FTAG);
4639 4645 return (error);
4640 4646 }
4641 4647 }
4642 4648
4643 4649 error = dmu_send_estimate(tosnap, fromsnap, compressok,
4644 4650 &zc->zc_objset_type);
4645 4651
4646 4652 if (fromsnap != NULL)
4647 4653 dsl_dataset_rele(fromsnap, FTAG);
4648 4654 dsl_dataset_rele(tosnap, FTAG);
4649 4655 dsl_pool_rele(dp, FTAG);
4650 4656 } else {
4651 4657 file_t *fp = getf(zc->zc_cookie);
4652 4658 if (fp == NULL)
4653 4659 return (SET_ERROR(EBADF));
4654 4660
4655 4661 off = fp->f_offset;
4656 4662 error = dmu_send_obj(zc->zc_name, zc->zc_sendobj,
4657 4663 zc->zc_fromobj, embedok, large_block_ok, compressok,
4658 4664 zc->zc_cookie, fp->f_vnode, &off);
4659 4665
4660 4666 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
4661 4667 fp->f_offset = off;
4662 4668 releasef(zc->zc_cookie);
4663 4669 }
4664 4670 return (error);
4665 4671 }
4666 4672
4667 4673 /*
4668 4674 * inputs:
4669 4675 * zc_name name of snapshot on which to report progress
4670 4676 * zc_cookie file descriptor of send stream
4671 4677 *
4672 4678 * outputs:
4673 4679 * zc_cookie number of bytes written in send stream thus far
4674 4680 */
4675 4681 static int
4676 4682 zfs_ioc_send_progress(zfs_cmd_t *zc)
4677 4683 {
4678 4684 dsl_pool_t *dp;
4679 4685 dsl_dataset_t *ds;
4680 4686 dmu_sendarg_t *dsp = NULL;
4681 4687 int error;
4682 4688
4683 4689 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
4684 4690 if (error != 0)
4685 4691 return (error);
4686 4692
4687 4693 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &ds);
4688 4694 if (error != 0) {
4689 4695 dsl_pool_rele(dp, FTAG);
4690 4696 return (error);
4691 4697 }
4692 4698
4693 4699 mutex_enter(&ds->ds_sendstream_lock);
4694 4700
4695 4701 /*
4696 4702 * Iterate over all the send streams currently active on this dataset.
4697 4703 * If there's one which matches the specified file descriptor _and_ the
4698 4704 * stream was started by the current process, return the progress of
4699 4705 * that stream.
4700 4706 */
4701 4707 for (dsp = list_head(&ds->ds_sendstreams); dsp != NULL;
4702 4708 dsp = list_next(&ds->ds_sendstreams, dsp)) {
4703 4709 if (dsp->dsa_outfd == zc->zc_cookie &&
4704 4710 dsp->dsa_proc == curproc)
4705 4711 break;
4706 4712 }
4707 4713
4708 4714 if (dsp != NULL)
4709 4715 zc->zc_cookie = *(dsp->dsa_off);
4710 4716 else
4711 4717 error = SET_ERROR(ENOENT);
4712 4718
4713 4719 mutex_exit(&ds->ds_sendstream_lock);
4714 4720 dsl_dataset_rele(ds, FTAG);
4715 4721 dsl_pool_rele(dp, FTAG);
4716 4722 return (error);
4717 4723 }
4718 4724
4719 4725 static int
4720 4726 zfs_ioc_inject_fault(zfs_cmd_t *zc)
4721 4727 {
4722 4728 int id, error;
4723 4729
4724 4730 error = zio_inject_fault(zc->zc_name, (int)zc->zc_guid, &id,
4725 4731 &zc->zc_inject_record);
4726 4732
4727 4733 if (error == 0)
4728 4734 zc->zc_guid = (uint64_t)id;
4729 4735
4730 4736 return (error);
4731 4737 }
4732 4738
4733 4739 static int
4734 4740 zfs_ioc_clear_fault(zfs_cmd_t *zc)
4735 4741 {
4736 4742 return (zio_clear_fault((int)zc->zc_guid));
4737 4743 }
4738 4744
4739 4745 static int
4740 4746 zfs_ioc_inject_list_next(zfs_cmd_t *zc)
4741 4747 {
4742 4748 int id = (int)zc->zc_guid;
4743 4749 int error;
4744 4750
4745 4751 error = zio_inject_list_next(&id, zc->zc_name, sizeof (zc->zc_name),
4746 4752 &zc->zc_inject_record);
4747 4753
4748 4754 zc->zc_guid = id;
4749 4755
4750 4756 return (error);
4751 4757 }
4752 4758
4753 4759 static int
4754 4760 zfs_ioc_error_log(zfs_cmd_t *zc)
4755 4761 {
4756 4762 spa_t *spa;
4757 4763 int error;
4758 4764 size_t count = (size_t)zc->zc_nvlist_dst_size;
4759 4765
4760 4766 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
4761 4767 return (error);
4762 4768
4763 4769 error = spa_get_errlog(spa, (void *)(uintptr_t)zc->zc_nvlist_dst,
4764 4770 &count);
4765 4771 if (error == 0)
4766 4772 zc->zc_nvlist_dst_size = count;
4767 4773 else
4768 4774 zc->zc_nvlist_dst_size = spa_get_errlog_size(spa);
4769 4775
4770 4776 spa_close(spa, FTAG);
4771 4777
4772 4778 return (error);
4773 4779 }
4774 4780
4775 4781 static int
4776 4782 zfs_ioc_clear(zfs_cmd_t *zc)
4777 4783 {
4778 4784 spa_t *spa;
4779 4785 vdev_t *vd;
4780 4786 int error;
4781 4787
4782 4788 /*
4783 4789 * On zpool clear we also fix up missing slogs
4784 4790 */
4785 4791 mutex_enter(&spa_namespace_lock);
4786 4792 spa = spa_lookup(zc->zc_name);
4787 4793 if (spa == NULL) {
4788 4794 mutex_exit(&spa_namespace_lock);
4789 4795 return (SET_ERROR(EIO));
4790 4796 }
4791 4797 if (spa_get_log_state(spa) == SPA_LOG_MISSING) {
4792 4798 /* we need to let spa_open/spa_load clear the chains */
4793 4799 spa_set_log_state(spa, SPA_LOG_CLEAR);
4794 4800 }
4795 4801 spa->spa_last_open_failed = 0;
4796 4802 mutex_exit(&spa_namespace_lock);
4797 4803
4798 4804 if (zc->zc_cookie & ZPOOL_NO_REWIND) {
4799 4805 error = spa_open(zc->zc_name, &spa, FTAG);
4800 4806 } else {
4801 4807 nvlist_t *policy;
4802 4808 nvlist_t *config = NULL;
4803 4809
4804 4810 if (zc->zc_nvlist_src == NULL)
4805 4811 return (SET_ERROR(EINVAL));
4806 4812
4807 4813 if ((error = get_nvlist(zc->zc_nvlist_src,
4808 4814 zc->zc_nvlist_src_size, zc->zc_iflags, &policy)) == 0) {
4809 4815 error = spa_open_rewind(zc->zc_name, &spa, FTAG,
4810 4816 policy, &config);
4811 4817 if (config != NULL) {
4812 4818 int err;
4813 4819
4814 4820 if ((err = put_nvlist(zc, config)) != 0)
4815 4821 error = err;
4816 4822 nvlist_free(config);
4817 4823 }
4818 4824 nvlist_free(policy);
4819 4825 }
4820 4826 }
4821 4827
4822 4828 if (error != 0)
4823 4829 return (error);
4824 4830
4825 4831 spa_vdev_state_enter(spa, SCL_NONE);
4826 4832
4827 4833 if (zc->zc_guid == 0) {
4828 4834 vd = NULL;
4829 4835 } else {
4830 4836 vd = spa_lookup_by_guid(spa, zc->zc_guid, B_TRUE);
4831 4837 if (vd == NULL) {
4832 4838 (void) spa_vdev_state_exit(spa, NULL, ENODEV);
4833 4839 spa_close(spa, FTAG);
4834 4840 return (SET_ERROR(ENODEV));
4835 4841 }
4836 4842 }
4837 4843
4838 4844 vdev_clear(spa, vd);
4839 4845
4840 4846 (void) spa_vdev_state_exit(spa, NULL, 0);
4841 4847
4842 4848 /*
4843 4849 * Resume any suspended I/Os.
4844 4850 */
4845 4851 if (zio_resume(spa) != 0)
4846 4852 error = SET_ERROR(EIO);
4847 4853
4848 4854 spa_close(spa, FTAG);
4849 4855
4850 4856 return (error);
4851 4857 }
4852 4858
4853 4859 static int
4854 4860 zfs_ioc_pool_reopen(zfs_cmd_t *zc)
4855 4861 {
4856 4862 spa_t *spa;
4857 4863 int error;
4858 4864
4859 4865 error = spa_open(zc->zc_name, &spa, FTAG);
4860 4866 if (error != 0)
4861 4867 return (error);
4862 4868
4863 4869 spa_vdev_state_enter(spa, SCL_NONE);
4864 4870
4865 4871 /*
4866 4872 * If a resilver is already in progress then set the
4867 4873 * spa_scrub_reopen flag to B_TRUE so that we don't restart
4868 4874 * the scan as a side effect of the reopen. Otherwise, let
4869 4875 * vdev_open() decided if a resilver is required.
4870 4876 */
4871 4877 spa->spa_scrub_reopen = dsl_scan_resilvering(spa->spa_dsl_pool);
4872 4878 vdev_reopen(spa->spa_root_vdev);
4873 4879 spa->spa_scrub_reopen = B_FALSE;
4874 4880
4875 4881 (void) spa_vdev_state_exit(spa, NULL, 0);
4876 4882 spa_close(spa, FTAG);
4877 4883 return (0);
4878 4884 }
4879 4885 /*
4880 4886 * inputs:
4881 4887 * zc_name name of filesystem
4882 4888 *
4883 4889 * outputs:
4884 4890 * zc_string name of conflicting snapshot, if there is one
4885 4891 */
4886 4892 static int
4887 4893 zfs_ioc_promote(zfs_cmd_t *zc)
4888 4894 {
4889 4895 dsl_pool_t *dp;
4890 4896 dsl_dataset_t *ds, *ods;
4891 4897 char origin[ZFS_MAX_DATASET_NAME_LEN];
4892 4898 char *cp;
4893 4899 int error;
4894 4900
4895 4901 zc->zc_name[sizeof (zc->zc_name) - 1] = '\0';
4896 4902 if (dataset_namecheck(zc->zc_name, NULL, NULL) != 0 ||
4897 4903 strchr(zc->zc_name, '%'))
4898 4904 return (SET_ERROR(EINVAL));
4899 4905
4900 4906 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
4901 4907 if (error != 0)
4902 4908 return (error);
4903 4909
4904 4910 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &ds);
4905 4911 if (error != 0) {
4906 4912 dsl_pool_rele(dp, FTAG);
4907 4913 return (error);
4908 4914 }
4909 4915
4910 4916 if (!dsl_dir_is_clone(ds->ds_dir)) {
4911 4917 dsl_dataset_rele(ds, FTAG);
4912 4918 dsl_pool_rele(dp, FTAG);
4913 4919 return (SET_ERROR(EINVAL));
4914 4920 }
4915 4921
4916 4922 error = dsl_dataset_hold_obj(dp,
4917 4923 dsl_dir_phys(ds->ds_dir)->dd_origin_obj, FTAG, &ods);
4918 4924 if (error != 0) {
4919 4925 dsl_dataset_rele(ds, FTAG);
4920 4926 dsl_pool_rele(dp, FTAG);
4921 4927 return (error);
4922 4928 }
4923 4929
4924 4930 dsl_dataset_name(ods, origin);
4925 4931 dsl_dataset_rele(ods, FTAG);
4926 4932 dsl_dataset_rele(ds, FTAG);
4927 4933 dsl_pool_rele(dp, FTAG);
4928 4934
4929 4935 /*
4930 4936 * We don't need to unmount *all* the origin fs's snapshots, but
4931 4937 * it's easier.
4932 4938 */
4933 4939 cp = strchr(origin, '@');
4934 4940 if (cp)
4935 4941 *cp = '\0';
4936 4942 (void) dmu_objset_find(origin,
4937 4943 zfs_unmount_snap_cb, NULL, DS_FIND_SNAPSHOTS);
4938 4944 return (dsl_dataset_promote(zc->zc_name, zc->zc_string));
4939 4945 }
4940 4946
4941 4947 /*
4942 4948 * Retrieve a single {user|group}{used|quota}@... property.
4943 4949 *
4944 4950 * inputs:
4945 4951 * zc_name name of filesystem
4946 4952 * zc_objset_type zfs_userquota_prop_t
4947 4953 * zc_value domain name (eg. "S-1-234-567-89")
4948 4954 * zc_guid RID/UID/GID
4949 4955 *
4950 4956 * outputs:
4951 4957 * zc_cookie property value
4952 4958 */
4953 4959 static int
4954 4960 zfs_ioc_userspace_one(zfs_cmd_t *zc)
4955 4961 {
4956 4962 zfsvfs_t *zfsvfs;
4957 4963 int error;
4958 4964
4959 4965 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
4960 4966 return (SET_ERROR(EINVAL));
4961 4967
4962 4968 error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE);
4963 4969 if (error != 0)
4964 4970 return (error);
4965 4971
4966 4972 error = zfs_userspace_one(zfsvfs,
4967 4973 zc->zc_objset_type, zc->zc_value, zc->zc_guid, &zc->zc_cookie);
4968 4974 zfsvfs_rele(zfsvfs, FTAG);
4969 4975
4970 4976 return (error);
4971 4977 }
4972 4978
4973 4979 /*
4974 4980 * inputs:
4975 4981 * zc_name name of filesystem
4976 4982 * zc_cookie zap cursor
4977 4983 * zc_objset_type zfs_userquota_prop_t
4978 4984 * zc_nvlist_dst[_size] buffer to fill (not really an nvlist)
4979 4985 *
4980 4986 * outputs:
4981 4987 * zc_nvlist_dst[_size] data buffer (array of zfs_useracct_t)
4982 4988 * zc_cookie zap cursor
4983 4989 */
4984 4990 static int
4985 4991 zfs_ioc_userspace_many(zfs_cmd_t *zc)
4986 4992 {
4987 4993 zfsvfs_t *zfsvfs;
4988 4994 int bufsize = zc->zc_nvlist_dst_size;
4989 4995
4990 4996 if (bufsize <= 0)
4991 4997 return (SET_ERROR(ENOMEM));
4992 4998
4993 4999 int error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE);
4994 5000 if (error != 0)
4995 5001 return (error);
4996 5002
4997 5003 void *buf = kmem_alloc(bufsize, KM_SLEEP);
4998 5004
4999 5005 error = zfs_userspace_many(zfsvfs, zc->zc_objset_type, &zc->zc_cookie,
5000 5006 buf, &zc->zc_nvlist_dst_size);
5001 5007
5002 5008 if (error == 0) {
5003 5009 error = xcopyout(buf,
5004 5010 (void *)(uintptr_t)zc->zc_nvlist_dst,
5005 5011 zc->zc_nvlist_dst_size);
5006 5012 }
5007 5013 kmem_free(buf, bufsize);
5008 5014 zfsvfs_rele(zfsvfs, FTAG);
5009 5015
5010 5016 return (error);
5011 5017 }
5012 5018
5013 5019 /*
5014 5020 * inputs:
5015 5021 * zc_name name of filesystem
5016 5022 *
5017 5023 * outputs:
5018 5024 * none
5019 5025 */
5020 5026 static int
5021 5027 zfs_ioc_userspace_upgrade(zfs_cmd_t *zc)
5022 5028 {
5023 5029 objset_t *os;
5024 5030 int error = 0;
5025 5031 zfsvfs_t *zfsvfs;
5026 5032
5027 5033 if (getzfsvfs(zc->zc_name, &zfsvfs) == 0) {
5028 5034 if (!dmu_objset_userused_enabled(zfsvfs->z_os)) {
5029 5035 /*
5030 5036 * If userused is not enabled, it may be because the
5031 5037 * objset needs to be closed & reopened (to grow the
5032 5038 * objset_phys_t). Suspend/resume the fs will do that.
5033 5039 */
5034 5040 dsl_dataset_t *ds, *newds;
5035 5041
5036 5042 ds = dmu_objset_ds(zfsvfs->z_os);
5037 5043 error = zfs_suspend_fs(zfsvfs);
5038 5044 if (error == 0) {
5039 5045 dmu_objset_refresh_ownership(ds, &newds,
5040 5046 zfsvfs);
5041 5047 error = zfs_resume_fs(zfsvfs, newds);
5042 5048 }
5043 5049 }
5044 5050 if (error == 0)
5045 5051 error = dmu_objset_userspace_upgrade(zfsvfs->z_os);
5046 5052 VFS_RELE(zfsvfs->z_vfs);
5047 5053 } else {
5048 5054 /* XXX kind of reading contents without owning */
5049 5055 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
5050 5056 if (error != 0)
5051 5057 return (error);
5052 5058
5053 5059 error = dmu_objset_userspace_upgrade(os);
5054 5060 dmu_objset_rele(os, FTAG);
5055 5061 }
5056 5062
5057 5063 return (error);
5058 5064 }
5059 5065
5060 5066 /*
5061 5067 * We don't want to have a hard dependency
5062 5068 * against some special symbols in sharefs
5063 5069 * nfs, and smbsrv. Determine them if needed when
5064 5070 * the first file system is shared.
5065 5071 * Neither sharefs, nfs or smbsrv are unloadable modules.
5066 5072 */
5067 5073 int (*znfsexport_fs)(void *arg);
5068 5074 int (*zshare_fs)(enum sharefs_sys_op, share_t *, uint32_t);
5069 5075 int (*zsmbexport_fs)(void *arg, boolean_t add_share);
5070 5076
5071 5077 int zfs_nfsshare_inited;
5072 5078 int zfs_smbshare_inited;
5073 5079
5074 5080 ddi_modhandle_t nfs_mod;
5075 5081 ddi_modhandle_t sharefs_mod;
5076 5082 ddi_modhandle_t smbsrv_mod;
5077 5083 kmutex_t zfs_share_lock;
5078 5084
5079 5085 static int
5080 5086 zfs_init_sharefs()
5081 5087 {
5082 5088 int error;
5083 5089
5084 5090 ASSERT(MUTEX_HELD(&zfs_share_lock));
5085 5091 /* Both NFS and SMB shares also require sharetab support. */
5086 5092 if (sharefs_mod == NULL && ((sharefs_mod =
5087 5093 ddi_modopen("fs/sharefs",
5088 5094 KRTLD_MODE_FIRST, &error)) == NULL)) {
5089 5095 return (SET_ERROR(ENOSYS));
5090 5096 }
5091 5097 if (zshare_fs == NULL && ((zshare_fs =
5092 5098 (int (*)(enum sharefs_sys_op, share_t *, uint32_t))
5093 5099 ddi_modsym(sharefs_mod, "sharefs_impl", &error)) == NULL)) {
5094 5100 return (SET_ERROR(ENOSYS));
5095 5101 }
5096 5102 return (0);
5097 5103 }
5098 5104
5099 5105 static int
5100 5106 zfs_ioc_share(zfs_cmd_t *zc)
5101 5107 {
5102 5108 int error;
5103 5109 int opcode;
5104 5110
5105 5111 switch (zc->zc_share.z_sharetype) {
5106 5112 case ZFS_SHARE_NFS:
5107 5113 case ZFS_UNSHARE_NFS:
5108 5114 if (zfs_nfsshare_inited == 0) {
5109 5115 mutex_enter(&zfs_share_lock);
5110 5116 if (nfs_mod == NULL && ((nfs_mod = ddi_modopen("fs/nfs",
5111 5117 KRTLD_MODE_FIRST, &error)) == NULL)) {
5112 5118 mutex_exit(&zfs_share_lock);
5113 5119 return (SET_ERROR(ENOSYS));
5114 5120 }
5115 5121 if (znfsexport_fs == NULL &&
5116 5122 ((znfsexport_fs = (int (*)(void *))
5117 5123 ddi_modsym(nfs_mod,
5118 5124 "nfs_export", &error)) == NULL)) {
5119 5125 mutex_exit(&zfs_share_lock);
5120 5126 return (SET_ERROR(ENOSYS));
5121 5127 }
5122 5128 error = zfs_init_sharefs();
5123 5129 if (error != 0) {
5124 5130 mutex_exit(&zfs_share_lock);
5125 5131 return (SET_ERROR(ENOSYS));
5126 5132 }
5127 5133 zfs_nfsshare_inited = 1;
5128 5134 mutex_exit(&zfs_share_lock);
5129 5135 }
5130 5136 break;
5131 5137 case ZFS_SHARE_SMB:
5132 5138 case ZFS_UNSHARE_SMB:
5133 5139 if (zfs_smbshare_inited == 0) {
5134 5140 mutex_enter(&zfs_share_lock);
5135 5141 if (smbsrv_mod == NULL && ((smbsrv_mod =
5136 5142 ddi_modopen("drv/smbsrv",
5137 5143 KRTLD_MODE_FIRST, &error)) == NULL)) {
5138 5144 mutex_exit(&zfs_share_lock);
5139 5145 return (SET_ERROR(ENOSYS));
5140 5146 }
5141 5147 if (zsmbexport_fs == NULL && ((zsmbexport_fs =
5142 5148 (int (*)(void *, boolean_t))ddi_modsym(smbsrv_mod,
5143 5149 "smb_server_share", &error)) == NULL)) {
5144 5150 mutex_exit(&zfs_share_lock);
5145 5151 return (SET_ERROR(ENOSYS));
5146 5152 }
5147 5153 error = zfs_init_sharefs();
5148 5154 if (error != 0) {
5149 5155 mutex_exit(&zfs_share_lock);
5150 5156 return (SET_ERROR(ENOSYS));
5151 5157 }
5152 5158 zfs_smbshare_inited = 1;
5153 5159 mutex_exit(&zfs_share_lock);
5154 5160 }
5155 5161 break;
5156 5162 default:
5157 5163 return (SET_ERROR(EINVAL));
5158 5164 }
5159 5165
5160 5166 switch (zc->zc_share.z_sharetype) {
5161 5167 case ZFS_SHARE_NFS:
5162 5168 case ZFS_UNSHARE_NFS:
5163 5169 if (error =
5164 5170 znfsexport_fs((void *)
5165 5171 (uintptr_t)zc->zc_share.z_exportdata))
5166 5172 return (error);
5167 5173 break;
5168 5174 case ZFS_SHARE_SMB:
5169 5175 case ZFS_UNSHARE_SMB:
5170 5176 if (error = zsmbexport_fs((void *)
5171 5177 (uintptr_t)zc->zc_share.z_exportdata,
5172 5178 zc->zc_share.z_sharetype == ZFS_SHARE_SMB ?
5173 5179 B_TRUE: B_FALSE)) {
5174 5180 return (error);
5175 5181 }
5176 5182 break;
5177 5183 }
5178 5184
5179 5185 opcode = (zc->zc_share.z_sharetype == ZFS_SHARE_NFS ||
5180 5186 zc->zc_share.z_sharetype == ZFS_SHARE_SMB) ?
5181 5187 SHAREFS_ADD : SHAREFS_REMOVE;
5182 5188
5183 5189 /*
5184 5190 * Add or remove share from sharetab
5185 5191 */
5186 5192 error = zshare_fs(opcode,
5187 5193 (void *)(uintptr_t)zc->zc_share.z_sharedata,
5188 5194 zc->zc_share.z_sharemax);
5189 5195
5190 5196 return (error);
5191 5197
5192 5198 }
5193 5199
5194 5200 ace_t full_access[] = {
5195 5201 {(uid_t)-1, ACE_ALL_PERMS, ACE_EVERYONE, 0}
5196 5202 };
5197 5203
5198 5204 /*
5199 5205 * inputs:
5200 5206 * zc_name name of containing filesystem
5201 5207 * zc_obj object # beyond which we want next in-use object #
5202 5208 *
5203 5209 * outputs:
5204 5210 * zc_obj next in-use object #
5205 5211 */
5206 5212 static int
5207 5213 zfs_ioc_next_obj(zfs_cmd_t *zc)
5208 5214 {
5209 5215 objset_t *os = NULL;
5210 5216 int error;
5211 5217
5212 5218 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
5213 5219 if (error != 0)
5214 5220 return (error);
5215 5221
5216 5222 error = dmu_object_next(os, &zc->zc_obj, B_FALSE,
5217 5223 dsl_dataset_phys(os->os_dsl_dataset)->ds_prev_snap_txg);
5218 5224
5219 5225 dmu_objset_rele(os, FTAG);
5220 5226 return (error);
5221 5227 }
5222 5228
5223 5229 /*
5224 5230 * inputs:
5225 5231 * zc_name name of filesystem
5226 5232 * zc_value prefix name for snapshot
5227 5233 * zc_cleanup_fd cleanup-on-exit file descriptor for calling process
5228 5234 *
5229 5235 * outputs:
5230 5236 * zc_value short name of new snapshot
5231 5237 */
5232 5238 static int
5233 5239 zfs_ioc_tmp_snapshot(zfs_cmd_t *zc)
5234 5240 {
5235 5241 char *snap_name;
5236 5242 char *hold_name;
5237 5243 int error;
5238 5244 minor_t minor;
5239 5245
5240 5246 error = zfs_onexit_fd_hold(zc->zc_cleanup_fd, &minor);
5241 5247 if (error != 0)
5242 5248 return (error);
5243 5249
5244 5250 snap_name = kmem_asprintf("%s-%016llx", zc->zc_value,
5245 5251 (u_longlong_t)ddi_get_lbolt64());
5246 5252 hold_name = kmem_asprintf("%%%s", zc->zc_value);
5247 5253
5248 5254 error = dsl_dataset_snapshot_tmp(zc->zc_name, snap_name, minor,
5249 5255 hold_name);
5250 5256 if (error == 0)
5251 5257 (void) strcpy(zc->zc_value, snap_name);
5252 5258 strfree(snap_name);
5253 5259 strfree(hold_name);
5254 5260 zfs_onexit_fd_rele(zc->zc_cleanup_fd);
5255 5261 return (error);
5256 5262 }
5257 5263
5258 5264 /*
5259 5265 * inputs:
5260 5266 * zc_name name of "to" snapshot
5261 5267 * zc_value name of "from" snapshot
5262 5268 * zc_cookie file descriptor to write diff data on
5263 5269 *
5264 5270 * outputs:
5265 5271 * dmu_diff_record_t's to the file descriptor
5266 5272 */
5267 5273 static int
5268 5274 zfs_ioc_diff(zfs_cmd_t *zc)
5269 5275 {
5270 5276 file_t *fp;
5271 5277 offset_t off;
5272 5278 int error;
5273 5279
5274 5280 fp = getf(zc->zc_cookie);
5275 5281 if (fp == NULL)
5276 5282 return (SET_ERROR(EBADF));
5277 5283
5278 5284 off = fp->f_offset;
5279 5285
5280 5286 error = dmu_diff(zc->zc_name, zc->zc_value, fp->f_vnode, &off);
5281 5287
5282 5288 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
5283 5289 fp->f_offset = off;
5284 5290 releasef(zc->zc_cookie);
5285 5291
5286 5292 return (error);
5287 5293 }
5288 5294
5289 5295 /*
5290 5296 * Remove all ACL files in shares dir
5291 5297 */
5292 5298 static int
5293 5299 zfs_smb_acl_purge(znode_t *dzp)
5294 5300 {
5295 5301 zap_cursor_t zc;
5296 5302 zap_attribute_t zap;
5297 5303 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
5298 5304 int error;
5299 5305
5300 5306 for (zap_cursor_init(&zc, zfsvfs->z_os, dzp->z_id);
5301 5307 (error = zap_cursor_retrieve(&zc, &zap)) == 0;
5302 5308 zap_cursor_advance(&zc)) {
5303 5309 if ((error = VOP_REMOVE(ZTOV(dzp), zap.za_name, kcred,
5304 5310 NULL, 0)) != 0)
5305 5311 break;
5306 5312 }
5307 5313 zap_cursor_fini(&zc);
5308 5314 return (error);
5309 5315 }
5310 5316
5311 5317 static int
5312 5318 zfs_ioc_smb_acl(zfs_cmd_t *zc)
5313 5319 {
5314 5320 vnode_t *vp;
5315 5321 znode_t *dzp;
5316 5322 vnode_t *resourcevp = NULL;
5317 5323 znode_t *sharedir;
5318 5324 zfsvfs_t *zfsvfs;
5319 5325 nvlist_t *nvlist;
5320 5326 char *src, *target;
5321 5327 vattr_t vattr;
5322 5328 vsecattr_t vsec;
5323 5329 int error = 0;
5324 5330
5325 5331 if ((error = lookupname(zc->zc_value, UIO_SYSSPACE,
5326 5332 NO_FOLLOW, NULL, &vp)) != 0)
5327 5333 return (error);
5328 5334
5329 5335 /* Now make sure mntpnt and dataset are ZFS */
5330 5336
5331 5337 if (vp->v_vfsp->vfs_fstype != zfsfstype ||
5332 5338 (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource),
5333 5339 zc->zc_name) != 0)) {
5334 5340 VN_RELE(vp);
5335 5341 return (SET_ERROR(EINVAL));
5336 5342 }
5337 5343
5338 5344 dzp = VTOZ(vp);
5339 5345 zfsvfs = dzp->z_zfsvfs;
5340 5346 ZFS_ENTER(zfsvfs);
5341 5347
5342 5348 /*
5343 5349 * Create share dir if its missing.
5344 5350 */
5345 5351 mutex_enter(&zfsvfs->z_lock);
5346 5352 if (zfsvfs->z_shares_dir == 0) {
5347 5353 dmu_tx_t *tx;
5348 5354
5349 5355 tx = dmu_tx_create(zfsvfs->z_os);
5350 5356 dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, TRUE,
5351 5357 ZFS_SHARES_DIR);
5352 5358 dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL);
5353 5359 error = dmu_tx_assign(tx, TXG_WAIT);
5354 5360 if (error != 0) {
5355 5361 dmu_tx_abort(tx);
5356 5362 } else {
5357 5363 error = zfs_create_share_dir(zfsvfs, tx);
5358 5364 dmu_tx_commit(tx);
5359 5365 }
5360 5366 if (error != 0) {
5361 5367 mutex_exit(&zfsvfs->z_lock);
5362 5368 VN_RELE(vp);
5363 5369 ZFS_EXIT(zfsvfs);
5364 5370 return (error);
5365 5371 }
5366 5372 }
5367 5373 mutex_exit(&zfsvfs->z_lock);
5368 5374
5369 5375 ASSERT(zfsvfs->z_shares_dir);
5370 5376 if ((error = zfs_zget(zfsvfs, zfsvfs->z_shares_dir, &sharedir)) != 0) {
5371 5377 VN_RELE(vp);
5372 5378 ZFS_EXIT(zfsvfs);
5373 5379 return (error);
5374 5380 }
5375 5381
5376 5382 switch (zc->zc_cookie) {
5377 5383 case ZFS_SMB_ACL_ADD:
5378 5384 vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE;
5379 5385 vattr.va_type = VREG;
5380 5386 vattr.va_mode = S_IFREG|0777;
5381 5387 vattr.va_uid = 0;
5382 5388 vattr.va_gid = 0;
5383 5389
5384 5390 vsec.vsa_mask = VSA_ACE;
5385 5391 vsec.vsa_aclentp = &full_access;
5386 5392 vsec.vsa_aclentsz = sizeof (full_access);
5387 5393 vsec.vsa_aclcnt = 1;
5388 5394
5389 5395 error = VOP_CREATE(ZTOV(sharedir), zc->zc_string,
5390 5396 &vattr, EXCL, 0, &resourcevp, kcred, 0, NULL, &vsec);
5391 5397 if (resourcevp)
5392 5398 VN_RELE(resourcevp);
5393 5399 break;
5394 5400
5395 5401 case ZFS_SMB_ACL_REMOVE:
5396 5402 error = VOP_REMOVE(ZTOV(sharedir), zc->zc_string, kcred,
5397 5403 NULL, 0);
5398 5404 break;
5399 5405
5400 5406 case ZFS_SMB_ACL_RENAME:
5401 5407 if ((error = get_nvlist(zc->zc_nvlist_src,
5402 5408 zc->zc_nvlist_src_size, zc->zc_iflags, &nvlist)) != 0) {
5403 5409 VN_RELE(vp);
5404 5410 VN_RELE(ZTOV(sharedir));
5405 5411 ZFS_EXIT(zfsvfs);
5406 5412 return (error);
5407 5413 }
5408 5414 if (nvlist_lookup_string(nvlist, ZFS_SMB_ACL_SRC, &src) ||
5409 5415 nvlist_lookup_string(nvlist, ZFS_SMB_ACL_TARGET,
5410 5416 &target)) {
5411 5417 VN_RELE(vp);
5412 5418 VN_RELE(ZTOV(sharedir));
5413 5419 ZFS_EXIT(zfsvfs);
5414 5420 nvlist_free(nvlist);
5415 5421 return (error);
5416 5422 }
5417 5423 error = VOP_RENAME(ZTOV(sharedir), src, ZTOV(sharedir), target,
5418 5424 kcred, NULL, 0);
5419 5425 nvlist_free(nvlist);
5420 5426 break;
5421 5427
5422 5428 case ZFS_SMB_ACL_PURGE:
5423 5429 error = zfs_smb_acl_purge(sharedir);
5424 5430 break;
5425 5431
5426 5432 default:
5427 5433 error = SET_ERROR(EINVAL);
5428 5434 break;
5429 5435 }
5430 5436
5431 5437 VN_RELE(vp);
5432 5438 VN_RELE(ZTOV(sharedir));
5433 5439
5434 5440 ZFS_EXIT(zfsvfs);
5435 5441
5436 5442 return (error);
5437 5443 }
5438 5444
5439 5445 /*
5440 5446 * innvl: {
5441 5447 * "holds" -> { snapname -> holdname (string), ... }
5442 5448 * (optional) "cleanup_fd" -> fd (int32)
5443 5449 * }
5444 5450 *
5445 5451 * outnvl: {
5446 5452 * snapname -> error value (int32)
5447 5453 * ...
5448 5454 * }
5449 5455 */
5450 5456 /* ARGSUSED */
5451 5457 static int
5452 5458 zfs_ioc_hold(const char *pool, nvlist_t *args, nvlist_t *errlist)
5453 5459 {
5454 5460 nvpair_t *pair;
5455 5461 nvlist_t *holds;
5456 5462 int cleanup_fd = -1;
5457 5463 int error;
5458 5464 minor_t minor = 0;
5459 5465
5460 5466 error = nvlist_lookup_nvlist(args, "holds", &holds);
5461 5467 if (error != 0)
5462 5468 return (SET_ERROR(EINVAL));
5463 5469
5464 5470 /* make sure the user didn't pass us any invalid (empty) tags */
5465 5471 for (pair = nvlist_next_nvpair(holds, NULL); pair != NULL;
5466 5472 pair = nvlist_next_nvpair(holds, pair)) {
5467 5473 char *htag;
5468 5474
5469 5475 error = nvpair_value_string(pair, &htag);
5470 5476 if (error != 0)
5471 5477 return (SET_ERROR(error));
5472 5478
5473 5479 if (strlen(htag) == 0)
5474 5480 return (SET_ERROR(EINVAL));
5475 5481 }
5476 5482
5477 5483 if (nvlist_lookup_int32(args, "cleanup_fd", &cleanup_fd) == 0) {
5478 5484 error = zfs_onexit_fd_hold(cleanup_fd, &minor);
5479 5485 if (error != 0)
5480 5486 return (error);
5481 5487 }
5482 5488
5483 5489 error = dsl_dataset_user_hold(holds, minor, errlist);
5484 5490 if (minor != 0)
5485 5491 zfs_onexit_fd_rele(cleanup_fd);
5486 5492 return (error);
5487 5493 }
5488 5494
5489 5495 /*
5490 5496 * innvl is not used.
5491 5497 *
5492 5498 * outnvl: {
5493 5499 * holdname -> time added (uint64 seconds since epoch)
5494 5500 * ...
5495 5501 * }
5496 5502 */
5497 5503 /* ARGSUSED */
5498 5504 static int
5499 5505 zfs_ioc_get_holds(const char *snapname, nvlist_t *args, nvlist_t *outnvl)
5500 5506 {
5501 5507 return (dsl_dataset_get_holds(snapname, outnvl));
5502 5508 }
5503 5509
5504 5510 /*
5505 5511 * innvl: {
5506 5512 * snapname -> { holdname, ... }
5507 5513 * ...
5508 5514 * }
5509 5515 *
5510 5516 * outnvl: {
5511 5517 * snapname -> error value (int32)
5512 5518 * ...
5513 5519 * }
5514 5520 */
5515 5521 /* ARGSUSED */
5516 5522 static int
5517 5523 zfs_ioc_release(const char *pool, nvlist_t *holds, nvlist_t *errlist)
5518 5524 {
5519 5525 return (dsl_dataset_user_release(holds, errlist));
5520 5526 }
5521 5527
5522 5528 /*
5523 5529 * inputs:
5524 5530 * zc_name name of new filesystem or snapshot
5525 5531 * zc_value full name of old snapshot
5526 5532 *
5527 5533 * outputs:
5528 5534 * zc_cookie space in bytes
5529 5535 * zc_objset_type compressed space in bytes
5530 5536 * zc_perm_action uncompressed space in bytes
5531 5537 */
5532 5538 static int
5533 5539 zfs_ioc_space_written(zfs_cmd_t *zc)
5534 5540 {
5535 5541 int error;
5536 5542 dsl_pool_t *dp;
5537 5543 dsl_dataset_t *new, *old;
5538 5544
5539 5545 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
5540 5546 if (error != 0)
5541 5547 return (error);
5542 5548 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &new);
5543 5549 if (error != 0) {
5544 5550 dsl_pool_rele(dp, FTAG);
5545 5551 return (error);
5546 5552 }
5547 5553 error = dsl_dataset_hold(dp, zc->zc_value, FTAG, &old);
5548 5554 if (error != 0) {
5549 5555 dsl_dataset_rele(new, FTAG);
5550 5556 dsl_pool_rele(dp, FTAG);
5551 5557 return (error);
5552 5558 }
5553 5559
5554 5560 error = dsl_dataset_space_written(old, new, &zc->zc_cookie,
5555 5561 &zc->zc_objset_type, &zc->zc_perm_action);
5556 5562 dsl_dataset_rele(old, FTAG);
5557 5563 dsl_dataset_rele(new, FTAG);
5558 5564 dsl_pool_rele(dp, FTAG);
5559 5565 return (error);
5560 5566 }
5561 5567
5562 5568 /*
5563 5569 * innvl: {
5564 5570 * "firstsnap" -> snapshot name
5565 5571 * }
5566 5572 *
5567 5573 * outnvl: {
5568 5574 * "used" -> space in bytes
5569 5575 * "compressed" -> compressed space in bytes
5570 5576 * "uncompressed" -> uncompressed space in bytes
5571 5577 * }
5572 5578 */
5573 5579 static int
5574 5580 zfs_ioc_space_snaps(const char *lastsnap, nvlist_t *innvl, nvlist_t *outnvl)
5575 5581 {
5576 5582 int error;
5577 5583 dsl_pool_t *dp;
5578 5584 dsl_dataset_t *new, *old;
5579 5585 char *firstsnap;
5580 5586 uint64_t used, comp, uncomp;
5581 5587
5582 5588 if (nvlist_lookup_string(innvl, "firstsnap", &firstsnap) != 0)
5583 5589 return (SET_ERROR(EINVAL));
5584 5590
5585 5591 error = dsl_pool_hold(lastsnap, FTAG, &dp);
5586 5592 if (error != 0)
5587 5593 return (error);
5588 5594
5589 5595 error = dsl_dataset_hold(dp, lastsnap, FTAG, &new);
5590 5596 if (error == 0 && !new->ds_is_snapshot) {
5591 5597 dsl_dataset_rele(new, FTAG);
5592 5598 error = SET_ERROR(EINVAL);
5593 5599 }
5594 5600 if (error != 0) {
5595 5601 dsl_pool_rele(dp, FTAG);
5596 5602 return (error);
5597 5603 }
5598 5604 error = dsl_dataset_hold(dp, firstsnap, FTAG, &old);
5599 5605 if (error == 0 && !old->ds_is_snapshot) {
5600 5606 dsl_dataset_rele(old, FTAG);
5601 5607 error = SET_ERROR(EINVAL);
5602 5608 }
5603 5609 if (error != 0) {
5604 5610 dsl_dataset_rele(new, FTAG);
5605 5611 dsl_pool_rele(dp, FTAG);
5606 5612 return (error);
5607 5613 }
5608 5614
5609 5615 error = dsl_dataset_space_wouldfree(old, new, &used, &comp, &uncomp);
5610 5616 dsl_dataset_rele(old, FTAG);
5611 5617 dsl_dataset_rele(new, FTAG);
5612 5618 dsl_pool_rele(dp, FTAG);
5613 5619 fnvlist_add_uint64(outnvl, "used", used);
5614 5620 fnvlist_add_uint64(outnvl, "compressed", comp);
5615 5621 fnvlist_add_uint64(outnvl, "uncompressed", uncomp);
5616 5622 return (error);
5617 5623 }
5618 5624
5619 5625 /*
5620 5626 * innvl: {
5621 5627 * "fd" -> file descriptor to write stream to (int32)
5622 5628 * (optional) "fromsnap" -> full snap name to send an incremental from
5623 5629 * (optional) "largeblockok" -> (value ignored)
5624 5630 * indicates that blocks > 128KB are permitted
5625 5631 * (optional) "embedok" -> (value ignored)
5626 5632 * presence indicates DRR_WRITE_EMBEDDED records are permitted
5627 5633 * (optional) "compressok" -> (value ignored)
5628 5634 * presence indicates compressed DRR_WRITE records are permitted
5629 5635 * (optional) "resume_object" and "resume_offset" -> (uint64)
5630 5636 * if present, resume send stream from specified object and offset.
5631 5637 * }
5632 5638 *
5633 5639 * outnvl is unused
5634 5640 */
5635 5641 /* ARGSUSED */
5636 5642 static int
5637 5643 zfs_ioc_send_new(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl)
5638 5644 {
5639 5645 int error;
5640 5646 offset_t off;
5641 5647 char *fromname = NULL;
5642 5648 int fd;
5643 5649 boolean_t largeblockok;
5644 5650 boolean_t embedok;
5645 5651 boolean_t compressok;
5646 5652 uint64_t resumeobj = 0;
5647 5653 uint64_t resumeoff = 0;
5648 5654
5649 5655 error = nvlist_lookup_int32(innvl, "fd", &fd);
5650 5656 if (error != 0)
5651 5657 return (SET_ERROR(EINVAL));
5652 5658
5653 5659 (void) nvlist_lookup_string(innvl, "fromsnap", &fromname);
5654 5660
5655 5661 largeblockok = nvlist_exists(innvl, "largeblockok");
5656 5662 embedok = nvlist_exists(innvl, "embedok");
5657 5663 compressok = nvlist_exists(innvl, "compressok");
5658 5664
5659 5665 (void) nvlist_lookup_uint64(innvl, "resume_object", &resumeobj);
5660 5666 (void) nvlist_lookup_uint64(innvl, "resume_offset", &resumeoff);
5661 5667
5662 5668 file_t *fp = getf(fd);
5663 5669 if (fp == NULL)
5664 5670 return (SET_ERROR(EBADF));
5665 5671
5666 5672 off = fp->f_offset;
5667 5673 error = dmu_send(snapname, fromname, embedok, largeblockok, compressok,
5668 5674 fd, resumeobj, resumeoff, fp->f_vnode, &off);
5669 5675
5670 5676 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
5671 5677 fp->f_offset = off;
5672 5678 releasef(fd);
5673 5679 return (error);
5674 5680 }
5675 5681
5676 5682 /*
5677 5683 * Determine approximately how large a zfs send stream will be -- the number
5678 5684 * of bytes that will be written to the fd supplied to zfs_ioc_send_new().
5679 5685 *
5680 5686 * innvl: {
5681 5687 * (optional) "from" -> full snap or bookmark name to send an incremental
5682 5688 * from
5683 5689 * (optional) "largeblockok" -> (value ignored)
5684 5690 * indicates that blocks > 128KB are permitted
5685 5691 * (optional) "embedok" -> (value ignored)
5686 5692 * presence indicates DRR_WRITE_EMBEDDED records are permitted
5687 5693 * (optional) "compressok" -> (value ignored)
5688 5694 * presence indicates compressed DRR_WRITE records are permitted
5689 5695 * }
5690 5696 *
5691 5697 * outnvl: {
5692 5698 * "space" -> bytes of space (uint64)
5693 5699 * }
5694 5700 */
5695 5701 static int
5696 5702 zfs_ioc_send_space(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl)
5697 5703 {
5698 5704 dsl_pool_t *dp;
5699 5705 dsl_dataset_t *tosnap;
5700 5706 int error;
5701 5707 char *fromname;
5702 5708 boolean_t compressok;
5703 5709 uint64_t space;
5704 5710
5705 5711 error = dsl_pool_hold(snapname, FTAG, &dp);
5706 5712 if (error != 0)
5707 5713 return (error);
5708 5714
5709 5715 error = dsl_dataset_hold(dp, snapname, FTAG, &tosnap);
5710 5716 if (error != 0) {
5711 5717 dsl_pool_rele(dp, FTAG);
5712 5718 return (error);
5713 5719 }
5714 5720
5715 5721 compressok = nvlist_exists(innvl, "compressok");
5716 5722
5717 5723 error = nvlist_lookup_string(innvl, "from", &fromname);
5718 5724 if (error == 0) {
5719 5725 if (strchr(fromname, '@') != NULL) {
5720 5726 /*
5721 5727 * If from is a snapshot, hold it and use the more
5722 5728 * efficient dmu_send_estimate to estimate send space
5723 5729 * size using deadlists.
5724 5730 */
5725 5731 dsl_dataset_t *fromsnap;
5726 5732 error = dsl_dataset_hold(dp, fromname, FTAG, &fromsnap);
5727 5733 if (error != 0)
5728 5734 goto out;
5729 5735 error = dmu_send_estimate(tosnap, fromsnap, compressok,
5730 5736 &space);
5731 5737 dsl_dataset_rele(fromsnap, FTAG);
5732 5738 } else if (strchr(fromname, '#') != NULL) {
5733 5739 /*
5734 5740 * If from is a bookmark, fetch the creation TXG of the
5735 5741 * snapshot it was created from and use that to find
5736 5742 * blocks that were born after it.
5737 5743 */
5738 5744 zfs_bookmark_phys_t frombm;
5739 5745
5740 5746 error = dsl_bookmark_lookup(dp, fromname, tosnap,
5741 5747 &frombm);
5742 5748 if (error != 0)
5743 5749 goto out;
5744 5750 error = dmu_send_estimate_from_txg(tosnap,
5745 5751 frombm.zbm_creation_txg, compressok, &space);
5746 5752 } else {
5747 5753 /*
5748 5754 * from is not properly formatted as a snapshot or
5749 5755 * bookmark
5750 5756 */
5751 5757 error = SET_ERROR(EINVAL);
5752 5758 goto out;
5753 5759 }
5754 5760 } else {
5755 5761 /*
5756 5762 * If estimating the size of a full send, use dmu_send_estimate.
5757 5763 */
5758 5764 error = dmu_send_estimate(tosnap, NULL, compressok, &space);
5759 5765 }
5760 5766
5761 5767 fnvlist_add_uint64(outnvl, "space", space);
5762 5768
5763 5769 out:
5764 5770 dsl_dataset_rele(tosnap, FTAG);
5765 5771 dsl_pool_rele(dp, FTAG);
5766 5772 return (error);
5767 5773 }
5768 5774
5769 5775 static zfs_ioc_vec_t zfs_ioc_vec[ZFS_IOC_LAST - ZFS_IOC_FIRST];
5770 5776
5771 5777 static void
5772 5778 zfs_ioctl_register_legacy(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5773 5779 zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck,
5774 5780 boolean_t log_history, zfs_ioc_poolcheck_t pool_check)
5775 5781 {
5776 5782 zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST];
5777 5783
5778 5784 ASSERT3U(ioc, >=, ZFS_IOC_FIRST);
5779 5785 ASSERT3U(ioc, <, ZFS_IOC_LAST);
5780 5786 ASSERT3P(vec->zvec_legacy_func, ==, NULL);
5781 5787 ASSERT3P(vec->zvec_func, ==, NULL);
5782 5788
5783 5789 vec->zvec_legacy_func = func;
5784 5790 vec->zvec_secpolicy = secpolicy;
5785 5791 vec->zvec_namecheck = namecheck;
5786 5792 vec->zvec_allow_log = log_history;
5787 5793 vec->zvec_pool_check = pool_check;
5788 5794 }
5789 5795
5790 5796 /*
5791 5797 * See the block comment at the beginning of this file for details on
5792 5798 * each argument to this function.
5793 5799 */
5794 5800 static void
5795 5801 zfs_ioctl_register(const char *name, zfs_ioc_t ioc, zfs_ioc_func_t *func,
5796 5802 zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck,
5797 5803 zfs_ioc_poolcheck_t pool_check, boolean_t smush_outnvlist,
5798 5804 boolean_t allow_log)
5799 5805 {
5800 5806 zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST];
5801 5807
5802 5808 ASSERT3U(ioc, >=, ZFS_IOC_FIRST);
5803 5809 ASSERT3U(ioc, <, ZFS_IOC_LAST);
5804 5810 ASSERT3P(vec->zvec_legacy_func, ==, NULL);
5805 5811 ASSERT3P(vec->zvec_func, ==, NULL);
5806 5812
5807 5813 /* if we are logging, the name must be valid */
5808 5814 ASSERT(!allow_log || namecheck != NO_NAME);
5809 5815
5810 5816 vec->zvec_name = name;
5811 5817 vec->zvec_func = func;
5812 5818 vec->zvec_secpolicy = secpolicy;
5813 5819 vec->zvec_namecheck = namecheck;
5814 5820 vec->zvec_pool_check = pool_check;
5815 5821 vec->zvec_smush_outnvlist = smush_outnvlist;
5816 5822 vec->zvec_allow_log = allow_log;
5817 5823 }
5818 5824
5819 5825 static void
5820 5826 zfs_ioctl_register_pool(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5821 5827 zfs_secpolicy_func_t *secpolicy, boolean_t log_history,
5822 5828 zfs_ioc_poolcheck_t pool_check)
5823 5829 {
5824 5830 zfs_ioctl_register_legacy(ioc, func, secpolicy,
5825 5831 POOL_NAME, log_history, pool_check);
5826 5832 }
5827 5833
5828 5834 static void
5829 5835 zfs_ioctl_register_dataset_nolog(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5830 5836 zfs_secpolicy_func_t *secpolicy, zfs_ioc_poolcheck_t pool_check)
5831 5837 {
5832 5838 zfs_ioctl_register_legacy(ioc, func, secpolicy,
5833 5839 DATASET_NAME, B_FALSE, pool_check);
5834 5840 }
5835 5841
5836 5842 static void
5837 5843 zfs_ioctl_register_pool_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func)
5838 5844 {
5839 5845 zfs_ioctl_register_legacy(ioc, func, zfs_secpolicy_config,
5840 5846 POOL_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5841 5847 }
5842 5848
5843 5849 static void
5844 5850 zfs_ioctl_register_pool_meta(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5845 5851 zfs_secpolicy_func_t *secpolicy)
5846 5852 {
5847 5853 zfs_ioctl_register_legacy(ioc, func, secpolicy,
5848 5854 NO_NAME, B_FALSE, POOL_CHECK_NONE);
5849 5855 }
5850 5856
5851 5857 static void
5852 5858 zfs_ioctl_register_dataset_read_secpolicy(zfs_ioc_t ioc,
5853 5859 zfs_ioc_legacy_func_t *func, zfs_secpolicy_func_t *secpolicy)
5854 5860 {
5855 5861 zfs_ioctl_register_legacy(ioc, func, secpolicy,
5856 5862 DATASET_NAME, B_FALSE, POOL_CHECK_SUSPENDED);
5857 5863 }
5858 5864
5859 5865 static void
5860 5866 zfs_ioctl_register_dataset_read(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func)
5861 5867 {
5862 5868 zfs_ioctl_register_dataset_read_secpolicy(ioc, func,
5863 5869 zfs_secpolicy_read);
5864 5870 }
5865 5871
5866 5872 static void
5867 5873 zfs_ioctl_register_dataset_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5868 5874 zfs_secpolicy_func_t *secpolicy)
5869 5875 {
5870 5876 zfs_ioctl_register_legacy(ioc, func, secpolicy,
5871 5877 DATASET_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5872 5878 }
5873 5879
5874 5880 static void
5875 5881 zfs_ioctl_init(void)
5876 5882 {
5877 5883 zfs_ioctl_register("snapshot", ZFS_IOC_SNAPSHOT,
5878 5884 zfs_ioc_snapshot, zfs_secpolicy_snapshot, POOL_NAME,
5879 5885 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5880 5886
5881 5887 zfs_ioctl_register("log_history", ZFS_IOC_LOG_HISTORY,
5882 5888 zfs_ioc_log_history, zfs_secpolicy_log_history, NO_NAME,
5883 5889 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_FALSE);
5884 5890
5885 5891 zfs_ioctl_register("space_snaps", ZFS_IOC_SPACE_SNAPS,
5886 5892 zfs_ioc_space_snaps, zfs_secpolicy_read, DATASET_NAME,
5887 5893 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5888 5894
5889 5895 zfs_ioctl_register("send", ZFS_IOC_SEND_NEW,
5890 5896 zfs_ioc_send_new, zfs_secpolicy_send_new, DATASET_NAME,
5891 5897 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5892 5898
5893 5899 zfs_ioctl_register("send_space", ZFS_IOC_SEND_SPACE,
5894 5900 zfs_ioc_send_space, zfs_secpolicy_read, DATASET_NAME,
5895 5901 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5896 5902
5897 5903 zfs_ioctl_register("create", ZFS_IOC_CREATE,
5898 5904 zfs_ioc_create, zfs_secpolicy_create_clone, DATASET_NAME,
5899 5905 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5900 5906
5901 5907 zfs_ioctl_register("clone", ZFS_IOC_CLONE,
5902 5908 zfs_ioc_clone, zfs_secpolicy_create_clone, DATASET_NAME,
5903 5909 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5904 5910
5905 5911 zfs_ioctl_register("remap", ZFS_IOC_REMAP,
5906 5912 zfs_ioc_remap, zfs_secpolicy_remap, DATASET_NAME,
5907 5913 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_TRUE);
5908 5914
5909 5915 zfs_ioctl_register("destroy_snaps", ZFS_IOC_DESTROY_SNAPS,
5910 5916 zfs_ioc_destroy_snaps, zfs_secpolicy_destroy_snaps, POOL_NAME,
5911 5917 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5912 5918
5913 5919 zfs_ioctl_register("hold", ZFS_IOC_HOLD,
5914 5920 zfs_ioc_hold, zfs_secpolicy_hold, POOL_NAME,
5915 5921 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5916 5922 zfs_ioctl_register("release", ZFS_IOC_RELEASE,
5917 5923 zfs_ioc_release, zfs_secpolicy_release, POOL_NAME,
5918 5924 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5919 5925
5920 5926 zfs_ioctl_register("get_holds", ZFS_IOC_GET_HOLDS,
5921 5927 zfs_ioc_get_holds, zfs_secpolicy_read, DATASET_NAME,
5922 5928 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5923 5929
5924 5930 zfs_ioctl_register("rollback", ZFS_IOC_ROLLBACK,
5925 5931 zfs_ioc_rollback, zfs_secpolicy_rollback, DATASET_NAME,
5926 5932 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_TRUE);
5927 5933
5928 5934 zfs_ioctl_register("bookmark", ZFS_IOC_BOOKMARK,
5929 5935 zfs_ioc_bookmark, zfs_secpolicy_bookmark, POOL_NAME,
5930 5936 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5931 5937
5932 5938 zfs_ioctl_register("get_bookmarks", ZFS_IOC_GET_BOOKMARKS,
5933 5939 zfs_ioc_get_bookmarks, zfs_secpolicy_read, DATASET_NAME,
5934 5940 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5935 5941
5936 5942 zfs_ioctl_register("destroy_bookmarks", ZFS_IOC_DESTROY_BOOKMARKS,
5937 5943 zfs_ioc_destroy_bookmarks, zfs_secpolicy_destroy_bookmarks,
5938 5944 POOL_NAME,
5939 5945 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5940 5946
5941 5947 zfs_ioctl_register("channel_program", ZFS_IOC_CHANNEL_PROGRAM,
5942 5948 zfs_ioc_channel_program, zfs_secpolicy_config,
5943 5949 POOL_NAME, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE,
5944 5950 B_TRUE);
5945 5951
5946 5952 zfs_ioctl_register("zpool_checkpoint", ZFS_IOC_POOL_CHECKPOINT,
5947 5953 zfs_ioc_pool_checkpoint, zfs_secpolicy_config, POOL_NAME,
5948 5954 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5949 5955
5950 5956 zfs_ioctl_register("zpool_discard_checkpoint",
5951 5957 ZFS_IOC_POOL_DISCARD_CHECKPOINT, zfs_ioc_pool_discard_checkpoint,
5952 5958 zfs_secpolicy_config, POOL_NAME,
5953 5959 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5954 5960
5955 5961 zfs_ioctl_register("initialize", ZFS_IOC_POOL_INITIALIZE,
5956 5962 zfs_ioc_pool_initialize, zfs_secpolicy_config, POOL_NAME,
5957 5963 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5958 5964
5959 5965 /* IOCTLS that use the legacy function signature */
5960 5966
5961 5967 zfs_ioctl_register_legacy(ZFS_IOC_POOL_FREEZE, zfs_ioc_pool_freeze,
5962 5968 zfs_secpolicy_config, NO_NAME, B_FALSE, POOL_CHECK_READONLY);
5963 5969
5964 5970 zfs_ioctl_register_pool(ZFS_IOC_POOL_CREATE, zfs_ioc_pool_create,
5965 5971 zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE);
5966 5972 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SCAN,
5967 5973 zfs_ioc_pool_scan);
5968 5974 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_UPGRADE,
5969 5975 zfs_ioc_pool_upgrade);
5970 5976 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ADD,
5971 5977 zfs_ioc_vdev_add);
5972 5978 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_REMOVE,
5973 5979 zfs_ioc_vdev_remove);
5974 5980 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SET_STATE,
5975 5981 zfs_ioc_vdev_set_state);
5976 5982 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ATTACH,
5977 5983 zfs_ioc_vdev_attach);
5978 5984 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_DETACH,
5979 5985 zfs_ioc_vdev_detach);
5980 5986 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETPATH,
5981 5987 zfs_ioc_vdev_setpath);
5982 5988 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETFRU,
5983 5989 zfs_ioc_vdev_setfru);
5984 5990 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SET_PROPS,
5985 5991 zfs_ioc_pool_set_props);
5986 5992 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SPLIT,
5987 5993 zfs_ioc_vdev_split);
5988 5994 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_REGUID,
5989 5995 zfs_ioc_pool_reguid);
5990 5996
5991 5997 zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_CONFIGS,
5992 5998 zfs_ioc_pool_configs, zfs_secpolicy_none);
5993 5999 zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_TRYIMPORT,
5994 6000 zfs_ioc_pool_tryimport, zfs_secpolicy_config);
5995 6001 zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_FAULT,
5996 6002 zfs_ioc_inject_fault, zfs_secpolicy_inject);
5997 6003 zfs_ioctl_register_pool_meta(ZFS_IOC_CLEAR_FAULT,
5998 6004 zfs_ioc_clear_fault, zfs_secpolicy_inject);
5999 6005 zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_LIST_NEXT,
6000 6006 zfs_ioc_inject_list_next, zfs_secpolicy_inject);
6001 6007
6002 6008 /*
6003 6009 * pool destroy, and export don't log the history as part of
6004 6010 * zfsdev_ioctl, but rather zfs_ioc_pool_export
6005 6011 * does the logging of those commands.
6006 6012 */
6007 6013 zfs_ioctl_register_pool(ZFS_IOC_POOL_DESTROY, zfs_ioc_pool_destroy,
6008 6014 zfs_secpolicy_config, B_FALSE, POOL_CHECK_NONE);
6009 6015 zfs_ioctl_register_pool(ZFS_IOC_POOL_EXPORT, zfs_ioc_pool_export,
6010 6016 zfs_secpolicy_config, B_FALSE, POOL_CHECK_NONE);
6011 6017
6012 6018 zfs_ioctl_register_pool(ZFS_IOC_POOL_STATS, zfs_ioc_pool_stats,
6013 6019 zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE);
6014 6020 zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_PROPS, zfs_ioc_pool_get_props,
6015 6021 zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE);
6016 6022
6017 6023 zfs_ioctl_register_pool(ZFS_IOC_ERROR_LOG, zfs_ioc_error_log,
6018 6024 zfs_secpolicy_inject, B_FALSE, POOL_CHECK_SUSPENDED);
6019 6025 zfs_ioctl_register_pool(ZFS_IOC_DSOBJ_TO_DSNAME,
6020 6026 zfs_ioc_dsobj_to_dsname,
6021 6027 zfs_secpolicy_diff, B_FALSE, POOL_CHECK_SUSPENDED);
6022 6028 zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_HISTORY,
6023 6029 zfs_ioc_pool_get_history,
6024 6030 zfs_secpolicy_config, B_FALSE, POOL_CHECK_SUSPENDED);
6025 6031
6026 6032 zfs_ioctl_register_pool(ZFS_IOC_POOL_IMPORT, zfs_ioc_pool_import,
6027 6033 zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE);
6028 6034
6029 6035 zfs_ioctl_register_pool(ZFS_IOC_CLEAR, zfs_ioc_clear,
6030 6036 zfs_secpolicy_config, B_TRUE, POOL_CHECK_READONLY);
6031 6037 zfs_ioctl_register_pool(ZFS_IOC_POOL_REOPEN, zfs_ioc_pool_reopen,
6032 6038 zfs_secpolicy_config, B_TRUE, POOL_CHECK_SUSPENDED);
6033 6039
6034 6040 zfs_ioctl_register_dataset_read(ZFS_IOC_SPACE_WRITTEN,
6035 6041 zfs_ioc_space_written);
6036 6042 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_RECVD_PROPS,
6037 6043 zfs_ioc_objset_recvd_props);
6038 6044 zfs_ioctl_register_dataset_read(ZFS_IOC_NEXT_OBJ,
6039 6045 zfs_ioc_next_obj);
6040 6046 zfs_ioctl_register_dataset_read(ZFS_IOC_GET_FSACL,
6041 6047 zfs_ioc_get_fsacl);
6042 6048 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_STATS,
6043 6049 zfs_ioc_objset_stats);
6044 6050 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_ZPLPROPS,
6045 6051 zfs_ioc_objset_zplprops);
6046 6052 zfs_ioctl_register_dataset_read(ZFS_IOC_DATASET_LIST_NEXT,
6047 6053 zfs_ioc_dataset_list_next);
6048 6054 zfs_ioctl_register_dataset_read(ZFS_IOC_SNAPSHOT_LIST_NEXT,
6049 6055 zfs_ioc_snapshot_list_next);
6050 6056 zfs_ioctl_register_dataset_read(ZFS_IOC_SEND_PROGRESS,
6051 6057 zfs_ioc_send_progress);
6052 6058
6053 6059 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_DIFF,
6054 6060 zfs_ioc_diff, zfs_secpolicy_diff);
6055 6061 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_STATS,
6056 6062 zfs_ioc_obj_to_stats, zfs_secpolicy_diff);
6057 6063 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_PATH,
6058 6064 zfs_ioc_obj_to_path, zfs_secpolicy_diff);
6059 6065 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_ONE,
6060 6066 zfs_ioc_userspace_one, zfs_secpolicy_userspace_one);
6061 6067 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_MANY,
6062 6068 zfs_ioc_userspace_many, zfs_secpolicy_userspace_many);
6063 6069 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_SEND,
6064 6070 zfs_ioc_send, zfs_secpolicy_send);
6065 6071
6066 6072 zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_PROP, zfs_ioc_set_prop,
6067 6073 zfs_secpolicy_none);
6068 6074 zfs_ioctl_register_dataset_modify(ZFS_IOC_DESTROY, zfs_ioc_destroy,
6069 6075 zfs_secpolicy_destroy);
6070 6076 zfs_ioctl_register_dataset_modify(ZFS_IOC_RENAME, zfs_ioc_rename,
6071 6077 zfs_secpolicy_rename);
6072 6078 zfs_ioctl_register_dataset_modify(ZFS_IOC_RECV, zfs_ioc_recv,
6073 6079 zfs_secpolicy_recv);
6074 6080 zfs_ioctl_register_dataset_modify(ZFS_IOC_PROMOTE, zfs_ioc_promote,
6075 6081 zfs_secpolicy_promote);
6076 6082 zfs_ioctl_register_dataset_modify(ZFS_IOC_INHERIT_PROP,
6077 6083 zfs_ioc_inherit_prop, zfs_secpolicy_inherit_prop);
6078 6084 zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_FSACL, zfs_ioc_set_fsacl,
6079 6085 zfs_secpolicy_set_fsacl);
6080 6086
6081 6087 zfs_ioctl_register_dataset_nolog(ZFS_IOC_SHARE, zfs_ioc_share,
6082 6088 zfs_secpolicy_share, POOL_CHECK_NONE);
6083 6089 zfs_ioctl_register_dataset_nolog(ZFS_IOC_SMB_ACL, zfs_ioc_smb_acl,
6084 6090 zfs_secpolicy_smb_acl, POOL_CHECK_NONE);
6085 6091 zfs_ioctl_register_dataset_nolog(ZFS_IOC_USERSPACE_UPGRADE,
6086 6092 zfs_ioc_userspace_upgrade, zfs_secpolicy_userspace_upgrade,
6087 6093 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
6088 6094 zfs_ioctl_register_dataset_nolog(ZFS_IOC_TMP_SNAPSHOT,
6089 6095 zfs_ioc_tmp_snapshot, zfs_secpolicy_tmp_snapshot,
6090 6096 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
6091 6097 }
6092 6098
6093 6099 int
6094 6100 pool_status_check(const char *name, zfs_ioc_namecheck_t type,
6095 6101 zfs_ioc_poolcheck_t check)
6096 6102 {
6097 6103 spa_t *spa;
6098 6104 int error;
6099 6105
6100 6106 ASSERT(type == POOL_NAME || type == DATASET_NAME);
6101 6107
6102 6108 if (check & POOL_CHECK_NONE)
6103 6109 return (0);
6104 6110
6105 6111 error = spa_open(name, &spa, FTAG);
6106 6112 if (error == 0) {
6107 6113 if ((check & POOL_CHECK_SUSPENDED) && spa_suspended(spa))
6108 6114 error = SET_ERROR(EAGAIN);
6109 6115 else if ((check & POOL_CHECK_READONLY) && !spa_writeable(spa))
6110 6116 error = SET_ERROR(EROFS);
6111 6117 spa_close(spa, FTAG);
6112 6118 }
6113 6119 return (error);
6114 6120 }
6115 6121
6116 6122 /*
6117 6123 * Find a free minor number.
6118 6124 */
6119 6125 minor_t
6120 6126 zfsdev_minor_alloc(void)
6121 6127 {
6122 6128 static minor_t last_minor;
6123 6129 minor_t m;
6124 6130
6125 6131 ASSERT(MUTEX_HELD(&zfsdev_state_lock));
6126 6132
6127 6133 for (m = last_minor + 1; m != last_minor; m++) {
6128 6134 if (m > ZFSDEV_MAX_MINOR)
6129 6135 m = 1;
6130 6136 if (ddi_get_soft_state(zfsdev_state, m) == NULL) {
6131 6137 last_minor = m;
6132 6138 return (m);
6133 6139 }
6134 6140 }
6135 6141
6136 6142 return (0);
6137 6143 }
6138 6144
6139 6145 static int
6140 6146 zfs_ctldev_init(dev_t *devp)
6141 6147 {
6142 6148 minor_t minor;
6143 6149 zfs_soft_state_t *zs;
6144 6150
6145 6151 ASSERT(MUTEX_HELD(&zfsdev_state_lock));
6146 6152 ASSERT(getminor(*devp) == 0);
6147 6153
6148 6154 minor = zfsdev_minor_alloc();
6149 6155 if (minor == 0)
6150 6156 return (SET_ERROR(ENXIO));
6151 6157
6152 6158 if (ddi_soft_state_zalloc(zfsdev_state, minor) != DDI_SUCCESS)
6153 6159 return (SET_ERROR(EAGAIN));
6154 6160
6155 6161 *devp = makedevice(getemajor(*devp), minor);
6156 6162
6157 6163 zs = ddi_get_soft_state(zfsdev_state, minor);
6158 6164 zs->zss_type = ZSST_CTLDEV;
6159 6165 zfs_onexit_init((zfs_onexit_t **)&zs->zss_data);
6160 6166
6161 6167 return (0);
6162 6168 }
6163 6169
6164 6170 static void
6165 6171 zfs_ctldev_destroy(zfs_onexit_t *zo, minor_t minor)
6166 6172 {
6167 6173 ASSERT(MUTEX_HELD(&zfsdev_state_lock));
6168 6174
6169 6175 zfs_onexit_destroy(zo);
6170 6176 ddi_soft_state_free(zfsdev_state, minor);
6171 6177 }
6172 6178
6173 6179 void *
6174 6180 zfsdev_get_soft_state(minor_t minor, enum zfs_soft_state_type which)
6175 6181 {
6176 6182 zfs_soft_state_t *zp;
6177 6183
6178 6184 zp = ddi_get_soft_state(zfsdev_state, minor);
6179 6185 if (zp == NULL || zp->zss_type != which)
6180 6186 return (NULL);
6181 6187
6182 6188 return (zp->zss_data);
6183 6189 }
6184 6190
6185 6191 static int
6186 6192 zfsdev_open(dev_t *devp, int flag, int otyp, cred_t *cr)
6187 6193 {
6188 6194 int error = 0;
6189 6195
6190 6196 if (getminor(*devp) != 0)
6191 6197 return (zvol_open(devp, flag, otyp, cr));
6192 6198
6193 6199 /* This is the control device. Allocate a new minor if requested. */
6194 6200 if (flag & FEXCL) {
6195 6201 mutex_enter(&zfsdev_state_lock);
6196 6202 error = zfs_ctldev_init(devp);
6197 6203 mutex_exit(&zfsdev_state_lock);
6198 6204 }
6199 6205
6200 6206 return (error);
6201 6207 }
6202 6208
6203 6209 static int
6204 6210 zfsdev_close(dev_t dev, int flag, int otyp, cred_t *cr)
6205 6211 {
6206 6212 zfs_onexit_t *zo;
6207 6213 minor_t minor = getminor(dev);
6208 6214
6209 6215 if (minor == 0)
6210 6216 return (0);
6211 6217
6212 6218 mutex_enter(&zfsdev_state_lock);
6213 6219 zo = zfsdev_get_soft_state(minor, ZSST_CTLDEV);
6214 6220 if (zo == NULL) {
6215 6221 mutex_exit(&zfsdev_state_lock);
6216 6222 return (zvol_close(dev, flag, otyp, cr));
6217 6223 }
6218 6224 zfs_ctldev_destroy(zo, minor);
6219 6225 mutex_exit(&zfsdev_state_lock);
6220 6226
6221 6227 return (0);
6222 6228 }
6223 6229
6224 6230 static int
6225 6231 zfsdev_ioctl(dev_t dev, int cmd, intptr_t arg, int flag, cred_t *cr, int *rvalp)
6226 6232 {
6227 6233 zfs_cmd_t *zc;
6228 6234 uint_t vecnum;
6229 6235 int error, rc, len;
6230 6236 minor_t minor = getminor(dev);
6231 6237 const zfs_ioc_vec_t *vec;
6232 6238 char *saved_poolname = NULL;
6233 6239 nvlist_t *innvl = NULL;
6234 6240
6235 6241 if (minor != 0 &&
6236 6242 zfsdev_get_soft_state(minor, ZSST_CTLDEV) == NULL)
6237 6243 return (zvol_ioctl(dev, cmd, arg, flag, cr, rvalp));
6238 6244
6239 6245 vecnum = cmd - ZFS_IOC_FIRST;
6240 6246 ASSERT3U(getmajor(dev), ==, ddi_driver_major(zfs_dip));
6241 6247
6242 6248 if (vecnum >= sizeof (zfs_ioc_vec) / sizeof (zfs_ioc_vec[0]))
6243 6249 return (SET_ERROR(EINVAL));
6244 6250 vec = &zfs_ioc_vec[vecnum];
6245 6251
6246 6252 zc = kmem_zalloc(sizeof (zfs_cmd_t), KM_SLEEP);
6247 6253
6248 6254 error = ddi_copyin((void *)arg, zc, sizeof (zfs_cmd_t), flag);
6249 6255 if (error != 0) {
6250 6256 error = SET_ERROR(EFAULT);
6251 6257 goto out;
6252 6258 }
6253 6259
6254 6260 zc->zc_iflags = flag & FKIOCTL;
6255 6261 if (zc->zc_nvlist_src_size != 0) {
6256 6262 error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
6257 6263 zc->zc_iflags, &innvl);
6258 6264 if (error != 0)
6259 6265 goto out;
6260 6266 }
6261 6267
6262 6268 /*
6263 6269 * Ensure that all pool/dataset names are valid before we pass down to
6264 6270 * the lower layers.
6265 6271 */
6266 6272 zc->zc_name[sizeof (zc->zc_name) - 1] = '\0';
6267 6273 switch (vec->zvec_namecheck) {
6268 6274 case POOL_NAME:
6269 6275 if (pool_namecheck(zc->zc_name, NULL, NULL) != 0)
6270 6276 error = SET_ERROR(EINVAL);
6271 6277 else
6272 6278 error = pool_status_check(zc->zc_name,
6273 6279 vec->zvec_namecheck, vec->zvec_pool_check);
6274 6280 break;
6275 6281
6276 6282 case DATASET_NAME:
6277 6283 if (dataset_namecheck(zc->zc_name, NULL, NULL) != 0)
6278 6284 error = SET_ERROR(EINVAL);
6279 6285 else
6280 6286 error = pool_status_check(zc->zc_name,
6281 6287 vec->zvec_namecheck, vec->zvec_pool_check);
6282 6288 break;
6283 6289
6284 6290 case NO_NAME:
6285 6291 break;
6286 6292 }
6287 6293
6288 6294
6289 6295 if (error == 0)
6290 6296 error = vec->zvec_secpolicy(zc, innvl, cr);
6291 6297
6292 6298 if (error != 0)
6293 6299 goto out;
6294 6300
6295 6301 /* legacy ioctls can modify zc_name */
6296 6302 len = strcspn(zc->zc_name, "/@#") + 1;
6297 6303 saved_poolname = kmem_alloc(len, KM_SLEEP);
6298 6304 (void) strlcpy(saved_poolname, zc->zc_name, len);
6299 6305
6300 6306 if (vec->zvec_func != NULL) {
6301 6307 nvlist_t *outnvl;
6302 6308 int puterror = 0;
6303 6309 spa_t *spa;
6304 6310 nvlist_t *lognv = NULL;
6305 6311
6306 6312 ASSERT(vec->zvec_legacy_func == NULL);
6307 6313
6308 6314 /*
6309 6315 * Add the innvl to the lognv before calling the func,
6310 6316 * in case the func changes the innvl.
6311 6317 */
6312 6318 if (vec->zvec_allow_log) {
6313 6319 lognv = fnvlist_alloc();
6314 6320 fnvlist_add_string(lognv, ZPOOL_HIST_IOCTL,
6315 6321 vec->zvec_name);
6316 6322 if (!nvlist_empty(innvl)) {
6317 6323 fnvlist_add_nvlist(lognv, ZPOOL_HIST_INPUT_NVL,
6318 6324 innvl);
6319 6325 }
6320 6326 }
6321 6327
6322 6328 outnvl = fnvlist_alloc();
6323 6329 error = vec->zvec_func(zc->zc_name, innvl, outnvl);
6324 6330
6325 6331 /*
6326 6332 * Some commands can partially execute, modfiy state, and still
6327 6333 * return an error. In these cases, attempt to record what
6328 6334 * was modified.
6329 6335 */
6330 6336 if ((error == 0 ||
6331 6337 (cmd == ZFS_IOC_CHANNEL_PROGRAM && error != EINVAL)) &&
6332 6338 vec->zvec_allow_log &&
6333 6339 spa_open(zc->zc_name, &spa, FTAG) == 0) {
6334 6340 if (!nvlist_empty(outnvl)) {
6335 6341 fnvlist_add_nvlist(lognv, ZPOOL_HIST_OUTPUT_NVL,
6336 6342 outnvl);
6337 6343 }
6338 6344 if (error != 0) {
6339 6345 fnvlist_add_int64(lognv, ZPOOL_HIST_ERRNO,
6340 6346 error);
6341 6347 }
6342 6348 (void) spa_history_log_nvl(spa, lognv);
6343 6349 spa_close(spa, FTAG);
6344 6350 }
6345 6351 fnvlist_free(lognv);
6346 6352
6347 6353 if (!nvlist_empty(outnvl) || zc->zc_nvlist_dst_size != 0) {
6348 6354 int smusherror = 0;
6349 6355 if (vec->zvec_smush_outnvlist) {
6350 6356 smusherror = nvlist_smush(outnvl,
6351 6357 zc->zc_nvlist_dst_size);
6352 6358 }
6353 6359 if (smusherror == 0)
6354 6360 puterror = put_nvlist(zc, outnvl);
6355 6361 }
6356 6362
6357 6363 if (puterror != 0)
6358 6364 error = puterror;
6359 6365
6360 6366 nvlist_free(outnvl);
6361 6367 } else {
6362 6368 error = vec->zvec_legacy_func(zc);
6363 6369 }
6364 6370
6365 6371 out:
6366 6372 nvlist_free(innvl);
6367 6373 rc = ddi_copyout(zc, (void *)arg, sizeof (zfs_cmd_t), flag);
6368 6374 if (error == 0 && rc != 0)
6369 6375 error = SET_ERROR(EFAULT);
6370 6376 if (error == 0 && vec->zvec_allow_log) {
6371 6377 char *s = tsd_get(zfs_allow_log_key);
6372 6378 if (s != NULL)
6373 6379 strfree(s);
6374 6380 (void) tsd_set(zfs_allow_log_key, saved_poolname);
6375 6381 } else {
6376 6382 if (saved_poolname != NULL)
6377 6383 strfree(saved_poolname);
6378 6384 }
6379 6385
6380 6386 kmem_free(zc, sizeof (zfs_cmd_t));
6381 6387 return (error);
6382 6388 }
6383 6389
6384 6390 static int
6385 6391 zfs_attach(dev_info_t *dip, ddi_attach_cmd_t cmd)
6386 6392 {
6387 6393 if (cmd != DDI_ATTACH)
6388 6394 return (DDI_FAILURE);
6389 6395
6390 6396 if (ddi_create_minor_node(dip, "zfs", S_IFCHR, 0,
6391 6397 DDI_PSEUDO, 0) == DDI_FAILURE)
6392 6398 return (DDI_FAILURE);
6393 6399
6394 6400 zfs_dip = dip;
6395 6401
6396 6402 ddi_report_dev(dip);
6397 6403
6398 6404 return (DDI_SUCCESS);
6399 6405 }
6400 6406
6401 6407 static int
6402 6408 zfs_detach(dev_info_t *dip, ddi_detach_cmd_t cmd)
6403 6409 {
6404 6410 if (spa_busy() || zfs_busy() || zvol_busy())
6405 6411 return (DDI_FAILURE);
6406 6412
6407 6413 if (cmd != DDI_DETACH)
6408 6414 return (DDI_FAILURE);
6409 6415
6410 6416 zfs_dip = NULL;
6411 6417
6412 6418 ddi_prop_remove_all(dip);
6413 6419 ddi_remove_minor_node(dip, NULL);
6414 6420
6415 6421 return (DDI_SUCCESS);
6416 6422 }
6417 6423
6418 6424 /*ARGSUSED*/
6419 6425 static int
6420 6426 zfs_info(dev_info_t *dip, ddi_info_cmd_t infocmd, void *arg, void **result)
6421 6427 {
6422 6428 switch (infocmd) {
6423 6429 case DDI_INFO_DEVT2DEVINFO:
6424 6430 *result = zfs_dip;
6425 6431 return (DDI_SUCCESS);
6426 6432
6427 6433 case DDI_INFO_DEVT2INSTANCE:
6428 6434 *result = (void *)0;
6429 6435 return (DDI_SUCCESS);
6430 6436 }
6431 6437
6432 6438 return (DDI_FAILURE);
6433 6439 }
6434 6440
6435 6441 /*
6436 6442 * OK, so this is a little weird.
6437 6443 *
6438 6444 * /dev/zfs is the control node, i.e. minor 0.
6439 6445 * /dev/zvol/[r]dsk/pool/dataset are the zvols, minor > 0.
6440 6446 *
6441 6447 * /dev/zfs has basically nothing to do except serve up ioctls,
6442 6448 * so most of the standard driver entry points are in zvol.c.
6443 6449 */
6444 6450 static struct cb_ops zfs_cb_ops = {
6445 6451 zfsdev_open, /* open */
6446 6452 zfsdev_close, /* close */
6447 6453 zvol_strategy, /* strategy */
6448 6454 nodev, /* print */
6449 6455 zvol_dump, /* dump */
6450 6456 zvol_read, /* read */
6451 6457 zvol_write, /* write */
6452 6458 zfsdev_ioctl, /* ioctl */
6453 6459 nodev, /* devmap */
6454 6460 nodev, /* mmap */
6455 6461 nodev, /* segmap */
6456 6462 nochpoll, /* poll */
6457 6463 ddi_prop_op, /* prop_op */
6458 6464 NULL, /* streamtab */
6459 6465 D_NEW | D_MP | D_64BIT, /* Driver compatibility flag */
6460 6466 CB_REV, /* version */
6461 6467 nodev, /* async read */
6462 6468 nodev, /* async write */
6463 6469 };
6464 6470
6465 6471 static struct dev_ops zfs_dev_ops = {
6466 6472 DEVO_REV, /* version */
6467 6473 0, /* refcnt */
6468 6474 zfs_info, /* info */
6469 6475 nulldev, /* identify */
6470 6476 nulldev, /* probe */
6471 6477 zfs_attach, /* attach */
6472 6478 zfs_detach, /* detach */
6473 6479 nodev, /* reset */
6474 6480 &zfs_cb_ops, /* driver operations */
6475 6481 NULL, /* no bus operations */
6476 6482 NULL, /* power */
6477 6483 ddi_quiesce_not_needed, /* quiesce */
6478 6484 };
6479 6485
6480 6486 static struct modldrv zfs_modldrv = {
6481 6487 &mod_driverops,
6482 6488 "ZFS storage pool",
6483 6489 &zfs_dev_ops
6484 6490 };
6485 6491
6486 6492 static struct modlinkage modlinkage = {
6487 6493 MODREV_1,
6488 6494 (void *)&zfs_modlfs,
6489 6495 (void *)&zfs_modldrv,
6490 6496 NULL
6491 6497 };
6492 6498
6493 6499 static void
6494 6500 zfs_allow_log_destroy(void *arg)
6495 6501 {
6496 6502 char *poolname = arg;
6497 6503 strfree(poolname);
6498 6504 }
6499 6505
6500 6506 int
6501 6507 _init(void)
6502 6508 {
6503 6509 int error;
6504 6510
6505 6511 spa_init(FREAD | FWRITE);
6506 6512 zfs_init();
6507 6513 zvol_init();
6508 6514 zfs_ioctl_init();
6509 6515
6510 6516 if ((error = mod_install(&modlinkage)) != 0) {
6511 6517 zvol_fini();
6512 6518 zfs_fini();
6513 6519 spa_fini();
6514 6520 return (error);
6515 6521 }
6516 6522
6517 6523 tsd_create(&zfs_fsyncer_key, NULL);
6518 6524 tsd_create(&rrw_tsd_key, rrw_tsd_destroy);
6519 6525 tsd_create(&zfs_allow_log_key, zfs_allow_log_destroy);
6520 6526
6521 6527 error = ldi_ident_from_mod(&modlinkage, &zfs_li);
6522 6528 ASSERT(error == 0);
6523 6529 mutex_init(&zfs_share_lock, NULL, MUTEX_DEFAULT, NULL);
6524 6530
6525 6531 return (0);
6526 6532 }
6527 6533
6528 6534 int
6529 6535 _fini(void)
6530 6536 {
6531 6537 int error;
6532 6538
6533 6539 if (spa_busy() || zfs_busy() || zvol_busy() || zio_injection_enabled)
6534 6540 return (SET_ERROR(EBUSY));
6535 6541
6536 6542 if ((error = mod_remove(&modlinkage)) != 0)
6537 6543 return (error);
6538 6544
6539 6545 zvol_fini();
6540 6546 zfs_fini();
6541 6547 spa_fini();
6542 6548 if (zfs_nfsshare_inited)
6543 6549 (void) ddi_modclose(nfs_mod);
6544 6550 if (zfs_smbshare_inited)
6545 6551 (void) ddi_modclose(smbsrv_mod);
6546 6552 if (zfs_nfsshare_inited || zfs_smbshare_inited)
6547 6553 (void) ddi_modclose(sharefs_mod);
6548 6554
6549 6555 tsd_destroy(&zfs_fsyncer_key);
6550 6556 ldi_ident_release(zfs_li);
6551 6557 zfs_li = NULL;
6552 6558 mutex_destroy(&zfs_share_lock);
6553 6559
6554 6560 return (error);
6555 6561 }
6556 6562
6557 6563 int
6558 6564 _info(struct modinfo *modinfop)
6559 6565 {
6560 6566 return (mod_info(&modlinkage, modinfop));
6561 6567 }
|
↓ open down ↓ |
5010 lines elided |
↑ open up ↑ |
XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX