Print this page
Optimize creation and removal of temporary "user holds" placed on
snapshots by a zfs send, by ensuring all the required holds and
releases are done in a single dsl_sync_task.
Creation now collates the required holds during a dry run and
then uses a single lzc_hold call via zfs_hold_apply instead of
processing each snapshot in turn.
Defered (on exit) cleanup by the kernel is also now done in
dsl_sync_task by reusing dsl_dataset_user_release.
On a test with 11 volumes in a tree each with 8 snapshots on a
single HDD zpool this reduces the time required to perform a full
send from 20 seconds to under 0.8 seconds.
For reference eliminating the hold entirely reduces this 0.15
seconds.
While I'm here:-
* Remove some unused structures
* Fix nvlist_t leak in zfs_release_one
| 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 * Portions Copyright 2011 Martin Matuska
25 25 * Copyright 2011 Nexenta Systems, Inc. All rights reserved.
26 26 * Copyright (c) 2012, Joyent, Inc. All rights reserved.
27 27 * Copyright (c) 2013 by Delphix. All rights reserved.
28 28 * Copyright (c) 2013 by Saso Kiselkov. All rights reserved.
29 29 */
30 30
31 31 /*
32 32 * ZFS ioctls.
33 33 *
34 34 * This file handles the ioctls to /dev/zfs, used for configuring ZFS storage
35 35 * pools and filesystems, e.g. with /sbin/zfs and /sbin/zpool.
36 36 *
37 37 * There are two ways that we handle ioctls: the legacy way where almost
38 38 * all of the logic is in the ioctl callback, and the new way where most
39 39 * of the marshalling is handled in the common entry point, zfsdev_ioctl().
40 40 *
41 41 * Non-legacy ioctls should be registered by calling
42 42 * zfs_ioctl_register() from zfs_ioctl_init(). The ioctl is invoked
43 43 * from userland by lzc_ioctl().
44 44 *
45 45 * The registration arguments are as follows:
46 46 *
47 47 * const char *name
48 48 * The name of the ioctl. This is used for history logging. If the
49 49 * ioctl returns successfully (the callback returns 0), and allow_log
50 50 * is true, then a history log entry will be recorded with the input &
51 51 * output nvlists. The log entry can be printed with "zpool history -i".
52 52 *
53 53 * zfs_ioc_t ioc
54 54 * The ioctl request number, which userland will pass to ioctl(2).
55 55 * The ioctl numbers can change from release to release, because
56 56 * the caller (libzfs) must be matched to the kernel.
57 57 *
58 58 * zfs_secpolicy_func_t *secpolicy
59 59 * This function will be called before the zfs_ioc_func_t, to
60 60 * determine if this operation is permitted. It should return EPERM
61 61 * on failure, and 0 on success. Checks include determining if the
62 62 * dataset is visible in this zone, and if the user has either all
63 63 * zfs privileges in the zone (SYS_MOUNT), or has been granted permission
64 64 * to do this operation on this dataset with "zfs allow".
65 65 *
66 66 * zfs_ioc_namecheck_t namecheck
67 67 * This specifies what to expect in the zfs_cmd_t:zc_name -- a pool
68 68 * name, a dataset name, or nothing. If the name is not well-formed,
69 69 * the ioctl will fail and the callback will not be called.
70 70 * Therefore, the callback can assume that the name is well-formed
71 71 * (e.g. is null-terminated, doesn't have more than one '@' character,
72 72 * doesn't have invalid characters).
73 73 *
74 74 * zfs_ioc_poolcheck_t pool_check
75 75 * This specifies requirements on the pool state. If the pool does
76 76 * not meet them (is suspended or is readonly), the ioctl will fail
77 77 * and the callback will not be called. If any checks are specified
78 78 * (i.e. it is not POOL_CHECK_NONE), namecheck must not be NO_NAME.
79 79 * Multiple checks can be or-ed together (e.g. POOL_CHECK_SUSPENDED |
80 80 * POOL_CHECK_READONLY).
81 81 *
82 82 * boolean_t smush_outnvlist
83 83 * If smush_outnvlist is true, then the output is presumed to be a
84 84 * list of errors, and it will be "smushed" down to fit into the
85 85 * caller's buffer, by removing some entries and replacing them with a
86 86 * single "N_MORE_ERRORS" entry indicating how many were removed. See
87 87 * nvlist_smush() for details. If smush_outnvlist is false, and the
88 88 * outnvlist does not fit into the userland-provided buffer, then the
89 89 * ioctl will fail with ENOMEM.
90 90 *
91 91 * zfs_ioc_func_t *func
92 92 * The callback function that will perform the operation.
93 93 *
94 94 * The callback should return 0 on success, or an error number on
95 95 * failure. If the function fails, the userland ioctl will return -1,
96 96 * and errno will be set to the callback's return value. The callback
97 97 * will be called with the following arguments:
98 98 *
99 99 * const char *name
100 100 * The name of the pool or dataset to operate on, from
101 101 * zfs_cmd_t:zc_name. The 'namecheck' argument specifies the
102 102 * expected type (pool, dataset, or none).
103 103 *
104 104 * nvlist_t *innvl
105 105 * The input nvlist, deserialized from zfs_cmd_t:zc_nvlist_src. Or
106 106 * NULL if no input nvlist was provided. Changes to this nvlist are
107 107 * ignored. If the input nvlist could not be deserialized, the
108 108 * ioctl will fail and the callback will not be called.
109 109 *
110 110 * nvlist_t *outnvl
111 111 * The output nvlist, initially empty. The callback can fill it in,
112 112 * and it will be returned to userland by serializing it into
113 113 * zfs_cmd_t:zc_nvlist_dst. If it is non-empty, and serialization
114 114 * fails (e.g. because the caller didn't supply a large enough
115 115 * buffer), then the overall ioctl will fail. See the
116 116 * 'smush_nvlist' argument above for additional behaviors.
117 117 *
118 118 * There are two typical uses of the output nvlist:
119 119 * - To return state, e.g. property values. In this case,
120 120 * smush_outnvlist should be false. If the buffer was not large
121 121 * enough, the caller will reallocate a larger buffer and try
122 122 * the ioctl again.
123 123 *
124 124 * - To return multiple errors from an ioctl which makes on-disk
125 125 * changes. In this case, smush_outnvlist should be true.
126 126 * Ioctls which make on-disk modifications should generally not
127 127 * use the outnvl if they succeed, because the caller can not
128 128 * distinguish between the operation failing, and
129 129 * deserialization failing.
130 130 */
131 131
132 132 #include <sys/types.h>
133 133 #include <sys/param.h>
134 134 #include <sys/errno.h>
135 135 #include <sys/uio.h>
136 136 #include <sys/buf.h>
137 137 #include <sys/modctl.h>
138 138 #include <sys/open.h>
139 139 #include <sys/file.h>
140 140 #include <sys/kmem.h>
141 141 #include <sys/conf.h>
142 142 #include <sys/cmn_err.h>
143 143 #include <sys/stat.h>
144 144 #include <sys/zfs_ioctl.h>
145 145 #include <sys/zfs_vfsops.h>
146 146 #include <sys/zfs_znode.h>
147 147 #include <sys/zap.h>
148 148 #include <sys/spa.h>
149 149 #include <sys/spa_impl.h>
150 150 #include <sys/vdev.h>
151 151 #include <sys/priv_impl.h>
152 152 #include <sys/dmu.h>
153 153 #include <sys/dsl_dir.h>
154 154 #include <sys/dsl_dataset.h>
155 155 #include <sys/dsl_prop.h>
156 156 #include <sys/dsl_deleg.h>
157 157 #include <sys/dmu_objset.h>
158 158 #include <sys/dmu_impl.h>
159 159 #include <sys/dmu_tx.h>
160 160 #include <sys/ddi.h>
161 161 #include <sys/sunddi.h>
162 162 #include <sys/sunldi.h>
163 163 #include <sys/policy.h>
164 164 #include <sys/zone.h>
165 165 #include <sys/nvpair.h>
166 166 #include <sys/pathname.h>
167 167 #include <sys/mount.h>
168 168 #include <sys/sdt.h>
169 169 #include <sys/fs/zfs.h>
170 170 #include <sys/zfs_ctldir.h>
171 171 #include <sys/zfs_dir.h>
172 172 #include <sys/zfs_onexit.h>
173 173 #include <sys/zvol.h>
174 174 #include <sys/dsl_scan.h>
175 175 #include <sharefs/share.h>
176 176 #include <sys/dmu_objset.h>
177 177 #include <sys/dmu_send.h>
178 178 #include <sys/dsl_destroy.h>
179 179 #include <sys/dsl_userhold.h>
180 180 #include <sys/zfeature.h>
181 181
182 182 #include "zfs_namecheck.h"
183 183 #include "zfs_prop.h"
184 184 #include "zfs_deleg.h"
185 185 #include "zfs_comutil.h"
186 186
187 187 extern struct modlfs zfs_modlfs;
188 188
189 189 extern void zfs_init(void);
190 190 extern void zfs_fini(void);
191 191
192 192 ldi_ident_t zfs_li = NULL;
193 193 dev_info_t *zfs_dip;
194 194
195 195 uint_t zfs_fsyncer_key;
196 196 extern uint_t rrw_tsd_key;
197 197 static uint_t zfs_allow_log_key;
198 198
199 199 typedef int zfs_ioc_legacy_func_t(zfs_cmd_t *);
200 200 typedef int zfs_ioc_func_t(const char *, nvlist_t *, nvlist_t *);
201 201 typedef int zfs_secpolicy_func_t(zfs_cmd_t *, nvlist_t *, cred_t *);
202 202
203 203 typedef enum {
204 204 NO_NAME,
205 205 POOL_NAME,
206 206 DATASET_NAME
207 207 } zfs_ioc_namecheck_t;
208 208
209 209 typedef enum {
210 210 POOL_CHECK_NONE = 1 << 0,
211 211 POOL_CHECK_SUSPENDED = 1 << 1,
212 212 POOL_CHECK_READONLY = 1 << 2,
213 213 } zfs_ioc_poolcheck_t;
214 214
215 215 typedef struct zfs_ioc_vec {
216 216 zfs_ioc_legacy_func_t *zvec_legacy_func;
217 217 zfs_ioc_func_t *zvec_func;
218 218 zfs_secpolicy_func_t *zvec_secpolicy;
219 219 zfs_ioc_namecheck_t zvec_namecheck;
220 220 boolean_t zvec_allow_log;
221 221 zfs_ioc_poolcheck_t zvec_pool_check;
222 222 boolean_t zvec_smush_outnvlist;
223 223 const char *zvec_name;
224 224 } zfs_ioc_vec_t;
225 225
226 226 /* This array is indexed by zfs_userquota_prop_t */
227 227 static const char *userquota_perms[] = {
228 228 ZFS_DELEG_PERM_USERUSED,
229 229 ZFS_DELEG_PERM_USERQUOTA,
230 230 ZFS_DELEG_PERM_GROUPUSED,
231 231 ZFS_DELEG_PERM_GROUPQUOTA,
232 232 };
233 233
234 234 static int zfs_ioc_userspace_upgrade(zfs_cmd_t *zc);
235 235 static int zfs_check_settable(const char *name, nvpair_t *property,
236 236 cred_t *cr);
237 237 static int zfs_check_clearable(char *dataset, nvlist_t *props,
238 238 nvlist_t **errors);
239 239 static int zfs_fill_zplprops_root(uint64_t, nvlist_t *, nvlist_t *,
240 240 boolean_t *);
241 241 int zfs_set_prop_nvlist(const char *, zprop_source_t, nvlist_t *, nvlist_t *);
242 242 static int get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp);
243 243
244 244 static int zfs_prop_activate_feature(spa_t *spa, zfeature_info_t *feature);
245 245
246 246 /* _NOTE(PRINTFLIKE(4)) - this is printf-like, but lint is too whiney */
247 247 void
248 248 __dprintf(const char *file, const char *func, int line, const char *fmt, ...)
249 249 {
250 250 const char *newfile;
251 251 char buf[512];
252 252 va_list adx;
253 253
254 254 /*
255 255 * Get rid of annoying "../common/" prefix to filename.
256 256 */
257 257 newfile = strrchr(file, '/');
258 258 if (newfile != NULL) {
259 259 newfile = newfile + 1; /* Get rid of leading / */
260 260 } else {
261 261 newfile = file;
262 262 }
263 263
264 264 va_start(adx, fmt);
265 265 (void) vsnprintf(buf, sizeof (buf), fmt, adx);
266 266 va_end(adx);
267 267
268 268 /*
269 269 * To get this data, use the zfs-dprintf probe as so:
270 270 * dtrace -q -n 'zfs-dprintf \
271 271 * /stringof(arg0) == "dbuf.c"/ \
272 272 * {printf("%s: %s", stringof(arg1), stringof(arg3))}'
273 273 * arg0 = file name
274 274 * arg1 = function name
275 275 * arg2 = line number
276 276 * arg3 = message
277 277 */
278 278 DTRACE_PROBE4(zfs__dprintf,
279 279 char *, newfile, char *, func, int, line, char *, buf);
280 280 }
281 281
282 282 static void
283 283 history_str_free(char *buf)
284 284 {
285 285 kmem_free(buf, HIS_MAX_RECORD_LEN);
286 286 }
287 287
288 288 static char *
289 289 history_str_get(zfs_cmd_t *zc)
290 290 {
291 291 char *buf;
292 292
293 293 if (zc->zc_history == NULL)
294 294 return (NULL);
295 295
296 296 buf = kmem_alloc(HIS_MAX_RECORD_LEN, KM_SLEEP);
297 297 if (copyinstr((void *)(uintptr_t)zc->zc_history,
298 298 buf, HIS_MAX_RECORD_LEN, NULL) != 0) {
299 299 history_str_free(buf);
300 300 return (NULL);
301 301 }
302 302
303 303 buf[HIS_MAX_RECORD_LEN -1] = '\0';
304 304
305 305 return (buf);
306 306 }
307 307
308 308 /*
309 309 * Check to see if the named dataset is currently defined as bootable
310 310 */
311 311 static boolean_t
312 312 zfs_is_bootfs(const char *name)
313 313 {
314 314 objset_t *os;
315 315
316 316 if (dmu_objset_hold(name, FTAG, &os) == 0) {
317 317 boolean_t ret;
318 318 ret = (dmu_objset_id(os) == spa_bootfs(dmu_objset_spa(os)));
319 319 dmu_objset_rele(os, FTAG);
320 320 return (ret);
321 321 }
322 322 return (B_FALSE);
323 323 }
324 324
325 325 /*
326 326 * zfs_earlier_version
327 327 *
328 328 * Return non-zero if the spa version is less than requested version.
329 329 */
330 330 static int
331 331 zfs_earlier_version(const char *name, int version)
332 332 {
333 333 spa_t *spa;
334 334
335 335 if (spa_open(name, &spa, FTAG) == 0) {
336 336 if (spa_version(spa) < version) {
337 337 spa_close(spa, FTAG);
338 338 return (1);
339 339 }
340 340 spa_close(spa, FTAG);
341 341 }
342 342 return (0);
343 343 }
344 344
345 345 /*
346 346 * zpl_earlier_version
347 347 *
348 348 * Return TRUE if the ZPL version is less than requested version.
349 349 */
350 350 static boolean_t
351 351 zpl_earlier_version(const char *name, int version)
352 352 {
353 353 objset_t *os;
354 354 boolean_t rc = B_TRUE;
355 355
356 356 if (dmu_objset_hold(name, FTAG, &os) == 0) {
357 357 uint64_t zplversion;
358 358
359 359 if (dmu_objset_type(os) != DMU_OST_ZFS) {
360 360 dmu_objset_rele(os, FTAG);
361 361 return (B_TRUE);
362 362 }
363 363 /* XXX reading from non-owned objset */
364 364 if (zfs_get_zplprop(os, ZFS_PROP_VERSION, &zplversion) == 0)
365 365 rc = zplversion < version;
366 366 dmu_objset_rele(os, FTAG);
367 367 }
368 368 return (rc);
369 369 }
370 370
371 371 static void
372 372 zfs_log_history(zfs_cmd_t *zc)
373 373 {
374 374 spa_t *spa;
375 375 char *buf;
376 376
377 377 if ((buf = history_str_get(zc)) == NULL)
378 378 return;
379 379
380 380 if (spa_open(zc->zc_name, &spa, FTAG) == 0) {
381 381 if (spa_version(spa) >= SPA_VERSION_ZPOOL_HISTORY)
382 382 (void) spa_history_log(spa, buf);
383 383 spa_close(spa, FTAG);
384 384 }
385 385 history_str_free(buf);
386 386 }
387 387
388 388 /*
389 389 * Policy for top-level read operations (list pools). Requires no privileges,
390 390 * and can be used in the local zone, as there is no associated dataset.
391 391 */
392 392 /* ARGSUSED */
393 393 static int
394 394 zfs_secpolicy_none(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
395 395 {
396 396 return (0);
397 397 }
398 398
399 399 /*
400 400 * Policy for dataset read operations (list children, get statistics). Requires
401 401 * no privileges, but must be visible in the local zone.
402 402 */
403 403 /* ARGSUSED */
404 404 static int
405 405 zfs_secpolicy_read(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
406 406 {
407 407 if (INGLOBALZONE(curproc) ||
408 408 zone_dataset_visible(zc->zc_name, NULL))
409 409 return (0);
410 410
411 411 return (SET_ERROR(ENOENT));
412 412 }
413 413
414 414 static int
415 415 zfs_dozonecheck_impl(const char *dataset, uint64_t zoned, cred_t *cr)
416 416 {
417 417 int writable = 1;
418 418
419 419 /*
420 420 * The dataset must be visible by this zone -- check this first
421 421 * so they don't see EPERM on something they shouldn't know about.
422 422 */
423 423 if (!INGLOBALZONE(curproc) &&
424 424 !zone_dataset_visible(dataset, &writable))
425 425 return (SET_ERROR(ENOENT));
426 426
427 427 if (INGLOBALZONE(curproc)) {
428 428 /*
429 429 * If the fs is zoned, only root can access it from the
430 430 * global zone.
431 431 */
432 432 if (secpolicy_zfs(cr) && zoned)
433 433 return (SET_ERROR(EPERM));
434 434 } else {
435 435 /*
436 436 * If we are in a local zone, the 'zoned' property must be set.
437 437 */
438 438 if (!zoned)
439 439 return (SET_ERROR(EPERM));
440 440
441 441 /* must be writable by this zone */
442 442 if (!writable)
443 443 return (SET_ERROR(EPERM));
444 444 }
445 445 return (0);
446 446 }
447 447
448 448 static int
449 449 zfs_dozonecheck(const char *dataset, cred_t *cr)
450 450 {
451 451 uint64_t zoned;
452 452
453 453 if (dsl_prop_get_integer(dataset, "zoned", &zoned, NULL))
454 454 return (SET_ERROR(ENOENT));
455 455
456 456 return (zfs_dozonecheck_impl(dataset, zoned, cr));
457 457 }
458 458
459 459 static int
460 460 zfs_dozonecheck_ds(const char *dataset, dsl_dataset_t *ds, cred_t *cr)
461 461 {
462 462 uint64_t zoned;
463 463
464 464 if (dsl_prop_get_int_ds(ds, "zoned", &zoned))
465 465 return (SET_ERROR(ENOENT));
466 466
467 467 return (zfs_dozonecheck_impl(dataset, zoned, cr));
468 468 }
469 469
470 470 static int
471 471 zfs_secpolicy_write_perms_ds(const char *name, dsl_dataset_t *ds,
472 472 const char *perm, cred_t *cr)
473 473 {
474 474 int error;
475 475
476 476 error = zfs_dozonecheck_ds(name, ds, cr);
477 477 if (error == 0) {
478 478 error = secpolicy_zfs(cr);
479 479 if (error != 0)
480 480 error = dsl_deleg_access_impl(ds, perm, cr);
481 481 }
482 482 return (error);
483 483 }
484 484
485 485 static int
486 486 zfs_secpolicy_write_perms(const char *name, const char *perm, cred_t *cr)
487 487 {
488 488 int error;
489 489 dsl_dataset_t *ds;
490 490 dsl_pool_t *dp;
491 491
492 492 error = dsl_pool_hold(name, FTAG, &dp);
493 493 if (error != 0)
494 494 return (error);
495 495
496 496 error = dsl_dataset_hold(dp, name, FTAG, &ds);
497 497 if (error != 0) {
498 498 dsl_pool_rele(dp, FTAG);
499 499 return (error);
500 500 }
501 501
502 502 error = zfs_secpolicy_write_perms_ds(name, ds, perm, cr);
503 503
504 504 dsl_dataset_rele(ds, FTAG);
505 505 dsl_pool_rele(dp, FTAG);
506 506 return (error);
507 507 }
508 508
509 509 /*
510 510 * Policy for setting the security label property.
511 511 *
512 512 * Returns 0 for success, non-zero for access and other errors.
513 513 */
514 514 static int
515 515 zfs_set_slabel_policy(const char *name, char *strval, cred_t *cr)
516 516 {
517 517 char ds_hexsl[MAXNAMELEN];
518 518 bslabel_t ds_sl, new_sl;
519 519 boolean_t new_default = FALSE;
520 520 uint64_t zoned;
521 521 int needed_priv = -1;
522 522 int error;
523 523
524 524 /* First get the existing dataset label. */
525 525 error = dsl_prop_get(name, zfs_prop_to_name(ZFS_PROP_MLSLABEL),
526 526 1, sizeof (ds_hexsl), &ds_hexsl, NULL);
527 527 if (error != 0)
528 528 return (SET_ERROR(EPERM));
529 529
530 530 if (strcasecmp(strval, ZFS_MLSLABEL_DEFAULT) == 0)
531 531 new_default = TRUE;
532 532
533 533 /* The label must be translatable */
534 534 if (!new_default && (hexstr_to_label(strval, &new_sl) != 0))
535 535 return (SET_ERROR(EINVAL));
536 536
537 537 /*
538 538 * In a non-global zone, disallow attempts to set a label that
539 539 * doesn't match that of the zone; otherwise no other checks
540 540 * are needed.
541 541 */
542 542 if (!INGLOBALZONE(curproc)) {
543 543 if (new_default || !blequal(&new_sl, CR_SL(CRED())))
544 544 return (SET_ERROR(EPERM));
545 545 return (0);
546 546 }
547 547
548 548 /*
549 549 * For global-zone datasets (i.e., those whose zoned property is
550 550 * "off", verify that the specified new label is valid for the
551 551 * global zone.
552 552 */
553 553 if (dsl_prop_get_integer(name,
554 554 zfs_prop_to_name(ZFS_PROP_ZONED), &zoned, NULL))
555 555 return (SET_ERROR(EPERM));
556 556 if (!zoned) {
557 557 if (zfs_check_global_label(name, strval) != 0)
558 558 return (SET_ERROR(EPERM));
559 559 }
560 560
561 561 /*
562 562 * If the existing dataset label is nondefault, check if the
563 563 * dataset is mounted (label cannot be changed while mounted).
564 564 * Get the zfsvfs; if there isn't one, then the dataset isn't
565 565 * mounted (or isn't a dataset, doesn't exist, ...).
566 566 */
567 567 if (strcasecmp(ds_hexsl, ZFS_MLSLABEL_DEFAULT) != 0) {
568 568 objset_t *os;
569 569 static char *setsl_tag = "setsl_tag";
570 570
571 571 /*
572 572 * Try to own the dataset; abort if there is any error,
573 573 * (e.g., already mounted, in use, or other error).
574 574 */
575 575 error = dmu_objset_own(name, DMU_OST_ZFS, B_TRUE,
576 576 setsl_tag, &os);
577 577 if (error != 0)
578 578 return (SET_ERROR(EPERM));
579 579
580 580 dmu_objset_disown(os, setsl_tag);
581 581
582 582 if (new_default) {
583 583 needed_priv = PRIV_FILE_DOWNGRADE_SL;
584 584 goto out_check;
585 585 }
586 586
587 587 if (hexstr_to_label(strval, &new_sl) != 0)
588 588 return (SET_ERROR(EPERM));
589 589
590 590 if (blstrictdom(&ds_sl, &new_sl))
591 591 needed_priv = PRIV_FILE_DOWNGRADE_SL;
592 592 else if (blstrictdom(&new_sl, &ds_sl))
593 593 needed_priv = PRIV_FILE_UPGRADE_SL;
594 594 } else {
595 595 /* dataset currently has a default label */
596 596 if (!new_default)
597 597 needed_priv = PRIV_FILE_UPGRADE_SL;
598 598 }
599 599
600 600 out_check:
601 601 if (needed_priv != -1)
602 602 return (PRIV_POLICY(cr, needed_priv, B_FALSE, EPERM, NULL));
603 603 return (0);
604 604 }
605 605
606 606 static int
607 607 zfs_secpolicy_setprop(const char *dsname, zfs_prop_t prop, nvpair_t *propval,
608 608 cred_t *cr)
609 609 {
610 610 char *strval;
611 611
612 612 /*
613 613 * Check permissions for special properties.
614 614 */
615 615 switch (prop) {
616 616 case ZFS_PROP_ZONED:
617 617 /*
618 618 * Disallow setting of 'zoned' from within a local zone.
619 619 */
620 620 if (!INGLOBALZONE(curproc))
621 621 return (SET_ERROR(EPERM));
622 622 break;
623 623
624 624 case ZFS_PROP_QUOTA:
625 625 if (!INGLOBALZONE(curproc)) {
626 626 uint64_t zoned;
627 627 char setpoint[MAXNAMELEN];
628 628 /*
629 629 * Unprivileged users are allowed to modify the
630 630 * quota on things *under* (ie. contained by)
631 631 * the thing they own.
632 632 */
633 633 if (dsl_prop_get_integer(dsname, "zoned", &zoned,
634 634 setpoint))
635 635 return (SET_ERROR(EPERM));
636 636 if (!zoned || strlen(dsname) <= strlen(setpoint))
637 637 return (SET_ERROR(EPERM));
638 638 }
639 639 break;
640 640
641 641 case ZFS_PROP_MLSLABEL:
642 642 if (!is_system_labeled())
643 643 return (SET_ERROR(EPERM));
644 644
645 645 if (nvpair_value_string(propval, &strval) == 0) {
646 646 int err;
647 647
648 648 err = zfs_set_slabel_policy(dsname, strval, CRED());
649 649 if (err != 0)
650 650 return (err);
651 651 }
652 652 break;
653 653 }
654 654
655 655 return (zfs_secpolicy_write_perms(dsname, zfs_prop_to_name(prop), cr));
656 656 }
657 657
658 658 /* ARGSUSED */
659 659 static int
660 660 zfs_secpolicy_set_fsacl(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
661 661 {
662 662 int error;
663 663
664 664 error = zfs_dozonecheck(zc->zc_name, cr);
665 665 if (error != 0)
666 666 return (error);
667 667
668 668 /*
669 669 * permission to set permissions will be evaluated later in
670 670 * dsl_deleg_can_allow()
671 671 */
672 672 return (0);
673 673 }
674 674
675 675 /* ARGSUSED */
676 676 static int
677 677 zfs_secpolicy_rollback(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
678 678 {
679 679 return (zfs_secpolicy_write_perms(zc->zc_name,
680 680 ZFS_DELEG_PERM_ROLLBACK, cr));
681 681 }
682 682
683 683 /* ARGSUSED */
684 684 static int
685 685 zfs_secpolicy_send(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
686 686 {
687 687 dsl_pool_t *dp;
688 688 dsl_dataset_t *ds;
689 689 char *cp;
690 690 int error;
691 691
692 692 /*
693 693 * Generate the current snapshot name from the given objsetid, then
694 694 * use that name for the secpolicy/zone checks.
695 695 */
696 696 cp = strchr(zc->zc_name, '@');
697 697 if (cp == NULL)
698 698 return (SET_ERROR(EINVAL));
699 699 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
700 700 if (error != 0)
701 701 return (error);
702 702
703 703 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &ds);
704 704 if (error != 0) {
705 705 dsl_pool_rele(dp, FTAG);
706 706 return (error);
707 707 }
708 708
709 709 dsl_dataset_name(ds, zc->zc_name);
710 710
711 711 error = zfs_secpolicy_write_perms_ds(zc->zc_name, ds,
712 712 ZFS_DELEG_PERM_SEND, cr);
713 713 dsl_dataset_rele(ds, FTAG);
714 714 dsl_pool_rele(dp, FTAG);
715 715
716 716 return (error);
717 717 }
718 718
719 719 /* ARGSUSED */
720 720 static int
721 721 zfs_secpolicy_send_new(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
722 722 {
723 723 return (zfs_secpolicy_write_perms(zc->zc_name,
724 724 ZFS_DELEG_PERM_SEND, cr));
725 725 }
726 726
727 727 /* ARGSUSED */
728 728 static int
729 729 zfs_secpolicy_deleg_share(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
730 730 {
731 731 vnode_t *vp;
732 732 int error;
733 733
734 734 if ((error = lookupname(zc->zc_value, UIO_SYSSPACE,
735 735 NO_FOLLOW, NULL, &vp)) != 0)
736 736 return (error);
737 737
738 738 /* Now make sure mntpnt and dataset are ZFS */
739 739
740 740 if (vp->v_vfsp->vfs_fstype != zfsfstype ||
741 741 (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource),
742 742 zc->zc_name) != 0)) {
743 743 VN_RELE(vp);
744 744 return (SET_ERROR(EPERM));
745 745 }
746 746
747 747 VN_RELE(vp);
748 748 return (dsl_deleg_access(zc->zc_name,
749 749 ZFS_DELEG_PERM_SHARE, cr));
750 750 }
751 751
752 752 int
753 753 zfs_secpolicy_share(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
754 754 {
755 755 if (!INGLOBALZONE(curproc))
756 756 return (SET_ERROR(EPERM));
757 757
758 758 if (secpolicy_nfs(cr) == 0) {
759 759 return (0);
760 760 } else {
761 761 return (zfs_secpolicy_deleg_share(zc, innvl, cr));
762 762 }
763 763 }
764 764
765 765 int
766 766 zfs_secpolicy_smb_acl(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
767 767 {
768 768 if (!INGLOBALZONE(curproc))
769 769 return (SET_ERROR(EPERM));
770 770
771 771 if (secpolicy_smb(cr) == 0) {
772 772 return (0);
773 773 } else {
774 774 return (zfs_secpolicy_deleg_share(zc, innvl, cr));
775 775 }
776 776 }
777 777
778 778 static int
779 779 zfs_get_parent(const char *datasetname, char *parent, int parentsize)
780 780 {
781 781 char *cp;
782 782
783 783 /*
784 784 * Remove the @bla or /bla from the end of the name to get the parent.
785 785 */
786 786 (void) strncpy(parent, datasetname, parentsize);
787 787 cp = strrchr(parent, '@');
788 788 if (cp != NULL) {
789 789 cp[0] = '\0';
790 790 } else {
791 791 cp = strrchr(parent, '/');
792 792 if (cp == NULL)
793 793 return (SET_ERROR(ENOENT));
794 794 cp[0] = '\0';
795 795 }
796 796
797 797 return (0);
798 798 }
799 799
800 800 int
801 801 zfs_secpolicy_destroy_perms(const char *name, cred_t *cr)
802 802 {
803 803 int error;
804 804
805 805 if ((error = zfs_secpolicy_write_perms(name,
806 806 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
807 807 return (error);
808 808
809 809 return (zfs_secpolicy_write_perms(name, ZFS_DELEG_PERM_DESTROY, cr));
810 810 }
811 811
812 812 /* ARGSUSED */
813 813 static int
814 814 zfs_secpolicy_destroy(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
815 815 {
816 816 return (zfs_secpolicy_destroy_perms(zc->zc_name, cr));
817 817 }
818 818
819 819 /*
820 820 * Destroying snapshots with delegated permissions requires
821 821 * descendant mount and destroy permissions.
822 822 */
823 823 /* ARGSUSED */
824 824 static int
825 825 zfs_secpolicy_destroy_snaps(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
826 826 {
827 827 nvlist_t *snaps;
828 828 nvpair_t *pair, *nextpair;
829 829 int error = 0;
830 830
831 831 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
832 832 return (SET_ERROR(EINVAL));
833 833 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
834 834 pair = nextpair) {
835 835 dsl_pool_t *dp;
836 836 dsl_dataset_t *ds;
837 837
838 838 error = dsl_pool_hold(nvpair_name(pair), FTAG, &dp);
839 839 if (error != 0)
840 840 break;
841 841 nextpair = nvlist_next_nvpair(snaps, pair);
842 842 error = dsl_dataset_hold(dp, nvpair_name(pair), FTAG, &ds);
843 843 if (error == 0)
844 844 dsl_dataset_rele(ds, FTAG);
845 845 dsl_pool_rele(dp, FTAG);
846 846
847 847 if (error == 0) {
848 848 error = zfs_secpolicy_destroy_perms(nvpair_name(pair),
849 849 cr);
850 850 } else if (error == ENOENT) {
851 851 /*
852 852 * Ignore any snapshots that don't exist (we consider
853 853 * them "already destroyed"). Remove the name from the
854 854 * nvl here in case the snapshot is created between
855 855 * now and when we try to destroy it (in which case
856 856 * we don't want to destroy it since we haven't
857 857 * checked for permission).
858 858 */
859 859 fnvlist_remove_nvpair(snaps, pair);
860 860 error = 0;
861 861 }
862 862 if (error != 0)
863 863 break;
864 864 }
865 865
866 866 return (error);
867 867 }
868 868
869 869 int
870 870 zfs_secpolicy_rename_perms(const char *from, const char *to, cred_t *cr)
871 871 {
872 872 char parentname[MAXNAMELEN];
873 873 int error;
874 874
875 875 if ((error = zfs_secpolicy_write_perms(from,
876 876 ZFS_DELEG_PERM_RENAME, cr)) != 0)
877 877 return (error);
878 878
879 879 if ((error = zfs_secpolicy_write_perms(from,
880 880 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
881 881 return (error);
882 882
883 883 if ((error = zfs_get_parent(to, parentname,
884 884 sizeof (parentname))) != 0)
885 885 return (error);
886 886
887 887 if ((error = zfs_secpolicy_write_perms(parentname,
888 888 ZFS_DELEG_PERM_CREATE, cr)) != 0)
889 889 return (error);
890 890
891 891 if ((error = zfs_secpolicy_write_perms(parentname,
892 892 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
893 893 return (error);
894 894
895 895 return (error);
896 896 }
897 897
898 898 /* ARGSUSED */
899 899 static int
900 900 zfs_secpolicy_rename(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
901 901 {
902 902 return (zfs_secpolicy_rename_perms(zc->zc_name, zc->zc_value, cr));
903 903 }
904 904
905 905 /* ARGSUSED */
906 906 static int
907 907 zfs_secpolicy_promote(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
908 908 {
909 909 dsl_pool_t *dp;
910 910 dsl_dataset_t *clone;
911 911 int error;
912 912
913 913 error = zfs_secpolicy_write_perms(zc->zc_name,
914 914 ZFS_DELEG_PERM_PROMOTE, cr);
915 915 if (error != 0)
916 916 return (error);
917 917
918 918 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
919 919 if (error != 0)
920 920 return (error);
921 921
922 922 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &clone);
923 923
924 924 if (error == 0) {
925 925 char parentname[MAXNAMELEN];
926 926 dsl_dataset_t *origin = NULL;
927 927 dsl_dir_t *dd;
928 928 dd = clone->ds_dir;
929 929
930 930 error = dsl_dataset_hold_obj(dd->dd_pool,
931 931 dd->dd_phys->dd_origin_obj, FTAG, &origin);
932 932 if (error != 0) {
933 933 dsl_dataset_rele(clone, FTAG);
934 934 dsl_pool_rele(dp, FTAG);
935 935 return (error);
936 936 }
937 937
938 938 error = zfs_secpolicy_write_perms_ds(zc->zc_name, clone,
939 939 ZFS_DELEG_PERM_MOUNT, cr);
940 940
941 941 dsl_dataset_name(origin, parentname);
942 942 if (error == 0) {
943 943 error = zfs_secpolicy_write_perms_ds(parentname, origin,
944 944 ZFS_DELEG_PERM_PROMOTE, cr);
945 945 }
946 946 dsl_dataset_rele(clone, FTAG);
947 947 dsl_dataset_rele(origin, FTAG);
948 948 }
949 949 dsl_pool_rele(dp, FTAG);
950 950 return (error);
951 951 }
952 952
953 953 /* ARGSUSED */
954 954 static int
955 955 zfs_secpolicy_recv(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
956 956 {
957 957 int error;
958 958
959 959 if ((error = zfs_secpolicy_write_perms(zc->zc_name,
960 960 ZFS_DELEG_PERM_RECEIVE, cr)) != 0)
961 961 return (error);
962 962
963 963 if ((error = zfs_secpolicy_write_perms(zc->zc_name,
964 964 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
965 965 return (error);
966 966
967 967 return (zfs_secpolicy_write_perms(zc->zc_name,
968 968 ZFS_DELEG_PERM_CREATE, cr));
969 969 }
970 970
971 971 int
972 972 zfs_secpolicy_snapshot_perms(const char *name, cred_t *cr)
973 973 {
974 974 return (zfs_secpolicy_write_perms(name,
975 975 ZFS_DELEG_PERM_SNAPSHOT, cr));
976 976 }
977 977
978 978 /*
979 979 * Check for permission to create each snapshot in the nvlist.
980 980 */
981 981 /* ARGSUSED */
982 982 static int
983 983 zfs_secpolicy_snapshot(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
984 984 {
985 985 nvlist_t *snaps;
986 986 int error = 0;
987 987 nvpair_t *pair;
988 988
989 989 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
990 990 return (SET_ERROR(EINVAL));
991 991 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
992 992 pair = nvlist_next_nvpair(snaps, pair)) {
993 993 char *name = nvpair_name(pair);
994 994 char *atp = strchr(name, '@');
995 995
996 996 if (atp == NULL) {
997 997 error = SET_ERROR(EINVAL);
998 998 break;
999 999 }
1000 1000 *atp = '\0';
1001 1001 error = zfs_secpolicy_snapshot_perms(name, cr);
1002 1002 *atp = '@';
1003 1003 if (error != 0)
1004 1004 break;
1005 1005 }
1006 1006 return (error);
1007 1007 }
1008 1008
1009 1009 /* ARGSUSED */
1010 1010 static int
1011 1011 zfs_secpolicy_log_history(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1012 1012 {
1013 1013 /*
1014 1014 * Even root must have a proper TSD so that we know what pool
1015 1015 * to log to.
1016 1016 */
1017 1017 if (tsd_get(zfs_allow_log_key) == NULL)
1018 1018 return (SET_ERROR(EPERM));
1019 1019 return (0);
1020 1020 }
1021 1021
1022 1022 static int
1023 1023 zfs_secpolicy_create_clone(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1024 1024 {
1025 1025 char parentname[MAXNAMELEN];
1026 1026 int error;
1027 1027 char *origin;
1028 1028
1029 1029 if ((error = zfs_get_parent(zc->zc_name, parentname,
1030 1030 sizeof (parentname))) != 0)
1031 1031 return (error);
1032 1032
1033 1033 if (nvlist_lookup_string(innvl, "origin", &origin) == 0 &&
1034 1034 (error = zfs_secpolicy_write_perms(origin,
1035 1035 ZFS_DELEG_PERM_CLONE, cr)) != 0)
1036 1036 return (error);
1037 1037
1038 1038 if ((error = zfs_secpolicy_write_perms(parentname,
1039 1039 ZFS_DELEG_PERM_CREATE, cr)) != 0)
1040 1040 return (error);
1041 1041
1042 1042 return (zfs_secpolicy_write_perms(parentname,
1043 1043 ZFS_DELEG_PERM_MOUNT, cr));
1044 1044 }
1045 1045
1046 1046 /*
1047 1047 * Policy for pool operations - create/destroy pools, add vdevs, etc. Requires
1048 1048 * SYS_CONFIG privilege, which is not available in a local zone.
1049 1049 */
1050 1050 /* ARGSUSED */
1051 1051 static int
1052 1052 zfs_secpolicy_config(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1053 1053 {
1054 1054 if (secpolicy_sys_config(cr, B_FALSE) != 0)
1055 1055 return (SET_ERROR(EPERM));
1056 1056
1057 1057 return (0);
1058 1058 }
1059 1059
1060 1060 /*
1061 1061 * Policy for object to name lookups.
1062 1062 */
1063 1063 /* ARGSUSED */
1064 1064 static int
1065 1065 zfs_secpolicy_diff(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1066 1066 {
1067 1067 int error;
1068 1068
1069 1069 if ((error = secpolicy_sys_config(cr, B_FALSE)) == 0)
1070 1070 return (0);
1071 1071
1072 1072 error = zfs_secpolicy_write_perms(zc->zc_name, ZFS_DELEG_PERM_DIFF, cr);
1073 1073 return (error);
1074 1074 }
1075 1075
1076 1076 /*
1077 1077 * Policy for fault injection. Requires all privileges.
1078 1078 */
1079 1079 /* ARGSUSED */
1080 1080 static int
1081 1081 zfs_secpolicy_inject(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1082 1082 {
1083 1083 return (secpolicy_zinject(cr));
1084 1084 }
1085 1085
1086 1086 /* ARGSUSED */
1087 1087 static int
1088 1088 zfs_secpolicy_inherit_prop(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1089 1089 {
1090 1090 zfs_prop_t prop = zfs_name_to_prop(zc->zc_value);
1091 1091
1092 1092 if (prop == ZPROP_INVAL) {
1093 1093 if (!zfs_prop_user(zc->zc_value))
1094 1094 return (SET_ERROR(EINVAL));
1095 1095 return (zfs_secpolicy_write_perms(zc->zc_name,
1096 1096 ZFS_DELEG_PERM_USERPROP, cr));
1097 1097 } else {
1098 1098 return (zfs_secpolicy_setprop(zc->zc_name, prop,
1099 1099 NULL, cr));
1100 1100 }
1101 1101 }
1102 1102
1103 1103 static int
1104 1104 zfs_secpolicy_userspace_one(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1105 1105 {
1106 1106 int err = zfs_secpolicy_read(zc, innvl, cr);
1107 1107 if (err)
1108 1108 return (err);
1109 1109
1110 1110 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
1111 1111 return (SET_ERROR(EINVAL));
1112 1112
1113 1113 if (zc->zc_value[0] == 0) {
1114 1114 /*
1115 1115 * They are asking about a posix uid/gid. If it's
1116 1116 * themself, allow it.
1117 1117 */
1118 1118 if (zc->zc_objset_type == ZFS_PROP_USERUSED ||
1119 1119 zc->zc_objset_type == ZFS_PROP_USERQUOTA) {
1120 1120 if (zc->zc_guid == crgetuid(cr))
1121 1121 return (0);
1122 1122 } else {
1123 1123 if (groupmember(zc->zc_guid, cr))
1124 1124 return (0);
1125 1125 }
1126 1126 }
1127 1127
1128 1128 return (zfs_secpolicy_write_perms(zc->zc_name,
1129 1129 userquota_perms[zc->zc_objset_type], cr));
1130 1130 }
1131 1131
1132 1132 static int
1133 1133 zfs_secpolicy_userspace_many(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1134 1134 {
1135 1135 int err = zfs_secpolicy_read(zc, innvl, cr);
1136 1136 if (err)
1137 1137 return (err);
1138 1138
1139 1139 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
1140 1140 return (SET_ERROR(EINVAL));
1141 1141
1142 1142 return (zfs_secpolicy_write_perms(zc->zc_name,
1143 1143 userquota_perms[zc->zc_objset_type], cr));
1144 1144 }
1145 1145
1146 1146 /* ARGSUSED */
1147 1147 static int
1148 1148 zfs_secpolicy_userspace_upgrade(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1149 1149 {
1150 1150 return (zfs_secpolicy_setprop(zc->zc_name, ZFS_PROP_VERSION,
1151 1151 NULL, cr));
1152 1152 }
1153 1153
1154 1154 /* ARGSUSED */
1155 1155 static int
1156 1156 zfs_secpolicy_hold(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1157 1157 {
1158 1158 nvpair_t *pair;
1159 1159 nvlist_t *holds;
1160 1160 int error;
1161 1161
1162 1162 error = nvlist_lookup_nvlist(innvl, "holds", &holds);
1163 1163 if (error != 0)
1164 1164 return (SET_ERROR(EINVAL));
1165 1165
1166 1166 for (pair = nvlist_next_nvpair(holds, NULL); pair != NULL;
1167 1167 pair = nvlist_next_nvpair(holds, pair)) {
1168 1168 char fsname[MAXNAMELEN];
1169 1169 error = dmu_fsname(nvpair_name(pair), fsname);
1170 1170 if (error != 0)
1171 1171 return (error);
1172 1172 error = zfs_secpolicy_write_perms(fsname,
1173 1173 ZFS_DELEG_PERM_HOLD, cr);
1174 1174 if (error != 0)
1175 1175 return (error);
1176 1176 }
1177 1177 return (0);
1178 1178 }
1179 1179
1180 1180 /* ARGSUSED */
1181 1181 static int
1182 1182 zfs_secpolicy_release(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1183 1183 {
1184 1184 nvpair_t *pair;
1185 1185 int error;
1186 1186
1187 1187 for (pair = nvlist_next_nvpair(innvl, NULL); pair != NULL;
1188 1188 pair = nvlist_next_nvpair(innvl, pair)) {
1189 1189 char fsname[MAXNAMELEN];
1190 1190 error = dmu_fsname(nvpair_name(pair), fsname);
1191 1191 if (error != 0)
1192 1192 return (error);
1193 1193 error = zfs_secpolicy_write_perms(fsname,
1194 1194 ZFS_DELEG_PERM_RELEASE, cr);
1195 1195 if (error != 0)
1196 1196 return (error);
1197 1197 }
1198 1198 return (0);
1199 1199 }
1200 1200
1201 1201 /*
1202 1202 * Policy for allowing temporary snapshots to be taken or released
1203 1203 */
1204 1204 static int
1205 1205 zfs_secpolicy_tmp_snapshot(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1206 1206 {
1207 1207 /*
1208 1208 * A temporary snapshot is the same as a snapshot,
1209 1209 * hold, destroy and release all rolled into one.
1210 1210 * Delegated diff alone is sufficient that we allow this.
1211 1211 */
1212 1212 int error;
1213 1213
1214 1214 if ((error = zfs_secpolicy_write_perms(zc->zc_name,
1215 1215 ZFS_DELEG_PERM_DIFF, cr)) == 0)
1216 1216 return (0);
1217 1217
1218 1218 error = zfs_secpolicy_snapshot_perms(zc->zc_name, cr);
1219 1219 if (error == 0)
1220 1220 error = zfs_secpolicy_hold(zc, innvl, cr);
1221 1221 if (error == 0)
1222 1222 error = zfs_secpolicy_release(zc, innvl, cr);
1223 1223 if (error == 0)
1224 1224 error = zfs_secpolicy_destroy(zc, innvl, cr);
1225 1225 return (error);
1226 1226 }
1227 1227
1228 1228 /*
1229 1229 * Returns the nvlist as specified by the user in the zfs_cmd_t.
1230 1230 */
1231 1231 static int
1232 1232 get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp)
1233 1233 {
1234 1234 char *packed;
1235 1235 int error;
1236 1236 nvlist_t *list = NULL;
1237 1237
1238 1238 /*
1239 1239 * Read in and unpack the user-supplied nvlist.
1240 1240 */
1241 1241 if (size == 0)
1242 1242 return (SET_ERROR(EINVAL));
1243 1243
1244 1244 packed = kmem_alloc(size, KM_SLEEP);
1245 1245
1246 1246 if ((error = ddi_copyin((void *)(uintptr_t)nvl, packed, size,
1247 1247 iflag)) != 0) {
1248 1248 kmem_free(packed, size);
1249 1249 return (error);
1250 1250 }
1251 1251
1252 1252 if ((error = nvlist_unpack(packed, size, &list, 0)) != 0) {
1253 1253 kmem_free(packed, size);
1254 1254 return (error);
1255 1255 }
1256 1256
1257 1257 kmem_free(packed, size);
1258 1258
1259 1259 *nvp = list;
1260 1260 return (0);
1261 1261 }
1262 1262
1263 1263 /*
1264 1264 * Reduce the size of this nvlist until it can be serialized in 'max' bytes.
1265 1265 * Entries will be removed from the end of the nvlist, and one int32 entry
1266 1266 * named "N_MORE_ERRORS" will be added indicating how many entries were
1267 1267 * removed.
1268 1268 */
1269 1269 static int
1270 1270 nvlist_smush(nvlist_t *errors, size_t max)
1271 1271 {
1272 1272 size_t size;
1273 1273
1274 1274 size = fnvlist_size(errors);
1275 1275
1276 1276 if (size > max) {
1277 1277 nvpair_t *more_errors;
1278 1278 int n = 0;
1279 1279
1280 1280 if (max < 1024)
1281 1281 return (SET_ERROR(ENOMEM));
1282 1282
1283 1283 fnvlist_add_int32(errors, ZPROP_N_MORE_ERRORS, 0);
1284 1284 more_errors = nvlist_prev_nvpair(errors, NULL);
1285 1285
1286 1286 do {
1287 1287 nvpair_t *pair = nvlist_prev_nvpair(errors,
1288 1288 more_errors);
1289 1289 fnvlist_remove_nvpair(errors, pair);
1290 1290 n++;
1291 1291 size = fnvlist_size(errors);
1292 1292 } while (size > max);
1293 1293
1294 1294 fnvlist_remove_nvpair(errors, more_errors);
1295 1295 fnvlist_add_int32(errors, ZPROP_N_MORE_ERRORS, n);
1296 1296 ASSERT3U(fnvlist_size(errors), <=, max);
1297 1297 }
1298 1298
1299 1299 return (0);
1300 1300 }
1301 1301
1302 1302 static int
1303 1303 put_nvlist(zfs_cmd_t *zc, nvlist_t *nvl)
1304 1304 {
1305 1305 char *packed = NULL;
1306 1306 int error = 0;
1307 1307 size_t size;
1308 1308
1309 1309 size = fnvlist_size(nvl);
1310 1310
1311 1311 if (size > zc->zc_nvlist_dst_size) {
1312 1312 error = SET_ERROR(ENOMEM);
1313 1313 } else {
1314 1314 packed = fnvlist_pack(nvl, &size);
1315 1315 if (ddi_copyout(packed, (void *)(uintptr_t)zc->zc_nvlist_dst,
1316 1316 size, zc->zc_iflags) != 0)
1317 1317 error = SET_ERROR(EFAULT);
1318 1318 fnvlist_pack_free(packed, size);
1319 1319 }
1320 1320
1321 1321 zc->zc_nvlist_dst_size = size;
1322 1322 zc->zc_nvlist_dst_filled = B_TRUE;
1323 1323 return (error);
1324 1324 }
1325 1325
1326 1326 static int
1327 1327 getzfsvfs(const char *dsname, zfsvfs_t **zfvp)
1328 1328 {
1329 1329 objset_t *os;
1330 1330 int error;
1331 1331
1332 1332 error = dmu_objset_hold(dsname, FTAG, &os);
1333 1333 if (error != 0)
1334 1334 return (error);
1335 1335 if (dmu_objset_type(os) != DMU_OST_ZFS) {
1336 1336 dmu_objset_rele(os, FTAG);
1337 1337 return (SET_ERROR(EINVAL));
1338 1338 }
1339 1339
1340 1340 mutex_enter(&os->os_user_ptr_lock);
1341 1341 *zfvp = dmu_objset_get_user(os);
1342 1342 if (*zfvp) {
1343 1343 VFS_HOLD((*zfvp)->z_vfs);
1344 1344 } else {
1345 1345 error = SET_ERROR(ESRCH);
1346 1346 }
1347 1347 mutex_exit(&os->os_user_ptr_lock);
1348 1348 dmu_objset_rele(os, FTAG);
1349 1349 return (error);
1350 1350 }
1351 1351
1352 1352 /*
1353 1353 * Find a zfsvfs_t for a mounted filesystem, or create our own, in which
1354 1354 * case its z_vfs will be NULL, and it will be opened as the owner.
1355 1355 * If 'writer' is set, the z_teardown_lock will be held for RW_WRITER,
1356 1356 * which prevents all vnode ops from running.
1357 1357 */
1358 1358 static int
1359 1359 zfsvfs_hold(const char *name, void *tag, zfsvfs_t **zfvp, boolean_t writer)
1360 1360 {
1361 1361 int error = 0;
1362 1362
1363 1363 if (getzfsvfs(name, zfvp) != 0)
1364 1364 error = zfsvfs_create(name, zfvp);
1365 1365 if (error == 0) {
1366 1366 rrw_enter(&(*zfvp)->z_teardown_lock, (writer) ? RW_WRITER :
1367 1367 RW_READER, tag);
1368 1368 if ((*zfvp)->z_unmounted) {
1369 1369 /*
1370 1370 * XXX we could probably try again, since the unmounting
1371 1371 * thread should be just about to disassociate the
1372 1372 * objset from the zfsvfs.
1373 1373 */
1374 1374 rrw_exit(&(*zfvp)->z_teardown_lock, tag);
1375 1375 return (SET_ERROR(EBUSY));
1376 1376 }
1377 1377 }
1378 1378 return (error);
1379 1379 }
1380 1380
1381 1381 static void
1382 1382 zfsvfs_rele(zfsvfs_t *zfsvfs, void *tag)
1383 1383 {
1384 1384 rrw_exit(&zfsvfs->z_teardown_lock, tag);
1385 1385
1386 1386 if (zfsvfs->z_vfs) {
1387 1387 VFS_RELE(zfsvfs->z_vfs);
1388 1388 } else {
1389 1389 dmu_objset_disown(zfsvfs->z_os, zfsvfs);
1390 1390 zfsvfs_free(zfsvfs);
1391 1391 }
1392 1392 }
1393 1393
1394 1394 static int
1395 1395 zfs_ioc_pool_create(zfs_cmd_t *zc)
1396 1396 {
1397 1397 int error;
1398 1398 nvlist_t *config, *props = NULL;
1399 1399 nvlist_t *rootprops = NULL;
1400 1400 nvlist_t *zplprops = NULL;
1401 1401
1402 1402 if (error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1403 1403 zc->zc_iflags, &config))
1404 1404 return (error);
1405 1405
1406 1406 if (zc->zc_nvlist_src_size != 0 && (error =
1407 1407 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1408 1408 zc->zc_iflags, &props))) {
1409 1409 nvlist_free(config);
1410 1410 return (error);
1411 1411 }
1412 1412
1413 1413 if (props) {
1414 1414 nvlist_t *nvl = NULL;
1415 1415 uint64_t version = SPA_VERSION;
1416 1416
1417 1417 (void) nvlist_lookup_uint64(props,
1418 1418 zpool_prop_to_name(ZPOOL_PROP_VERSION), &version);
1419 1419 if (!SPA_VERSION_IS_SUPPORTED(version)) {
1420 1420 error = SET_ERROR(EINVAL);
1421 1421 goto pool_props_bad;
1422 1422 }
1423 1423 (void) nvlist_lookup_nvlist(props, ZPOOL_ROOTFS_PROPS, &nvl);
1424 1424 if (nvl) {
1425 1425 error = nvlist_dup(nvl, &rootprops, KM_SLEEP);
1426 1426 if (error != 0) {
1427 1427 nvlist_free(config);
1428 1428 nvlist_free(props);
1429 1429 return (error);
1430 1430 }
1431 1431 (void) nvlist_remove_all(props, ZPOOL_ROOTFS_PROPS);
1432 1432 }
1433 1433 VERIFY(nvlist_alloc(&zplprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
1434 1434 error = zfs_fill_zplprops_root(version, rootprops,
1435 1435 zplprops, NULL);
1436 1436 if (error != 0)
1437 1437 goto pool_props_bad;
1438 1438 }
1439 1439
1440 1440 error = spa_create(zc->zc_name, config, props, zplprops);
1441 1441
1442 1442 /*
1443 1443 * Set the remaining root properties
1444 1444 */
1445 1445 if (!error && (error = zfs_set_prop_nvlist(zc->zc_name,
1446 1446 ZPROP_SRC_LOCAL, rootprops, NULL)) != 0)
1447 1447 (void) spa_destroy(zc->zc_name);
1448 1448
1449 1449 pool_props_bad:
1450 1450 nvlist_free(rootprops);
1451 1451 nvlist_free(zplprops);
1452 1452 nvlist_free(config);
1453 1453 nvlist_free(props);
1454 1454
1455 1455 return (error);
1456 1456 }
1457 1457
1458 1458 static int
1459 1459 zfs_ioc_pool_destroy(zfs_cmd_t *zc)
1460 1460 {
1461 1461 int error;
1462 1462 zfs_log_history(zc);
1463 1463 error = spa_destroy(zc->zc_name);
1464 1464 if (error == 0)
1465 1465 zvol_remove_minors(zc->zc_name);
1466 1466 return (error);
1467 1467 }
1468 1468
1469 1469 static int
1470 1470 zfs_ioc_pool_import(zfs_cmd_t *zc)
1471 1471 {
1472 1472 nvlist_t *config, *props = NULL;
1473 1473 uint64_t guid;
1474 1474 int error;
1475 1475
1476 1476 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1477 1477 zc->zc_iflags, &config)) != 0)
1478 1478 return (error);
1479 1479
1480 1480 if (zc->zc_nvlist_src_size != 0 && (error =
1481 1481 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1482 1482 zc->zc_iflags, &props))) {
1483 1483 nvlist_free(config);
1484 1484 return (error);
1485 1485 }
1486 1486
1487 1487 if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, &guid) != 0 ||
1488 1488 guid != zc->zc_guid)
1489 1489 error = SET_ERROR(EINVAL);
1490 1490 else
1491 1491 error = spa_import(zc->zc_name, config, props, zc->zc_cookie);
1492 1492
1493 1493 if (zc->zc_nvlist_dst != 0) {
1494 1494 int err;
1495 1495
1496 1496 if ((err = put_nvlist(zc, config)) != 0)
1497 1497 error = err;
1498 1498 }
1499 1499
1500 1500 nvlist_free(config);
1501 1501
1502 1502 if (props)
1503 1503 nvlist_free(props);
1504 1504
1505 1505 return (error);
1506 1506 }
1507 1507
1508 1508 static int
1509 1509 zfs_ioc_pool_export(zfs_cmd_t *zc)
1510 1510 {
1511 1511 int error;
1512 1512 boolean_t force = (boolean_t)zc->zc_cookie;
1513 1513 boolean_t hardforce = (boolean_t)zc->zc_guid;
1514 1514
1515 1515 zfs_log_history(zc);
1516 1516 error = spa_export(zc->zc_name, NULL, force, hardforce);
1517 1517 if (error == 0)
1518 1518 zvol_remove_minors(zc->zc_name);
1519 1519 return (error);
1520 1520 }
1521 1521
1522 1522 static int
1523 1523 zfs_ioc_pool_configs(zfs_cmd_t *zc)
1524 1524 {
1525 1525 nvlist_t *configs;
1526 1526 int error;
1527 1527
1528 1528 if ((configs = spa_all_configs(&zc->zc_cookie)) == NULL)
1529 1529 return (SET_ERROR(EEXIST));
1530 1530
1531 1531 error = put_nvlist(zc, configs);
1532 1532
1533 1533 nvlist_free(configs);
1534 1534
1535 1535 return (error);
1536 1536 }
1537 1537
1538 1538 /*
1539 1539 * inputs:
1540 1540 * zc_name name of the pool
1541 1541 *
1542 1542 * outputs:
1543 1543 * zc_cookie real errno
1544 1544 * zc_nvlist_dst config nvlist
1545 1545 * zc_nvlist_dst_size size of config nvlist
1546 1546 */
1547 1547 static int
1548 1548 zfs_ioc_pool_stats(zfs_cmd_t *zc)
1549 1549 {
1550 1550 nvlist_t *config;
1551 1551 int error;
1552 1552 int ret = 0;
1553 1553
1554 1554 error = spa_get_stats(zc->zc_name, &config, zc->zc_value,
1555 1555 sizeof (zc->zc_value));
1556 1556
1557 1557 if (config != NULL) {
1558 1558 ret = put_nvlist(zc, config);
1559 1559 nvlist_free(config);
1560 1560
1561 1561 /*
1562 1562 * The config may be present even if 'error' is non-zero.
1563 1563 * In this case we return success, and preserve the real errno
1564 1564 * in 'zc_cookie'.
1565 1565 */
1566 1566 zc->zc_cookie = error;
1567 1567 } else {
1568 1568 ret = error;
1569 1569 }
1570 1570
1571 1571 return (ret);
1572 1572 }
1573 1573
1574 1574 /*
1575 1575 * Try to import the given pool, returning pool stats as appropriate so that
1576 1576 * user land knows which devices are available and overall pool health.
1577 1577 */
1578 1578 static int
1579 1579 zfs_ioc_pool_tryimport(zfs_cmd_t *zc)
1580 1580 {
1581 1581 nvlist_t *tryconfig, *config;
1582 1582 int error;
1583 1583
1584 1584 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1585 1585 zc->zc_iflags, &tryconfig)) != 0)
1586 1586 return (error);
1587 1587
1588 1588 config = spa_tryimport(tryconfig);
1589 1589
1590 1590 nvlist_free(tryconfig);
1591 1591
1592 1592 if (config == NULL)
1593 1593 return (SET_ERROR(EINVAL));
1594 1594
1595 1595 error = put_nvlist(zc, config);
1596 1596 nvlist_free(config);
1597 1597
1598 1598 return (error);
1599 1599 }
1600 1600
1601 1601 /*
1602 1602 * inputs:
1603 1603 * zc_name name of the pool
1604 1604 * zc_cookie scan func (pool_scan_func_t)
1605 1605 */
1606 1606 static int
1607 1607 zfs_ioc_pool_scan(zfs_cmd_t *zc)
1608 1608 {
1609 1609 spa_t *spa;
1610 1610 int error;
1611 1611
1612 1612 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1613 1613 return (error);
1614 1614
1615 1615 if (zc->zc_cookie == POOL_SCAN_NONE)
1616 1616 error = spa_scan_stop(spa);
1617 1617 else
1618 1618 error = spa_scan(spa, zc->zc_cookie);
1619 1619
1620 1620 spa_close(spa, FTAG);
1621 1621
1622 1622 return (error);
1623 1623 }
1624 1624
1625 1625 static int
1626 1626 zfs_ioc_pool_freeze(zfs_cmd_t *zc)
1627 1627 {
1628 1628 spa_t *spa;
1629 1629 int error;
1630 1630
1631 1631 error = spa_open(zc->zc_name, &spa, FTAG);
1632 1632 if (error == 0) {
1633 1633 spa_freeze(spa);
1634 1634 spa_close(spa, FTAG);
1635 1635 }
1636 1636 return (error);
1637 1637 }
1638 1638
1639 1639 static int
1640 1640 zfs_ioc_pool_upgrade(zfs_cmd_t *zc)
1641 1641 {
1642 1642 spa_t *spa;
1643 1643 int error;
1644 1644
1645 1645 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1646 1646 return (error);
1647 1647
1648 1648 if (zc->zc_cookie < spa_version(spa) ||
1649 1649 !SPA_VERSION_IS_SUPPORTED(zc->zc_cookie)) {
1650 1650 spa_close(spa, FTAG);
1651 1651 return (SET_ERROR(EINVAL));
1652 1652 }
1653 1653
1654 1654 spa_upgrade(spa, zc->zc_cookie);
1655 1655 spa_close(spa, FTAG);
1656 1656
1657 1657 return (error);
1658 1658 }
1659 1659
1660 1660 static int
1661 1661 zfs_ioc_pool_get_history(zfs_cmd_t *zc)
1662 1662 {
1663 1663 spa_t *spa;
1664 1664 char *hist_buf;
1665 1665 uint64_t size;
1666 1666 int error;
1667 1667
1668 1668 if ((size = zc->zc_history_len) == 0)
1669 1669 return (SET_ERROR(EINVAL));
1670 1670
1671 1671 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1672 1672 return (error);
1673 1673
1674 1674 if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) {
1675 1675 spa_close(spa, FTAG);
1676 1676 return (SET_ERROR(ENOTSUP));
1677 1677 }
1678 1678
1679 1679 hist_buf = kmem_alloc(size, KM_SLEEP);
1680 1680 if ((error = spa_history_get(spa, &zc->zc_history_offset,
1681 1681 &zc->zc_history_len, hist_buf)) == 0) {
1682 1682 error = ddi_copyout(hist_buf,
1683 1683 (void *)(uintptr_t)zc->zc_history,
1684 1684 zc->zc_history_len, zc->zc_iflags);
1685 1685 }
1686 1686
1687 1687 spa_close(spa, FTAG);
1688 1688 kmem_free(hist_buf, size);
1689 1689 return (error);
1690 1690 }
1691 1691
1692 1692 static int
1693 1693 zfs_ioc_pool_reguid(zfs_cmd_t *zc)
1694 1694 {
1695 1695 spa_t *spa;
1696 1696 int error;
1697 1697
1698 1698 error = spa_open(zc->zc_name, &spa, FTAG);
1699 1699 if (error == 0) {
1700 1700 error = spa_change_guid(spa);
1701 1701 spa_close(spa, FTAG);
1702 1702 }
1703 1703 return (error);
1704 1704 }
1705 1705
1706 1706 static int
1707 1707 zfs_ioc_dsobj_to_dsname(zfs_cmd_t *zc)
1708 1708 {
1709 1709 return (dsl_dsobj_to_dsname(zc->zc_name, zc->zc_obj, zc->zc_value));
1710 1710 }
1711 1711
1712 1712 /*
1713 1713 * inputs:
1714 1714 * zc_name name of filesystem
1715 1715 * zc_obj object to find
1716 1716 *
1717 1717 * outputs:
1718 1718 * zc_value name of object
1719 1719 */
1720 1720 static int
1721 1721 zfs_ioc_obj_to_path(zfs_cmd_t *zc)
1722 1722 {
1723 1723 objset_t *os;
1724 1724 int error;
1725 1725
1726 1726 /* XXX reading from objset not owned */
1727 1727 if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os)) != 0)
1728 1728 return (error);
1729 1729 if (dmu_objset_type(os) != DMU_OST_ZFS) {
1730 1730 dmu_objset_rele(os, FTAG);
1731 1731 return (SET_ERROR(EINVAL));
1732 1732 }
1733 1733 error = zfs_obj_to_path(os, zc->zc_obj, zc->zc_value,
1734 1734 sizeof (zc->zc_value));
1735 1735 dmu_objset_rele(os, FTAG);
1736 1736
1737 1737 return (error);
1738 1738 }
1739 1739
1740 1740 /*
1741 1741 * inputs:
1742 1742 * zc_name name of filesystem
1743 1743 * zc_obj object to find
1744 1744 *
1745 1745 * outputs:
1746 1746 * zc_stat stats on object
1747 1747 * zc_value path to object
1748 1748 */
1749 1749 static int
1750 1750 zfs_ioc_obj_to_stats(zfs_cmd_t *zc)
1751 1751 {
1752 1752 objset_t *os;
1753 1753 int error;
1754 1754
1755 1755 /* XXX reading from objset not owned */
1756 1756 if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os)) != 0)
1757 1757 return (error);
1758 1758 if (dmu_objset_type(os) != DMU_OST_ZFS) {
1759 1759 dmu_objset_rele(os, FTAG);
1760 1760 return (SET_ERROR(EINVAL));
1761 1761 }
1762 1762 error = zfs_obj_to_stats(os, zc->zc_obj, &zc->zc_stat, zc->zc_value,
1763 1763 sizeof (zc->zc_value));
1764 1764 dmu_objset_rele(os, FTAG);
1765 1765
1766 1766 return (error);
1767 1767 }
1768 1768
1769 1769 static int
1770 1770 zfs_ioc_vdev_add(zfs_cmd_t *zc)
1771 1771 {
1772 1772 spa_t *spa;
1773 1773 int error;
1774 1774 nvlist_t *config, **l2cache, **spares;
1775 1775 uint_t nl2cache = 0, nspares = 0;
1776 1776
1777 1777 error = spa_open(zc->zc_name, &spa, FTAG);
1778 1778 if (error != 0)
1779 1779 return (error);
1780 1780
1781 1781 error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1782 1782 zc->zc_iflags, &config);
1783 1783 (void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_L2CACHE,
1784 1784 &l2cache, &nl2cache);
1785 1785
1786 1786 (void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_SPARES,
1787 1787 &spares, &nspares);
1788 1788
1789 1789 /*
1790 1790 * A root pool with concatenated devices is not supported.
1791 1791 * Thus, can not add a device to a root pool.
1792 1792 *
1793 1793 * Intent log device can not be added to a rootpool because
1794 1794 * during mountroot, zil is replayed, a seperated log device
1795 1795 * can not be accessed during the mountroot time.
1796 1796 *
1797 1797 * l2cache and spare devices are ok to be added to a rootpool.
1798 1798 */
1799 1799 if (spa_bootfs(spa) != 0 && nl2cache == 0 && nspares == 0) {
1800 1800 nvlist_free(config);
1801 1801 spa_close(spa, FTAG);
1802 1802 return (SET_ERROR(EDOM));
1803 1803 }
1804 1804
1805 1805 if (error == 0) {
1806 1806 error = spa_vdev_add(spa, config);
1807 1807 nvlist_free(config);
1808 1808 }
1809 1809 spa_close(spa, FTAG);
1810 1810 return (error);
1811 1811 }
1812 1812
1813 1813 /*
1814 1814 * inputs:
1815 1815 * zc_name name of the pool
1816 1816 * zc_nvlist_conf nvlist of devices to remove
1817 1817 * zc_cookie to stop the remove?
1818 1818 */
1819 1819 static int
1820 1820 zfs_ioc_vdev_remove(zfs_cmd_t *zc)
1821 1821 {
1822 1822 spa_t *spa;
1823 1823 int error;
1824 1824
1825 1825 error = spa_open(zc->zc_name, &spa, FTAG);
1826 1826 if (error != 0)
1827 1827 return (error);
1828 1828 error = spa_vdev_remove(spa, zc->zc_guid, B_FALSE);
1829 1829 spa_close(spa, FTAG);
1830 1830 return (error);
1831 1831 }
1832 1832
1833 1833 static int
1834 1834 zfs_ioc_vdev_set_state(zfs_cmd_t *zc)
1835 1835 {
1836 1836 spa_t *spa;
1837 1837 int error;
1838 1838 vdev_state_t newstate = VDEV_STATE_UNKNOWN;
1839 1839
1840 1840 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1841 1841 return (error);
1842 1842 switch (zc->zc_cookie) {
1843 1843 case VDEV_STATE_ONLINE:
1844 1844 error = vdev_online(spa, zc->zc_guid, zc->zc_obj, &newstate);
1845 1845 break;
1846 1846
1847 1847 case VDEV_STATE_OFFLINE:
1848 1848 error = vdev_offline(spa, zc->zc_guid, zc->zc_obj);
1849 1849 break;
1850 1850
1851 1851 case VDEV_STATE_FAULTED:
1852 1852 if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED &&
1853 1853 zc->zc_obj != VDEV_AUX_EXTERNAL)
1854 1854 zc->zc_obj = VDEV_AUX_ERR_EXCEEDED;
1855 1855
1856 1856 error = vdev_fault(spa, zc->zc_guid, zc->zc_obj);
1857 1857 break;
1858 1858
1859 1859 case VDEV_STATE_DEGRADED:
1860 1860 if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED &&
1861 1861 zc->zc_obj != VDEV_AUX_EXTERNAL)
1862 1862 zc->zc_obj = VDEV_AUX_ERR_EXCEEDED;
1863 1863
1864 1864 error = vdev_degrade(spa, zc->zc_guid, zc->zc_obj);
1865 1865 break;
1866 1866
1867 1867 default:
1868 1868 error = SET_ERROR(EINVAL);
1869 1869 }
1870 1870 zc->zc_cookie = newstate;
1871 1871 spa_close(spa, FTAG);
1872 1872 return (error);
1873 1873 }
1874 1874
1875 1875 static int
1876 1876 zfs_ioc_vdev_attach(zfs_cmd_t *zc)
1877 1877 {
1878 1878 spa_t *spa;
1879 1879 int replacing = zc->zc_cookie;
1880 1880 nvlist_t *config;
1881 1881 int error;
1882 1882
1883 1883 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1884 1884 return (error);
1885 1885
1886 1886 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1887 1887 zc->zc_iflags, &config)) == 0) {
1888 1888 error = spa_vdev_attach(spa, zc->zc_guid, config, replacing);
1889 1889 nvlist_free(config);
1890 1890 }
1891 1891
1892 1892 spa_close(spa, FTAG);
1893 1893 return (error);
1894 1894 }
1895 1895
1896 1896 static int
1897 1897 zfs_ioc_vdev_detach(zfs_cmd_t *zc)
1898 1898 {
1899 1899 spa_t *spa;
1900 1900 int error;
1901 1901
1902 1902 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1903 1903 return (error);
1904 1904
1905 1905 error = spa_vdev_detach(spa, zc->zc_guid, 0, B_FALSE);
1906 1906
1907 1907 spa_close(spa, FTAG);
1908 1908 return (error);
1909 1909 }
1910 1910
1911 1911 static int
1912 1912 zfs_ioc_vdev_split(zfs_cmd_t *zc)
1913 1913 {
1914 1914 spa_t *spa;
1915 1915 nvlist_t *config, *props = NULL;
1916 1916 int error;
1917 1917 boolean_t exp = !!(zc->zc_cookie & ZPOOL_EXPORT_AFTER_SPLIT);
1918 1918
1919 1919 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1920 1920 return (error);
1921 1921
1922 1922 if (error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1923 1923 zc->zc_iflags, &config)) {
1924 1924 spa_close(spa, FTAG);
1925 1925 return (error);
1926 1926 }
1927 1927
1928 1928 if (zc->zc_nvlist_src_size != 0 && (error =
1929 1929 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1930 1930 zc->zc_iflags, &props))) {
1931 1931 spa_close(spa, FTAG);
1932 1932 nvlist_free(config);
1933 1933 return (error);
1934 1934 }
1935 1935
1936 1936 error = spa_vdev_split_mirror(spa, zc->zc_string, config, props, exp);
1937 1937
1938 1938 spa_close(spa, FTAG);
1939 1939
1940 1940 nvlist_free(config);
1941 1941 nvlist_free(props);
1942 1942
1943 1943 return (error);
1944 1944 }
1945 1945
1946 1946 static int
1947 1947 zfs_ioc_vdev_setpath(zfs_cmd_t *zc)
1948 1948 {
1949 1949 spa_t *spa;
1950 1950 char *path = zc->zc_value;
1951 1951 uint64_t guid = zc->zc_guid;
1952 1952 int error;
1953 1953
1954 1954 error = spa_open(zc->zc_name, &spa, FTAG);
1955 1955 if (error != 0)
1956 1956 return (error);
1957 1957
1958 1958 error = spa_vdev_setpath(spa, guid, path);
1959 1959 spa_close(spa, FTAG);
1960 1960 return (error);
1961 1961 }
1962 1962
1963 1963 static int
1964 1964 zfs_ioc_vdev_setfru(zfs_cmd_t *zc)
1965 1965 {
1966 1966 spa_t *spa;
1967 1967 char *fru = zc->zc_value;
1968 1968 uint64_t guid = zc->zc_guid;
1969 1969 int error;
1970 1970
1971 1971 error = spa_open(zc->zc_name, &spa, FTAG);
1972 1972 if (error != 0)
1973 1973 return (error);
1974 1974
1975 1975 error = spa_vdev_setfru(spa, guid, fru);
1976 1976 spa_close(spa, FTAG);
1977 1977 return (error);
1978 1978 }
1979 1979
1980 1980 static int
1981 1981 zfs_ioc_objset_stats_impl(zfs_cmd_t *zc, objset_t *os)
1982 1982 {
1983 1983 int error = 0;
1984 1984 nvlist_t *nv;
1985 1985
1986 1986 dmu_objset_fast_stat(os, &zc->zc_objset_stats);
1987 1987
1988 1988 if (zc->zc_nvlist_dst != 0 &&
1989 1989 (error = dsl_prop_get_all(os, &nv)) == 0) {
1990 1990 dmu_objset_stats(os, nv);
1991 1991 /*
1992 1992 * NB: zvol_get_stats() will read the objset contents,
1993 1993 * which we aren't supposed to do with a
1994 1994 * DS_MODE_USER hold, because it could be
1995 1995 * inconsistent. So this is a bit of a workaround...
1996 1996 * XXX reading with out owning
1997 1997 */
1998 1998 if (!zc->zc_objset_stats.dds_inconsistent &&
1999 1999 dmu_objset_type(os) == DMU_OST_ZVOL) {
2000 2000 error = zvol_get_stats(os, nv);
2001 2001 if (error == EIO)
2002 2002 return (error);
2003 2003 VERIFY0(error);
2004 2004 }
2005 2005 error = put_nvlist(zc, nv);
2006 2006 nvlist_free(nv);
2007 2007 }
2008 2008
2009 2009 return (error);
2010 2010 }
2011 2011
2012 2012 /*
2013 2013 * inputs:
2014 2014 * zc_name name of filesystem
2015 2015 * zc_nvlist_dst_size size of buffer for property nvlist
2016 2016 *
2017 2017 * outputs:
2018 2018 * zc_objset_stats stats
2019 2019 * zc_nvlist_dst property nvlist
2020 2020 * zc_nvlist_dst_size size of property nvlist
2021 2021 */
2022 2022 static int
2023 2023 zfs_ioc_objset_stats(zfs_cmd_t *zc)
2024 2024 {
2025 2025 objset_t *os;
2026 2026 int error;
2027 2027
2028 2028 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
2029 2029 if (error == 0) {
2030 2030 error = zfs_ioc_objset_stats_impl(zc, os);
2031 2031 dmu_objset_rele(os, FTAG);
2032 2032 }
2033 2033
2034 2034 return (error);
2035 2035 }
2036 2036
2037 2037 /*
2038 2038 * inputs:
2039 2039 * zc_name name of filesystem
2040 2040 * zc_nvlist_dst_size size of buffer for property nvlist
2041 2041 *
2042 2042 * outputs:
2043 2043 * zc_nvlist_dst received property nvlist
2044 2044 * zc_nvlist_dst_size size of received property nvlist
2045 2045 *
2046 2046 * Gets received properties (distinct from local properties on or after
2047 2047 * SPA_VERSION_RECVD_PROPS) for callers who want to differentiate received from
2048 2048 * local property values.
2049 2049 */
2050 2050 static int
2051 2051 zfs_ioc_objset_recvd_props(zfs_cmd_t *zc)
2052 2052 {
2053 2053 int error = 0;
2054 2054 nvlist_t *nv;
2055 2055
2056 2056 /*
2057 2057 * Without this check, we would return local property values if the
2058 2058 * caller has not already received properties on or after
2059 2059 * SPA_VERSION_RECVD_PROPS.
2060 2060 */
2061 2061 if (!dsl_prop_get_hasrecvd(zc->zc_name))
2062 2062 return (SET_ERROR(ENOTSUP));
2063 2063
2064 2064 if (zc->zc_nvlist_dst != 0 &&
2065 2065 (error = dsl_prop_get_received(zc->zc_name, &nv)) == 0) {
2066 2066 error = put_nvlist(zc, nv);
2067 2067 nvlist_free(nv);
2068 2068 }
2069 2069
2070 2070 return (error);
2071 2071 }
2072 2072
2073 2073 static int
2074 2074 nvl_add_zplprop(objset_t *os, nvlist_t *props, zfs_prop_t prop)
2075 2075 {
2076 2076 uint64_t value;
2077 2077 int error;
2078 2078
2079 2079 /*
2080 2080 * zfs_get_zplprop() will either find a value or give us
2081 2081 * the default value (if there is one).
2082 2082 */
2083 2083 if ((error = zfs_get_zplprop(os, prop, &value)) != 0)
2084 2084 return (error);
2085 2085 VERIFY(nvlist_add_uint64(props, zfs_prop_to_name(prop), value) == 0);
2086 2086 return (0);
2087 2087 }
2088 2088
2089 2089 /*
2090 2090 * inputs:
2091 2091 * zc_name name of filesystem
2092 2092 * zc_nvlist_dst_size size of buffer for zpl property nvlist
2093 2093 *
2094 2094 * outputs:
2095 2095 * zc_nvlist_dst zpl property nvlist
2096 2096 * zc_nvlist_dst_size size of zpl property nvlist
2097 2097 */
2098 2098 static int
2099 2099 zfs_ioc_objset_zplprops(zfs_cmd_t *zc)
2100 2100 {
2101 2101 objset_t *os;
2102 2102 int err;
2103 2103
2104 2104 /* XXX reading without owning */
2105 2105 if (err = dmu_objset_hold(zc->zc_name, FTAG, &os))
2106 2106 return (err);
2107 2107
2108 2108 dmu_objset_fast_stat(os, &zc->zc_objset_stats);
2109 2109
2110 2110 /*
2111 2111 * NB: nvl_add_zplprop() will read the objset contents,
2112 2112 * which we aren't supposed to do with a DS_MODE_USER
2113 2113 * hold, because it could be inconsistent.
2114 2114 */
2115 2115 if (zc->zc_nvlist_dst != NULL &&
2116 2116 !zc->zc_objset_stats.dds_inconsistent &&
2117 2117 dmu_objset_type(os) == DMU_OST_ZFS) {
2118 2118 nvlist_t *nv;
2119 2119
2120 2120 VERIFY(nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2121 2121 if ((err = nvl_add_zplprop(os, nv, ZFS_PROP_VERSION)) == 0 &&
2122 2122 (err = nvl_add_zplprop(os, nv, ZFS_PROP_NORMALIZE)) == 0 &&
2123 2123 (err = nvl_add_zplprop(os, nv, ZFS_PROP_UTF8ONLY)) == 0 &&
2124 2124 (err = nvl_add_zplprop(os, nv, ZFS_PROP_CASE)) == 0)
2125 2125 err = put_nvlist(zc, nv);
2126 2126 nvlist_free(nv);
2127 2127 } else {
2128 2128 err = SET_ERROR(ENOENT);
2129 2129 }
2130 2130 dmu_objset_rele(os, FTAG);
2131 2131 return (err);
2132 2132 }
2133 2133
2134 2134 static boolean_t
2135 2135 dataset_name_hidden(const char *name)
2136 2136 {
2137 2137 /*
2138 2138 * Skip over datasets that are not visible in this zone,
2139 2139 * internal datasets (which have a $ in their name), and
2140 2140 * temporary datasets (which have a % in their name).
2141 2141 */
2142 2142 if (strchr(name, '$') != NULL)
2143 2143 return (B_TRUE);
2144 2144 if (strchr(name, '%') != NULL)
2145 2145 return (B_TRUE);
2146 2146 if (!INGLOBALZONE(curproc) && !zone_dataset_visible(name, NULL))
2147 2147 return (B_TRUE);
2148 2148 return (B_FALSE);
2149 2149 }
2150 2150
2151 2151 /*
2152 2152 * inputs:
2153 2153 * zc_name name of filesystem
2154 2154 * zc_cookie zap cursor
2155 2155 * zc_nvlist_dst_size size of buffer for property nvlist
2156 2156 *
2157 2157 * outputs:
2158 2158 * zc_name name of next filesystem
2159 2159 * zc_cookie zap cursor
2160 2160 * zc_objset_stats stats
2161 2161 * zc_nvlist_dst property nvlist
2162 2162 * zc_nvlist_dst_size size of property nvlist
2163 2163 */
2164 2164 static int
2165 2165 zfs_ioc_dataset_list_next(zfs_cmd_t *zc)
2166 2166 {
2167 2167 objset_t *os;
2168 2168 int error;
2169 2169 char *p;
2170 2170 size_t orig_len = strlen(zc->zc_name);
2171 2171
2172 2172 top:
2173 2173 if (error = dmu_objset_hold(zc->zc_name, FTAG, &os)) {
2174 2174 if (error == ENOENT)
2175 2175 error = SET_ERROR(ESRCH);
2176 2176 return (error);
2177 2177 }
2178 2178
2179 2179 p = strrchr(zc->zc_name, '/');
2180 2180 if (p == NULL || p[1] != '\0')
2181 2181 (void) strlcat(zc->zc_name, "/", sizeof (zc->zc_name));
2182 2182 p = zc->zc_name + strlen(zc->zc_name);
2183 2183
2184 2184 do {
2185 2185 error = dmu_dir_list_next(os,
2186 2186 sizeof (zc->zc_name) - (p - zc->zc_name), p,
2187 2187 NULL, &zc->zc_cookie);
2188 2188 if (error == ENOENT)
2189 2189 error = SET_ERROR(ESRCH);
2190 2190 } while (error == 0 && dataset_name_hidden(zc->zc_name));
2191 2191 dmu_objset_rele(os, FTAG);
2192 2192
2193 2193 /*
2194 2194 * If it's an internal dataset (ie. with a '$' in its name),
2195 2195 * don't try to get stats for it, otherwise we'll return ENOENT.
2196 2196 */
2197 2197 if (error == 0 && strchr(zc->zc_name, '$') == NULL) {
2198 2198 error = zfs_ioc_objset_stats(zc); /* fill in the stats */
2199 2199 if (error == ENOENT) {
2200 2200 /* We lost a race with destroy, get the next one. */
2201 2201 zc->zc_name[orig_len] = '\0';
2202 2202 goto top;
2203 2203 }
2204 2204 }
2205 2205 return (error);
2206 2206 }
2207 2207
2208 2208 /*
2209 2209 * inputs:
2210 2210 * zc_name name of filesystem
2211 2211 * zc_cookie zap cursor
2212 2212 * zc_nvlist_dst_size size of buffer for property nvlist
2213 2213 *
2214 2214 * outputs:
2215 2215 * zc_name name of next snapshot
2216 2216 * zc_objset_stats stats
2217 2217 * zc_nvlist_dst property nvlist
2218 2218 * zc_nvlist_dst_size size of property nvlist
2219 2219 */
2220 2220 static int
2221 2221 zfs_ioc_snapshot_list_next(zfs_cmd_t *zc)
2222 2222 {
2223 2223 objset_t *os;
2224 2224 int error;
2225 2225
2226 2226 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
2227 2227 if (error != 0) {
2228 2228 return (error == ENOENT ? ESRCH : error);
2229 2229 }
2230 2230
2231 2231 /*
2232 2232 * A dataset name of maximum length cannot have any snapshots,
2233 2233 * so exit immediately.
2234 2234 */
2235 2235 if (strlcat(zc->zc_name, "@", sizeof (zc->zc_name)) >= MAXNAMELEN) {
2236 2236 dmu_objset_rele(os, FTAG);
2237 2237 return (SET_ERROR(ESRCH));
2238 2238 }
2239 2239
2240 2240 error = dmu_snapshot_list_next(os,
2241 2241 sizeof (zc->zc_name) - strlen(zc->zc_name),
2242 2242 zc->zc_name + strlen(zc->zc_name), &zc->zc_obj, &zc->zc_cookie,
2243 2243 NULL);
2244 2244
2245 2245 if (error == 0) {
2246 2246 dsl_dataset_t *ds;
2247 2247 dsl_pool_t *dp = os->os_dsl_dataset->ds_dir->dd_pool;
2248 2248
2249 2249 error = dsl_dataset_hold_obj(dp, zc->zc_obj, FTAG, &ds);
2250 2250 if (error == 0) {
2251 2251 objset_t *ossnap;
2252 2252
2253 2253 error = dmu_objset_from_ds(ds, &ossnap);
2254 2254 if (error == 0)
2255 2255 error = zfs_ioc_objset_stats_impl(zc, ossnap);
2256 2256 dsl_dataset_rele(ds, FTAG);
2257 2257 }
2258 2258 } else if (error == ENOENT) {
2259 2259 error = SET_ERROR(ESRCH);
2260 2260 }
2261 2261
2262 2262 dmu_objset_rele(os, FTAG);
2263 2263 /* if we failed, undo the @ that we tacked on to zc_name */
2264 2264 if (error != 0)
2265 2265 *strchr(zc->zc_name, '@') = '\0';
2266 2266 return (error);
2267 2267 }
2268 2268
2269 2269 static int
2270 2270 zfs_prop_set_userquota(const char *dsname, nvpair_t *pair)
2271 2271 {
2272 2272 const char *propname = nvpair_name(pair);
2273 2273 uint64_t *valary;
2274 2274 unsigned int vallen;
2275 2275 const char *domain;
2276 2276 char *dash;
2277 2277 zfs_userquota_prop_t type;
2278 2278 uint64_t rid;
2279 2279 uint64_t quota;
2280 2280 zfsvfs_t *zfsvfs;
2281 2281 int err;
2282 2282
2283 2283 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2284 2284 nvlist_t *attrs;
2285 2285 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
2286 2286 if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2287 2287 &pair) != 0)
2288 2288 return (SET_ERROR(EINVAL));
2289 2289 }
2290 2290
2291 2291 /*
2292 2292 * A correctly constructed propname is encoded as
2293 2293 * userquota@<rid>-<domain>.
2294 2294 */
2295 2295 if ((dash = strchr(propname, '-')) == NULL ||
2296 2296 nvpair_value_uint64_array(pair, &valary, &vallen) != 0 ||
2297 2297 vallen != 3)
2298 2298 return (SET_ERROR(EINVAL));
2299 2299
2300 2300 domain = dash + 1;
2301 2301 type = valary[0];
2302 2302 rid = valary[1];
2303 2303 quota = valary[2];
2304 2304
2305 2305 err = zfsvfs_hold(dsname, FTAG, &zfsvfs, B_FALSE);
2306 2306 if (err == 0) {
2307 2307 err = zfs_set_userquota(zfsvfs, type, domain, rid, quota);
2308 2308 zfsvfs_rele(zfsvfs, FTAG);
2309 2309 }
2310 2310
2311 2311 return (err);
2312 2312 }
2313 2313
2314 2314 /*
2315 2315 * If the named property is one that has a special function to set its value,
2316 2316 * return 0 on success and a positive error code on failure; otherwise if it is
2317 2317 * not one of the special properties handled by this function, return -1.
2318 2318 *
2319 2319 * XXX: It would be better for callers of the property interface if we handled
2320 2320 * these special cases in dsl_prop.c (in the dsl layer).
2321 2321 */
2322 2322 static int
2323 2323 zfs_prop_set_special(const char *dsname, zprop_source_t source,
2324 2324 nvpair_t *pair)
2325 2325 {
2326 2326 const char *propname = nvpair_name(pair);
2327 2327 zfs_prop_t prop = zfs_name_to_prop(propname);
2328 2328 uint64_t intval;
2329 2329 int err;
2330 2330
2331 2331 if (prop == ZPROP_INVAL) {
2332 2332 if (zfs_prop_userquota(propname))
2333 2333 return (zfs_prop_set_userquota(dsname, pair));
2334 2334 return (-1);
2335 2335 }
2336 2336
2337 2337 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2338 2338 nvlist_t *attrs;
2339 2339 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
2340 2340 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2341 2341 &pair) == 0);
2342 2342 }
2343 2343
2344 2344 if (zfs_prop_get_type(prop) == PROP_TYPE_STRING)
2345 2345 return (-1);
2346 2346
2347 2347 VERIFY(0 == nvpair_value_uint64(pair, &intval));
2348 2348
2349 2349 switch (prop) {
2350 2350 case ZFS_PROP_QUOTA:
2351 2351 err = dsl_dir_set_quota(dsname, source, intval);
2352 2352 break;
2353 2353 case ZFS_PROP_REFQUOTA:
2354 2354 err = dsl_dataset_set_refquota(dsname, source, intval);
2355 2355 break;
2356 2356 case ZFS_PROP_RESERVATION:
2357 2357 err = dsl_dir_set_reservation(dsname, source, intval);
2358 2358 break;
2359 2359 case ZFS_PROP_REFRESERVATION:
2360 2360 err = dsl_dataset_set_refreservation(dsname, source, intval);
2361 2361 break;
2362 2362 case ZFS_PROP_VOLSIZE:
2363 2363 err = zvol_set_volsize(dsname, intval);
2364 2364 break;
2365 2365 case ZFS_PROP_VERSION:
2366 2366 {
2367 2367 zfsvfs_t *zfsvfs;
2368 2368
2369 2369 if ((err = zfsvfs_hold(dsname, FTAG, &zfsvfs, B_TRUE)) != 0)
2370 2370 break;
2371 2371
2372 2372 err = zfs_set_version(zfsvfs, intval);
2373 2373 zfsvfs_rele(zfsvfs, FTAG);
2374 2374
2375 2375 if (err == 0 && intval >= ZPL_VERSION_USERSPACE) {
2376 2376 zfs_cmd_t *zc;
2377 2377
2378 2378 zc = kmem_zalloc(sizeof (zfs_cmd_t), KM_SLEEP);
2379 2379 (void) strcpy(zc->zc_name, dsname);
2380 2380 (void) zfs_ioc_userspace_upgrade(zc);
2381 2381 kmem_free(zc, sizeof (zfs_cmd_t));
2382 2382 }
2383 2383 break;
2384 2384 }
2385 2385 case ZFS_PROP_COMPRESSION:
2386 2386 {
2387 2387 if (intval == ZIO_COMPRESS_LZ4) {
2388 2388 zfeature_info_t *feature =
2389 2389 &spa_feature_table[SPA_FEATURE_LZ4_COMPRESS];
2390 2390 spa_t *spa;
2391 2391
2392 2392 if ((err = spa_open(dsname, &spa, FTAG)) != 0)
2393 2393 return (err);
2394 2394
2395 2395 /*
2396 2396 * Setting the LZ4 compression algorithm activates
2397 2397 * the feature.
2398 2398 */
2399 2399 if (!spa_feature_is_active(spa, feature)) {
2400 2400 if ((err = zfs_prop_activate_feature(spa,
2401 2401 feature)) != 0) {
2402 2402 spa_close(spa, FTAG);
2403 2403 return (err);
2404 2404 }
2405 2405 }
2406 2406
2407 2407 spa_close(spa, FTAG);
2408 2408 }
2409 2409 /*
2410 2410 * We still want the default set action to be performed in the
2411 2411 * caller, we only performed zfeature settings here.
2412 2412 */
2413 2413 err = -1;
2414 2414 break;
2415 2415 }
2416 2416
2417 2417 default:
2418 2418 err = -1;
2419 2419 }
2420 2420
2421 2421 return (err);
2422 2422 }
2423 2423
2424 2424 /*
2425 2425 * This function is best effort. If it fails to set any of the given properties,
2426 2426 * it continues to set as many as it can and returns the last error
2427 2427 * encountered. If the caller provides a non-NULL errlist, it will be filled in
2428 2428 * with the list of names of all the properties that failed along with the
2429 2429 * corresponding error numbers.
2430 2430 *
2431 2431 * If every property is set successfully, zero is returned and errlist is not
2432 2432 * modified.
2433 2433 */
2434 2434 int
2435 2435 zfs_set_prop_nvlist(const char *dsname, zprop_source_t source, nvlist_t *nvl,
2436 2436 nvlist_t *errlist)
2437 2437 {
2438 2438 nvpair_t *pair;
2439 2439 nvpair_t *propval;
2440 2440 int rv = 0;
2441 2441 uint64_t intval;
2442 2442 char *strval;
2443 2443 nvlist_t *genericnvl = fnvlist_alloc();
2444 2444 nvlist_t *retrynvl = fnvlist_alloc();
2445 2445
2446 2446 retry:
2447 2447 pair = NULL;
2448 2448 while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) {
2449 2449 const char *propname = nvpair_name(pair);
2450 2450 zfs_prop_t prop = zfs_name_to_prop(propname);
2451 2451 int err = 0;
2452 2452
2453 2453 /* decode the property value */
2454 2454 propval = pair;
2455 2455 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2456 2456 nvlist_t *attrs;
2457 2457 attrs = fnvpair_value_nvlist(pair);
2458 2458 if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2459 2459 &propval) != 0)
2460 2460 err = SET_ERROR(EINVAL);
2461 2461 }
2462 2462
2463 2463 /* Validate value type */
2464 2464 if (err == 0 && prop == ZPROP_INVAL) {
2465 2465 if (zfs_prop_user(propname)) {
2466 2466 if (nvpair_type(propval) != DATA_TYPE_STRING)
2467 2467 err = SET_ERROR(EINVAL);
2468 2468 } else if (zfs_prop_userquota(propname)) {
2469 2469 if (nvpair_type(propval) !=
2470 2470 DATA_TYPE_UINT64_ARRAY)
2471 2471 err = SET_ERROR(EINVAL);
2472 2472 } else {
2473 2473 err = SET_ERROR(EINVAL);
2474 2474 }
2475 2475 } else if (err == 0) {
2476 2476 if (nvpair_type(propval) == DATA_TYPE_STRING) {
2477 2477 if (zfs_prop_get_type(prop) != PROP_TYPE_STRING)
2478 2478 err = SET_ERROR(EINVAL);
2479 2479 } else if (nvpair_type(propval) == DATA_TYPE_UINT64) {
2480 2480 const char *unused;
2481 2481
2482 2482 intval = fnvpair_value_uint64(propval);
2483 2483
2484 2484 switch (zfs_prop_get_type(prop)) {
2485 2485 case PROP_TYPE_NUMBER:
2486 2486 break;
2487 2487 case PROP_TYPE_STRING:
2488 2488 err = SET_ERROR(EINVAL);
2489 2489 break;
2490 2490 case PROP_TYPE_INDEX:
2491 2491 if (zfs_prop_index_to_string(prop,
2492 2492 intval, &unused) != 0)
2493 2493 err = SET_ERROR(EINVAL);
2494 2494 break;
2495 2495 default:
2496 2496 cmn_err(CE_PANIC,
2497 2497 "unknown property type");
2498 2498 }
2499 2499 } else {
2500 2500 err = SET_ERROR(EINVAL);
2501 2501 }
2502 2502 }
2503 2503
2504 2504 /* Validate permissions */
2505 2505 if (err == 0)
2506 2506 err = zfs_check_settable(dsname, pair, CRED());
2507 2507
2508 2508 if (err == 0) {
2509 2509 err = zfs_prop_set_special(dsname, source, pair);
2510 2510 if (err == -1) {
2511 2511 /*
2512 2512 * For better performance we build up a list of
2513 2513 * properties to set in a single transaction.
2514 2514 */
2515 2515 err = nvlist_add_nvpair(genericnvl, pair);
2516 2516 } else if (err != 0 && nvl != retrynvl) {
2517 2517 /*
2518 2518 * This may be a spurious error caused by
2519 2519 * receiving quota and reservation out of order.
2520 2520 * Try again in a second pass.
2521 2521 */
2522 2522 err = nvlist_add_nvpair(retrynvl, pair);
2523 2523 }
2524 2524 }
2525 2525
2526 2526 if (err != 0) {
2527 2527 if (errlist != NULL)
2528 2528 fnvlist_add_int32(errlist, propname, err);
2529 2529 rv = err;
2530 2530 }
2531 2531 }
2532 2532
2533 2533 if (nvl != retrynvl && !nvlist_empty(retrynvl)) {
2534 2534 nvl = retrynvl;
2535 2535 goto retry;
2536 2536 }
2537 2537
2538 2538 if (!nvlist_empty(genericnvl) &&
2539 2539 dsl_props_set(dsname, source, genericnvl) != 0) {
2540 2540 /*
2541 2541 * If this fails, we still want to set as many properties as we
2542 2542 * can, so try setting them individually.
2543 2543 */
2544 2544 pair = NULL;
2545 2545 while ((pair = nvlist_next_nvpair(genericnvl, pair)) != NULL) {
2546 2546 const char *propname = nvpair_name(pair);
2547 2547 int err = 0;
2548 2548
2549 2549 propval = pair;
2550 2550 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2551 2551 nvlist_t *attrs;
2552 2552 attrs = fnvpair_value_nvlist(pair);
2553 2553 propval = fnvlist_lookup_nvpair(attrs,
2554 2554 ZPROP_VALUE);
2555 2555 }
2556 2556
2557 2557 if (nvpair_type(propval) == DATA_TYPE_STRING) {
2558 2558 strval = fnvpair_value_string(propval);
2559 2559 err = dsl_prop_set_string(dsname, propname,
2560 2560 source, strval);
2561 2561 } else {
2562 2562 intval = fnvpair_value_uint64(propval);
2563 2563 err = dsl_prop_set_int(dsname, propname, source,
2564 2564 intval);
2565 2565 }
2566 2566
2567 2567 if (err != 0) {
2568 2568 if (errlist != NULL) {
2569 2569 fnvlist_add_int32(errlist, propname,
2570 2570 err);
2571 2571 }
2572 2572 rv = err;
2573 2573 }
2574 2574 }
2575 2575 }
2576 2576 nvlist_free(genericnvl);
2577 2577 nvlist_free(retrynvl);
2578 2578
2579 2579 return (rv);
2580 2580 }
2581 2581
2582 2582 /*
2583 2583 * Check that all the properties are valid user properties.
2584 2584 */
2585 2585 static int
2586 2586 zfs_check_userprops(const char *fsname, nvlist_t *nvl)
2587 2587 {
2588 2588 nvpair_t *pair = NULL;
2589 2589 int error = 0;
2590 2590
2591 2591 while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) {
2592 2592 const char *propname = nvpair_name(pair);
2593 2593 char *valstr;
2594 2594
2595 2595 if (!zfs_prop_user(propname) ||
2596 2596 nvpair_type(pair) != DATA_TYPE_STRING)
2597 2597 return (SET_ERROR(EINVAL));
2598 2598
2599 2599 if (error = zfs_secpolicy_write_perms(fsname,
2600 2600 ZFS_DELEG_PERM_USERPROP, CRED()))
2601 2601 return (error);
2602 2602
2603 2603 if (strlen(propname) >= ZAP_MAXNAMELEN)
2604 2604 return (SET_ERROR(ENAMETOOLONG));
2605 2605
2606 2606 VERIFY(nvpair_value_string(pair, &valstr) == 0);
2607 2607 if (strlen(valstr) >= ZAP_MAXVALUELEN)
2608 2608 return (E2BIG);
2609 2609 }
2610 2610 return (0);
2611 2611 }
2612 2612
2613 2613 static void
2614 2614 props_skip(nvlist_t *props, nvlist_t *skipped, nvlist_t **newprops)
2615 2615 {
2616 2616 nvpair_t *pair;
2617 2617
2618 2618 VERIFY(nvlist_alloc(newprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2619 2619
2620 2620 pair = NULL;
2621 2621 while ((pair = nvlist_next_nvpair(props, pair)) != NULL) {
2622 2622 if (nvlist_exists(skipped, nvpair_name(pair)))
2623 2623 continue;
2624 2624
2625 2625 VERIFY(nvlist_add_nvpair(*newprops, pair) == 0);
2626 2626 }
2627 2627 }
2628 2628
2629 2629 static int
2630 2630 clear_received_props(const char *dsname, nvlist_t *props,
2631 2631 nvlist_t *skipped)
2632 2632 {
2633 2633 int err = 0;
2634 2634 nvlist_t *cleared_props = NULL;
2635 2635 props_skip(props, skipped, &cleared_props);
2636 2636 if (!nvlist_empty(cleared_props)) {
2637 2637 /*
2638 2638 * Acts on local properties until the dataset has received
2639 2639 * properties at least once on or after SPA_VERSION_RECVD_PROPS.
2640 2640 */
2641 2641 zprop_source_t flags = (ZPROP_SRC_NONE |
2642 2642 (dsl_prop_get_hasrecvd(dsname) ? ZPROP_SRC_RECEIVED : 0));
2643 2643 err = zfs_set_prop_nvlist(dsname, flags, cleared_props, NULL);
2644 2644 }
2645 2645 nvlist_free(cleared_props);
2646 2646 return (err);
2647 2647 }
2648 2648
2649 2649 /*
2650 2650 * inputs:
2651 2651 * zc_name name of filesystem
2652 2652 * zc_value name of property to set
2653 2653 * zc_nvlist_src{_size} nvlist of properties to apply
2654 2654 * zc_cookie received properties flag
2655 2655 *
2656 2656 * outputs:
2657 2657 * zc_nvlist_dst{_size} error for each unapplied received property
2658 2658 */
2659 2659 static int
2660 2660 zfs_ioc_set_prop(zfs_cmd_t *zc)
2661 2661 {
2662 2662 nvlist_t *nvl;
2663 2663 boolean_t received = zc->zc_cookie;
2664 2664 zprop_source_t source = (received ? ZPROP_SRC_RECEIVED :
2665 2665 ZPROP_SRC_LOCAL);
2666 2666 nvlist_t *errors;
2667 2667 int error;
2668 2668
2669 2669 if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2670 2670 zc->zc_iflags, &nvl)) != 0)
2671 2671 return (error);
2672 2672
2673 2673 if (received) {
2674 2674 nvlist_t *origprops;
2675 2675
2676 2676 if (dsl_prop_get_received(zc->zc_name, &origprops) == 0) {
2677 2677 (void) clear_received_props(zc->zc_name,
2678 2678 origprops, nvl);
2679 2679 nvlist_free(origprops);
2680 2680 }
2681 2681
2682 2682 error = dsl_prop_set_hasrecvd(zc->zc_name);
2683 2683 }
2684 2684
2685 2685 errors = fnvlist_alloc();
2686 2686 if (error == 0)
2687 2687 error = zfs_set_prop_nvlist(zc->zc_name, source, nvl, errors);
2688 2688
2689 2689 if (zc->zc_nvlist_dst != NULL && errors != NULL) {
2690 2690 (void) put_nvlist(zc, errors);
2691 2691 }
2692 2692
2693 2693 nvlist_free(errors);
2694 2694 nvlist_free(nvl);
2695 2695 return (error);
2696 2696 }
2697 2697
2698 2698 /*
2699 2699 * inputs:
2700 2700 * zc_name name of filesystem
2701 2701 * zc_value name of property to inherit
2702 2702 * zc_cookie revert to received value if TRUE
2703 2703 *
2704 2704 * outputs: none
2705 2705 */
2706 2706 static int
2707 2707 zfs_ioc_inherit_prop(zfs_cmd_t *zc)
2708 2708 {
2709 2709 const char *propname = zc->zc_value;
2710 2710 zfs_prop_t prop = zfs_name_to_prop(propname);
2711 2711 boolean_t received = zc->zc_cookie;
2712 2712 zprop_source_t source = (received
2713 2713 ? ZPROP_SRC_NONE /* revert to received value, if any */
2714 2714 : ZPROP_SRC_INHERITED); /* explicitly inherit */
2715 2715
2716 2716 if (received) {
2717 2717 nvlist_t *dummy;
2718 2718 nvpair_t *pair;
2719 2719 zprop_type_t type;
2720 2720 int err;
2721 2721
2722 2722 /*
2723 2723 * zfs_prop_set_special() expects properties in the form of an
2724 2724 * nvpair with type info.
2725 2725 */
2726 2726 if (prop == ZPROP_INVAL) {
2727 2727 if (!zfs_prop_user(propname))
2728 2728 return (SET_ERROR(EINVAL));
2729 2729
2730 2730 type = PROP_TYPE_STRING;
2731 2731 } else if (prop == ZFS_PROP_VOLSIZE ||
2732 2732 prop == ZFS_PROP_VERSION) {
2733 2733 return (SET_ERROR(EINVAL));
2734 2734 } else {
2735 2735 type = zfs_prop_get_type(prop);
2736 2736 }
2737 2737
2738 2738 VERIFY(nvlist_alloc(&dummy, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2739 2739
2740 2740 switch (type) {
2741 2741 case PROP_TYPE_STRING:
2742 2742 VERIFY(0 == nvlist_add_string(dummy, propname, ""));
2743 2743 break;
2744 2744 case PROP_TYPE_NUMBER:
2745 2745 case PROP_TYPE_INDEX:
2746 2746 VERIFY(0 == nvlist_add_uint64(dummy, propname, 0));
2747 2747 break;
2748 2748 default:
2749 2749 nvlist_free(dummy);
2750 2750 return (SET_ERROR(EINVAL));
2751 2751 }
2752 2752
2753 2753 pair = nvlist_next_nvpair(dummy, NULL);
2754 2754 err = zfs_prop_set_special(zc->zc_name, source, pair);
2755 2755 nvlist_free(dummy);
2756 2756 if (err != -1)
2757 2757 return (err); /* special property already handled */
2758 2758 } else {
2759 2759 /*
2760 2760 * Only check this in the non-received case. We want to allow
2761 2761 * 'inherit -S' to revert non-inheritable properties like quota
2762 2762 * and reservation to the received or default values even though
2763 2763 * they are not considered inheritable.
2764 2764 */
2765 2765 if (prop != ZPROP_INVAL && !zfs_prop_inheritable(prop))
2766 2766 return (SET_ERROR(EINVAL));
2767 2767 }
2768 2768
2769 2769 /* property name has been validated by zfs_secpolicy_inherit_prop() */
2770 2770 return (dsl_prop_inherit(zc->zc_name, zc->zc_value, source));
2771 2771 }
2772 2772
2773 2773 static int
2774 2774 zfs_ioc_pool_set_props(zfs_cmd_t *zc)
2775 2775 {
2776 2776 nvlist_t *props;
2777 2777 spa_t *spa;
2778 2778 int error;
2779 2779 nvpair_t *pair;
2780 2780
2781 2781 if (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2782 2782 zc->zc_iflags, &props))
2783 2783 return (error);
2784 2784
2785 2785 /*
2786 2786 * If the only property is the configfile, then just do a spa_lookup()
2787 2787 * to handle the faulted case.
2788 2788 */
2789 2789 pair = nvlist_next_nvpair(props, NULL);
2790 2790 if (pair != NULL && strcmp(nvpair_name(pair),
2791 2791 zpool_prop_to_name(ZPOOL_PROP_CACHEFILE)) == 0 &&
2792 2792 nvlist_next_nvpair(props, pair) == NULL) {
2793 2793 mutex_enter(&spa_namespace_lock);
2794 2794 if ((spa = spa_lookup(zc->zc_name)) != NULL) {
2795 2795 spa_configfile_set(spa, props, B_FALSE);
2796 2796 spa_config_sync(spa, B_FALSE, B_TRUE);
2797 2797 }
2798 2798 mutex_exit(&spa_namespace_lock);
2799 2799 if (spa != NULL) {
2800 2800 nvlist_free(props);
2801 2801 return (0);
2802 2802 }
2803 2803 }
2804 2804
2805 2805 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) {
2806 2806 nvlist_free(props);
2807 2807 return (error);
2808 2808 }
2809 2809
2810 2810 error = spa_prop_set(spa, props);
2811 2811
2812 2812 nvlist_free(props);
2813 2813 spa_close(spa, FTAG);
2814 2814
2815 2815 return (error);
2816 2816 }
2817 2817
2818 2818 static int
2819 2819 zfs_ioc_pool_get_props(zfs_cmd_t *zc)
2820 2820 {
2821 2821 spa_t *spa;
2822 2822 int error;
2823 2823 nvlist_t *nvp = NULL;
2824 2824
2825 2825 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) {
2826 2826 /*
2827 2827 * If the pool is faulted, there may be properties we can still
2828 2828 * get (such as altroot and cachefile), so attempt to get them
2829 2829 * anyway.
2830 2830 */
2831 2831 mutex_enter(&spa_namespace_lock);
2832 2832 if ((spa = spa_lookup(zc->zc_name)) != NULL)
2833 2833 error = spa_prop_get(spa, &nvp);
2834 2834 mutex_exit(&spa_namespace_lock);
2835 2835 } else {
2836 2836 error = spa_prop_get(spa, &nvp);
2837 2837 spa_close(spa, FTAG);
2838 2838 }
2839 2839
2840 2840 if (error == 0 && zc->zc_nvlist_dst != NULL)
2841 2841 error = put_nvlist(zc, nvp);
2842 2842 else
2843 2843 error = SET_ERROR(EFAULT);
2844 2844
2845 2845 nvlist_free(nvp);
2846 2846 return (error);
2847 2847 }
2848 2848
2849 2849 /*
2850 2850 * inputs:
2851 2851 * zc_name name of filesystem
2852 2852 * zc_nvlist_src{_size} nvlist of delegated permissions
2853 2853 * zc_perm_action allow/unallow flag
2854 2854 *
2855 2855 * outputs: none
2856 2856 */
2857 2857 static int
2858 2858 zfs_ioc_set_fsacl(zfs_cmd_t *zc)
2859 2859 {
2860 2860 int error;
2861 2861 nvlist_t *fsaclnv = NULL;
2862 2862
2863 2863 if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2864 2864 zc->zc_iflags, &fsaclnv)) != 0)
2865 2865 return (error);
2866 2866
2867 2867 /*
2868 2868 * Verify nvlist is constructed correctly
2869 2869 */
2870 2870 if ((error = zfs_deleg_verify_nvlist(fsaclnv)) != 0) {
2871 2871 nvlist_free(fsaclnv);
2872 2872 return (SET_ERROR(EINVAL));
2873 2873 }
2874 2874
2875 2875 /*
2876 2876 * If we don't have PRIV_SYS_MOUNT, then validate
2877 2877 * that user is allowed to hand out each permission in
2878 2878 * the nvlist(s)
2879 2879 */
2880 2880
2881 2881 error = secpolicy_zfs(CRED());
2882 2882 if (error != 0) {
2883 2883 if (zc->zc_perm_action == B_FALSE) {
2884 2884 error = dsl_deleg_can_allow(zc->zc_name,
2885 2885 fsaclnv, CRED());
2886 2886 } else {
2887 2887 error = dsl_deleg_can_unallow(zc->zc_name,
2888 2888 fsaclnv, CRED());
2889 2889 }
2890 2890 }
2891 2891
2892 2892 if (error == 0)
2893 2893 error = dsl_deleg_set(zc->zc_name, fsaclnv, zc->zc_perm_action);
2894 2894
2895 2895 nvlist_free(fsaclnv);
2896 2896 return (error);
2897 2897 }
2898 2898
2899 2899 /*
2900 2900 * inputs:
2901 2901 * zc_name name of filesystem
2902 2902 *
2903 2903 * outputs:
2904 2904 * zc_nvlist_src{_size} nvlist of delegated permissions
2905 2905 */
2906 2906 static int
2907 2907 zfs_ioc_get_fsacl(zfs_cmd_t *zc)
2908 2908 {
2909 2909 nvlist_t *nvp;
2910 2910 int error;
2911 2911
2912 2912 if ((error = dsl_deleg_get(zc->zc_name, &nvp)) == 0) {
2913 2913 error = put_nvlist(zc, nvp);
2914 2914 nvlist_free(nvp);
2915 2915 }
2916 2916
2917 2917 return (error);
2918 2918 }
2919 2919
2920 2920 /*
2921 2921 * Search the vfs list for a specified resource. Returns a pointer to it
2922 2922 * or NULL if no suitable entry is found. The caller of this routine
2923 2923 * is responsible for releasing the returned vfs pointer.
2924 2924 */
2925 2925 static vfs_t *
2926 2926 zfs_get_vfs(const char *resource)
2927 2927 {
2928 2928 struct vfs *vfsp;
2929 2929 struct vfs *vfs_found = NULL;
2930 2930
2931 2931 vfs_list_read_lock();
2932 2932 vfsp = rootvfs;
2933 2933 do {
2934 2934 if (strcmp(refstr_value(vfsp->vfs_resource), resource) == 0) {
2935 2935 VFS_HOLD(vfsp);
2936 2936 vfs_found = vfsp;
2937 2937 break;
2938 2938 }
2939 2939 vfsp = vfsp->vfs_next;
2940 2940 } while (vfsp != rootvfs);
2941 2941 vfs_list_unlock();
2942 2942 return (vfs_found);
2943 2943 }
2944 2944
2945 2945 /* ARGSUSED */
2946 2946 static void
2947 2947 zfs_create_cb(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx)
2948 2948 {
2949 2949 zfs_creat_t *zct = arg;
2950 2950
2951 2951 zfs_create_fs(os, cr, zct->zct_zplprops, tx);
2952 2952 }
2953 2953
2954 2954 #define ZFS_PROP_UNDEFINED ((uint64_t)-1)
2955 2955
2956 2956 /*
2957 2957 * inputs:
2958 2958 * createprops list of properties requested by creator
2959 2959 * default_zplver zpl version to use if unspecified in createprops
2960 2960 * fuids_ok fuids allowed in this version of the spa?
2961 2961 * os parent objset pointer (NULL if root fs)
2962 2962 *
2963 2963 * outputs:
2964 2964 * zplprops values for the zplprops we attach to the master node object
2965 2965 * is_ci true if requested file system will be purely case-insensitive
2966 2966 *
2967 2967 * Determine the settings for utf8only, normalization and
2968 2968 * casesensitivity. Specific values may have been requested by the
2969 2969 * creator and/or we can inherit values from the parent dataset. If
2970 2970 * the file system is of too early a vintage, a creator can not
2971 2971 * request settings for these properties, even if the requested
2972 2972 * setting is the default value. We don't actually want to create dsl
2973 2973 * properties for these, so remove them from the source nvlist after
2974 2974 * processing.
2975 2975 */
2976 2976 static int
2977 2977 zfs_fill_zplprops_impl(objset_t *os, uint64_t zplver,
2978 2978 boolean_t fuids_ok, boolean_t sa_ok, nvlist_t *createprops,
2979 2979 nvlist_t *zplprops, boolean_t *is_ci)
2980 2980 {
2981 2981 uint64_t sense = ZFS_PROP_UNDEFINED;
2982 2982 uint64_t norm = ZFS_PROP_UNDEFINED;
2983 2983 uint64_t u8 = ZFS_PROP_UNDEFINED;
2984 2984
2985 2985 ASSERT(zplprops != NULL);
2986 2986
2987 2987 /*
2988 2988 * Pull out creator prop choices, if any.
2989 2989 */
2990 2990 if (createprops) {
2991 2991 (void) nvlist_lookup_uint64(createprops,
2992 2992 zfs_prop_to_name(ZFS_PROP_VERSION), &zplver);
2993 2993 (void) nvlist_lookup_uint64(createprops,
2994 2994 zfs_prop_to_name(ZFS_PROP_NORMALIZE), &norm);
2995 2995 (void) nvlist_remove_all(createprops,
2996 2996 zfs_prop_to_name(ZFS_PROP_NORMALIZE));
2997 2997 (void) nvlist_lookup_uint64(createprops,
2998 2998 zfs_prop_to_name(ZFS_PROP_UTF8ONLY), &u8);
2999 2999 (void) nvlist_remove_all(createprops,
3000 3000 zfs_prop_to_name(ZFS_PROP_UTF8ONLY));
3001 3001 (void) nvlist_lookup_uint64(createprops,
3002 3002 zfs_prop_to_name(ZFS_PROP_CASE), &sense);
3003 3003 (void) nvlist_remove_all(createprops,
3004 3004 zfs_prop_to_name(ZFS_PROP_CASE));
3005 3005 }
3006 3006
3007 3007 /*
3008 3008 * If the zpl version requested is whacky or the file system
3009 3009 * or pool is version is too "young" to support normalization
3010 3010 * and the creator tried to set a value for one of the props,
3011 3011 * error out.
3012 3012 */
3013 3013 if ((zplver < ZPL_VERSION_INITIAL || zplver > ZPL_VERSION) ||
3014 3014 (zplver >= ZPL_VERSION_FUID && !fuids_ok) ||
3015 3015 (zplver >= ZPL_VERSION_SA && !sa_ok) ||
3016 3016 (zplver < ZPL_VERSION_NORMALIZATION &&
3017 3017 (norm != ZFS_PROP_UNDEFINED || u8 != ZFS_PROP_UNDEFINED ||
3018 3018 sense != ZFS_PROP_UNDEFINED)))
3019 3019 return (SET_ERROR(ENOTSUP));
3020 3020
3021 3021 /*
3022 3022 * Put the version in the zplprops
3023 3023 */
3024 3024 VERIFY(nvlist_add_uint64(zplprops,
3025 3025 zfs_prop_to_name(ZFS_PROP_VERSION), zplver) == 0);
3026 3026
3027 3027 if (norm == ZFS_PROP_UNDEFINED)
3028 3028 VERIFY(zfs_get_zplprop(os, ZFS_PROP_NORMALIZE, &norm) == 0);
3029 3029 VERIFY(nvlist_add_uint64(zplprops,
3030 3030 zfs_prop_to_name(ZFS_PROP_NORMALIZE), norm) == 0);
3031 3031
3032 3032 /*
3033 3033 * If we're normalizing, names must always be valid UTF-8 strings.
3034 3034 */
3035 3035 if (norm)
3036 3036 u8 = 1;
3037 3037 if (u8 == ZFS_PROP_UNDEFINED)
3038 3038 VERIFY(zfs_get_zplprop(os, ZFS_PROP_UTF8ONLY, &u8) == 0);
3039 3039 VERIFY(nvlist_add_uint64(zplprops,
3040 3040 zfs_prop_to_name(ZFS_PROP_UTF8ONLY), u8) == 0);
3041 3041
3042 3042 if (sense == ZFS_PROP_UNDEFINED)
3043 3043 VERIFY(zfs_get_zplprop(os, ZFS_PROP_CASE, &sense) == 0);
3044 3044 VERIFY(nvlist_add_uint64(zplprops,
3045 3045 zfs_prop_to_name(ZFS_PROP_CASE), sense) == 0);
3046 3046
3047 3047 if (is_ci)
3048 3048 *is_ci = (sense == ZFS_CASE_INSENSITIVE);
3049 3049
3050 3050 return (0);
3051 3051 }
3052 3052
3053 3053 static int
3054 3054 zfs_fill_zplprops(const char *dataset, nvlist_t *createprops,
3055 3055 nvlist_t *zplprops, boolean_t *is_ci)
3056 3056 {
3057 3057 boolean_t fuids_ok, sa_ok;
3058 3058 uint64_t zplver = ZPL_VERSION;
3059 3059 objset_t *os = NULL;
3060 3060 char parentname[MAXNAMELEN];
3061 3061 char *cp;
3062 3062 spa_t *spa;
3063 3063 uint64_t spa_vers;
3064 3064 int error;
3065 3065
3066 3066 (void) strlcpy(parentname, dataset, sizeof (parentname));
3067 3067 cp = strrchr(parentname, '/');
3068 3068 ASSERT(cp != NULL);
3069 3069 cp[0] = '\0';
3070 3070
3071 3071 if ((error = spa_open(dataset, &spa, FTAG)) != 0)
3072 3072 return (error);
3073 3073
3074 3074 spa_vers = spa_version(spa);
3075 3075 spa_close(spa, FTAG);
3076 3076
3077 3077 zplver = zfs_zpl_version_map(spa_vers);
3078 3078 fuids_ok = (zplver >= ZPL_VERSION_FUID);
3079 3079 sa_ok = (zplver >= ZPL_VERSION_SA);
3080 3080
3081 3081 /*
3082 3082 * Open parent object set so we can inherit zplprop values.
3083 3083 */
3084 3084 if ((error = dmu_objset_hold(parentname, FTAG, &os)) != 0)
3085 3085 return (error);
3086 3086
3087 3087 error = zfs_fill_zplprops_impl(os, zplver, fuids_ok, sa_ok, createprops,
3088 3088 zplprops, is_ci);
3089 3089 dmu_objset_rele(os, FTAG);
3090 3090 return (error);
3091 3091 }
3092 3092
3093 3093 static int
3094 3094 zfs_fill_zplprops_root(uint64_t spa_vers, nvlist_t *createprops,
3095 3095 nvlist_t *zplprops, boolean_t *is_ci)
3096 3096 {
3097 3097 boolean_t fuids_ok;
3098 3098 boolean_t sa_ok;
3099 3099 uint64_t zplver = ZPL_VERSION;
3100 3100 int error;
3101 3101
3102 3102 zplver = zfs_zpl_version_map(spa_vers);
3103 3103 fuids_ok = (zplver >= ZPL_VERSION_FUID);
3104 3104 sa_ok = (zplver >= ZPL_VERSION_SA);
3105 3105
3106 3106 error = zfs_fill_zplprops_impl(NULL, zplver, fuids_ok, sa_ok,
3107 3107 createprops, zplprops, is_ci);
3108 3108 return (error);
3109 3109 }
3110 3110
3111 3111 /*
3112 3112 * innvl: {
3113 3113 * "type" -> dmu_objset_type_t (int32)
3114 3114 * (optional) "props" -> { prop -> value }
3115 3115 * }
3116 3116 *
3117 3117 * outnvl: propname -> error code (int32)
3118 3118 */
3119 3119 static int
3120 3120 zfs_ioc_create(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3121 3121 {
3122 3122 int error = 0;
3123 3123 zfs_creat_t zct = { 0 };
3124 3124 nvlist_t *nvprops = NULL;
3125 3125 void (*cbfunc)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx);
3126 3126 int32_t type32;
3127 3127 dmu_objset_type_t type;
3128 3128 boolean_t is_insensitive = B_FALSE;
3129 3129
3130 3130 if (nvlist_lookup_int32(innvl, "type", &type32) != 0)
3131 3131 return (SET_ERROR(EINVAL));
3132 3132 type = type32;
3133 3133 (void) nvlist_lookup_nvlist(innvl, "props", &nvprops);
3134 3134
3135 3135 switch (type) {
3136 3136 case DMU_OST_ZFS:
3137 3137 cbfunc = zfs_create_cb;
3138 3138 break;
3139 3139
3140 3140 case DMU_OST_ZVOL:
3141 3141 cbfunc = zvol_create_cb;
3142 3142 break;
3143 3143
3144 3144 default:
3145 3145 cbfunc = NULL;
3146 3146 break;
3147 3147 }
3148 3148 if (strchr(fsname, '@') ||
3149 3149 strchr(fsname, '%'))
3150 3150 return (SET_ERROR(EINVAL));
3151 3151
3152 3152 zct.zct_props = nvprops;
3153 3153
3154 3154 if (cbfunc == NULL)
3155 3155 return (SET_ERROR(EINVAL));
3156 3156
3157 3157 if (type == DMU_OST_ZVOL) {
3158 3158 uint64_t volsize, volblocksize;
3159 3159
3160 3160 if (nvprops == NULL)
3161 3161 return (SET_ERROR(EINVAL));
3162 3162 if (nvlist_lookup_uint64(nvprops,
3163 3163 zfs_prop_to_name(ZFS_PROP_VOLSIZE), &volsize) != 0)
3164 3164 return (SET_ERROR(EINVAL));
3165 3165
3166 3166 if ((error = nvlist_lookup_uint64(nvprops,
3167 3167 zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE),
3168 3168 &volblocksize)) != 0 && error != ENOENT)
3169 3169 return (SET_ERROR(EINVAL));
3170 3170
3171 3171 if (error != 0)
3172 3172 volblocksize = zfs_prop_default_numeric(
3173 3173 ZFS_PROP_VOLBLOCKSIZE);
3174 3174
3175 3175 if ((error = zvol_check_volblocksize(
3176 3176 volblocksize)) != 0 ||
3177 3177 (error = zvol_check_volsize(volsize,
3178 3178 volblocksize)) != 0)
3179 3179 return (error);
3180 3180 } else if (type == DMU_OST_ZFS) {
3181 3181 int error;
3182 3182
3183 3183 /*
3184 3184 * We have to have normalization and
3185 3185 * case-folding flags correct when we do the
3186 3186 * file system creation, so go figure them out
3187 3187 * now.
3188 3188 */
3189 3189 VERIFY(nvlist_alloc(&zct.zct_zplprops,
3190 3190 NV_UNIQUE_NAME, KM_SLEEP) == 0);
3191 3191 error = zfs_fill_zplprops(fsname, nvprops,
3192 3192 zct.zct_zplprops, &is_insensitive);
3193 3193 if (error != 0) {
3194 3194 nvlist_free(zct.zct_zplprops);
3195 3195 return (error);
3196 3196 }
3197 3197 }
3198 3198
3199 3199 error = dmu_objset_create(fsname, type,
3200 3200 is_insensitive ? DS_FLAG_CI_DATASET : 0, cbfunc, &zct);
3201 3201 nvlist_free(zct.zct_zplprops);
3202 3202
3203 3203 /*
3204 3204 * It would be nice to do this atomically.
3205 3205 */
3206 3206 if (error == 0) {
3207 3207 error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL,
3208 3208 nvprops, outnvl);
3209 3209 if (error != 0)
3210 3210 (void) dsl_destroy_head(fsname);
3211 3211 }
3212 3212 return (error);
3213 3213 }
3214 3214
3215 3215 /*
3216 3216 * innvl: {
3217 3217 * "origin" -> name of origin snapshot
3218 3218 * (optional) "props" -> { prop -> value }
3219 3219 * }
3220 3220 *
3221 3221 * outnvl: propname -> error code (int32)
3222 3222 */
3223 3223 static int
3224 3224 zfs_ioc_clone(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3225 3225 {
3226 3226 int error = 0;
3227 3227 nvlist_t *nvprops = NULL;
3228 3228 char *origin_name;
3229 3229
3230 3230 if (nvlist_lookup_string(innvl, "origin", &origin_name) != 0)
3231 3231 return (SET_ERROR(EINVAL));
3232 3232 (void) nvlist_lookup_nvlist(innvl, "props", &nvprops);
3233 3233
3234 3234 if (strchr(fsname, '@') ||
3235 3235 strchr(fsname, '%'))
3236 3236 return (SET_ERROR(EINVAL));
3237 3237
3238 3238 if (dataset_namecheck(origin_name, NULL, NULL) != 0)
3239 3239 return (SET_ERROR(EINVAL));
3240 3240 error = dmu_objset_clone(fsname, origin_name);
3241 3241 if (error != 0)
3242 3242 return (error);
3243 3243
3244 3244 /*
3245 3245 * It would be nice to do this atomically.
3246 3246 */
3247 3247 if (error == 0) {
3248 3248 error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL,
3249 3249 nvprops, outnvl);
3250 3250 if (error != 0)
3251 3251 (void) dsl_destroy_head(fsname);
3252 3252 }
3253 3253 return (error);
3254 3254 }
3255 3255
3256 3256 /*
3257 3257 * innvl: {
3258 3258 * "snaps" -> { snapshot1, snapshot2 }
3259 3259 * (optional) "props" -> { prop -> value (string) }
3260 3260 * }
3261 3261 *
3262 3262 * outnvl: snapshot -> error code (int32)
3263 3263 */
3264 3264 static int
3265 3265 zfs_ioc_snapshot(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3266 3266 {
3267 3267 nvlist_t *snaps;
3268 3268 nvlist_t *props = NULL;
3269 3269 int error, poollen;
3270 3270 nvpair_t *pair;
3271 3271
3272 3272 (void) nvlist_lookup_nvlist(innvl, "props", &props);
3273 3273 if ((error = zfs_check_userprops(poolname, props)) != 0)
3274 3274 return (error);
3275 3275
3276 3276 if (!nvlist_empty(props) &&
3277 3277 zfs_earlier_version(poolname, SPA_VERSION_SNAP_PROPS))
3278 3278 return (SET_ERROR(ENOTSUP));
3279 3279
3280 3280 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
3281 3281 return (SET_ERROR(EINVAL));
3282 3282 poollen = strlen(poolname);
3283 3283 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
3284 3284 pair = nvlist_next_nvpair(snaps, pair)) {
3285 3285 const char *name = nvpair_name(pair);
3286 3286 const char *cp = strchr(name, '@');
3287 3287
3288 3288 /*
3289 3289 * The snap name must contain an @, and the part after it must
3290 3290 * contain only valid characters.
3291 3291 */
3292 3292 if (cp == NULL || snapshot_namecheck(cp + 1, NULL, NULL) != 0)
3293 3293 return (SET_ERROR(EINVAL));
3294 3294
3295 3295 /*
3296 3296 * The snap must be in the specified pool.
3297 3297 */
3298 3298 if (strncmp(name, poolname, poollen) != 0 ||
3299 3299 (name[poollen] != '/' && name[poollen] != '@'))
3300 3300 return (SET_ERROR(EXDEV));
3301 3301
3302 3302 /* This must be the only snap of this fs. */
3303 3303 for (nvpair_t *pair2 = nvlist_next_nvpair(snaps, pair);
3304 3304 pair2 != NULL; pair2 = nvlist_next_nvpair(snaps, pair2)) {
3305 3305 if (strncmp(name, nvpair_name(pair2), cp - name + 1)
3306 3306 == 0) {
3307 3307 return (SET_ERROR(EXDEV));
3308 3308 }
3309 3309 }
3310 3310 }
3311 3311
3312 3312 error = dsl_dataset_snapshot(snaps, props, outnvl);
3313 3313 return (error);
3314 3314 }
3315 3315
3316 3316 /*
3317 3317 * innvl: "message" -> string
3318 3318 */
3319 3319 /* ARGSUSED */
3320 3320 static int
3321 3321 zfs_ioc_log_history(const char *unused, nvlist_t *innvl, nvlist_t *outnvl)
3322 3322 {
3323 3323 char *message;
3324 3324 spa_t *spa;
3325 3325 int error;
3326 3326 char *poolname;
3327 3327
3328 3328 /*
3329 3329 * The poolname in the ioctl is not set, we get it from the TSD,
3330 3330 * which was set at the end of the last successful ioctl that allows
3331 3331 * logging. The secpolicy func already checked that it is set.
3332 3332 * Only one log ioctl is allowed after each successful ioctl, so
3333 3333 * we clear the TSD here.
3334 3334 */
3335 3335 poolname = tsd_get(zfs_allow_log_key);
3336 3336 (void) tsd_set(zfs_allow_log_key, NULL);
3337 3337 error = spa_open(poolname, &spa, FTAG);
3338 3338 strfree(poolname);
3339 3339 if (error != 0)
3340 3340 return (error);
3341 3341
3342 3342 if (nvlist_lookup_string(innvl, "message", &message) != 0) {
3343 3343 spa_close(spa, FTAG);
3344 3344 return (SET_ERROR(EINVAL));
3345 3345 }
3346 3346
3347 3347 if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) {
3348 3348 spa_close(spa, FTAG);
3349 3349 return (SET_ERROR(ENOTSUP));
3350 3350 }
3351 3351
3352 3352 error = spa_history_log(spa, message);
3353 3353 spa_close(spa, FTAG);
3354 3354 return (error);
3355 3355 }
3356 3356
3357 3357 /*
3358 3358 * The dp_config_rwlock must not be held when calling this, because the
3359 3359 * unmount may need to write out data.
3360 3360 *
3361 3361 * This function is best-effort. Callers must deal gracefully if it
3362 3362 * remains mounted (or is remounted after this call).
3363 3363 */
3364 3364 void
3365 3365 zfs_unmount_snap(const char *snapname)
3366 3366 {
3367 3367 vfs_t *vfsp;
3368 3368 zfsvfs_t *zfsvfs;
3369 3369
3370 3370 if (strchr(snapname, '@') == NULL)
3371 3371 return;
3372 3372
3373 3373 vfsp = zfs_get_vfs(snapname);
3374 3374 if (vfsp == NULL)
3375 3375 return;
3376 3376
3377 3377 zfsvfs = vfsp->vfs_data;
3378 3378 ASSERT(!dsl_pool_config_held(dmu_objset_pool(zfsvfs->z_os)));
3379 3379
3380 3380 if (vn_vfswlock(vfsp->vfs_vnodecovered) != 0) {
3381 3381 VFS_RELE(vfsp);
3382 3382 return;
3383 3383 }
3384 3384 VFS_RELE(vfsp);
3385 3385
3386 3386 /*
3387 3387 * Always force the unmount for snapshots.
3388 3388 */
3389 3389 (void) dounmount(vfsp, MS_FORCE, kcred);
3390 3390 }
3391 3391
3392 3392 /* ARGSUSED */
3393 3393 static int
3394 3394 zfs_unmount_snap_cb(const char *snapname, void *arg)
3395 3395 {
3396 3396 zfs_unmount_snap(snapname);
3397 3397 return (0);
3398 3398 }
3399 3399
3400 3400 /*
3401 3401 * When a clone is destroyed, its origin may also need to be destroyed,
3402 3402 * in which case it must be unmounted. This routine will do that unmount
3403 3403 * if necessary.
3404 3404 */
3405 3405 void
3406 3406 zfs_destroy_unmount_origin(const char *fsname)
3407 3407 {
3408 3408 int error;
3409 3409 objset_t *os;
3410 3410 dsl_dataset_t *ds;
3411 3411
3412 3412 error = dmu_objset_hold(fsname, FTAG, &os);
3413 3413 if (error != 0)
3414 3414 return;
3415 3415 ds = dmu_objset_ds(os);
3416 3416 if (dsl_dir_is_clone(ds->ds_dir) && DS_IS_DEFER_DESTROY(ds->ds_prev)) {
3417 3417 char originname[MAXNAMELEN];
3418 3418 dsl_dataset_name(ds->ds_prev, originname);
3419 3419 dmu_objset_rele(os, FTAG);
3420 3420 zfs_unmount_snap(originname);
3421 3421 } else {
3422 3422 dmu_objset_rele(os, FTAG);
3423 3423 }
3424 3424 }
3425 3425
3426 3426 /*
3427 3427 * innvl: {
3428 3428 * "snaps" -> { snapshot1, snapshot2 }
3429 3429 * (optional boolean) "defer"
3430 3430 * }
3431 3431 *
3432 3432 * outnvl: snapshot -> error code (int32)
3433 3433 *
3434 3434 */
3435 3435 static int
3436 3436 zfs_ioc_destroy_snaps(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3437 3437 {
3438 3438 int poollen;
3439 3439 nvlist_t *snaps;
3440 3440 nvpair_t *pair;
3441 3441 boolean_t defer;
3442 3442
3443 3443 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
3444 3444 return (SET_ERROR(EINVAL));
3445 3445 defer = nvlist_exists(innvl, "defer");
3446 3446
3447 3447 poollen = strlen(poolname);
3448 3448 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
3449 3449 pair = nvlist_next_nvpair(snaps, pair)) {
3450 3450 const char *name = nvpair_name(pair);
3451 3451
3452 3452 /*
3453 3453 * The snap must be in the specified pool.
3454 3454 */
3455 3455 if (strncmp(name, poolname, poollen) != 0 ||
3456 3456 (name[poollen] != '/' && name[poollen] != '@'))
3457 3457 return (SET_ERROR(EXDEV));
3458 3458
3459 3459 zfs_unmount_snap(name);
3460 3460 }
3461 3461
3462 3462 return (dsl_destroy_snapshots_nvl(snaps, defer, outnvl));
3463 3463 }
3464 3464
3465 3465 /*
3466 3466 * inputs:
3467 3467 * zc_name name of dataset to destroy
3468 3468 * zc_objset_type type of objset
3469 3469 * zc_defer_destroy mark for deferred destroy
3470 3470 *
3471 3471 * outputs: none
3472 3472 */
3473 3473 static int
3474 3474 zfs_ioc_destroy(zfs_cmd_t *zc)
3475 3475 {
3476 3476 int err;
3477 3477 if (strchr(zc->zc_name, '@') && zc->zc_objset_type == DMU_OST_ZFS)
3478 3478 zfs_unmount_snap(zc->zc_name);
3479 3479
3480 3480 if (strchr(zc->zc_name, '@'))
3481 3481 err = dsl_destroy_snapshot(zc->zc_name, zc->zc_defer_destroy);
3482 3482 else
3483 3483 err = dsl_destroy_head(zc->zc_name);
3484 3484 if (zc->zc_objset_type == DMU_OST_ZVOL && err == 0)
3485 3485 (void) zvol_remove_minor(zc->zc_name);
3486 3486 return (err);
3487 3487 }
3488 3488
3489 3489 /*
3490 3490 * inputs:
3491 3491 * zc_name name of dataset to rollback (to most recent snapshot)
3492 3492 *
3493 3493 * outputs: none
3494 3494 */
3495 3495 static int
3496 3496 zfs_ioc_rollback(zfs_cmd_t *zc)
3497 3497 {
3498 3498 zfsvfs_t *zfsvfs;
3499 3499 int error;
3500 3500
3501 3501 if (getzfsvfs(zc->zc_name, &zfsvfs) == 0) {
3502 3502 error = zfs_suspend_fs(zfsvfs);
3503 3503 if (error == 0) {
3504 3504 int resume_err;
3505 3505
3506 3506 error = dsl_dataset_rollback(zc->zc_name);
3507 3507 resume_err = zfs_resume_fs(zfsvfs, zc->zc_name);
3508 3508 error = error ? error : resume_err;
3509 3509 }
3510 3510 VFS_RELE(zfsvfs->z_vfs);
3511 3511 } else {
3512 3512 error = dsl_dataset_rollback(zc->zc_name);
3513 3513 }
3514 3514 return (error);
3515 3515 }
3516 3516
3517 3517 static int
3518 3518 recursive_unmount(const char *fsname, void *arg)
3519 3519 {
3520 3520 const char *snapname = arg;
3521 3521 char fullname[MAXNAMELEN];
3522 3522
3523 3523 (void) snprintf(fullname, sizeof (fullname), "%s@%s", fsname, snapname);
3524 3524 zfs_unmount_snap(fullname);
3525 3525 return (0);
3526 3526 }
3527 3527
3528 3528 /*
3529 3529 * inputs:
3530 3530 * zc_name old name of dataset
3531 3531 * zc_value new name of dataset
3532 3532 * zc_cookie recursive flag (only valid for snapshots)
3533 3533 *
3534 3534 * outputs: none
3535 3535 */
3536 3536 static int
3537 3537 zfs_ioc_rename(zfs_cmd_t *zc)
3538 3538 {
3539 3539 boolean_t recursive = zc->zc_cookie & 1;
3540 3540 char *at;
3541 3541
3542 3542 zc->zc_value[sizeof (zc->zc_value) - 1] = '\0';
3543 3543 if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
3544 3544 strchr(zc->zc_value, '%'))
3545 3545 return (SET_ERROR(EINVAL));
3546 3546
3547 3547 at = strchr(zc->zc_name, '@');
3548 3548 if (at != NULL) {
3549 3549 /* snaps must be in same fs */
3550 3550 if (strncmp(zc->zc_name, zc->zc_value, at - zc->zc_name + 1))
3551 3551 return (SET_ERROR(EXDEV));
3552 3552 *at = '\0';
3553 3553 if (zc->zc_objset_type == DMU_OST_ZFS) {
3554 3554 int error = dmu_objset_find(zc->zc_name,
3555 3555 recursive_unmount, at + 1,
3556 3556 recursive ? DS_FIND_CHILDREN : 0);
3557 3557 if (error != 0)
3558 3558 return (error);
3559 3559 }
3560 3560 return (dsl_dataset_rename_snapshot(zc->zc_name,
3561 3561 at + 1, strchr(zc->zc_value, '@') + 1, recursive));
3562 3562 } else {
3563 3563 if (zc->zc_objset_type == DMU_OST_ZVOL)
3564 3564 (void) zvol_remove_minor(zc->zc_name);
3565 3565 return (dsl_dir_rename(zc->zc_name, zc->zc_value));
3566 3566 }
3567 3567 }
3568 3568
3569 3569 static int
3570 3570 zfs_check_settable(const char *dsname, nvpair_t *pair, cred_t *cr)
3571 3571 {
3572 3572 const char *propname = nvpair_name(pair);
3573 3573 boolean_t issnap = (strchr(dsname, '@') != NULL);
3574 3574 zfs_prop_t prop = zfs_name_to_prop(propname);
3575 3575 uint64_t intval;
3576 3576 int err;
3577 3577
3578 3578 if (prop == ZPROP_INVAL) {
3579 3579 if (zfs_prop_user(propname)) {
3580 3580 if (err = zfs_secpolicy_write_perms(dsname,
3581 3581 ZFS_DELEG_PERM_USERPROP, cr))
3582 3582 return (err);
3583 3583 return (0);
3584 3584 }
3585 3585
3586 3586 if (!issnap && zfs_prop_userquota(propname)) {
3587 3587 const char *perm = NULL;
3588 3588 const char *uq_prefix =
3589 3589 zfs_userquota_prop_prefixes[ZFS_PROP_USERQUOTA];
3590 3590 const char *gq_prefix =
3591 3591 zfs_userquota_prop_prefixes[ZFS_PROP_GROUPQUOTA];
3592 3592
3593 3593 if (strncmp(propname, uq_prefix,
3594 3594 strlen(uq_prefix)) == 0) {
3595 3595 perm = ZFS_DELEG_PERM_USERQUOTA;
3596 3596 } else if (strncmp(propname, gq_prefix,
3597 3597 strlen(gq_prefix)) == 0) {
3598 3598 perm = ZFS_DELEG_PERM_GROUPQUOTA;
3599 3599 } else {
3600 3600 /* USERUSED and GROUPUSED are read-only */
3601 3601 return (SET_ERROR(EINVAL));
3602 3602 }
3603 3603
3604 3604 if (err = zfs_secpolicy_write_perms(dsname, perm, cr))
3605 3605 return (err);
3606 3606 return (0);
3607 3607 }
3608 3608
3609 3609 return (SET_ERROR(EINVAL));
3610 3610 }
3611 3611
3612 3612 if (issnap)
3613 3613 return (SET_ERROR(EINVAL));
3614 3614
3615 3615 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
3616 3616 /*
3617 3617 * dsl_prop_get_all_impl() returns properties in this
3618 3618 * format.
3619 3619 */
3620 3620 nvlist_t *attrs;
3621 3621 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
3622 3622 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
3623 3623 &pair) == 0);
3624 3624 }
3625 3625
3626 3626 /*
3627 3627 * Check that this value is valid for this pool version
3628 3628 */
3629 3629 switch (prop) {
3630 3630 case ZFS_PROP_COMPRESSION:
3631 3631 /*
3632 3632 * If the user specified gzip compression, make sure
3633 3633 * the SPA supports it. We ignore any errors here since
3634 3634 * we'll catch them later.
3635 3635 */
3636 3636 if (nvpair_type(pair) == DATA_TYPE_UINT64 &&
3637 3637 nvpair_value_uint64(pair, &intval) == 0) {
3638 3638 if (intval >= ZIO_COMPRESS_GZIP_1 &&
3639 3639 intval <= ZIO_COMPRESS_GZIP_9 &&
3640 3640 zfs_earlier_version(dsname,
3641 3641 SPA_VERSION_GZIP_COMPRESSION)) {
3642 3642 return (SET_ERROR(ENOTSUP));
3643 3643 }
3644 3644
3645 3645 if (intval == ZIO_COMPRESS_ZLE &&
3646 3646 zfs_earlier_version(dsname,
3647 3647 SPA_VERSION_ZLE_COMPRESSION))
3648 3648 return (SET_ERROR(ENOTSUP));
3649 3649
3650 3650 if (intval == ZIO_COMPRESS_LZ4) {
3651 3651 zfeature_info_t *feature =
3652 3652 &spa_feature_table[
3653 3653 SPA_FEATURE_LZ4_COMPRESS];
3654 3654 spa_t *spa;
3655 3655
3656 3656 if ((err = spa_open(dsname, &spa, FTAG)) != 0)
3657 3657 return (err);
3658 3658
3659 3659 if (!spa_feature_is_enabled(spa, feature)) {
3660 3660 spa_close(spa, FTAG);
3661 3661 return (SET_ERROR(ENOTSUP));
3662 3662 }
3663 3663 spa_close(spa, FTAG);
3664 3664 }
3665 3665
3666 3666 /*
3667 3667 * If this is a bootable dataset then
3668 3668 * verify that the compression algorithm
3669 3669 * is supported for booting. We must return
3670 3670 * something other than ENOTSUP since it
3671 3671 * implies a downrev pool version.
3672 3672 */
3673 3673 if (zfs_is_bootfs(dsname) &&
3674 3674 !BOOTFS_COMPRESS_VALID(intval)) {
3675 3675 return (SET_ERROR(ERANGE));
3676 3676 }
3677 3677 }
3678 3678 break;
3679 3679
3680 3680 case ZFS_PROP_COPIES:
3681 3681 if (zfs_earlier_version(dsname, SPA_VERSION_DITTO_BLOCKS))
3682 3682 return (SET_ERROR(ENOTSUP));
3683 3683 break;
3684 3684
3685 3685 case ZFS_PROP_DEDUP:
3686 3686 if (zfs_earlier_version(dsname, SPA_VERSION_DEDUP))
3687 3687 return (SET_ERROR(ENOTSUP));
3688 3688 break;
3689 3689
3690 3690 case ZFS_PROP_SHARESMB:
3691 3691 if (zpl_earlier_version(dsname, ZPL_VERSION_FUID))
3692 3692 return (SET_ERROR(ENOTSUP));
3693 3693 break;
3694 3694
3695 3695 case ZFS_PROP_ACLINHERIT:
3696 3696 if (nvpair_type(pair) == DATA_TYPE_UINT64 &&
3697 3697 nvpair_value_uint64(pair, &intval) == 0) {
3698 3698 if (intval == ZFS_ACL_PASSTHROUGH_X &&
3699 3699 zfs_earlier_version(dsname,
3700 3700 SPA_VERSION_PASSTHROUGH_X))
3701 3701 return (SET_ERROR(ENOTSUP));
3702 3702 }
3703 3703 break;
3704 3704 }
3705 3705
3706 3706 return (zfs_secpolicy_setprop(dsname, prop, pair, CRED()));
3707 3707 }
3708 3708
3709 3709 /*
3710 3710 * Checks for a race condition to make sure we don't increment a feature flag
3711 3711 * multiple times.
3712 3712 */
3713 3713 static int
3714 3714 zfs_prop_activate_feature_check(void *arg, dmu_tx_t *tx)
3715 3715 {
3716 3716 spa_t *spa = dmu_tx_pool(tx)->dp_spa;
3717 3717 zfeature_info_t *feature = arg;
3718 3718
3719 3719 if (!spa_feature_is_active(spa, feature))
3720 3720 return (0);
3721 3721 else
3722 3722 return (SET_ERROR(EBUSY));
3723 3723 }
3724 3724
3725 3725 /*
3726 3726 * The callback invoked on feature activation in the sync task caused by
3727 3727 * zfs_prop_activate_feature.
3728 3728 */
3729 3729 static void
3730 3730 zfs_prop_activate_feature_sync(void *arg, dmu_tx_t *tx)
3731 3731 {
3732 3732 spa_t *spa = dmu_tx_pool(tx)->dp_spa;
3733 3733 zfeature_info_t *feature = arg;
3734 3734
3735 3735 spa_feature_incr(spa, feature, tx);
3736 3736 }
3737 3737
3738 3738 /*
3739 3739 * Activates a feature on a pool in response to a property setting. This
3740 3740 * creates a new sync task which modifies the pool to reflect the feature
3741 3741 * as being active.
3742 3742 */
3743 3743 static int
3744 3744 zfs_prop_activate_feature(spa_t *spa, zfeature_info_t *feature)
3745 3745 {
3746 3746 int err;
3747 3747
3748 3748 /* EBUSY here indicates that the feature is already active */
3749 3749 err = dsl_sync_task(spa_name(spa),
3750 3750 zfs_prop_activate_feature_check, zfs_prop_activate_feature_sync,
3751 3751 feature, 2);
3752 3752
3753 3753 if (err != 0 && err != EBUSY)
3754 3754 return (err);
3755 3755 else
3756 3756 return (0);
3757 3757 }
3758 3758
3759 3759 /*
3760 3760 * Removes properties from the given props list that fail permission checks
3761 3761 * needed to clear them and to restore them in case of a receive error. For each
3762 3762 * property, make sure we have both set and inherit permissions.
3763 3763 *
3764 3764 * Returns the first error encountered if any permission checks fail. If the
3765 3765 * caller provides a non-NULL errlist, it also gives the complete list of names
3766 3766 * of all the properties that failed a permission check along with the
3767 3767 * corresponding error numbers. The caller is responsible for freeing the
3768 3768 * returned errlist.
3769 3769 *
3770 3770 * If every property checks out successfully, zero is returned and the list
3771 3771 * pointed at by errlist is NULL.
3772 3772 */
3773 3773 static int
3774 3774 zfs_check_clearable(char *dataset, nvlist_t *props, nvlist_t **errlist)
3775 3775 {
3776 3776 zfs_cmd_t *zc;
3777 3777 nvpair_t *pair, *next_pair;
3778 3778 nvlist_t *errors;
3779 3779 int err, rv = 0;
3780 3780
3781 3781 if (props == NULL)
3782 3782 return (0);
3783 3783
3784 3784 VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0);
3785 3785
3786 3786 zc = kmem_alloc(sizeof (zfs_cmd_t), KM_SLEEP);
3787 3787 (void) strcpy(zc->zc_name, dataset);
3788 3788 pair = nvlist_next_nvpair(props, NULL);
3789 3789 while (pair != NULL) {
3790 3790 next_pair = nvlist_next_nvpair(props, pair);
3791 3791
3792 3792 (void) strcpy(zc->zc_value, nvpair_name(pair));
3793 3793 if ((err = zfs_check_settable(dataset, pair, CRED())) != 0 ||
3794 3794 (err = zfs_secpolicy_inherit_prop(zc, NULL, CRED())) != 0) {
3795 3795 VERIFY(nvlist_remove_nvpair(props, pair) == 0);
3796 3796 VERIFY(nvlist_add_int32(errors,
3797 3797 zc->zc_value, err) == 0);
3798 3798 }
3799 3799 pair = next_pair;
3800 3800 }
3801 3801 kmem_free(zc, sizeof (zfs_cmd_t));
3802 3802
3803 3803 if ((pair = nvlist_next_nvpair(errors, NULL)) == NULL) {
3804 3804 nvlist_free(errors);
3805 3805 errors = NULL;
3806 3806 } else {
3807 3807 VERIFY(nvpair_value_int32(pair, &rv) == 0);
3808 3808 }
3809 3809
3810 3810 if (errlist == NULL)
3811 3811 nvlist_free(errors);
3812 3812 else
3813 3813 *errlist = errors;
3814 3814
3815 3815 return (rv);
3816 3816 }
3817 3817
3818 3818 static boolean_t
3819 3819 propval_equals(nvpair_t *p1, nvpair_t *p2)
3820 3820 {
3821 3821 if (nvpair_type(p1) == DATA_TYPE_NVLIST) {
3822 3822 /* dsl_prop_get_all_impl() format */
3823 3823 nvlist_t *attrs;
3824 3824 VERIFY(nvpair_value_nvlist(p1, &attrs) == 0);
3825 3825 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
3826 3826 &p1) == 0);
3827 3827 }
3828 3828
3829 3829 if (nvpair_type(p2) == DATA_TYPE_NVLIST) {
3830 3830 nvlist_t *attrs;
3831 3831 VERIFY(nvpair_value_nvlist(p2, &attrs) == 0);
3832 3832 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
3833 3833 &p2) == 0);
3834 3834 }
3835 3835
3836 3836 if (nvpair_type(p1) != nvpair_type(p2))
3837 3837 return (B_FALSE);
3838 3838
3839 3839 if (nvpair_type(p1) == DATA_TYPE_STRING) {
3840 3840 char *valstr1, *valstr2;
3841 3841
3842 3842 VERIFY(nvpair_value_string(p1, (char **)&valstr1) == 0);
3843 3843 VERIFY(nvpair_value_string(p2, (char **)&valstr2) == 0);
3844 3844 return (strcmp(valstr1, valstr2) == 0);
3845 3845 } else {
3846 3846 uint64_t intval1, intval2;
3847 3847
3848 3848 VERIFY(nvpair_value_uint64(p1, &intval1) == 0);
3849 3849 VERIFY(nvpair_value_uint64(p2, &intval2) == 0);
3850 3850 return (intval1 == intval2);
3851 3851 }
3852 3852 }
3853 3853
3854 3854 /*
3855 3855 * Remove properties from props if they are not going to change (as determined
3856 3856 * by comparison with origprops). Remove them from origprops as well, since we
3857 3857 * do not need to clear or restore properties that won't change.
3858 3858 */
3859 3859 static void
3860 3860 props_reduce(nvlist_t *props, nvlist_t *origprops)
3861 3861 {
3862 3862 nvpair_t *pair, *next_pair;
3863 3863
3864 3864 if (origprops == NULL)
3865 3865 return; /* all props need to be received */
3866 3866
3867 3867 pair = nvlist_next_nvpair(props, NULL);
3868 3868 while (pair != NULL) {
3869 3869 const char *propname = nvpair_name(pair);
3870 3870 nvpair_t *match;
3871 3871
3872 3872 next_pair = nvlist_next_nvpair(props, pair);
3873 3873
3874 3874 if ((nvlist_lookup_nvpair(origprops, propname,
3875 3875 &match) != 0) || !propval_equals(pair, match))
3876 3876 goto next; /* need to set received value */
3877 3877
3878 3878 /* don't clear the existing received value */
3879 3879 (void) nvlist_remove_nvpair(origprops, match);
3880 3880 /* don't bother receiving the property */
3881 3881 (void) nvlist_remove_nvpair(props, pair);
3882 3882 next:
3883 3883 pair = next_pair;
3884 3884 }
3885 3885 }
3886 3886
3887 3887 #ifdef DEBUG
3888 3888 static boolean_t zfs_ioc_recv_inject_err;
3889 3889 #endif
3890 3890
3891 3891 /*
3892 3892 * inputs:
3893 3893 * zc_name name of containing filesystem
3894 3894 * zc_nvlist_src{_size} nvlist of properties to apply
3895 3895 * zc_value name of snapshot to create
3896 3896 * zc_string name of clone origin (if DRR_FLAG_CLONE)
3897 3897 * zc_cookie file descriptor to recv from
3898 3898 * zc_begin_record the BEGIN record of the stream (not byteswapped)
3899 3899 * zc_guid force flag
3900 3900 * zc_cleanup_fd cleanup-on-exit file descriptor
3901 3901 * zc_action_handle handle for this guid/ds mapping (or zero on first call)
3902 3902 *
3903 3903 * outputs:
3904 3904 * zc_cookie number of bytes read
3905 3905 * zc_nvlist_dst{_size} error for each unapplied received property
3906 3906 * zc_obj zprop_errflags_t
3907 3907 * zc_action_handle handle for this guid/ds mapping
3908 3908 */
3909 3909 static int
3910 3910 zfs_ioc_recv(zfs_cmd_t *zc)
3911 3911 {
3912 3912 file_t *fp;
3913 3913 dmu_recv_cookie_t drc;
3914 3914 boolean_t force = (boolean_t)zc->zc_guid;
3915 3915 int fd;
3916 3916 int error = 0;
3917 3917 int props_error = 0;
3918 3918 nvlist_t *errors;
3919 3919 offset_t off;
3920 3920 nvlist_t *props = NULL; /* sent properties */
3921 3921 nvlist_t *origprops = NULL; /* existing properties */
3922 3922 char *origin = NULL;
3923 3923 char *tosnap;
3924 3924 char tofs[ZFS_MAXNAMELEN];
3925 3925 boolean_t first_recvd_props = B_FALSE;
3926 3926
3927 3927 if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
3928 3928 strchr(zc->zc_value, '@') == NULL ||
3929 3929 strchr(zc->zc_value, '%'))
3930 3930 return (SET_ERROR(EINVAL));
3931 3931
3932 3932 (void) strcpy(tofs, zc->zc_value);
3933 3933 tosnap = strchr(tofs, '@');
3934 3934 *tosnap++ = '\0';
3935 3935
3936 3936 if (zc->zc_nvlist_src != NULL &&
3937 3937 (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
3938 3938 zc->zc_iflags, &props)) != 0)
3939 3939 return (error);
3940 3940
3941 3941 fd = zc->zc_cookie;
3942 3942 fp = getf(fd);
3943 3943 if (fp == NULL) {
3944 3944 nvlist_free(props);
3945 3945 return (SET_ERROR(EBADF));
3946 3946 }
3947 3947
3948 3948 VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0);
3949 3949
3950 3950 if (zc->zc_string[0])
3951 3951 origin = zc->zc_string;
3952 3952
3953 3953 error = dmu_recv_begin(tofs, tosnap,
3954 3954 &zc->zc_begin_record, force, origin, &drc);
3955 3955 if (error != 0)
3956 3956 goto out;
3957 3957
3958 3958 /*
3959 3959 * Set properties before we receive the stream so that they are applied
3960 3960 * to the new data. Note that we must call dmu_recv_stream() if
3961 3961 * dmu_recv_begin() succeeds.
3962 3962 */
3963 3963 if (props != NULL && !drc.drc_newfs) {
3964 3964 if (spa_version(dsl_dataset_get_spa(drc.drc_ds)) >=
3965 3965 SPA_VERSION_RECVD_PROPS &&
3966 3966 !dsl_prop_get_hasrecvd(tofs))
3967 3967 first_recvd_props = B_TRUE;
3968 3968
3969 3969 /*
3970 3970 * If new received properties are supplied, they are to
3971 3971 * completely replace the existing received properties, so stash
3972 3972 * away the existing ones.
3973 3973 */
3974 3974 if (dsl_prop_get_received(tofs, &origprops) == 0) {
3975 3975 nvlist_t *errlist = NULL;
3976 3976 /*
3977 3977 * Don't bother writing a property if its value won't
3978 3978 * change (and avoid the unnecessary security checks).
3979 3979 *
3980 3980 * The first receive after SPA_VERSION_RECVD_PROPS is a
3981 3981 * special case where we blow away all local properties
3982 3982 * regardless.
3983 3983 */
3984 3984 if (!first_recvd_props)
3985 3985 props_reduce(props, origprops);
3986 3986 if (zfs_check_clearable(tofs, origprops, &errlist) != 0)
3987 3987 (void) nvlist_merge(errors, errlist, 0);
3988 3988 nvlist_free(errlist);
3989 3989
3990 3990 if (clear_received_props(tofs, origprops,
3991 3991 first_recvd_props ? NULL : props) != 0)
3992 3992 zc->zc_obj |= ZPROP_ERR_NOCLEAR;
3993 3993 } else {
3994 3994 zc->zc_obj |= ZPROP_ERR_NOCLEAR;
3995 3995 }
3996 3996 }
3997 3997
3998 3998 if (props != NULL) {
3999 3999 props_error = dsl_prop_set_hasrecvd(tofs);
4000 4000
4001 4001 if (props_error == 0) {
4002 4002 (void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_RECEIVED,
4003 4003 props, errors);
4004 4004 }
4005 4005 }
4006 4006
4007 4007 if (zc->zc_nvlist_dst_size != 0 &&
4008 4008 (nvlist_smush(errors, zc->zc_nvlist_dst_size) != 0 ||
4009 4009 put_nvlist(zc, errors) != 0)) {
4010 4010 /*
4011 4011 * Caller made zc->zc_nvlist_dst less than the minimum expected
4012 4012 * size or supplied an invalid address.
4013 4013 */
4014 4014 props_error = SET_ERROR(EINVAL);
4015 4015 }
4016 4016
4017 4017 off = fp->f_offset;
4018 4018 error = dmu_recv_stream(&drc, fp->f_vnode, &off, zc->zc_cleanup_fd,
4019 4019 &zc->zc_action_handle);
4020 4020
4021 4021 if (error == 0) {
4022 4022 zfsvfs_t *zfsvfs = NULL;
4023 4023
4024 4024 if (getzfsvfs(tofs, &zfsvfs) == 0) {
4025 4025 /* online recv */
4026 4026 int end_err;
4027 4027
4028 4028 error = zfs_suspend_fs(zfsvfs);
4029 4029 /*
4030 4030 * If the suspend fails, then the recv_end will
4031 4031 * likely also fail, and clean up after itself.
4032 4032 */
4033 4033 end_err = dmu_recv_end(&drc);
4034 4034 if (error == 0)
4035 4035 error = zfs_resume_fs(zfsvfs, tofs);
4036 4036 error = error ? error : end_err;
4037 4037 VFS_RELE(zfsvfs->z_vfs);
4038 4038 } else {
4039 4039 error = dmu_recv_end(&drc);
4040 4040 }
4041 4041 }
4042 4042
4043 4043 zc->zc_cookie = off - fp->f_offset;
4044 4044 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
4045 4045 fp->f_offset = off;
4046 4046
4047 4047 #ifdef DEBUG
4048 4048 if (zfs_ioc_recv_inject_err) {
4049 4049 zfs_ioc_recv_inject_err = B_FALSE;
4050 4050 error = 1;
4051 4051 }
4052 4052 #endif
4053 4053 /*
4054 4054 * On error, restore the original props.
4055 4055 */
4056 4056 if (error != 0 && props != NULL && !drc.drc_newfs) {
4057 4057 if (clear_received_props(tofs, props, NULL) != 0) {
4058 4058 /*
4059 4059 * We failed to clear the received properties.
4060 4060 * Since we may have left a $recvd value on the
4061 4061 * system, we can't clear the $hasrecvd flag.
4062 4062 */
4063 4063 zc->zc_obj |= ZPROP_ERR_NORESTORE;
4064 4064 } else if (first_recvd_props) {
4065 4065 dsl_prop_unset_hasrecvd(tofs);
4066 4066 }
4067 4067
4068 4068 if (origprops == NULL && !drc.drc_newfs) {
4069 4069 /* We failed to stash the original properties. */
4070 4070 zc->zc_obj |= ZPROP_ERR_NORESTORE;
4071 4071 }
4072 4072
4073 4073 /*
4074 4074 * dsl_props_set() will not convert RECEIVED to LOCAL on or
4075 4075 * after SPA_VERSION_RECVD_PROPS, so we need to specify LOCAL
4076 4076 * explictly if we're restoring local properties cleared in the
4077 4077 * first new-style receive.
4078 4078 */
4079 4079 if (origprops != NULL &&
4080 4080 zfs_set_prop_nvlist(tofs, (first_recvd_props ?
4081 4081 ZPROP_SRC_LOCAL : ZPROP_SRC_RECEIVED),
4082 4082 origprops, NULL) != 0) {
4083 4083 /*
4084 4084 * We stashed the original properties but failed to
4085 4085 * restore them.
4086 4086 */
4087 4087 zc->zc_obj |= ZPROP_ERR_NORESTORE;
4088 4088 }
4089 4089 }
4090 4090 out:
4091 4091 nvlist_free(props);
4092 4092 nvlist_free(origprops);
4093 4093 nvlist_free(errors);
4094 4094 releasef(fd);
4095 4095
4096 4096 if (error == 0)
4097 4097 error = props_error;
4098 4098
4099 4099 return (error);
4100 4100 }
4101 4101
4102 4102 /*
4103 4103 * inputs:
4104 4104 * zc_name name of snapshot to send
4105 4105 * zc_cookie file descriptor to send stream to
4106 4106 * zc_obj fromorigin flag (mutually exclusive with zc_fromobj)
4107 4107 * zc_sendobj objsetid of snapshot to send
4108 4108 * zc_fromobj objsetid of incremental fromsnap (may be zero)
4109 4109 * zc_guid if set, estimate size of stream only. zc_cookie is ignored.
4110 4110 * output size in zc_objset_type.
4111 4111 *
4112 4112 * outputs: none
4113 4113 */
4114 4114 static int
4115 4115 zfs_ioc_send(zfs_cmd_t *zc)
4116 4116 {
4117 4117 int error;
4118 4118 offset_t off;
4119 4119 boolean_t estimate = (zc->zc_guid != 0);
4120 4120
4121 4121 if (zc->zc_obj != 0) {
4122 4122 dsl_pool_t *dp;
4123 4123 dsl_dataset_t *tosnap;
4124 4124
4125 4125 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
4126 4126 if (error != 0)
4127 4127 return (error);
4128 4128
4129 4129 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &tosnap);
4130 4130 if (error != 0) {
4131 4131 dsl_pool_rele(dp, FTAG);
4132 4132 return (error);
4133 4133 }
4134 4134
4135 4135 if (dsl_dir_is_clone(tosnap->ds_dir))
4136 4136 zc->zc_fromobj = tosnap->ds_dir->dd_phys->dd_origin_obj;
4137 4137 dsl_dataset_rele(tosnap, FTAG);
4138 4138 dsl_pool_rele(dp, FTAG);
4139 4139 }
4140 4140
4141 4141 if (estimate) {
4142 4142 dsl_pool_t *dp;
4143 4143 dsl_dataset_t *tosnap;
4144 4144 dsl_dataset_t *fromsnap = NULL;
4145 4145
4146 4146 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
4147 4147 if (error != 0)
4148 4148 return (error);
4149 4149
4150 4150 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &tosnap);
4151 4151 if (error != 0) {
4152 4152 dsl_pool_rele(dp, FTAG);
4153 4153 return (error);
4154 4154 }
4155 4155
4156 4156 if (zc->zc_fromobj != 0) {
4157 4157 error = dsl_dataset_hold_obj(dp, zc->zc_fromobj,
4158 4158 FTAG, &fromsnap);
4159 4159 if (error != 0) {
4160 4160 dsl_dataset_rele(tosnap, FTAG);
4161 4161 dsl_pool_rele(dp, FTAG);
4162 4162 return (error);
4163 4163 }
4164 4164 }
4165 4165
4166 4166 error = dmu_send_estimate(tosnap, fromsnap,
4167 4167 &zc->zc_objset_type);
4168 4168
4169 4169 if (fromsnap != NULL)
4170 4170 dsl_dataset_rele(fromsnap, FTAG);
4171 4171 dsl_dataset_rele(tosnap, FTAG);
4172 4172 dsl_pool_rele(dp, FTAG);
4173 4173 } else {
4174 4174 file_t *fp = getf(zc->zc_cookie);
4175 4175 if (fp == NULL)
4176 4176 return (SET_ERROR(EBADF));
4177 4177
4178 4178 off = fp->f_offset;
4179 4179 error = dmu_send_obj(zc->zc_name, zc->zc_sendobj,
4180 4180 zc->zc_fromobj, zc->zc_cookie, fp->f_vnode, &off);
4181 4181
4182 4182 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
4183 4183 fp->f_offset = off;
4184 4184 releasef(zc->zc_cookie);
4185 4185 }
4186 4186 return (error);
4187 4187 }
4188 4188
4189 4189 /*
4190 4190 * inputs:
4191 4191 * zc_name name of snapshot on which to report progress
4192 4192 * zc_cookie file descriptor of send stream
4193 4193 *
4194 4194 * outputs:
4195 4195 * zc_cookie number of bytes written in send stream thus far
4196 4196 */
4197 4197 static int
4198 4198 zfs_ioc_send_progress(zfs_cmd_t *zc)
4199 4199 {
4200 4200 dsl_pool_t *dp;
4201 4201 dsl_dataset_t *ds;
4202 4202 dmu_sendarg_t *dsp = NULL;
4203 4203 int error;
4204 4204
4205 4205 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
4206 4206 if (error != 0)
4207 4207 return (error);
4208 4208
4209 4209 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &ds);
4210 4210 if (error != 0) {
4211 4211 dsl_pool_rele(dp, FTAG);
4212 4212 return (error);
4213 4213 }
4214 4214
4215 4215 mutex_enter(&ds->ds_sendstream_lock);
4216 4216
4217 4217 /*
4218 4218 * Iterate over all the send streams currently active on this dataset.
4219 4219 * If there's one which matches the specified file descriptor _and_ the
4220 4220 * stream was started by the current process, return the progress of
4221 4221 * that stream.
4222 4222 */
4223 4223 for (dsp = list_head(&ds->ds_sendstreams); dsp != NULL;
4224 4224 dsp = list_next(&ds->ds_sendstreams, dsp)) {
4225 4225 if (dsp->dsa_outfd == zc->zc_cookie &&
4226 4226 dsp->dsa_proc == curproc)
4227 4227 break;
4228 4228 }
4229 4229
4230 4230 if (dsp != NULL)
4231 4231 zc->zc_cookie = *(dsp->dsa_off);
4232 4232 else
4233 4233 error = SET_ERROR(ENOENT);
4234 4234
4235 4235 mutex_exit(&ds->ds_sendstream_lock);
4236 4236 dsl_dataset_rele(ds, FTAG);
4237 4237 dsl_pool_rele(dp, FTAG);
4238 4238 return (error);
4239 4239 }
4240 4240
4241 4241 static int
4242 4242 zfs_ioc_inject_fault(zfs_cmd_t *zc)
4243 4243 {
4244 4244 int id, error;
4245 4245
4246 4246 error = zio_inject_fault(zc->zc_name, (int)zc->zc_guid, &id,
4247 4247 &zc->zc_inject_record);
4248 4248
4249 4249 if (error == 0)
4250 4250 zc->zc_guid = (uint64_t)id;
4251 4251
4252 4252 return (error);
4253 4253 }
4254 4254
4255 4255 static int
4256 4256 zfs_ioc_clear_fault(zfs_cmd_t *zc)
4257 4257 {
4258 4258 return (zio_clear_fault((int)zc->zc_guid));
4259 4259 }
4260 4260
4261 4261 static int
4262 4262 zfs_ioc_inject_list_next(zfs_cmd_t *zc)
4263 4263 {
4264 4264 int id = (int)zc->zc_guid;
4265 4265 int error;
4266 4266
4267 4267 error = zio_inject_list_next(&id, zc->zc_name, sizeof (zc->zc_name),
4268 4268 &zc->zc_inject_record);
4269 4269
4270 4270 zc->zc_guid = id;
4271 4271
4272 4272 return (error);
4273 4273 }
4274 4274
4275 4275 static int
4276 4276 zfs_ioc_error_log(zfs_cmd_t *zc)
4277 4277 {
4278 4278 spa_t *spa;
4279 4279 int error;
4280 4280 size_t count = (size_t)zc->zc_nvlist_dst_size;
4281 4281
4282 4282 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
4283 4283 return (error);
4284 4284
4285 4285 error = spa_get_errlog(spa, (void *)(uintptr_t)zc->zc_nvlist_dst,
4286 4286 &count);
4287 4287 if (error == 0)
4288 4288 zc->zc_nvlist_dst_size = count;
4289 4289 else
4290 4290 zc->zc_nvlist_dst_size = spa_get_errlog_size(spa);
4291 4291
4292 4292 spa_close(spa, FTAG);
4293 4293
4294 4294 return (error);
4295 4295 }
4296 4296
4297 4297 static int
4298 4298 zfs_ioc_clear(zfs_cmd_t *zc)
4299 4299 {
4300 4300 spa_t *spa;
4301 4301 vdev_t *vd;
4302 4302 int error;
4303 4303
4304 4304 /*
4305 4305 * On zpool clear we also fix up missing slogs
4306 4306 */
4307 4307 mutex_enter(&spa_namespace_lock);
4308 4308 spa = spa_lookup(zc->zc_name);
4309 4309 if (spa == NULL) {
4310 4310 mutex_exit(&spa_namespace_lock);
4311 4311 return (SET_ERROR(EIO));
4312 4312 }
4313 4313 if (spa_get_log_state(spa) == SPA_LOG_MISSING) {
4314 4314 /* we need to let spa_open/spa_load clear the chains */
4315 4315 spa_set_log_state(spa, SPA_LOG_CLEAR);
4316 4316 }
4317 4317 spa->spa_last_open_failed = 0;
4318 4318 mutex_exit(&spa_namespace_lock);
4319 4319
4320 4320 if (zc->zc_cookie & ZPOOL_NO_REWIND) {
4321 4321 error = spa_open(zc->zc_name, &spa, FTAG);
4322 4322 } else {
4323 4323 nvlist_t *policy;
4324 4324 nvlist_t *config = NULL;
4325 4325
4326 4326 if (zc->zc_nvlist_src == NULL)
4327 4327 return (SET_ERROR(EINVAL));
4328 4328
4329 4329 if ((error = get_nvlist(zc->zc_nvlist_src,
4330 4330 zc->zc_nvlist_src_size, zc->zc_iflags, &policy)) == 0) {
4331 4331 error = spa_open_rewind(zc->zc_name, &spa, FTAG,
4332 4332 policy, &config);
4333 4333 if (config != NULL) {
4334 4334 int err;
4335 4335
4336 4336 if ((err = put_nvlist(zc, config)) != 0)
4337 4337 error = err;
4338 4338 nvlist_free(config);
4339 4339 }
4340 4340 nvlist_free(policy);
4341 4341 }
4342 4342 }
4343 4343
4344 4344 if (error != 0)
4345 4345 return (error);
4346 4346
4347 4347 spa_vdev_state_enter(spa, SCL_NONE);
4348 4348
4349 4349 if (zc->zc_guid == 0) {
4350 4350 vd = NULL;
4351 4351 } else {
4352 4352 vd = spa_lookup_by_guid(spa, zc->zc_guid, B_TRUE);
4353 4353 if (vd == NULL) {
4354 4354 (void) spa_vdev_state_exit(spa, NULL, ENODEV);
4355 4355 spa_close(spa, FTAG);
4356 4356 return (SET_ERROR(ENODEV));
4357 4357 }
4358 4358 }
4359 4359
4360 4360 vdev_clear(spa, vd);
4361 4361
4362 4362 (void) spa_vdev_state_exit(spa, NULL, 0);
4363 4363
4364 4364 /*
4365 4365 * Resume any suspended I/Os.
4366 4366 */
4367 4367 if (zio_resume(spa) != 0)
4368 4368 error = SET_ERROR(EIO);
4369 4369
4370 4370 spa_close(spa, FTAG);
4371 4371
4372 4372 return (error);
4373 4373 }
4374 4374
4375 4375 static int
4376 4376 zfs_ioc_pool_reopen(zfs_cmd_t *zc)
4377 4377 {
4378 4378 spa_t *spa;
4379 4379 int error;
4380 4380
4381 4381 error = spa_open(zc->zc_name, &spa, FTAG);
4382 4382 if (error != 0)
4383 4383 return (error);
4384 4384
4385 4385 spa_vdev_state_enter(spa, SCL_NONE);
4386 4386
4387 4387 /*
4388 4388 * If a resilver is already in progress then set the
4389 4389 * spa_scrub_reopen flag to B_TRUE so that we don't restart
4390 4390 * the scan as a side effect of the reopen. Otherwise, let
4391 4391 * vdev_open() decided if a resilver is required.
4392 4392 */
4393 4393 spa->spa_scrub_reopen = dsl_scan_resilvering(spa->spa_dsl_pool);
4394 4394 vdev_reopen(spa->spa_root_vdev);
4395 4395 spa->spa_scrub_reopen = B_FALSE;
4396 4396
4397 4397 (void) spa_vdev_state_exit(spa, NULL, 0);
4398 4398 spa_close(spa, FTAG);
4399 4399 return (0);
4400 4400 }
4401 4401 /*
4402 4402 * inputs:
4403 4403 * zc_name name of filesystem
4404 4404 * zc_value name of origin snapshot
4405 4405 *
4406 4406 * outputs:
4407 4407 * zc_string name of conflicting snapshot, if there is one
4408 4408 */
4409 4409 static int
4410 4410 zfs_ioc_promote(zfs_cmd_t *zc)
4411 4411 {
4412 4412 char *cp;
4413 4413
4414 4414 /*
4415 4415 * We don't need to unmount *all* the origin fs's snapshots, but
4416 4416 * it's easier.
4417 4417 */
4418 4418 cp = strchr(zc->zc_value, '@');
4419 4419 if (cp)
4420 4420 *cp = '\0';
4421 4421 (void) dmu_objset_find(zc->zc_value,
4422 4422 zfs_unmount_snap_cb, NULL, DS_FIND_SNAPSHOTS);
4423 4423 return (dsl_dataset_promote(zc->zc_name, zc->zc_string));
4424 4424 }
4425 4425
4426 4426 /*
4427 4427 * Retrieve a single {user|group}{used|quota}@... property.
4428 4428 *
4429 4429 * inputs:
4430 4430 * zc_name name of filesystem
4431 4431 * zc_objset_type zfs_userquota_prop_t
4432 4432 * zc_value domain name (eg. "S-1-234-567-89")
4433 4433 * zc_guid RID/UID/GID
4434 4434 *
4435 4435 * outputs:
4436 4436 * zc_cookie property value
4437 4437 */
4438 4438 static int
4439 4439 zfs_ioc_userspace_one(zfs_cmd_t *zc)
4440 4440 {
4441 4441 zfsvfs_t *zfsvfs;
4442 4442 int error;
4443 4443
4444 4444 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
4445 4445 return (SET_ERROR(EINVAL));
4446 4446
4447 4447 error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE);
4448 4448 if (error != 0)
4449 4449 return (error);
4450 4450
4451 4451 error = zfs_userspace_one(zfsvfs,
4452 4452 zc->zc_objset_type, zc->zc_value, zc->zc_guid, &zc->zc_cookie);
4453 4453 zfsvfs_rele(zfsvfs, FTAG);
4454 4454
4455 4455 return (error);
4456 4456 }
4457 4457
4458 4458 /*
4459 4459 * inputs:
4460 4460 * zc_name name of filesystem
4461 4461 * zc_cookie zap cursor
4462 4462 * zc_objset_type zfs_userquota_prop_t
4463 4463 * zc_nvlist_dst[_size] buffer to fill (not really an nvlist)
4464 4464 *
4465 4465 * outputs:
4466 4466 * zc_nvlist_dst[_size] data buffer (array of zfs_useracct_t)
4467 4467 * zc_cookie zap cursor
4468 4468 */
4469 4469 static int
4470 4470 zfs_ioc_userspace_many(zfs_cmd_t *zc)
4471 4471 {
4472 4472 zfsvfs_t *zfsvfs;
4473 4473 int bufsize = zc->zc_nvlist_dst_size;
4474 4474
4475 4475 if (bufsize <= 0)
4476 4476 return (SET_ERROR(ENOMEM));
4477 4477
4478 4478 int error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE);
4479 4479 if (error != 0)
4480 4480 return (error);
4481 4481
4482 4482 void *buf = kmem_alloc(bufsize, KM_SLEEP);
4483 4483
4484 4484 error = zfs_userspace_many(zfsvfs, zc->zc_objset_type, &zc->zc_cookie,
4485 4485 buf, &zc->zc_nvlist_dst_size);
4486 4486
4487 4487 if (error == 0) {
4488 4488 error = xcopyout(buf,
4489 4489 (void *)(uintptr_t)zc->zc_nvlist_dst,
4490 4490 zc->zc_nvlist_dst_size);
4491 4491 }
4492 4492 kmem_free(buf, bufsize);
4493 4493 zfsvfs_rele(zfsvfs, FTAG);
4494 4494
4495 4495 return (error);
4496 4496 }
4497 4497
4498 4498 /*
4499 4499 * inputs:
4500 4500 * zc_name name of filesystem
4501 4501 *
4502 4502 * outputs:
4503 4503 * none
4504 4504 */
4505 4505 static int
4506 4506 zfs_ioc_userspace_upgrade(zfs_cmd_t *zc)
4507 4507 {
4508 4508 objset_t *os;
4509 4509 int error = 0;
4510 4510 zfsvfs_t *zfsvfs;
4511 4511
4512 4512 if (getzfsvfs(zc->zc_name, &zfsvfs) == 0) {
4513 4513 if (!dmu_objset_userused_enabled(zfsvfs->z_os)) {
4514 4514 /*
4515 4515 * If userused is not enabled, it may be because the
4516 4516 * objset needs to be closed & reopened (to grow the
4517 4517 * objset_phys_t). Suspend/resume the fs will do that.
4518 4518 */
4519 4519 error = zfs_suspend_fs(zfsvfs);
4520 4520 if (error == 0)
4521 4521 error = zfs_resume_fs(zfsvfs, zc->zc_name);
4522 4522 }
4523 4523 if (error == 0)
4524 4524 error = dmu_objset_userspace_upgrade(zfsvfs->z_os);
4525 4525 VFS_RELE(zfsvfs->z_vfs);
4526 4526 } else {
4527 4527 /* XXX kind of reading contents without owning */
4528 4528 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
4529 4529 if (error != 0)
4530 4530 return (error);
4531 4531
4532 4532 error = dmu_objset_userspace_upgrade(os);
4533 4533 dmu_objset_rele(os, FTAG);
4534 4534 }
4535 4535
4536 4536 return (error);
4537 4537 }
4538 4538
4539 4539 /*
4540 4540 * We don't want to have a hard dependency
4541 4541 * against some special symbols in sharefs
4542 4542 * nfs, and smbsrv. Determine them if needed when
4543 4543 * the first file system is shared.
4544 4544 * Neither sharefs, nfs or smbsrv are unloadable modules.
4545 4545 */
4546 4546 int (*znfsexport_fs)(void *arg);
4547 4547 int (*zshare_fs)(enum sharefs_sys_op, share_t *, uint32_t);
4548 4548 int (*zsmbexport_fs)(void *arg, boolean_t add_share);
4549 4549
4550 4550 int zfs_nfsshare_inited;
4551 4551 int zfs_smbshare_inited;
4552 4552
4553 4553 ddi_modhandle_t nfs_mod;
4554 4554 ddi_modhandle_t sharefs_mod;
4555 4555 ddi_modhandle_t smbsrv_mod;
4556 4556 kmutex_t zfs_share_lock;
4557 4557
4558 4558 static int
4559 4559 zfs_init_sharefs()
4560 4560 {
4561 4561 int error;
4562 4562
4563 4563 ASSERT(MUTEX_HELD(&zfs_share_lock));
4564 4564 /* Both NFS and SMB shares also require sharetab support. */
4565 4565 if (sharefs_mod == NULL && ((sharefs_mod =
4566 4566 ddi_modopen("fs/sharefs",
4567 4567 KRTLD_MODE_FIRST, &error)) == NULL)) {
4568 4568 return (SET_ERROR(ENOSYS));
4569 4569 }
4570 4570 if (zshare_fs == NULL && ((zshare_fs =
4571 4571 (int (*)(enum sharefs_sys_op, share_t *, uint32_t))
4572 4572 ddi_modsym(sharefs_mod, "sharefs_impl", &error)) == NULL)) {
4573 4573 return (SET_ERROR(ENOSYS));
4574 4574 }
4575 4575 return (0);
4576 4576 }
4577 4577
4578 4578 static int
4579 4579 zfs_ioc_share(zfs_cmd_t *zc)
4580 4580 {
4581 4581 int error;
4582 4582 int opcode;
4583 4583
4584 4584 switch (zc->zc_share.z_sharetype) {
4585 4585 case ZFS_SHARE_NFS:
4586 4586 case ZFS_UNSHARE_NFS:
4587 4587 if (zfs_nfsshare_inited == 0) {
4588 4588 mutex_enter(&zfs_share_lock);
4589 4589 if (nfs_mod == NULL && ((nfs_mod = ddi_modopen("fs/nfs",
4590 4590 KRTLD_MODE_FIRST, &error)) == NULL)) {
4591 4591 mutex_exit(&zfs_share_lock);
4592 4592 return (SET_ERROR(ENOSYS));
4593 4593 }
4594 4594 if (znfsexport_fs == NULL &&
4595 4595 ((znfsexport_fs = (int (*)(void *))
4596 4596 ddi_modsym(nfs_mod,
4597 4597 "nfs_export", &error)) == NULL)) {
4598 4598 mutex_exit(&zfs_share_lock);
4599 4599 return (SET_ERROR(ENOSYS));
4600 4600 }
4601 4601 error = zfs_init_sharefs();
4602 4602 if (error != 0) {
4603 4603 mutex_exit(&zfs_share_lock);
4604 4604 return (SET_ERROR(ENOSYS));
4605 4605 }
4606 4606 zfs_nfsshare_inited = 1;
4607 4607 mutex_exit(&zfs_share_lock);
4608 4608 }
4609 4609 break;
4610 4610 case ZFS_SHARE_SMB:
4611 4611 case ZFS_UNSHARE_SMB:
4612 4612 if (zfs_smbshare_inited == 0) {
4613 4613 mutex_enter(&zfs_share_lock);
4614 4614 if (smbsrv_mod == NULL && ((smbsrv_mod =
4615 4615 ddi_modopen("drv/smbsrv",
4616 4616 KRTLD_MODE_FIRST, &error)) == NULL)) {
4617 4617 mutex_exit(&zfs_share_lock);
4618 4618 return (SET_ERROR(ENOSYS));
4619 4619 }
4620 4620 if (zsmbexport_fs == NULL && ((zsmbexport_fs =
4621 4621 (int (*)(void *, boolean_t))ddi_modsym(smbsrv_mod,
4622 4622 "smb_server_share", &error)) == NULL)) {
4623 4623 mutex_exit(&zfs_share_lock);
4624 4624 return (SET_ERROR(ENOSYS));
4625 4625 }
4626 4626 error = zfs_init_sharefs();
4627 4627 if (error != 0) {
4628 4628 mutex_exit(&zfs_share_lock);
4629 4629 return (SET_ERROR(ENOSYS));
4630 4630 }
4631 4631 zfs_smbshare_inited = 1;
4632 4632 mutex_exit(&zfs_share_lock);
4633 4633 }
4634 4634 break;
4635 4635 default:
4636 4636 return (SET_ERROR(EINVAL));
4637 4637 }
4638 4638
4639 4639 switch (zc->zc_share.z_sharetype) {
4640 4640 case ZFS_SHARE_NFS:
4641 4641 case ZFS_UNSHARE_NFS:
4642 4642 if (error =
4643 4643 znfsexport_fs((void *)
4644 4644 (uintptr_t)zc->zc_share.z_exportdata))
4645 4645 return (error);
4646 4646 break;
4647 4647 case ZFS_SHARE_SMB:
4648 4648 case ZFS_UNSHARE_SMB:
4649 4649 if (error = zsmbexport_fs((void *)
4650 4650 (uintptr_t)zc->zc_share.z_exportdata,
4651 4651 zc->zc_share.z_sharetype == ZFS_SHARE_SMB ?
4652 4652 B_TRUE: B_FALSE)) {
4653 4653 return (error);
4654 4654 }
4655 4655 break;
4656 4656 }
4657 4657
4658 4658 opcode = (zc->zc_share.z_sharetype == ZFS_SHARE_NFS ||
4659 4659 zc->zc_share.z_sharetype == ZFS_SHARE_SMB) ?
4660 4660 SHAREFS_ADD : SHAREFS_REMOVE;
4661 4661
4662 4662 /*
4663 4663 * Add or remove share from sharetab
4664 4664 */
4665 4665 error = zshare_fs(opcode,
4666 4666 (void *)(uintptr_t)zc->zc_share.z_sharedata,
4667 4667 zc->zc_share.z_sharemax);
4668 4668
4669 4669 return (error);
4670 4670
4671 4671 }
4672 4672
4673 4673 ace_t full_access[] = {
4674 4674 {(uid_t)-1, ACE_ALL_PERMS, ACE_EVERYONE, 0}
4675 4675 };
4676 4676
4677 4677 /*
4678 4678 * inputs:
4679 4679 * zc_name name of containing filesystem
4680 4680 * zc_obj object # beyond which we want next in-use object #
4681 4681 *
4682 4682 * outputs:
4683 4683 * zc_obj next in-use object #
4684 4684 */
4685 4685 static int
4686 4686 zfs_ioc_next_obj(zfs_cmd_t *zc)
4687 4687 {
4688 4688 objset_t *os = NULL;
4689 4689 int error;
4690 4690
4691 4691 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
4692 4692 if (error != 0)
4693 4693 return (error);
4694 4694
4695 4695 error = dmu_object_next(os, &zc->zc_obj, B_FALSE,
4696 4696 os->os_dsl_dataset->ds_phys->ds_prev_snap_txg);
4697 4697
4698 4698 dmu_objset_rele(os, FTAG);
4699 4699 return (error);
4700 4700 }
4701 4701
4702 4702 /*
4703 4703 * inputs:
4704 4704 * zc_name name of filesystem
4705 4705 * zc_value prefix name for snapshot
4706 4706 * zc_cleanup_fd cleanup-on-exit file descriptor for calling process
4707 4707 *
4708 4708 * outputs:
4709 4709 * zc_value short name of new snapshot
4710 4710 */
4711 4711 static int
4712 4712 zfs_ioc_tmp_snapshot(zfs_cmd_t *zc)
4713 4713 {
4714 4714 char *snap_name;
4715 4715 char *hold_name;
4716 4716 int error;
4717 4717 minor_t minor;
4718 4718
4719 4719 error = zfs_onexit_fd_hold(zc->zc_cleanup_fd, &minor);
4720 4720 if (error != 0)
4721 4721 return (error);
4722 4722
4723 4723 snap_name = kmem_asprintf("%s-%016llx", zc->zc_value,
4724 4724 (u_longlong_t)ddi_get_lbolt64());
4725 4725 hold_name = kmem_asprintf("%%%s", zc->zc_value);
4726 4726
4727 4727 error = dsl_dataset_snapshot_tmp(zc->zc_name, snap_name, minor,
4728 4728 hold_name);
4729 4729 if (error == 0)
4730 4730 (void) strcpy(zc->zc_value, snap_name);
4731 4731 strfree(snap_name);
4732 4732 strfree(hold_name);
4733 4733 zfs_onexit_fd_rele(zc->zc_cleanup_fd);
4734 4734 return (error);
4735 4735 }
4736 4736
4737 4737 /*
4738 4738 * inputs:
4739 4739 * zc_name name of "to" snapshot
4740 4740 * zc_value name of "from" snapshot
4741 4741 * zc_cookie file descriptor to write diff data on
4742 4742 *
4743 4743 * outputs:
4744 4744 * dmu_diff_record_t's to the file descriptor
4745 4745 */
4746 4746 static int
4747 4747 zfs_ioc_diff(zfs_cmd_t *zc)
4748 4748 {
4749 4749 file_t *fp;
4750 4750 offset_t off;
4751 4751 int error;
4752 4752
4753 4753 fp = getf(zc->zc_cookie);
4754 4754 if (fp == NULL)
4755 4755 return (SET_ERROR(EBADF));
4756 4756
4757 4757 off = fp->f_offset;
4758 4758
4759 4759 error = dmu_diff(zc->zc_name, zc->zc_value, fp->f_vnode, &off);
4760 4760
4761 4761 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
4762 4762 fp->f_offset = off;
4763 4763 releasef(zc->zc_cookie);
4764 4764
4765 4765 return (error);
4766 4766 }
4767 4767
4768 4768 /*
4769 4769 * Remove all ACL files in shares dir
4770 4770 */
4771 4771 static int
4772 4772 zfs_smb_acl_purge(znode_t *dzp)
4773 4773 {
4774 4774 zap_cursor_t zc;
4775 4775 zap_attribute_t zap;
4776 4776 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
4777 4777 int error;
4778 4778
4779 4779 for (zap_cursor_init(&zc, zfsvfs->z_os, dzp->z_id);
4780 4780 (error = zap_cursor_retrieve(&zc, &zap)) == 0;
4781 4781 zap_cursor_advance(&zc)) {
4782 4782 if ((error = VOP_REMOVE(ZTOV(dzp), zap.za_name, kcred,
4783 4783 NULL, 0)) != 0)
4784 4784 break;
4785 4785 }
4786 4786 zap_cursor_fini(&zc);
4787 4787 return (error);
4788 4788 }
4789 4789
4790 4790 static int
4791 4791 zfs_ioc_smb_acl(zfs_cmd_t *zc)
4792 4792 {
4793 4793 vnode_t *vp;
4794 4794 znode_t *dzp;
4795 4795 vnode_t *resourcevp = NULL;
4796 4796 znode_t *sharedir;
4797 4797 zfsvfs_t *zfsvfs;
4798 4798 nvlist_t *nvlist;
4799 4799 char *src, *target;
4800 4800 vattr_t vattr;
4801 4801 vsecattr_t vsec;
4802 4802 int error = 0;
4803 4803
4804 4804 if ((error = lookupname(zc->zc_value, UIO_SYSSPACE,
4805 4805 NO_FOLLOW, NULL, &vp)) != 0)
4806 4806 return (error);
4807 4807
4808 4808 /* Now make sure mntpnt and dataset are ZFS */
4809 4809
4810 4810 if (vp->v_vfsp->vfs_fstype != zfsfstype ||
4811 4811 (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource),
4812 4812 zc->zc_name) != 0)) {
4813 4813 VN_RELE(vp);
4814 4814 return (SET_ERROR(EINVAL));
4815 4815 }
4816 4816
4817 4817 dzp = VTOZ(vp);
4818 4818 zfsvfs = dzp->z_zfsvfs;
4819 4819 ZFS_ENTER(zfsvfs);
4820 4820
4821 4821 /*
4822 4822 * Create share dir if its missing.
4823 4823 */
4824 4824 mutex_enter(&zfsvfs->z_lock);
4825 4825 if (zfsvfs->z_shares_dir == 0) {
4826 4826 dmu_tx_t *tx;
4827 4827
4828 4828 tx = dmu_tx_create(zfsvfs->z_os);
4829 4829 dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, TRUE,
4830 4830 ZFS_SHARES_DIR);
4831 4831 dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL);
4832 4832 error = dmu_tx_assign(tx, TXG_WAIT);
4833 4833 if (error != 0) {
4834 4834 dmu_tx_abort(tx);
4835 4835 } else {
4836 4836 error = zfs_create_share_dir(zfsvfs, tx);
4837 4837 dmu_tx_commit(tx);
4838 4838 }
4839 4839 if (error != 0) {
4840 4840 mutex_exit(&zfsvfs->z_lock);
4841 4841 VN_RELE(vp);
4842 4842 ZFS_EXIT(zfsvfs);
4843 4843 return (error);
4844 4844 }
4845 4845 }
4846 4846 mutex_exit(&zfsvfs->z_lock);
4847 4847
4848 4848 ASSERT(zfsvfs->z_shares_dir);
4849 4849 if ((error = zfs_zget(zfsvfs, zfsvfs->z_shares_dir, &sharedir)) != 0) {
4850 4850 VN_RELE(vp);
4851 4851 ZFS_EXIT(zfsvfs);
4852 4852 return (error);
4853 4853 }
4854 4854
4855 4855 switch (zc->zc_cookie) {
4856 4856 case ZFS_SMB_ACL_ADD:
4857 4857 vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE;
4858 4858 vattr.va_type = VREG;
4859 4859 vattr.va_mode = S_IFREG|0777;
4860 4860 vattr.va_uid = 0;
4861 4861 vattr.va_gid = 0;
4862 4862
4863 4863 vsec.vsa_mask = VSA_ACE;
4864 4864 vsec.vsa_aclentp = &full_access;
4865 4865 vsec.vsa_aclentsz = sizeof (full_access);
4866 4866 vsec.vsa_aclcnt = 1;
4867 4867
4868 4868 error = VOP_CREATE(ZTOV(sharedir), zc->zc_string,
4869 4869 &vattr, EXCL, 0, &resourcevp, kcred, 0, NULL, &vsec);
4870 4870 if (resourcevp)
4871 4871 VN_RELE(resourcevp);
4872 4872 break;
4873 4873
4874 4874 case ZFS_SMB_ACL_REMOVE:
4875 4875 error = VOP_REMOVE(ZTOV(sharedir), zc->zc_string, kcred,
4876 4876 NULL, 0);
4877 4877 break;
4878 4878
4879 4879 case ZFS_SMB_ACL_RENAME:
4880 4880 if ((error = get_nvlist(zc->zc_nvlist_src,
4881 4881 zc->zc_nvlist_src_size, zc->zc_iflags, &nvlist)) != 0) {
4882 4882 VN_RELE(vp);
4883 4883 ZFS_EXIT(zfsvfs);
4884 4884 return (error);
4885 4885 }
4886 4886 if (nvlist_lookup_string(nvlist, ZFS_SMB_ACL_SRC, &src) ||
4887 4887 nvlist_lookup_string(nvlist, ZFS_SMB_ACL_TARGET,
4888 4888 &target)) {
4889 4889 VN_RELE(vp);
4890 4890 VN_RELE(ZTOV(sharedir));
4891 4891 ZFS_EXIT(zfsvfs);
4892 4892 nvlist_free(nvlist);
4893 4893 return (error);
4894 4894 }
4895 4895 error = VOP_RENAME(ZTOV(sharedir), src, ZTOV(sharedir), target,
4896 4896 kcred, NULL, 0);
4897 4897 nvlist_free(nvlist);
4898 4898 break;
4899 4899
4900 4900 case ZFS_SMB_ACL_PURGE:
4901 4901 error = zfs_smb_acl_purge(sharedir);
4902 4902 break;
4903 4903
4904 4904 default:
4905 4905 error = SET_ERROR(EINVAL);
4906 4906 break;
4907 4907 }
4908 4908
4909 4909 VN_RELE(vp);
4910 4910 VN_RELE(ZTOV(sharedir));
4911 4911
4912 4912 ZFS_EXIT(zfsvfs);
4913 4913
4914 4914 return (error);
4915 4915 }
4916 4916
4917 4917 /*
4918 4918 * innvl: {
4919 4919 * "holds" -> { snapname -> holdname (string), ... }
4920 4920 * (optional) "cleanup_fd" -> fd (int32)
4921 4921 * }
4922 4922 *
4923 4923 * outnvl: {
4924 4924 * snapname -> error value (int32)
4925 4925 * ...
4926 4926 * }
4927 4927 */
4928 4928 /* ARGSUSED */
4929 4929 static int
4930 4930 zfs_ioc_hold(const char *pool, nvlist_t *args, nvlist_t *errlist)
4931 4931 {
4932 4932 nvlist_t *holds;
4933 4933 int cleanup_fd = -1;
4934 4934 int error;
4935 4935 minor_t minor = 0;
4936 4936
4937 4937 error = nvlist_lookup_nvlist(args, "holds", &holds);
4938 4938 if (error != 0)
4939 4939 return (SET_ERROR(EINVAL));
4940 4940
4941 4941 if (nvlist_lookup_int32(args, "cleanup_fd", &cleanup_fd) == 0) {
4942 4942 error = zfs_onexit_fd_hold(cleanup_fd, &minor);
4943 4943 if (error != 0)
4944 4944 return (error);
4945 4945 }
4946 4946
4947 4947 error = dsl_dataset_user_hold(holds, minor, errlist);
4948 4948 if (minor != 0)
4949 4949 zfs_onexit_fd_rele(cleanup_fd);
4950 4950 return (error);
4951 4951 }
4952 4952
4953 4953 /*
4954 4954 * innvl is not used.
4955 4955 *
4956 4956 * outnvl: {
4957 4957 * holdname -> time added (uint64 seconds since epoch)
4958 4958 * ...
4959 4959 * }
4960 4960 */
4961 4961 /* ARGSUSED */
4962 4962 static int
4963 4963 zfs_ioc_get_holds(const char *snapname, nvlist_t *args, nvlist_t *outnvl)
4964 4964 {
4965 4965 return (dsl_dataset_get_holds(snapname, outnvl));
4966 4966 }
4967 4967
4968 4968 /*
4969 4969 * innvl: {
4970 4970 * snapname -> { holdname, ... }
4971 4971 * ...
4972 4972 * }
|
↓ open down ↓ |
4972 lines elided |
↑ open up ↑ |
4973 4973 *
4974 4974 * outnvl: {
4975 4975 * snapname -> error value (int32)
4976 4976 * ...
4977 4977 * }
4978 4978 */
4979 4979 /* ARGSUSED */
4980 4980 static int
4981 4981 zfs_ioc_release(const char *pool, nvlist_t *holds, nvlist_t *errlist)
4982 4982 {
4983 - nvpair_t *pair;
4984 -
4985 - /*
4986 - * The release may cause the snapshot to be destroyed; make sure it
4987 - * is not mounted.
4988 - */
4989 - for (pair = nvlist_next_nvpair(holds, NULL); pair != NULL;
4990 - pair = nvlist_next_nvpair(holds, pair))
4991 - zfs_unmount_snap(nvpair_name(pair));
4992 -
4993 4983 return (dsl_dataset_user_release(holds, errlist));
4994 4984 }
4995 4985
4996 4986 /*
4997 4987 * inputs:
4998 4988 * zc_name name of new filesystem or snapshot
4999 4989 * zc_value full name of old snapshot
5000 4990 *
5001 4991 * outputs:
5002 4992 * zc_cookie space in bytes
5003 4993 * zc_objset_type compressed space in bytes
5004 4994 * zc_perm_action uncompressed space in bytes
5005 4995 */
5006 4996 static int
5007 4997 zfs_ioc_space_written(zfs_cmd_t *zc)
5008 4998 {
5009 4999 int error;
5010 5000 dsl_pool_t *dp;
5011 5001 dsl_dataset_t *new, *old;
5012 5002
5013 5003 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
5014 5004 if (error != 0)
5015 5005 return (error);
5016 5006 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &new);
5017 5007 if (error != 0) {
5018 5008 dsl_pool_rele(dp, FTAG);
5019 5009 return (error);
5020 5010 }
5021 5011 error = dsl_dataset_hold(dp, zc->zc_value, FTAG, &old);
5022 5012 if (error != 0) {
5023 5013 dsl_dataset_rele(new, FTAG);
5024 5014 dsl_pool_rele(dp, FTAG);
5025 5015 return (error);
5026 5016 }
5027 5017
5028 5018 error = dsl_dataset_space_written(old, new, &zc->zc_cookie,
5029 5019 &zc->zc_objset_type, &zc->zc_perm_action);
5030 5020 dsl_dataset_rele(old, FTAG);
5031 5021 dsl_dataset_rele(new, FTAG);
5032 5022 dsl_pool_rele(dp, FTAG);
5033 5023 return (error);
5034 5024 }
5035 5025
5036 5026 /*
5037 5027 * innvl: {
5038 5028 * "firstsnap" -> snapshot name
5039 5029 * }
5040 5030 *
5041 5031 * outnvl: {
5042 5032 * "used" -> space in bytes
5043 5033 * "compressed" -> compressed space in bytes
5044 5034 * "uncompressed" -> uncompressed space in bytes
5045 5035 * }
5046 5036 */
5047 5037 static int
5048 5038 zfs_ioc_space_snaps(const char *lastsnap, nvlist_t *innvl, nvlist_t *outnvl)
5049 5039 {
5050 5040 int error;
5051 5041 dsl_pool_t *dp;
5052 5042 dsl_dataset_t *new, *old;
5053 5043 char *firstsnap;
5054 5044 uint64_t used, comp, uncomp;
5055 5045
5056 5046 if (nvlist_lookup_string(innvl, "firstsnap", &firstsnap) != 0)
5057 5047 return (SET_ERROR(EINVAL));
5058 5048
5059 5049 error = dsl_pool_hold(lastsnap, FTAG, &dp);
5060 5050 if (error != 0)
5061 5051 return (error);
5062 5052
5063 5053 error = dsl_dataset_hold(dp, lastsnap, FTAG, &new);
5064 5054 if (error != 0) {
5065 5055 dsl_pool_rele(dp, FTAG);
5066 5056 return (error);
5067 5057 }
5068 5058 error = dsl_dataset_hold(dp, firstsnap, FTAG, &old);
5069 5059 if (error != 0) {
5070 5060 dsl_dataset_rele(new, FTAG);
5071 5061 dsl_pool_rele(dp, FTAG);
5072 5062 return (error);
5073 5063 }
5074 5064
5075 5065 error = dsl_dataset_space_wouldfree(old, new, &used, &comp, &uncomp);
5076 5066 dsl_dataset_rele(old, FTAG);
5077 5067 dsl_dataset_rele(new, FTAG);
5078 5068 dsl_pool_rele(dp, FTAG);
5079 5069 fnvlist_add_uint64(outnvl, "used", used);
5080 5070 fnvlist_add_uint64(outnvl, "compressed", comp);
5081 5071 fnvlist_add_uint64(outnvl, "uncompressed", uncomp);
5082 5072 return (error);
5083 5073 }
5084 5074
5085 5075 /*
5086 5076 * innvl: {
5087 5077 * "fd" -> file descriptor to write stream to (int32)
5088 5078 * (optional) "fromsnap" -> full snap name to send an incremental from
5089 5079 * }
5090 5080 *
5091 5081 * outnvl is unused
5092 5082 */
5093 5083 /* ARGSUSED */
5094 5084 static int
5095 5085 zfs_ioc_send_new(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl)
5096 5086 {
5097 5087 int error;
5098 5088 offset_t off;
5099 5089 char *fromname = NULL;
5100 5090 int fd;
5101 5091
5102 5092 error = nvlist_lookup_int32(innvl, "fd", &fd);
5103 5093 if (error != 0)
5104 5094 return (SET_ERROR(EINVAL));
5105 5095
5106 5096 (void) nvlist_lookup_string(innvl, "fromsnap", &fromname);
5107 5097
5108 5098 file_t *fp = getf(fd);
5109 5099 if (fp == NULL)
5110 5100 return (SET_ERROR(EBADF));
5111 5101
5112 5102 off = fp->f_offset;
5113 5103 error = dmu_send(snapname, fromname, fd, fp->f_vnode, &off);
5114 5104
5115 5105 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
5116 5106 fp->f_offset = off;
5117 5107 releasef(fd);
5118 5108 return (error);
5119 5109 }
5120 5110
5121 5111 /*
5122 5112 * Determine approximately how large a zfs send stream will be -- the number
5123 5113 * of bytes that will be written to the fd supplied to zfs_ioc_send_new().
5124 5114 *
5125 5115 * innvl: {
5126 5116 * (optional) "fromsnap" -> full snap name to send an incremental from
5127 5117 * }
5128 5118 *
5129 5119 * outnvl: {
5130 5120 * "space" -> bytes of space (uint64)
5131 5121 * }
5132 5122 */
5133 5123 static int
5134 5124 zfs_ioc_send_space(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl)
5135 5125 {
5136 5126 dsl_pool_t *dp;
5137 5127 dsl_dataset_t *fromsnap = NULL;
5138 5128 dsl_dataset_t *tosnap;
5139 5129 int error;
5140 5130 char *fromname;
5141 5131 uint64_t space;
5142 5132
5143 5133 error = dsl_pool_hold(snapname, FTAG, &dp);
5144 5134 if (error != 0)
5145 5135 return (error);
5146 5136
5147 5137 error = dsl_dataset_hold(dp, snapname, FTAG, &tosnap);
5148 5138 if (error != 0) {
5149 5139 dsl_pool_rele(dp, FTAG);
5150 5140 return (error);
5151 5141 }
5152 5142
5153 5143 error = nvlist_lookup_string(innvl, "fromsnap", &fromname);
5154 5144 if (error == 0) {
5155 5145 error = dsl_dataset_hold(dp, fromname, FTAG, &fromsnap);
5156 5146 if (error != 0) {
5157 5147 dsl_dataset_rele(tosnap, FTAG);
5158 5148 dsl_pool_rele(dp, FTAG);
5159 5149 return (error);
5160 5150 }
5161 5151 }
5162 5152
5163 5153 error = dmu_send_estimate(tosnap, fromsnap, &space);
5164 5154 fnvlist_add_uint64(outnvl, "space", space);
5165 5155
5166 5156 if (fromsnap != NULL)
5167 5157 dsl_dataset_rele(fromsnap, FTAG);
5168 5158 dsl_dataset_rele(tosnap, FTAG);
5169 5159 dsl_pool_rele(dp, FTAG);
5170 5160 return (error);
5171 5161 }
5172 5162
5173 5163
5174 5164 static zfs_ioc_vec_t zfs_ioc_vec[ZFS_IOC_LAST - ZFS_IOC_FIRST];
5175 5165
5176 5166 static void
5177 5167 zfs_ioctl_register_legacy(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5178 5168 zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck,
5179 5169 boolean_t log_history, zfs_ioc_poolcheck_t pool_check)
5180 5170 {
5181 5171 zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST];
5182 5172
5183 5173 ASSERT3U(ioc, >=, ZFS_IOC_FIRST);
5184 5174 ASSERT3U(ioc, <, ZFS_IOC_LAST);
5185 5175 ASSERT3P(vec->zvec_legacy_func, ==, NULL);
5186 5176 ASSERT3P(vec->zvec_func, ==, NULL);
5187 5177
5188 5178 vec->zvec_legacy_func = func;
5189 5179 vec->zvec_secpolicy = secpolicy;
5190 5180 vec->zvec_namecheck = namecheck;
5191 5181 vec->zvec_allow_log = log_history;
5192 5182 vec->zvec_pool_check = pool_check;
5193 5183 }
5194 5184
5195 5185 /*
5196 5186 * See the block comment at the beginning of this file for details on
5197 5187 * each argument to this function.
5198 5188 */
5199 5189 static void
5200 5190 zfs_ioctl_register(const char *name, zfs_ioc_t ioc, zfs_ioc_func_t *func,
5201 5191 zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck,
5202 5192 zfs_ioc_poolcheck_t pool_check, boolean_t smush_outnvlist,
5203 5193 boolean_t allow_log)
5204 5194 {
5205 5195 zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST];
5206 5196
5207 5197 ASSERT3U(ioc, >=, ZFS_IOC_FIRST);
5208 5198 ASSERT3U(ioc, <, ZFS_IOC_LAST);
5209 5199 ASSERT3P(vec->zvec_legacy_func, ==, NULL);
5210 5200 ASSERT3P(vec->zvec_func, ==, NULL);
5211 5201
5212 5202 /* if we are logging, the name must be valid */
5213 5203 ASSERT(!allow_log || namecheck != NO_NAME);
5214 5204
5215 5205 vec->zvec_name = name;
5216 5206 vec->zvec_func = func;
5217 5207 vec->zvec_secpolicy = secpolicy;
5218 5208 vec->zvec_namecheck = namecheck;
5219 5209 vec->zvec_pool_check = pool_check;
5220 5210 vec->zvec_smush_outnvlist = smush_outnvlist;
5221 5211 vec->zvec_allow_log = allow_log;
5222 5212 }
5223 5213
5224 5214 static void
5225 5215 zfs_ioctl_register_pool(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5226 5216 zfs_secpolicy_func_t *secpolicy, boolean_t log_history,
5227 5217 zfs_ioc_poolcheck_t pool_check)
5228 5218 {
5229 5219 zfs_ioctl_register_legacy(ioc, func, secpolicy,
5230 5220 POOL_NAME, log_history, pool_check);
5231 5221 }
5232 5222
5233 5223 static void
5234 5224 zfs_ioctl_register_dataset_nolog(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5235 5225 zfs_secpolicy_func_t *secpolicy, zfs_ioc_poolcheck_t pool_check)
5236 5226 {
5237 5227 zfs_ioctl_register_legacy(ioc, func, secpolicy,
5238 5228 DATASET_NAME, B_FALSE, pool_check);
5239 5229 }
5240 5230
5241 5231 static void
5242 5232 zfs_ioctl_register_pool_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func)
5243 5233 {
5244 5234 zfs_ioctl_register_legacy(ioc, func, zfs_secpolicy_config,
5245 5235 POOL_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5246 5236 }
5247 5237
5248 5238 static void
5249 5239 zfs_ioctl_register_pool_meta(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5250 5240 zfs_secpolicy_func_t *secpolicy)
5251 5241 {
5252 5242 zfs_ioctl_register_legacy(ioc, func, secpolicy,
5253 5243 NO_NAME, B_FALSE, POOL_CHECK_NONE);
5254 5244 }
5255 5245
5256 5246 static void
5257 5247 zfs_ioctl_register_dataset_read_secpolicy(zfs_ioc_t ioc,
5258 5248 zfs_ioc_legacy_func_t *func, zfs_secpolicy_func_t *secpolicy)
5259 5249 {
5260 5250 zfs_ioctl_register_legacy(ioc, func, secpolicy,
5261 5251 DATASET_NAME, B_FALSE, POOL_CHECK_SUSPENDED);
5262 5252 }
5263 5253
5264 5254 static void
5265 5255 zfs_ioctl_register_dataset_read(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func)
5266 5256 {
5267 5257 zfs_ioctl_register_dataset_read_secpolicy(ioc, func,
5268 5258 zfs_secpolicy_read);
5269 5259 }
5270 5260
5271 5261 static void
5272 5262 zfs_ioctl_register_dataset_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5273 5263 zfs_secpolicy_func_t *secpolicy)
5274 5264 {
5275 5265 zfs_ioctl_register_legacy(ioc, func, secpolicy,
5276 5266 DATASET_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5277 5267 }
5278 5268
5279 5269 static void
5280 5270 zfs_ioctl_init(void)
5281 5271 {
5282 5272 zfs_ioctl_register("snapshot", ZFS_IOC_SNAPSHOT,
5283 5273 zfs_ioc_snapshot, zfs_secpolicy_snapshot, POOL_NAME,
5284 5274 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5285 5275
5286 5276 zfs_ioctl_register("log_history", ZFS_IOC_LOG_HISTORY,
5287 5277 zfs_ioc_log_history, zfs_secpolicy_log_history, NO_NAME,
5288 5278 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_FALSE);
5289 5279
5290 5280 zfs_ioctl_register("space_snaps", ZFS_IOC_SPACE_SNAPS,
5291 5281 zfs_ioc_space_snaps, zfs_secpolicy_read, DATASET_NAME,
5292 5282 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5293 5283
5294 5284 zfs_ioctl_register("send", ZFS_IOC_SEND_NEW,
5295 5285 zfs_ioc_send_new, zfs_secpolicy_send_new, DATASET_NAME,
5296 5286 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5297 5287
5298 5288 zfs_ioctl_register("send_space", ZFS_IOC_SEND_SPACE,
5299 5289 zfs_ioc_send_space, zfs_secpolicy_read, DATASET_NAME,
5300 5290 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5301 5291
5302 5292 zfs_ioctl_register("create", ZFS_IOC_CREATE,
5303 5293 zfs_ioc_create, zfs_secpolicy_create_clone, DATASET_NAME,
5304 5294 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5305 5295
5306 5296 zfs_ioctl_register("clone", ZFS_IOC_CLONE,
5307 5297 zfs_ioc_clone, zfs_secpolicy_create_clone, DATASET_NAME,
5308 5298 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5309 5299
5310 5300 zfs_ioctl_register("destroy_snaps", ZFS_IOC_DESTROY_SNAPS,
5311 5301 zfs_ioc_destroy_snaps, zfs_secpolicy_destroy_snaps, POOL_NAME,
5312 5302 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5313 5303
5314 5304 zfs_ioctl_register("hold", ZFS_IOC_HOLD,
5315 5305 zfs_ioc_hold, zfs_secpolicy_hold, POOL_NAME,
5316 5306 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5317 5307 zfs_ioctl_register("release", ZFS_IOC_RELEASE,
5318 5308 zfs_ioc_release, zfs_secpolicy_release, POOL_NAME,
5319 5309 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5320 5310
5321 5311 zfs_ioctl_register("get_holds", ZFS_IOC_GET_HOLDS,
5322 5312 zfs_ioc_get_holds, zfs_secpolicy_read, DATASET_NAME,
5323 5313 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5324 5314
5325 5315 /* IOCTLS that use the legacy function signature */
5326 5316
5327 5317 zfs_ioctl_register_legacy(ZFS_IOC_POOL_FREEZE, zfs_ioc_pool_freeze,
5328 5318 zfs_secpolicy_config, NO_NAME, B_FALSE, POOL_CHECK_READONLY);
5329 5319
5330 5320 zfs_ioctl_register_pool(ZFS_IOC_POOL_CREATE, zfs_ioc_pool_create,
5331 5321 zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE);
5332 5322 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SCAN,
5333 5323 zfs_ioc_pool_scan);
5334 5324 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_UPGRADE,
5335 5325 zfs_ioc_pool_upgrade);
5336 5326 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ADD,
5337 5327 zfs_ioc_vdev_add);
5338 5328 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_REMOVE,
5339 5329 zfs_ioc_vdev_remove);
5340 5330 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SET_STATE,
5341 5331 zfs_ioc_vdev_set_state);
5342 5332 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ATTACH,
5343 5333 zfs_ioc_vdev_attach);
5344 5334 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_DETACH,
5345 5335 zfs_ioc_vdev_detach);
5346 5336 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETPATH,
5347 5337 zfs_ioc_vdev_setpath);
5348 5338 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETFRU,
5349 5339 zfs_ioc_vdev_setfru);
5350 5340 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SET_PROPS,
5351 5341 zfs_ioc_pool_set_props);
5352 5342 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SPLIT,
5353 5343 zfs_ioc_vdev_split);
5354 5344 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_REGUID,
5355 5345 zfs_ioc_pool_reguid);
5356 5346
5357 5347 zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_CONFIGS,
5358 5348 zfs_ioc_pool_configs, zfs_secpolicy_none);
5359 5349 zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_TRYIMPORT,
5360 5350 zfs_ioc_pool_tryimport, zfs_secpolicy_config);
5361 5351 zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_FAULT,
5362 5352 zfs_ioc_inject_fault, zfs_secpolicy_inject);
5363 5353 zfs_ioctl_register_pool_meta(ZFS_IOC_CLEAR_FAULT,
5364 5354 zfs_ioc_clear_fault, zfs_secpolicy_inject);
5365 5355 zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_LIST_NEXT,
5366 5356 zfs_ioc_inject_list_next, zfs_secpolicy_inject);
5367 5357
5368 5358 /*
5369 5359 * pool destroy, and export don't log the history as part of
5370 5360 * zfsdev_ioctl, but rather zfs_ioc_pool_export
5371 5361 * does the logging of those commands.
5372 5362 */
5373 5363 zfs_ioctl_register_pool(ZFS_IOC_POOL_DESTROY, zfs_ioc_pool_destroy,
5374 5364 zfs_secpolicy_config, B_FALSE, POOL_CHECK_NONE);
5375 5365 zfs_ioctl_register_pool(ZFS_IOC_POOL_EXPORT, zfs_ioc_pool_export,
5376 5366 zfs_secpolicy_config, B_FALSE, POOL_CHECK_NONE);
5377 5367
5378 5368 zfs_ioctl_register_pool(ZFS_IOC_POOL_STATS, zfs_ioc_pool_stats,
5379 5369 zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE);
5380 5370 zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_PROPS, zfs_ioc_pool_get_props,
5381 5371 zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE);
5382 5372
5383 5373 zfs_ioctl_register_pool(ZFS_IOC_ERROR_LOG, zfs_ioc_error_log,
5384 5374 zfs_secpolicy_inject, B_FALSE, POOL_CHECK_SUSPENDED);
5385 5375 zfs_ioctl_register_pool(ZFS_IOC_DSOBJ_TO_DSNAME,
5386 5376 zfs_ioc_dsobj_to_dsname,
5387 5377 zfs_secpolicy_diff, B_FALSE, POOL_CHECK_SUSPENDED);
5388 5378 zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_HISTORY,
5389 5379 zfs_ioc_pool_get_history,
5390 5380 zfs_secpolicy_config, B_FALSE, POOL_CHECK_SUSPENDED);
5391 5381
5392 5382 zfs_ioctl_register_pool(ZFS_IOC_POOL_IMPORT, zfs_ioc_pool_import,
5393 5383 zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE);
5394 5384
5395 5385 zfs_ioctl_register_pool(ZFS_IOC_CLEAR, zfs_ioc_clear,
5396 5386 zfs_secpolicy_config, B_TRUE, POOL_CHECK_SUSPENDED);
5397 5387 zfs_ioctl_register_pool(ZFS_IOC_POOL_REOPEN, zfs_ioc_pool_reopen,
5398 5388 zfs_secpolicy_config, B_TRUE, POOL_CHECK_SUSPENDED);
5399 5389
5400 5390 zfs_ioctl_register_dataset_read(ZFS_IOC_SPACE_WRITTEN,
5401 5391 zfs_ioc_space_written);
5402 5392 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_RECVD_PROPS,
5403 5393 zfs_ioc_objset_recvd_props);
5404 5394 zfs_ioctl_register_dataset_read(ZFS_IOC_NEXT_OBJ,
5405 5395 zfs_ioc_next_obj);
5406 5396 zfs_ioctl_register_dataset_read(ZFS_IOC_GET_FSACL,
5407 5397 zfs_ioc_get_fsacl);
5408 5398 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_STATS,
5409 5399 zfs_ioc_objset_stats);
5410 5400 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_ZPLPROPS,
5411 5401 zfs_ioc_objset_zplprops);
5412 5402 zfs_ioctl_register_dataset_read(ZFS_IOC_DATASET_LIST_NEXT,
5413 5403 zfs_ioc_dataset_list_next);
5414 5404 zfs_ioctl_register_dataset_read(ZFS_IOC_SNAPSHOT_LIST_NEXT,
5415 5405 zfs_ioc_snapshot_list_next);
5416 5406 zfs_ioctl_register_dataset_read(ZFS_IOC_SEND_PROGRESS,
5417 5407 zfs_ioc_send_progress);
5418 5408
5419 5409 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_DIFF,
5420 5410 zfs_ioc_diff, zfs_secpolicy_diff);
5421 5411 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_STATS,
5422 5412 zfs_ioc_obj_to_stats, zfs_secpolicy_diff);
5423 5413 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_PATH,
5424 5414 zfs_ioc_obj_to_path, zfs_secpolicy_diff);
5425 5415 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_ONE,
5426 5416 zfs_ioc_userspace_one, zfs_secpolicy_userspace_one);
5427 5417 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_MANY,
5428 5418 zfs_ioc_userspace_many, zfs_secpolicy_userspace_many);
5429 5419 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_SEND,
5430 5420 zfs_ioc_send, zfs_secpolicy_send);
5431 5421
5432 5422 zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_PROP, zfs_ioc_set_prop,
5433 5423 zfs_secpolicy_none);
5434 5424 zfs_ioctl_register_dataset_modify(ZFS_IOC_DESTROY, zfs_ioc_destroy,
5435 5425 zfs_secpolicy_destroy);
5436 5426 zfs_ioctl_register_dataset_modify(ZFS_IOC_ROLLBACK, zfs_ioc_rollback,
5437 5427 zfs_secpolicy_rollback);
5438 5428 zfs_ioctl_register_dataset_modify(ZFS_IOC_RENAME, zfs_ioc_rename,
5439 5429 zfs_secpolicy_rename);
5440 5430 zfs_ioctl_register_dataset_modify(ZFS_IOC_RECV, zfs_ioc_recv,
5441 5431 zfs_secpolicy_recv);
5442 5432 zfs_ioctl_register_dataset_modify(ZFS_IOC_PROMOTE, zfs_ioc_promote,
5443 5433 zfs_secpolicy_promote);
5444 5434 zfs_ioctl_register_dataset_modify(ZFS_IOC_INHERIT_PROP,
5445 5435 zfs_ioc_inherit_prop, zfs_secpolicy_inherit_prop);
5446 5436 zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_FSACL, zfs_ioc_set_fsacl,
5447 5437 zfs_secpolicy_set_fsacl);
5448 5438
5449 5439 zfs_ioctl_register_dataset_nolog(ZFS_IOC_SHARE, zfs_ioc_share,
5450 5440 zfs_secpolicy_share, POOL_CHECK_NONE);
5451 5441 zfs_ioctl_register_dataset_nolog(ZFS_IOC_SMB_ACL, zfs_ioc_smb_acl,
5452 5442 zfs_secpolicy_smb_acl, POOL_CHECK_NONE);
5453 5443 zfs_ioctl_register_dataset_nolog(ZFS_IOC_USERSPACE_UPGRADE,
5454 5444 zfs_ioc_userspace_upgrade, zfs_secpolicy_userspace_upgrade,
5455 5445 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5456 5446 zfs_ioctl_register_dataset_nolog(ZFS_IOC_TMP_SNAPSHOT,
5457 5447 zfs_ioc_tmp_snapshot, zfs_secpolicy_tmp_snapshot,
5458 5448 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5459 5449 }
5460 5450
5461 5451 int
5462 5452 pool_status_check(const char *name, zfs_ioc_namecheck_t type,
5463 5453 zfs_ioc_poolcheck_t check)
5464 5454 {
5465 5455 spa_t *spa;
5466 5456 int error;
5467 5457
5468 5458 ASSERT(type == POOL_NAME || type == DATASET_NAME);
5469 5459
5470 5460 if (check & POOL_CHECK_NONE)
5471 5461 return (0);
5472 5462
5473 5463 error = spa_open(name, &spa, FTAG);
5474 5464 if (error == 0) {
5475 5465 if ((check & POOL_CHECK_SUSPENDED) && spa_suspended(spa))
5476 5466 error = SET_ERROR(EAGAIN);
5477 5467 else if ((check & POOL_CHECK_READONLY) && !spa_writeable(spa))
5478 5468 error = SET_ERROR(EROFS);
5479 5469 spa_close(spa, FTAG);
5480 5470 }
5481 5471 return (error);
5482 5472 }
5483 5473
5484 5474 /*
5485 5475 * Find a free minor number.
5486 5476 */
5487 5477 minor_t
5488 5478 zfsdev_minor_alloc(void)
5489 5479 {
5490 5480 static minor_t last_minor;
5491 5481 minor_t m;
5492 5482
5493 5483 ASSERT(MUTEX_HELD(&zfsdev_state_lock));
5494 5484
5495 5485 for (m = last_minor + 1; m != last_minor; m++) {
5496 5486 if (m > ZFSDEV_MAX_MINOR)
5497 5487 m = 1;
5498 5488 if (ddi_get_soft_state(zfsdev_state, m) == NULL) {
5499 5489 last_minor = m;
5500 5490 return (m);
5501 5491 }
5502 5492 }
5503 5493
5504 5494 return (0);
5505 5495 }
5506 5496
5507 5497 static int
5508 5498 zfs_ctldev_init(dev_t *devp)
5509 5499 {
5510 5500 minor_t minor;
5511 5501 zfs_soft_state_t *zs;
5512 5502
5513 5503 ASSERT(MUTEX_HELD(&zfsdev_state_lock));
5514 5504 ASSERT(getminor(*devp) == 0);
5515 5505
5516 5506 minor = zfsdev_minor_alloc();
5517 5507 if (minor == 0)
5518 5508 return (SET_ERROR(ENXIO));
5519 5509
5520 5510 if (ddi_soft_state_zalloc(zfsdev_state, minor) != DDI_SUCCESS)
5521 5511 return (SET_ERROR(EAGAIN));
5522 5512
5523 5513 *devp = makedevice(getemajor(*devp), minor);
5524 5514
5525 5515 zs = ddi_get_soft_state(zfsdev_state, minor);
5526 5516 zs->zss_type = ZSST_CTLDEV;
5527 5517 zfs_onexit_init((zfs_onexit_t **)&zs->zss_data);
5528 5518
5529 5519 return (0);
5530 5520 }
5531 5521
5532 5522 static void
5533 5523 zfs_ctldev_destroy(zfs_onexit_t *zo, minor_t minor)
5534 5524 {
5535 5525 ASSERT(MUTEX_HELD(&zfsdev_state_lock));
5536 5526
5537 5527 zfs_onexit_destroy(zo);
5538 5528 ddi_soft_state_free(zfsdev_state, minor);
5539 5529 }
5540 5530
5541 5531 void *
5542 5532 zfsdev_get_soft_state(minor_t minor, enum zfs_soft_state_type which)
5543 5533 {
5544 5534 zfs_soft_state_t *zp;
5545 5535
5546 5536 zp = ddi_get_soft_state(zfsdev_state, minor);
5547 5537 if (zp == NULL || zp->zss_type != which)
5548 5538 return (NULL);
5549 5539
5550 5540 return (zp->zss_data);
5551 5541 }
5552 5542
5553 5543 static int
5554 5544 zfsdev_open(dev_t *devp, int flag, int otyp, cred_t *cr)
5555 5545 {
5556 5546 int error = 0;
5557 5547
5558 5548 if (getminor(*devp) != 0)
5559 5549 return (zvol_open(devp, flag, otyp, cr));
5560 5550
5561 5551 /* This is the control device. Allocate a new minor if requested. */
5562 5552 if (flag & FEXCL) {
5563 5553 mutex_enter(&zfsdev_state_lock);
5564 5554 error = zfs_ctldev_init(devp);
5565 5555 mutex_exit(&zfsdev_state_lock);
5566 5556 }
5567 5557
5568 5558 return (error);
5569 5559 }
5570 5560
5571 5561 static int
5572 5562 zfsdev_close(dev_t dev, int flag, int otyp, cred_t *cr)
5573 5563 {
5574 5564 zfs_onexit_t *zo;
5575 5565 minor_t minor = getminor(dev);
5576 5566
5577 5567 if (minor == 0)
5578 5568 return (0);
5579 5569
5580 5570 mutex_enter(&zfsdev_state_lock);
5581 5571 zo = zfsdev_get_soft_state(minor, ZSST_CTLDEV);
5582 5572 if (zo == NULL) {
5583 5573 mutex_exit(&zfsdev_state_lock);
5584 5574 return (zvol_close(dev, flag, otyp, cr));
5585 5575 }
5586 5576 zfs_ctldev_destroy(zo, minor);
5587 5577 mutex_exit(&zfsdev_state_lock);
5588 5578
5589 5579 return (0);
5590 5580 }
5591 5581
5592 5582 static int
5593 5583 zfsdev_ioctl(dev_t dev, int cmd, intptr_t arg, int flag, cred_t *cr, int *rvalp)
5594 5584 {
5595 5585 zfs_cmd_t *zc;
5596 5586 uint_t vecnum;
5597 5587 int error, rc, len;
5598 5588 minor_t minor = getminor(dev);
5599 5589 const zfs_ioc_vec_t *vec;
5600 5590 char *saved_poolname = NULL;
5601 5591 nvlist_t *innvl = NULL;
5602 5592
5603 5593 if (minor != 0 &&
5604 5594 zfsdev_get_soft_state(minor, ZSST_CTLDEV) == NULL)
5605 5595 return (zvol_ioctl(dev, cmd, arg, flag, cr, rvalp));
5606 5596
5607 5597 vecnum = cmd - ZFS_IOC_FIRST;
5608 5598 ASSERT3U(getmajor(dev), ==, ddi_driver_major(zfs_dip));
5609 5599
5610 5600 if (vecnum >= sizeof (zfs_ioc_vec) / sizeof (zfs_ioc_vec[0]))
5611 5601 return (SET_ERROR(EINVAL));
5612 5602 vec = &zfs_ioc_vec[vecnum];
5613 5603
5614 5604 zc = kmem_zalloc(sizeof (zfs_cmd_t), KM_SLEEP);
5615 5605
5616 5606 error = ddi_copyin((void *)arg, zc, sizeof (zfs_cmd_t), flag);
5617 5607 if (error != 0) {
5618 5608 error = SET_ERROR(EFAULT);
5619 5609 goto out;
5620 5610 }
5621 5611
5622 5612 zc->zc_iflags = flag & FKIOCTL;
5623 5613 if (zc->zc_nvlist_src_size != 0) {
5624 5614 error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
5625 5615 zc->zc_iflags, &innvl);
5626 5616 if (error != 0)
5627 5617 goto out;
5628 5618 }
5629 5619
5630 5620 /*
5631 5621 * Ensure that all pool/dataset names are valid before we pass down to
5632 5622 * the lower layers.
5633 5623 */
5634 5624 zc->zc_name[sizeof (zc->zc_name) - 1] = '\0';
5635 5625 switch (vec->zvec_namecheck) {
5636 5626 case POOL_NAME:
5637 5627 if (pool_namecheck(zc->zc_name, NULL, NULL) != 0)
5638 5628 error = SET_ERROR(EINVAL);
5639 5629 else
5640 5630 error = pool_status_check(zc->zc_name,
5641 5631 vec->zvec_namecheck, vec->zvec_pool_check);
5642 5632 break;
5643 5633
5644 5634 case DATASET_NAME:
5645 5635 if (dataset_namecheck(zc->zc_name, NULL, NULL) != 0)
5646 5636 error = SET_ERROR(EINVAL);
5647 5637 else
5648 5638 error = pool_status_check(zc->zc_name,
5649 5639 vec->zvec_namecheck, vec->zvec_pool_check);
5650 5640 break;
5651 5641
5652 5642 case NO_NAME:
5653 5643 break;
5654 5644 }
5655 5645
5656 5646
5657 5647 if (error == 0 && !(flag & FKIOCTL))
5658 5648 error = vec->zvec_secpolicy(zc, innvl, cr);
5659 5649
5660 5650 if (error != 0)
5661 5651 goto out;
5662 5652
5663 5653 /* legacy ioctls can modify zc_name */
5664 5654 len = strcspn(zc->zc_name, "/@") + 1;
5665 5655 saved_poolname = kmem_alloc(len, KM_SLEEP);
5666 5656 (void) strlcpy(saved_poolname, zc->zc_name, len);
5667 5657
5668 5658 if (vec->zvec_func != NULL) {
5669 5659 nvlist_t *outnvl;
5670 5660 int puterror = 0;
5671 5661 spa_t *spa;
5672 5662 nvlist_t *lognv = NULL;
5673 5663
5674 5664 ASSERT(vec->zvec_legacy_func == NULL);
5675 5665
5676 5666 /*
5677 5667 * Add the innvl to the lognv before calling the func,
5678 5668 * in case the func changes the innvl.
5679 5669 */
5680 5670 if (vec->zvec_allow_log) {
5681 5671 lognv = fnvlist_alloc();
5682 5672 fnvlist_add_string(lognv, ZPOOL_HIST_IOCTL,
5683 5673 vec->zvec_name);
5684 5674 if (!nvlist_empty(innvl)) {
5685 5675 fnvlist_add_nvlist(lognv, ZPOOL_HIST_INPUT_NVL,
5686 5676 innvl);
5687 5677 }
5688 5678 }
5689 5679
5690 5680 outnvl = fnvlist_alloc();
5691 5681 error = vec->zvec_func(zc->zc_name, innvl, outnvl);
5692 5682
5693 5683 if (error == 0 && vec->zvec_allow_log &&
5694 5684 spa_open(zc->zc_name, &spa, FTAG) == 0) {
5695 5685 if (!nvlist_empty(outnvl)) {
5696 5686 fnvlist_add_nvlist(lognv, ZPOOL_HIST_OUTPUT_NVL,
5697 5687 outnvl);
5698 5688 }
5699 5689 (void) spa_history_log_nvl(spa, lognv);
5700 5690 spa_close(spa, FTAG);
5701 5691 }
5702 5692 fnvlist_free(lognv);
5703 5693
5704 5694 if (!nvlist_empty(outnvl) || zc->zc_nvlist_dst_size != 0) {
5705 5695 int smusherror = 0;
5706 5696 if (vec->zvec_smush_outnvlist) {
5707 5697 smusherror = nvlist_smush(outnvl,
5708 5698 zc->zc_nvlist_dst_size);
5709 5699 }
5710 5700 if (smusherror == 0)
5711 5701 puterror = put_nvlist(zc, outnvl);
5712 5702 }
5713 5703
5714 5704 if (puterror != 0)
5715 5705 error = puterror;
5716 5706
5717 5707 nvlist_free(outnvl);
5718 5708 } else {
5719 5709 error = vec->zvec_legacy_func(zc);
5720 5710 }
5721 5711
5722 5712 out:
5723 5713 nvlist_free(innvl);
5724 5714 rc = ddi_copyout(zc, (void *)arg, sizeof (zfs_cmd_t), flag);
5725 5715 if (error == 0 && rc != 0)
5726 5716 error = SET_ERROR(EFAULT);
5727 5717 if (error == 0 && vec->zvec_allow_log) {
5728 5718 char *s = tsd_get(zfs_allow_log_key);
5729 5719 if (s != NULL)
5730 5720 strfree(s);
5731 5721 (void) tsd_set(zfs_allow_log_key, saved_poolname);
5732 5722 } else {
5733 5723 if (saved_poolname != NULL)
5734 5724 strfree(saved_poolname);
5735 5725 }
5736 5726
5737 5727 kmem_free(zc, sizeof (zfs_cmd_t));
5738 5728 return (error);
5739 5729 }
5740 5730
5741 5731 static int
5742 5732 zfs_attach(dev_info_t *dip, ddi_attach_cmd_t cmd)
5743 5733 {
5744 5734 if (cmd != DDI_ATTACH)
5745 5735 return (DDI_FAILURE);
5746 5736
5747 5737 if (ddi_create_minor_node(dip, "zfs", S_IFCHR, 0,
5748 5738 DDI_PSEUDO, 0) == DDI_FAILURE)
5749 5739 return (DDI_FAILURE);
5750 5740
5751 5741 zfs_dip = dip;
5752 5742
5753 5743 ddi_report_dev(dip);
5754 5744
5755 5745 return (DDI_SUCCESS);
5756 5746 }
5757 5747
5758 5748 static int
5759 5749 zfs_detach(dev_info_t *dip, ddi_detach_cmd_t cmd)
5760 5750 {
5761 5751 if (spa_busy() || zfs_busy() || zvol_busy())
5762 5752 return (DDI_FAILURE);
5763 5753
5764 5754 if (cmd != DDI_DETACH)
5765 5755 return (DDI_FAILURE);
5766 5756
5767 5757 zfs_dip = NULL;
5768 5758
5769 5759 ddi_prop_remove_all(dip);
5770 5760 ddi_remove_minor_node(dip, NULL);
5771 5761
5772 5762 return (DDI_SUCCESS);
5773 5763 }
5774 5764
5775 5765 /*ARGSUSED*/
5776 5766 static int
5777 5767 zfs_info(dev_info_t *dip, ddi_info_cmd_t infocmd, void *arg, void **result)
5778 5768 {
5779 5769 switch (infocmd) {
5780 5770 case DDI_INFO_DEVT2DEVINFO:
5781 5771 *result = zfs_dip;
5782 5772 return (DDI_SUCCESS);
5783 5773
5784 5774 case DDI_INFO_DEVT2INSTANCE:
5785 5775 *result = (void *)0;
5786 5776 return (DDI_SUCCESS);
5787 5777 }
5788 5778
5789 5779 return (DDI_FAILURE);
5790 5780 }
5791 5781
5792 5782 /*
5793 5783 * OK, so this is a little weird.
5794 5784 *
5795 5785 * /dev/zfs is the control node, i.e. minor 0.
5796 5786 * /dev/zvol/[r]dsk/pool/dataset are the zvols, minor > 0.
5797 5787 *
5798 5788 * /dev/zfs has basically nothing to do except serve up ioctls,
5799 5789 * so most of the standard driver entry points are in zvol.c.
5800 5790 */
5801 5791 static struct cb_ops zfs_cb_ops = {
5802 5792 zfsdev_open, /* open */
5803 5793 zfsdev_close, /* close */
5804 5794 zvol_strategy, /* strategy */
5805 5795 nodev, /* print */
5806 5796 zvol_dump, /* dump */
5807 5797 zvol_read, /* read */
5808 5798 zvol_write, /* write */
5809 5799 zfsdev_ioctl, /* ioctl */
5810 5800 nodev, /* devmap */
5811 5801 nodev, /* mmap */
5812 5802 nodev, /* segmap */
5813 5803 nochpoll, /* poll */
5814 5804 ddi_prop_op, /* prop_op */
5815 5805 NULL, /* streamtab */
5816 5806 D_NEW | D_MP | D_64BIT, /* Driver compatibility flag */
5817 5807 CB_REV, /* version */
5818 5808 nodev, /* async read */
5819 5809 nodev, /* async write */
5820 5810 };
5821 5811
5822 5812 static struct dev_ops zfs_dev_ops = {
5823 5813 DEVO_REV, /* version */
5824 5814 0, /* refcnt */
5825 5815 zfs_info, /* info */
5826 5816 nulldev, /* identify */
5827 5817 nulldev, /* probe */
5828 5818 zfs_attach, /* attach */
5829 5819 zfs_detach, /* detach */
5830 5820 nodev, /* reset */
5831 5821 &zfs_cb_ops, /* driver operations */
5832 5822 NULL, /* no bus operations */
5833 5823 NULL, /* power */
5834 5824 ddi_quiesce_not_needed, /* quiesce */
5835 5825 };
5836 5826
5837 5827 static struct modldrv zfs_modldrv = {
5838 5828 &mod_driverops,
5839 5829 "ZFS storage pool",
5840 5830 &zfs_dev_ops
5841 5831 };
5842 5832
5843 5833 static struct modlinkage modlinkage = {
5844 5834 MODREV_1,
5845 5835 (void *)&zfs_modlfs,
5846 5836 (void *)&zfs_modldrv,
5847 5837 NULL
5848 5838 };
5849 5839
5850 5840 static void
5851 5841 zfs_allow_log_destroy(void *arg)
5852 5842 {
5853 5843 char *poolname = arg;
5854 5844 strfree(poolname);
5855 5845 }
5856 5846
5857 5847 int
5858 5848 _init(void)
5859 5849 {
5860 5850 int error;
5861 5851
5862 5852 spa_init(FREAD | FWRITE);
5863 5853 zfs_init();
5864 5854 zvol_init();
5865 5855 zfs_ioctl_init();
5866 5856
5867 5857 if ((error = mod_install(&modlinkage)) != 0) {
5868 5858 zvol_fini();
5869 5859 zfs_fini();
5870 5860 spa_fini();
5871 5861 return (error);
5872 5862 }
5873 5863
5874 5864 tsd_create(&zfs_fsyncer_key, NULL);
5875 5865 tsd_create(&rrw_tsd_key, rrw_tsd_destroy);
5876 5866 tsd_create(&zfs_allow_log_key, zfs_allow_log_destroy);
5877 5867
5878 5868 error = ldi_ident_from_mod(&modlinkage, &zfs_li);
5879 5869 ASSERT(error == 0);
5880 5870 mutex_init(&zfs_share_lock, NULL, MUTEX_DEFAULT, NULL);
5881 5871
5882 5872 return (0);
5883 5873 }
5884 5874
5885 5875 int
5886 5876 _fini(void)
5887 5877 {
5888 5878 int error;
5889 5879
5890 5880 if (spa_busy() || zfs_busy() || zvol_busy() || zio_injection_enabled)
5891 5881 return (SET_ERROR(EBUSY));
5892 5882
5893 5883 if ((error = mod_remove(&modlinkage)) != 0)
5894 5884 return (error);
5895 5885
5896 5886 zvol_fini();
5897 5887 zfs_fini();
5898 5888 spa_fini();
5899 5889 if (zfs_nfsshare_inited)
5900 5890 (void) ddi_modclose(nfs_mod);
5901 5891 if (zfs_smbshare_inited)
5902 5892 (void) ddi_modclose(smbsrv_mod);
5903 5893 if (zfs_nfsshare_inited || zfs_smbshare_inited)
5904 5894 (void) ddi_modclose(sharefs_mod);
5905 5895
5906 5896 tsd_destroy(&zfs_fsyncer_key);
5907 5897 ldi_ident_release(zfs_li);
5908 5898 zfs_li = NULL;
5909 5899 mutex_destroy(&zfs_share_lock);
5910 5900
5911 5901 return (error);
5912 5902 }
5913 5903
5914 5904 int
5915 5905 _info(struct modinfo *modinfop)
5916 5906 {
5917 5907 return (mod_info(&modlinkage, modinfop));
5918 5908 }
|
↓ open down ↓ |
916 lines elided |
↑ open up ↑ |
XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX