Print this page
6659 nvlist_free(NULL) is a no-op
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 2015, OmniTI Computer Consulting, Inc. All rights reserved.
26 26 * Copyright 2015 Nexenta Systems, Inc. All rights reserved.
27 27 * Copyright (c) 2014, Joyent, Inc. All rights reserved.
28 28 * Copyright (c) 2011, 2015 by Delphix. All rights reserved.
29 29 * Copyright (c) 2013 by Saso Kiselkov. All rights reserved.
30 30 * Copyright (c) 2013 Steven Hartland. All rights reserved.
31 31 * Copyright (c) 2014 Integros [integros.com]
32 32 */
33 33
34 34 /*
35 35 * ZFS ioctls.
36 36 *
37 37 * This file handles the ioctls to /dev/zfs, used for configuring ZFS storage
38 38 * pools and filesystems, e.g. with /sbin/zfs and /sbin/zpool.
39 39 *
40 40 * There are two ways that we handle ioctls: the legacy way where almost
41 41 * all of the logic is in the ioctl callback, and the new way where most
42 42 * of the marshalling is handled in the common entry point, zfsdev_ioctl().
43 43 *
44 44 * Non-legacy ioctls should be registered by calling
45 45 * zfs_ioctl_register() from zfs_ioctl_init(). The ioctl is invoked
46 46 * from userland by lzc_ioctl().
47 47 *
48 48 * The registration arguments are as follows:
49 49 *
50 50 * const char *name
51 51 * The name of the ioctl. This is used for history logging. If the
52 52 * ioctl returns successfully (the callback returns 0), and allow_log
53 53 * is true, then a history log entry will be recorded with the input &
54 54 * output nvlists. The log entry can be printed with "zpool history -i".
55 55 *
56 56 * zfs_ioc_t ioc
57 57 * The ioctl request number, which userland will pass to ioctl(2).
58 58 * The ioctl numbers can change from release to release, because
59 59 * the caller (libzfs) must be matched to the kernel.
60 60 *
61 61 * zfs_secpolicy_func_t *secpolicy
62 62 * This function will be called before the zfs_ioc_func_t, to
63 63 * determine if this operation is permitted. It should return EPERM
64 64 * on failure, and 0 on success. Checks include determining if the
65 65 * dataset is visible in this zone, and if the user has either all
66 66 * zfs privileges in the zone (SYS_MOUNT), or has been granted permission
67 67 * to do this operation on this dataset with "zfs allow".
68 68 *
69 69 * zfs_ioc_namecheck_t namecheck
70 70 * This specifies what to expect in the zfs_cmd_t:zc_name -- a pool
71 71 * name, a dataset name, or nothing. If the name is not well-formed,
72 72 * the ioctl will fail and the callback will not be called.
73 73 * Therefore, the callback can assume that the name is well-formed
74 74 * (e.g. is null-terminated, doesn't have more than one '@' character,
75 75 * doesn't have invalid characters).
76 76 *
77 77 * zfs_ioc_poolcheck_t pool_check
78 78 * This specifies requirements on the pool state. If the pool does
79 79 * not meet them (is suspended or is readonly), the ioctl will fail
80 80 * and the callback will not be called. If any checks are specified
81 81 * (i.e. it is not POOL_CHECK_NONE), namecheck must not be NO_NAME.
82 82 * Multiple checks can be or-ed together (e.g. POOL_CHECK_SUSPENDED |
83 83 * POOL_CHECK_READONLY).
84 84 *
85 85 * boolean_t smush_outnvlist
86 86 * If smush_outnvlist is true, then the output is presumed to be a
87 87 * list of errors, and it will be "smushed" down to fit into the
88 88 * caller's buffer, by removing some entries and replacing them with a
89 89 * single "N_MORE_ERRORS" entry indicating how many were removed. See
90 90 * nvlist_smush() for details. If smush_outnvlist is false, and the
91 91 * outnvlist does not fit into the userland-provided buffer, then the
92 92 * ioctl will fail with ENOMEM.
93 93 *
94 94 * zfs_ioc_func_t *func
95 95 * The callback function that will perform the operation.
96 96 *
97 97 * The callback should return 0 on success, or an error number on
98 98 * failure. If the function fails, the userland ioctl will return -1,
99 99 * and errno will be set to the callback's return value. The callback
100 100 * will be called with the following arguments:
101 101 *
102 102 * const char *name
103 103 * The name of the pool or dataset to operate on, from
104 104 * zfs_cmd_t:zc_name. The 'namecheck' argument specifies the
105 105 * expected type (pool, dataset, or none).
106 106 *
107 107 * nvlist_t *innvl
108 108 * The input nvlist, deserialized from zfs_cmd_t:zc_nvlist_src. Or
109 109 * NULL if no input nvlist was provided. Changes to this nvlist are
110 110 * ignored. If the input nvlist could not be deserialized, the
111 111 * ioctl will fail and the callback will not be called.
112 112 *
113 113 * nvlist_t *outnvl
114 114 * The output nvlist, initially empty. The callback can fill it in,
115 115 * and it will be returned to userland by serializing it into
116 116 * zfs_cmd_t:zc_nvlist_dst. If it is non-empty, and serialization
117 117 * fails (e.g. because the caller didn't supply a large enough
118 118 * buffer), then the overall ioctl will fail. See the
119 119 * 'smush_nvlist' argument above for additional behaviors.
120 120 *
121 121 * There are two typical uses of the output nvlist:
122 122 * - To return state, e.g. property values. In this case,
123 123 * smush_outnvlist should be false. If the buffer was not large
124 124 * enough, the caller will reallocate a larger buffer and try
125 125 * the ioctl again.
126 126 *
127 127 * - To return multiple errors from an ioctl which makes on-disk
128 128 * changes. In this case, smush_outnvlist should be true.
129 129 * Ioctls which make on-disk modifications should generally not
130 130 * use the outnvl if they succeed, because the caller can not
131 131 * distinguish between the operation failing, and
132 132 * deserialization failing.
133 133 */
134 134
135 135 #include <sys/types.h>
136 136 #include <sys/param.h>
137 137 #include <sys/errno.h>
138 138 #include <sys/uio.h>
139 139 #include <sys/buf.h>
140 140 #include <sys/modctl.h>
141 141 #include <sys/open.h>
142 142 #include <sys/file.h>
143 143 #include <sys/kmem.h>
144 144 #include <sys/conf.h>
145 145 #include <sys/cmn_err.h>
146 146 #include <sys/stat.h>
147 147 #include <sys/zfs_ioctl.h>
148 148 #include <sys/zfs_vfsops.h>
149 149 #include <sys/zfs_znode.h>
150 150 #include <sys/zap.h>
151 151 #include <sys/spa.h>
152 152 #include <sys/spa_impl.h>
153 153 #include <sys/vdev.h>
154 154 #include <sys/priv_impl.h>
155 155 #include <sys/dmu.h>
156 156 #include <sys/dsl_dir.h>
157 157 #include <sys/dsl_dataset.h>
158 158 #include <sys/dsl_prop.h>
159 159 #include <sys/dsl_deleg.h>
160 160 #include <sys/dmu_objset.h>
161 161 #include <sys/dmu_impl.h>
162 162 #include <sys/dmu_tx.h>
163 163 #include <sys/ddi.h>
164 164 #include <sys/sunddi.h>
165 165 #include <sys/sunldi.h>
166 166 #include <sys/policy.h>
167 167 #include <sys/zone.h>
168 168 #include <sys/nvpair.h>
169 169 #include <sys/pathname.h>
170 170 #include <sys/mount.h>
171 171 #include <sys/sdt.h>
172 172 #include <sys/fs/zfs.h>
173 173 #include <sys/zfs_ctldir.h>
174 174 #include <sys/zfs_dir.h>
175 175 #include <sys/zfs_onexit.h>
176 176 #include <sys/zvol.h>
177 177 #include <sys/dsl_scan.h>
178 178 #include <sharefs/share.h>
179 179 #include <sys/dmu_objset.h>
180 180 #include <sys/dmu_send.h>
181 181 #include <sys/dsl_destroy.h>
182 182 #include <sys/dsl_bookmark.h>
183 183 #include <sys/dsl_userhold.h>
184 184 #include <sys/zfeature.h>
185 185 #include <sys/zio_checksum.h>
186 186
187 187 #include "zfs_namecheck.h"
188 188 #include "zfs_prop.h"
189 189 #include "zfs_deleg.h"
190 190 #include "zfs_comutil.h"
191 191
192 192 extern struct modlfs zfs_modlfs;
193 193
194 194 extern void zfs_init(void);
195 195 extern void zfs_fini(void);
196 196
197 197 ldi_ident_t zfs_li = NULL;
198 198 dev_info_t *zfs_dip;
199 199
200 200 uint_t zfs_fsyncer_key;
201 201 extern uint_t rrw_tsd_key;
202 202 static uint_t zfs_allow_log_key;
203 203
204 204 typedef int zfs_ioc_legacy_func_t(zfs_cmd_t *);
205 205 typedef int zfs_ioc_func_t(const char *, nvlist_t *, nvlist_t *);
206 206 typedef int zfs_secpolicy_func_t(zfs_cmd_t *, nvlist_t *, cred_t *);
207 207
208 208 typedef enum {
209 209 NO_NAME,
210 210 POOL_NAME,
211 211 DATASET_NAME
212 212 } zfs_ioc_namecheck_t;
213 213
214 214 typedef enum {
215 215 POOL_CHECK_NONE = 1 << 0,
216 216 POOL_CHECK_SUSPENDED = 1 << 1,
217 217 POOL_CHECK_READONLY = 1 << 2,
218 218 } zfs_ioc_poolcheck_t;
219 219
220 220 typedef struct zfs_ioc_vec {
221 221 zfs_ioc_legacy_func_t *zvec_legacy_func;
222 222 zfs_ioc_func_t *zvec_func;
223 223 zfs_secpolicy_func_t *zvec_secpolicy;
224 224 zfs_ioc_namecheck_t zvec_namecheck;
225 225 boolean_t zvec_allow_log;
226 226 zfs_ioc_poolcheck_t zvec_pool_check;
227 227 boolean_t zvec_smush_outnvlist;
228 228 const char *zvec_name;
229 229 } zfs_ioc_vec_t;
230 230
231 231 /* This array is indexed by zfs_userquota_prop_t */
232 232 static const char *userquota_perms[] = {
233 233 ZFS_DELEG_PERM_USERUSED,
234 234 ZFS_DELEG_PERM_USERQUOTA,
235 235 ZFS_DELEG_PERM_GROUPUSED,
236 236 ZFS_DELEG_PERM_GROUPQUOTA,
237 237 };
238 238
239 239 static int zfs_ioc_userspace_upgrade(zfs_cmd_t *zc);
240 240 static int zfs_check_settable(const char *name, nvpair_t *property,
241 241 cred_t *cr);
242 242 static int zfs_check_clearable(char *dataset, nvlist_t *props,
243 243 nvlist_t **errors);
244 244 static int zfs_fill_zplprops_root(uint64_t, nvlist_t *, nvlist_t *,
245 245 boolean_t *);
246 246 int zfs_set_prop_nvlist(const char *, zprop_source_t, nvlist_t *, nvlist_t *);
247 247 static int get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp);
248 248
249 249 static int zfs_prop_activate_feature(spa_t *spa, spa_feature_t feature);
250 250
251 251 /* _NOTE(PRINTFLIKE(4)) - this is printf-like, but lint is too whiney */
252 252 void
253 253 __dprintf(const char *file, const char *func, int line, const char *fmt, ...)
254 254 {
255 255 const char *newfile;
256 256 char buf[512];
257 257 va_list adx;
258 258
259 259 /*
260 260 * Get rid of annoying "../common/" prefix to filename.
261 261 */
262 262 newfile = strrchr(file, '/');
263 263 if (newfile != NULL) {
264 264 newfile = newfile + 1; /* Get rid of leading / */
265 265 } else {
266 266 newfile = file;
267 267 }
268 268
269 269 va_start(adx, fmt);
270 270 (void) vsnprintf(buf, sizeof (buf), fmt, adx);
271 271 va_end(adx);
272 272
273 273 /*
274 274 * To get this data, use the zfs-dprintf probe as so:
275 275 * dtrace -q -n 'zfs-dprintf \
276 276 * /stringof(arg0) == "dbuf.c"/ \
277 277 * {printf("%s: %s", stringof(arg1), stringof(arg3))}'
278 278 * arg0 = file name
279 279 * arg1 = function name
280 280 * arg2 = line number
281 281 * arg3 = message
282 282 */
283 283 DTRACE_PROBE4(zfs__dprintf,
284 284 char *, newfile, char *, func, int, line, char *, buf);
285 285 }
286 286
287 287 static void
288 288 history_str_free(char *buf)
289 289 {
290 290 kmem_free(buf, HIS_MAX_RECORD_LEN);
291 291 }
292 292
293 293 static char *
294 294 history_str_get(zfs_cmd_t *zc)
295 295 {
296 296 char *buf;
297 297
298 298 if (zc->zc_history == NULL)
299 299 return (NULL);
300 300
301 301 buf = kmem_alloc(HIS_MAX_RECORD_LEN, KM_SLEEP);
302 302 if (copyinstr((void *)(uintptr_t)zc->zc_history,
303 303 buf, HIS_MAX_RECORD_LEN, NULL) != 0) {
304 304 history_str_free(buf);
305 305 return (NULL);
306 306 }
307 307
308 308 buf[HIS_MAX_RECORD_LEN -1] = '\0';
309 309
310 310 return (buf);
311 311 }
312 312
313 313 /*
314 314 * Check to see if the named dataset is currently defined as bootable
315 315 */
316 316 static boolean_t
317 317 zfs_is_bootfs(const char *name)
318 318 {
319 319 objset_t *os;
320 320
321 321 if (dmu_objset_hold(name, FTAG, &os) == 0) {
322 322 boolean_t ret;
323 323 ret = (dmu_objset_id(os) == spa_bootfs(dmu_objset_spa(os)));
324 324 dmu_objset_rele(os, FTAG);
325 325 return (ret);
326 326 }
327 327 return (B_FALSE);
328 328 }
329 329
330 330 /*
331 331 * Return non-zero if the spa version is less than requested version.
332 332 */
333 333 static int
334 334 zfs_earlier_version(const char *name, int version)
335 335 {
336 336 spa_t *spa;
337 337
338 338 if (spa_open(name, &spa, FTAG) == 0) {
339 339 if (spa_version(spa) < version) {
340 340 spa_close(spa, FTAG);
341 341 return (1);
342 342 }
343 343 spa_close(spa, FTAG);
344 344 }
345 345 return (0);
346 346 }
347 347
348 348 /*
349 349 * Return TRUE if the ZPL version is less than requested version.
350 350 */
351 351 static boolean_t
352 352 zpl_earlier_version(const char *name, int version)
353 353 {
354 354 objset_t *os;
355 355 boolean_t rc = B_TRUE;
356 356
357 357 if (dmu_objset_hold(name, FTAG, &os) == 0) {
358 358 uint64_t zplversion;
359 359
360 360 if (dmu_objset_type(os) != DMU_OST_ZFS) {
361 361 dmu_objset_rele(os, FTAG);
362 362 return (B_TRUE);
363 363 }
364 364 /* XXX reading from non-owned objset */
365 365 if (zfs_get_zplprop(os, ZFS_PROP_VERSION, &zplversion) == 0)
366 366 rc = zplversion < version;
367 367 dmu_objset_rele(os, FTAG);
368 368 }
369 369 return (rc);
370 370 }
371 371
372 372 static void
373 373 zfs_log_history(zfs_cmd_t *zc)
374 374 {
375 375 spa_t *spa;
376 376 char *buf;
377 377
378 378 if ((buf = history_str_get(zc)) == NULL)
379 379 return;
380 380
381 381 if (spa_open(zc->zc_name, &spa, FTAG) == 0) {
382 382 if (spa_version(spa) >= SPA_VERSION_ZPOOL_HISTORY)
383 383 (void) spa_history_log(spa, buf);
384 384 spa_close(spa, FTAG);
385 385 }
386 386 history_str_free(buf);
387 387 }
388 388
389 389 /*
390 390 * Policy for top-level read operations (list pools). Requires no privileges,
391 391 * and can be used in the local zone, as there is no associated dataset.
392 392 */
393 393 /* ARGSUSED */
394 394 static int
395 395 zfs_secpolicy_none(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
396 396 {
397 397 return (0);
398 398 }
399 399
400 400 /*
401 401 * Policy for dataset read operations (list children, get statistics). Requires
402 402 * no privileges, but must be visible in the local zone.
403 403 */
404 404 /* ARGSUSED */
405 405 static int
406 406 zfs_secpolicy_read(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
407 407 {
408 408 if (INGLOBALZONE(curproc) ||
409 409 zone_dataset_visible(zc->zc_name, NULL))
410 410 return (0);
411 411
412 412 return (SET_ERROR(ENOENT));
413 413 }
414 414
415 415 static int
416 416 zfs_dozonecheck_impl(const char *dataset, uint64_t zoned, cred_t *cr)
417 417 {
418 418 int writable = 1;
419 419
420 420 /*
421 421 * The dataset must be visible by this zone -- check this first
422 422 * so they don't see EPERM on something they shouldn't know about.
423 423 */
424 424 if (!INGLOBALZONE(curproc) &&
425 425 !zone_dataset_visible(dataset, &writable))
426 426 return (SET_ERROR(ENOENT));
427 427
428 428 if (INGLOBALZONE(curproc)) {
429 429 /*
430 430 * If the fs is zoned, only root can access it from the
431 431 * global zone.
432 432 */
433 433 if (secpolicy_zfs(cr) && zoned)
434 434 return (SET_ERROR(EPERM));
435 435 } else {
436 436 /*
437 437 * If we are in a local zone, the 'zoned' property must be set.
438 438 */
439 439 if (!zoned)
440 440 return (SET_ERROR(EPERM));
441 441
442 442 /* must be writable by this zone */
443 443 if (!writable)
444 444 return (SET_ERROR(EPERM));
445 445 }
446 446 return (0);
447 447 }
448 448
449 449 static int
450 450 zfs_dozonecheck(const char *dataset, cred_t *cr)
451 451 {
452 452 uint64_t zoned;
453 453
454 454 if (dsl_prop_get_integer(dataset, "zoned", &zoned, NULL))
455 455 return (SET_ERROR(ENOENT));
456 456
457 457 return (zfs_dozonecheck_impl(dataset, zoned, cr));
458 458 }
459 459
460 460 static int
461 461 zfs_dozonecheck_ds(const char *dataset, dsl_dataset_t *ds, cred_t *cr)
462 462 {
463 463 uint64_t zoned;
464 464
465 465 if (dsl_prop_get_int_ds(ds, "zoned", &zoned))
466 466 return (SET_ERROR(ENOENT));
467 467
468 468 return (zfs_dozonecheck_impl(dataset, zoned, cr));
469 469 }
470 470
471 471 static int
472 472 zfs_secpolicy_write_perms_ds(const char *name, dsl_dataset_t *ds,
473 473 const char *perm, cred_t *cr)
474 474 {
475 475 int error;
476 476
477 477 error = zfs_dozonecheck_ds(name, ds, cr);
478 478 if (error == 0) {
479 479 error = secpolicy_zfs(cr);
480 480 if (error != 0)
481 481 error = dsl_deleg_access_impl(ds, perm, cr);
482 482 }
483 483 return (error);
484 484 }
485 485
486 486 static int
487 487 zfs_secpolicy_write_perms(const char *name, const char *perm, cred_t *cr)
488 488 {
489 489 int error;
490 490 dsl_dataset_t *ds;
491 491 dsl_pool_t *dp;
492 492
493 493 error = dsl_pool_hold(name, FTAG, &dp);
494 494 if (error != 0)
495 495 return (error);
496 496
497 497 error = dsl_dataset_hold(dp, name, FTAG, &ds);
498 498 if (error != 0) {
499 499 dsl_pool_rele(dp, FTAG);
500 500 return (error);
501 501 }
502 502
503 503 error = zfs_secpolicy_write_perms_ds(name, ds, perm, cr);
504 504
505 505 dsl_dataset_rele(ds, FTAG);
506 506 dsl_pool_rele(dp, FTAG);
507 507 return (error);
508 508 }
509 509
510 510 /*
511 511 * Policy for setting the security label property.
512 512 *
513 513 * Returns 0 for success, non-zero for access and other errors.
514 514 */
515 515 static int
516 516 zfs_set_slabel_policy(const char *name, char *strval, cred_t *cr)
517 517 {
518 518 char ds_hexsl[MAXNAMELEN];
519 519 bslabel_t ds_sl, new_sl;
520 520 boolean_t new_default = FALSE;
521 521 uint64_t zoned;
522 522 int needed_priv = -1;
523 523 int error;
524 524
525 525 /* First get the existing dataset label. */
526 526 error = dsl_prop_get(name, zfs_prop_to_name(ZFS_PROP_MLSLABEL),
527 527 1, sizeof (ds_hexsl), &ds_hexsl, NULL);
528 528 if (error != 0)
529 529 return (SET_ERROR(EPERM));
530 530
531 531 if (strcasecmp(strval, ZFS_MLSLABEL_DEFAULT) == 0)
532 532 new_default = TRUE;
533 533
534 534 /* The label must be translatable */
535 535 if (!new_default && (hexstr_to_label(strval, &new_sl) != 0))
536 536 return (SET_ERROR(EINVAL));
537 537
538 538 /*
539 539 * In a non-global zone, disallow attempts to set a label that
540 540 * doesn't match that of the zone; otherwise no other checks
541 541 * are needed.
542 542 */
543 543 if (!INGLOBALZONE(curproc)) {
544 544 if (new_default || !blequal(&new_sl, CR_SL(CRED())))
545 545 return (SET_ERROR(EPERM));
546 546 return (0);
547 547 }
548 548
549 549 /*
550 550 * For global-zone datasets (i.e., those whose zoned property is
551 551 * "off", verify that the specified new label is valid for the
552 552 * global zone.
553 553 */
554 554 if (dsl_prop_get_integer(name,
555 555 zfs_prop_to_name(ZFS_PROP_ZONED), &zoned, NULL))
556 556 return (SET_ERROR(EPERM));
557 557 if (!zoned) {
558 558 if (zfs_check_global_label(name, strval) != 0)
559 559 return (SET_ERROR(EPERM));
560 560 }
561 561
562 562 /*
563 563 * If the existing dataset label is nondefault, check if the
564 564 * dataset is mounted (label cannot be changed while mounted).
565 565 * Get the zfsvfs; if there isn't one, then the dataset isn't
566 566 * mounted (or isn't a dataset, doesn't exist, ...).
567 567 */
568 568 if (strcasecmp(ds_hexsl, ZFS_MLSLABEL_DEFAULT) != 0) {
569 569 objset_t *os;
570 570 static char *setsl_tag = "setsl_tag";
571 571
572 572 /*
573 573 * Try to own the dataset; abort if there is any error,
574 574 * (e.g., already mounted, in use, or other error).
575 575 */
576 576 error = dmu_objset_own(name, DMU_OST_ZFS, B_TRUE,
577 577 setsl_tag, &os);
578 578 if (error != 0)
579 579 return (SET_ERROR(EPERM));
580 580
581 581 dmu_objset_disown(os, setsl_tag);
582 582
583 583 if (new_default) {
584 584 needed_priv = PRIV_FILE_DOWNGRADE_SL;
585 585 goto out_check;
586 586 }
587 587
588 588 if (hexstr_to_label(strval, &new_sl) != 0)
589 589 return (SET_ERROR(EPERM));
590 590
591 591 if (blstrictdom(&ds_sl, &new_sl))
592 592 needed_priv = PRIV_FILE_DOWNGRADE_SL;
593 593 else if (blstrictdom(&new_sl, &ds_sl))
594 594 needed_priv = PRIV_FILE_UPGRADE_SL;
595 595 } else {
596 596 /* dataset currently has a default label */
597 597 if (!new_default)
598 598 needed_priv = PRIV_FILE_UPGRADE_SL;
599 599 }
600 600
601 601 out_check:
602 602 if (needed_priv != -1)
603 603 return (PRIV_POLICY(cr, needed_priv, B_FALSE, EPERM, NULL));
604 604 return (0);
605 605 }
606 606
607 607 static int
608 608 zfs_secpolicy_setprop(const char *dsname, zfs_prop_t prop, nvpair_t *propval,
609 609 cred_t *cr)
610 610 {
611 611 char *strval;
612 612
613 613 /*
614 614 * Check permissions for special properties.
615 615 */
616 616 switch (prop) {
617 617 case ZFS_PROP_ZONED:
618 618 /*
619 619 * Disallow setting of 'zoned' from within a local zone.
620 620 */
621 621 if (!INGLOBALZONE(curproc))
622 622 return (SET_ERROR(EPERM));
623 623 break;
624 624
625 625 case ZFS_PROP_QUOTA:
626 626 case ZFS_PROP_FILESYSTEM_LIMIT:
627 627 case ZFS_PROP_SNAPSHOT_LIMIT:
628 628 if (!INGLOBALZONE(curproc)) {
629 629 uint64_t zoned;
630 630 char setpoint[MAXNAMELEN];
631 631 /*
632 632 * Unprivileged users are allowed to modify the
633 633 * limit on things *under* (ie. contained by)
634 634 * the thing they own.
635 635 */
636 636 if (dsl_prop_get_integer(dsname, "zoned", &zoned,
637 637 setpoint))
638 638 return (SET_ERROR(EPERM));
639 639 if (!zoned || strlen(dsname) <= strlen(setpoint))
640 640 return (SET_ERROR(EPERM));
641 641 }
642 642 break;
643 643
644 644 case ZFS_PROP_MLSLABEL:
645 645 if (!is_system_labeled())
646 646 return (SET_ERROR(EPERM));
647 647
648 648 if (nvpair_value_string(propval, &strval) == 0) {
649 649 int err;
650 650
651 651 err = zfs_set_slabel_policy(dsname, strval, CRED());
652 652 if (err != 0)
653 653 return (err);
654 654 }
655 655 break;
656 656 }
657 657
658 658 return (zfs_secpolicy_write_perms(dsname, zfs_prop_to_name(prop), cr));
659 659 }
660 660
661 661 /* ARGSUSED */
662 662 static int
663 663 zfs_secpolicy_set_fsacl(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
664 664 {
665 665 int error;
666 666
667 667 error = zfs_dozonecheck(zc->zc_name, cr);
668 668 if (error != 0)
669 669 return (error);
670 670
671 671 /*
672 672 * permission to set permissions will be evaluated later in
673 673 * dsl_deleg_can_allow()
674 674 */
675 675 return (0);
676 676 }
677 677
678 678 /* ARGSUSED */
679 679 static int
680 680 zfs_secpolicy_rollback(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
681 681 {
682 682 return (zfs_secpolicy_write_perms(zc->zc_name,
683 683 ZFS_DELEG_PERM_ROLLBACK, cr));
684 684 }
685 685
686 686 /* ARGSUSED */
687 687 static int
688 688 zfs_secpolicy_send(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
689 689 {
690 690 dsl_pool_t *dp;
691 691 dsl_dataset_t *ds;
692 692 char *cp;
693 693 int error;
694 694
695 695 /*
696 696 * Generate the current snapshot name from the given objsetid, then
697 697 * use that name for the secpolicy/zone checks.
698 698 */
699 699 cp = strchr(zc->zc_name, '@');
700 700 if (cp == NULL)
701 701 return (SET_ERROR(EINVAL));
702 702 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
703 703 if (error != 0)
704 704 return (error);
705 705
706 706 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &ds);
707 707 if (error != 0) {
708 708 dsl_pool_rele(dp, FTAG);
709 709 return (error);
710 710 }
711 711
712 712 dsl_dataset_name(ds, zc->zc_name);
713 713
714 714 error = zfs_secpolicy_write_perms_ds(zc->zc_name, ds,
715 715 ZFS_DELEG_PERM_SEND, cr);
716 716 dsl_dataset_rele(ds, FTAG);
717 717 dsl_pool_rele(dp, FTAG);
718 718
719 719 return (error);
720 720 }
721 721
722 722 /* ARGSUSED */
723 723 static int
724 724 zfs_secpolicy_send_new(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
725 725 {
726 726 return (zfs_secpolicy_write_perms(zc->zc_name,
727 727 ZFS_DELEG_PERM_SEND, cr));
728 728 }
729 729
730 730 /* ARGSUSED */
731 731 static int
732 732 zfs_secpolicy_deleg_share(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
733 733 {
734 734 vnode_t *vp;
735 735 int error;
736 736
737 737 if ((error = lookupname(zc->zc_value, UIO_SYSSPACE,
738 738 NO_FOLLOW, NULL, &vp)) != 0)
739 739 return (error);
740 740
741 741 /* Now make sure mntpnt and dataset are ZFS */
742 742
743 743 if (vp->v_vfsp->vfs_fstype != zfsfstype ||
744 744 (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource),
745 745 zc->zc_name) != 0)) {
746 746 VN_RELE(vp);
747 747 return (SET_ERROR(EPERM));
748 748 }
749 749
750 750 VN_RELE(vp);
751 751 return (dsl_deleg_access(zc->zc_name,
752 752 ZFS_DELEG_PERM_SHARE, cr));
753 753 }
754 754
755 755 int
756 756 zfs_secpolicy_share(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
757 757 {
758 758 if (!INGLOBALZONE(curproc))
759 759 return (SET_ERROR(EPERM));
760 760
761 761 if (secpolicy_nfs(cr) == 0) {
762 762 return (0);
763 763 } else {
764 764 return (zfs_secpolicy_deleg_share(zc, innvl, cr));
765 765 }
766 766 }
767 767
768 768 int
769 769 zfs_secpolicy_smb_acl(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
770 770 {
771 771 if (!INGLOBALZONE(curproc))
772 772 return (SET_ERROR(EPERM));
773 773
774 774 if (secpolicy_smb(cr) == 0) {
775 775 return (0);
776 776 } else {
777 777 return (zfs_secpolicy_deleg_share(zc, innvl, cr));
778 778 }
779 779 }
780 780
781 781 static int
782 782 zfs_get_parent(const char *datasetname, char *parent, int parentsize)
783 783 {
784 784 char *cp;
785 785
786 786 /*
787 787 * Remove the @bla or /bla from the end of the name to get the parent.
788 788 */
789 789 (void) strncpy(parent, datasetname, parentsize);
790 790 cp = strrchr(parent, '@');
791 791 if (cp != NULL) {
792 792 cp[0] = '\0';
793 793 } else {
794 794 cp = strrchr(parent, '/');
795 795 if (cp == NULL)
796 796 return (SET_ERROR(ENOENT));
797 797 cp[0] = '\0';
798 798 }
799 799
800 800 return (0);
801 801 }
802 802
803 803 int
804 804 zfs_secpolicy_destroy_perms(const char *name, cred_t *cr)
805 805 {
806 806 int error;
807 807
808 808 if ((error = zfs_secpolicy_write_perms(name,
809 809 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
810 810 return (error);
811 811
812 812 return (zfs_secpolicy_write_perms(name, ZFS_DELEG_PERM_DESTROY, cr));
813 813 }
814 814
815 815 /* ARGSUSED */
816 816 static int
817 817 zfs_secpolicy_destroy(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
818 818 {
819 819 return (zfs_secpolicy_destroy_perms(zc->zc_name, cr));
820 820 }
821 821
822 822 /*
823 823 * Destroying snapshots with delegated permissions requires
824 824 * descendant mount and destroy permissions.
825 825 */
826 826 /* ARGSUSED */
827 827 static int
828 828 zfs_secpolicy_destroy_snaps(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
829 829 {
830 830 nvlist_t *snaps;
831 831 nvpair_t *pair, *nextpair;
832 832 int error = 0;
833 833
834 834 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
835 835 return (SET_ERROR(EINVAL));
836 836 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
837 837 pair = nextpair) {
838 838 nextpair = nvlist_next_nvpair(snaps, pair);
839 839 error = zfs_secpolicy_destroy_perms(nvpair_name(pair), cr);
840 840 if (error == ENOENT) {
841 841 /*
842 842 * Ignore any snapshots that don't exist (we consider
843 843 * them "already destroyed"). Remove the name from the
844 844 * nvl here in case the snapshot is created between
845 845 * now and when we try to destroy it (in which case
846 846 * we don't want to destroy it since we haven't
847 847 * checked for permission).
848 848 */
849 849 fnvlist_remove_nvpair(snaps, pair);
850 850 error = 0;
851 851 }
852 852 if (error != 0)
853 853 break;
854 854 }
855 855
856 856 return (error);
857 857 }
858 858
859 859 int
860 860 zfs_secpolicy_rename_perms(const char *from, const char *to, cred_t *cr)
861 861 {
862 862 char parentname[MAXNAMELEN];
863 863 int error;
864 864
865 865 if ((error = zfs_secpolicy_write_perms(from,
866 866 ZFS_DELEG_PERM_RENAME, cr)) != 0)
867 867 return (error);
868 868
869 869 if ((error = zfs_secpolicy_write_perms(from,
870 870 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
871 871 return (error);
872 872
873 873 if ((error = zfs_get_parent(to, parentname,
874 874 sizeof (parentname))) != 0)
875 875 return (error);
876 876
877 877 if ((error = zfs_secpolicy_write_perms(parentname,
878 878 ZFS_DELEG_PERM_CREATE, cr)) != 0)
879 879 return (error);
880 880
881 881 if ((error = zfs_secpolicy_write_perms(parentname,
882 882 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
883 883 return (error);
884 884
885 885 return (error);
886 886 }
887 887
888 888 /* ARGSUSED */
889 889 static int
890 890 zfs_secpolicy_rename(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
891 891 {
892 892 return (zfs_secpolicy_rename_perms(zc->zc_name, zc->zc_value, cr));
893 893 }
894 894
895 895 /* ARGSUSED */
896 896 static int
897 897 zfs_secpolicy_promote(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
898 898 {
899 899 dsl_pool_t *dp;
900 900 dsl_dataset_t *clone;
901 901 int error;
902 902
903 903 error = zfs_secpolicy_write_perms(zc->zc_name,
904 904 ZFS_DELEG_PERM_PROMOTE, cr);
905 905 if (error != 0)
906 906 return (error);
907 907
908 908 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
909 909 if (error != 0)
910 910 return (error);
911 911
912 912 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &clone);
913 913
914 914 if (error == 0) {
915 915 char parentname[MAXNAMELEN];
916 916 dsl_dataset_t *origin = NULL;
917 917 dsl_dir_t *dd;
918 918 dd = clone->ds_dir;
919 919
920 920 error = dsl_dataset_hold_obj(dd->dd_pool,
921 921 dsl_dir_phys(dd)->dd_origin_obj, FTAG, &origin);
922 922 if (error != 0) {
923 923 dsl_dataset_rele(clone, FTAG);
924 924 dsl_pool_rele(dp, FTAG);
925 925 return (error);
926 926 }
927 927
928 928 error = zfs_secpolicy_write_perms_ds(zc->zc_name, clone,
929 929 ZFS_DELEG_PERM_MOUNT, cr);
930 930
931 931 dsl_dataset_name(origin, parentname);
932 932 if (error == 0) {
933 933 error = zfs_secpolicy_write_perms_ds(parentname, origin,
934 934 ZFS_DELEG_PERM_PROMOTE, cr);
935 935 }
936 936 dsl_dataset_rele(clone, FTAG);
937 937 dsl_dataset_rele(origin, FTAG);
938 938 }
939 939 dsl_pool_rele(dp, FTAG);
940 940 return (error);
941 941 }
942 942
943 943 /* ARGSUSED */
944 944 static int
945 945 zfs_secpolicy_recv(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
946 946 {
947 947 int error;
948 948
949 949 if ((error = zfs_secpolicy_write_perms(zc->zc_name,
950 950 ZFS_DELEG_PERM_RECEIVE, cr)) != 0)
951 951 return (error);
952 952
953 953 if ((error = zfs_secpolicy_write_perms(zc->zc_name,
954 954 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
955 955 return (error);
956 956
957 957 return (zfs_secpolicy_write_perms(zc->zc_name,
958 958 ZFS_DELEG_PERM_CREATE, cr));
959 959 }
960 960
961 961 int
962 962 zfs_secpolicy_snapshot_perms(const char *name, cred_t *cr)
963 963 {
964 964 return (zfs_secpolicy_write_perms(name,
965 965 ZFS_DELEG_PERM_SNAPSHOT, cr));
966 966 }
967 967
968 968 /*
969 969 * Check for permission to create each snapshot in the nvlist.
970 970 */
971 971 /* ARGSUSED */
972 972 static int
973 973 zfs_secpolicy_snapshot(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
974 974 {
975 975 nvlist_t *snaps;
976 976 int error = 0;
977 977 nvpair_t *pair;
978 978
979 979 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
980 980 return (SET_ERROR(EINVAL));
981 981 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
982 982 pair = nvlist_next_nvpair(snaps, pair)) {
983 983 char *name = nvpair_name(pair);
984 984 char *atp = strchr(name, '@');
985 985
986 986 if (atp == NULL) {
987 987 error = SET_ERROR(EINVAL);
988 988 break;
989 989 }
990 990 *atp = '\0';
991 991 error = zfs_secpolicy_snapshot_perms(name, cr);
992 992 *atp = '@';
993 993 if (error != 0)
994 994 break;
995 995 }
996 996 return (error);
997 997 }
998 998
999 999 /*
1000 1000 * Check for permission to create each snapshot in the nvlist.
1001 1001 */
1002 1002 /* ARGSUSED */
1003 1003 static int
1004 1004 zfs_secpolicy_bookmark(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1005 1005 {
1006 1006 int error = 0;
1007 1007
1008 1008 for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL);
1009 1009 pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) {
1010 1010 char *name = nvpair_name(pair);
1011 1011 char *hashp = strchr(name, '#');
1012 1012
1013 1013 if (hashp == NULL) {
1014 1014 error = SET_ERROR(EINVAL);
1015 1015 break;
1016 1016 }
1017 1017 *hashp = '\0';
1018 1018 error = zfs_secpolicy_write_perms(name,
1019 1019 ZFS_DELEG_PERM_BOOKMARK, cr);
1020 1020 *hashp = '#';
1021 1021 if (error != 0)
1022 1022 break;
1023 1023 }
1024 1024 return (error);
1025 1025 }
1026 1026
1027 1027 /* ARGSUSED */
1028 1028 static int
1029 1029 zfs_secpolicy_destroy_bookmarks(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1030 1030 {
1031 1031 nvpair_t *pair, *nextpair;
1032 1032 int error = 0;
1033 1033
1034 1034 for (pair = nvlist_next_nvpair(innvl, NULL); pair != NULL;
1035 1035 pair = nextpair) {
1036 1036 char *name = nvpair_name(pair);
1037 1037 char *hashp = strchr(name, '#');
1038 1038 nextpair = nvlist_next_nvpair(innvl, pair);
1039 1039
1040 1040 if (hashp == NULL) {
1041 1041 error = SET_ERROR(EINVAL);
1042 1042 break;
1043 1043 }
1044 1044
1045 1045 *hashp = '\0';
1046 1046 error = zfs_secpolicy_write_perms(name,
1047 1047 ZFS_DELEG_PERM_DESTROY, cr);
1048 1048 *hashp = '#';
1049 1049 if (error == ENOENT) {
1050 1050 /*
1051 1051 * Ignore any filesystems that don't exist (we consider
1052 1052 * their bookmarks "already destroyed"). Remove
1053 1053 * the name from the nvl here in case the filesystem
1054 1054 * is created between now and when we try to destroy
1055 1055 * the bookmark (in which case we don't want to
1056 1056 * destroy it since we haven't checked for permission).
1057 1057 */
1058 1058 fnvlist_remove_nvpair(innvl, pair);
1059 1059 error = 0;
1060 1060 }
1061 1061 if (error != 0)
1062 1062 break;
1063 1063 }
1064 1064
1065 1065 return (error);
1066 1066 }
1067 1067
1068 1068 /* ARGSUSED */
1069 1069 static int
1070 1070 zfs_secpolicy_log_history(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1071 1071 {
1072 1072 /*
1073 1073 * Even root must have a proper TSD so that we know what pool
1074 1074 * to log to.
1075 1075 */
1076 1076 if (tsd_get(zfs_allow_log_key) == NULL)
1077 1077 return (SET_ERROR(EPERM));
1078 1078 return (0);
1079 1079 }
1080 1080
1081 1081 static int
1082 1082 zfs_secpolicy_create_clone(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1083 1083 {
1084 1084 char parentname[MAXNAMELEN];
1085 1085 int error;
1086 1086 char *origin;
1087 1087
1088 1088 if ((error = zfs_get_parent(zc->zc_name, parentname,
1089 1089 sizeof (parentname))) != 0)
1090 1090 return (error);
1091 1091
1092 1092 if (nvlist_lookup_string(innvl, "origin", &origin) == 0 &&
1093 1093 (error = zfs_secpolicy_write_perms(origin,
1094 1094 ZFS_DELEG_PERM_CLONE, cr)) != 0)
1095 1095 return (error);
1096 1096
1097 1097 if ((error = zfs_secpolicy_write_perms(parentname,
1098 1098 ZFS_DELEG_PERM_CREATE, cr)) != 0)
1099 1099 return (error);
1100 1100
1101 1101 return (zfs_secpolicy_write_perms(parentname,
1102 1102 ZFS_DELEG_PERM_MOUNT, cr));
1103 1103 }
1104 1104
1105 1105 /*
1106 1106 * Policy for pool operations - create/destroy pools, add vdevs, etc. Requires
1107 1107 * SYS_CONFIG privilege, which is not available in a local zone.
1108 1108 */
1109 1109 /* ARGSUSED */
1110 1110 static int
1111 1111 zfs_secpolicy_config(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1112 1112 {
1113 1113 if (secpolicy_sys_config(cr, B_FALSE) != 0)
1114 1114 return (SET_ERROR(EPERM));
1115 1115
1116 1116 return (0);
1117 1117 }
1118 1118
1119 1119 /*
1120 1120 * Policy for object to name lookups.
1121 1121 */
1122 1122 /* ARGSUSED */
1123 1123 static int
1124 1124 zfs_secpolicy_diff(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1125 1125 {
1126 1126 int error;
1127 1127
1128 1128 if ((error = secpolicy_sys_config(cr, B_FALSE)) == 0)
1129 1129 return (0);
1130 1130
1131 1131 error = zfs_secpolicy_write_perms(zc->zc_name, ZFS_DELEG_PERM_DIFF, cr);
1132 1132 return (error);
1133 1133 }
1134 1134
1135 1135 /*
1136 1136 * Policy for fault injection. Requires all privileges.
1137 1137 */
1138 1138 /* ARGSUSED */
1139 1139 static int
1140 1140 zfs_secpolicy_inject(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1141 1141 {
1142 1142 return (secpolicy_zinject(cr));
1143 1143 }
1144 1144
1145 1145 /* ARGSUSED */
1146 1146 static int
1147 1147 zfs_secpolicy_inherit_prop(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1148 1148 {
1149 1149 zfs_prop_t prop = zfs_name_to_prop(zc->zc_value);
1150 1150
1151 1151 if (prop == ZPROP_INVAL) {
1152 1152 if (!zfs_prop_user(zc->zc_value))
1153 1153 return (SET_ERROR(EINVAL));
1154 1154 return (zfs_secpolicy_write_perms(zc->zc_name,
1155 1155 ZFS_DELEG_PERM_USERPROP, cr));
1156 1156 } else {
1157 1157 return (zfs_secpolicy_setprop(zc->zc_name, prop,
1158 1158 NULL, cr));
1159 1159 }
1160 1160 }
1161 1161
1162 1162 static int
1163 1163 zfs_secpolicy_userspace_one(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1164 1164 {
1165 1165 int err = zfs_secpolicy_read(zc, innvl, cr);
1166 1166 if (err)
1167 1167 return (err);
1168 1168
1169 1169 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
1170 1170 return (SET_ERROR(EINVAL));
1171 1171
1172 1172 if (zc->zc_value[0] == 0) {
1173 1173 /*
1174 1174 * They are asking about a posix uid/gid. If it's
1175 1175 * themself, allow it.
1176 1176 */
1177 1177 if (zc->zc_objset_type == ZFS_PROP_USERUSED ||
1178 1178 zc->zc_objset_type == ZFS_PROP_USERQUOTA) {
1179 1179 if (zc->zc_guid == crgetuid(cr))
1180 1180 return (0);
1181 1181 } else {
1182 1182 if (groupmember(zc->zc_guid, cr))
1183 1183 return (0);
1184 1184 }
1185 1185 }
1186 1186
1187 1187 return (zfs_secpolicy_write_perms(zc->zc_name,
1188 1188 userquota_perms[zc->zc_objset_type], cr));
1189 1189 }
1190 1190
1191 1191 static int
1192 1192 zfs_secpolicy_userspace_many(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1193 1193 {
1194 1194 int err = zfs_secpolicy_read(zc, innvl, cr);
1195 1195 if (err)
1196 1196 return (err);
1197 1197
1198 1198 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
1199 1199 return (SET_ERROR(EINVAL));
1200 1200
1201 1201 return (zfs_secpolicy_write_perms(zc->zc_name,
1202 1202 userquota_perms[zc->zc_objset_type], cr));
1203 1203 }
1204 1204
1205 1205 /* ARGSUSED */
1206 1206 static int
1207 1207 zfs_secpolicy_userspace_upgrade(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1208 1208 {
1209 1209 return (zfs_secpolicy_setprop(zc->zc_name, ZFS_PROP_VERSION,
1210 1210 NULL, cr));
1211 1211 }
1212 1212
1213 1213 /* ARGSUSED */
1214 1214 static int
1215 1215 zfs_secpolicy_hold(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1216 1216 {
1217 1217 nvpair_t *pair;
1218 1218 nvlist_t *holds;
1219 1219 int error;
1220 1220
1221 1221 error = nvlist_lookup_nvlist(innvl, "holds", &holds);
1222 1222 if (error != 0)
1223 1223 return (SET_ERROR(EINVAL));
1224 1224
1225 1225 for (pair = nvlist_next_nvpair(holds, NULL); pair != NULL;
1226 1226 pair = nvlist_next_nvpair(holds, pair)) {
1227 1227 char fsname[MAXNAMELEN];
1228 1228 error = dmu_fsname(nvpair_name(pair), fsname);
1229 1229 if (error != 0)
1230 1230 return (error);
1231 1231 error = zfs_secpolicy_write_perms(fsname,
1232 1232 ZFS_DELEG_PERM_HOLD, cr);
1233 1233 if (error != 0)
1234 1234 return (error);
1235 1235 }
1236 1236 return (0);
1237 1237 }
1238 1238
1239 1239 /* ARGSUSED */
1240 1240 static int
1241 1241 zfs_secpolicy_release(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1242 1242 {
1243 1243 nvpair_t *pair;
1244 1244 int error;
1245 1245
1246 1246 for (pair = nvlist_next_nvpair(innvl, NULL); pair != NULL;
1247 1247 pair = nvlist_next_nvpair(innvl, pair)) {
1248 1248 char fsname[MAXNAMELEN];
1249 1249 error = dmu_fsname(nvpair_name(pair), fsname);
1250 1250 if (error != 0)
1251 1251 return (error);
1252 1252 error = zfs_secpolicy_write_perms(fsname,
1253 1253 ZFS_DELEG_PERM_RELEASE, cr);
1254 1254 if (error != 0)
1255 1255 return (error);
1256 1256 }
1257 1257 return (0);
1258 1258 }
1259 1259
1260 1260 /*
1261 1261 * Policy for allowing temporary snapshots to be taken or released
1262 1262 */
1263 1263 static int
1264 1264 zfs_secpolicy_tmp_snapshot(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1265 1265 {
1266 1266 /*
1267 1267 * A temporary snapshot is the same as a snapshot,
1268 1268 * hold, destroy and release all rolled into one.
1269 1269 * Delegated diff alone is sufficient that we allow this.
1270 1270 */
1271 1271 int error;
1272 1272
1273 1273 if ((error = zfs_secpolicy_write_perms(zc->zc_name,
1274 1274 ZFS_DELEG_PERM_DIFF, cr)) == 0)
1275 1275 return (0);
1276 1276
1277 1277 error = zfs_secpolicy_snapshot_perms(zc->zc_name, cr);
1278 1278 if (error == 0)
1279 1279 error = zfs_secpolicy_hold(zc, innvl, cr);
1280 1280 if (error == 0)
1281 1281 error = zfs_secpolicy_release(zc, innvl, cr);
1282 1282 if (error == 0)
1283 1283 error = zfs_secpolicy_destroy(zc, innvl, cr);
1284 1284 return (error);
1285 1285 }
1286 1286
1287 1287 /*
1288 1288 * Returns the nvlist as specified by the user in the zfs_cmd_t.
1289 1289 */
1290 1290 static int
1291 1291 get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp)
1292 1292 {
1293 1293 char *packed;
1294 1294 int error;
1295 1295 nvlist_t *list = NULL;
1296 1296
1297 1297 /*
1298 1298 * Read in and unpack the user-supplied nvlist.
1299 1299 */
1300 1300 if (size == 0)
1301 1301 return (SET_ERROR(EINVAL));
1302 1302
1303 1303 packed = kmem_alloc(size, KM_SLEEP);
1304 1304
1305 1305 if ((error = ddi_copyin((void *)(uintptr_t)nvl, packed, size,
1306 1306 iflag)) != 0) {
1307 1307 kmem_free(packed, size);
1308 1308 return (SET_ERROR(EFAULT));
1309 1309 }
1310 1310
1311 1311 if ((error = nvlist_unpack(packed, size, &list, 0)) != 0) {
1312 1312 kmem_free(packed, size);
1313 1313 return (error);
1314 1314 }
1315 1315
1316 1316 kmem_free(packed, size);
1317 1317
1318 1318 *nvp = list;
1319 1319 return (0);
1320 1320 }
1321 1321
1322 1322 /*
1323 1323 * Reduce the size of this nvlist until it can be serialized in 'max' bytes.
1324 1324 * Entries will be removed from the end of the nvlist, and one int32 entry
1325 1325 * named "N_MORE_ERRORS" will be added indicating how many entries were
1326 1326 * removed.
1327 1327 */
1328 1328 static int
1329 1329 nvlist_smush(nvlist_t *errors, size_t max)
1330 1330 {
1331 1331 size_t size;
1332 1332
1333 1333 size = fnvlist_size(errors);
1334 1334
1335 1335 if (size > max) {
1336 1336 nvpair_t *more_errors;
1337 1337 int n = 0;
1338 1338
1339 1339 if (max < 1024)
1340 1340 return (SET_ERROR(ENOMEM));
1341 1341
1342 1342 fnvlist_add_int32(errors, ZPROP_N_MORE_ERRORS, 0);
1343 1343 more_errors = nvlist_prev_nvpair(errors, NULL);
1344 1344
1345 1345 do {
1346 1346 nvpair_t *pair = nvlist_prev_nvpair(errors,
1347 1347 more_errors);
1348 1348 fnvlist_remove_nvpair(errors, pair);
1349 1349 n++;
1350 1350 size = fnvlist_size(errors);
1351 1351 } while (size > max);
1352 1352
1353 1353 fnvlist_remove_nvpair(errors, more_errors);
1354 1354 fnvlist_add_int32(errors, ZPROP_N_MORE_ERRORS, n);
1355 1355 ASSERT3U(fnvlist_size(errors), <=, max);
1356 1356 }
1357 1357
1358 1358 return (0);
1359 1359 }
1360 1360
1361 1361 static int
1362 1362 put_nvlist(zfs_cmd_t *zc, nvlist_t *nvl)
1363 1363 {
1364 1364 char *packed = NULL;
1365 1365 int error = 0;
1366 1366 size_t size;
1367 1367
1368 1368 size = fnvlist_size(nvl);
1369 1369
1370 1370 if (size > zc->zc_nvlist_dst_size) {
1371 1371 error = SET_ERROR(ENOMEM);
1372 1372 } else {
1373 1373 packed = fnvlist_pack(nvl, &size);
1374 1374 if (ddi_copyout(packed, (void *)(uintptr_t)zc->zc_nvlist_dst,
1375 1375 size, zc->zc_iflags) != 0)
1376 1376 error = SET_ERROR(EFAULT);
1377 1377 fnvlist_pack_free(packed, size);
1378 1378 }
1379 1379
1380 1380 zc->zc_nvlist_dst_size = size;
1381 1381 zc->zc_nvlist_dst_filled = B_TRUE;
1382 1382 return (error);
1383 1383 }
1384 1384
1385 1385 static int
1386 1386 getzfsvfs(const char *dsname, zfsvfs_t **zfvp)
1387 1387 {
1388 1388 objset_t *os;
1389 1389 int error;
1390 1390
1391 1391 error = dmu_objset_hold(dsname, FTAG, &os);
1392 1392 if (error != 0)
1393 1393 return (error);
1394 1394 if (dmu_objset_type(os) != DMU_OST_ZFS) {
1395 1395 dmu_objset_rele(os, FTAG);
1396 1396 return (SET_ERROR(EINVAL));
1397 1397 }
1398 1398
1399 1399 mutex_enter(&os->os_user_ptr_lock);
1400 1400 *zfvp = dmu_objset_get_user(os);
1401 1401 if (*zfvp) {
1402 1402 VFS_HOLD((*zfvp)->z_vfs);
1403 1403 } else {
1404 1404 error = SET_ERROR(ESRCH);
1405 1405 }
1406 1406 mutex_exit(&os->os_user_ptr_lock);
1407 1407 dmu_objset_rele(os, FTAG);
1408 1408 return (error);
1409 1409 }
1410 1410
1411 1411 /*
1412 1412 * Find a zfsvfs_t for a mounted filesystem, or create our own, in which
1413 1413 * case its z_vfs will be NULL, and it will be opened as the owner.
1414 1414 * If 'writer' is set, the z_teardown_lock will be held for RW_WRITER,
1415 1415 * which prevents all vnode ops from running.
1416 1416 */
1417 1417 static int
1418 1418 zfsvfs_hold(const char *name, void *tag, zfsvfs_t **zfvp, boolean_t writer)
1419 1419 {
1420 1420 int error = 0;
1421 1421
1422 1422 if (getzfsvfs(name, zfvp) != 0)
1423 1423 error = zfsvfs_create(name, zfvp);
1424 1424 if (error == 0) {
1425 1425 rrm_enter(&(*zfvp)->z_teardown_lock, (writer) ? RW_WRITER :
1426 1426 RW_READER, tag);
1427 1427 if ((*zfvp)->z_unmounted) {
1428 1428 /*
1429 1429 * XXX we could probably try again, since the unmounting
1430 1430 * thread should be just about to disassociate the
1431 1431 * objset from the zfsvfs.
1432 1432 */
1433 1433 rrm_exit(&(*zfvp)->z_teardown_lock, tag);
1434 1434 return (SET_ERROR(EBUSY));
1435 1435 }
1436 1436 }
1437 1437 return (error);
1438 1438 }
1439 1439
1440 1440 static void
1441 1441 zfsvfs_rele(zfsvfs_t *zfsvfs, void *tag)
1442 1442 {
1443 1443 rrm_exit(&zfsvfs->z_teardown_lock, tag);
1444 1444
1445 1445 if (zfsvfs->z_vfs) {
1446 1446 VFS_RELE(zfsvfs->z_vfs);
1447 1447 } else {
1448 1448 dmu_objset_disown(zfsvfs->z_os, zfsvfs);
1449 1449 zfsvfs_free(zfsvfs);
1450 1450 }
1451 1451 }
1452 1452
1453 1453 static int
1454 1454 zfs_ioc_pool_create(zfs_cmd_t *zc)
1455 1455 {
1456 1456 int error;
1457 1457 nvlist_t *config, *props = NULL;
1458 1458 nvlist_t *rootprops = NULL;
1459 1459 nvlist_t *zplprops = NULL;
1460 1460
1461 1461 if (error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1462 1462 zc->zc_iflags, &config))
1463 1463 return (error);
1464 1464
1465 1465 if (zc->zc_nvlist_src_size != 0 && (error =
1466 1466 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1467 1467 zc->zc_iflags, &props))) {
1468 1468 nvlist_free(config);
1469 1469 return (error);
1470 1470 }
1471 1471
1472 1472 if (props) {
1473 1473 nvlist_t *nvl = NULL;
1474 1474 uint64_t version = SPA_VERSION;
1475 1475
1476 1476 (void) nvlist_lookup_uint64(props,
1477 1477 zpool_prop_to_name(ZPOOL_PROP_VERSION), &version);
1478 1478 if (!SPA_VERSION_IS_SUPPORTED(version)) {
1479 1479 error = SET_ERROR(EINVAL);
1480 1480 goto pool_props_bad;
1481 1481 }
1482 1482 (void) nvlist_lookup_nvlist(props, ZPOOL_ROOTFS_PROPS, &nvl);
1483 1483 if (nvl) {
1484 1484 error = nvlist_dup(nvl, &rootprops, KM_SLEEP);
1485 1485 if (error != 0) {
1486 1486 nvlist_free(config);
1487 1487 nvlist_free(props);
1488 1488 return (error);
1489 1489 }
1490 1490 (void) nvlist_remove_all(props, ZPOOL_ROOTFS_PROPS);
1491 1491 }
1492 1492 VERIFY(nvlist_alloc(&zplprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
1493 1493 error = zfs_fill_zplprops_root(version, rootprops,
1494 1494 zplprops, NULL);
1495 1495 if (error != 0)
1496 1496 goto pool_props_bad;
1497 1497 }
1498 1498
1499 1499 error = spa_create(zc->zc_name, config, props, zplprops);
1500 1500
1501 1501 /*
1502 1502 * Set the remaining root properties
1503 1503 */
1504 1504 if (!error && (error = zfs_set_prop_nvlist(zc->zc_name,
1505 1505 ZPROP_SRC_LOCAL, rootprops, NULL)) != 0)
1506 1506 (void) spa_destroy(zc->zc_name);
1507 1507
1508 1508 pool_props_bad:
1509 1509 nvlist_free(rootprops);
1510 1510 nvlist_free(zplprops);
1511 1511 nvlist_free(config);
1512 1512 nvlist_free(props);
1513 1513
1514 1514 return (error);
1515 1515 }
1516 1516
1517 1517 static int
1518 1518 zfs_ioc_pool_destroy(zfs_cmd_t *zc)
1519 1519 {
1520 1520 int error;
1521 1521 zfs_log_history(zc);
1522 1522 error = spa_destroy(zc->zc_name);
1523 1523 if (error == 0)
1524 1524 zvol_remove_minors(zc->zc_name);
1525 1525 return (error);
1526 1526 }
1527 1527
1528 1528 static int
1529 1529 zfs_ioc_pool_import(zfs_cmd_t *zc)
1530 1530 {
1531 1531 nvlist_t *config, *props = NULL;
1532 1532 uint64_t guid;
1533 1533 int error;
1534 1534
1535 1535 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1536 1536 zc->zc_iflags, &config)) != 0)
1537 1537 return (error);
1538 1538
1539 1539 if (zc->zc_nvlist_src_size != 0 && (error =
1540 1540 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1541 1541 zc->zc_iflags, &props))) {
1542 1542 nvlist_free(config);
1543 1543 return (error);
1544 1544 }
1545 1545
1546 1546 if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, &guid) != 0 ||
1547 1547 guid != zc->zc_guid)
1548 1548 error = SET_ERROR(EINVAL);
1549 1549 else
1550 1550 error = spa_import(zc->zc_name, config, props, zc->zc_cookie);
↓ open down ↓ |
1550 lines elided |
↑ open up ↑ |
1551 1551
1552 1552 if (zc->zc_nvlist_dst != 0) {
1553 1553 int err;
1554 1554
1555 1555 if ((err = put_nvlist(zc, config)) != 0)
1556 1556 error = err;
1557 1557 }
1558 1558
1559 1559 nvlist_free(config);
1560 1560
1561 - if (props)
1562 - nvlist_free(props);
1561 + nvlist_free(props);
1563 1562
1564 1563 return (error);
1565 1564 }
1566 1565
1567 1566 static int
1568 1567 zfs_ioc_pool_export(zfs_cmd_t *zc)
1569 1568 {
1570 1569 int error;
1571 1570 boolean_t force = (boolean_t)zc->zc_cookie;
1572 1571 boolean_t hardforce = (boolean_t)zc->zc_guid;
1573 1572
1574 1573 zfs_log_history(zc);
1575 1574 error = spa_export(zc->zc_name, NULL, force, hardforce);
1576 1575 if (error == 0)
1577 1576 zvol_remove_minors(zc->zc_name);
1578 1577 return (error);
1579 1578 }
1580 1579
1581 1580 static int
1582 1581 zfs_ioc_pool_configs(zfs_cmd_t *zc)
1583 1582 {
1584 1583 nvlist_t *configs;
1585 1584 int error;
1586 1585
1587 1586 if ((configs = spa_all_configs(&zc->zc_cookie)) == NULL)
1588 1587 return (SET_ERROR(EEXIST));
1589 1588
1590 1589 error = put_nvlist(zc, configs);
1591 1590
1592 1591 nvlist_free(configs);
1593 1592
1594 1593 return (error);
1595 1594 }
1596 1595
1597 1596 /*
1598 1597 * inputs:
1599 1598 * zc_name name of the pool
1600 1599 *
1601 1600 * outputs:
1602 1601 * zc_cookie real errno
1603 1602 * zc_nvlist_dst config nvlist
1604 1603 * zc_nvlist_dst_size size of config nvlist
1605 1604 */
1606 1605 static int
1607 1606 zfs_ioc_pool_stats(zfs_cmd_t *zc)
1608 1607 {
1609 1608 nvlist_t *config;
1610 1609 int error;
1611 1610 int ret = 0;
1612 1611
1613 1612 error = spa_get_stats(zc->zc_name, &config, zc->zc_value,
1614 1613 sizeof (zc->zc_value));
1615 1614
1616 1615 if (config != NULL) {
1617 1616 ret = put_nvlist(zc, config);
1618 1617 nvlist_free(config);
1619 1618
1620 1619 /*
1621 1620 * The config may be present even if 'error' is non-zero.
1622 1621 * In this case we return success, and preserve the real errno
1623 1622 * in 'zc_cookie'.
1624 1623 */
1625 1624 zc->zc_cookie = error;
1626 1625 } else {
1627 1626 ret = error;
1628 1627 }
1629 1628
1630 1629 return (ret);
1631 1630 }
1632 1631
1633 1632 /*
1634 1633 * Try to import the given pool, returning pool stats as appropriate so that
1635 1634 * user land knows which devices are available and overall pool health.
1636 1635 */
1637 1636 static int
1638 1637 zfs_ioc_pool_tryimport(zfs_cmd_t *zc)
1639 1638 {
1640 1639 nvlist_t *tryconfig, *config;
1641 1640 int error;
1642 1641
1643 1642 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1644 1643 zc->zc_iflags, &tryconfig)) != 0)
1645 1644 return (error);
1646 1645
1647 1646 config = spa_tryimport(tryconfig);
1648 1647
1649 1648 nvlist_free(tryconfig);
1650 1649
1651 1650 if (config == NULL)
1652 1651 return (SET_ERROR(EINVAL));
1653 1652
1654 1653 error = put_nvlist(zc, config);
1655 1654 nvlist_free(config);
1656 1655
1657 1656 return (error);
1658 1657 }
1659 1658
1660 1659 /*
1661 1660 * inputs:
1662 1661 * zc_name name of the pool
1663 1662 * zc_cookie scan func (pool_scan_func_t)
1664 1663 */
1665 1664 static int
1666 1665 zfs_ioc_pool_scan(zfs_cmd_t *zc)
1667 1666 {
1668 1667 spa_t *spa;
1669 1668 int error;
1670 1669
1671 1670 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1672 1671 return (error);
1673 1672
1674 1673 if (zc->zc_cookie == POOL_SCAN_NONE)
1675 1674 error = spa_scan_stop(spa);
1676 1675 else
1677 1676 error = spa_scan(spa, zc->zc_cookie);
1678 1677
1679 1678 spa_close(spa, FTAG);
1680 1679
1681 1680 return (error);
1682 1681 }
1683 1682
1684 1683 static int
1685 1684 zfs_ioc_pool_freeze(zfs_cmd_t *zc)
1686 1685 {
1687 1686 spa_t *spa;
1688 1687 int error;
1689 1688
1690 1689 error = spa_open(zc->zc_name, &spa, FTAG);
1691 1690 if (error == 0) {
1692 1691 spa_freeze(spa);
1693 1692 spa_close(spa, FTAG);
1694 1693 }
1695 1694 return (error);
1696 1695 }
1697 1696
1698 1697 static int
1699 1698 zfs_ioc_pool_upgrade(zfs_cmd_t *zc)
1700 1699 {
1701 1700 spa_t *spa;
1702 1701 int error;
1703 1702
1704 1703 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1705 1704 return (error);
1706 1705
1707 1706 if (zc->zc_cookie < spa_version(spa) ||
1708 1707 !SPA_VERSION_IS_SUPPORTED(zc->zc_cookie)) {
1709 1708 spa_close(spa, FTAG);
1710 1709 return (SET_ERROR(EINVAL));
1711 1710 }
1712 1711
1713 1712 spa_upgrade(spa, zc->zc_cookie);
1714 1713 spa_close(spa, FTAG);
1715 1714
1716 1715 return (error);
1717 1716 }
1718 1717
1719 1718 static int
1720 1719 zfs_ioc_pool_get_history(zfs_cmd_t *zc)
1721 1720 {
1722 1721 spa_t *spa;
1723 1722 char *hist_buf;
1724 1723 uint64_t size;
1725 1724 int error;
1726 1725
1727 1726 if ((size = zc->zc_history_len) == 0)
1728 1727 return (SET_ERROR(EINVAL));
1729 1728
1730 1729 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1731 1730 return (error);
1732 1731
1733 1732 if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) {
1734 1733 spa_close(spa, FTAG);
1735 1734 return (SET_ERROR(ENOTSUP));
1736 1735 }
1737 1736
1738 1737 hist_buf = kmem_alloc(size, KM_SLEEP);
1739 1738 if ((error = spa_history_get(spa, &zc->zc_history_offset,
1740 1739 &zc->zc_history_len, hist_buf)) == 0) {
1741 1740 error = ddi_copyout(hist_buf,
1742 1741 (void *)(uintptr_t)zc->zc_history,
1743 1742 zc->zc_history_len, zc->zc_iflags);
1744 1743 }
1745 1744
1746 1745 spa_close(spa, FTAG);
1747 1746 kmem_free(hist_buf, size);
1748 1747 return (error);
1749 1748 }
1750 1749
1751 1750 static int
1752 1751 zfs_ioc_pool_reguid(zfs_cmd_t *zc)
1753 1752 {
1754 1753 spa_t *spa;
1755 1754 int error;
1756 1755
1757 1756 error = spa_open(zc->zc_name, &spa, FTAG);
1758 1757 if (error == 0) {
1759 1758 error = spa_change_guid(spa);
1760 1759 spa_close(spa, FTAG);
1761 1760 }
1762 1761 return (error);
1763 1762 }
1764 1763
1765 1764 static int
1766 1765 zfs_ioc_dsobj_to_dsname(zfs_cmd_t *zc)
1767 1766 {
1768 1767 return (dsl_dsobj_to_dsname(zc->zc_name, zc->zc_obj, zc->zc_value));
1769 1768 }
1770 1769
1771 1770 /*
1772 1771 * inputs:
1773 1772 * zc_name name of filesystem
1774 1773 * zc_obj object to find
1775 1774 *
1776 1775 * outputs:
1777 1776 * zc_value name of object
1778 1777 */
1779 1778 static int
1780 1779 zfs_ioc_obj_to_path(zfs_cmd_t *zc)
1781 1780 {
1782 1781 objset_t *os;
1783 1782 int error;
1784 1783
1785 1784 /* XXX reading from objset not owned */
1786 1785 if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os)) != 0)
1787 1786 return (error);
1788 1787 if (dmu_objset_type(os) != DMU_OST_ZFS) {
1789 1788 dmu_objset_rele(os, FTAG);
1790 1789 return (SET_ERROR(EINVAL));
1791 1790 }
1792 1791 error = zfs_obj_to_path(os, zc->zc_obj, zc->zc_value,
1793 1792 sizeof (zc->zc_value));
1794 1793 dmu_objset_rele(os, FTAG);
1795 1794
1796 1795 return (error);
1797 1796 }
1798 1797
1799 1798 /*
1800 1799 * inputs:
1801 1800 * zc_name name of filesystem
1802 1801 * zc_obj object to find
1803 1802 *
1804 1803 * outputs:
1805 1804 * zc_stat stats on object
1806 1805 * zc_value path to object
1807 1806 */
1808 1807 static int
1809 1808 zfs_ioc_obj_to_stats(zfs_cmd_t *zc)
1810 1809 {
1811 1810 objset_t *os;
1812 1811 int error;
1813 1812
1814 1813 /* XXX reading from objset not owned */
1815 1814 if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os)) != 0)
1816 1815 return (error);
1817 1816 if (dmu_objset_type(os) != DMU_OST_ZFS) {
1818 1817 dmu_objset_rele(os, FTAG);
1819 1818 return (SET_ERROR(EINVAL));
1820 1819 }
1821 1820 error = zfs_obj_to_stats(os, zc->zc_obj, &zc->zc_stat, zc->zc_value,
1822 1821 sizeof (zc->zc_value));
1823 1822 dmu_objset_rele(os, FTAG);
1824 1823
1825 1824 return (error);
1826 1825 }
1827 1826
1828 1827 static int
1829 1828 zfs_ioc_vdev_add(zfs_cmd_t *zc)
1830 1829 {
1831 1830 spa_t *spa;
1832 1831 int error;
1833 1832 nvlist_t *config, **l2cache, **spares;
1834 1833 uint_t nl2cache = 0, nspares = 0;
1835 1834
1836 1835 error = spa_open(zc->zc_name, &spa, FTAG);
1837 1836 if (error != 0)
1838 1837 return (error);
1839 1838
1840 1839 error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1841 1840 zc->zc_iflags, &config);
1842 1841 (void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_L2CACHE,
1843 1842 &l2cache, &nl2cache);
1844 1843
1845 1844 (void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_SPARES,
1846 1845 &spares, &nspares);
1847 1846
1848 1847 /*
1849 1848 * A root pool with concatenated devices is not supported.
1850 1849 * Thus, can not add a device to a root pool.
1851 1850 *
1852 1851 * Intent log device can not be added to a rootpool because
1853 1852 * during mountroot, zil is replayed, a seperated log device
1854 1853 * can not be accessed during the mountroot time.
1855 1854 *
1856 1855 * l2cache and spare devices are ok to be added to a rootpool.
1857 1856 */
1858 1857 if (spa_bootfs(spa) != 0 && nl2cache == 0 && nspares == 0) {
1859 1858 nvlist_free(config);
1860 1859 spa_close(spa, FTAG);
1861 1860 return (SET_ERROR(EDOM));
1862 1861 }
1863 1862
1864 1863 if (error == 0) {
1865 1864 error = spa_vdev_add(spa, config);
1866 1865 nvlist_free(config);
1867 1866 }
1868 1867 spa_close(spa, FTAG);
1869 1868 return (error);
1870 1869 }
1871 1870
1872 1871 /*
1873 1872 * inputs:
1874 1873 * zc_name name of the pool
1875 1874 * zc_nvlist_conf nvlist of devices to remove
1876 1875 * zc_cookie to stop the remove?
1877 1876 */
1878 1877 static int
1879 1878 zfs_ioc_vdev_remove(zfs_cmd_t *zc)
1880 1879 {
1881 1880 spa_t *spa;
1882 1881 int error;
1883 1882
1884 1883 error = spa_open(zc->zc_name, &spa, FTAG);
1885 1884 if (error != 0)
1886 1885 return (error);
1887 1886 error = spa_vdev_remove(spa, zc->zc_guid, B_FALSE);
1888 1887 spa_close(spa, FTAG);
1889 1888 return (error);
1890 1889 }
1891 1890
1892 1891 static int
1893 1892 zfs_ioc_vdev_set_state(zfs_cmd_t *zc)
1894 1893 {
1895 1894 spa_t *spa;
1896 1895 int error;
1897 1896 vdev_state_t newstate = VDEV_STATE_UNKNOWN;
1898 1897
1899 1898 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1900 1899 return (error);
1901 1900 switch (zc->zc_cookie) {
1902 1901 case VDEV_STATE_ONLINE:
1903 1902 error = vdev_online(spa, zc->zc_guid, zc->zc_obj, &newstate);
1904 1903 break;
1905 1904
1906 1905 case VDEV_STATE_OFFLINE:
1907 1906 error = vdev_offline(spa, zc->zc_guid, zc->zc_obj);
1908 1907 break;
1909 1908
1910 1909 case VDEV_STATE_FAULTED:
1911 1910 if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED &&
1912 1911 zc->zc_obj != VDEV_AUX_EXTERNAL)
1913 1912 zc->zc_obj = VDEV_AUX_ERR_EXCEEDED;
1914 1913
1915 1914 error = vdev_fault(spa, zc->zc_guid, zc->zc_obj);
1916 1915 break;
1917 1916
1918 1917 case VDEV_STATE_DEGRADED:
1919 1918 if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED &&
1920 1919 zc->zc_obj != VDEV_AUX_EXTERNAL)
1921 1920 zc->zc_obj = VDEV_AUX_ERR_EXCEEDED;
1922 1921
1923 1922 error = vdev_degrade(spa, zc->zc_guid, zc->zc_obj);
1924 1923 break;
1925 1924
1926 1925 default:
1927 1926 error = SET_ERROR(EINVAL);
1928 1927 }
1929 1928 zc->zc_cookie = newstate;
1930 1929 spa_close(spa, FTAG);
1931 1930 return (error);
1932 1931 }
1933 1932
1934 1933 static int
1935 1934 zfs_ioc_vdev_attach(zfs_cmd_t *zc)
1936 1935 {
1937 1936 spa_t *spa;
1938 1937 int replacing = zc->zc_cookie;
1939 1938 nvlist_t *config;
1940 1939 int error;
1941 1940
1942 1941 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1943 1942 return (error);
1944 1943
1945 1944 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1946 1945 zc->zc_iflags, &config)) == 0) {
1947 1946 error = spa_vdev_attach(spa, zc->zc_guid, config, replacing);
1948 1947 nvlist_free(config);
1949 1948 }
1950 1949
1951 1950 spa_close(spa, FTAG);
1952 1951 return (error);
1953 1952 }
1954 1953
1955 1954 static int
1956 1955 zfs_ioc_vdev_detach(zfs_cmd_t *zc)
1957 1956 {
1958 1957 spa_t *spa;
1959 1958 int error;
1960 1959
1961 1960 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1962 1961 return (error);
1963 1962
1964 1963 error = spa_vdev_detach(spa, zc->zc_guid, 0, B_FALSE);
1965 1964
1966 1965 spa_close(spa, FTAG);
1967 1966 return (error);
1968 1967 }
1969 1968
1970 1969 static int
1971 1970 zfs_ioc_vdev_split(zfs_cmd_t *zc)
1972 1971 {
1973 1972 spa_t *spa;
1974 1973 nvlist_t *config, *props = NULL;
1975 1974 int error;
1976 1975 boolean_t exp = !!(zc->zc_cookie & ZPOOL_EXPORT_AFTER_SPLIT);
1977 1976
1978 1977 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1979 1978 return (error);
1980 1979
1981 1980 if (error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1982 1981 zc->zc_iflags, &config)) {
1983 1982 spa_close(spa, FTAG);
1984 1983 return (error);
1985 1984 }
1986 1985
1987 1986 if (zc->zc_nvlist_src_size != 0 && (error =
1988 1987 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1989 1988 zc->zc_iflags, &props))) {
1990 1989 spa_close(spa, FTAG);
1991 1990 nvlist_free(config);
1992 1991 return (error);
1993 1992 }
1994 1993
1995 1994 error = spa_vdev_split_mirror(spa, zc->zc_string, config, props, exp);
1996 1995
1997 1996 spa_close(spa, FTAG);
1998 1997
1999 1998 nvlist_free(config);
2000 1999 nvlist_free(props);
2001 2000
2002 2001 return (error);
2003 2002 }
2004 2003
2005 2004 static int
2006 2005 zfs_ioc_vdev_setpath(zfs_cmd_t *zc)
2007 2006 {
2008 2007 spa_t *spa;
2009 2008 char *path = zc->zc_value;
2010 2009 uint64_t guid = zc->zc_guid;
2011 2010 int error;
2012 2011
2013 2012 error = spa_open(zc->zc_name, &spa, FTAG);
2014 2013 if (error != 0)
2015 2014 return (error);
2016 2015
2017 2016 error = spa_vdev_setpath(spa, guid, path);
2018 2017 spa_close(spa, FTAG);
2019 2018 return (error);
2020 2019 }
2021 2020
2022 2021 static int
2023 2022 zfs_ioc_vdev_setfru(zfs_cmd_t *zc)
2024 2023 {
2025 2024 spa_t *spa;
2026 2025 char *fru = zc->zc_value;
2027 2026 uint64_t guid = zc->zc_guid;
2028 2027 int error;
2029 2028
2030 2029 error = spa_open(zc->zc_name, &spa, FTAG);
2031 2030 if (error != 0)
2032 2031 return (error);
2033 2032
2034 2033 error = spa_vdev_setfru(spa, guid, fru);
2035 2034 spa_close(spa, FTAG);
2036 2035 return (error);
2037 2036 }
2038 2037
2039 2038 static int
2040 2039 zfs_ioc_objset_stats_impl(zfs_cmd_t *zc, objset_t *os)
2041 2040 {
2042 2041 int error = 0;
2043 2042 nvlist_t *nv;
2044 2043
2045 2044 dmu_objset_fast_stat(os, &zc->zc_objset_stats);
2046 2045
2047 2046 if (zc->zc_nvlist_dst != 0 &&
2048 2047 (error = dsl_prop_get_all(os, &nv)) == 0) {
2049 2048 dmu_objset_stats(os, nv);
2050 2049 /*
2051 2050 * NB: zvol_get_stats() will read the objset contents,
2052 2051 * which we aren't supposed to do with a
2053 2052 * DS_MODE_USER hold, because it could be
2054 2053 * inconsistent. So this is a bit of a workaround...
2055 2054 * XXX reading with out owning
2056 2055 */
2057 2056 if (!zc->zc_objset_stats.dds_inconsistent &&
2058 2057 dmu_objset_type(os) == DMU_OST_ZVOL) {
2059 2058 error = zvol_get_stats(os, nv);
2060 2059 if (error == EIO)
2061 2060 return (error);
2062 2061 VERIFY0(error);
2063 2062 }
2064 2063 error = put_nvlist(zc, nv);
2065 2064 nvlist_free(nv);
2066 2065 }
2067 2066
2068 2067 return (error);
2069 2068 }
2070 2069
2071 2070 /*
2072 2071 * inputs:
2073 2072 * zc_name name of filesystem
2074 2073 * zc_nvlist_dst_size size of buffer for property nvlist
2075 2074 *
2076 2075 * outputs:
2077 2076 * zc_objset_stats stats
2078 2077 * zc_nvlist_dst property nvlist
2079 2078 * zc_nvlist_dst_size size of property nvlist
2080 2079 */
2081 2080 static int
2082 2081 zfs_ioc_objset_stats(zfs_cmd_t *zc)
2083 2082 {
2084 2083 objset_t *os;
2085 2084 int error;
2086 2085
2087 2086 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
2088 2087 if (error == 0) {
2089 2088 error = zfs_ioc_objset_stats_impl(zc, os);
2090 2089 dmu_objset_rele(os, FTAG);
2091 2090 }
2092 2091
2093 2092 return (error);
2094 2093 }
2095 2094
2096 2095 /*
2097 2096 * inputs:
2098 2097 * zc_name name of filesystem
2099 2098 * zc_nvlist_dst_size size of buffer for property nvlist
2100 2099 *
2101 2100 * outputs:
2102 2101 * zc_nvlist_dst received property nvlist
2103 2102 * zc_nvlist_dst_size size of received property nvlist
2104 2103 *
2105 2104 * Gets received properties (distinct from local properties on or after
2106 2105 * SPA_VERSION_RECVD_PROPS) for callers who want to differentiate received from
2107 2106 * local property values.
2108 2107 */
2109 2108 static int
2110 2109 zfs_ioc_objset_recvd_props(zfs_cmd_t *zc)
2111 2110 {
2112 2111 int error = 0;
2113 2112 nvlist_t *nv;
2114 2113
2115 2114 /*
2116 2115 * Without this check, we would return local property values if the
2117 2116 * caller has not already received properties on or after
2118 2117 * SPA_VERSION_RECVD_PROPS.
2119 2118 */
2120 2119 if (!dsl_prop_get_hasrecvd(zc->zc_name))
2121 2120 return (SET_ERROR(ENOTSUP));
2122 2121
2123 2122 if (zc->zc_nvlist_dst != 0 &&
2124 2123 (error = dsl_prop_get_received(zc->zc_name, &nv)) == 0) {
2125 2124 error = put_nvlist(zc, nv);
2126 2125 nvlist_free(nv);
2127 2126 }
2128 2127
2129 2128 return (error);
2130 2129 }
2131 2130
2132 2131 static int
2133 2132 nvl_add_zplprop(objset_t *os, nvlist_t *props, zfs_prop_t prop)
2134 2133 {
2135 2134 uint64_t value;
2136 2135 int error;
2137 2136
2138 2137 /*
2139 2138 * zfs_get_zplprop() will either find a value or give us
2140 2139 * the default value (if there is one).
2141 2140 */
2142 2141 if ((error = zfs_get_zplprop(os, prop, &value)) != 0)
2143 2142 return (error);
2144 2143 VERIFY(nvlist_add_uint64(props, zfs_prop_to_name(prop), value) == 0);
2145 2144 return (0);
2146 2145 }
2147 2146
2148 2147 /*
2149 2148 * inputs:
2150 2149 * zc_name name of filesystem
2151 2150 * zc_nvlist_dst_size size of buffer for zpl property nvlist
2152 2151 *
2153 2152 * outputs:
2154 2153 * zc_nvlist_dst zpl property nvlist
2155 2154 * zc_nvlist_dst_size size of zpl property nvlist
2156 2155 */
2157 2156 static int
2158 2157 zfs_ioc_objset_zplprops(zfs_cmd_t *zc)
2159 2158 {
2160 2159 objset_t *os;
2161 2160 int err;
2162 2161
2163 2162 /* XXX reading without owning */
2164 2163 if (err = dmu_objset_hold(zc->zc_name, FTAG, &os))
2165 2164 return (err);
2166 2165
2167 2166 dmu_objset_fast_stat(os, &zc->zc_objset_stats);
2168 2167
2169 2168 /*
2170 2169 * NB: nvl_add_zplprop() will read the objset contents,
2171 2170 * which we aren't supposed to do with a DS_MODE_USER
2172 2171 * hold, because it could be inconsistent.
2173 2172 */
2174 2173 if (zc->zc_nvlist_dst != NULL &&
2175 2174 !zc->zc_objset_stats.dds_inconsistent &&
2176 2175 dmu_objset_type(os) == DMU_OST_ZFS) {
2177 2176 nvlist_t *nv;
2178 2177
2179 2178 VERIFY(nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2180 2179 if ((err = nvl_add_zplprop(os, nv, ZFS_PROP_VERSION)) == 0 &&
2181 2180 (err = nvl_add_zplprop(os, nv, ZFS_PROP_NORMALIZE)) == 0 &&
2182 2181 (err = nvl_add_zplprop(os, nv, ZFS_PROP_UTF8ONLY)) == 0 &&
2183 2182 (err = nvl_add_zplprop(os, nv, ZFS_PROP_CASE)) == 0)
2184 2183 err = put_nvlist(zc, nv);
2185 2184 nvlist_free(nv);
2186 2185 } else {
2187 2186 err = SET_ERROR(ENOENT);
2188 2187 }
2189 2188 dmu_objset_rele(os, FTAG);
2190 2189 return (err);
2191 2190 }
2192 2191
2193 2192 static boolean_t
2194 2193 dataset_name_hidden(const char *name)
2195 2194 {
2196 2195 /*
2197 2196 * Skip over datasets that are not visible in this zone,
2198 2197 * internal datasets (which have a $ in their name), and
2199 2198 * temporary datasets (which have a % in their name).
2200 2199 */
2201 2200 if (strchr(name, '$') != NULL)
2202 2201 return (B_TRUE);
2203 2202 if (strchr(name, '%') != NULL)
2204 2203 return (B_TRUE);
2205 2204 if (!INGLOBALZONE(curproc) && !zone_dataset_visible(name, NULL))
2206 2205 return (B_TRUE);
2207 2206 return (B_FALSE);
2208 2207 }
2209 2208
2210 2209 /*
2211 2210 * inputs:
2212 2211 * zc_name name of filesystem
2213 2212 * zc_cookie zap cursor
2214 2213 * zc_nvlist_dst_size size of buffer for property nvlist
2215 2214 *
2216 2215 * outputs:
2217 2216 * zc_name name of next filesystem
2218 2217 * zc_cookie zap cursor
2219 2218 * zc_objset_stats stats
2220 2219 * zc_nvlist_dst property nvlist
2221 2220 * zc_nvlist_dst_size size of property nvlist
2222 2221 */
2223 2222 static int
2224 2223 zfs_ioc_dataset_list_next(zfs_cmd_t *zc)
2225 2224 {
2226 2225 objset_t *os;
2227 2226 int error;
2228 2227 char *p;
2229 2228 size_t orig_len = strlen(zc->zc_name);
2230 2229
2231 2230 top:
2232 2231 if (error = dmu_objset_hold(zc->zc_name, FTAG, &os)) {
2233 2232 if (error == ENOENT)
2234 2233 error = SET_ERROR(ESRCH);
2235 2234 return (error);
2236 2235 }
2237 2236
2238 2237 p = strrchr(zc->zc_name, '/');
2239 2238 if (p == NULL || p[1] != '\0')
2240 2239 (void) strlcat(zc->zc_name, "/", sizeof (zc->zc_name));
2241 2240 p = zc->zc_name + strlen(zc->zc_name);
2242 2241
2243 2242 do {
2244 2243 error = dmu_dir_list_next(os,
2245 2244 sizeof (zc->zc_name) - (p - zc->zc_name), p,
2246 2245 NULL, &zc->zc_cookie);
2247 2246 if (error == ENOENT)
2248 2247 error = SET_ERROR(ESRCH);
2249 2248 } while (error == 0 && dataset_name_hidden(zc->zc_name));
2250 2249 dmu_objset_rele(os, FTAG);
2251 2250
2252 2251 /*
2253 2252 * If it's an internal dataset (ie. with a '$' in its name),
2254 2253 * don't try to get stats for it, otherwise we'll return ENOENT.
2255 2254 */
2256 2255 if (error == 0 && strchr(zc->zc_name, '$') == NULL) {
2257 2256 error = zfs_ioc_objset_stats(zc); /* fill in the stats */
2258 2257 if (error == ENOENT) {
2259 2258 /* We lost a race with destroy, get the next one. */
2260 2259 zc->zc_name[orig_len] = '\0';
2261 2260 goto top;
2262 2261 }
2263 2262 }
2264 2263 return (error);
2265 2264 }
2266 2265
2267 2266 /*
2268 2267 * inputs:
2269 2268 * zc_name name of filesystem
2270 2269 * zc_cookie zap cursor
2271 2270 * zc_nvlist_dst_size size of buffer for property nvlist
2272 2271 *
2273 2272 * outputs:
2274 2273 * zc_name name of next snapshot
2275 2274 * zc_objset_stats stats
2276 2275 * zc_nvlist_dst property nvlist
2277 2276 * zc_nvlist_dst_size size of property nvlist
2278 2277 */
2279 2278 static int
2280 2279 zfs_ioc_snapshot_list_next(zfs_cmd_t *zc)
2281 2280 {
2282 2281 objset_t *os;
2283 2282 int error;
2284 2283
2285 2284 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
2286 2285 if (error != 0) {
2287 2286 return (error == ENOENT ? ESRCH : error);
2288 2287 }
2289 2288
2290 2289 /*
2291 2290 * A dataset name of maximum length cannot have any snapshots,
2292 2291 * so exit immediately.
2293 2292 */
2294 2293 if (strlcat(zc->zc_name, "@", sizeof (zc->zc_name)) >= MAXNAMELEN) {
2295 2294 dmu_objset_rele(os, FTAG);
2296 2295 return (SET_ERROR(ESRCH));
2297 2296 }
2298 2297
2299 2298 error = dmu_snapshot_list_next(os,
2300 2299 sizeof (zc->zc_name) - strlen(zc->zc_name),
2301 2300 zc->zc_name + strlen(zc->zc_name), &zc->zc_obj, &zc->zc_cookie,
2302 2301 NULL);
2303 2302
2304 2303 if (error == 0) {
2305 2304 dsl_dataset_t *ds;
2306 2305 dsl_pool_t *dp = os->os_dsl_dataset->ds_dir->dd_pool;
2307 2306
2308 2307 error = dsl_dataset_hold_obj(dp, zc->zc_obj, FTAG, &ds);
2309 2308 if (error == 0) {
2310 2309 objset_t *ossnap;
2311 2310
2312 2311 error = dmu_objset_from_ds(ds, &ossnap);
2313 2312 if (error == 0)
2314 2313 error = zfs_ioc_objset_stats_impl(zc, ossnap);
2315 2314 dsl_dataset_rele(ds, FTAG);
2316 2315 }
2317 2316 } else if (error == ENOENT) {
2318 2317 error = SET_ERROR(ESRCH);
2319 2318 }
2320 2319
2321 2320 dmu_objset_rele(os, FTAG);
2322 2321 /* if we failed, undo the @ that we tacked on to zc_name */
2323 2322 if (error != 0)
2324 2323 *strchr(zc->zc_name, '@') = '\0';
2325 2324 return (error);
2326 2325 }
2327 2326
2328 2327 static int
2329 2328 zfs_prop_set_userquota(const char *dsname, nvpair_t *pair)
2330 2329 {
2331 2330 const char *propname = nvpair_name(pair);
2332 2331 uint64_t *valary;
2333 2332 unsigned int vallen;
2334 2333 const char *domain;
2335 2334 char *dash;
2336 2335 zfs_userquota_prop_t type;
2337 2336 uint64_t rid;
2338 2337 uint64_t quota;
2339 2338 zfsvfs_t *zfsvfs;
2340 2339 int err;
2341 2340
2342 2341 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2343 2342 nvlist_t *attrs;
2344 2343 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
2345 2344 if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2346 2345 &pair) != 0)
2347 2346 return (SET_ERROR(EINVAL));
2348 2347 }
2349 2348
2350 2349 /*
2351 2350 * A correctly constructed propname is encoded as
2352 2351 * userquota@<rid>-<domain>.
2353 2352 */
2354 2353 if ((dash = strchr(propname, '-')) == NULL ||
2355 2354 nvpair_value_uint64_array(pair, &valary, &vallen) != 0 ||
2356 2355 vallen != 3)
2357 2356 return (SET_ERROR(EINVAL));
2358 2357
2359 2358 domain = dash + 1;
2360 2359 type = valary[0];
2361 2360 rid = valary[1];
2362 2361 quota = valary[2];
2363 2362
2364 2363 err = zfsvfs_hold(dsname, FTAG, &zfsvfs, B_FALSE);
2365 2364 if (err == 0) {
2366 2365 err = zfs_set_userquota(zfsvfs, type, domain, rid, quota);
2367 2366 zfsvfs_rele(zfsvfs, FTAG);
2368 2367 }
2369 2368
2370 2369 return (err);
2371 2370 }
2372 2371
2373 2372 /*
2374 2373 * If the named property is one that has a special function to set its value,
2375 2374 * return 0 on success and a positive error code on failure; otherwise if it is
2376 2375 * not one of the special properties handled by this function, return -1.
2377 2376 *
2378 2377 * XXX: It would be better for callers of the property interface if we handled
2379 2378 * these special cases in dsl_prop.c (in the dsl layer).
2380 2379 */
2381 2380 static int
2382 2381 zfs_prop_set_special(const char *dsname, zprop_source_t source,
2383 2382 nvpair_t *pair)
2384 2383 {
2385 2384 const char *propname = nvpair_name(pair);
2386 2385 zfs_prop_t prop = zfs_name_to_prop(propname);
2387 2386 uint64_t intval;
2388 2387 int err = -1;
2389 2388
2390 2389 if (prop == ZPROP_INVAL) {
2391 2390 if (zfs_prop_userquota(propname))
2392 2391 return (zfs_prop_set_userquota(dsname, pair));
2393 2392 return (-1);
2394 2393 }
2395 2394
2396 2395 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2397 2396 nvlist_t *attrs;
2398 2397 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
2399 2398 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2400 2399 &pair) == 0);
2401 2400 }
2402 2401
2403 2402 if (zfs_prop_get_type(prop) == PROP_TYPE_STRING)
2404 2403 return (-1);
2405 2404
2406 2405 VERIFY(0 == nvpair_value_uint64(pair, &intval));
2407 2406
2408 2407 switch (prop) {
2409 2408 case ZFS_PROP_QUOTA:
2410 2409 err = dsl_dir_set_quota(dsname, source, intval);
2411 2410 break;
2412 2411 case ZFS_PROP_REFQUOTA:
2413 2412 err = dsl_dataset_set_refquota(dsname, source, intval);
2414 2413 break;
2415 2414 case ZFS_PROP_FILESYSTEM_LIMIT:
2416 2415 case ZFS_PROP_SNAPSHOT_LIMIT:
2417 2416 if (intval == UINT64_MAX) {
2418 2417 /* clearing the limit, just do it */
2419 2418 err = 0;
2420 2419 } else {
2421 2420 err = dsl_dir_activate_fs_ss_limit(dsname);
2422 2421 }
2423 2422 /*
2424 2423 * Set err to -1 to force the zfs_set_prop_nvlist code down the
2425 2424 * default path to set the value in the nvlist.
2426 2425 */
2427 2426 if (err == 0)
2428 2427 err = -1;
2429 2428 break;
2430 2429 case ZFS_PROP_RESERVATION:
2431 2430 err = dsl_dir_set_reservation(dsname, source, intval);
2432 2431 break;
2433 2432 case ZFS_PROP_REFRESERVATION:
2434 2433 err = dsl_dataset_set_refreservation(dsname, source, intval);
2435 2434 break;
2436 2435 case ZFS_PROP_VOLSIZE:
2437 2436 err = zvol_set_volsize(dsname, intval);
2438 2437 break;
2439 2438 case ZFS_PROP_VERSION:
2440 2439 {
2441 2440 zfsvfs_t *zfsvfs;
2442 2441
2443 2442 if ((err = zfsvfs_hold(dsname, FTAG, &zfsvfs, B_TRUE)) != 0)
2444 2443 break;
2445 2444
2446 2445 err = zfs_set_version(zfsvfs, intval);
2447 2446 zfsvfs_rele(zfsvfs, FTAG);
2448 2447
2449 2448 if (err == 0 && intval >= ZPL_VERSION_USERSPACE) {
2450 2449 zfs_cmd_t *zc;
2451 2450
2452 2451 zc = kmem_zalloc(sizeof (zfs_cmd_t), KM_SLEEP);
2453 2452 (void) strcpy(zc->zc_name, dsname);
2454 2453 (void) zfs_ioc_userspace_upgrade(zc);
2455 2454 kmem_free(zc, sizeof (zfs_cmd_t));
2456 2455 }
2457 2456 break;
2458 2457 }
2459 2458 default:
2460 2459 err = -1;
2461 2460 }
2462 2461
2463 2462 return (err);
2464 2463 }
2465 2464
2466 2465 /*
2467 2466 * This function is best effort. If it fails to set any of the given properties,
2468 2467 * it continues to set as many as it can and returns the last error
2469 2468 * encountered. If the caller provides a non-NULL errlist, it will be filled in
2470 2469 * with the list of names of all the properties that failed along with the
2471 2470 * corresponding error numbers.
2472 2471 *
2473 2472 * If every property is set successfully, zero is returned and errlist is not
2474 2473 * modified.
2475 2474 */
2476 2475 int
2477 2476 zfs_set_prop_nvlist(const char *dsname, zprop_source_t source, nvlist_t *nvl,
2478 2477 nvlist_t *errlist)
2479 2478 {
2480 2479 nvpair_t *pair;
2481 2480 nvpair_t *propval;
2482 2481 int rv = 0;
2483 2482 uint64_t intval;
2484 2483 char *strval;
2485 2484 nvlist_t *genericnvl = fnvlist_alloc();
2486 2485 nvlist_t *retrynvl = fnvlist_alloc();
2487 2486
2488 2487 retry:
2489 2488 pair = NULL;
2490 2489 while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) {
2491 2490 const char *propname = nvpair_name(pair);
2492 2491 zfs_prop_t prop = zfs_name_to_prop(propname);
2493 2492 int err = 0;
2494 2493
2495 2494 /* decode the property value */
2496 2495 propval = pair;
2497 2496 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2498 2497 nvlist_t *attrs;
2499 2498 attrs = fnvpair_value_nvlist(pair);
2500 2499 if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2501 2500 &propval) != 0)
2502 2501 err = SET_ERROR(EINVAL);
2503 2502 }
2504 2503
2505 2504 /* Validate value type */
2506 2505 if (err == 0 && prop == ZPROP_INVAL) {
2507 2506 if (zfs_prop_user(propname)) {
2508 2507 if (nvpair_type(propval) != DATA_TYPE_STRING)
2509 2508 err = SET_ERROR(EINVAL);
2510 2509 } else if (zfs_prop_userquota(propname)) {
2511 2510 if (nvpair_type(propval) !=
2512 2511 DATA_TYPE_UINT64_ARRAY)
2513 2512 err = SET_ERROR(EINVAL);
2514 2513 } else {
2515 2514 err = SET_ERROR(EINVAL);
2516 2515 }
2517 2516 } else if (err == 0) {
2518 2517 if (nvpair_type(propval) == DATA_TYPE_STRING) {
2519 2518 if (zfs_prop_get_type(prop) != PROP_TYPE_STRING)
2520 2519 err = SET_ERROR(EINVAL);
2521 2520 } else if (nvpair_type(propval) == DATA_TYPE_UINT64) {
2522 2521 const char *unused;
2523 2522
2524 2523 intval = fnvpair_value_uint64(propval);
2525 2524
2526 2525 switch (zfs_prop_get_type(prop)) {
2527 2526 case PROP_TYPE_NUMBER:
2528 2527 break;
2529 2528 case PROP_TYPE_STRING:
2530 2529 err = SET_ERROR(EINVAL);
2531 2530 break;
2532 2531 case PROP_TYPE_INDEX:
2533 2532 if (zfs_prop_index_to_string(prop,
2534 2533 intval, &unused) != 0)
2535 2534 err = SET_ERROR(EINVAL);
2536 2535 break;
2537 2536 default:
2538 2537 cmn_err(CE_PANIC,
2539 2538 "unknown property type");
2540 2539 }
2541 2540 } else {
2542 2541 err = SET_ERROR(EINVAL);
2543 2542 }
2544 2543 }
2545 2544
2546 2545 /* Validate permissions */
2547 2546 if (err == 0)
2548 2547 err = zfs_check_settable(dsname, pair, CRED());
2549 2548
2550 2549 if (err == 0) {
2551 2550 err = zfs_prop_set_special(dsname, source, pair);
2552 2551 if (err == -1) {
2553 2552 /*
2554 2553 * For better performance we build up a list of
2555 2554 * properties to set in a single transaction.
2556 2555 */
2557 2556 err = nvlist_add_nvpair(genericnvl, pair);
2558 2557 } else if (err != 0 && nvl != retrynvl) {
2559 2558 /*
2560 2559 * This may be a spurious error caused by
2561 2560 * receiving quota and reservation out of order.
2562 2561 * Try again in a second pass.
2563 2562 */
2564 2563 err = nvlist_add_nvpair(retrynvl, pair);
2565 2564 }
2566 2565 }
2567 2566
2568 2567 if (err != 0) {
2569 2568 if (errlist != NULL)
2570 2569 fnvlist_add_int32(errlist, propname, err);
2571 2570 rv = err;
2572 2571 }
2573 2572 }
2574 2573
2575 2574 if (nvl != retrynvl && !nvlist_empty(retrynvl)) {
2576 2575 nvl = retrynvl;
2577 2576 goto retry;
2578 2577 }
2579 2578
2580 2579 if (!nvlist_empty(genericnvl) &&
2581 2580 dsl_props_set(dsname, source, genericnvl) != 0) {
2582 2581 /*
2583 2582 * If this fails, we still want to set as many properties as we
2584 2583 * can, so try setting them individually.
2585 2584 */
2586 2585 pair = NULL;
2587 2586 while ((pair = nvlist_next_nvpair(genericnvl, pair)) != NULL) {
2588 2587 const char *propname = nvpair_name(pair);
2589 2588 int err = 0;
2590 2589
2591 2590 propval = pair;
2592 2591 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2593 2592 nvlist_t *attrs;
2594 2593 attrs = fnvpair_value_nvlist(pair);
2595 2594 propval = fnvlist_lookup_nvpair(attrs,
2596 2595 ZPROP_VALUE);
2597 2596 }
2598 2597
2599 2598 if (nvpair_type(propval) == DATA_TYPE_STRING) {
2600 2599 strval = fnvpair_value_string(propval);
2601 2600 err = dsl_prop_set_string(dsname, propname,
2602 2601 source, strval);
2603 2602 } else {
2604 2603 intval = fnvpair_value_uint64(propval);
2605 2604 err = dsl_prop_set_int(dsname, propname, source,
2606 2605 intval);
2607 2606 }
2608 2607
2609 2608 if (err != 0) {
2610 2609 if (errlist != NULL) {
2611 2610 fnvlist_add_int32(errlist, propname,
2612 2611 err);
2613 2612 }
2614 2613 rv = err;
2615 2614 }
2616 2615 }
2617 2616 }
2618 2617 nvlist_free(genericnvl);
2619 2618 nvlist_free(retrynvl);
2620 2619
2621 2620 return (rv);
2622 2621 }
2623 2622
2624 2623 /*
2625 2624 * Check that all the properties are valid user properties.
2626 2625 */
2627 2626 static int
2628 2627 zfs_check_userprops(const char *fsname, nvlist_t *nvl)
2629 2628 {
2630 2629 nvpair_t *pair = NULL;
2631 2630 int error = 0;
2632 2631
2633 2632 while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) {
2634 2633 const char *propname = nvpair_name(pair);
2635 2634
2636 2635 if (!zfs_prop_user(propname) ||
2637 2636 nvpair_type(pair) != DATA_TYPE_STRING)
2638 2637 return (SET_ERROR(EINVAL));
2639 2638
2640 2639 if (error = zfs_secpolicy_write_perms(fsname,
2641 2640 ZFS_DELEG_PERM_USERPROP, CRED()))
2642 2641 return (error);
2643 2642
2644 2643 if (strlen(propname) >= ZAP_MAXNAMELEN)
2645 2644 return (SET_ERROR(ENAMETOOLONG));
2646 2645
2647 2646 if (strlen(fnvpair_value_string(pair)) >= ZAP_MAXVALUELEN)
2648 2647 return (E2BIG);
2649 2648 }
2650 2649 return (0);
2651 2650 }
2652 2651
2653 2652 static void
2654 2653 props_skip(nvlist_t *props, nvlist_t *skipped, nvlist_t **newprops)
2655 2654 {
2656 2655 nvpair_t *pair;
2657 2656
2658 2657 VERIFY(nvlist_alloc(newprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2659 2658
2660 2659 pair = NULL;
2661 2660 while ((pair = nvlist_next_nvpair(props, pair)) != NULL) {
2662 2661 if (nvlist_exists(skipped, nvpair_name(pair)))
2663 2662 continue;
2664 2663
2665 2664 VERIFY(nvlist_add_nvpair(*newprops, pair) == 0);
2666 2665 }
2667 2666 }
2668 2667
2669 2668 static int
2670 2669 clear_received_props(const char *dsname, nvlist_t *props,
2671 2670 nvlist_t *skipped)
2672 2671 {
2673 2672 int err = 0;
2674 2673 nvlist_t *cleared_props = NULL;
2675 2674 props_skip(props, skipped, &cleared_props);
2676 2675 if (!nvlist_empty(cleared_props)) {
2677 2676 /*
2678 2677 * Acts on local properties until the dataset has received
2679 2678 * properties at least once on or after SPA_VERSION_RECVD_PROPS.
2680 2679 */
2681 2680 zprop_source_t flags = (ZPROP_SRC_NONE |
2682 2681 (dsl_prop_get_hasrecvd(dsname) ? ZPROP_SRC_RECEIVED : 0));
2683 2682 err = zfs_set_prop_nvlist(dsname, flags, cleared_props, NULL);
2684 2683 }
2685 2684 nvlist_free(cleared_props);
2686 2685 return (err);
2687 2686 }
2688 2687
2689 2688 /*
2690 2689 * inputs:
2691 2690 * zc_name name of filesystem
2692 2691 * zc_value name of property to set
2693 2692 * zc_nvlist_src{_size} nvlist of properties to apply
2694 2693 * zc_cookie received properties flag
2695 2694 *
2696 2695 * outputs:
2697 2696 * zc_nvlist_dst{_size} error for each unapplied received property
2698 2697 */
2699 2698 static int
2700 2699 zfs_ioc_set_prop(zfs_cmd_t *zc)
2701 2700 {
2702 2701 nvlist_t *nvl;
2703 2702 boolean_t received = zc->zc_cookie;
2704 2703 zprop_source_t source = (received ? ZPROP_SRC_RECEIVED :
2705 2704 ZPROP_SRC_LOCAL);
2706 2705 nvlist_t *errors;
2707 2706 int error;
2708 2707
2709 2708 if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2710 2709 zc->zc_iflags, &nvl)) != 0)
2711 2710 return (error);
2712 2711
2713 2712 if (received) {
2714 2713 nvlist_t *origprops;
2715 2714
2716 2715 if (dsl_prop_get_received(zc->zc_name, &origprops) == 0) {
2717 2716 (void) clear_received_props(zc->zc_name,
2718 2717 origprops, nvl);
2719 2718 nvlist_free(origprops);
2720 2719 }
2721 2720
2722 2721 error = dsl_prop_set_hasrecvd(zc->zc_name);
2723 2722 }
2724 2723
2725 2724 errors = fnvlist_alloc();
2726 2725 if (error == 0)
2727 2726 error = zfs_set_prop_nvlist(zc->zc_name, source, nvl, errors);
2728 2727
2729 2728 if (zc->zc_nvlist_dst != NULL && errors != NULL) {
2730 2729 (void) put_nvlist(zc, errors);
2731 2730 }
2732 2731
2733 2732 nvlist_free(errors);
2734 2733 nvlist_free(nvl);
2735 2734 return (error);
2736 2735 }
2737 2736
2738 2737 /*
2739 2738 * inputs:
2740 2739 * zc_name name of filesystem
2741 2740 * zc_value name of property to inherit
2742 2741 * zc_cookie revert to received value if TRUE
2743 2742 *
2744 2743 * outputs: none
2745 2744 */
2746 2745 static int
2747 2746 zfs_ioc_inherit_prop(zfs_cmd_t *zc)
2748 2747 {
2749 2748 const char *propname = zc->zc_value;
2750 2749 zfs_prop_t prop = zfs_name_to_prop(propname);
2751 2750 boolean_t received = zc->zc_cookie;
2752 2751 zprop_source_t source = (received
2753 2752 ? ZPROP_SRC_NONE /* revert to received value, if any */
2754 2753 : ZPROP_SRC_INHERITED); /* explicitly inherit */
2755 2754
2756 2755 if (received) {
2757 2756 nvlist_t *dummy;
2758 2757 nvpair_t *pair;
2759 2758 zprop_type_t type;
2760 2759 int err;
2761 2760
2762 2761 /*
2763 2762 * zfs_prop_set_special() expects properties in the form of an
2764 2763 * nvpair with type info.
2765 2764 */
2766 2765 if (prop == ZPROP_INVAL) {
2767 2766 if (!zfs_prop_user(propname))
2768 2767 return (SET_ERROR(EINVAL));
2769 2768
2770 2769 type = PROP_TYPE_STRING;
2771 2770 } else if (prop == ZFS_PROP_VOLSIZE ||
2772 2771 prop == ZFS_PROP_VERSION) {
2773 2772 return (SET_ERROR(EINVAL));
2774 2773 } else {
2775 2774 type = zfs_prop_get_type(prop);
2776 2775 }
2777 2776
2778 2777 VERIFY(nvlist_alloc(&dummy, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2779 2778
2780 2779 switch (type) {
2781 2780 case PROP_TYPE_STRING:
2782 2781 VERIFY(0 == nvlist_add_string(dummy, propname, ""));
2783 2782 break;
2784 2783 case PROP_TYPE_NUMBER:
2785 2784 case PROP_TYPE_INDEX:
2786 2785 VERIFY(0 == nvlist_add_uint64(dummy, propname, 0));
2787 2786 break;
2788 2787 default:
2789 2788 nvlist_free(dummy);
2790 2789 return (SET_ERROR(EINVAL));
2791 2790 }
2792 2791
2793 2792 pair = nvlist_next_nvpair(dummy, NULL);
2794 2793 err = zfs_prop_set_special(zc->zc_name, source, pair);
2795 2794 nvlist_free(dummy);
2796 2795 if (err != -1)
2797 2796 return (err); /* special property already handled */
2798 2797 } else {
2799 2798 /*
2800 2799 * Only check this in the non-received case. We want to allow
2801 2800 * 'inherit -S' to revert non-inheritable properties like quota
2802 2801 * and reservation to the received or default values even though
2803 2802 * they are not considered inheritable.
2804 2803 */
2805 2804 if (prop != ZPROP_INVAL && !zfs_prop_inheritable(prop))
2806 2805 return (SET_ERROR(EINVAL));
2807 2806 }
2808 2807
2809 2808 /* property name has been validated by zfs_secpolicy_inherit_prop() */
2810 2809 return (dsl_prop_inherit(zc->zc_name, zc->zc_value, source));
2811 2810 }
2812 2811
2813 2812 static int
2814 2813 zfs_ioc_pool_set_props(zfs_cmd_t *zc)
2815 2814 {
2816 2815 nvlist_t *props;
2817 2816 spa_t *spa;
2818 2817 int error;
2819 2818 nvpair_t *pair;
2820 2819
2821 2820 if (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2822 2821 zc->zc_iflags, &props))
2823 2822 return (error);
2824 2823
2825 2824 /*
2826 2825 * If the only property is the configfile, then just do a spa_lookup()
2827 2826 * to handle the faulted case.
2828 2827 */
2829 2828 pair = nvlist_next_nvpair(props, NULL);
2830 2829 if (pair != NULL && strcmp(nvpair_name(pair),
2831 2830 zpool_prop_to_name(ZPOOL_PROP_CACHEFILE)) == 0 &&
2832 2831 nvlist_next_nvpair(props, pair) == NULL) {
2833 2832 mutex_enter(&spa_namespace_lock);
2834 2833 if ((spa = spa_lookup(zc->zc_name)) != NULL) {
2835 2834 spa_configfile_set(spa, props, B_FALSE);
2836 2835 spa_config_sync(spa, B_FALSE, B_TRUE);
2837 2836 }
2838 2837 mutex_exit(&spa_namespace_lock);
2839 2838 if (spa != NULL) {
2840 2839 nvlist_free(props);
2841 2840 return (0);
2842 2841 }
2843 2842 }
2844 2843
2845 2844 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) {
2846 2845 nvlist_free(props);
2847 2846 return (error);
2848 2847 }
2849 2848
2850 2849 error = spa_prop_set(spa, props);
2851 2850
2852 2851 nvlist_free(props);
2853 2852 spa_close(spa, FTAG);
2854 2853
2855 2854 return (error);
2856 2855 }
2857 2856
2858 2857 static int
2859 2858 zfs_ioc_pool_get_props(zfs_cmd_t *zc)
2860 2859 {
2861 2860 spa_t *spa;
2862 2861 int error;
2863 2862 nvlist_t *nvp = NULL;
2864 2863
2865 2864 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) {
2866 2865 /*
2867 2866 * If the pool is faulted, there may be properties we can still
2868 2867 * get (such as altroot and cachefile), so attempt to get them
2869 2868 * anyway.
2870 2869 */
2871 2870 mutex_enter(&spa_namespace_lock);
2872 2871 if ((spa = spa_lookup(zc->zc_name)) != NULL)
2873 2872 error = spa_prop_get(spa, &nvp);
2874 2873 mutex_exit(&spa_namespace_lock);
2875 2874 } else {
2876 2875 error = spa_prop_get(spa, &nvp);
2877 2876 spa_close(spa, FTAG);
2878 2877 }
2879 2878
2880 2879 if (error == 0 && zc->zc_nvlist_dst != NULL)
2881 2880 error = put_nvlist(zc, nvp);
2882 2881 else
2883 2882 error = SET_ERROR(EFAULT);
2884 2883
2885 2884 nvlist_free(nvp);
2886 2885 return (error);
2887 2886 }
2888 2887
2889 2888 /*
2890 2889 * inputs:
2891 2890 * zc_name name of filesystem
2892 2891 * zc_nvlist_src{_size} nvlist of delegated permissions
2893 2892 * zc_perm_action allow/unallow flag
2894 2893 *
2895 2894 * outputs: none
2896 2895 */
2897 2896 static int
2898 2897 zfs_ioc_set_fsacl(zfs_cmd_t *zc)
2899 2898 {
2900 2899 int error;
2901 2900 nvlist_t *fsaclnv = NULL;
2902 2901
2903 2902 if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2904 2903 zc->zc_iflags, &fsaclnv)) != 0)
2905 2904 return (error);
2906 2905
2907 2906 /*
2908 2907 * Verify nvlist is constructed correctly
2909 2908 */
2910 2909 if ((error = zfs_deleg_verify_nvlist(fsaclnv)) != 0) {
2911 2910 nvlist_free(fsaclnv);
2912 2911 return (SET_ERROR(EINVAL));
2913 2912 }
2914 2913
2915 2914 /*
2916 2915 * If we don't have PRIV_SYS_MOUNT, then validate
2917 2916 * that user is allowed to hand out each permission in
2918 2917 * the nvlist(s)
2919 2918 */
2920 2919
2921 2920 error = secpolicy_zfs(CRED());
2922 2921 if (error != 0) {
2923 2922 if (zc->zc_perm_action == B_FALSE) {
2924 2923 error = dsl_deleg_can_allow(zc->zc_name,
2925 2924 fsaclnv, CRED());
2926 2925 } else {
2927 2926 error = dsl_deleg_can_unallow(zc->zc_name,
2928 2927 fsaclnv, CRED());
2929 2928 }
2930 2929 }
2931 2930
2932 2931 if (error == 0)
2933 2932 error = dsl_deleg_set(zc->zc_name, fsaclnv, zc->zc_perm_action);
2934 2933
2935 2934 nvlist_free(fsaclnv);
2936 2935 return (error);
2937 2936 }
2938 2937
2939 2938 /*
2940 2939 * inputs:
2941 2940 * zc_name name of filesystem
2942 2941 *
2943 2942 * outputs:
2944 2943 * zc_nvlist_src{_size} nvlist of delegated permissions
2945 2944 */
2946 2945 static int
2947 2946 zfs_ioc_get_fsacl(zfs_cmd_t *zc)
2948 2947 {
2949 2948 nvlist_t *nvp;
2950 2949 int error;
2951 2950
2952 2951 if ((error = dsl_deleg_get(zc->zc_name, &nvp)) == 0) {
2953 2952 error = put_nvlist(zc, nvp);
2954 2953 nvlist_free(nvp);
2955 2954 }
2956 2955
2957 2956 return (error);
2958 2957 }
2959 2958
2960 2959 /*
2961 2960 * Search the vfs list for a specified resource. Returns a pointer to it
2962 2961 * or NULL if no suitable entry is found. The caller of this routine
2963 2962 * is responsible for releasing the returned vfs pointer.
2964 2963 */
2965 2964 static vfs_t *
2966 2965 zfs_get_vfs(const char *resource)
2967 2966 {
2968 2967 struct vfs *vfsp;
2969 2968 struct vfs *vfs_found = NULL;
2970 2969
2971 2970 vfs_list_read_lock();
2972 2971 vfsp = rootvfs;
2973 2972 do {
2974 2973 if (strcmp(refstr_value(vfsp->vfs_resource), resource) == 0) {
2975 2974 VFS_HOLD(vfsp);
2976 2975 vfs_found = vfsp;
2977 2976 break;
2978 2977 }
2979 2978 vfsp = vfsp->vfs_next;
2980 2979 } while (vfsp != rootvfs);
2981 2980 vfs_list_unlock();
2982 2981 return (vfs_found);
2983 2982 }
2984 2983
2985 2984 /* ARGSUSED */
2986 2985 static void
2987 2986 zfs_create_cb(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx)
2988 2987 {
2989 2988 zfs_creat_t *zct = arg;
2990 2989
2991 2990 zfs_create_fs(os, cr, zct->zct_zplprops, tx);
2992 2991 }
2993 2992
2994 2993 #define ZFS_PROP_UNDEFINED ((uint64_t)-1)
2995 2994
2996 2995 /*
2997 2996 * inputs:
2998 2997 * os parent objset pointer (NULL if root fs)
2999 2998 * fuids_ok fuids allowed in this version of the spa?
3000 2999 * sa_ok SAs allowed in this version of the spa?
3001 3000 * createprops list of properties requested by creator
3002 3001 *
3003 3002 * outputs:
3004 3003 * zplprops values for the zplprops we attach to the master node object
3005 3004 * is_ci true if requested file system will be purely case-insensitive
3006 3005 *
3007 3006 * Determine the settings for utf8only, normalization and
3008 3007 * casesensitivity. Specific values may have been requested by the
3009 3008 * creator and/or we can inherit values from the parent dataset. If
3010 3009 * the file system is of too early a vintage, a creator can not
3011 3010 * request settings for these properties, even if the requested
3012 3011 * setting is the default value. We don't actually want to create dsl
3013 3012 * properties for these, so remove them from the source nvlist after
3014 3013 * processing.
3015 3014 */
3016 3015 static int
3017 3016 zfs_fill_zplprops_impl(objset_t *os, uint64_t zplver,
3018 3017 boolean_t fuids_ok, boolean_t sa_ok, nvlist_t *createprops,
3019 3018 nvlist_t *zplprops, boolean_t *is_ci)
3020 3019 {
3021 3020 uint64_t sense = ZFS_PROP_UNDEFINED;
3022 3021 uint64_t norm = ZFS_PROP_UNDEFINED;
3023 3022 uint64_t u8 = ZFS_PROP_UNDEFINED;
3024 3023
3025 3024 ASSERT(zplprops != NULL);
3026 3025
3027 3026 /*
3028 3027 * Pull out creator prop choices, if any.
3029 3028 */
3030 3029 if (createprops) {
3031 3030 (void) nvlist_lookup_uint64(createprops,
3032 3031 zfs_prop_to_name(ZFS_PROP_VERSION), &zplver);
3033 3032 (void) nvlist_lookup_uint64(createprops,
3034 3033 zfs_prop_to_name(ZFS_PROP_NORMALIZE), &norm);
3035 3034 (void) nvlist_remove_all(createprops,
3036 3035 zfs_prop_to_name(ZFS_PROP_NORMALIZE));
3037 3036 (void) nvlist_lookup_uint64(createprops,
3038 3037 zfs_prop_to_name(ZFS_PROP_UTF8ONLY), &u8);
3039 3038 (void) nvlist_remove_all(createprops,
3040 3039 zfs_prop_to_name(ZFS_PROP_UTF8ONLY));
3041 3040 (void) nvlist_lookup_uint64(createprops,
3042 3041 zfs_prop_to_name(ZFS_PROP_CASE), &sense);
3043 3042 (void) nvlist_remove_all(createprops,
3044 3043 zfs_prop_to_name(ZFS_PROP_CASE));
3045 3044 }
3046 3045
3047 3046 /*
3048 3047 * If the zpl version requested is whacky or the file system
3049 3048 * or pool is version is too "young" to support normalization
3050 3049 * and the creator tried to set a value for one of the props,
3051 3050 * error out.
3052 3051 */
3053 3052 if ((zplver < ZPL_VERSION_INITIAL || zplver > ZPL_VERSION) ||
3054 3053 (zplver >= ZPL_VERSION_FUID && !fuids_ok) ||
3055 3054 (zplver >= ZPL_VERSION_SA && !sa_ok) ||
3056 3055 (zplver < ZPL_VERSION_NORMALIZATION &&
3057 3056 (norm != ZFS_PROP_UNDEFINED || u8 != ZFS_PROP_UNDEFINED ||
3058 3057 sense != ZFS_PROP_UNDEFINED)))
3059 3058 return (SET_ERROR(ENOTSUP));
3060 3059
3061 3060 /*
3062 3061 * Put the version in the zplprops
3063 3062 */
3064 3063 VERIFY(nvlist_add_uint64(zplprops,
3065 3064 zfs_prop_to_name(ZFS_PROP_VERSION), zplver) == 0);
3066 3065
3067 3066 if (norm == ZFS_PROP_UNDEFINED)
3068 3067 VERIFY(zfs_get_zplprop(os, ZFS_PROP_NORMALIZE, &norm) == 0);
3069 3068 VERIFY(nvlist_add_uint64(zplprops,
3070 3069 zfs_prop_to_name(ZFS_PROP_NORMALIZE), norm) == 0);
3071 3070
3072 3071 /*
3073 3072 * If we're normalizing, names must always be valid UTF-8 strings.
3074 3073 */
3075 3074 if (norm)
3076 3075 u8 = 1;
3077 3076 if (u8 == ZFS_PROP_UNDEFINED)
3078 3077 VERIFY(zfs_get_zplprop(os, ZFS_PROP_UTF8ONLY, &u8) == 0);
3079 3078 VERIFY(nvlist_add_uint64(zplprops,
3080 3079 zfs_prop_to_name(ZFS_PROP_UTF8ONLY), u8) == 0);
3081 3080
3082 3081 if (sense == ZFS_PROP_UNDEFINED)
3083 3082 VERIFY(zfs_get_zplprop(os, ZFS_PROP_CASE, &sense) == 0);
3084 3083 VERIFY(nvlist_add_uint64(zplprops,
3085 3084 zfs_prop_to_name(ZFS_PROP_CASE), sense) == 0);
3086 3085
3087 3086 if (is_ci)
3088 3087 *is_ci = (sense == ZFS_CASE_INSENSITIVE);
3089 3088
3090 3089 return (0);
3091 3090 }
3092 3091
3093 3092 static int
3094 3093 zfs_fill_zplprops(const char *dataset, nvlist_t *createprops,
3095 3094 nvlist_t *zplprops, boolean_t *is_ci)
3096 3095 {
3097 3096 boolean_t fuids_ok, sa_ok;
3098 3097 uint64_t zplver = ZPL_VERSION;
3099 3098 objset_t *os = NULL;
3100 3099 char parentname[MAXNAMELEN];
3101 3100 char *cp;
3102 3101 spa_t *spa;
3103 3102 uint64_t spa_vers;
3104 3103 int error;
3105 3104
3106 3105 (void) strlcpy(parentname, dataset, sizeof (parentname));
3107 3106 cp = strrchr(parentname, '/');
3108 3107 ASSERT(cp != NULL);
3109 3108 cp[0] = '\0';
3110 3109
3111 3110 if ((error = spa_open(dataset, &spa, FTAG)) != 0)
3112 3111 return (error);
3113 3112
3114 3113 spa_vers = spa_version(spa);
3115 3114 spa_close(spa, FTAG);
3116 3115
3117 3116 zplver = zfs_zpl_version_map(spa_vers);
3118 3117 fuids_ok = (zplver >= ZPL_VERSION_FUID);
3119 3118 sa_ok = (zplver >= ZPL_VERSION_SA);
3120 3119
3121 3120 /*
3122 3121 * Open parent object set so we can inherit zplprop values.
3123 3122 */
3124 3123 if ((error = dmu_objset_hold(parentname, FTAG, &os)) != 0)
3125 3124 return (error);
3126 3125
3127 3126 error = zfs_fill_zplprops_impl(os, zplver, fuids_ok, sa_ok, createprops,
3128 3127 zplprops, is_ci);
3129 3128 dmu_objset_rele(os, FTAG);
3130 3129 return (error);
3131 3130 }
3132 3131
3133 3132 static int
3134 3133 zfs_fill_zplprops_root(uint64_t spa_vers, nvlist_t *createprops,
3135 3134 nvlist_t *zplprops, boolean_t *is_ci)
3136 3135 {
3137 3136 boolean_t fuids_ok;
3138 3137 boolean_t sa_ok;
3139 3138 uint64_t zplver = ZPL_VERSION;
3140 3139 int error;
3141 3140
3142 3141 zplver = zfs_zpl_version_map(spa_vers);
3143 3142 fuids_ok = (zplver >= ZPL_VERSION_FUID);
3144 3143 sa_ok = (zplver >= ZPL_VERSION_SA);
3145 3144
3146 3145 error = zfs_fill_zplprops_impl(NULL, zplver, fuids_ok, sa_ok,
3147 3146 createprops, zplprops, is_ci);
3148 3147 return (error);
3149 3148 }
3150 3149
3151 3150 /*
3152 3151 * innvl: {
3153 3152 * "type" -> dmu_objset_type_t (int32)
3154 3153 * (optional) "props" -> { prop -> value }
3155 3154 * }
3156 3155 *
3157 3156 * outnvl: propname -> error code (int32)
3158 3157 */
3159 3158 static int
3160 3159 zfs_ioc_create(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3161 3160 {
3162 3161 int error = 0;
3163 3162 zfs_creat_t zct = { 0 };
3164 3163 nvlist_t *nvprops = NULL;
3165 3164 void (*cbfunc)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx);
3166 3165 int32_t type32;
3167 3166 dmu_objset_type_t type;
3168 3167 boolean_t is_insensitive = B_FALSE;
3169 3168
3170 3169 if (nvlist_lookup_int32(innvl, "type", &type32) != 0)
3171 3170 return (SET_ERROR(EINVAL));
3172 3171 type = type32;
3173 3172 (void) nvlist_lookup_nvlist(innvl, "props", &nvprops);
3174 3173
3175 3174 switch (type) {
3176 3175 case DMU_OST_ZFS:
3177 3176 cbfunc = zfs_create_cb;
3178 3177 break;
3179 3178
3180 3179 case DMU_OST_ZVOL:
3181 3180 cbfunc = zvol_create_cb;
3182 3181 break;
3183 3182
3184 3183 default:
3185 3184 cbfunc = NULL;
3186 3185 break;
3187 3186 }
3188 3187 if (strchr(fsname, '@') ||
3189 3188 strchr(fsname, '%'))
3190 3189 return (SET_ERROR(EINVAL));
3191 3190
3192 3191 zct.zct_props = nvprops;
3193 3192
3194 3193 if (cbfunc == NULL)
3195 3194 return (SET_ERROR(EINVAL));
3196 3195
3197 3196 if (type == DMU_OST_ZVOL) {
3198 3197 uint64_t volsize, volblocksize;
3199 3198
3200 3199 if (nvprops == NULL)
3201 3200 return (SET_ERROR(EINVAL));
3202 3201 if (nvlist_lookup_uint64(nvprops,
3203 3202 zfs_prop_to_name(ZFS_PROP_VOLSIZE), &volsize) != 0)
3204 3203 return (SET_ERROR(EINVAL));
3205 3204
3206 3205 if ((error = nvlist_lookup_uint64(nvprops,
3207 3206 zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE),
3208 3207 &volblocksize)) != 0 && error != ENOENT)
3209 3208 return (SET_ERROR(EINVAL));
3210 3209
3211 3210 if (error != 0)
3212 3211 volblocksize = zfs_prop_default_numeric(
3213 3212 ZFS_PROP_VOLBLOCKSIZE);
3214 3213
3215 3214 if ((error = zvol_check_volblocksize(
3216 3215 volblocksize)) != 0 ||
3217 3216 (error = zvol_check_volsize(volsize,
3218 3217 volblocksize)) != 0)
3219 3218 return (error);
3220 3219 } else if (type == DMU_OST_ZFS) {
3221 3220 int error;
3222 3221
3223 3222 /*
3224 3223 * We have to have normalization and
3225 3224 * case-folding flags correct when we do the
3226 3225 * file system creation, so go figure them out
3227 3226 * now.
3228 3227 */
3229 3228 VERIFY(nvlist_alloc(&zct.zct_zplprops,
3230 3229 NV_UNIQUE_NAME, KM_SLEEP) == 0);
3231 3230 error = zfs_fill_zplprops(fsname, nvprops,
3232 3231 zct.zct_zplprops, &is_insensitive);
3233 3232 if (error != 0) {
3234 3233 nvlist_free(zct.zct_zplprops);
3235 3234 return (error);
3236 3235 }
3237 3236 }
3238 3237
3239 3238 error = dmu_objset_create(fsname, type,
3240 3239 is_insensitive ? DS_FLAG_CI_DATASET : 0, cbfunc, &zct);
3241 3240 nvlist_free(zct.zct_zplprops);
3242 3241
3243 3242 /*
3244 3243 * It would be nice to do this atomically.
3245 3244 */
3246 3245 if (error == 0) {
3247 3246 error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL,
3248 3247 nvprops, outnvl);
3249 3248 if (error != 0)
3250 3249 (void) dsl_destroy_head(fsname);
3251 3250 }
3252 3251 return (error);
3253 3252 }
3254 3253
3255 3254 /*
3256 3255 * innvl: {
3257 3256 * "origin" -> name of origin snapshot
3258 3257 * (optional) "props" -> { prop -> value }
3259 3258 * }
3260 3259 *
3261 3260 * outnvl: propname -> error code (int32)
3262 3261 */
3263 3262 static int
3264 3263 zfs_ioc_clone(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3265 3264 {
3266 3265 int error = 0;
3267 3266 nvlist_t *nvprops = NULL;
3268 3267 char *origin_name;
3269 3268
3270 3269 if (nvlist_lookup_string(innvl, "origin", &origin_name) != 0)
3271 3270 return (SET_ERROR(EINVAL));
3272 3271 (void) nvlist_lookup_nvlist(innvl, "props", &nvprops);
3273 3272
3274 3273 if (strchr(fsname, '@') ||
3275 3274 strchr(fsname, '%'))
3276 3275 return (SET_ERROR(EINVAL));
3277 3276
3278 3277 if (dataset_namecheck(origin_name, NULL, NULL) != 0)
3279 3278 return (SET_ERROR(EINVAL));
3280 3279 error = dmu_objset_clone(fsname, origin_name);
3281 3280 if (error != 0)
3282 3281 return (error);
3283 3282
3284 3283 /*
3285 3284 * It would be nice to do this atomically.
3286 3285 */
3287 3286 if (error == 0) {
3288 3287 error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL,
3289 3288 nvprops, outnvl);
3290 3289 if (error != 0)
3291 3290 (void) dsl_destroy_head(fsname);
3292 3291 }
3293 3292 return (error);
3294 3293 }
3295 3294
3296 3295 /*
3297 3296 * innvl: {
3298 3297 * "snaps" -> { snapshot1, snapshot2 }
3299 3298 * (optional) "props" -> { prop -> value (string) }
3300 3299 * }
3301 3300 *
3302 3301 * outnvl: snapshot -> error code (int32)
3303 3302 */
3304 3303 static int
3305 3304 zfs_ioc_snapshot(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3306 3305 {
3307 3306 nvlist_t *snaps;
3308 3307 nvlist_t *props = NULL;
3309 3308 int error, poollen;
3310 3309 nvpair_t *pair;
3311 3310
3312 3311 (void) nvlist_lookup_nvlist(innvl, "props", &props);
3313 3312 if ((error = zfs_check_userprops(poolname, props)) != 0)
3314 3313 return (error);
3315 3314
3316 3315 if (!nvlist_empty(props) &&
3317 3316 zfs_earlier_version(poolname, SPA_VERSION_SNAP_PROPS))
3318 3317 return (SET_ERROR(ENOTSUP));
3319 3318
3320 3319 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
3321 3320 return (SET_ERROR(EINVAL));
3322 3321 poollen = strlen(poolname);
3323 3322 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
3324 3323 pair = nvlist_next_nvpair(snaps, pair)) {
3325 3324 const char *name = nvpair_name(pair);
3326 3325 const char *cp = strchr(name, '@');
3327 3326
3328 3327 /*
3329 3328 * The snap name must contain an @, and the part after it must
3330 3329 * contain only valid characters.
3331 3330 */
3332 3331 if (cp == NULL ||
3333 3332 zfs_component_namecheck(cp + 1, NULL, NULL) != 0)
3334 3333 return (SET_ERROR(EINVAL));
3335 3334
3336 3335 /*
3337 3336 * The snap must be in the specified pool.
3338 3337 */
3339 3338 if (strncmp(name, poolname, poollen) != 0 ||
3340 3339 (name[poollen] != '/' && name[poollen] != '@'))
3341 3340 return (SET_ERROR(EXDEV));
3342 3341
3343 3342 /* This must be the only snap of this fs. */
3344 3343 for (nvpair_t *pair2 = nvlist_next_nvpair(snaps, pair);
3345 3344 pair2 != NULL; pair2 = nvlist_next_nvpair(snaps, pair2)) {
3346 3345 if (strncmp(name, nvpair_name(pair2), cp - name + 1)
3347 3346 == 0) {
3348 3347 return (SET_ERROR(EXDEV));
3349 3348 }
3350 3349 }
3351 3350 }
3352 3351
3353 3352 error = dsl_dataset_snapshot(snaps, props, outnvl);
3354 3353 return (error);
3355 3354 }
3356 3355
3357 3356 /*
3358 3357 * innvl: "message" -> string
3359 3358 */
3360 3359 /* ARGSUSED */
3361 3360 static int
3362 3361 zfs_ioc_log_history(const char *unused, nvlist_t *innvl, nvlist_t *outnvl)
3363 3362 {
3364 3363 char *message;
3365 3364 spa_t *spa;
3366 3365 int error;
3367 3366 char *poolname;
3368 3367
3369 3368 /*
3370 3369 * The poolname in the ioctl is not set, we get it from the TSD,
3371 3370 * which was set at the end of the last successful ioctl that allows
3372 3371 * logging. The secpolicy func already checked that it is set.
3373 3372 * Only one log ioctl is allowed after each successful ioctl, so
3374 3373 * we clear the TSD here.
3375 3374 */
3376 3375 poolname = tsd_get(zfs_allow_log_key);
3377 3376 (void) tsd_set(zfs_allow_log_key, NULL);
3378 3377 error = spa_open(poolname, &spa, FTAG);
3379 3378 strfree(poolname);
3380 3379 if (error != 0)
3381 3380 return (error);
3382 3381
3383 3382 if (nvlist_lookup_string(innvl, "message", &message) != 0) {
3384 3383 spa_close(spa, FTAG);
3385 3384 return (SET_ERROR(EINVAL));
3386 3385 }
3387 3386
3388 3387 if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) {
3389 3388 spa_close(spa, FTAG);
3390 3389 return (SET_ERROR(ENOTSUP));
3391 3390 }
3392 3391
3393 3392 error = spa_history_log(spa, message);
3394 3393 spa_close(spa, FTAG);
3395 3394 return (error);
3396 3395 }
3397 3396
3398 3397 /*
3399 3398 * The dp_config_rwlock must not be held when calling this, because the
3400 3399 * unmount may need to write out data.
3401 3400 *
3402 3401 * This function is best-effort. Callers must deal gracefully if it
3403 3402 * remains mounted (or is remounted after this call).
3404 3403 *
3405 3404 * Returns 0 if the argument is not a snapshot, or it is not currently a
3406 3405 * filesystem, or we were able to unmount it. Returns error code otherwise.
3407 3406 */
3408 3407 int
3409 3408 zfs_unmount_snap(const char *snapname)
3410 3409 {
3411 3410 vfs_t *vfsp;
3412 3411 zfsvfs_t *zfsvfs;
3413 3412 int err;
3414 3413
3415 3414 if (strchr(snapname, '@') == NULL)
3416 3415 return (0);
3417 3416
3418 3417 vfsp = zfs_get_vfs(snapname);
3419 3418 if (vfsp == NULL)
3420 3419 return (0);
3421 3420
3422 3421 zfsvfs = vfsp->vfs_data;
3423 3422 ASSERT(!dsl_pool_config_held(dmu_objset_pool(zfsvfs->z_os)));
3424 3423
3425 3424 err = vn_vfswlock(vfsp->vfs_vnodecovered);
3426 3425 VFS_RELE(vfsp);
3427 3426 if (err != 0)
3428 3427 return (SET_ERROR(err));
3429 3428
3430 3429 /*
3431 3430 * Always force the unmount for snapshots.
3432 3431 */
3433 3432 (void) dounmount(vfsp, MS_FORCE, kcred);
3434 3433 return (0);
3435 3434 }
3436 3435
3437 3436 /* ARGSUSED */
3438 3437 static int
3439 3438 zfs_unmount_snap_cb(const char *snapname, void *arg)
3440 3439 {
3441 3440 return (zfs_unmount_snap(snapname));
3442 3441 }
3443 3442
3444 3443 /*
3445 3444 * When a clone is destroyed, its origin may also need to be destroyed,
3446 3445 * in which case it must be unmounted. This routine will do that unmount
3447 3446 * if necessary.
3448 3447 */
3449 3448 void
3450 3449 zfs_destroy_unmount_origin(const char *fsname)
3451 3450 {
3452 3451 int error;
3453 3452 objset_t *os;
3454 3453 dsl_dataset_t *ds;
3455 3454
3456 3455 error = dmu_objset_hold(fsname, FTAG, &os);
3457 3456 if (error != 0)
3458 3457 return;
3459 3458 ds = dmu_objset_ds(os);
3460 3459 if (dsl_dir_is_clone(ds->ds_dir) && DS_IS_DEFER_DESTROY(ds->ds_prev)) {
3461 3460 char originname[MAXNAMELEN];
3462 3461 dsl_dataset_name(ds->ds_prev, originname);
3463 3462 dmu_objset_rele(os, FTAG);
3464 3463 (void) zfs_unmount_snap(originname);
3465 3464 } else {
3466 3465 dmu_objset_rele(os, FTAG);
3467 3466 }
3468 3467 }
3469 3468
3470 3469 /*
3471 3470 * innvl: {
3472 3471 * "snaps" -> { snapshot1, snapshot2 }
3473 3472 * (optional boolean) "defer"
3474 3473 * }
3475 3474 *
3476 3475 * outnvl: snapshot -> error code (int32)
3477 3476 *
3478 3477 */
3479 3478 /* ARGSUSED */
3480 3479 static int
3481 3480 zfs_ioc_destroy_snaps(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3482 3481 {
3483 3482 nvlist_t *snaps;
3484 3483 nvpair_t *pair;
3485 3484 boolean_t defer;
3486 3485
3487 3486 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
3488 3487 return (SET_ERROR(EINVAL));
3489 3488 defer = nvlist_exists(innvl, "defer");
3490 3489
3491 3490 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
3492 3491 pair = nvlist_next_nvpair(snaps, pair)) {
3493 3492 (void) zfs_unmount_snap(nvpair_name(pair));
3494 3493 }
3495 3494
3496 3495 return (dsl_destroy_snapshots_nvl(snaps, defer, outnvl));
3497 3496 }
3498 3497
3499 3498 /*
3500 3499 * Create bookmarks. Bookmark names are of the form <fs>#<bmark>.
3501 3500 * All bookmarks must be in the same pool.
3502 3501 *
3503 3502 * innvl: {
3504 3503 * bookmark1 -> snapshot1, bookmark2 -> snapshot2
3505 3504 * }
3506 3505 *
3507 3506 * outnvl: bookmark -> error code (int32)
3508 3507 *
3509 3508 */
3510 3509 /* ARGSUSED */
3511 3510 static int
3512 3511 zfs_ioc_bookmark(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3513 3512 {
3514 3513 for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL);
3515 3514 pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) {
3516 3515 char *snap_name;
3517 3516
3518 3517 /*
3519 3518 * Verify the snapshot argument.
3520 3519 */
3521 3520 if (nvpair_value_string(pair, &snap_name) != 0)
3522 3521 return (SET_ERROR(EINVAL));
3523 3522
3524 3523
3525 3524 /* Verify that the keys (bookmarks) are unique */
3526 3525 for (nvpair_t *pair2 = nvlist_next_nvpair(innvl, pair);
3527 3526 pair2 != NULL; pair2 = nvlist_next_nvpair(innvl, pair2)) {
3528 3527 if (strcmp(nvpair_name(pair), nvpair_name(pair2)) == 0)
3529 3528 return (SET_ERROR(EINVAL));
3530 3529 }
3531 3530 }
3532 3531
3533 3532 return (dsl_bookmark_create(innvl, outnvl));
3534 3533 }
3535 3534
3536 3535 /*
3537 3536 * innvl: {
3538 3537 * property 1, property 2, ...
3539 3538 * }
3540 3539 *
3541 3540 * outnvl: {
3542 3541 * bookmark name 1 -> { property 1, property 2, ... },
3543 3542 * bookmark name 2 -> { property 1, property 2, ... }
3544 3543 * }
3545 3544 *
3546 3545 */
3547 3546 static int
3548 3547 zfs_ioc_get_bookmarks(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3549 3548 {
3550 3549 return (dsl_get_bookmarks(fsname, innvl, outnvl));
3551 3550 }
3552 3551
3553 3552 /*
3554 3553 * innvl: {
3555 3554 * bookmark name 1, bookmark name 2
3556 3555 * }
3557 3556 *
3558 3557 * outnvl: bookmark -> error code (int32)
3559 3558 *
3560 3559 */
3561 3560 static int
3562 3561 zfs_ioc_destroy_bookmarks(const char *poolname, nvlist_t *innvl,
3563 3562 nvlist_t *outnvl)
3564 3563 {
3565 3564 int error, poollen;
3566 3565
3567 3566 poollen = strlen(poolname);
3568 3567 for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL);
3569 3568 pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) {
3570 3569 const char *name = nvpair_name(pair);
3571 3570 const char *cp = strchr(name, '#');
3572 3571
3573 3572 /*
3574 3573 * The bookmark name must contain an #, and the part after it
3575 3574 * must contain only valid characters.
3576 3575 */
3577 3576 if (cp == NULL ||
3578 3577 zfs_component_namecheck(cp + 1, NULL, NULL) != 0)
3579 3578 return (SET_ERROR(EINVAL));
3580 3579
3581 3580 /*
3582 3581 * The bookmark must be in the specified pool.
3583 3582 */
3584 3583 if (strncmp(name, poolname, poollen) != 0 ||
3585 3584 (name[poollen] != '/' && name[poollen] != '#'))
3586 3585 return (SET_ERROR(EXDEV));
3587 3586 }
3588 3587
3589 3588 error = dsl_bookmark_destroy(innvl, outnvl);
3590 3589 return (error);
3591 3590 }
3592 3591
3593 3592 /*
3594 3593 * inputs:
3595 3594 * zc_name name of dataset to destroy
3596 3595 * zc_objset_type type of objset
3597 3596 * zc_defer_destroy mark for deferred destroy
3598 3597 *
3599 3598 * outputs: none
3600 3599 */
3601 3600 static int
3602 3601 zfs_ioc_destroy(zfs_cmd_t *zc)
3603 3602 {
3604 3603 int err;
3605 3604
3606 3605 if (zc->zc_objset_type == DMU_OST_ZFS) {
3607 3606 err = zfs_unmount_snap(zc->zc_name);
3608 3607 if (err != 0)
3609 3608 return (err);
3610 3609 }
3611 3610
3612 3611 if (strchr(zc->zc_name, '@'))
3613 3612 err = dsl_destroy_snapshot(zc->zc_name, zc->zc_defer_destroy);
3614 3613 else
3615 3614 err = dsl_destroy_head(zc->zc_name);
3616 3615 if (zc->zc_objset_type == DMU_OST_ZVOL && err == 0)
3617 3616 (void) zvol_remove_minor(zc->zc_name);
3618 3617 return (err);
3619 3618 }
3620 3619
3621 3620 /*
3622 3621 * fsname is name of dataset to rollback (to most recent snapshot)
3623 3622 *
3624 3623 * innvl is not used.
3625 3624 *
3626 3625 * outnvl: "target" -> name of most recent snapshot
3627 3626 * }
3628 3627 */
3629 3628 /* ARGSUSED */
3630 3629 static int
3631 3630 zfs_ioc_rollback(const char *fsname, nvlist_t *args, nvlist_t *outnvl)
3632 3631 {
3633 3632 zfsvfs_t *zfsvfs;
3634 3633 int error;
3635 3634
3636 3635 if (getzfsvfs(fsname, &zfsvfs) == 0) {
3637 3636 error = zfs_suspend_fs(zfsvfs);
3638 3637 if (error == 0) {
3639 3638 int resume_err;
3640 3639
3641 3640 error = dsl_dataset_rollback(fsname, zfsvfs, outnvl);
3642 3641 resume_err = zfs_resume_fs(zfsvfs, fsname);
3643 3642 error = error ? error : resume_err;
3644 3643 }
3645 3644 VFS_RELE(zfsvfs->z_vfs);
3646 3645 } else {
3647 3646 error = dsl_dataset_rollback(fsname, NULL, outnvl);
3648 3647 }
3649 3648 return (error);
3650 3649 }
3651 3650
3652 3651 static int
3653 3652 recursive_unmount(const char *fsname, void *arg)
3654 3653 {
3655 3654 const char *snapname = arg;
3656 3655 char fullname[MAXNAMELEN];
3657 3656
3658 3657 (void) snprintf(fullname, sizeof (fullname), "%s@%s", fsname, snapname);
3659 3658 return (zfs_unmount_snap(fullname));
3660 3659 }
3661 3660
3662 3661 /*
3663 3662 * inputs:
3664 3663 * zc_name old name of dataset
3665 3664 * zc_value new name of dataset
3666 3665 * zc_cookie recursive flag (only valid for snapshots)
3667 3666 *
3668 3667 * outputs: none
3669 3668 */
3670 3669 static int
3671 3670 zfs_ioc_rename(zfs_cmd_t *zc)
3672 3671 {
3673 3672 boolean_t recursive = zc->zc_cookie & 1;
3674 3673 char *at;
3675 3674
3676 3675 zc->zc_value[sizeof (zc->zc_value) - 1] = '\0';
3677 3676 if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
3678 3677 strchr(zc->zc_value, '%'))
3679 3678 return (SET_ERROR(EINVAL));
3680 3679
3681 3680 at = strchr(zc->zc_name, '@');
3682 3681 if (at != NULL) {
3683 3682 /* snaps must be in same fs */
3684 3683 int error;
3685 3684
3686 3685 if (strncmp(zc->zc_name, zc->zc_value, at - zc->zc_name + 1))
3687 3686 return (SET_ERROR(EXDEV));
3688 3687 *at = '\0';
3689 3688 if (zc->zc_objset_type == DMU_OST_ZFS) {
3690 3689 error = dmu_objset_find(zc->zc_name,
3691 3690 recursive_unmount, at + 1,
3692 3691 recursive ? DS_FIND_CHILDREN : 0);
3693 3692 if (error != 0) {
3694 3693 *at = '@';
3695 3694 return (error);
3696 3695 }
3697 3696 }
3698 3697 error = dsl_dataset_rename_snapshot(zc->zc_name,
3699 3698 at + 1, strchr(zc->zc_value, '@') + 1, recursive);
3700 3699 *at = '@';
3701 3700
3702 3701 return (error);
3703 3702 } else {
3704 3703 if (zc->zc_objset_type == DMU_OST_ZVOL)
3705 3704 (void) zvol_remove_minor(zc->zc_name);
3706 3705 return (dsl_dir_rename(zc->zc_name, zc->zc_value));
3707 3706 }
3708 3707 }
3709 3708
3710 3709 static int
3711 3710 zfs_check_settable(const char *dsname, nvpair_t *pair, cred_t *cr)
3712 3711 {
3713 3712 const char *propname = nvpair_name(pair);
3714 3713 boolean_t issnap = (strchr(dsname, '@') != NULL);
3715 3714 zfs_prop_t prop = zfs_name_to_prop(propname);
3716 3715 uint64_t intval;
3717 3716 int err;
3718 3717
3719 3718 if (prop == ZPROP_INVAL) {
3720 3719 if (zfs_prop_user(propname)) {
3721 3720 if (err = zfs_secpolicy_write_perms(dsname,
3722 3721 ZFS_DELEG_PERM_USERPROP, cr))
3723 3722 return (err);
3724 3723 return (0);
3725 3724 }
3726 3725
3727 3726 if (!issnap && zfs_prop_userquota(propname)) {
3728 3727 const char *perm = NULL;
3729 3728 const char *uq_prefix =
3730 3729 zfs_userquota_prop_prefixes[ZFS_PROP_USERQUOTA];
3731 3730 const char *gq_prefix =
3732 3731 zfs_userquota_prop_prefixes[ZFS_PROP_GROUPQUOTA];
3733 3732
3734 3733 if (strncmp(propname, uq_prefix,
3735 3734 strlen(uq_prefix)) == 0) {
3736 3735 perm = ZFS_DELEG_PERM_USERQUOTA;
3737 3736 } else if (strncmp(propname, gq_prefix,
3738 3737 strlen(gq_prefix)) == 0) {
3739 3738 perm = ZFS_DELEG_PERM_GROUPQUOTA;
3740 3739 } else {
3741 3740 /* USERUSED and GROUPUSED are read-only */
3742 3741 return (SET_ERROR(EINVAL));
3743 3742 }
3744 3743
3745 3744 if (err = zfs_secpolicy_write_perms(dsname, perm, cr))
3746 3745 return (err);
3747 3746 return (0);
3748 3747 }
3749 3748
3750 3749 return (SET_ERROR(EINVAL));
3751 3750 }
3752 3751
3753 3752 if (issnap)
3754 3753 return (SET_ERROR(EINVAL));
3755 3754
3756 3755 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
3757 3756 /*
3758 3757 * dsl_prop_get_all_impl() returns properties in this
3759 3758 * format.
3760 3759 */
3761 3760 nvlist_t *attrs;
3762 3761 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
3763 3762 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
3764 3763 &pair) == 0);
3765 3764 }
3766 3765
3767 3766 /*
3768 3767 * Check that this value is valid for this pool version
3769 3768 */
3770 3769 switch (prop) {
3771 3770 case ZFS_PROP_COMPRESSION:
3772 3771 /*
3773 3772 * If the user specified gzip compression, make sure
3774 3773 * the SPA supports it. We ignore any errors here since
3775 3774 * we'll catch them later.
3776 3775 */
3777 3776 if (nvpair_value_uint64(pair, &intval) == 0) {
3778 3777 if (intval >= ZIO_COMPRESS_GZIP_1 &&
3779 3778 intval <= ZIO_COMPRESS_GZIP_9 &&
3780 3779 zfs_earlier_version(dsname,
3781 3780 SPA_VERSION_GZIP_COMPRESSION)) {
3782 3781 return (SET_ERROR(ENOTSUP));
3783 3782 }
3784 3783
3785 3784 if (intval == ZIO_COMPRESS_ZLE &&
3786 3785 zfs_earlier_version(dsname,
3787 3786 SPA_VERSION_ZLE_COMPRESSION))
3788 3787 return (SET_ERROR(ENOTSUP));
3789 3788
3790 3789 if (intval == ZIO_COMPRESS_LZ4) {
3791 3790 spa_t *spa;
3792 3791
3793 3792 if ((err = spa_open(dsname, &spa, FTAG)) != 0)
3794 3793 return (err);
3795 3794
3796 3795 if (!spa_feature_is_enabled(spa,
3797 3796 SPA_FEATURE_LZ4_COMPRESS)) {
3798 3797 spa_close(spa, FTAG);
3799 3798 return (SET_ERROR(ENOTSUP));
3800 3799 }
3801 3800 spa_close(spa, FTAG);
3802 3801 }
3803 3802
3804 3803 /*
3805 3804 * If this is a bootable dataset then
3806 3805 * verify that the compression algorithm
3807 3806 * is supported for booting. We must return
3808 3807 * something other than ENOTSUP since it
3809 3808 * implies a downrev pool version.
3810 3809 */
3811 3810 if (zfs_is_bootfs(dsname) &&
3812 3811 !BOOTFS_COMPRESS_VALID(intval)) {
3813 3812 return (SET_ERROR(ERANGE));
3814 3813 }
3815 3814 }
3816 3815 break;
3817 3816
3818 3817 case ZFS_PROP_COPIES:
3819 3818 if (zfs_earlier_version(dsname, SPA_VERSION_DITTO_BLOCKS))
3820 3819 return (SET_ERROR(ENOTSUP));
3821 3820 break;
3822 3821
3823 3822 case ZFS_PROP_RECORDSIZE:
3824 3823 /* Record sizes above 128k need the feature to be enabled */
3825 3824 if (nvpair_value_uint64(pair, &intval) == 0 &&
3826 3825 intval > SPA_OLD_MAXBLOCKSIZE) {
3827 3826 spa_t *spa;
3828 3827
3829 3828 /*
3830 3829 * If this is a bootable dataset then
3831 3830 * the we don't allow large (>128K) blocks,
3832 3831 * because GRUB doesn't support them.
3833 3832 */
3834 3833 if (zfs_is_bootfs(dsname) &&
3835 3834 intval > SPA_OLD_MAXBLOCKSIZE) {
3836 3835 return (SET_ERROR(ERANGE));
3837 3836 }
3838 3837
3839 3838 /*
3840 3839 * We don't allow setting the property above 1MB,
3841 3840 * unless the tunable has been changed.
3842 3841 */
3843 3842 if (intval > zfs_max_recordsize ||
3844 3843 intval > SPA_MAXBLOCKSIZE)
3845 3844 return (SET_ERROR(ERANGE));
3846 3845
3847 3846 if ((err = spa_open(dsname, &spa, FTAG)) != 0)
3848 3847 return (err);
3849 3848
3850 3849 if (!spa_feature_is_enabled(spa,
3851 3850 SPA_FEATURE_LARGE_BLOCKS)) {
3852 3851 spa_close(spa, FTAG);
3853 3852 return (SET_ERROR(ENOTSUP));
3854 3853 }
3855 3854 spa_close(spa, FTAG);
3856 3855 }
3857 3856 break;
3858 3857
3859 3858 case ZFS_PROP_SHARESMB:
3860 3859 if (zpl_earlier_version(dsname, ZPL_VERSION_FUID))
3861 3860 return (SET_ERROR(ENOTSUP));
3862 3861 break;
3863 3862
3864 3863 case ZFS_PROP_ACLINHERIT:
3865 3864 if (nvpair_type(pair) == DATA_TYPE_UINT64 &&
3866 3865 nvpair_value_uint64(pair, &intval) == 0) {
3867 3866 if (intval == ZFS_ACL_PASSTHROUGH_X &&
3868 3867 zfs_earlier_version(dsname,
3869 3868 SPA_VERSION_PASSTHROUGH_X))
3870 3869 return (SET_ERROR(ENOTSUP));
3871 3870 }
3872 3871 break;
3873 3872
3874 3873 case ZFS_PROP_CHECKSUM:
3875 3874 case ZFS_PROP_DEDUP:
3876 3875 {
3877 3876 spa_feature_t feature;
3878 3877 spa_t *spa;
3879 3878
3880 3879 /* dedup feature version checks */
3881 3880 if (prop == ZFS_PROP_DEDUP &&
3882 3881 zfs_earlier_version(dsname, SPA_VERSION_DEDUP))
3883 3882 return (SET_ERROR(ENOTSUP));
3884 3883
3885 3884 if (nvpair_value_uint64(pair, &intval) != 0)
3886 3885 return (SET_ERROR(EINVAL));
3887 3886
3888 3887 /* check prop value is enabled in features */
3889 3888 feature = zio_checksum_to_feature(intval & ZIO_CHECKSUM_MASK);
3890 3889 if (feature == SPA_FEATURE_NONE)
3891 3890 break;
3892 3891
3893 3892 if ((err = spa_open(dsname, &spa, FTAG)) != 0)
3894 3893 return (err);
3895 3894 /*
3896 3895 * Salted checksums are not supported on root pools.
3897 3896 */
3898 3897 if (spa_bootfs(spa) != 0 &&
3899 3898 intval < ZIO_CHECKSUM_FUNCTIONS &&
3900 3899 (zio_checksum_table[intval].ci_flags &
3901 3900 ZCHECKSUM_FLAG_SALTED)) {
3902 3901 spa_close(spa, FTAG);
3903 3902 return (SET_ERROR(ERANGE));
3904 3903 }
3905 3904 if (!spa_feature_is_enabled(spa, feature)) {
3906 3905 spa_close(spa, FTAG);
3907 3906 return (SET_ERROR(ENOTSUP));
3908 3907 }
3909 3908 spa_close(spa, FTAG);
3910 3909 break;
3911 3910 }
3912 3911 }
3913 3912
3914 3913 return (zfs_secpolicy_setprop(dsname, prop, pair, CRED()));
3915 3914 }
3916 3915
3917 3916 /*
3918 3917 * Checks for a race condition to make sure we don't increment a feature flag
3919 3918 * multiple times.
3920 3919 */
3921 3920 static int
3922 3921 zfs_prop_activate_feature_check(void *arg, dmu_tx_t *tx)
3923 3922 {
3924 3923 spa_t *spa = dmu_tx_pool(tx)->dp_spa;
3925 3924 spa_feature_t *featurep = arg;
3926 3925
3927 3926 if (!spa_feature_is_active(spa, *featurep))
3928 3927 return (0);
3929 3928 else
3930 3929 return (SET_ERROR(EBUSY));
3931 3930 }
3932 3931
3933 3932 /*
3934 3933 * The callback invoked on feature activation in the sync task caused by
3935 3934 * zfs_prop_activate_feature.
3936 3935 */
3937 3936 static void
3938 3937 zfs_prop_activate_feature_sync(void *arg, dmu_tx_t *tx)
3939 3938 {
3940 3939 spa_t *spa = dmu_tx_pool(tx)->dp_spa;
3941 3940 spa_feature_t *featurep = arg;
3942 3941
3943 3942 spa_feature_incr(spa, *featurep, tx);
3944 3943 }
3945 3944
3946 3945 /*
3947 3946 * Activates a feature on a pool in response to a property setting. This
3948 3947 * creates a new sync task which modifies the pool to reflect the feature
3949 3948 * as being active.
3950 3949 */
3951 3950 static int
3952 3951 zfs_prop_activate_feature(spa_t *spa, spa_feature_t feature)
3953 3952 {
3954 3953 int err;
3955 3954
3956 3955 /* EBUSY here indicates that the feature is already active */
3957 3956 err = dsl_sync_task(spa_name(spa),
3958 3957 zfs_prop_activate_feature_check, zfs_prop_activate_feature_sync,
3959 3958 &feature, 2, ZFS_SPACE_CHECK_RESERVED);
3960 3959
3961 3960 if (err != 0 && err != EBUSY)
3962 3961 return (err);
3963 3962 else
3964 3963 return (0);
3965 3964 }
3966 3965
3967 3966 /*
3968 3967 * Removes properties from the given props list that fail permission checks
3969 3968 * needed to clear them and to restore them in case of a receive error. For each
3970 3969 * property, make sure we have both set and inherit permissions.
3971 3970 *
3972 3971 * Returns the first error encountered if any permission checks fail. If the
3973 3972 * caller provides a non-NULL errlist, it also gives the complete list of names
3974 3973 * of all the properties that failed a permission check along with the
3975 3974 * corresponding error numbers. The caller is responsible for freeing the
3976 3975 * returned errlist.
3977 3976 *
3978 3977 * If every property checks out successfully, zero is returned and the list
3979 3978 * pointed at by errlist is NULL.
3980 3979 */
3981 3980 static int
3982 3981 zfs_check_clearable(char *dataset, nvlist_t *props, nvlist_t **errlist)
3983 3982 {
3984 3983 zfs_cmd_t *zc;
3985 3984 nvpair_t *pair, *next_pair;
3986 3985 nvlist_t *errors;
3987 3986 int err, rv = 0;
3988 3987
3989 3988 if (props == NULL)
3990 3989 return (0);
3991 3990
3992 3991 VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0);
3993 3992
3994 3993 zc = kmem_alloc(sizeof (zfs_cmd_t), KM_SLEEP);
3995 3994 (void) strcpy(zc->zc_name, dataset);
3996 3995 pair = nvlist_next_nvpair(props, NULL);
3997 3996 while (pair != NULL) {
3998 3997 next_pair = nvlist_next_nvpair(props, pair);
3999 3998
4000 3999 (void) strcpy(zc->zc_value, nvpair_name(pair));
4001 4000 if ((err = zfs_check_settable(dataset, pair, CRED())) != 0 ||
4002 4001 (err = zfs_secpolicy_inherit_prop(zc, NULL, CRED())) != 0) {
4003 4002 VERIFY(nvlist_remove_nvpair(props, pair) == 0);
4004 4003 VERIFY(nvlist_add_int32(errors,
4005 4004 zc->zc_value, err) == 0);
4006 4005 }
4007 4006 pair = next_pair;
4008 4007 }
4009 4008 kmem_free(zc, sizeof (zfs_cmd_t));
4010 4009
4011 4010 if ((pair = nvlist_next_nvpair(errors, NULL)) == NULL) {
4012 4011 nvlist_free(errors);
4013 4012 errors = NULL;
4014 4013 } else {
4015 4014 VERIFY(nvpair_value_int32(pair, &rv) == 0);
4016 4015 }
4017 4016
4018 4017 if (errlist == NULL)
4019 4018 nvlist_free(errors);
4020 4019 else
4021 4020 *errlist = errors;
4022 4021
4023 4022 return (rv);
4024 4023 }
4025 4024
4026 4025 static boolean_t
4027 4026 propval_equals(nvpair_t *p1, nvpair_t *p2)
4028 4027 {
4029 4028 if (nvpair_type(p1) == DATA_TYPE_NVLIST) {
4030 4029 /* dsl_prop_get_all_impl() format */
4031 4030 nvlist_t *attrs;
4032 4031 VERIFY(nvpair_value_nvlist(p1, &attrs) == 0);
4033 4032 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
4034 4033 &p1) == 0);
4035 4034 }
4036 4035
4037 4036 if (nvpair_type(p2) == DATA_TYPE_NVLIST) {
4038 4037 nvlist_t *attrs;
4039 4038 VERIFY(nvpair_value_nvlist(p2, &attrs) == 0);
4040 4039 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
4041 4040 &p2) == 0);
4042 4041 }
4043 4042
4044 4043 if (nvpair_type(p1) != nvpair_type(p2))
4045 4044 return (B_FALSE);
4046 4045
4047 4046 if (nvpair_type(p1) == DATA_TYPE_STRING) {
4048 4047 char *valstr1, *valstr2;
4049 4048
4050 4049 VERIFY(nvpair_value_string(p1, (char **)&valstr1) == 0);
4051 4050 VERIFY(nvpair_value_string(p2, (char **)&valstr2) == 0);
4052 4051 return (strcmp(valstr1, valstr2) == 0);
4053 4052 } else {
4054 4053 uint64_t intval1, intval2;
4055 4054
4056 4055 VERIFY(nvpair_value_uint64(p1, &intval1) == 0);
4057 4056 VERIFY(nvpair_value_uint64(p2, &intval2) == 0);
4058 4057 return (intval1 == intval2);
4059 4058 }
4060 4059 }
4061 4060
4062 4061 /*
4063 4062 * Remove properties from props if they are not going to change (as determined
4064 4063 * by comparison with origprops). Remove them from origprops as well, since we
4065 4064 * do not need to clear or restore properties that won't change.
4066 4065 */
4067 4066 static void
4068 4067 props_reduce(nvlist_t *props, nvlist_t *origprops)
4069 4068 {
4070 4069 nvpair_t *pair, *next_pair;
4071 4070
4072 4071 if (origprops == NULL)
4073 4072 return; /* all props need to be received */
4074 4073
4075 4074 pair = nvlist_next_nvpair(props, NULL);
4076 4075 while (pair != NULL) {
4077 4076 const char *propname = nvpair_name(pair);
4078 4077 nvpair_t *match;
4079 4078
4080 4079 next_pair = nvlist_next_nvpair(props, pair);
4081 4080
4082 4081 if ((nvlist_lookup_nvpair(origprops, propname,
4083 4082 &match) != 0) || !propval_equals(pair, match))
4084 4083 goto next; /* need to set received value */
4085 4084
4086 4085 /* don't clear the existing received value */
4087 4086 (void) nvlist_remove_nvpair(origprops, match);
4088 4087 /* don't bother receiving the property */
4089 4088 (void) nvlist_remove_nvpair(props, pair);
4090 4089 next:
4091 4090 pair = next_pair;
4092 4091 }
4093 4092 }
4094 4093
4095 4094 /*
4096 4095 * Extract properties that cannot be set PRIOR to the receipt of a dataset.
4097 4096 * For example, refquota cannot be set until after the receipt of a dataset,
4098 4097 * because in replication streams, an older/earlier snapshot may exceed the
4099 4098 * refquota. We want to receive the older/earlier snapshot, but setting
4100 4099 * refquota pre-receipt will set the dsl's ACTUAL quota, which will prevent
4101 4100 * the older/earlier snapshot from being received (with EDQUOT).
4102 4101 *
4103 4102 * The ZFS test "zfs_receive_011_pos" demonstrates such a scenario.
4104 4103 *
4105 4104 * libzfs will need to be judicious handling errors encountered by props
4106 4105 * extracted by this function.
4107 4106 */
4108 4107 static nvlist_t *
4109 4108 extract_delay_props(nvlist_t *props)
4110 4109 {
4111 4110 nvlist_t *delayprops;
4112 4111 nvpair_t *nvp, *tmp;
4113 4112 static const zfs_prop_t delayable[] = { ZFS_PROP_REFQUOTA, 0 };
4114 4113 int i;
4115 4114
4116 4115 VERIFY(nvlist_alloc(&delayprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
4117 4116
4118 4117 for (nvp = nvlist_next_nvpair(props, NULL); nvp != NULL;
4119 4118 nvp = nvlist_next_nvpair(props, nvp)) {
4120 4119 /*
4121 4120 * strcmp() is safe because zfs_prop_to_name() always returns
4122 4121 * a bounded string.
4123 4122 */
4124 4123 for (i = 0; delayable[i] != 0; i++) {
4125 4124 if (strcmp(zfs_prop_to_name(delayable[i]),
4126 4125 nvpair_name(nvp)) == 0) {
4127 4126 break;
4128 4127 }
4129 4128 }
4130 4129 if (delayable[i] != 0) {
4131 4130 tmp = nvlist_prev_nvpair(props, nvp);
4132 4131 VERIFY(nvlist_add_nvpair(delayprops, nvp) == 0);
4133 4132 VERIFY(nvlist_remove_nvpair(props, nvp) == 0);
4134 4133 nvp = tmp;
4135 4134 }
4136 4135 }
4137 4136
4138 4137 if (nvlist_empty(delayprops)) {
4139 4138 nvlist_free(delayprops);
4140 4139 delayprops = NULL;
4141 4140 }
4142 4141 return (delayprops);
4143 4142 }
4144 4143
4145 4144 #ifdef DEBUG
4146 4145 static boolean_t zfs_ioc_recv_inject_err;
4147 4146 #endif
4148 4147
4149 4148 /*
4150 4149 * inputs:
4151 4150 * zc_name name of containing filesystem
4152 4151 * zc_nvlist_src{_size} nvlist of properties to apply
4153 4152 * zc_value name of snapshot to create
4154 4153 * zc_string name of clone origin (if DRR_FLAG_CLONE)
4155 4154 * zc_cookie file descriptor to recv from
4156 4155 * zc_begin_record the BEGIN record of the stream (not byteswapped)
4157 4156 * zc_guid force flag
4158 4157 * zc_cleanup_fd cleanup-on-exit file descriptor
4159 4158 * zc_action_handle handle for this guid/ds mapping (or zero on first call)
4160 4159 * zc_resumable if data is incomplete assume sender will resume
4161 4160 *
4162 4161 * outputs:
4163 4162 * zc_cookie number of bytes read
4164 4163 * zc_nvlist_dst{_size} error for each unapplied received property
4165 4164 * zc_obj zprop_errflags_t
4166 4165 * zc_action_handle handle for this guid/ds mapping
4167 4166 */
4168 4167 static int
4169 4168 zfs_ioc_recv(zfs_cmd_t *zc)
4170 4169 {
4171 4170 file_t *fp;
4172 4171 dmu_recv_cookie_t drc;
4173 4172 boolean_t force = (boolean_t)zc->zc_guid;
4174 4173 int fd;
4175 4174 int error = 0;
4176 4175 int props_error = 0;
4177 4176 nvlist_t *errors;
4178 4177 offset_t off;
4179 4178 nvlist_t *props = NULL; /* sent properties */
4180 4179 nvlist_t *origprops = NULL; /* existing properties */
4181 4180 nvlist_t *delayprops = NULL; /* sent properties applied post-receive */
4182 4181 char *origin = NULL;
4183 4182 char *tosnap;
4184 4183 char tofs[ZFS_MAXNAMELEN];
4185 4184 boolean_t first_recvd_props = B_FALSE;
4186 4185
4187 4186 if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
4188 4187 strchr(zc->zc_value, '@') == NULL ||
4189 4188 strchr(zc->zc_value, '%'))
4190 4189 return (SET_ERROR(EINVAL));
4191 4190
4192 4191 (void) strcpy(tofs, zc->zc_value);
4193 4192 tosnap = strchr(tofs, '@');
4194 4193 *tosnap++ = '\0';
4195 4194
4196 4195 if (zc->zc_nvlist_src != NULL &&
4197 4196 (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
4198 4197 zc->zc_iflags, &props)) != 0)
4199 4198 return (error);
4200 4199
4201 4200 fd = zc->zc_cookie;
4202 4201 fp = getf(fd);
4203 4202 if (fp == NULL) {
4204 4203 nvlist_free(props);
4205 4204 return (SET_ERROR(EBADF));
4206 4205 }
4207 4206
4208 4207 errors = fnvlist_alloc();
4209 4208
4210 4209 if (zc->zc_string[0])
4211 4210 origin = zc->zc_string;
4212 4211
4213 4212 error = dmu_recv_begin(tofs, tosnap,
4214 4213 &zc->zc_begin_record, force, zc->zc_resumable, origin, &drc);
4215 4214 if (error != 0)
4216 4215 goto out;
4217 4216
4218 4217 /*
4219 4218 * Set properties before we receive the stream so that they are applied
4220 4219 * to the new data. Note that we must call dmu_recv_stream() if
4221 4220 * dmu_recv_begin() succeeds.
4222 4221 */
4223 4222 if (props != NULL && !drc.drc_newfs) {
4224 4223 if (spa_version(dsl_dataset_get_spa(drc.drc_ds)) >=
4225 4224 SPA_VERSION_RECVD_PROPS &&
4226 4225 !dsl_prop_get_hasrecvd(tofs))
4227 4226 first_recvd_props = B_TRUE;
4228 4227
4229 4228 /*
4230 4229 * If new received properties are supplied, they are to
4231 4230 * completely replace the existing received properties, so stash
4232 4231 * away the existing ones.
4233 4232 */
4234 4233 if (dsl_prop_get_received(tofs, &origprops) == 0) {
4235 4234 nvlist_t *errlist = NULL;
4236 4235 /*
4237 4236 * Don't bother writing a property if its value won't
4238 4237 * change (and avoid the unnecessary security checks).
4239 4238 *
4240 4239 * The first receive after SPA_VERSION_RECVD_PROPS is a
4241 4240 * special case where we blow away all local properties
4242 4241 * regardless.
4243 4242 */
4244 4243 if (!first_recvd_props)
4245 4244 props_reduce(props, origprops);
4246 4245 if (zfs_check_clearable(tofs, origprops, &errlist) != 0)
4247 4246 (void) nvlist_merge(errors, errlist, 0);
4248 4247 nvlist_free(errlist);
4249 4248
4250 4249 if (clear_received_props(tofs, origprops,
4251 4250 first_recvd_props ? NULL : props) != 0)
4252 4251 zc->zc_obj |= ZPROP_ERR_NOCLEAR;
4253 4252 } else {
4254 4253 zc->zc_obj |= ZPROP_ERR_NOCLEAR;
4255 4254 }
4256 4255 }
4257 4256
4258 4257 if (props != NULL) {
4259 4258 props_error = dsl_prop_set_hasrecvd(tofs);
4260 4259
4261 4260 if (props_error == 0) {
4262 4261 delayprops = extract_delay_props(props);
4263 4262 (void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_RECEIVED,
4264 4263 props, errors);
4265 4264 }
4266 4265 }
4267 4266
4268 4267 off = fp->f_offset;
4269 4268 error = dmu_recv_stream(&drc, fp->f_vnode, &off, zc->zc_cleanup_fd,
4270 4269 &zc->zc_action_handle);
4271 4270
4272 4271 if (error == 0) {
4273 4272 zfsvfs_t *zfsvfs = NULL;
4274 4273
4275 4274 if (getzfsvfs(tofs, &zfsvfs) == 0) {
4276 4275 /* online recv */
4277 4276 int end_err;
4278 4277
4279 4278 error = zfs_suspend_fs(zfsvfs);
4280 4279 /*
4281 4280 * If the suspend fails, then the recv_end will
4282 4281 * likely also fail, and clean up after itself.
4283 4282 */
4284 4283 end_err = dmu_recv_end(&drc, zfsvfs);
4285 4284 if (error == 0)
4286 4285 error = zfs_resume_fs(zfsvfs, tofs);
4287 4286 error = error ? error : end_err;
4288 4287 VFS_RELE(zfsvfs->z_vfs);
4289 4288 } else {
4290 4289 error = dmu_recv_end(&drc, NULL);
4291 4290 }
4292 4291
4293 4292 /* Set delayed properties now, after we're done receiving. */
4294 4293 if (delayprops != NULL && error == 0) {
4295 4294 (void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_RECEIVED,
4296 4295 delayprops, errors);
4297 4296 }
4298 4297 }
4299 4298
4300 4299 if (delayprops != NULL) {
4301 4300 /*
4302 4301 * Merge delayed props back in with initial props, in case
4303 4302 * we're DEBUG and zfs_ioc_recv_inject_err is set (which means
4304 4303 * we have to make sure clear_received_props() includes
4305 4304 * the delayed properties).
4306 4305 *
4307 4306 * Since zfs_ioc_recv_inject_err is only in DEBUG kernels,
4308 4307 * using ASSERT() will be just like a VERIFY.
4309 4308 */
4310 4309 ASSERT(nvlist_merge(props, delayprops, 0) == 0);
4311 4310 nvlist_free(delayprops);
4312 4311 }
4313 4312
4314 4313 /*
4315 4314 * Now that all props, initial and delayed, are set, report the prop
4316 4315 * errors to the caller.
4317 4316 */
4318 4317 if (zc->zc_nvlist_dst_size != 0 &&
4319 4318 (nvlist_smush(errors, zc->zc_nvlist_dst_size) != 0 ||
4320 4319 put_nvlist(zc, errors) != 0)) {
4321 4320 /*
4322 4321 * Caller made zc->zc_nvlist_dst less than the minimum expected
4323 4322 * size or supplied an invalid address.
4324 4323 */
4325 4324 props_error = SET_ERROR(EINVAL);
4326 4325 }
4327 4326
4328 4327 zc->zc_cookie = off - fp->f_offset;
4329 4328 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
4330 4329 fp->f_offset = off;
4331 4330
4332 4331 #ifdef DEBUG
4333 4332 if (zfs_ioc_recv_inject_err) {
4334 4333 zfs_ioc_recv_inject_err = B_FALSE;
4335 4334 error = 1;
4336 4335 }
4337 4336 #endif
4338 4337 /*
4339 4338 * On error, restore the original props.
4340 4339 */
4341 4340 if (error != 0 && props != NULL && !drc.drc_newfs) {
4342 4341 if (clear_received_props(tofs, props, NULL) != 0) {
4343 4342 /*
4344 4343 * We failed to clear the received properties.
4345 4344 * Since we may have left a $recvd value on the
4346 4345 * system, we can't clear the $hasrecvd flag.
4347 4346 */
4348 4347 zc->zc_obj |= ZPROP_ERR_NORESTORE;
4349 4348 } else if (first_recvd_props) {
4350 4349 dsl_prop_unset_hasrecvd(tofs);
4351 4350 }
4352 4351
4353 4352 if (origprops == NULL && !drc.drc_newfs) {
4354 4353 /* We failed to stash the original properties. */
4355 4354 zc->zc_obj |= ZPROP_ERR_NORESTORE;
4356 4355 }
4357 4356
4358 4357 /*
4359 4358 * dsl_props_set() will not convert RECEIVED to LOCAL on or
4360 4359 * after SPA_VERSION_RECVD_PROPS, so we need to specify LOCAL
4361 4360 * explictly if we're restoring local properties cleared in the
4362 4361 * first new-style receive.
4363 4362 */
4364 4363 if (origprops != NULL &&
4365 4364 zfs_set_prop_nvlist(tofs, (first_recvd_props ?
4366 4365 ZPROP_SRC_LOCAL : ZPROP_SRC_RECEIVED),
4367 4366 origprops, NULL) != 0) {
4368 4367 /*
4369 4368 * We stashed the original properties but failed to
4370 4369 * restore them.
4371 4370 */
4372 4371 zc->zc_obj |= ZPROP_ERR_NORESTORE;
4373 4372 }
4374 4373 }
4375 4374 out:
4376 4375 nvlist_free(props);
4377 4376 nvlist_free(origprops);
4378 4377 nvlist_free(errors);
4379 4378 releasef(fd);
4380 4379
4381 4380 if (error == 0)
4382 4381 error = props_error;
4383 4382
4384 4383 return (error);
4385 4384 }
4386 4385
4387 4386 /*
4388 4387 * inputs:
4389 4388 * zc_name name of snapshot to send
4390 4389 * zc_cookie file descriptor to send stream to
4391 4390 * zc_obj fromorigin flag (mutually exclusive with zc_fromobj)
4392 4391 * zc_sendobj objsetid of snapshot to send
4393 4392 * zc_fromobj objsetid of incremental fromsnap (may be zero)
4394 4393 * zc_guid if set, estimate size of stream only. zc_cookie is ignored.
4395 4394 * output size in zc_objset_type.
4396 4395 * zc_flags lzc_send_flags
4397 4396 *
4398 4397 * outputs:
4399 4398 * zc_objset_type estimated size, if zc_guid is set
4400 4399 */
4401 4400 static int
4402 4401 zfs_ioc_send(zfs_cmd_t *zc)
4403 4402 {
4404 4403 int error;
4405 4404 offset_t off;
4406 4405 boolean_t estimate = (zc->zc_guid != 0);
4407 4406 boolean_t embedok = (zc->zc_flags & 0x1);
4408 4407 boolean_t large_block_ok = (zc->zc_flags & 0x2);
4409 4408
4410 4409 if (zc->zc_obj != 0) {
4411 4410 dsl_pool_t *dp;
4412 4411 dsl_dataset_t *tosnap;
4413 4412
4414 4413 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
4415 4414 if (error != 0)
4416 4415 return (error);
4417 4416
4418 4417 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &tosnap);
4419 4418 if (error != 0) {
4420 4419 dsl_pool_rele(dp, FTAG);
4421 4420 return (error);
4422 4421 }
4423 4422
4424 4423 if (dsl_dir_is_clone(tosnap->ds_dir))
4425 4424 zc->zc_fromobj =
4426 4425 dsl_dir_phys(tosnap->ds_dir)->dd_origin_obj;
4427 4426 dsl_dataset_rele(tosnap, FTAG);
4428 4427 dsl_pool_rele(dp, FTAG);
4429 4428 }
4430 4429
4431 4430 if (estimate) {
4432 4431 dsl_pool_t *dp;
4433 4432 dsl_dataset_t *tosnap;
4434 4433 dsl_dataset_t *fromsnap = NULL;
4435 4434
4436 4435 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
4437 4436 if (error != 0)
4438 4437 return (error);
4439 4438
4440 4439 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &tosnap);
4441 4440 if (error != 0) {
4442 4441 dsl_pool_rele(dp, FTAG);
4443 4442 return (error);
4444 4443 }
4445 4444
4446 4445 if (zc->zc_fromobj != 0) {
4447 4446 error = dsl_dataset_hold_obj(dp, zc->zc_fromobj,
4448 4447 FTAG, &fromsnap);
4449 4448 if (error != 0) {
4450 4449 dsl_dataset_rele(tosnap, FTAG);
4451 4450 dsl_pool_rele(dp, FTAG);
4452 4451 return (error);
4453 4452 }
4454 4453 }
4455 4454
4456 4455 error = dmu_send_estimate(tosnap, fromsnap,
4457 4456 &zc->zc_objset_type);
4458 4457
4459 4458 if (fromsnap != NULL)
4460 4459 dsl_dataset_rele(fromsnap, FTAG);
4461 4460 dsl_dataset_rele(tosnap, FTAG);
4462 4461 dsl_pool_rele(dp, FTAG);
4463 4462 } else {
4464 4463 file_t *fp = getf(zc->zc_cookie);
4465 4464 if (fp == NULL)
4466 4465 return (SET_ERROR(EBADF));
4467 4466
4468 4467 off = fp->f_offset;
4469 4468 error = dmu_send_obj(zc->zc_name, zc->zc_sendobj,
4470 4469 zc->zc_fromobj, embedok, large_block_ok,
4471 4470 zc->zc_cookie, fp->f_vnode, &off);
4472 4471
4473 4472 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
4474 4473 fp->f_offset = off;
4475 4474 releasef(zc->zc_cookie);
4476 4475 }
4477 4476 return (error);
4478 4477 }
4479 4478
4480 4479 /*
4481 4480 * inputs:
4482 4481 * zc_name name of snapshot on which to report progress
4483 4482 * zc_cookie file descriptor of send stream
4484 4483 *
4485 4484 * outputs:
4486 4485 * zc_cookie number of bytes written in send stream thus far
4487 4486 */
4488 4487 static int
4489 4488 zfs_ioc_send_progress(zfs_cmd_t *zc)
4490 4489 {
4491 4490 dsl_pool_t *dp;
4492 4491 dsl_dataset_t *ds;
4493 4492 dmu_sendarg_t *dsp = NULL;
4494 4493 int error;
4495 4494
4496 4495 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
4497 4496 if (error != 0)
4498 4497 return (error);
4499 4498
4500 4499 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &ds);
4501 4500 if (error != 0) {
4502 4501 dsl_pool_rele(dp, FTAG);
4503 4502 return (error);
4504 4503 }
4505 4504
4506 4505 mutex_enter(&ds->ds_sendstream_lock);
4507 4506
4508 4507 /*
4509 4508 * Iterate over all the send streams currently active on this dataset.
4510 4509 * If there's one which matches the specified file descriptor _and_ the
4511 4510 * stream was started by the current process, return the progress of
4512 4511 * that stream.
4513 4512 */
4514 4513 for (dsp = list_head(&ds->ds_sendstreams); dsp != NULL;
4515 4514 dsp = list_next(&ds->ds_sendstreams, dsp)) {
4516 4515 if (dsp->dsa_outfd == zc->zc_cookie &&
4517 4516 dsp->dsa_proc == curproc)
4518 4517 break;
4519 4518 }
4520 4519
4521 4520 if (dsp != NULL)
4522 4521 zc->zc_cookie = *(dsp->dsa_off);
4523 4522 else
4524 4523 error = SET_ERROR(ENOENT);
4525 4524
4526 4525 mutex_exit(&ds->ds_sendstream_lock);
4527 4526 dsl_dataset_rele(ds, FTAG);
4528 4527 dsl_pool_rele(dp, FTAG);
4529 4528 return (error);
4530 4529 }
4531 4530
4532 4531 static int
4533 4532 zfs_ioc_inject_fault(zfs_cmd_t *zc)
4534 4533 {
4535 4534 int id, error;
4536 4535
4537 4536 error = zio_inject_fault(zc->zc_name, (int)zc->zc_guid, &id,
4538 4537 &zc->zc_inject_record);
4539 4538
4540 4539 if (error == 0)
4541 4540 zc->zc_guid = (uint64_t)id;
4542 4541
4543 4542 return (error);
4544 4543 }
4545 4544
4546 4545 static int
4547 4546 zfs_ioc_clear_fault(zfs_cmd_t *zc)
4548 4547 {
4549 4548 return (zio_clear_fault((int)zc->zc_guid));
4550 4549 }
4551 4550
4552 4551 static int
4553 4552 zfs_ioc_inject_list_next(zfs_cmd_t *zc)
4554 4553 {
4555 4554 int id = (int)zc->zc_guid;
4556 4555 int error;
4557 4556
4558 4557 error = zio_inject_list_next(&id, zc->zc_name, sizeof (zc->zc_name),
4559 4558 &zc->zc_inject_record);
4560 4559
4561 4560 zc->zc_guid = id;
4562 4561
4563 4562 return (error);
4564 4563 }
4565 4564
4566 4565 static int
4567 4566 zfs_ioc_error_log(zfs_cmd_t *zc)
4568 4567 {
4569 4568 spa_t *spa;
4570 4569 int error;
4571 4570 size_t count = (size_t)zc->zc_nvlist_dst_size;
4572 4571
4573 4572 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
4574 4573 return (error);
4575 4574
4576 4575 error = spa_get_errlog(spa, (void *)(uintptr_t)zc->zc_nvlist_dst,
4577 4576 &count);
4578 4577 if (error == 0)
4579 4578 zc->zc_nvlist_dst_size = count;
4580 4579 else
4581 4580 zc->zc_nvlist_dst_size = spa_get_errlog_size(spa);
4582 4581
4583 4582 spa_close(spa, FTAG);
4584 4583
4585 4584 return (error);
4586 4585 }
4587 4586
4588 4587 static int
4589 4588 zfs_ioc_clear(zfs_cmd_t *zc)
4590 4589 {
4591 4590 spa_t *spa;
4592 4591 vdev_t *vd;
4593 4592 int error;
4594 4593
4595 4594 /*
4596 4595 * On zpool clear we also fix up missing slogs
4597 4596 */
4598 4597 mutex_enter(&spa_namespace_lock);
4599 4598 spa = spa_lookup(zc->zc_name);
4600 4599 if (spa == NULL) {
4601 4600 mutex_exit(&spa_namespace_lock);
4602 4601 return (SET_ERROR(EIO));
4603 4602 }
4604 4603 if (spa_get_log_state(spa) == SPA_LOG_MISSING) {
4605 4604 /* we need to let spa_open/spa_load clear the chains */
4606 4605 spa_set_log_state(spa, SPA_LOG_CLEAR);
4607 4606 }
4608 4607 spa->spa_last_open_failed = 0;
4609 4608 mutex_exit(&spa_namespace_lock);
4610 4609
4611 4610 if (zc->zc_cookie & ZPOOL_NO_REWIND) {
4612 4611 error = spa_open(zc->zc_name, &spa, FTAG);
4613 4612 } else {
4614 4613 nvlist_t *policy;
4615 4614 nvlist_t *config = NULL;
4616 4615
4617 4616 if (zc->zc_nvlist_src == NULL)
4618 4617 return (SET_ERROR(EINVAL));
4619 4618
4620 4619 if ((error = get_nvlist(zc->zc_nvlist_src,
4621 4620 zc->zc_nvlist_src_size, zc->zc_iflags, &policy)) == 0) {
4622 4621 error = spa_open_rewind(zc->zc_name, &spa, FTAG,
4623 4622 policy, &config);
4624 4623 if (config != NULL) {
4625 4624 int err;
4626 4625
4627 4626 if ((err = put_nvlist(zc, config)) != 0)
4628 4627 error = err;
4629 4628 nvlist_free(config);
4630 4629 }
4631 4630 nvlist_free(policy);
4632 4631 }
4633 4632 }
4634 4633
4635 4634 if (error != 0)
4636 4635 return (error);
4637 4636
4638 4637 spa_vdev_state_enter(spa, SCL_NONE);
4639 4638
4640 4639 if (zc->zc_guid == 0) {
4641 4640 vd = NULL;
4642 4641 } else {
4643 4642 vd = spa_lookup_by_guid(spa, zc->zc_guid, B_TRUE);
4644 4643 if (vd == NULL) {
4645 4644 (void) spa_vdev_state_exit(spa, NULL, ENODEV);
4646 4645 spa_close(spa, FTAG);
4647 4646 return (SET_ERROR(ENODEV));
4648 4647 }
4649 4648 }
4650 4649
4651 4650 vdev_clear(spa, vd);
4652 4651
4653 4652 (void) spa_vdev_state_exit(spa, NULL, 0);
4654 4653
4655 4654 /*
4656 4655 * Resume any suspended I/Os.
4657 4656 */
4658 4657 if (zio_resume(spa) != 0)
4659 4658 error = SET_ERROR(EIO);
4660 4659
4661 4660 spa_close(spa, FTAG);
4662 4661
4663 4662 return (error);
4664 4663 }
4665 4664
4666 4665 static int
4667 4666 zfs_ioc_pool_reopen(zfs_cmd_t *zc)
4668 4667 {
4669 4668 spa_t *spa;
4670 4669 int error;
4671 4670
4672 4671 error = spa_open(zc->zc_name, &spa, FTAG);
4673 4672 if (error != 0)
4674 4673 return (error);
4675 4674
4676 4675 spa_vdev_state_enter(spa, SCL_NONE);
4677 4676
4678 4677 /*
4679 4678 * If a resilver is already in progress then set the
4680 4679 * spa_scrub_reopen flag to B_TRUE so that we don't restart
4681 4680 * the scan as a side effect of the reopen. Otherwise, let
4682 4681 * vdev_open() decided if a resilver is required.
4683 4682 */
4684 4683 spa->spa_scrub_reopen = dsl_scan_resilvering(spa->spa_dsl_pool);
4685 4684 vdev_reopen(spa->spa_root_vdev);
4686 4685 spa->spa_scrub_reopen = B_FALSE;
4687 4686
4688 4687 (void) spa_vdev_state_exit(spa, NULL, 0);
4689 4688 spa_close(spa, FTAG);
4690 4689 return (0);
4691 4690 }
4692 4691 /*
4693 4692 * inputs:
4694 4693 * zc_name name of filesystem
4695 4694 * zc_value name of origin snapshot
4696 4695 *
4697 4696 * outputs:
4698 4697 * zc_string name of conflicting snapshot, if there is one
4699 4698 */
4700 4699 static int
4701 4700 zfs_ioc_promote(zfs_cmd_t *zc)
4702 4701 {
4703 4702 char *cp;
4704 4703
4705 4704 /*
4706 4705 * We don't need to unmount *all* the origin fs's snapshots, but
4707 4706 * it's easier.
4708 4707 */
4709 4708 cp = strchr(zc->zc_value, '@');
4710 4709 if (cp)
4711 4710 *cp = '\0';
4712 4711 (void) dmu_objset_find(zc->zc_value,
4713 4712 zfs_unmount_snap_cb, NULL, DS_FIND_SNAPSHOTS);
4714 4713 return (dsl_dataset_promote(zc->zc_name, zc->zc_string));
4715 4714 }
4716 4715
4717 4716 /*
4718 4717 * Retrieve a single {user|group}{used|quota}@... property.
4719 4718 *
4720 4719 * inputs:
4721 4720 * zc_name name of filesystem
4722 4721 * zc_objset_type zfs_userquota_prop_t
4723 4722 * zc_value domain name (eg. "S-1-234-567-89")
4724 4723 * zc_guid RID/UID/GID
4725 4724 *
4726 4725 * outputs:
4727 4726 * zc_cookie property value
4728 4727 */
4729 4728 static int
4730 4729 zfs_ioc_userspace_one(zfs_cmd_t *zc)
4731 4730 {
4732 4731 zfsvfs_t *zfsvfs;
4733 4732 int error;
4734 4733
4735 4734 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
4736 4735 return (SET_ERROR(EINVAL));
4737 4736
4738 4737 error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE);
4739 4738 if (error != 0)
4740 4739 return (error);
4741 4740
4742 4741 error = zfs_userspace_one(zfsvfs,
4743 4742 zc->zc_objset_type, zc->zc_value, zc->zc_guid, &zc->zc_cookie);
4744 4743 zfsvfs_rele(zfsvfs, FTAG);
4745 4744
4746 4745 return (error);
4747 4746 }
4748 4747
4749 4748 /*
4750 4749 * inputs:
4751 4750 * zc_name name of filesystem
4752 4751 * zc_cookie zap cursor
4753 4752 * zc_objset_type zfs_userquota_prop_t
4754 4753 * zc_nvlist_dst[_size] buffer to fill (not really an nvlist)
4755 4754 *
4756 4755 * outputs:
4757 4756 * zc_nvlist_dst[_size] data buffer (array of zfs_useracct_t)
4758 4757 * zc_cookie zap cursor
4759 4758 */
4760 4759 static int
4761 4760 zfs_ioc_userspace_many(zfs_cmd_t *zc)
4762 4761 {
4763 4762 zfsvfs_t *zfsvfs;
4764 4763 int bufsize = zc->zc_nvlist_dst_size;
4765 4764
4766 4765 if (bufsize <= 0)
4767 4766 return (SET_ERROR(ENOMEM));
4768 4767
4769 4768 int error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE);
4770 4769 if (error != 0)
4771 4770 return (error);
4772 4771
4773 4772 void *buf = kmem_alloc(bufsize, KM_SLEEP);
4774 4773
4775 4774 error = zfs_userspace_many(zfsvfs, zc->zc_objset_type, &zc->zc_cookie,
4776 4775 buf, &zc->zc_nvlist_dst_size);
4777 4776
4778 4777 if (error == 0) {
4779 4778 error = xcopyout(buf,
4780 4779 (void *)(uintptr_t)zc->zc_nvlist_dst,
4781 4780 zc->zc_nvlist_dst_size);
4782 4781 }
4783 4782 kmem_free(buf, bufsize);
4784 4783 zfsvfs_rele(zfsvfs, FTAG);
4785 4784
4786 4785 return (error);
4787 4786 }
4788 4787
4789 4788 /*
4790 4789 * inputs:
4791 4790 * zc_name name of filesystem
4792 4791 *
4793 4792 * outputs:
4794 4793 * none
4795 4794 */
4796 4795 static int
4797 4796 zfs_ioc_userspace_upgrade(zfs_cmd_t *zc)
4798 4797 {
4799 4798 objset_t *os;
4800 4799 int error = 0;
4801 4800 zfsvfs_t *zfsvfs;
4802 4801
4803 4802 if (getzfsvfs(zc->zc_name, &zfsvfs) == 0) {
4804 4803 if (!dmu_objset_userused_enabled(zfsvfs->z_os)) {
4805 4804 /*
4806 4805 * If userused is not enabled, it may be because the
4807 4806 * objset needs to be closed & reopened (to grow the
4808 4807 * objset_phys_t). Suspend/resume the fs will do that.
4809 4808 */
4810 4809 error = zfs_suspend_fs(zfsvfs);
4811 4810 if (error == 0) {
4812 4811 dmu_objset_refresh_ownership(zfsvfs->z_os,
4813 4812 zfsvfs);
4814 4813 error = zfs_resume_fs(zfsvfs, zc->zc_name);
4815 4814 }
4816 4815 }
4817 4816 if (error == 0)
4818 4817 error = dmu_objset_userspace_upgrade(zfsvfs->z_os);
4819 4818 VFS_RELE(zfsvfs->z_vfs);
4820 4819 } else {
4821 4820 /* XXX kind of reading contents without owning */
4822 4821 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
4823 4822 if (error != 0)
4824 4823 return (error);
4825 4824
4826 4825 error = dmu_objset_userspace_upgrade(os);
4827 4826 dmu_objset_rele(os, FTAG);
4828 4827 }
4829 4828
4830 4829 return (error);
4831 4830 }
4832 4831
4833 4832 /*
4834 4833 * We don't want to have a hard dependency
4835 4834 * against some special symbols in sharefs
4836 4835 * nfs, and smbsrv. Determine them if needed when
4837 4836 * the first file system is shared.
4838 4837 * Neither sharefs, nfs or smbsrv are unloadable modules.
4839 4838 */
4840 4839 int (*znfsexport_fs)(void *arg);
4841 4840 int (*zshare_fs)(enum sharefs_sys_op, share_t *, uint32_t);
4842 4841 int (*zsmbexport_fs)(void *arg, boolean_t add_share);
4843 4842
4844 4843 int zfs_nfsshare_inited;
4845 4844 int zfs_smbshare_inited;
4846 4845
4847 4846 ddi_modhandle_t nfs_mod;
4848 4847 ddi_modhandle_t sharefs_mod;
4849 4848 ddi_modhandle_t smbsrv_mod;
4850 4849 kmutex_t zfs_share_lock;
4851 4850
4852 4851 static int
4853 4852 zfs_init_sharefs()
4854 4853 {
4855 4854 int error;
4856 4855
4857 4856 ASSERT(MUTEX_HELD(&zfs_share_lock));
4858 4857 /* Both NFS and SMB shares also require sharetab support. */
4859 4858 if (sharefs_mod == NULL && ((sharefs_mod =
4860 4859 ddi_modopen("fs/sharefs",
4861 4860 KRTLD_MODE_FIRST, &error)) == NULL)) {
4862 4861 return (SET_ERROR(ENOSYS));
4863 4862 }
4864 4863 if (zshare_fs == NULL && ((zshare_fs =
4865 4864 (int (*)(enum sharefs_sys_op, share_t *, uint32_t))
4866 4865 ddi_modsym(sharefs_mod, "sharefs_impl", &error)) == NULL)) {
4867 4866 return (SET_ERROR(ENOSYS));
4868 4867 }
4869 4868 return (0);
4870 4869 }
4871 4870
4872 4871 static int
4873 4872 zfs_ioc_share(zfs_cmd_t *zc)
4874 4873 {
4875 4874 int error;
4876 4875 int opcode;
4877 4876
4878 4877 switch (zc->zc_share.z_sharetype) {
4879 4878 case ZFS_SHARE_NFS:
4880 4879 case ZFS_UNSHARE_NFS:
4881 4880 if (zfs_nfsshare_inited == 0) {
4882 4881 mutex_enter(&zfs_share_lock);
4883 4882 if (nfs_mod == NULL && ((nfs_mod = ddi_modopen("fs/nfs",
4884 4883 KRTLD_MODE_FIRST, &error)) == NULL)) {
4885 4884 mutex_exit(&zfs_share_lock);
4886 4885 return (SET_ERROR(ENOSYS));
4887 4886 }
4888 4887 if (znfsexport_fs == NULL &&
4889 4888 ((znfsexport_fs = (int (*)(void *))
4890 4889 ddi_modsym(nfs_mod,
4891 4890 "nfs_export", &error)) == NULL)) {
4892 4891 mutex_exit(&zfs_share_lock);
4893 4892 return (SET_ERROR(ENOSYS));
4894 4893 }
4895 4894 error = zfs_init_sharefs();
4896 4895 if (error != 0) {
4897 4896 mutex_exit(&zfs_share_lock);
4898 4897 return (SET_ERROR(ENOSYS));
4899 4898 }
4900 4899 zfs_nfsshare_inited = 1;
4901 4900 mutex_exit(&zfs_share_lock);
4902 4901 }
4903 4902 break;
4904 4903 case ZFS_SHARE_SMB:
4905 4904 case ZFS_UNSHARE_SMB:
4906 4905 if (zfs_smbshare_inited == 0) {
4907 4906 mutex_enter(&zfs_share_lock);
4908 4907 if (smbsrv_mod == NULL && ((smbsrv_mod =
4909 4908 ddi_modopen("drv/smbsrv",
4910 4909 KRTLD_MODE_FIRST, &error)) == NULL)) {
4911 4910 mutex_exit(&zfs_share_lock);
4912 4911 return (SET_ERROR(ENOSYS));
4913 4912 }
4914 4913 if (zsmbexport_fs == NULL && ((zsmbexport_fs =
4915 4914 (int (*)(void *, boolean_t))ddi_modsym(smbsrv_mod,
4916 4915 "smb_server_share", &error)) == NULL)) {
4917 4916 mutex_exit(&zfs_share_lock);
4918 4917 return (SET_ERROR(ENOSYS));
4919 4918 }
4920 4919 error = zfs_init_sharefs();
4921 4920 if (error != 0) {
4922 4921 mutex_exit(&zfs_share_lock);
4923 4922 return (SET_ERROR(ENOSYS));
4924 4923 }
4925 4924 zfs_smbshare_inited = 1;
4926 4925 mutex_exit(&zfs_share_lock);
4927 4926 }
4928 4927 break;
4929 4928 default:
4930 4929 return (SET_ERROR(EINVAL));
4931 4930 }
4932 4931
4933 4932 switch (zc->zc_share.z_sharetype) {
4934 4933 case ZFS_SHARE_NFS:
4935 4934 case ZFS_UNSHARE_NFS:
4936 4935 if (error =
4937 4936 znfsexport_fs((void *)
4938 4937 (uintptr_t)zc->zc_share.z_exportdata))
4939 4938 return (error);
4940 4939 break;
4941 4940 case ZFS_SHARE_SMB:
4942 4941 case ZFS_UNSHARE_SMB:
4943 4942 if (error = zsmbexport_fs((void *)
4944 4943 (uintptr_t)zc->zc_share.z_exportdata,
4945 4944 zc->zc_share.z_sharetype == ZFS_SHARE_SMB ?
4946 4945 B_TRUE: B_FALSE)) {
4947 4946 return (error);
4948 4947 }
4949 4948 break;
4950 4949 }
4951 4950
4952 4951 opcode = (zc->zc_share.z_sharetype == ZFS_SHARE_NFS ||
4953 4952 zc->zc_share.z_sharetype == ZFS_SHARE_SMB) ?
4954 4953 SHAREFS_ADD : SHAREFS_REMOVE;
4955 4954
4956 4955 /*
4957 4956 * Add or remove share from sharetab
4958 4957 */
4959 4958 error = zshare_fs(opcode,
4960 4959 (void *)(uintptr_t)zc->zc_share.z_sharedata,
4961 4960 zc->zc_share.z_sharemax);
4962 4961
4963 4962 return (error);
4964 4963
4965 4964 }
4966 4965
4967 4966 ace_t full_access[] = {
4968 4967 {(uid_t)-1, ACE_ALL_PERMS, ACE_EVERYONE, 0}
4969 4968 };
4970 4969
4971 4970 /*
4972 4971 * inputs:
4973 4972 * zc_name name of containing filesystem
4974 4973 * zc_obj object # beyond which we want next in-use object #
4975 4974 *
4976 4975 * outputs:
4977 4976 * zc_obj next in-use object #
4978 4977 */
4979 4978 static int
4980 4979 zfs_ioc_next_obj(zfs_cmd_t *zc)
4981 4980 {
4982 4981 objset_t *os = NULL;
4983 4982 int error;
4984 4983
4985 4984 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
4986 4985 if (error != 0)
4987 4986 return (error);
4988 4987
4989 4988 error = dmu_object_next(os, &zc->zc_obj, B_FALSE,
4990 4989 dsl_dataset_phys(os->os_dsl_dataset)->ds_prev_snap_txg);
4991 4990
4992 4991 dmu_objset_rele(os, FTAG);
4993 4992 return (error);
4994 4993 }
4995 4994
4996 4995 /*
4997 4996 * inputs:
4998 4997 * zc_name name of filesystem
4999 4998 * zc_value prefix name for snapshot
5000 4999 * zc_cleanup_fd cleanup-on-exit file descriptor for calling process
5001 5000 *
5002 5001 * outputs:
5003 5002 * zc_value short name of new snapshot
5004 5003 */
5005 5004 static int
5006 5005 zfs_ioc_tmp_snapshot(zfs_cmd_t *zc)
5007 5006 {
5008 5007 char *snap_name;
5009 5008 char *hold_name;
5010 5009 int error;
5011 5010 minor_t minor;
5012 5011
5013 5012 error = zfs_onexit_fd_hold(zc->zc_cleanup_fd, &minor);
5014 5013 if (error != 0)
5015 5014 return (error);
5016 5015
5017 5016 snap_name = kmem_asprintf("%s-%016llx", zc->zc_value,
5018 5017 (u_longlong_t)ddi_get_lbolt64());
5019 5018 hold_name = kmem_asprintf("%%%s", zc->zc_value);
5020 5019
5021 5020 error = dsl_dataset_snapshot_tmp(zc->zc_name, snap_name, minor,
5022 5021 hold_name);
5023 5022 if (error == 0)
5024 5023 (void) strcpy(zc->zc_value, snap_name);
5025 5024 strfree(snap_name);
5026 5025 strfree(hold_name);
5027 5026 zfs_onexit_fd_rele(zc->zc_cleanup_fd);
5028 5027 return (error);
5029 5028 }
5030 5029
5031 5030 /*
5032 5031 * inputs:
5033 5032 * zc_name name of "to" snapshot
5034 5033 * zc_value name of "from" snapshot
5035 5034 * zc_cookie file descriptor to write diff data on
5036 5035 *
5037 5036 * outputs:
5038 5037 * dmu_diff_record_t's to the file descriptor
5039 5038 */
5040 5039 static int
5041 5040 zfs_ioc_diff(zfs_cmd_t *zc)
5042 5041 {
5043 5042 file_t *fp;
5044 5043 offset_t off;
5045 5044 int error;
5046 5045
5047 5046 fp = getf(zc->zc_cookie);
5048 5047 if (fp == NULL)
5049 5048 return (SET_ERROR(EBADF));
5050 5049
5051 5050 off = fp->f_offset;
5052 5051
5053 5052 error = dmu_diff(zc->zc_name, zc->zc_value, fp->f_vnode, &off);
5054 5053
5055 5054 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
5056 5055 fp->f_offset = off;
5057 5056 releasef(zc->zc_cookie);
5058 5057
5059 5058 return (error);
5060 5059 }
5061 5060
5062 5061 /*
5063 5062 * Remove all ACL files in shares dir
5064 5063 */
5065 5064 static int
5066 5065 zfs_smb_acl_purge(znode_t *dzp)
5067 5066 {
5068 5067 zap_cursor_t zc;
5069 5068 zap_attribute_t zap;
5070 5069 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
5071 5070 int error;
5072 5071
5073 5072 for (zap_cursor_init(&zc, zfsvfs->z_os, dzp->z_id);
5074 5073 (error = zap_cursor_retrieve(&zc, &zap)) == 0;
5075 5074 zap_cursor_advance(&zc)) {
5076 5075 if ((error = VOP_REMOVE(ZTOV(dzp), zap.za_name, kcred,
5077 5076 NULL, 0)) != 0)
5078 5077 break;
5079 5078 }
5080 5079 zap_cursor_fini(&zc);
5081 5080 return (error);
5082 5081 }
5083 5082
5084 5083 static int
5085 5084 zfs_ioc_smb_acl(zfs_cmd_t *zc)
5086 5085 {
5087 5086 vnode_t *vp;
5088 5087 znode_t *dzp;
5089 5088 vnode_t *resourcevp = NULL;
5090 5089 znode_t *sharedir;
5091 5090 zfsvfs_t *zfsvfs;
5092 5091 nvlist_t *nvlist;
5093 5092 char *src, *target;
5094 5093 vattr_t vattr;
5095 5094 vsecattr_t vsec;
5096 5095 int error = 0;
5097 5096
5098 5097 if ((error = lookupname(zc->zc_value, UIO_SYSSPACE,
5099 5098 NO_FOLLOW, NULL, &vp)) != 0)
5100 5099 return (error);
5101 5100
5102 5101 /* Now make sure mntpnt and dataset are ZFS */
5103 5102
5104 5103 if (vp->v_vfsp->vfs_fstype != zfsfstype ||
5105 5104 (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource),
5106 5105 zc->zc_name) != 0)) {
5107 5106 VN_RELE(vp);
5108 5107 return (SET_ERROR(EINVAL));
5109 5108 }
5110 5109
5111 5110 dzp = VTOZ(vp);
5112 5111 zfsvfs = dzp->z_zfsvfs;
5113 5112 ZFS_ENTER(zfsvfs);
5114 5113
5115 5114 /*
5116 5115 * Create share dir if its missing.
5117 5116 */
5118 5117 mutex_enter(&zfsvfs->z_lock);
5119 5118 if (zfsvfs->z_shares_dir == 0) {
5120 5119 dmu_tx_t *tx;
5121 5120
5122 5121 tx = dmu_tx_create(zfsvfs->z_os);
5123 5122 dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, TRUE,
5124 5123 ZFS_SHARES_DIR);
5125 5124 dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL);
5126 5125 error = dmu_tx_assign(tx, TXG_WAIT);
5127 5126 if (error != 0) {
5128 5127 dmu_tx_abort(tx);
5129 5128 } else {
5130 5129 error = zfs_create_share_dir(zfsvfs, tx);
5131 5130 dmu_tx_commit(tx);
5132 5131 }
5133 5132 if (error != 0) {
5134 5133 mutex_exit(&zfsvfs->z_lock);
5135 5134 VN_RELE(vp);
5136 5135 ZFS_EXIT(zfsvfs);
5137 5136 return (error);
5138 5137 }
5139 5138 }
5140 5139 mutex_exit(&zfsvfs->z_lock);
5141 5140
5142 5141 ASSERT(zfsvfs->z_shares_dir);
5143 5142 if ((error = zfs_zget(zfsvfs, zfsvfs->z_shares_dir, &sharedir)) != 0) {
5144 5143 VN_RELE(vp);
5145 5144 ZFS_EXIT(zfsvfs);
5146 5145 return (error);
5147 5146 }
5148 5147
5149 5148 switch (zc->zc_cookie) {
5150 5149 case ZFS_SMB_ACL_ADD:
5151 5150 vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE;
5152 5151 vattr.va_type = VREG;
5153 5152 vattr.va_mode = S_IFREG|0777;
5154 5153 vattr.va_uid = 0;
5155 5154 vattr.va_gid = 0;
5156 5155
5157 5156 vsec.vsa_mask = VSA_ACE;
5158 5157 vsec.vsa_aclentp = &full_access;
5159 5158 vsec.vsa_aclentsz = sizeof (full_access);
5160 5159 vsec.vsa_aclcnt = 1;
5161 5160
5162 5161 error = VOP_CREATE(ZTOV(sharedir), zc->zc_string,
5163 5162 &vattr, EXCL, 0, &resourcevp, kcred, 0, NULL, &vsec);
5164 5163 if (resourcevp)
5165 5164 VN_RELE(resourcevp);
5166 5165 break;
5167 5166
5168 5167 case ZFS_SMB_ACL_REMOVE:
5169 5168 error = VOP_REMOVE(ZTOV(sharedir), zc->zc_string, kcred,
5170 5169 NULL, 0);
5171 5170 break;
5172 5171
5173 5172 case ZFS_SMB_ACL_RENAME:
5174 5173 if ((error = get_nvlist(zc->zc_nvlist_src,
5175 5174 zc->zc_nvlist_src_size, zc->zc_iflags, &nvlist)) != 0) {
5176 5175 VN_RELE(vp);
5177 5176 VN_RELE(ZTOV(sharedir));
5178 5177 ZFS_EXIT(zfsvfs);
5179 5178 return (error);
5180 5179 }
5181 5180 if (nvlist_lookup_string(nvlist, ZFS_SMB_ACL_SRC, &src) ||
5182 5181 nvlist_lookup_string(nvlist, ZFS_SMB_ACL_TARGET,
5183 5182 &target)) {
5184 5183 VN_RELE(vp);
5185 5184 VN_RELE(ZTOV(sharedir));
5186 5185 ZFS_EXIT(zfsvfs);
5187 5186 nvlist_free(nvlist);
5188 5187 return (error);
5189 5188 }
5190 5189 error = VOP_RENAME(ZTOV(sharedir), src, ZTOV(sharedir), target,
5191 5190 kcred, NULL, 0);
5192 5191 nvlist_free(nvlist);
5193 5192 break;
5194 5193
5195 5194 case ZFS_SMB_ACL_PURGE:
5196 5195 error = zfs_smb_acl_purge(sharedir);
5197 5196 break;
5198 5197
5199 5198 default:
5200 5199 error = SET_ERROR(EINVAL);
5201 5200 break;
5202 5201 }
5203 5202
5204 5203 VN_RELE(vp);
5205 5204 VN_RELE(ZTOV(sharedir));
5206 5205
5207 5206 ZFS_EXIT(zfsvfs);
5208 5207
5209 5208 return (error);
5210 5209 }
5211 5210
5212 5211 /*
5213 5212 * innvl: {
5214 5213 * "holds" -> { snapname -> holdname (string), ... }
5215 5214 * (optional) "cleanup_fd" -> fd (int32)
5216 5215 * }
5217 5216 *
5218 5217 * outnvl: {
5219 5218 * snapname -> error value (int32)
5220 5219 * ...
5221 5220 * }
5222 5221 */
5223 5222 /* ARGSUSED */
5224 5223 static int
5225 5224 zfs_ioc_hold(const char *pool, nvlist_t *args, nvlist_t *errlist)
5226 5225 {
5227 5226 nvpair_t *pair;
5228 5227 nvlist_t *holds;
5229 5228 int cleanup_fd = -1;
5230 5229 int error;
5231 5230 minor_t minor = 0;
5232 5231
5233 5232 error = nvlist_lookup_nvlist(args, "holds", &holds);
5234 5233 if (error != 0)
5235 5234 return (SET_ERROR(EINVAL));
5236 5235
5237 5236 /* make sure the user didn't pass us any invalid (empty) tags */
5238 5237 for (pair = nvlist_next_nvpair(holds, NULL); pair != NULL;
5239 5238 pair = nvlist_next_nvpair(holds, pair)) {
5240 5239 char *htag;
5241 5240
5242 5241 error = nvpair_value_string(pair, &htag);
5243 5242 if (error != 0)
5244 5243 return (SET_ERROR(error));
5245 5244
5246 5245 if (strlen(htag) == 0)
5247 5246 return (SET_ERROR(EINVAL));
5248 5247 }
5249 5248
5250 5249 if (nvlist_lookup_int32(args, "cleanup_fd", &cleanup_fd) == 0) {
5251 5250 error = zfs_onexit_fd_hold(cleanup_fd, &minor);
5252 5251 if (error != 0)
5253 5252 return (error);
5254 5253 }
5255 5254
5256 5255 error = dsl_dataset_user_hold(holds, minor, errlist);
5257 5256 if (minor != 0)
5258 5257 zfs_onexit_fd_rele(cleanup_fd);
5259 5258 return (error);
5260 5259 }
5261 5260
5262 5261 /*
5263 5262 * innvl is not used.
5264 5263 *
5265 5264 * outnvl: {
5266 5265 * holdname -> time added (uint64 seconds since epoch)
5267 5266 * ...
5268 5267 * }
5269 5268 */
5270 5269 /* ARGSUSED */
5271 5270 static int
5272 5271 zfs_ioc_get_holds(const char *snapname, nvlist_t *args, nvlist_t *outnvl)
5273 5272 {
5274 5273 return (dsl_dataset_get_holds(snapname, outnvl));
5275 5274 }
5276 5275
5277 5276 /*
5278 5277 * innvl: {
5279 5278 * snapname -> { holdname, ... }
5280 5279 * ...
5281 5280 * }
5282 5281 *
5283 5282 * outnvl: {
5284 5283 * snapname -> error value (int32)
5285 5284 * ...
5286 5285 * }
5287 5286 */
5288 5287 /* ARGSUSED */
5289 5288 static int
5290 5289 zfs_ioc_release(const char *pool, nvlist_t *holds, nvlist_t *errlist)
5291 5290 {
5292 5291 return (dsl_dataset_user_release(holds, errlist));
5293 5292 }
5294 5293
5295 5294 /*
5296 5295 * inputs:
5297 5296 * zc_name name of new filesystem or snapshot
5298 5297 * zc_value full name of old snapshot
5299 5298 *
5300 5299 * outputs:
5301 5300 * zc_cookie space in bytes
5302 5301 * zc_objset_type compressed space in bytes
5303 5302 * zc_perm_action uncompressed space in bytes
5304 5303 */
5305 5304 static int
5306 5305 zfs_ioc_space_written(zfs_cmd_t *zc)
5307 5306 {
5308 5307 int error;
5309 5308 dsl_pool_t *dp;
5310 5309 dsl_dataset_t *new, *old;
5311 5310
5312 5311 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
5313 5312 if (error != 0)
5314 5313 return (error);
5315 5314 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &new);
5316 5315 if (error != 0) {
5317 5316 dsl_pool_rele(dp, FTAG);
5318 5317 return (error);
5319 5318 }
5320 5319 error = dsl_dataset_hold(dp, zc->zc_value, FTAG, &old);
5321 5320 if (error != 0) {
5322 5321 dsl_dataset_rele(new, FTAG);
5323 5322 dsl_pool_rele(dp, FTAG);
5324 5323 return (error);
5325 5324 }
5326 5325
5327 5326 error = dsl_dataset_space_written(old, new, &zc->zc_cookie,
5328 5327 &zc->zc_objset_type, &zc->zc_perm_action);
5329 5328 dsl_dataset_rele(old, FTAG);
5330 5329 dsl_dataset_rele(new, FTAG);
5331 5330 dsl_pool_rele(dp, FTAG);
5332 5331 return (error);
5333 5332 }
5334 5333
5335 5334 /*
5336 5335 * innvl: {
5337 5336 * "firstsnap" -> snapshot name
5338 5337 * }
5339 5338 *
5340 5339 * outnvl: {
5341 5340 * "used" -> space in bytes
5342 5341 * "compressed" -> compressed space in bytes
5343 5342 * "uncompressed" -> uncompressed space in bytes
5344 5343 * }
5345 5344 */
5346 5345 static int
5347 5346 zfs_ioc_space_snaps(const char *lastsnap, nvlist_t *innvl, nvlist_t *outnvl)
5348 5347 {
5349 5348 int error;
5350 5349 dsl_pool_t *dp;
5351 5350 dsl_dataset_t *new, *old;
5352 5351 char *firstsnap;
5353 5352 uint64_t used, comp, uncomp;
5354 5353
5355 5354 if (nvlist_lookup_string(innvl, "firstsnap", &firstsnap) != 0)
5356 5355 return (SET_ERROR(EINVAL));
5357 5356
5358 5357 error = dsl_pool_hold(lastsnap, FTAG, &dp);
5359 5358 if (error != 0)
5360 5359 return (error);
5361 5360
5362 5361 error = dsl_dataset_hold(dp, lastsnap, FTAG, &new);
5363 5362 if (error == 0 && !new->ds_is_snapshot) {
5364 5363 dsl_dataset_rele(new, FTAG);
5365 5364 error = SET_ERROR(EINVAL);
5366 5365 }
5367 5366 if (error != 0) {
5368 5367 dsl_pool_rele(dp, FTAG);
5369 5368 return (error);
5370 5369 }
5371 5370 error = dsl_dataset_hold(dp, firstsnap, FTAG, &old);
5372 5371 if (error == 0 && !old->ds_is_snapshot) {
5373 5372 dsl_dataset_rele(old, FTAG);
5374 5373 error = SET_ERROR(EINVAL);
5375 5374 }
5376 5375 if (error != 0) {
5377 5376 dsl_dataset_rele(new, FTAG);
5378 5377 dsl_pool_rele(dp, FTAG);
5379 5378 return (error);
5380 5379 }
5381 5380
5382 5381 error = dsl_dataset_space_wouldfree(old, new, &used, &comp, &uncomp);
5383 5382 dsl_dataset_rele(old, FTAG);
5384 5383 dsl_dataset_rele(new, FTAG);
5385 5384 dsl_pool_rele(dp, FTAG);
5386 5385 fnvlist_add_uint64(outnvl, "used", used);
5387 5386 fnvlist_add_uint64(outnvl, "compressed", comp);
5388 5387 fnvlist_add_uint64(outnvl, "uncompressed", uncomp);
5389 5388 return (error);
5390 5389 }
5391 5390
5392 5391 /*
5393 5392 * innvl: {
5394 5393 * "fd" -> file descriptor to write stream to (int32)
5395 5394 * (optional) "fromsnap" -> full snap name to send an incremental from
5396 5395 * (optional) "largeblockok" -> (value ignored)
5397 5396 * indicates that blocks > 128KB are permitted
5398 5397 * (optional) "embedok" -> (value ignored)
5399 5398 * presence indicates DRR_WRITE_EMBEDDED records are permitted
5400 5399 * (optional) "resume_object" and "resume_offset" -> (uint64)
5401 5400 * if present, resume send stream from specified object and offset.
5402 5401 * }
5403 5402 *
5404 5403 * outnvl is unused
5405 5404 */
5406 5405 /* ARGSUSED */
5407 5406 static int
5408 5407 zfs_ioc_send_new(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl)
5409 5408 {
5410 5409 int error;
5411 5410 offset_t off;
5412 5411 char *fromname = NULL;
5413 5412 int fd;
5414 5413 boolean_t largeblockok;
5415 5414 boolean_t embedok;
5416 5415 uint64_t resumeobj = 0;
5417 5416 uint64_t resumeoff = 0;
5418 5417
5419 5418 error = nvlist_lookup_int32(innvl, "fd", &fd);
5420 5419 if (error != 0)
5421 5420 return (SET_ERROR(EINVAL));
5422 5421
5423 5422 (void) nvlist_lookup_string(innvl, "fromsnap", &fromname);
5424 5423
5425 5424 largeblockok = nvlist_exists(innvl, "largeblockok");
5426 5425 embedok = nvlist_exists(innvl, "embedok");
5427 5426
5428 5427 (void) nvlist_lookup_uint64(innvl, "resume_object", &resumeobj);
5429 5428 (void) nvlist_lookup_uint64(innvl, "resume_offset", &resumeoff);
5430 5429
5431 5430 file_t *fp = getf(fd);
5432 5431 if (fp == NULL)
5433 5432 return (SET_ERROR(EBADF));
5434 5433
5435 5434 off = fp->f_offset;
5436 5435 error = dmu_send(snapname, fromname, embedok, largeblockok, fd,
5437 5436 resumeobj, resumeoff, fp->f_vnode, &off);
5438 5437
5439 5438 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
5440 5439 fp->f_offset = off;
5441 5440 releasef(fd);
5442 5441 return (error);
5443 5442 }
5444 5443
5445 5444 /*
5446 5445 * Determine approximately how large a zfs send stream will be -- the number
5447 5446 * of bytes that will be written to the fd supplied to zfs_ioc_send_new().
5448 5447 *
5449 5448 * innvl: {
5450 5449 * (optional) "from" -> full snap or bookmark name to send an incremental
5451 5450 * from
5452 5451 * }
5453 5452 *
5454 5453 * outnvl: {
5455 5454 * "space" -> bytes of space (uint64)
5456 5455 * }
5457 5456 */
5458 5457 static int
5459 5458 zfs_ioc_send_space(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl)
5460 5459 {
5461 5460 dsl_pool_t *dp;
5462 5461 dsl_dataset_t *tosnap;
5463 5462 int error;
5464 5463 char *fromname;
5465 5464 uint64_t space;
5466 5465
5467 5466 error = dsl_pool_hold(snapname, FTAG, &dp);
5468 5467 if (error != 0)
5469 5468 return (error);
5470 5469
5471 5470 error = dsl_dataset_hold(dp, snapname, FTAG, &tosnap);
5472 5471 if (error != 0) {
5473 5472 dsl_pool_rele(dp, FTAG);
5474 5473 return (error);
5475 5474 }
5476 5475
5477 5476 error = nvlist_lookup_string(innvl, "from", &fromname);
5478 5477 if (error == 0) {
5479 5478 if (strchr(fromname, '@') != NULL) {
5480 5479 /*
5481 5480 * If from is a snapshot, hold it and use the more
5482 5481 * efficient dmu_send_estimate to estimate send space
5483 5482 * size using deadlists.
5484 5483 */
5485 5484 dsl_dataset_t *fromsnap;
5486 5485 error = dsl_dataset_hold(dp, fromname, FTAG, &fromsnap);
5487 5486 if (error != 0)
5488 5487 goto out;
5489 5488 error = dmu_send_estimate(tosnap, fromsnap, &space);
5490 5489 dsl_dataset_rele(fromsnap, FTAG);
5491 5490 } else if (strchr(fromname, '#') != NULL) {
5492 5491 /*
5493 5492 * If from is a bookmark, fetch the creation TXG of the
5494 5493 * snapshot it was created from and use that to find
5495 5494 * blocks that were born after it.
5496 5495 */
5497 5496 zfs_bookmark_phys_t frombm;
5498 5497
5499 5498 error = dsl_bookmark_lookup(dp, fromname, tosnap,
5500 5499 &frombm);
5501 5500 if (error != 0)
5502 5501 goto out;
5503 5502 error = dmu_send_estimate_from_txg(tosnap,
5504 5503 frombm.zbm_creation_txg, &space);
5505 5504 } else {
5506 5505 /*
5507 5506 * from is not properly formatted as a snapshot or
5508 5507 * bookmark
5509 5508 */
5510 5509 error = SET_ERROR(EINVAL);
5511 5510 goto out;
5512 5511 }
5513 5512 } else {
5514 5513 // If estimating the size of a full send, use dmu_send_estimate
5515 5514 error = dmu_send_estimate(tosnap, NULL, &space);
5516 5515 }
5517 5516
5518 5517 fnvlist_add_uint64(outnvl, "space", space);
5519 5518
5520 5519 out:
5521 5520 dsl_dataset_rele(tosnap, FTAG);
5522 5521 dsl_pool_rele(dp, FTAG);
5523 5522 return (error);
5524 5523 }
5525 5524
5526 5525 static zfs_ioc_vec_t zfs_ioc_vec[ZFS_IOC_LAST - ZFS_IOC_FIRST];
5527 5526
5528 5527 static void
5529 5528 zfs_ioctl_register_legacy(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5530 5529 zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck,
5531 5530 boolean_t log_history, zfs_ioc_poolcheck_t pool_check)
5532 5531 {
5533 5532 zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST];
5534 5533
5535 5534 ASSERT3U(ioc, >=, ZFS_IOC_FIRST);
5536 5535 ASSERT3U(ioc, <, ZFS_IOC_LAST);
5537 5536 ASSERT3P(vec->zvec_legacy_func, ==, NULL);
5538 5537 ASSERT3P(vec->zvec_func, ==, NULL);
5539 5538
5540 5539 vec->zvec_legacy_func = func;
5541 5540 vec->zvec_secpolicy = secpolicy;
5542 5541 vec->zvec_namecheck = namecheck;
5543 5542 vec->zvec_allow_log = log_history;
5544 5543 vec->zvec_pool_check = pool_check;
5545 5544 }
5546 5545
5547 5546 /*
5548 5547 * See the block comment at the beginning of this file for details on
5549 5548 * each argument to this function.
5550 5549 */
5551 5550 static void
5552 5551 zfs_ioctl_register(const char *name, zfs_ioc_t ioc, zfs_ioc_func_t *func,
5553 5552 zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck,
5554 5553 zfs_ioc_poolcheck_t pool_check, boolean_t smush_outnvlist,
5555 5554 boolean_t allow_log)
5556 5555 {
5557 5556 zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST];
5558 5557
5559 5558 ASSERT3U(ioc, >=, ZFS_IOC_FIRST);
5560 5559 ASSERT3U(ioc, <, ZFS_IOC_LAST);
5561 5560 ASSERT3P(vec->zvec_legacy_func, ==, NULL);
5562 5561 ASSERT3P(vec->zvec_func, ==, NULL);
5563 5562
5564 5563 /* if we are logging, the name must be valid */
5565 5564 ASSERT(!allow_log || namecheck != NO_NAME);
5566 5565
5567 5566 vec->zvec_name = name;
5568 5567 vec->zvec_func = func;
5569 5568 vec->zvec_secpolicy = secpolicy;
5570 5569 vec->zvec_namecheck = namecheck;
5571 5570 vec->zvec_pool_check = pool_check;
5572 5571 vec->zvec_smush_outnvlist = smush_outnvlist;
5573 5572 vec->zvec_allow_log = allow_log;
5574 5573 }
5575 5574
5576 5575 static void
5577 5576 zfs_ioctl_register_pool(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5578 5577 zfs_secpolicy_func_t *secpolicy, boolean_t log_history,
5579 5578 zfs_ioc_poolcheck_t pool_check)
5580 5579 {
5581 5580 zfs_ioctl_register_legacy(ioc, func, secpolicy,
5582 5581 POOL_NAME, log_history, pool_check);
5583 5582 }
5584 5583
5585 5584 static void
5586 5585 zfs_ioctl_register_dataset_nolog(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5587 5586 zfs_secpolicy_func_t *secpolicy, zfs_ioc_poolcheck_t pool_check)
5588 5587 {
5589 5588 zfs_ioctl_register_legacy(ioc, func, secpolicy,
5590 5589 DATASET_NAME, B_FALSE, pool_check);
5591 5590 }
5592 5591
5593 5592 static void
5594 5593 zfs_ioctl_register_pool_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func)
5595 5594 {
5596 5595 zfs_ioctl_register_legacy(ioc, func, zfs_secpolicy_config,
5597 5596 POOL_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5598 5597 }
5599 5598
5600 5599 static void
5601 5600 zfs_ioctl_register_pool_meta(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5602 5601 zfs_secpolicy_func_t *secpolicy)
5603 5602 {
5604 5603 zfs_ioctl_register_legacy(ioc, func, secpolicy,
5605 5604 NO_NAME, B_FALSE, POOL_CHECK_NONE);
5606 5605 }
5607 5606
5608 5607 static void
5609 5608 zfs_ioctl_register_dataset_read_secpolicy(zfs_ioc_t ioc,
5610 5609 zfs_ioc_legacy_func_t *func, zfs_secpolicy_func_t *secpolicy)
5611 5610 {
5612 5611 zfs_ioctl_register_legacy(ioc, func, secpolicy,
5613 5612 DATASET_NAME, B_FALSE, POOL_CHECK_SUSPENDED);
5614 5613 }
5615 5614
5616 5615 static void
5617 5616 zfs_ioctl_register_dataset_read(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func)
5618 5617 {
5619 5618 zfs_ioctl_register_dataset_read_secpolicy(ioc, func,
5620 5619 zfs_secpolicy_read);
5621 5620 }
5622 5621
5623 5622 static void
5624 5623 zfs_ioctl_register_dataset_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5625 5624 zfs_secpolicy_func_t *secpolicy)
5626 5625 {
5627 5626 zfs_ioctl_register_legacy(ioc, func, secpolicy,
5628 5627 DATASET_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5629 5628 }
5630 5629
5631 5630 static void
5632 5631 zfs_ioctl_init(void)
5633 5632 {
5634 5633 zfs_ioctl_register("snapshot", ZFS_IOC_SNAPSHOT,
5635 5634 zfs_ioc_snapshot, zfs_secpolicy_snapshot, POOL_NAME,
5636 5635 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5637 5636
5638 5637 zfs_ioctl_register("log_history", ZFS_IOC_LOG_HISTORY,
5639 5638 zfs_ioc_log_history, zfs_secpolicy_log_history, NO_NAME,
5640 5639 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_FALSE);
5641 5640
5642 5641 zfs_ioctl_register("space_snaps", ZFS_IOC_SPACE_SNAPS,
5643 5642 zfs_ioc_space_snaps, zfs_secpolicy_read, DATASET_NAME,
5644 5643 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5645 5644
5646 5645 zfs_ioctl_register("send", ZFS_IOC_SEND_NEW,
5647 5646 zfs_ioc_send_new, zfs_secpolicy_send_new, DATASET_NAME,
5648 5647 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5649 5648
5650 5649 zfs_ioctl_register("send_space", ZFS_IOC_SEND_SPACE,
5651 5650 zfs_ioc_send_space, zfs_secpolicy_read, DATASET_NAME,
5652 5651 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5653 5652
5654 5653 zfs_ioctl_register("create", ZFS_IOC_CREATE,
5655 5654 zfs_ioc_create, zfs_secpolicy_create_clone, DATASET_NAME,
5656 5655 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5657 5656
5658 5657 zfs_ioctl_register("clone", ZFS_IOC_CLONE,
5659 5658 zfs_ioc_clone, zfs_secpolicy_create_clone, DATASET_NAME,
5660 5659 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5661 5660
5662 5661 zfs_ioctl_register("destroy_snaps", ZFS_IOC_DESTROY_SNAPS,
5663 5662 zfs_ioc_destroy_snaps, zfs_secpolicy_destroy_snaps, POOL_NAME,
5664 5663 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5665 5664
5666 5665 zfs_ioctl_register("hold", ZFS_IOC_HOLD,
5667 5666 zfs_ioc_hold, zfs_secpolicy_hold, POOL_NAME,
5668 5667 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5669 5668 zfs_ioctl_register("release", ZFS_IOC_RELEASE,
5670 5669 zfs_ioc_release, zfs_secpolicy_release, POOL_NAME,
5671 5670 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5672 5671
5673 5672 zfs_ioctl_register("get_holds", ZFS_IOC_GET_HOLDS,
5674 5673 zfs_ioc_get_holds, zfs_secpolicy_read, DATASET_NAME,
5675 5674 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5676 5675
5677 5676 zfs_ioctl_register("rollback", ZFS_IOC_ROLLBACK,
5678 5677 zfs_ioc_rollback, zfs_secpolicy_rollback, DATASET_NAME,
5679 5678 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_TRUE);
5680 5679
5681 5680 zfs_ioctl_register("bookmark", ZFS_IOC_BOOKMARK,
5682 5681 zfs_ioc_bookmark, zfs_secpolicy_bookmark, POOL_NAME,
5683 5682 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5684 5683
5685 5684 zfs_ioctl_register("get_bookmarks", ZFS_IOC_GET_BOOKMARKS,
5686 5685 zfs_ioc_get_bookmarks, zfs_secpolicy_read, DATASET_NAME,
5687 5686 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5688 5687
5689 5688 zfs_ioctl_register("destroy_bookmarks", ZFS_IOC_DESTROY_BOOKMARKS,
5690 5689 zfs_ioc_destroy_bookmarks, zfs_secpolicy_destroy_bookmarks,
5691 5690 POOL_NAME,
5692 5691 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5693 5692
5694 5693 /* IOCTLS that use the legacy function signature */
5695 5694
5696 5695 zfs_ioctl_register_legacy(ZFS_IOC_POOL_FREEZE, zfs_ioc_pool_freeze,
5697 5696 zfs_secpolicy_config, NO_NAME, B_FALSE, POOL_CHECK_READONLY);
5698 5697
5699 5698 zfs_ioctl_register_pool(ZFS_IOC_POOL_CREATE, zfs_ioc_pool_create,
5700 5699 zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE);
5701 5700 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SCAN,
5702 5701 zfs_ioc_pool_scan);
5703 5702 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_UPGRADE,
5704 5703 zfs_ioc_pool_upgrade);
5705 5704 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ADD,
5706 5705 zfs_ioc_vdev_add);
5707 5706 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_REMOVE,
5708 5707 zfs_ioc_vdev_remove);
5709 5708 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SET_STATE,
5710 5709 zfs_ioc_vdev_set_state);
5711 5710 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ATTACH,
5712 5711 zfs_ioc_vdev_attach);
5713 5712 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_DETACH,
5714 5713 zfs_ioc_vdev_detach);
5715 5714 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETPATH,
5716 5715 zfs_ioc_vdev_setpath);
5717 5716 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETFRU,
5718 5717 zfs_ioc_vdev_setfru);
5719 5718 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SET_PROPS,
5720 5719 zfs_ioc_pool_set_props);
5721 5720 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SPLIT,
5722 5721 zfs_ioc_vdev_split);
5723 5722 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_REGUID,
5724 5723 zfs_ioc_pool_reguid);
5725 5724
5726 5725 zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_CONFIGS,
5727 5726 zfs_ioc_pool_configs, zfs_secpolicy_none);
5728 5727 zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_TRYIMPORT,
5729 5728 zfs_ioc_pool_tryimport, zfs_secpolicy_config);
5730 5729 zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_FAULT,
5731 5730 zfs_ioc_inject_fault, zfs_secpolicy_inject);
5732 5731 zfs_ioctl_register_pool_meta(ZFS_IOC_CLEAR_FAULT,
5733 5732 zfs_ioc_clear_fault, zfs_secpolicy_inject);
5734 5733 zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_LIST_NEXT,
5735 5734 zfs_ioc_inject_list_next, zfs_secpolicy_inject);
5736 5735
5737 5736 /*
5738 5737 * pool destroy, and export don't log the history as part of
5739 5738 * zfsdev_ioctl, but rather zfs_ioc_pool_export
5740 5739 * does the logging of those commands.
5741 5740 */
5742 5741 zfs_ioctl_register_pool(ZFS_IOC_POOL_DESTROY, zfs_ioc_pool_destroy,
5743 5742 zfs_secpolicy_config, B_FALSE, POOL_CHECK_NONE);
5744 5743 zfs_ioctl_register_pool(ZFS_IOC_POOL_EXPORT, zfs_ioc_pool_export,
5745 5744 zfs_secpolicy_config, B_FALSE, POOL_CHECK_NONE);
5746 5745
5747 5746 zfs_ioctl_register_pool(ZFS_IOC_POOL_STATS, zfs_ioc_pool_stats,
5748 5747 zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE);
5749 5748 zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_PROPS, zfs_ioc_pool_get_props,
5750 5749 zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE);
5751 5750
5752 5751 zfs_ioctl_register_pool(ZFS_IOC_ERROR_LOG, zfs_ioc_error_log,
5753 5752 zfs_secpolicy_inject, B_FALSE, POOL_CHECK_SUSPENDED);
5754 5753 zfs_ioctl_register_pool(ZFS_IOC_DSOBJ_TO_DSNAME,
5755 5754 zfs_ioc_dsobj_to_dsname,
5756 5755 zfs_secpolicy_diff, B_FALSE, POOL_CHECK_SUSPENDED);
5757 5756 zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_HISTORY,
5758 5757 zfs_ioc_pool_get_history,
5759 5758 zfs_secpolicy_config, B_FALSE, POOL_CHECK_SUSPENDED);
5760 5759
5761 5760 zfs_ioctl_register_pool(ZFS_IOC_POOL_IMPORT, zfs_ioc_pool_import,
5762 5761 zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE);
5763 5762
5764 5763 zfs_ioctl_register_pool(ZFS_IOC_CLEAR, zfs_ioc_clear,
5765 5764 zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE);
5766 5765 zfs_ioctl_register_pool(ZFS_IOC_POOL_REOPEN, zfs_ioc_pool_reopen,
5767 5766 zfs_secpolicy_config, B_TRUE, POOL_CHECK_SUSPENDED);
5768 5767
5769 5768 zfs_ioctl_register_dataset_read(ZFS_IOC_SPACE_WRITTEN,
5770 5769 zfs_ioc_space_written);
5771 5770 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_RECVD_PROPS,
5772 5771 zfs_ioc_objset_recvd_props);
5773 5772 zfs_ioctl_register_dataset_read(ZFS_IOC_NEXT_OBJ,
5774 5773 zfs_ioc_next_obj);
5775 5774 zfs_ioctl_register_dataset_read(ZFS_IOC_GET_FSACL,
5776 5775 zfs_ioc_get_fsacl);
5777 5776 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_STATS,
5778 5777 zfs_ioc_objset_stats);
5779 5778 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_ZPLPROPS,
5780 5779 zfs_ioc_objset_zplprops);
5781 5780 zfs_ioctl_register_dataset_read(ZFS_IOC_DATASET_LIST_NEXT,
5782 5781 zfs_ioc_dataset_list_next);
5783 5782 zfs_ioctl_register_dataset_read(ZFS_IOC_SNAPSHOT_LIST_NEXT,
5784 5783 zfs_ioc_snapshot_list_next);
5785 5784 zfs_ioctl_register_dataset_read(ZFS_IOC_SEND_PROGRESS,
5786 5785 zfs_ioc_send_progress);
5787 5786
5788 5787 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_DIFF,
5789 5788 zfs_ioc_diff, zfs_secpolicy_diff);
5790 5789 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_STATS,
5791 5790 zfs_ioc_obj_to_stats, zfs_secpolicy_diff);
5792 5791 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_PATH,
5793 5792 zfs_ioc_obj_to_path, zfs_secpolicy_diff);
5794 5793 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_ONE,
5795 5794 zfs_ioc_userspace_one, zfs_secpolicy_userspace_one);
5796 5795 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_MANY,
5797 5796 zfs_ioc_userspace_many, zfs_secpolicy_userspace_many);
5798 5797 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_SEND,
5799 5798 zfs_ioc_send, zfs_secpolicy_send);
5800 5799
5801 5800 zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_PROP, zfs_ioc_set_prop,
5802 5801 zfs_secpolicy_none);
5803 5802 zfs_ioctl_register_dataset_modify(ZFS_IOC_DESTROY, zfs_ioc_destroy,
5804 5803 zfs_secpolicy_destroy);
5805 5804 zfs_ioctl_register_dataset_modify(ZFS_IOC_RENAME, zfs_ioc_rename,
5806 5805 zfs_secpolicy_rename);
5807 5806 zfs_ioctl_register_dataset_modify(ZFS_IOC_RECV, zfs_ioc_recv,
5808 5807 zfs_secpolicy_recv);
5809 5808 zfs_ioctl_register_dataset_modify(ZFS_IOC_PROMOTE, zfs_ioc_promote,
5810 5809 zfs_secpolicy_promote);
5811 5810 zfs_ioctl_register_dataset_modify(ZFS_IOC_INHERIT_PROP,
5812 5811 zfs_ioc_inherit_prop, zfs_secpolicy_inherit_prop);
5813 5812 zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_FSACL, zfs_ioc_set_fsacl,
5814 5813 zfs_secpolicy_set_fsacl);
5815 5814
5816 5815 zfs_ioctl_register_dataset_nolog(ZFS_IOC_SHARE, zfs_ioc_share,
5817 5816 zfs_secpolicy_share, POOL_CHECK_NONE);
5818 5817 zfs_ioctl_register_dataset_nolog(ZFS_IOC_SMB_ACL, zfs_ioc_smb_acl,
5819 5818 zfs_secpolicy_smb_acl, POOL_CHECK_NONE);
5820 5819 zfs_ioctl_register_dataset_nolog(ZFS_IOC_USERSPACE_UPGRADE,
5821 5820 zfs_ioc_userspace_upgrade, zfs_secpolicy_userspace_upgrade,
5822 5821 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5823 5822 zfs_ioctl_register_dataset_nolog(ZFS_IOC_TMP_SNAPSHOT,
5824 5823 zfs_ioc_tmp_snapshot, zfs_secpolicy_tmp_snapshot,
5825 5824 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5826 5825 }
5827 5826
5828 5827 int
5829 5828 pool_status_check(const char *name, zfs_ioc_namecheck_t type,
5830 5829 zfs_ioc_poolcheck_t check)
5831 5830 {
5832 5831 spa_t *spa;
5833 5832 int error;
5834 5833
5835 5834 ASSERT(type == POOL_NAME || type == DATASET_NAME);
5836 5835
5837 5836 if (check & POOL_CHECK_NONE)
5838 5837 return (0);
5839 5838
5840 5839 error = spa_open(name, &spa, FTAG);
5841 5840 if (error == 0) {
5842 5841 if ((check & POOL_CHECK_SUSPENDED) && spa_suspended(spa))
5843 5842 error = SET_ERROR(EAGAIN);
5844 5843 else if ((check & POOL_CHECK_READONLY) && !spa_writeable(spa))
5845 5844 error = SET_ERROR(EROFS);
5846 5845 spa_close(spa, FTAG);
5847 5846 }
5848 5847 return (error);
5849 5848 }
5850 5849
5851 5850 /*
5852 5851 * Find a free minor number.
5853 5852 */
5854 5853 minor_t
5855 5854 zfsdev_minor_alloc(void)
5856 5855 {
5857 5856 static minor_t last_minor;
5858 5857 minor_t m;
5859 5858
5860 5859 ASSERT(MUTEX_HELD(&zfsdev_state_lock));
5861 5860
5862 5861 for (m = last_minor + 1; m != last_minor; m++) {
5863 5862 if (m > ZFSDEV_MAX_MINOR)
5864 5863 m = 1;
5865 5864 if (ddi_get_soft_state(zfsdev_state, m) == NULL) {
5866 5865 last_minor = m;
5867 5866 return (m);
5868 5867 }
5869 5868 }
5870 5869
5871 5870 return (0);
5872 5871 }
5873 5872
5874 5873 static int
5875 5874 zfs_ctldev_init(dev_t *devp)
5876 5875 {
5877 5876 minor_t minor;
5878 5877 zfs_soft_state_t *zs;
5879 5878
5880 5879 ASSERT(MUTEX_HELD(&zfsdev_state_lock));
5881 5880 ASSERT(getminor(*devp) == 0);
5882 5881
5883 5882 minor = zfsdev_minor_alloc();
5884 5883 if (minor == 0)
5885 5884 return (SET_ERROR(ENXIO));
5886 5885
5887 5886 if (ddi_soft_state_zalloc(zfsdev_state, minor) != DDI_SUCCESS)
5888 5887 return (SET_ERROR(EAGAIN));
5889 5888
5890 5889 *devp = makedevice(getemajor(*devp), minor);
5891 5890
5892 5891 zs = ddi_get_soft_state(zfsdev_state, minor);
5893 5892 zs->zss_type = ZSST_CTLDEV;
5894 5893 zfs_onexit_init((zfs_onexit_t **)&zs->zss_data);
5895 5894
5896 5895 return (0);
5897 5896 }
5898 5897
5899 5898 static void
5900 5899 zfs_ctldev_destroy(zfs_onexit_t *zo, minor_t minor)
5901 5900 {
5902 5901 ASSERT(MUTEX_HELD(&zfsdev_state_lock));
5903 5902
5904 5903 zfs_onexit_destroy(zo);
5905 5904 ddi_soft_state_free(zfsdev_state, minor);
5906 5905 }
5907 5906
5908 5907 void *
5909 5908 zfsdev_get_soft_state(minor_t minor, enum zfs_soft_state_type which)
5910 5909 {
5911 5910 zfs_soft_state_t *zp;
5912 5911
5913 5912 zp = ddi_get_soft_state(zfsdev_state, minor);
5914 5913 if (zp == NULL || zp->zss_type != which)
5915 5914 return (NULL);
5916 5915
5917 5916 return (zp->zss_data);
5918 5917 }
5919 5918
5920 5919 static int
5921 5920 zfsdev_open(dev_t *devp, int flag, int otyp, cred_t *cr)
5922 5921 {
5923 5922 int error = 0;
5924 5923
5925 5924 if (getminor(*devp) != 0)
5926 5925 return (zvol_open(devp, flag, otyp, cr));
5927 5926
5928 5927 /* This is the control device. Allocate a new minor if requested. */
5929 5928 if (flag & FEXCL) {
5930 5929 mutex_enter(&zfsdev_state_lock);
5931 5930 error = zfs_ctldev_init(devp);
5932 5931 mutex_exit(&zfsdev_state_lock);
5933 5932 }
5934 5933
5935 5934 return (error);
5936 5935 }
5937 5936
5938 5937 static int
5939 5938 zfsdev_close(dev_t dev, int flag, int otyp, cred_t *cr)
5940 5939 {
5941 5940 zfs_onexit_t *zo;
5942 5941 minor_t minor = getminor(dev);
5943 5942
5944 5943 if (minor == 0)
5945 5944 return (0);
5946 5945
5947 5946 mutex_enter(&zfsdev_state_lock);
5948 5947 zo = zfsdev_get_soft_state(minor, ZSST_CTLDEV);
5949 5948 if (zo == NULL) {
5950 5949 mutex_exit(&zfsdev_state_lock);
5951 5950 return (zvol_close(dev, flag, otyp, cr));
5952 5951 }
5953 5952 zfs_ctldev_destroy(zo, minor);
5954 5953 mutex_exit(&zfsdev_state_lock);
5955 5954
5956 5955 return (0);
5957 5956 }
5958 5957
5959 5958 static int
5960 5959 zfsdev_ioctl(dev_t dev, int cmd, intptr_t arg, int flag, cred_t *cr, int *rvalp)
5961 5960 {
5962 5961 zfs_cmd_t *zc;
5963 5962 uint_t vecnum;
5964 5963 int error, rc, len;
5965 5964 minor_t minor = getminor(dev);
5966 5965 const zfs_ioc_vec_t *vec;
5967 5966 char *saved_poolname = NULL;
5968 5967 nvlist_t *innvl = NULL;
5969 5968
5970 5969 if (minor != 0 &&
5971 5970 zfsdev_get_soft_state(minor, ZSST_CTLDEV) == NULL)
5972 5971 return (zvol_ioctl(dev, cmd, arg, flag, cr, rvalp));
5973 5972
5974 5973 vecnum = cmd - ZFS_IOC_FIRST;
5975 5974 ASSERT3U(getmajor(dev), ==, ddi_driver_major(zfs_dip));
5976 5975
5977 5976 if (vecnum >= sizeof (zfs_ioc_vec) / sizeof (zfs_ioc_vec[0]))
5978 5977 return (SET_ERROR(EINVAL));
5979 5978 vec = &zfs_ioc_vec[vecnum];
5980 5979
5981 5980 zc = kmem_zalloc(sizeof (zfs_cmd_t), KM_SLEEP);
5982 5981
5983 5982 error = ddi_copyin((void *)arg, zc, sizeof (zfs_cmd_t), flag);
5984 5983 if (error != 0) {
5985 5984 error = SET_ERROR(EFAULT);
5986 5985 goto out;
5987 5986 }
5988 5987
5989 5988 zc->zc_iflags = flag & FKIOCTL;
5990 5989 if (zc->zc_nvlist_src_size != 0) {
5991 5990 error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
5992 5991 zc->zc_iflags, &innvl);
5993 5992 if (error != 0)
5994 5993 goto out;
5995 5994 }
5996 5995
5997 5996 /*
5998 5997 * Ensure that all pool/dataset names are valid before we pass down to
5999 5998 * the lower layers.
6000 5999 */
6001 6000 zc->zc_name[sizeof (zc->zc_name) - 1] = '\0';
6002 6001 switch (vec->zvec_namecheck) {
6003 6002 case POOL_NAME:
6004 6003 if (pool_namecheck(zc->zc_name, NULL, NULL) != 0)
6005 6004 error = SET_ERROR(EINVAL);
6006 6005 else
6007 6006 error = pool_status_check(zc->zc_name,
6008 6007 vec->zvec_namecheck, vec->zvec_pool_check);
6009 6008 break;
6010 6009
6011 6010 case DATASET_NAME:
6012 6011 if (dataset_namecheck(zc->zc_name, NULL, NULL) != 0)
6013 6012 error = SET_ERROR(EINVAL);
6014 6013 else
6015 6014 error = pool_status_check(zc->zc_name,
6016 6015 vec->zvec_namecheck, vec->zvec_pool_check);
6017 6016 break;
6018 6017
6019 6018 case NO_NAME:
6020 6019 break;
6021 6020 }
6022 6021
6023 6022
6024 6023 if (error == 0 && !(flag & FKIOCTL))
6025 6024 error = vec->zvec_secpolicy(zc, innvl, cr);
6026 6025
6027 6026 if (error != 0)
6028 6027 goto out;
6029 6028
6030 6029 /* legacy ioctls can modify zc_name */
6031 6030 len = strcspn(zc->zc_name, "/@#") + 1;
6032 6031 saved_poolname = kmem_alloc(len, KM_SLEEP);
6033 6032 (void) strlcpy(saved_poolname, zc->zc_name, len);
6034 6033
6035 6034 if (vec->zvec_func != NULL) {
6036 6035 nvlist_t *outnvl;
6037 6036 int puterror = 0;
6038 6037 spa_t *spa;
6039 6038 nvlist_t *lognv = NULL;
6040 6039
6041 6040 ASSERT(vec->zvec_legacy_func == NULL);
6042 6041
6043 6042 /*
6044 6043 * Add the innvl to the lognv before calling the func,
6045 6044 * in case the func changes the innvl.
6046 6045 */
6047 6046 if (vec->zvec_allow_log) {
6048 6047 lognv = fnvlist_alloc();
6049 6048 fnvlist_add_string(lognv, ZPOOL_HIST_IOCTL,
6050 6049 vec->zvec_name);
6051 6050 if (!nvlist_empty(innvl)) {
6052 6051 fnvlist_add_nvlist(lognv, ZPOOL_HIST_INPUT_NVL,
6053 6052 innvl);
6054 6053 }
6055 6054 }
6056 6055
6057 6056 outnvl = fnvlist_alloc();
6058 6057 error = vec->zvec_func(zc->zc_name, innvl, outnvl);
6059 6058
6060 6059 if (error == 0 && vec->zvec_allow_log &&
6061 6060 spa_open(zc->zc_name, &spa, FTAG) == 0) {
6062 6061 if (!nvlist_empty(outnvl)) {
6063 6062 fnvlist_add_nvlist(lognv, ZPOOL_HIST_OUTPUT_NVL,
6064 6063 outnvl);
6065 6064 }
6066 6065 (void) spa_history_log_nvl(spa, lognv);
6067 6066 spa_close(spa, FTAG);
6068 6067 }
6069 6068 fnvlist_free(lognv);
6070 6069
6071 6070 if (!nvlist_empty(outnvl) || zc->zc_nvlist_dst_size != 0) {
6072 6071 int smusherror = 0;
6073 6072 if (vec->zvec_smush_outnvlist) {
6074 6073 smusherror = nvlist_smush(outnvl,
6075 6074 zc->zc_nvlist_dst_size);
6076 6075 }
6077 6076 if (smusherror == 0)
6078 6077 puterror = put_nvlist(zc, outnvl);
6079 6078 }
6080 6079
6081 6080 if (puterror != 0)
6082 6081 error = puterror;
6083 6082
6084 6083 nvlist_free(outnvl);
6085 6084 } else {
6086 6085 error = vec->zvec_legacy_func(zc);
6087 6086 }
6088 6087
6089 6088 out:
6090 6089 nvlist_free(innvl);
6091 6090 rc = ddi_copyout(zc, (void *)arg, sizeof (zfs_cmd_t), flag);
6092 6091 if (error == 0 && rc != 0)
6093 6092 error = SET_ERROR(EFAULT);
6094 6093 if (error == 0 && vec->zvec_allow_log) {
6095 6094 char *s = tsd_get(zfs_allow_log_key);
6096 6095 if (s != NULL)
6097 6096 strfree(s);
6098 6097 (void) tsd_set(zfs_allow_log_key, saved_poolname);
6099 6098 } else {
6100 6099 if (saved_poolname != NULL)
6101 6100 strfree(saved_poolname);
6102 6101 }
6103 6102
6104 6103 kmem_free(zc, sizeof (zfs_cmd_t));
6105 6104 return (error);
6106 6105 }
6107 6106
6108 6107 static int
6109 6108 zfs_attach(dev_info_t *dip, ddi_attach_cmd_t cmd)
6110 6109 {
6111 6110 if (cmd != DDI_ATTACH)
6112 6111 return (DDI_FAILURE);
6113 6112
6114 6113 if (ddi_create_minor_node(dip, "zfs", S_IFCHR, 0,
6115 6114 DDI_PSEUDO, 0) == DDI_FAILURE)
6116 6115 return (DDI_FAILURE);
6117 6116
6118 6117 zfs_dip = dip;
6119 6118
6120 6119 ddi_report_dev(dip);
6121 6120
6122 6121 return (DDI_SUCCESS);
6123 6122 }
6124 6123
6125 6124 static int
6126 6125 zfs_detach(dev_info_t *dip, ddi_detach_cmd_t cmd)
6127 6126 {
6128 6127 if (spa_busy() || zfs_busy() || zvol_busy())
6129 6128 return (DDI_FAILURE);
6130 6129
6131 6130 if (cmd != DDI_DETACH)
6132 6131 return (DDI_FAILURE);
6133 6132
6134 6133 zfs_dip = NULL;
6135 6134
6136 6135 ddi_prop_remove_all(dip);
6137 6136 ddi_remove_minor_node(dip, NULL);
6138 6137
6139 6138 return (DDI_SUCCESS);
6140 6139 }
6141 6140
6142 6141 /*ARGSUSED*/
6143 6142 static int
6144 6143 zfs_info(dev_info_t *dip, ddi_info_cmd_t infocmd, void *arg, void **result)
6145 6144 {
6146 6145 switch (infocmd) {
6147 6146 case DDI_INFO_DEVT2DEVINFO:
6148 6147 *result = zfs_dip;
6149 6148 return (DDI_SUCCESS);
6150 6149
6151 6150 case DDI_INFO_DEVT2INSTANCE:
6152 6151 *result = (void *)0;
6153 6152 return (DDI_SUCCESS);
6154 6153 }
6155 6154
6156 6155 return (DDI_FAILURE);
6157 6156 }
6158 6157
6159 6158 /*
6160 6159 * OK, so this is a little weird.
6161 6160 *
6162 6161 * /dev/zfs is the control node, i.e. minor 0.
6163 6162 * /dev/zvol/[r]dsk/pool/dataset are the zvols, minor > 0.
6164 6163 *
6165 6164 * /dev/zfs has basically nothing to do except serve up ioctls,
6166 6165 * so most of the standard driver entry points are in zvol.c.
6167 6166 */
6168 6167 static struct cb_ops zfs_cb_ops = {
6169 6168 zfsdev_open, /* open */
6170 6169 zfsdev_close, /* close */
6171 6170 zvol_strategy, /* strategy */
6172 6171 nodev, /* print */
6173 6172 zvol_dump, /* dump */
6174 6173 zvol_read, /* read */
6175 6174 zvol_write, /* write */
6176 6175 zfsdev_ioctl, /* ioctl */
6177 6176 nodev, /* devmap */
6178 6177 nodev, /* mmap */
6179 6178 nodev, /* segmap */
6180 6179 nochpoll, /* poll */
6181 6180 ddi_prop_op, /* prop_op */
6182 6181 NULL, /* streamtab */
6183 6182 D_NEW | D_MP | D_64BIT, /* Driver compatibility flag */
6184 6183 CB_REV, /* version */
6185 6184 nodev, /* async read */
6186 6185 nodev, /* async write */
6187 6186 };
6188 6187
6189 6188 static struct dev_ops zfs_dev_ops = {
6190 6189 DEVO_REV, /* version */
6191 6190 0, /* refcnt */
6192 6191 zfs_info, /* info */
6193 6192 nulldev, /* identify */
6194 6193 nulldev, /* probe */
6195 6194 zfs_attach, /* attach */
6196 6195 zfs_detach, /* detach */
6197 6196 nodev, /* reset */
6198 6197 &zfs_cb_ops, /* driver operations */
6199 6198 NULL, /* no bus operations */
6200 6199 NULL, /* power */
6201 6200 ddi_quiesce_not_needed, /* quiesce */
6202 6201 };
6203 6202
6204 6203 static struct modldrv zfs_modldrv = {
6205 6204 &mod_driverops,
6206 6205 "ZFS storage pool",
6207 6206 &zfs_dev_ops
6208 6207 };
6209 6208
6210 6209 static struct modlinkage modlinkage = {
6211 6210 MODREV_1,
6212 6211 (void *)&zfs_modlfs,
6213 6212 (void *)&zfs_modldrv,
6214 6213 NULL
6215 6214 };
6216 6215
6217 6216 static void
6218 6217 zfs_allow_log_destroy(void *arg)
6219 6218 {
6220 6219 char *poolname = arg;
6221 6220 strfree(poolname);
6222 6221 }
6223 6222
6224 6223 int
6225 6224 _init(void)
6226 6225 {
6227 6226 int error;
6228 6227
6229 6228 spa_init(FREAD | FWRITE);
6230 6229 zfs_init();
6231 6230 zvol_init();
6232 6231 zfs_ioctl_init();
6233 6232
6234 6233 if ((error = mod_install(&modlinkage)) != 0) {
6235 6234 zvol_fini();
6236 6235 zfs_fini();
6237 6236 spa_fini();
6238 6237 return (error);
6239 6238 }
6240 6239
6241 6240 tsd_create(&zfs_fsyncer_key, NULL);
6242 6241 tsd_create(&rrw_tsd_key, rrw_tsd_destroy);
6243 6242 tsd_create(&zfs_allow_log_key, zfs_allow_log_destroy);
6244 6243
6245 6244 error = ldi_ident_from_mod(&modlinkage, &zfs_li);
6246 6245 ASSERT(error == 0);
6247 6246 mutex_init(&zfs_share_lock, NULL, MUTEX_DEFAULT, NULL);
6248 6247
6249 6248 return (0);
6250 6249 }
6251 6250
6252 6251 int
6253 6252 _fini(void)
6254 6253 {
6255 6254 int error;
6256 6255
6257 6256 if (spa_busy() || zfs_busy() || zvol_busy() || zio_injection_enabled)
6258 6257 return (SET_ERROR(EBUSY));
6259 6258
6260 6259 if ((error = mod_remove(&modlinkage)) != 0)
6261 6260 return (error);
6262 6261
6263 6262 zvol_fini();
6264 6263 zfs_fini();
6265 6264 spa_fini();
6266 6265 if (zfs_nfsshare_inited)
6267 6266 (void) ddi_modclose(nfs_mod);
6268 6267 if (zfs_smbshare_inited)
6269 6268 (void) ddi_modclose(smbsrv_mod);
6270 6269 if (zfs_nfsshare_inited || zfs_smbshare_inited)
6271 6270 (void) ddi_modclose(sharefs_mod);
6272 6271
6273 6272 tsd_destroy(&zfs_fsyncer_key);
6274 6273 ldi_ident_release(zfs_li);
6275 6274 zfs_li = NULL;
6276 6275 mutex_destroy(&zfs_share_lock);
6277 6276
6278 6277 return (error);
6279 6278 }
6280 6279
6281 6280 int
6282 6281 _info(struct modinfo *modinfop)
6283 6282 {
6284 6283 return (mod_info(&modlinkage, modinfop));
6285 6284 }
↓ open down ↓ |
4713 lines elided |
↑ open up ↑ |
XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX