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