Print this page
3973 zfs_ioc_rename alters passed in zc->zc_name
Reviewed by: Matthew Ahrens <mahrens@delphix.com>
| Split |
Close |
| Expand all |
| Collapse all |
--- old/usr/src/uts/common/fs/zfs/zfs_ioctl.c
+++ new/usr/src/uts/common/fs/zfs/zfs_ioctl.c
1 1 /*
2 2 * CDDL HEADER START
3 3 *
4 4 * The contents of this file are subject to the terms of the
5 5 * Common Development and Distribution License (the "License").
6 6 * You may not use this file except in compliance with the License.
7 7 *
8 8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9 9 * or http://www.opensolaris.org/os/licensing.
10 10 * See the License for the specific language governing permissions
11 11 * and limitations under the License.
12 12 *
13 13 * When distributing Covered Code, include this CDDL HEADER in each
14 14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15 15 * If applicable, add the following below this CDDL HEADER, with the
16 16 * fields enclosed by brackets "[]" replaced with your own identifying
17 17 * information: Portions Copyright [yyyy] [name of copyright owner]
18 18 *
19 19 * CDDL HEADER END
20 20 */
21 21
22 22 /*
23 23 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
24 24 * Portions Copyright 2011 Martin Matuska
25 25 * Copyright 2011 Nexenta Systems, Inc. All rights reserved.
26 26 * Copyright (c) 2012, Joyent, Inc. All rights reserved.
27 27 * Copyright (c) 2013 by Delphix. All rights reserved.
28 28 * Copyright (c) 2013 by Saso Kiselkov. All rights reserved.
29 29 * Copyright (c) 2013 Steven Hartland. All rights reserved.
30 30 */
31 31
32 32 /*
33 33 * ZFS ioctls.
34 34 *
35 35 * This file handles the ioctls to /dev/zfs, used for configuring ZFS storage
36 36 * pools and filesystems, e.g. with /sbin/zfs and /sbin/zpool.
37 37 *
38 38 * There are two ways that we handle ioctls: the legacy way where almost
39 39 * all of the logic is in the ioctl callback, and the new way where most
40 40 * of the marshalling is handled in the common entry point, zfsdev_ioctl().
41 41 *
42 42 * Non-legacy ioctls should be registered by calling
43 43 * zfs_ioctl_register() from zfs_ioctl_init(). The ioctl is invoked
44 44 * from userland by lzc_ioctl().
45 45 *
46 46 * The registration arguments are as follows:
47 47 *
48 48 * const char *name
49 49 * The name of the ioctl. This is used for history logging. If the
50 50 * ioctl returns successfully (the callback returns 0), and allow_log
51 51 * is true, then a history log entry will be recorded with the input &
52 52 * output nvlists. The log entry can be printed with "zpool history -i".
53 53 *
54 54 * zfs_ioc_t ioc
55 55 * The ioctl request number, which userland will pass to ioctl(2).
56 56 * The ioctl numbers can change from release to release, because
57 57 * the caller (libzfs) must be matched to the kernel.
58 58 *
59 59 * zfs_secpolicy_func_t *secpolicy
60 60 * This function will be called before the zfs_ioc_func_t, to
61 61 * determine if this operation is permitted. It should return EPERM
62 62 * on failure, and 0 on success. Checks include determining if the
63 63 * dataset is visible in this zone, and if the user has either all
64 64 * zfs privileges in the zone (SYS_MOUNT), or has been granted permission
65 65 * to do this operation on this dataset with "zfs allow".
66 66 *
67 67 * zfs_ioc_namecheck_t namecheck
68 68 * This specifies what to expect in the zfs_cmd_t:zc_name -- a pool
69 69 * name, a dataset name, or nothing. If the name is not well-formed,
70 70 * the ioctl will fail and the callback will not be called.
71 71 * Therefore, the callback can assume that the name is well-formed
72 72 * (e.g. is null-terminated, doesn't have more than one '@' character,
73 73 * doesn't have invalid characters).
74 74 *
75 75 * zfs_ioc_poolcheck_t pool_check
76 76 * This specifies requirements on the pool state. If the pool does
77 77 * not meet them (is suspended or is readonly), the ioctl will fail
78 78 * and the callback will not be called. If any checks are specified
79 79 * (i.e. it is not POOL_CHECK_NONE), namecheck must not be NO_NAME.
80 80 * Multiple checks can be or-ed together (e.g. POOL_CHECK_SUSPENDED |
81 81 * POOL_CHECK_READONLY).
82 82 *
83 83 * boolean_t smush_outnvlist
84 84 * If smush_outnvlist is true, then the output is presumed to be a
85 85 * list of errors, and it will be "smushed" down to fit into the
86 86 * caller's buffer, by removing some entries and replacing them with a
87 87 * single "N_MORE_ERRORS" entry indicating how many were removed. See
88 88 * nvlist_smush() for details. If smush_outnvlist is false, and the
89 89 * outnvlist does not fit into the userland-provided buffer, then the
90 90 * ioctl will fail with ENOMEM.
91 91 *
92 92 * zfs_ioc_func_t *func
93 93 * The callback function that will perform the operation.
94 94 *
95 95 * The callback should return 0 on success, or an error number on
96 96 * failure. If the function fails, the userland ioctl will return -1,
97 97 * and errno will be set to the callback's return value. The callback
98 98 * will be called with the following arguments:
99 99 *
100 100 * const char *name
101 101 * The name of the pool or dataset to operate on, from
102 102 * zfs_cmd_t:zc_name. The 'namecheck' argument specifies the
103 103 * expected type (pool, dataset, or none).
104 104 *
105 105 * nvlist_t *innvl
106 106 * The input nvlist, deserialized from zfs_cmd_t:zc_nvlist_src. Or
107 107 * NULL if no input nvlist was provided. Changes to this nvlist are
108 108 * ignored. If the input nvlist could not be deserialized, the
109 109 * ioctl will fail and the callback will not be called.
110 110 *
111 111 * nvlist_t *outnvl
112 112 * The output nvlist, initially empty. The callback can fill it in,
113 113 * and it will be returned to userland by serializing it into
114 114 * zfs_cmd_t:zc_nvlist_dst. If it is non-empty, and serialization
115 115 * fails (e.g. because the caller didn't supply a large enough
116 116 * buffer), then the overall ioctl will fail. See the
117 117 * 'smush_nvlist' argument above for additional behaviors.
118 118 *
119 119 * There are two typical uses of the output nvlist:
120 120 * - To return state, e.g. property values. In this case,
121 121 * smush_outnvlist should be false. If the buffer was not large
122 122 * enough, the caller will reallocate a larger buffer and try
123 123 * the ioctl again.
124 124 *
125 125 * - To return multiple errors from an ioctl which makes on-disk
126 126 * changes. In this case, smush_outnvlist should be true.
127 127 * Ioctls which make on-disk modifications should generally not
128 128 * use the outnvl if they succeed, because the caller can not
129 129 * distinguish between the operation failing, and
130 130 * deserialization failing.
131 131 */
132 132
133 133 #include <sys/types.h>
134 134 #include <sys/param.h>
135 135 #include <sys/errno.h>
136 136 #include <sys/uio.h>
137 137 #include <sys/buf.h>
138 138 #include <sys/modctl.h>
139 139 #include <sys/open.h>
140 140 #include <sys/file.h>
141 141 #include <sys/kmem.h>
142 142 #include <sys/conf.h>
143 143 #include <sys/cmn_err.h>
144 144 #include <sys/stat.h>
145 145 #include <sys/zfs_ioctl.h>
146 146 #include <sys/zfs_vfsops.h>
147 147 #include <sys/zfs_znode.h>
148 148 #include <sys/zap.h>
149 149 #include <sys/spa.h>
150 150 #include <sys/spa_impl.h>
151 151 #include <sys/vdev.h>
152 152 #include <sys/priv_impl.h>
153 153 #include <sys/dmu.h>
154 154 #include <sys/dsl_dir.h>
155 155 #include <sys/dsl_dataset.h>
156 156 #include <sys/dsl_prop.h>
157 157 #include <sys/dsl_deleg.h>
158 158 #include <sys/dmu_objset.h>
159 159 #include <sys/dmu_impl.h>
160 160 #include <sys/dmu_tx.h>
161 161 #include <sys/ddi.h>
162 162 #include <sys/sunddi.h>
163 163 #include <sys/sunldi.h>
164 164 #include <sys/policy.h>
165 165 #include <sys/zone.h>
166 166 #include <sys/nvpair.h>
167 167 #include <sys/pathname.h>
168 168 #include <sys/mount.h>
169 169 #include <sys/sdt.h>
170 170 #include <sys/fs/zfs.h>
171 171 #include <sys/zfs_ctldir.h>
172 172 #include <sys/zfs_dir.h>
173 173 #include <sys/zfs_onexit.h>
174 174 #include <sys/zvol.h>
175 175 #include <sys/dsl_scan.h>
176 176 #include <sharefs/share.h>
177 177 #include <sys/dmu_objset.h>
178 178 #include <sys/dmu_send.h>
179 179 #include <sys/dsl_destroy.h>
180 180 #include <sys/dsl_userhold.h>
181 181 #include <sys/zfeature.h>
182 182
183 183 #include "zfs_namecheck.h"
184 184 #include "zfs_prop.h"
185 185 #include "zfs_deleg.h"
186 186 #include "zfs_comutil.h"
187 187
188 188 extern struct modlfs zfs_modlfs;
189 189
190 190 extern void zfs_init(void);
191 191 extern void zfs_fini(void);
192 192
193 193 ldi_ident_t zfs_li = NULL;
194 194 dev_info_t *zfs_dip;
195 195
196 196 uint_t zfs_fsyncer_key;
197 197 extern uint_t rrw_tsd_key;
198 198 static uint_t zfs_allow_log_key;
199 199
200 200 typedef int zfs_ioc_legacy_func_t(zfs_cmd_t *);
201 201 typedef int zfs_ioc_func_t(const char *, nvlist_t *, nvlist_t *);
202 202 typedef int zfs_secpolicy_func_t(zfs_cmd_t *, nvlist_t *, cred_t *);
203 203
204 204 typedef enum {
205 205 NO_NAME,
206 206 POOL_NAME,
207 207 DATASET_NAME
208 208 } zfs_ioc_namecheck_t;
209 209
210 210 typedef enum {
211 211 POOL_CHECK_NONE = 1 << 0,
212 212 POOL_CHECK_SUSPENDED = 1 << 1,
213 213 POOL_CHECK_READONLY = 1 << 2,
214 214 } zfs_ioc_poolcheck_t;
215 215
216 216 typedef struct zfs_ioc_vec {
217 217 zfs_ioc_legacy_func_t *zvec_legacy_func;
218 218 zfs_ioc_func_t *zvec_func;
219 219 zfs_secpolicy_func_t *zvec_secpolicy;
220 220 zfs_ioc_namecheck_t zvec_namecheck;
221 221 boolean_t zvec_allow_log;
222 222 zfs_ioc_poolcheck_t zvec_pool_check;
223 223 boolean_t zvec_smush_outnvlist;
224 224 const char *zvec_name;
225 225 } zfs_ioc_vec_t;
226 226
227 227 /* This array is indexed by zfs_userquota_prop_t */
228 228 static const char *userquota_perms[] = {
229 229 ZFS_DELEG_PERM_USERUSED,
230 230 ZFS_DELEG_PERM_USERQUOTA,
231 231 ZFS_DELEG_PERM_GROUPUSED,
232 232 ZFS_DELEG_PERM_GROUPQUOTA,
233 233 };
234 234
235 235 static int zfs_ioc_userspace_upgrade(zfs_cmd_t *zc);
236 236 static int zfs_check_settable(const char *name, nvpair_t *property,
237 237 cred_t *cr);
238 238 static int zfs_check_clearable(char *dataset, nvlist_t *props,
239 239 nvlist_t **errors);
240 240 static int zfs_fill_zplprops_root(uint64_t, nvlist_t *, nvlist_t *,
241 241 boolean_t *);
242 242 int zfs_set_prop_nvlist(const char *, zprop_source_t, nvlist_t *, nvlist_t *);
243 243 static int get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp);
244 244
245 245 static int zfs_prop_activate_feature(spa_t *spa, zfeature_info_t *feature);
246 246
247 247 /* _NOTE(PRINTFLIKE(4)) - this is printf-like, but lint is too whiney */
248 248 void
249 249 __dprintf(const char *file, const char *func, int line, const char *fmt, ...)
250 250 {
251 251 const char *newfile;
252 252 char buf[512];
253 253 va_list adx;
254 254
255 255 /*
256 256 * Get rid of annoying "../common/" prefix to filename.
257 257 */
258 258 newfile = strrchr(file, '/');
259 259 if (newfile != NULL) {
260 260 newfile = newfile + 1; /* Get rid of leading / */
261 261 } else {
262 262 newfile = file;
263 263 }
264 264
265 265 va_start(adx, fmt);
266 266 (void) vsnprintf(buf, sizeof (buf), fmt, adx);
267 267 va_end(adx);
268 268
269 269 /*
270 270 * To get this data, use the zfs-dprintf probe as so:
271 271 * dtrace -q -n 'zfs-dprintf \
272 272 * /stringof(arg0) == "dbuf.c"/ \
273 273 * {printf("%s: %s", stringof(arg1), stringof(arg3))}'
274 274 * arg0 = file name
275 275 * arg1 = function name
276 276 * arg2 = line number
277 277 * arg3 = message
278 278 */
279 279 DTRACE_PROBE4(zfs__dprintf,
280 280 char *, newfile, char *, func, int, line, char *, buf);
281 281 }
282 282
283 283 static void
284 284 history_str_free(char *buf)
285 285 {
286 286 kmem_free(buf, HIS_MAX_RECORD_LEN);
287 287 }
288 288
289 289 static char *
290 290 history_str_get(zfs_cmd_t *zc)
291 291 {
292 292 char *buf;
293 293
294 294 if (zc->zc_history == NULL)
295 295 return (NULL);
296 296
297 297 buf = kmem_alloc(HIS_MAX_RECORD_LEN, KM_SLEEP);
298 298 if (copyinstr((void *)(uintptr_t)zc->zc_history,
299 299 buf, HIS_MAX_RECORD_LEN, NULL) != 0) {
300 300 history_str_free(buf);
301 301 return (NULL);
302 302 }
303 303
304 304 buf[HIS_MAX_RECORD_LEN -1] = '\0';
305 305
306 306 return (buf);
307 307 }
308 308
309 309 /*
310 310 * Check to see if the named dataset is currently defined as bootable
311 311 */
312 312 static boolean_t
313 313 zfs_is_bootfs(const char *name)
314 314 {
315 315 objset_t *os;
316 316
317 317 if (dmu_objset_hold(name, FTAG, &os) == 0) {
318 318 boolean_t ret;
319 319 ret = (dmu_objset_id(os) == spa_bootfs(dmu_objset_spa(os)));
320 320 dmu_objset_rele(os, FTAG);
321 321 return (ret);
322 322 }
323 323 return (B_FALSE);
324 324 }
325 325
326 326 /*
327 327 * Return non-zero if the spa version is less than requested version.
328 328 */
329 329 static int
330 330 zfs_earlier_version(const char *name, int version)
331 331 {
332 332 spa_t *spa;
333 333
334 334 if (spa_open(name, &spa, FTAG) == 0) {
335 335 if (spa_version(spa) < version) {
336 336 spa_close(spa, FTAG);
337 337 return (1);
338 338 }
339 339 spa_close(spa, FTAG);
340 340 }
341 341 return (0);
342 342 }
343 343
344 344 /*
345 345 * Return TRUE if the ZPL version is less than requested version.
346 346 */
347 347 static boolean_t
348 348 zpl_earlier_version(const char *name, int version)
349 349 {
350 350 objset_t *os;
351 351 boolean_t rc = B_TRUE;
352 352
353 353 if (dmu_objset_hold(name, FTAG, &os) == 0) {
354 354 uint64_t zplversion;
355 355
356 356 if (dmu_objset_type(os) != DMU_OST_ZFS) {
357 357 dmu_objset_rele(os, FTAG);
358 358 return (B_TRUE);
359 359 }
360 360 /* XXX reading from non-owned objset */
361 361 if (zfs_get_zplprop(os, ZFS_PROP_VERSION, &zplversion) == 0)
362 362 rc = zplversion < version;
363 363 dmu_objset_rele(os, FTAG);
364 364 }
365 365 return (rc);
366 366 }
367 367
368 368 static void
369 369 zfs_log_history(zfs_cmd_t *zc)
370 370 {
371 371 spa_t *spa;
372 372 char *buf;
373 373
374 374 if ((buf = history_str_get(zc)) == NULL)
375 375 return;
376 376
377 377 if (spa_open(zc->zc_name, &spa, FTAG) == 0) {
378 378 if (spa_version(spa) >= SPA_VERSION_ZPOOL_HISTORY)
379 379 (void) spa_history_log(spa, buf);
380 380 spa_close(spa, FTAG);
381 381 }
382 382 history_str_free(buf);
383 383 }
384 384
385 385 /*
386 386 * Policy for top-level read operations (list pools). Requires no privileges,
387 387 * and can be used in the local zone, as there is no associated dataset.
388 388 */
389 389 /* ARGSUSED */
390 390 static int
391 391 zfs_secpolicy_none(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
392 392 {
393 393 return (0);
394 394 }
395 395
396 396 /*
397 397 * Policy for dataset read operations (list children, get statistics). Requires
398 398 * no privileges, but must be visible in the local zone.
399 399 */
400 400 /* ARGSUSED */
401 401 static int
402 402 zfs_secpolicy_read(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
403 403 {
404 404 if (INGLOBALZONE(curproc) ||
405 405 zone_dataset_visible(zc->zc_name, NULL))
406 406 return (0);
407 407
408 408 return (SET_ERROR(ENOENT));
409 409 }
410 410
411 411 static int
412 412 zfs_dozonecheck_impl(const char *dataset, uint64_t zoned, cred_t *cr)
413 413 {
414 414 int writable = 1;
415 415
416 416 /*
417 417 * The dataset must be visible by this zone -- check this first
418 418 * so they don't see EPERM on something they shouldn't know about.
419 419 */
420 420 if (!INGLOBALZONE(curproc) &&
421 421 !zone_dataset_visible(dataset, &writable))
422 422 return (SET_ERROR(ENOENT));
423 423
424 424 if (INGLOBALZONE(curproc)) {
425 425 /*
426 426 * If the fs is zoned, only root can access it from the
427 427 * global zone.
428 428 */
429 429 if (secpolicy_zfs(cr) && zoned)
430 430 return (SET_ERROR(EPERM));
431 431 } else {
432 432 /*
433 433 * If we are in a local zone, the 'zoned' property must be set.
434 434 */
435 435 if (!zoned)
436 436 return (SET_ERROR(EPERM));
437 437
438 438 /* must be writable by this zone */
439 439 if (!writable)
440 440 return (SET_ERROR(EPERM));
441 441 }
442 442 return (0);
443 443 }
444 444
445 445 static int
446 446 zfs_dozonecheck(const char *dataset, cred_t *cr)
447 447 {
448 448 uint64_t zoned;
449 449
450 450 if (dsl_prop_get_integer(dataset, "zoned", &zoned, NULL))
451 451 return (SET_ERROR(ENOENT));
452 452
453 453 return (zfs_dozonecheck_impl(dataset, zoned, cr));
454 454 }
455 455
456 456 static int
457 457 zfs_dozonecheck_ds(const char *dataset, dsl_dataset_t *ds, cred_t *cr)
458 458 {
459 459 uint64_t zoned;
460 460
461 461 if (dsl_prop_get_int_ds(ds, "zoned", &zoned))
462 462 return (SET_ERROR(ENOENT));
463 463
464 464 return (zfs_dozonecheck_impl(dataset, zoned, cr));
465 465 }
466 466
467 467 static int
468 468 zfs_secpolicy_write_perms_ds(const char *name, dsl_dataset_t *ds,
469 469 const char *perm, cred_t *cr)
470 470 {
471 471 int error;
472 472
473 473 error = zfs_dozonecheck_ds(name, ds, cr);
474 474 if (error == 0) {
475 475 error = secpolicy_zfs(cr);
476 476 if (error != 0)
477 477 error = dsl_deleg_access_impl(ds, perm, cr);
478 478 }
479 479 return (error);
480 480 }
481 481
482 482 static int
483 483 zfs_secpolicy_write_perms(const char *name, const char *perm, cred_t *cr)
484 484 {
485 485 int error;
486 486 dsl_dataset_t *ds;
487 487 dsl_pool_t *dp;
488 488
489 489 error = dsl_pool_hold(name, FTAG, &dp);
490 490 if (error != 0)
491 491 return (error);
492 492
493 493 error = dsl_dataset_hold(dp, name, FTAG, &ds);
494 494 if (error != 0) {
495 495 dsl_pool_rele(dp, FTAG);
496 496 return (error);
497 497 }
498 498
499 499 error = zfs_secpolicy_write_perms_ds(name, ds, perm, cr);
500 500
501 501 dsl_dataset_rele(ds, FTAG);
502 502 dsl_pool_rele(dp, FTAG);
503 503 return (error);
504 504 }
505 505
506 506 /*
507 507 * Policy for setting the security label property.
508 508 *
509 509 * Returns 0 for success, non-zero for access and other errors.
510 510 */
511 511 static int
512 512 zfs_set_slabel_policy(const char *name, char *strval, cred_t *cr)
513 513 {
514 514 char ds_hexsl[MAXNAMELEN];
515 515 bslabel_t ds_sl, new_sl;
516 516 boolean_t new_default = FALSE;
517 517 uint64_t zoned;
518 518 int needed_priv = -1;
519 519 int error;
520 520
521 521 /* First get the existing dataset label. */
522 522 error = dsl_prop_get(name, zfs_prop_to_name(ZFS_PROP_MLSLABEL),
523 523 1, sizeof (ds_hexsl), &ds_hexsl, NULL);
524 524 if (error != 0)
525 525 return (SET_ERROR(EPERM));
526 526
527 527 if (strcasecmp(strval, ZFS_MLSLABEL_DEFAULT) == 0)
528 528 new_default = TRUE;
529 529
530 530 /* The label must be translatable */
531 531 if (!new_default && (hexstr_to_label(strval, &new_sl) != 0))
532 532 return (SET_ERROR(EINVAL));
533 533
534 534 /*
535 535 * In a non-global zone, disallow attempts to set a label that
536 536 * doesn't match that of the zone; otherwise no other checks
537 537 * are needed.
538 538 */
539 539 if (!INGLOBALZONE(curproc)) {
540 540 if (new_default || !blequal(&new_sl, CR_SL(CRED())))
541 541 return (SET_ERROR(EPERM));
542 542 return (0);
543 543 }
544 544
545 545 /*
546 546 * For global-zone datasets (i.e., those whose zoned property is
547 547 * "off", verify that the specified new label is valid for the
548 548 * global zone.
549 549 */
550 550 if (dsl_prop_get_integer(name,
551 551 zfs_prop_to_name(ZFS_PROP_ZONED), &zoned, NULL))
552 552 return (SET_ERROR(EPERM));
553 553 if (!zoned) {
554 554 if (zfs_check_global_label(name, strval) != 0)
555 555 return (SET_ERROR(EPERM));
556 556 }
557 557
558 558 /*
559 559 * If the existing dataset label is nondefault, check if the
560 560 * dataset is mounted (label cannot be changed while mounted).
561 561 * Get the zfsvfs; if there isn't one, then the dataset isn't
562 562 * mounted (or isn't a dataset, doesn't exist, ...).
563 563 */
564 564 if (strcasecmp(ds_hexsl, ZFS_MLSLABEL_DEFAULT) != 0) {
565 565 objset_t *os;
566 566 static char *setsl_tag = "setsl_tag";
567 567
568 568 /*
569 569 * Try to own the dataset; abort if there is any error,
570 570 * (e.g., already mounted, in use, or other error).
571 571 */
572 572 error = dmu_objset_own(name, DMU_OST_ZFS, B_TRUE,
573 573 setsl_tag, &os);
574 574 if (error != 0)
575 575 return (SET_ERROR(EPERM));
576 576
577 577 dmu_objset_disown(os, setsl_tag);
578 578
579 579 if (new_default) {
580 580 needed_priv = PRIV_FILE_DOWNGRADE_SL;
581 581 goto out_check;
582 582 }
583 583
584 584 if (hexstr_to_label(strval, &new_sl) != 0)
585 585 return (SET_ERROR(EPERM));
586 586
587 587 if (blstrictdom(&ds_sl, &new_sl))
588 588 needed_priv = PRIV_FILE_DOWNGRADE_SL;
589 589 else if (blstrictdom(&new_sl, &ds_sl))
590 590 needed_priv = PRIV_FILE_UPGRADE_SL;
591 591 } else {
592 592 /* dataset currently has a default label */
593 593 if (!new_default)
594 594 needed_priv = PRIV_FILE_UPGRADE_SL;
595 595 }
596 596
597 597 out_check:
598 598 if (needed_priv != -1)
599 599 return (PRIV_POLICY(cr, needed_priv, B_FALSE, EPERM, NULL));
600 600 return (0);
601 601 }
602 602
603 603 static int
604 604 zfs_secpolicy_setprop(const char *dsname, zfs_prop_t prop, nvpair_t *propval,
605 605 cred_t *cr)
606 606 {
607 607 char *strval;
608 608
609 609 /*
610 610 * Check permissions for special properties.
611 611 */
612 612 switch (prop) {
613 613 case ZFS_PROP_ZONED:
614 614 /*
615 615 * Disallow setting of 'zoned' from within a local zone.
616 616 */
617 617 if (!INGLOBALZONE(curproc))
618 618 return (SET_ERROR(EPERM));
619 619 break;
620 620
621 621 case ZFS_PROP_QUOTA:
622 622 if (!INGLOBALZONE(curproc)) {
623 623 uint64_t zoned;
624 624 char setpoint[MAXNAMELEN];
625 625 /*
626 626 * Unprivileged users are allowed to modify the
627 627 * quota on things *under* (ie. contained by)
628 628 * the thing they own.
629 629 */
630 630 if (dsl_prop_get_integer(dsname, "zoned", &zoned,
631 631 setpoint))
632 632 return (SET_ERROR(EPERM));
633 633 if (!zoned || strlen(dsname) <= strlen(setpoint))
634 634 return (SET_ERROR(EPERM));
635 635 }
636 636 break;
637 637
638 638 case ZFS_PROP_MLSLABEL:
639 639 if (!is_system_labeled())
640 640 return (SET_ERROR(EPERM));
641 641
642 642 if (nvpair_value_string(propval, &strval) == 0) {
643 643 int err;
644 644
645 645 err = zfs_set_slabel_policy(dsname, strval, CRED());
646 646 if (err != 0)
647 647 return (err);
648 648 }
649 649 break;
650 650 }
651 651
652 652 return (zfs_secpolicy_write_perms(dsname, zfs_prop_to_name(prop), cr));
653 653 }
654 654
655 655 /* ARGSUSED */
656 656 static int
657 657 zfs_secpolicy_set_fsacl(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
658 658 {
659 659 int error;
660 660
661 661 error = zfs_dozonecheck(zc->zc_name, cr);
662 662 if (error != 0)
663 663 return (error);
664 664
665 665 /*
666 666 * permission to set permissions will be evaluated later in
667 667 * dsl_deleg_can_allow()
668 668 */
669 669 return (0);
670 670 }
671 671
672 672 /* ARGSUSED */
673 673 static int
674 674 zfs_secpolicy_rollback(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
675 675 {
676 676 return (zfs_secpolicy_write_perms(zc->zc_name,
677 677 ZFS_DELEG_PERM_ROLLBACK, cr));
678 678 }
679 679
680 680 /* ARGSUSED */
681 681 static int
682 682 zfs_secpolicy_send(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
683 683 {
684 684 dsl_pool_t *dp;
685 685 dsl_dataset_t *ds;
686 686 char *cp;
687 687 int error;
688 688
689 689 /*
690 690 * Generate the current snapshot name from the given objsetid, then
691 691 * use that name for the secpolicy/zone checks.
692 692 */
693 693 cp = strchr(zc->zc_name, '@');
694 694 if (cp == NULL)
695 695 return (SET_ERROR(EINVAL));
696 696 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
697 697 if (error != 0)
698 698 return (error);
699 699
700 700 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &ds);
701 701 if (error != 0) {
702 702 dsl_pool_rele(dp, FTAG);
703 703 return (error);
704 704 }
705 705
706 706 dsl_dataset_name(ds, zc->zc_name);
707 707
708 708 error = zfs_secpolicy_write_perms_ds(zc->zc_name, ds,
709 709 ZFS_DELEG_PERM_SEND, cr);
710 710 dsl_dataset_rele(ds, FTAG);
711 711 dsl_pool_rele(dp, FTAG);
712 712
713 713 return (error);
714 714 }
715 715
716 716 /* ARGSUSED */
717 717 static int
718 718 zfs_secpolicy_send_new(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
719 719 {
720 720 return (zfs_secpolicy_write_perms(zc->zc_name,
721 721 ZFS_DELEG_PERM_SEND, cr));
722 722 }
723 723
724 724 /* ARGSUSED */
725 725 static int
726 726 zfs_secpolicy_deleg_share(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
727 727 {
728 728 vnode_t *vp;
729 729 int error;
730 730
731 731 if ((error = lookupname(zc->zc_value, UIO_SYSSPACE,
732 732 NO_FOLLOW, NULL, &vp)) != 0)
733 733 return (error);
734 734
735 735 /* Now make sure mntpnt and dataset are ZFS */
736 736
737 737 if (vp->v_vfsp->vfs_fstype != zfsfstype ||
738 738 (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource),
739 739 zc->zc_name) != 0)) {
740 740 VN_RELE(vp);
741 741 return (SET_ERROR(EPERM));
742 742 }
743 743
744 744 VN_RELE(vp);
745 745 return (dsl_deleg_access(zc->zc_name,
746 746 ZFS_DELEG_PERM_SHARE, cr));
747 747 }
748 748
749 749 int
750 750 zfs_secpolicy_share(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
751 751 {
752 752 if (!INGLOBALZONE(curproc))
753 753 return (SET_ERROR(EPERM));
754 754
755 755 if (secpolicy_nfs(cr) == 0) {
756 756 return (0);
757 757 } else {
758 758 return (zfs_secpolicy_deleg_share(zc, innvl, cr));
759 759 }
760 760 }
761 761
762 762 int
763 763 zfs_secpolicy_smb_acl(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
764 764 {
765 765 if (!INGLOBALZONE(curproc))
766 766 return (SET_ERROR(EPERM));
767 767
768 768 if (secpolicy_smb(cr) == 0) {
769 769 return (0);
770 770 } else {
771 771 return (zfs_secpolicy_deleg_share(zc, innvl, cr));
772 772 }
773 773 }
774 774
775 775 static int
776 776 zfs_get_parent(const char *datasetname, char *parent, int parentsize)
777 777 {
778 778 char *cp;
779 779
780 780 /*
781 781 * Remove the @bla or /bla from the end of the name to get the parent.
782 782 */
783 783 (void) strncpy(parent, datasetname, parentsize);
784 784 cp = strrchr(parent, '@');
785 785 if (cp != NULL) {
786 786 cp[0] = '\0';
787 787 } else {
788 788 cp = strrchr(parent, '/');
789 789 if (cp == NULL)
790 790 return (SET_ERROR(ENOENT));
791 791 cp[0] = '\0';
792 792 }
793 793
794 794 return (0);
795 795 }
796 796
797 797 int
798 798 zfs_secpolicy_destroy_perms(const char *name, cred_t *cr)
799 799 {
800 800 int error;
801 801
802 802 if ((error = zfs_secpolicy_write_perms(name,
803 803 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
804 804 return (error);
805 805
806 806 return (zfs_secpolicy_write_perms(name, ZFS_DELEG_PERM_DESTROY, cr));
807 807 }
808 808
809 809 /* ARGSUSED */
810 810 static int
811 811 zfs_secpolicy_destroy(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
812 812 {
813 813 return (zfs_secpolicy_destroy_perms(zc->zc_name, cr));
814 814 }
815 815
816 816 /*
817 817 * Destroying snapshots with delegated permissions requires
818 818 * descendant mount and destroy permissions.
819 819 */
820 820 /* ARGSUSED */
821 821 static int
822 822 zfs_secpolicy_destroy_snaps(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
823 823 {
824 824 nvlist_t *snaps;
825 825 nvpair_t *pair, *nextpair;
826 826 int error = 0;
827 827
828 828 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
829 829 return (SET_ERROR(EINVAL));
830 830 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
831 831 pair = nextpair) {
832 832 dsl_pool_t *dp;
833 833 dsl_dataset_t *ds;
834 834
835 835 error = dsl_pool_hold(nvpair_name(pair), FTAG, &dp);
836 836 if (error != 0)
837 837 break;
838 838 nextpair = nvlist_next_nvpair(snaps, pair);
839 839 error = dsl_dataset_hold(dp, nvpair_name(pair), FTAG, &ds);
840 840 if (error == 0)
841 841 dsl_dataset_rele(ds, FTAG);
842 842 dsl_pool_rele(dp, FTAG);
843 843
844 844 if (error == 0) {
845 845 error = zfs_secpolicy_destroy_perms(nvpair_name(pair),
846 846 cr);
847 847 } else if (error == ENOENT) {
848 848 /*
849 849 * Ignore any snapshots that don't exist (we consider
850 850 * them "already destroyed"). Remove the name from the
851 851 * nvl here in case the snapshot is created between
852 852 * now and when we try to destroy it (in which case
853 853 * we don't want to destroy it since we haven't
854 854 * checked for permission).
855 855 */
856 856 fnvlist_remove_nvpair(snaps, pair);
857 857 error = 0;
858 858 }
859 859 if (error != 0)
860 860 break;
861 861 }
862 862
863 863 return (error);
864 864 }
865 865
866 866 int
867 867 zfs_secpolicy_rename_perms(const char *from, const char *to, cred_t *cr)
868 868 {
869 869 char parentname[MAXNAMELEN];
870 870 int error;
871 871
872 872 if ((error = zfs_secpolicy_write_perms(from,
873 873 ZFS_DELEG_PERM_RENAME, cr)) != 0)
874 874 return (error);
875 875
876 876 if ((error = zfs_secpolicy_write_perms(from,
877 877 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
878 878 return (error);
879 879
880 880 if ((error = zfs_get_parent(to, parentname,
881 881 sizeof (parentname))) != 0)
882 882 return (error);
883 883
884 884 if ((error = zfs_secpolicy_write_perms(parentname,
885 885 ZFS_DELEG_PERM_CREATE, cr)) != 0)
886 886 return (error);
887 887
888 888 if ((error = zfs_secpolicy_write_perms(parentname,
889 889 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
890 890 return (error);
891 891
892 892 return (error);
893 893 }
894 894
895 895 /* ARGSUSED */
896 896 static int
897 897 zfs_secpolicy_rename(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
898 898 {
899 899 return (zfs_secpolicy_rename_perms(zc->zc_name, zc->zc_value, cr));
900 900 }
901 901
902 902 /* ARGSUSED */
903 903 static int
904 904 zfs_secpolicy_promote(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
905 905 {
906 906 dsl_pool_t *dp;
907 907 dsl_dataset_t *clone;
908 908 int error;
909 909
910 910 error = zfs_secpolicy_write_perms(zc->zc_name,
911 911 ZFS_DELEG_PERM_PROMOTE, cr);
912 912 if (error != 0)
913 913 return (error);
914 914
915 915 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
916 916 if (error != 0)
917 917 return (error);
918 918
919 919 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &clone);
920 920
921 921 if (error == 0) {
922 922 char parentname[MAXNAMELEN];
923 923 dsl_dataset_t *origin = NULL;
924 924 dsl_dir_t *dd;
925 925 dd = clone->ds_dir;
926 926
927 927 error = dsl_dataset_hold_obj(dd->dd_pool,
928 928 dd->dd_phys->dd_origin_obj, FTAG, &origin);
929 929 if (error != 0) {
930 930 dsl_dataset_rele(clone, FTAG);
931 931 dsl_pool_rele(dp, FTAG);
932 932 return (error);
933 933 }
934 934
935 935 error = zfs_secpolicy_write_perms_ds(zc->zc_name, clone,
936 936 ZFS_DELEG_PERM_MOUNT, cr);
937 937
938 938 dsl_dataset_name(origin, parentname);
939 939 if (error == 0) {
940 940 error = zfs_secpolicy_write_perms_ds(parentname, origin,
941 941 ZFS_DELEG_PERM_PROMOTE, cr);
942 942 }
943 943 dsl_dataset_rele(clone, FTAG);
944 944 dsl_dataset_rele(origin, FTAG);
945 945 }
946 946 dsl_pool_rele(dp, FTAG);
947 947 return (error);
948 948 }
949 949
950 950 /* ARGSUSED */
951 951 static int
952 952 zfs_secpolicy_recv(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
953 953 {
954 954 int error;
955 955
956 956 if ((error = zfs_secpolicy_write_perms(zc->zc_name,
957 957 ZFS_DELEG_PERM_RECEIVE, cr)) != 0)
958 958 return (error);
959 959
960 960 if ((error = zfs_secpolicy_write_perms(zc->zc_name,
961 961 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
962 962 return (error);
963 963
964 964 return (zfs_secpolicy_write_perms(zc->zc_name,
965 965 ZFS_DELEG_PERM_CREATE, cr));
966 966 }
967 967
968 968 int
969 969 zfs_secpolicy_snapshot_perms(const char *name, cred_t *cr)
970 970 {
971 971 return (zfs_secpolicy_write_perms(name,
972 972 ZFS_DELEG_PERM_SNAPSHOT, cr));
973 973 }
974 974
975 975 /*
976 976 * Check for permission to create each snapshot in the nvlist.
977 977 */
978 978 /* ARGSUSED */
979 979 static int
980 980 zfs_secpolicy_snapshot(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
981 981 {
982 982 nvlist_t *snaps;
983 983 int error = 0;
984 984 nvpair_t *pair;
985 985
986 986 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
987 987 return (SET_ERROR(EINVAL));
988 988 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
989 989 pair = nvlist_next_nvpair(snaps, pair)) {
990 990 char *name = nvpair_name(pair);
991 991 char *atp = strchr(name, '@');
992 992
993 993 if (atp == NULL) {
994 994 error = SET_ERROR(EINVAL);
995 995 break;
996 996 }
997 997 *atp = '\0';
998 998 error = zfs_secpolicy_snapshot_perms(name, cr);
999 999 *atp = '@';
1000 1000 if (error != 0)
1001 1001 break;
1002 1002 }
1003 1003 return (error);
1004 1004 }
1005 1005
1006 1006 /* ARGSUSED */
1007 1007 static int
1008 1008 zfs_secpolicy_log_history(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1009 1009 {
1010 1010 /*
1011 1011 * Even root must have a proper TSD so that we know what pool
1012 1012 * to log to.
1013 1013 */
1014 1014 if (tsd_get(zfs_allow_log_key) == NULL)
1015 1015 return (SET_ERROR(EPERM));
1016 1016 return (0);
1017 1017 }
1018 1018
1019 1019 static int
1020 1020 zfs_secpolicy_create_clone(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1021 1021 {
1022 1022 char parentname[MAXNAMELEN];
1023 1023 int error;
1024 1024 char *origin;
1025 1025
1026 1026 if ((error = zfs_get_parent(zc->zc_name, parentname,
1027 1027 sizeof (parentname))) != 0)
1028 1028 return (error);
1029 1029
1030 1030 if (nvlist_lookup_string(innvl, "origin", &origin) == 0 &&
1031 1031 (error = zfs_secpolicy_write_perms(origin,
1032 1032 ZFS_DELEG_PERM_CLONE, cr)) != 0)
1033 1033 return (error);
1034 1034
1035 1035 if ((error = zfs_secpolicy_write_perms(parentname,
1036 1036 ZFS_DELEG_PERM_CREATE, cr)) != 0)
1037 1037 return (error);
1038 1038
1039 1039 return (zfs_secpolicy_write_perms(parentname,
1040 1040 ZFS_DELEG_PERM_MOUNT, cr));
1041 1041 }
1042 1042
1043 1043 /*
1044 1044 * Policy for pool operations - create/destroy pools, add vdevs, etc. Requires
1045 1045 * SYS_CONFIG privilege, which is not available in a local zone.
1046 1046 */
1047 1047 /* ARGSUSED */
1048 1048 static int
1049 1049 zfs_secpolicy_config(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1050 1050 {
1051 1051 if (secpolicy_sys_config(cr, B_FALSE) != 0)
1052 1052 return (SET_ERROR(EPERM));
1053 1053
1054 1054 return (0);
1055 1055 }
1056 1056
1057 1057 /*
1058 1058 * Policy for object to name lookups.
1059 1059 */
1060 1060 /* ARGSUSED */
1061 1061 static int
1062 1062 zfs_secpolicy_diff(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1063 1063 {
1064 1064 int error;
1065 1065
1066 1066 if ((error = secpolicy_sys_config(cr, B_FALSE)) == 0)
1067 1067 return (0);
1068 1068
1069 1069 error = zfs_secpolicy_write_perms(zc->zc_name, ZFS_DELEG_PERM_DIFF, cr);
1070 1070 return (error);
1071 1071 }
1072 1072
1073 1073 /*
1074 1074 * Policy for fault injection. Requires all privileges.
1075 1075 */
1076 1076 /* ARGSUSED */
1077 1077 static int
1078 1078 zfs_secpolicy_inject(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1079 1079 {
1080 1080 return (secpolicy_zinject(cr));
1081 1081 }
1082 1082
1083 1083 /* ARGSUSED */
1084 1084 static int
1085 1085 zfs_secpolicy_inherit_prop(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1086 1086 {
1087 1087 zfs_prop_t prop = zfs_name_to_prop(zc->zc_value);
1088 1088
1089 1089 if (prop == ZPROP_INVAL) {
1090 1090 if (!zfs_prop_user(zc->zc_value))
1091 1091 return (SET_ERROR(EINVAL));
1092 1092 return (zfs_secpolicy_write_perms(zc->zc_name,
1093 1093 ZFS_DELEG_PERM_USERPROP, cr));
1094 1094 } else {
1095 1095 return (zfs_secpolicy_setprop(zc->zc_name, prop,
1096 1096 NULL, cr));
1097 1097 }
1098 1098 }
1099 1099
1100 1100 static int
1101 1101 zfs_secpolicy_userspace_one(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1102 1102 {
1103 1103 int err = zfs_secpolicy_read(zc, innvl, cr);
1104 1104 if (err)
1105 1105 return (err);
1106 1106
1107 1107 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
1108 1108 return (SET_ERROR(EINVAL));
1109 1109
1110 1110 if (zc->zc_value[0] == 0) {
1111 1111 /*
1112 1112 * They are asking about a posix uid/gid. If it's
1113 1113 * themself, allow it.
1114 1114 */
1115 1115 if (zc->zc_objset_type == ZFS_PROP_USERUSED ||
1116 1116 zc->zc_objset_type == ZFS_PROP_USERQUOTA) {
1117 1117 if (zc->zc_guid == crgetuid(cr))
1118 1118 return (0);
1119 1119 } else {
1120 1120 if (groupmember(zc->zc_guid, cr))
1121 1121 return (0);
1122 1122 }
1123 1123 }
1124 1124
1125 1125 return (zfs_secpolicy_write_perms(zc->zc_name,
1126 1126 userquota_perms[zc->zc_objset_type], cr));
1127 1127 }
1128 1128
1129 1129 static int
1130 1130 zfs_secpolicy_userspace_many(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1131 1131 {
1132 1132 int err = zfs_secpolicy_read(zc, innvl, cr);
1133 1133 if (err)
1134 1134 return (err);
1135 1135
1136 1136 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
1137 1137 return (SET_ERROR(EINVAL));
1138 1138
1139 1139 return (zfs_secpolicy_write_perms(zc->zc_name,
1140 1140 userquota_perms[zc->zc_objset_type], cr));
1141 1141 }
1142 1142
1143 1143 /* ARGSUSED */
1144 1144 static int
1145 1145 zfs_secpolicy_userspace_upgrade(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1146 1146 {
1147 1147 return (zfs_secpolicy_setprop(zc->zc_name, ZFS_PROP_VERSION,
1148 1148 NULL, cr));
1149 1149 }
1150 1150
1151 1151 /* ARGSUSED */
1152 1152 static int
1153 1153 zfs_secpolicy_hold(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1154 1154 {
1155 1155 nvpair_t *pair;
1156 1156 nvlist_t *holds;
1157 1157 int error;
1158 1158
1159 1159 error = nvlist_lookup_nvlist(innvl, "holds", &holds);
1160 1160 if (error != 0)
1161 1161 return (SET_ERROR(EINVAL));
1162 1162
1163 1163 for (pair = nvlist_next_nvpair(holds, NULL); pair != NULL;
1164 1164 pair = nvlist_next_nvpair(holds, pair)) {
1165 1165 char fsname[MAXNAMELEN];
1166 1166 error = dmu_fsname(nvpair_name(pair), fsname);
1167 1167 if (error != 0)
1168 1168 return (error);
1169 1169 error = zfs_secpolicy_write_perms(fsname,
1170 1170 ZFS_DELEG_PERM_HOLD, cr);
1171 1171 if (error != 0)
1172 1172 return (error);
1173 1173 }
1174 1174 return (0);
1175 1175 }
1176 1176
1177 1177 /* ARGSUSED */
1178 1178 static int
1179 1179 zfs_secpolicy_release(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1180 1180 {
1181 1181 nvpair_t *pair;
1182 1182 int error;
1183 1183
1184 1184 for (pair = nvlist_next_nvpair(innvl, NULL); pair != NULL;
1185 1185 pair = nvlist_next_nvpair(innvl, pair)) {
1186 1186 char fsname[MAXNAMELEN];
1187 1187 error = dmu_fsname(nvpair_name(pair), fsname);
1188 1188 if (error != 0)
1189 1189 return (error);
1190 1190 error = zfs_secpolicy_write_perms(fsname,
1191 1191 ZFS_DELEG_PERM_RELEASE, cr);
1192 1192 if (error != 0)
1193 1193 return (error);
1194 1194 }
1195 1195 return (0);
1196 1196 }
1197 1197
1198 1198 /*
1199 1199 * Policy for allowing temporary snapshots to be taken or released
1200 1200 */
1201 1201 static int
1202 1202 zfs_secpolicy_tmp_snapshot(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1203 1203 {
1204 1204 /*
1205 1205 * A temporary snapshot is the same as a snapshot,
1206 1206 * hold, destroy and release all rolled into one.
1207 1207 * Delegated diff alone is sufficient that we allow this.
1208 1208 */
1209 1209 int error;
1210 1210
1211 1211 if ((error = zfs_secpolicy_write_perms(zc->zc_name,
1212 1212 ZFS_DELEG_PERM_DIFF, cr)) == 0)
1213 1213 return (0);
1214 1214
1215 1215 error = zfs_secpolicy_snapshot_perms(zc->zc_name, cr);
1216 1216 if (error == 0)
1217 1217 error = zfs_secpolicy_hold(zc, innvl, cr);
1218 1218 if (error == 0)
1219 1219 error = zfs_secpolicy_release(zc, innvl, cr);
1220 1220 if (error == 0)
1221 1221 error = zfs_secpolicy_destroy(zc, innvl, cr);
1222 1222 return (error);
1223 1223 }
1224 1224
1225 1225 /*
1226 1226 * Returns the nvlist as specified by the user in the zfs_cmd_t.
1227 1227 */
1228 1228 static int
1229 1229 get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp)
1230 1230 {
1231 1231 char *packed;
1232 1232 int error;
1233 1233 nvlist_t *list = NULL;
1234 1234
1235 1235 /*
1236 1236 * Read in and unpack the user-supplied nvlist.
1237 1237 */
1238 1238 if (size == 0)
1239 1239 return (SET_ERROR(EINVAL));
1240 1240
1241 1241 packed = kmem_alloc(size, KM_SLEEP);
1242 1242
1243 1243 if ((error = ddi_copyin((void *)(uintptr_t)nvl, packed, size,
1244 1244 iflag)) != 0) {
1245 1245 kmem_free(packed, size);
1246 1246 return (error);
1247 1247 }
1248 1248
1249 1249 if ((error = nvlist_unpack(packed, size, &list, 0)) != 0) {
1250 1250 kmem_free(packed, size);
1251 1251 return (error);
1252 1252 }
1253 1253
1254 1254 kmem_free(packed, size);
1255 1255
1256 1256 *nvp = list;
1257 1257 return (0);
1258 1258 }
1259 1259
1260 1260 /*
1261 1261 * Reduce the size of this nvlist until it can be serialized in 'max' bytes.
1262 1262 * Entries will be removed from the end of the nvlist, and one int32 entry
1263 1263 * named "N_MORE_ERRORS" will be added indicating how many entries were
1264 1264 * removed.
1265 1265 */
1266 1266 static int
1267 1267 nvlist_smush(nvlist_t *errors, size_t max)
1268 1268 {
1269 1269 size_t size;
1270 1270
1271 1271 size = fnvlist_size(errors);
1272 1272
1273 1273 if (size > max) {
1274 1274 nvpair_t *more_errors;
1275 1275 int n = 0;
1276 1276
1277 1277 if (max < 1024)
1278 1278 return (SET_ERROR(ENOMEM));
1279 1279
1280 1280 fnvlist_add_int32(errors, ZPROP_N_MORE_ERRORS, 0);
1281 1281 more_errors = nvlist_prev_nvpair(errors, NULL);
1282 1282
1283 1283 do {
1284 1284 nvpair_t *pair = nvlist_prev_nvpair(errors,
1285 1285 more_errors);
1286 1286 fnvlist_remove_nvpair(errors, pair);
1287 1287 n++;
1288 1288 size = fnvlist_size(errors);
1289 1289 } while (size > max);
1290 1290
1291 1291 fnvlist_remove_nvpair(errors, more_errors);
1292 1292 fnvlist_add_int32(errors, ZPROP_N_MORE_ERRORS, n);
1293 1293 ASSERT3U(fnvlist_size(errors), <=, max);
1294 1294 }
1295 1295
1296 1296 return (0);
1297 1297 }
1298 1298
1299 1299 static int
1300 1300 put_nvlist(zfs_cmd_t *zc, nvlist_t *nvl)
1301 1301 {
1302 1302 char *packed = NULL;
1303 1303 int error = 0;
1304 1304 size_t size;
1305 1305
1306 1306 size = fnvlist_size(nvl);
1307 1307
1308 1308 if (size > zc->zc_nvlist_dst_size) {
1309 1309 error = SET_ERROR(ENOMEM);
1310 1310 } else {
1311 1311 packed = fnvlist_pack(nvl, &size);
1312 1312 if (ddi_copyout(packed, (void *)(uintptr_t)zc->zc_nvlist_dst,
1313 1313 size, zc->zc_iflags) != 0)
1314 1314 error = SET_ERROR(EFAULT);
1315 1315 fnvlist_pack_free(packed, size);
1316 1316 }
1317 1317
1318 1318 zc->zc_nvlist_dst_size = size;
1319 1319 zc->zc_nvlist_dst_filled = B_TRUE;
1320 1320 return (error);
1321 1321 }
1322 1322
1323 1323 static int
1324 1324 getzfsvfs(const char *dsname, zfsvfs_t **zfvp)
1325 1325 {
1326 1326 objset_t *os;
1327 1327 int error;
1328 1328
1329 1329 error = dmu_objset_hold(dsname, FTAG, &os);
1330 1330 if (error != 0)
1331 1331 return (error);
1332 1332 if (dmu_objset_type(os) != DMU_OST_ZFS) {
1333 1333 dmu_objset_rele(os, FTAG);
1334 1334 return (SET_ERROR(EINVAL));
1335 1335 }
1336 1336
1337 1337 mutex_enter(&os->os_user_ptr_lock);
1338 1338 *zfvp = dmu_objset_get_user(os);
1339 1339 if (*zfvp) {
1340 1340 VFS_HOLD((*zfvp)->z_vfs);
1341 1341 } else {
1342 1342 error = SET_ERROR(ESRCH);
1343 1343 }
1344 1344 mutex_exit(&os->os_user_ptr_lock);
1345 1345 dmu_objset_rele(os, FTAG);
1346 1346 return (error);
1347 1347 }
1348 1348
1349 1349 /*
1350 1350 * Find a zfsvfs_t for a mounted filesystem, or create our own, in which
1351 1351 * case its z_vfs will be NULL, and it will be opened as the owner.
1352 1352 * If 'writer' is set, the z_teardown_lock will be held for RW_WRITER,
1353 1353 * which prevents all vnode ops from running.
1354 1354 */
1355 1355 static int
1356 1356 zfsvfs_hold(const char *name, void *tag, zfsvfs_t **zfvp, boolean_t writer)
1357 1357 {
1358 1358 int error = 0;
1359 1359
1360 1360 if (getzfsvfs(name, zfvp) != 0)
1361 1361 error = zfsvfs_create(name, zfvp);
1362 1362 if (error == 0) {
1363 1363 rrw_enter(&(*zfvp)->z_teardown_lock, (writer) ? RW_WRITER :
1364 1364 RW_READER, tag);
1365 1365 if ((*zfvp)->z_unmounted) {
1366 1366 /*
1367 1367 * XXX we could probably try again, since the unmounting
1368 1368 * thread should be just about to disassociate the
1369 1369 * objset from the zfsvfs.
1370 1370 */
1371 1371 rrw_exit(&(*zfvp)->z_teardown_lock, tag);
1372 1372 return (SET_ERROR(EBUSY));
1373 1373 }
1374 1374 }
1375 1375 return (error);
1376 1376 }
1377 1377
1378 1378 static void
1379 1379 zfsvfs_rele(zfsvfs_t *zfsvfs, void *tag)
1380 1380 {
1381 1381 rrw_exit(&zfsvfs->z_teardown_lock, tag);
1382 1382
1383 1383 if (zfsvfs->z_vfs) {
1384 1384 VFS_RELE(zfsvfs->z_vfs);
1385 1385 } else {
1386 1386 dmu_objset_disown(zfsvfs->z_os, zfsvfs);
1387 1387 zfsvfs_free(zfsvfs);
1388 1388 }
1389 1389 }
1390 1390
1391 1391 static int
1392 1392 zfs_ioc_pool_create(zfs_cmd_t *zc)
1393 1393 {
1394 1394 int error;
1395 1395 nvlist_t *config, *props = NULL;
1396 1396 nvlist_t *rootprops = NULL;
1397 1397 nvlist_t *zplprops = NULL;
1398 1398
1399 1399 if (error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1400 1400 zc->zc_iflags, &config))
1401 1401 return (error);
1402 1402
1403 1403 if (zc->zc_nvlist_src_size != 0 && (error =
1404 1404 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1405 1405 zc->zc_iflags, &props))) {
1406 1406 nvlist_free(config);
1407 1407 return (error);
1408 1408 }
1409 1409
1410 1410 if (props) {
1411 1411 nvlist_t *nvl = NULL;
1412 1412 uint64_t version = SPA_VERSION;
1413 1413
1414 1414 (void) nvlist_lookup_uint64(props,
1415 1415 zpool_prop_to_name(ZPOOL_PROP_VERSION), &version);
1416 1416 if (!SPA_VERSION_IS_SUPPORTED(version)) {
1417 1417 error = SET_ERROR(EINVAL);
1418 1418 goto pool_props_bad;
1419 1419 }
1420 1420 (void) nvlist_lookup_nvlist(props, ZPOOL_ROOTFS_PROPS, &nvl);
1421 1421 if (nvl) {
1422 1422 error = nvlist_dup(nvl, &rootprops, KM_SLEEP);
1423 1423 if (error != 0) {
1424 1424 nvlist_free(config);
1425 1425 nvlist_free(props);
1426 1426 return (error);
1427 1427 }
1428 1428 (void) nvlist_remove_all(props, ZPOOL_ROOTFS_PROPS);
1429 1429 }
1430 1430 VERIFY(nvlist_alloc(&zplprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
1431 1431 error = zfs_fill_zplprops_root(version, rootprops,
1432 1432 zplprops, NULL);
1433 1433 if (error != 0)
1434 1434 goto pool_props_bad;
1435 1435 }
1436 1436
1437 1437 error = spa_create(zc->zc_name, config, props, zplprops);
1438 1438
1439 1439 /*
1440 1440 * Set the remaining root properties
1441 1441 */
1442 1442 if (!error && (error = zfs_set_prop_nvlist(zc->zc_name,
1443 1443 ZPROP_SRC_LOCAL, rootprops, NULL)) != 0)
1444 1444 (void) spa_destroy(zc->zc_name);
1445 1445
1446 1446 pool_props_bad:
1447 1447 nvlist_free(rootprops);
1448 1448 nvlist_free(zplprops);
1449 1449 nvlist_free(config);
1450 1450 nvlist_free(props);
1451 1451
1452 1452 return (error);
1453 1453 }
1454 1454
1455 1455 static int
1456 1456 zfs_ioc_pool_destroy(zfs_cmd_t *zc)
1457 1457 {
1458 1458 int error;
1459 1459 zfs_log_history(zc);
1460 1460 error = spa_destroy(zc->zc_name);
1461 1461 if (error == 0)
1462 1462 zvol_remove_minors(zc->zc_name);
1463 1463 return (error);
1464 1464 }
1465 1465
1466 1466 static int
1467 1467 zfs_ioc_pool_import(zfs_cmd_t *zc)
1468 1468 {
1469 1469 nvlist_t *config, *props = NULL;
1470 1470 uint64_t guid;
1471 1471 int error;
1472 1472
1473 1473 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1474 1474 zc->zc_iflags, &config)) != 0)
1475 1475 return (error);
1476 1476
1477 1477 if (zc->zc_nvlist_src_size != 0 && (error =
1478 1478 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1479 1479 zc->zc_iflags, &props))) {
1480 1480 nvlist_free(config);
1481 1481 return (error);
1482 1482 }
1483 1483
1484 1484 if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, &guid) != 0 ||
1485 1485 guid != zc->zc_guid)
1486 1486 error = SET_ERROR(EINVAL);
1487 1487 else
1488 1488 error = spa_import(zc->zc_name, config, props, zc->zc_cookie);
1489 1489
1490 1490 if (zc->zc_nvlist_dst != 0) {
1491 1491 int err;
1492 1492
1493 1493 if ((err = put_nvlist(zc, config)) != 0)
1494 1494 error = err;
1495 1495 }
1496 1496
1497 1497 nvlist_free(config);
1498 1498
1499 1499 if (props)
1500 1500 nvlist_free(props);
1501 1501
1502 1502 return (error);
1503 1503 }
1504 1504
1505 1505 static int
1506 1506 zfs_ioc_pool_export(zfs_cmd_t *zc)
1507 1507 {
1508 1508 int error;
1509 1509 boolean_t force = (boolean_t)zc->zc_cookie;
1510 1510 boolean_t hardforce = (boolean_t)zc->zc_guid;
1511 1511
1512 1512 zfs_log_history(zc);
1513 1513 error = spa_export(zc->zc_name, NULL, force, hardforce);
1514 1514 if (error == 0)
1515 1515 zvol_remove_minors(zc->zc_name);
1516 1516 return (error);
1517 1517 }
1518 1518
1519 1519 static int
1520 1520 zfs_ioc_pool_configs(zfs_cmd_t *zc)
1521 1521 {
1522 1522 nvlist_t *configs;
1523 1523 int error;
1524 1524
1525 1525 if ((configs = spa_all_configs(&zc->zc_cookie)) == NULL)
1526 1526 return (SET_ERROR(EEXIST));
1527 1527
1528 1528 error = put_nvlist(zc, configs);
1529 1529
1530 1530 nvlist_free(configs);
1531 1531
1532 1532 return (error);
1533 1533 }
1534 1534
1535 1535 /*
1536 1536 * inputs:
1537 1537 * zc_name name of the pool
1538 1538 *
1539 1539 * outputs:
1540 1540 * zc_cookie real errno
1541 1541 * zc_nvlist_dst config nvlist
1542 1542 * zc_nvlist_dst_size size of config nvlist
1543 1543 */
1544 1544 static int
1545 1545 zfs_ioc_pool_stats(zfs_cmd_t *zc)
1546 1546 {
1547 1547 nvlist_t *config;
1548 1548 int error;
1549 1549 int ret = 0;
1550 1550
1551 1551 error = spa_get_stats(zc->zc_name, &config, zc->zc_value,
1552 1552 sizeof (zc->zc_value));
1553 1553
1554 1554 if (config != NULL) {
1555 1555 ret = put_nvlist(zc, config);
1556 1556 nvlist_free(config);
1557 1557
1558 1558 /*
1559 1559 * The config may be present even if 'error' is non-zero.
1560 1560 * In this case we return success, and preserve the real errno
1561 1561 * in 'zc_cookie'.
1562 1562 */
1563 1563 zc->zc_cookie = error;
1564 1564 } else {
1565 1565 ret = error;
1566 1566 }
1567 1567
1568 1568 return (ret);
1569 1569 }
1570 1570
1571 1571 /*
1572 1572 * Try to import the given pool, returning pool stats as appropriate so that
1573 1573 * user land knows which devices are available and overall pool health.
1574 1574 */
1575 1575 static int
1576 1576 zfs_ioc_pool_tryimport(zfs_cmd_t *zc)
1577 1577 {
1578 1578 nvlist_t *tryconfig, *config;
1579 1579 int error;
1580 1580
1581 1581 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1582 1582 zc->zc_iflags, &tryconfig)) != 0)
1583 1583 return (error);
1584 1584
1585 1585 config = spa_tryimport(tryconfig);
1586 1586
1587 1587 nvlist_free(tryconfig);
1588 1588
1589 1589 if (config == NULL)
1590 1590 return (SET_ERROR(EINVAL));
1591 1591
1592 1592 error = put_nvlist(zc, config);
1593 1593 nvlist_free(config);
1594 1594
1595 1595 return (error);
1596 1596 }
1597 1597
1598 1598 /*
1599 1599 * inputs:
1600 1600 * zc_name name of the pool
1601 1601 * zc_cookie scan func (pool_scan_func_t)
1602 1602 */
1603 1603 static int
1604 1604 zfs_ioc_pool_scan(zfs_cmd_t *zc)
1605 1605 {
1606 1606 spa_t *spa;
1607 1607 int error;
1608 1608
1609 1609 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1610 1610 return (error);
1611 1611
1612 1612 if (zc->zc_cookie == POOL_SCAN_NONE)
1613 1613 error = spa_scan_stop(spa);
1614 1614 else
1615 1615 error = spa_scan(spa, zc->zc_cookie);
1616 1616
1617 1617 spa_close(spa, FTAG);
1618 1618
1619 1619 return (error);
1620 1620 }
1621 1621
1622 1622 static int
1623 1623 zfs_ioc_pool_freeze(zfs_cmd_t *zc)
1624 1624 {
1625 1625 spa_t *spa;
1626 1626 int error;
1627 1627
1628 1628 error = spa_open(zc->zc_name, &spa, FTAG);
1629 1629 if (error == 0) {
1630 1630 spa_freeze(spa);
1631 1631 spa_close(spa, FTAG);
1632 1632 }
1633 1633 return (error);
1634 1634 }
1635 1635
1636 1636 static int
1637 1637 zfs_ioc_pool_upgrade(zfs_cmd_t *zc)
1638 1638 {
1639 1639 spa_t *spa;
1640 1640 int error;
1641 1641
1642 1642 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1643 1643 return (error);
1644 1644
1645 1645 if (zc->zc_cookie < spa_version(spa) ||
1646 1646 !SPA_VERSION_IS_SUPPORTED(zc->zc_cookie)) {
1647 1647 spa_close(spa, FTAG);
1648 1648 return (SET_ERROR(EINVAL));
1649 1649 }
1650 1650
1651 1651 spa_upgrade(spa, zc->zc_cookie);
1652 1652 spa_close(spa, FTAG);
1653 1653
1654 1654 return (error);
1655 1655 }
1656 1656
1657 1657 static int
1658 1658 zfs_ioc_pool_get_history(zfs_cmd_t *zc)
1659 1659 {
1660 1660 spa_t *spa;
1661 1661 char *hist_buf;
1662 1662 uint64_t size;
1663 1663 int error;
1664 1664
1665 1665 if ((size = zc->zc_history_len) == 0)
1666 1666 return (SET_ERROR(EINVAL));
1667 1667
1668 1668 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1669 1669 return (error);
1670 1670
1671 1671 if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) {
1672 1672 spa_close(spa, FTAG);
1673 1673 return (SET_ERROR(ENOTSUP));
1674 1674 }
1675 1675
1676 1676 hist_buf = kmem_alloc(size, KM_SLEEP);
1677 1677 if ((error = spa_history_get(spa, &zc->zc_history_offset,
1678 1678 &zc->zc_history_len, hist_buf)) == 0) {
1679 1679 error = ddi_copyout(hist_buf,
1680 1680 (void *)(uintptr_t)zc->zc_history,
1681 1681 zc->zc_history_len, zc->zc_iflags);
1682 1682 }
1683 1683
1684 1684 spa_close(spa, FTAG);
1685 1685 kmem_free(hist_buf, size);
1686 1686 return (error);
1687 1687 }
1688 1688
1689 1689 static int
1690 1690 zfs_ioc_pool_reguid(zfs_cmd_t *zc)
1691 1691 {
1692 1692 spa_t *spa;
1693 1693 int error;
1694 1694
1695 1695 error = spa_open(zc->zc_name, &spa, FTAG);
1696 1696 if (error == 0) {
1697 1697 error = spa_change_guid(spa);
1698 1698 spa_close(spa, FTAG);
1699 1699 }
1700 1700 return (error);
1701 1701 }
1702 1702
1703 1703 static int
1704 1704 zfs_ioc_dsobj_to_dsname(zfs_cmd_t *zc)
1705 1705 {
1706 1706 return (dsl_dsobj_to_dsname(zc->zc_name, zc->zc_obj, zc->zc_value));
1707 1707 }
1708 1708
1709 1709 /*
1710 1710 * inputs:
1711 1711 * zc_name name of filesystem
1712 1712 * zc_obj object to find
1713 1713 *
1714 1714 * outputs:
1715 1715 * zc_value name of object
1716 1716 */
1717 1717 static int
1718 1718 zfs_ioc_obj_to_path(zfs_cmd_t *zc)
1719 1719 {
1720 1720 objset_t *os;
1721 1721 int error;
1722 1722
1723 1723 /* XXX reading from objset not owned */
1724 1724 if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os)) != 0)
1725 1725 return (error);
1726 1726 if (dmu_objset_type(os) != DMU_OST_ZFS) {
1727 1727 dmu_objset_rele(os, FTAG);
1728 1728 return (SET_ERROR(EINVAL));
1729 1729 }
1730 1730 error = zfs_obj_to_path(os, zc->zc_obj, zc->zc_value,
1731 1731 sizeof (zc->zc_value));
1732 1732 dmu_objset_rele(os, FTAG);
1733 1733
1734 1734 return (error);
1735 1735 }
1736 1736
1737 1737 /*
1738 1738 * inputs:
1739 1739 * zc_name name of filesystem
1740 1740 * zc_obj object to find
1741 1741 *
1742 1742 * outputs:
1743 1743 * zc_stat stats on object
1744 1744 * zc_value path to object
1745 1745 */
1746 1746 static int
1747 1747 zfs_ioc_obj_to_stats(zfs_cmd_t *zc)
1748 1748 {
1749 1749 objset_t *os;
1750 1750 int error;
1751 1751
1752 1752 /* XXX reading from objset not owned */
1753 1753 if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os)) != 0)
1754 1754 return (error);
1755 1755 if (dmu_objset_type(os) != DMU_OST_ZFS) {
1756 1756 dmu_objset_rele(os, FTAG);
1757 1757 return (SET_ERROR(EINVAL));
1758 1758 }
1759 1759 error = zfs_obj_to_stats(os, zc->zc_obj, &zc->zc_stat, zc->zc_value,
1760 1760 sizeof (zc->zc_value));
1761 1761 dmu_objset_rele(os, FTAG);
1762 1762
1763 1763 return (error);
1764 1764 }
1765 1765
1766 1766 static int
1767 1767 zfs_ioc_vdev_add(zfs_cmd_t *zc)
1768 1768 {
1769 1769 spa_t *spa;
1770 1770 int error;
1771 1771 nvlist_t *config, **l2cache, **spares;
1772 1772 uint_t nl2cache = 0, nspares = 0;
1773 1773
1774 1774 error = spa_open(zc->zc_name, &spa, FTAG);
1775 1775 if (error != 0)
1776 1776 return (error);
1777 1777
1778 1778 error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1779 1779 zc->zc_iflags, &config);
1780 1780 (void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_L2CACHE,
1781 1781 &l2cache, &nl2cache);
1782 1782
1783 1783 (void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_SPARES,
1784 1784 &spares, &nspares);
1785 1785
1786 1786 /*
1787 1787 * A root pool with concatenated devices is not supported.
1788 1788 * Thus, can not add a device to a root pool.
1789 1789 *
1790 1790 * Intent log device can not be added to a rootpool because
1791 1791 * during mountroot, zil is replayed, a seperated log device
1792 1792 * can not be accessed during the mountroot time.
1793 1793 *
1794 1794 * l2cache and spare devices are ok to be added to a rootpool.
1795 1795 */
1796 1796 if (spa_bootfs(spa) != 0 && nl2cache == 0 && nspares == 0) {
1797 1797 nvlist_free(config);
1798 1798 spa_close(spa, FTAG);
1799 1799 return (SET_ERROR(EDOM));
1800 1800 }
1801 1801
1802 1802 if (error == 0) {
1803 1803 error = spa_vdev_add(spa, config);
1804 1804 nvlist_free(config);
1805 1805 }
1806 1806 spa_close(spa, FTAG);
1807 1807 return (error);
1808 1808 }
1809 1809
1810 1810 /*
1811 1811 * inputs:
1812 1812 * zc_name name of the pool
1813 1813 * zc_nvlist_conf nvlist of devices to remove
1814 1814 * zc_cookie to stop the remove?
1815 1815 */
1816 1816 static int
1817 1817 zfs_ioc_vdev_remove(zfs_cmd_t *zc)
1818 1818 {
1819 1819 spa_t *spa;
1820 1820 int error;
1821 1821
1822 1822 error = spa_open(zc->zc_name, &spa, FTAG);
1823 1823 if (error != 0)
1824 1824 return (error);
1825 1825 error = spa_vdev_remove(spa, zc->zc_guid, B_FALSE);
1826 1826 spa_close(spa, FTAG);
1827 1827 return (error);
1828 1828 }
1829 1829
1830 1830 static int
1831 1831 zfs_ioc_vdev_set_state(zfs_cmd_t *zc)
1832 1832 {
1833 1833 spa_t *spa;
1834 1834 int error;
1835 1835 vdev_state_t newstate = VDEV_STATE_UNKNOWN;
1836 1836
1837 1837 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1838 1838 return (error);
1839 1839 switch (zc->zc_cookie) {
1840 1840 case VDEV_STATE_ONLINE:
1841 1841 error = vdev_online(spa, zc->zc_guid, zc->zc_obj, &newstate);
1842 1842 break;
1843 1843
1844 1844 case VDEV_STATE_OFFLINE:
1845 1845 error = vdev_offline(spa, zc->zc_guid, zc->zc_obj);
1846 1846 break;
1847 1847
1848 1848 case VDEV_STATE_FAULTED:
1849 1849 if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED &&
1850 1850 zc->zc_obj != VDEV_AUX_EXTERNAL)
1851 1851 zc->zc_obj = VDEV_AUX_ERR_EXCEEDED;
1852 1852
1853 1853 error = vdev_fault(spa, zc->zc_guid, zc->zc_obj);
1854 1854 break;
1855 1855
1856 1856 case VDEV_STATE_DEGRADED:
1857 1857 if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED &&
1858 1858 zc->zc_obj != VDEV_AUX_EXTERNAL)
1859 1859 zc->zc_obj = VDEV_AUX_ERR_EXCEEDED;
1860 1860
1861 1861 error = vdev_degrade(spa, zc->zc_guid, zc->zc_obj);
1862 1862 break;
1863 1863
1864 1864 default:
1865 1865 error = SET_ERROR(EINVAL);
1866 1866 }
1867 1867 zc->zc_cookie = newstate;
1868 1868 spa_close(spa, FTAG);
1869 1869 return (error);
1870 1870 }
1871 1871
1872 1872 static int
1873 1873 zfs_ioc_vdev_attach(zfs_cmd_t *zc)
1874 1874 {
1875 1875 spa_t *spa;
1876 1876 int replacing = zc->zc_cookie;
1877 1877 nvlist_t *config;
1878 1878 int error;
1879 1879
1880 1880 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1881 1881 return (error);
1882 1882
1883 1883 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1884 1884 zc->zc_iflags, &config)) == 0) {
1885 1885 error = spa_vdev_attach(spa, zc->zc_guid, config, replacing);
1886 1886 nvlist_free(config);
1887 1887 }
1888 1888
1889 1889 spa_close(spa, FTAG);
1890 1890 return (error);
1891 1891 }
1892 1892
1893 1893 static int
1894 1894 zfs_ioc_vdev_detach(zfs_cmd_t *zc)
1895 1895 {
1896 1896 spa_t *spa;
1897 1897 int error;
1898 1898
1899 1899 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1900 1900 return (error);
1901 1901
1902 1902 error = spa_vdev_detach(spa, zc->zc_guid, 0, B_FALSE);
1903 1903
1904 1904 spa_close(spa, FTAG);
1905 1905 return (error);
1906 1906 }
1907 1907
1908 1908 static int
1909 1909 zfs_ioc_vdev_split(zfs_cmd_t *zc)
1910 1910 {
1911 1911 spa_t *spa;
1912 1912 nvlist_t *config, *props = NULL;
1913 1913 int error;
1914 1914 boolean_t exp = !!(zc->zc_cookie & ZPOOL_EXPORT_AFTER_SPLIT);
1915 1915
1916 1916 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1917 1917 return (error);
1918 1918
1919 1919 if (error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1920 1920 zc->zc_iflags, &config)) {
1921 1921 spa_close(spa, FTAG);
1922 1922 return (error);
1923 1923 }
1924 1924
1925 1925 if (zc->zc_nvlist_src_size != 0 && (error =
1926 1926 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1927 1927 zc->zc_iflags, &props))) {
1928 1928 spa_close(spa, FTAG);
1929 1929 nvlist_free(config);
1930 1930 return (error);
1931 1931 }
1932 1932
1933 1933 error = spa_vdev_split_mirror(spa, zc->zc_string, config, props, exp);
1934 1934
1935 1935 spa_close(spa, FTAG);
1936 1936
1937 1937 nvlist_free(config);
1938 1938 nvlist_free(props);
1939 1939
1940 1940 return (error);
1941 1941 }
1942 1942
1943 1943 static int
1944 1944 zfs_ioc_vdev_setpath(zfs_cmd_t *zc)
1945 1945 {
1946 1946 spa_t *spa;
1947 1947 char *path = zc->zc_value;
1948 1948 uint64_t guid = zc->zc_guid;
1949 1949 int error;
1950 1950
1951 1951 error = spa_open(zc->zc_name, &spa, FTAG);
1952 1952 if (error != 0)
1953 1953 return (error);
1954 1954
1955 1955 error = spa_vdev_setpath(spa, guid, path);
1956 1956 spa_close(spa, FTAG);
1957 1957 return (error);
1958 1958 }
1959 1959
1960 1960 static int
1961 1961 zfs_ioc_vdev_setfru(zfs_cmd_t *zc)
1962 1962 {
1963 1963 spa_t *spa;
1964 1964 char *fru = zc->zc_value;
1965 1965 uint64_t guid = zc->zc_guid;
1966 1966 int error;
1967 1967
1968 1968 error = spa_open(zc->zc_name, &spa, FTAG);
1969 1969 if (error != 0)
1970 1970 return (error);
1971 1971
1972 1972 error = spa_vdev_setfru(spa, guid, fru);
1973 1973 spa_close(spa, FTAG);
1974 1974 return (error);
1975 1975 }
1976 1976
1977 1977 static int
1978 1978 zfs_ioc_objset_stats_impl(zfs_cmd_t *zc, objset_t *os)
1979 1979 {
1980 1980 int error = 0;
1981 1981 nvlist_t *nv;
1982 1982
1983 1983 dmu_objset_fast_stat(os, &zc->zc_objset_stats);
1984 1984
1985 1985 if (zc->zc_nvlist_dst != 0 &&
1986 1986 (error = dsl_prop_get_all(os, &nv)) == 0) {
1987 1987 dmu_objset_stats(os, nv);
1988 1988 /*
1989 1989 * NB: zvol_get_stats() will read the objset contents,
1990 1990 * which we aren't supposed to do with a
1991 1991 * DS_MODE_USER hold, because it could be
1992 1992 * inconsistent. So this is a bit of a workaround...
1993 1993 * XXX reading with out owning
1994 1994 */
1995 1995 if (!zc->zc_objset_stats.dds_inconsistent &&
1996 1996 dmu_objset_type(os) == DMU_OST_ZVOL) {
1997 1997 error = zvol_get_stats(os, nv);
1998 1998 if (error == EIO)
1999 1999 return (error);
2000 2000 VERIFY0(error);
2001 2001 }
2002 2002 error = put_nvlist(zc, nv);
2003 2003 nvlist_free(nv);
2004 2004 }
2005 2005
2006 2006 return (error);
2007 2007 }
2008 2008
2009 2009 /*
2010 2010 * inputs:
2011 2011 * zc_name name of filesystem
2012 2012 * zc_nvlist_dst_size size of buffer for property nvlist
2013 2013 *
2014 2014 * outputs:
2015 2015 * zc_objset_stats stats
2016 2016 * zc_nvlist_dst property nvlist
2017 2017 * zc_nvlist_dst_size size of property nvlist
2018 2018 */
2019 2019 static int
2020 2020 zfs_ioc_objset_stats(zfs_cmd_t *zc)
2021 2021 {
2022 2022 objset_t *os;
2023 2023 int error;
2024 2024
2025 2025 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
2026 2026 if (error == 0) {
2027 2027 error = zfs_ioc_objset_stats_impl(zc, os);
2028 2028 dmu_objset_rele(os, FTAG);
2029 2029 }
2030 2030
2031 2031 return (error);
2032 2032 }
2033 2033
2034 2034 /*
2035 2035 * inputs:
2036 2036 * zc_name name of filesystem
2037 2037 * zc_nvlist_dst_size size of buffer for property nvlist
2038 2038 *
2039 2039 * outputs:
2040 2040 * zc_nvlist_dst received property nvlist
2041 2041 * zc_nvlist_dst_size size of received property nvlist
2042 2042 *
2043 2043 * Gets received properties (distinct from local properties on or after
2044 2044 * SPA_VERSION_RECVD_PROPS) for callers who want to differentiate received from
2045 2045 * local property values.
2046 2046 */
2047 2047 static int
2048 2048 zfs_ioc_objset_recvd_props(zfs_cmd_t *zc)
2049 2049 {
2050 2050 int error = 0;
2051 2051 nvlist_t *nv;
2052 2052
2053 2053 /*
2054 2054 * Without this check, we would return local property values if the
2055 2055 * caller has not already received properties on or after
2056 2056 * SPA_VERSION_RECVD_PROPS.
2057 2057 */
2058 2058 if (!dsl_prop_get_hasrecvd(zc->zc_name))
2059 2059 return (SET_ERROR(ENOTSUP));
2060 2060
2061 2061 if (zc->zc_nvlist_dst != 0 &&
2062 2062 (error = dsl_prop_get_received(zc->zc_name, &nv)) == 0) {
2063 2063 error = put_nvlist(zc, nv);
2064 2064 nvlist_free(nv);
2065 2065 }
2066 2066
2067 2067 return (error);
2068 2068 }
2069 2069
2070 2070 static int
2071 2071 nvl_add_zplprop(objset_t *os, nvlist_t *props, zfs_prop_t prop)
2072 2072 {
2073 2073 uint64_t value;
2074 2074 int error;
2075 2075
2076 2076 /*
2077 2077 * zfs_get_zplprop() will either find a value or give us
2078 2078 * the default value (if there is one).
2079 2079 */
2080 2080 if ((error = zfs_get_zplprop(os, prop, &value)) != 0)
2081 2081 return (error);
2082 2082 VERIFY(nvlist_add_uint64(props, zfs_prop_to_name(prop), value) == 0);
2083 2083 return (0);
2084 2084 }
2085 2085
2086 2086 /*
2087 2087 * inputs:
2088 2088 * zc_name name of filesystem
2089 2089 * zc_nvlist_dst_size size of buffer for zpl property nvlist
2090 2090 *
2091 2091 * outputs:
2092 2092 * zc_nvlist_dst zpl property nvlist
2093 2093 * zc_nvlist_dst_size size of zpl property nvlist
2094 2094 */
2095 2095 static int
2096 2096 zfs_ioc_objset_zplprops(zfs_cmd_t *zc)
2097 2097 {
2098 2098 objset_t *os;
2099 2099 int err;
2100 2100
2101 2101 /* XXX reading without owning */
2102 2102 if (err = dmu_objset_hold(zc->zc_name, FTAG, &os))
2103 2103 return (err);
2104 2104
2105 2105 dmu_objset_fast_stat(os, &zc->zc_objset_stats);
2106 2106
2107 2107 /*
2108 2108 * NB: nvl_add_zplprop() will read the objset contents,
2109 2109 * which we aren't supposed to do with a DS_MODE_USER
2110 2110 * hold, because it could be inconsistent.
2111 2111 */
2112 2112 if (zc->zc_nvlist_dst != NULL &&
2113 2113 !zc->zc_objset_stats.dds_inconsistent &&
2114 2114 dmu_objset_type(os) == DMU_OST_ZFS) {
2115 2115 nvlist_t *nv;
2116 2116
2117 2117 VERIFY(nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2118 2118 if ((err = nvl_add_zplprop(os, nv, ZFS_PROP_VERSION)) == 0 &&
2119 2119 (err = nvl_add_zplprop(os, nv, ZFS_PROP_NORMALIZE)) == 0 &&
2120 2120 (err = nvl_add_zplprop(os, nv, ZFS_PROP_UTF8ONLY)) == 0 &&
2121 2121 (err = nvl_add_zplprop(os, nv, ZFS_PROP_CASE)) == 0)
2122 2122 err = put_nvlist(zc, nv);
2123 2123 nvlist_free(nv);
2124 2124 } else {
2125 2125 err = SET_ERROR(ENOENT);
2126 2126 }
2127 2127 dmu_objset_rele(os, FTAG);
2128 2128 return (err);
2129 2129 }
2130 2130
2131 2131 static boolean_t
2132 2132 dataset_name_hidden(const char *name)
2133 2133 {
2134 2134 /*
2135 2135 * Skip over datasets that are not visible in this zone,
2136 2136 * internal datasets (which have a $ in their name), and
2137 2137 * temporary datasets (which have a % in their name).
2138 2138 */
2139 2139 if (strchr(name, '$') != NULL)
2140 2140 return (B_TRUE);
2141 2141 if (strchr(name, '%') != NULL)
2142 2142 return (B_TRUE);
2143 2143 if (!INGLOBALZONE(curproc) && !zone_dataset_visible(name, NULL))
2144 2144 return (B_TRUE);
2145 2145 return (B_FALSE);
2146 2146 }
2147 2147
2148 2148 /*
2149 2149 * inputs:
2150 2150 * zc_name name of filesystem
2151 2151 * zc_cookie zap cursor
2152 2152 * zc_nvlist_dst_size size of buffer for property nvlist
2153 2153 *
2154 2154 * outputs:
2155 2155 * zc_name name of next filesystem
2156 2156 * zc_cookie zap cursor
2157 2157 * zc_objset_stats stats
2158 2158 * zc_nvlist_dst property nvlist
2159 2159 * zc_nvlist_dst_size size of property nvlist
2160 2160 */
2161 2161 static int
2162 2162 zfs_ioc_dataset_list_next(zfs_cmd_t *zc)
2163 2163 {
2164 2164 objset_t *os;
2165 2165 int error;
2166 2166 char *p;
2167 2167 size_t orig_len = strlen(zc->zc_name);
2168 2168
2169 2169 top:
2170 2170 if (error = dmu_objset_hold(zc->zc_name, FTAG, &os)) {
2171 2171 if (error == ENOENT)
2172 2172 error = SET_ERROR(ESRCH);
2173 2173 return (error);
2174 2174 }
2175 2175
2176 2176 p = strrchr(zc->zc_name, '/');
2177 2177 if (p == NULL || p[1] != '\0')
2178 2178 (void) strlcat(zc->zc_name, "/", sizeof (zc->zc_name));
2179 2179 p = zc->zc_name + strlen(zc->zc_name);
2180 2180
2181 2181 do {
2182 2182 error = dmu_dir_list_next(os,
2183 2183 sizeof (zc->zc_name) - (p - zc->zc_name), p,
2184 2184 NULL, &zc->zc_cookie);
2185 2185 if (error == ENOENT)
2186 2186 error = SET_ERROR(ESRCH);
2187 2187 } while (error == 0 && dataset_name_hidden(zc->zc_name));
2188 2188 dmu_objset_rele(os, FTAG);
2189 2189
2190 2190 /*
2191 2191 * If it's an internal dataset (ie. with a '$' in its name),
2192 2192 * don't try to get stats for it, otherwise we'll return ENOENT.
2193 2193 */
2194 2194 if (error == 0 && strchr(zc->zc_name, '$') == NULL) {
2195 2195 error = zfs_ioc_objset_stats(zc); /* fill in the stats */
2196 2196 if (error == ENOENT) {
2197 2197 /* We lost a race with destroy, get the next one. */
2198 2198 zc->zc_name[orig_len] = '\0';
2199 2199 goto top;
2200 2200 }
2201 2201 }
2202 2202 return (error);
2203 2203 }
2204 2204
2205 2205 /*
2206 2206 * inputs:
2207 2207 * zc_name name of filesystem
2208 2208 * zc_cookie zap cursor
2209 2209 * zc_nvlist_dst_size size of buffer for property nvlist
2210 2210 *
2211 2211 * outputs:
2212 2212 * zc_name name of next snapshot
2213 2213 * zc_objset_stats stats
2214 2214 * zc_nvlist_dst property nvlist
2215 2215 * zc_nvlist_dst_size size of property nvlist
2216 2216 */
2217 2217 static int
2218 2218 zfs_ioc_snapshot_list_next(zfs_cmd_t *zc)
2219 2219 {
2220 2220 objset_t *os;
2221 2221 int error;
2222 2222
2223 2223 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
2224 2224 if (error != 0) {
2225 2225 return (error == ENOENT ? ESRCH : error);
2226 2226 }
2227 2227
2228 2228 /*
2229 2229 * A dataset name of maximum length cannot have any snapshots,
2230 2230 * so exit immediately.
2231 2231 */
2232 2232 if (strlcat(zc->zc_name, "@", sizeof (zc->zc_name)) >= MAXNAMELEN) {
2233 2233 dmu_objset_rele(os, FTAG);
2234 2234 return (SET_ERROR(ESRCH));
2235 2235 }
2236 2236
2237 2237 error = dmu_snapshot_list_next(os,
2238 2238 sizeof (zc->zc_name) - strlen(zc->zc_name),
2239 2239 zc->zc_name + strlen(zc->zc_name), &zc->zc_obj, &zc->zc_cookie,
2240 2240 NULL);
2241 2241
2242 2242 if (error == 0) {
2243 2243 dsl_dataset_t *ds;
2244 2244 dsl_pool_t *dp = os->os_dsl_dataset->ds_dir->dd_pool;
2245 2245
2246 2246 error = dsl_dataset_hold_obj(dp, zc->zc_obj, FTAG, &ds);
2247 2247 if (error == 0) {
2248 2248 objset_t *ossnap;
2249 2249
2250 2250 error = dmu_objset_from_ds(ds, &ossnap);
2251 2251 if (error == 0)
2252 2252 error = zfs_ioc_objset_stats_impl(zc, ossnap);
2253 2253 dsl_dataset_rele(ds, FTAG);
2254 2254 }
2255 2255 } else if (error == ENOENT) {
2256 2256 error = SET_ERROR(ESRCH);
2257 2257 }
2258 2258
2259 2259 dmu_objset_rele(os, FTAG);
2260 2260 /* if we failed, undo the @ that we tacked on to zc_name */
2261 2261 if (error != 0)
2262 2262 *strchr(zc->zc_name, '@') = '\0';
2263 2263 return (error);
2264 2264 }
2265 2265
2266 2266 static int
2267 2267 zfs_prop_set_userquota(const char *dsname, nvpair_t *pair)
2268 2268 {
2269 2269 const char *propname = nvpair_name(pair);
2270 2270 uint64_t *valary;
2271 2271 unsigned int vallen;
2272 2272 const char *domain;
2273 2273 char *dash;
2274 2274 zfs_userquota_prop_t type;
2275 2275 uint64_t rid;
2276 2276 uint64_t quota;
2277 2277 zfsvfs_t *zfsvfs;
2278 2278 int err;
2279 2279
2280 2280 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2281 2281 nvlist_t *attrs;
2282 2282 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
2283 2283 if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2284 2284 &pair) != 0)
2285 2285 return (SET_ERROR(EINVAL));
2286 2286 }
2287 2287
2288 2288 /*
2289 2289 * A correctly constructed propname is encoded as
2290 2290 * userquota@<rid>-<domain>.
2291 2291 */
2292 2292 if ((dash = strchr(propname, '-')) == NULL ||
2293 2293 nvpair_value_uint64_array(pair, &valary, &vallen) != 0 ||
2294 2294 vallen != 3)
2295 2295 return (SET_ERROR(EINVAL));
2296 2296
2297 2297 domain = dash + 1;
2298 2298 type = valary[0];
2299 2299 rid = valary[1];
2300 2300 quota = valary[2];
2301 2301
2302 2302 err = zfsvfs_hold(dsname, FTAG, &zfsvfs, B_FALSE);
2303 2303 if (err == 0) {
2304 2304 err = zfs_set_userquota(zfsvfs, type, domain, rid, quota);
2305 2305 zfsvfs_rele(zfsvfs, FTAG);
2306 2306 }
2307 2307
2308 2308 return (err);
2309 2309 }
2310 2310
2311 2311 /*
2312 2312 * If the named property is one that has a special function to set its value,
2313 2313 * return 0 on success and a positive error code on failure; otherwise if it is
2314 2314 * not one of the special properties handled by this function, return -1.
2315 2315 *
2316 2316 * XXX: It would be better for callers of the property interface if we handled
2317 2317 * these special cases in dsl_prop.c (in the dsl layer).
2318 2318 */
2319 2319 static int
2320 2320 zfs_prop_set_special(const char *dsname, zprop_source_t source,
2321 2321 nvpair_t *pair)
2322 2322 {
2323 2323 const char *propname = nvpair_name(pair);
2324 2324 zfs_prop_t prop = zfs_name_to_prop(propname);
2325 2325 uint64_t intval;
2326 2326 int err;
2327 2327
2328 2328 if (prop == ZPROP_INVAL) {
2329 2329 if (zfs_prop_userquota(propname))
2330 2330 return (zfs_prop_set_userquota(dsname, pair));
2331 2331 return (-1);
2332 2332 }
2333 2333
2334 2334 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2335 2335 nvlist_t *attrs;
2336 2336 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
2337 2337 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2338 2338 &pair) == 0);
2339 2339 }
2340 2340
2341 2341 if (zfs_prop_get_type(prop) == PROP_TYPE_STRING)
2342 2342 return (-1);
2343 2343
2344 2344 VERIFY(0 == nvpair_value_uint64(pair, &intval));
2345 2345
2346 2346 switch (prop) {
2347 2347 case ZFS_PROP_QUOTA:
2348 2348 err = dsl_dir_set_quota(dsname, source, intval);
2349 2349 break;
2350 2350 case ZFS_PROP_REFQUOTA:
2351 2351 err = dsl_dataset_set_refquota(dsname, source, intval);
2352 2352 break;
2353 2353 case ZFS_PROP_RESERVATION:
2354 2354 err = dsl_dir_set_reservation(dsname, source, intval);
2355 2355 break;
2356 2356 case ZFS_PROP_REFRESERVATION:
2357 2357 err = dsl_dataset_set_refreservation(dsname, source, intval);
2358 2358 break;
2359 2359 case ZFS_PROP_VOLSIZE:
2360 2360 err = zvol_set_volsize(dsname, intval);
2361 2361 break;
2362 2362 case ZFS_PROP_VERSION:
2363 2363 {
2364 2364 zfsvfs_t *zfsvfs;
2365 2365
2366 2366 if ((err = zfsvfs_hold(dsname, FTAG, &zfsvfs, B_TRUE)) != 0)
2367 2367 break;
2368 2368
2369 2369 err = zfs_set_version(zfsvfs, intval);
2370 2370 zfsvfs_rele(zfsvfs, FTAG);
2371 2371
2372 2372 if (err == 0 && intval >= ZPL_VERSION_USERSPACE) {
2373 2373 zfs_cmd_t *zc;
2374 2374
2375 2375 zc = kmem_zalloc(sizeof (zfs_cmd_t), KM_SLEEP);
2376 2376 (void) strcpy(zc->zc_name, dsname);
2377 2377 (void) zfs_ioc_userspace_upgrade(zc);
2378 2378 kmem_free(zc, sizeof (zfs_cmd_t));
2379 2379 }
2380 2380 break;
2381 2381 }
2382 2382 case ZFS_PROP_COMPRESSION:
2383 2383 {
2384 2384 if (intval == ZIO_COMPRESS_LZ4) {
2385 2385 zfeature_info_t *feature =
2386 2386 &spa_feature_table[SPA_FEATURE_LZ4_COMPRESS];
2387 2387 spa_t *spa;
2388 2388
2389 2389 if ((err = spa_open(dsname, &spa, FTAG)) != 0)
2390 2390 return (err);
2391 2391
2392 2392 /*
2393 2393 * Setting the LZ4 compression algorithm activates
2394 2394 * the feature.
2395 2395 */
2396 2396 if (!spa_feature_is_active(spa, feature)) {
2397 2397 if ((err = zfs_prop_activate_feature(spa,
2398 2398 feature)) != 0) {
2399 2399 spa_close(spa, FTAG);
2400 2400 return (err);
2401 2401 }
2402 2402 }
2403 2403
2404 2404 spa_close(spa, FTAG);
2405 2405 }
2406 2406 /*
2407 2407 * We still want the default set action to be performed in the
2408 2408 * caller, we only performed zfeature settings here.
2409 2409 */
2410 2410 err = -1;
2411 2411 break;
2412 2412 }
2413 2413
2414 2414 default:
2415 2415 err = -1;
2416 2416 }
2417 2417
2418 2418 return (err);
2419 2419 }
2420 2420
2421 2421 /*
2422 2422 * This function is best effort. If it fails to set any of the given properties,
2423 2423 * it continues to set as many as it can and returns the last error
2424 2424 * encountered. If the caller provides a non-NULL errlist, it will be filled in
2425 2425 * with the list of names of all the properties that failed along with the
2426 2426 * corresponding error numbers.
2427 2427 *
2428 2428 * If every property is set successfully, zero is returned and errlist is not
2429 2429 * modified.
2430 2430 */
2431 2431 int
2432 2432 zfs_set_prop_nvlist(const char *dsname, zprop_source_t source, nvlist_t *nvl,
2433 2433 nvlist_t *errlist)
2434 2434 {
2435 2435 nvpair_t *pair;
2436 2436 nvpair_t *propval;
2437 2437 int rv = 0;
2438 2438 uint64_t intval;
2439 2439 char *strval;
2440 2440 nvlist_t *genericnvl = fnvlist_alloc();
2441 2441 nvlist_t *retrynvl = fnvlist_alloc();
2442 2442
2443 2443 retry:
2444 2444 pair = NULL;
2445 2445 while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) {
2446 2446 const char *propname = nvpair_name(pair);
2447 2447 zfs_prop_t prop = zfs_name_to_prop(propname);
2448 2448 int err = 0;
2449 2449
2450 2450 /* decode the property value */
2451 2451 propval = pair;
2452 2452 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2453 2453 nvlist_t *attrs;
2454 2454 attrs = fnvpair_value_nvlist(pair);
2455 2455 if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2456 2456 &propval) != 0)
2457 2457 err = SET_ERROR(EINVAL);
2458 2458 }
2459 2459
2460 2460 /* Validate value type */
2461 2461 if (err == 0 && prop == ZPROP_INVAL) {
2462 2462 if (zfs_prop_user(propname)) {
2463 2463 if (nvpair_type(propval) != DATA_TYPE_STRING)
2464 2464 err = SET_ERROR(EINVAL);
2465 2465 } else if (zfs_prop_userquota(propname)) {
2466 2466 if (nvpair_type(propval) !=
2467 2467 DATA_TYPE_UINT64_ARRAY)
2468 2468 err = SET_ERROR(EINVAL);
2469 2469 } else {
2470 2470 err = SET_ERROR(EINVAL);
2471 2471 }
2472 2472 } else if (err == 0) {
2473 2473 if (nvpair_type(propval) == DATA_TYPE_STRING) {
2474 2474 if (zfs_prop_get_type(prop) != PROP_TYPE_STRING)
2475 2475 err = SET_ERROR(EINVAL);
2476 2476 } else if (nvpair_type(propval) == DATA_TYPE_UINT64) {
2477 2477 const char *unused;
2478 2478
2479 2479 intval = fnvpair_value_uint64(propval);
2480 2480
2481 2481 switch (zfs_prop_get_type(prop)) {
2482 2482 case PROP_TYPE_NUMBER:
2483 2483 break;
2484 2484 case PROP_TYPE_STRING:
2485 2485 err = SET_ERROR(EINVAL);
2486 2486 break;
2487 2487 case PROP_TYPE_INDEX:
2488 2488 if (zfs_prop_index_to_string(prop,
2489 2489 intval, &unused) != 0)
2490 2490 err = SET_ERROR(EINVAL);
2491 2491 break;
2492 2492 default:
2493 2493 cmn_err(CE_PANIC,
2494 2494 "unknown property type");
2495 2495 }
2496 2496 } else {
2497 2497 err = SET_ERROR(EINVAL);
2498 2498 }
2499 2499 }
2500 2500
2501 2501 /* Validate permissions */
2502 2502 if (err == 0)
2503 2503 err = zfs_check_settable(dsname, pair, CRED());
2504 2504
2505 2505 if (err == 0) {
2506 2506 err = zfs_prop_set_special(dsname, source, pair);
2507 2507 if (err == -1) {
2508 2508 /*
2509 2509 * For better performance we build up a list of
2510 2510 * properties to set in a single transaction.
2511 2511 */
2512 2512 err = nvlist_add_nvpair(genericnvl, pair);
2513 2513 } else if (err != 0 && nvl != retrynvl) {
2514 2514 /*
2515 2515 * This may be a spurious error caused by
2516 2516 * receiving quota and reservation out of order.
2517 2517 * Try again in a second pass.
2518 2518 */
2519 2519 err = nvlist_add_nvpair(retrynvl, pair);
2520 2520 }
2521 2521 }
2522 2522
2523 2523 if (err != 0) {
2524 2524 if (errlist != NULL)
2525 2525 fnvlist_add_int32(errlist, propname, err);
2526 2526 rv = err;
2527 2527 }
2528 2528 }
2529 2529
2530 2530 if (nvl != retrynvl && !nvlist_empty(retrynvl)) {
2531 2531 nvl = retrynvl;
2532 2532 goto retry;
2533 2533 }
2534 2534
2535 2535 if (!nvlist_empty(genericnvl) &&
2536 2536 dsl_props_set(dsname, source, genericnvl) != 0) {
2537 2537 /*
2538 2538 * If this fails, we still want to set as many properties as we
2539 2539 * can, so try setting them individually.
2540 2540 */
2541 2541 pair = NULL;
2542 2542 while ((pair = nvlist_next_nvpair(genericnvl, pair)) != NULL) {
2543 2543 const char *propname = nvpair_name(pair);
2544 2544 int err = 0;
2545 2545
2546 2546 propval = pair;
2547 2547 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2548 2548 nvlist_t *attrs;
2549 2549 attrs = fnvpair_value_nvlist(pair);
2550 2550 propval = fnvlist_lookup_nvpair(attrs,
2551 2551 ZPROP_VALUE);
2552 2552 }
2553 2553
2554 2554 if (nvpair_type(propval) == DATA_TYPE_STRING) {
2555 2555 strval = fnvpair_value_string(propval);
2556 2556 err = dsl_prop_set_string(dsname, propname,
2557 2557 source, strval);
2558 2558 } else {
2559 2559 intval = fnvpair_value_uint64(propval);
2560 2560 err = dsl_prop_set_int(dsname, propname, source,
2561 2561 intval);
2562 2562 }
2563 2563
2564 2564 if (err != 0) {
2565 2565 if (errlist != NULL) {
2566 2566 fnvlist_add_int32(errlist, propname,
2567 2567 err);
2568 2568 }
2569 2569 rv = err;
2570 2570 }
2571 2571 }
2572 2572 }
2573 2573 nvlist_free(genericnvl);
2574 2574 nvlist_free(retrynvl);
2575 2575
2576 2576 return (rv);
2577 2577 }
2578 2578
2579 2579 /*
2580 2580 * Check that all the properties are valid user properties.
2581 2581 */
2582 2582 static int
2583 2583 zfs_check_userprops(const char *fsname, nvlist_t *nvl)
2584 2584 {
2585 2585 nvpair_t *pair = NULL;
2586 2586 int error = 0;
2587 2587
2588 2588 while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) {
2589 2589 const char *propname = nvpair_name(pair);
2590 2590 char *valstr;
2591 2591
2592 2592 if (!zfs_prop_user(propname) ||
2593 2593 nvpair_type(pair) != DATA_TYPE_STRING)
2594 2594 return (SET_ERROR(EINVAL));
2595 2595
2596 2596 if (error = zfs_secpolicy_write_perms(fsname,
2597 2597 ZFS_DELEG_PERM_USERPROP, CRED()))
2598 2598 return (error);
2599 2599
2600 2600 if (strlen(propname) >= ZAP_MAXNAMELEN)
2601 2601 return (SET_ERROR(ENAMETOOLONG));
2602 2602
2603 2603 VERIFY(nvpair_value_string(pair, &valstr) == 0);
2604 2604 if (strlen(valstr) >= ZAP_MAXVALUELEN)
2605 2605 return (E2BIG);
2606 2606 }
2607 2607 return (0);
2608 2608 }
2609 2609
2610 2610 static void
2611 2611 props_skip(nvlist_t *props, nvlist_t *skipped, nvlist_t **newprops)
2612 2612 {
2613 2613 nvpair_t *pair;
2614 2614
2615 2615 VERIFY(nvlist_alloc(newprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2616 2616
2617 2617 pair = NULL;
2618 2618 while ((pair = nvlist_next_nvpair(props, pair)) != NULL) {
2619 2619 if (nvlist_exists(skipped, nvpair_name(pair)))
2620 2620 continue;
2621 2621
2622 2622 VERIFY(nvlist_add_nvpair(*newprops, pair) == 0);
2623 2623 }
2624 2624 }
2625 2625
2626 2626 static int
2627 2627 clear_received_props(const char *dsname, nvlist_t *props,
2628 2628 nvlist_t *skipped)
2629 2629 {
2630 2630 int err = 0;
2631 2631 nvlist_t *cleared_props = NULL;
2632 2632 props_skip(props, skipped, &cleared_props);
2633 2633 if (!nvlist_empty(cleared_props)) {
2634 2634 /*
2635 2635 * Acts on local properties until the dataset has received
2636 2636 * properties at least once on or after SPA_VERSION_RECVD_PROPS.
2637 2637 */
2638 2638 zprop_source_t flags = (ZPROP_SRC_NONE |
2639 2639 (dsl_prop_get_hasrecvd(dsname) ? ZPROP_SRC_RECEIVED : 0));
2640 2640 err = zfs_set_prop_nvlist(dsname, flags, cleared_props, NULL);
2641 2641 }
2642 2642 nvlist_free(cleared_props);
2643 2643 return (err);
2644 2644 }
2645 2645
2646 2646 /*
2647 2647 * inputs:
2648 2648 * zc_name name of filesystem
2649 2649 * zc_value name of property to set
2650 2650 * zc_nvlist_src{_size} nvlist of properties to apply
2651 2651 * zc_cookie received properties flag
2652 2652 *
2653 2653 * outputs:
2654 2654 * zc_nvlist_dst{_size} error for each unapplied received property
2655 2655 */
2656 2656 static int
2657 2657 zfs_ioc_set_prop(zfs_cmd_t *zc)
2658 2658 {
2659 2659 nvlist_t *nvl;
2660 2660 boolean_t received = zc->zc_cookie;
2661 2661 zprop_source_t source = (received ? ZPROP_SRC_RECEIVED :
2662 2662 ZPROP_SRC_LOCAL);
2663 2663 nvlist_t *errors;
2664 2664 int error;
2665 2665
2666 2666 if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2667 2667 zc->zc_iflags, &nvl)) != 0)
2668 2668 return (error);
2669 2669
2670 2670 if (received) {
2671 2671 nvlist_t *origprops;
2672 2672
2673 2673 if (dsl_prop_get_received(zc->zc_name, &origprops) == 0) {
2674 2674 (void) clear_received_props(zc->zc_name,
2675 2675 origprops, nvl);
2676 2676 nvlist_free(origprops);
2677 2677 }
2678 2678
2679 2679 error = dsl_prop_set_hasrecvd(zc->zc_name);
2680 2680 }
2681 2681
2682 2682 errors = fnvlist_alloc();
2683 2683 if (error == 0)
2684 2684 error = zfs_set_prop_nvlist(zc->zc_name, source, nvl, errors);
2685 2685
2686 2686 if (zc->zc_nvlist_dst != NULL && errors != NULL) {
2687 2687 (void) put_nvlist(zc, errors);
2688 2688 }
2689 2689
2690 2690 nvlist_free(errors);
2691 2691 nvlist_free(nvl);
2692 2692 return (error);
2693 2693 }
2694 2694
2695 2695 /*
2696 2696 * inputs:
2697 2697 * zc_name name of filesystem
2698 2698 * zc_value name of property to inherit
2699 2699 * zc_cookie revert to received value if TRUE
2700 2700 *
2701 2701 * outputs: none
2702 2702 */
2703 2703 static int
2704 2704 zfs_ioc_inherit_prop(zfs_cmd_t *zc)
2705 2705 {
2706 2706 const char *propname = zc->zc_value;
2707 2707 zfs_prop_t prop = zfs_name_to_prop(propname);
2708 2708 boolean_t received = zc->zc_cookie;
2709 2709 zprop_source_t source = (received
2710 2710 ? ZPROP_SRC_NONE /* revert to received value, if any */
2711 2711 : ZPROP_SRC_INHERITED); /* explicitly inherit */
2712 2712
2713 2713 if (received) {
2714 2714 nvlist_t *dummy;
2715 2715 nvpair_t *pair;
2716 2716 zprop_type_t type;
2717 2717 int err;
2718 2718
2719 2719 /*
2720 2720 * zfs_prop_set_special() expects properties in the form of an
2721 2721 * nvpair with type info.
2722 2722 */
2723 2723 if (prop == ZPROP_INVAL) {
2724 2724 if (!zfs_prop_user(propname))
2725 2725 return (SET_ERROR(EINVAL));
2726 2726
2727 2727 type = PROP_TYPE_STRING;
2728 2728 } else if (prop == ZFS_PROP_VOLSIZE ||
2729 2729 prop == ZFS_PROP_VERSION) {
2730 2730 return (SET_ERROR(EINVAL));
2731 2731 } else {
2732 2732 type = zfs_prop_get_type(prop);
2733 2733 }
2734 2734
2735 2735 VERIFY(nvlist_alloc(&dummy, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2736 2736
2737 2737 switch (type) {
2738 2738 case PROP_TYPE_STRING:
2739 2739 VERIFY(0 == nvlist_add_string(dummy, propname, ""));
2740 2740 break;
2741 2741 case PROP_TYPE_NUMBER:
2742 2742 case PROP_TYPE_INDEX:
2743 2743 VERIFY(0 == nvlist_add_uint64(dummy, propname, 0));
2744 2744 break;
2745 2745 default:
2746 2746 nvlist_free(dummy);
2747 2747 return (SET_ERROR(EINVAL));
2748 2748 }
2749 2749
2750 2750 pair = nvlist_next_nvpair(dummy, NULL);
2751 2751 err = zfs_prop_set_special(zc->zc_name, source, pair);
2752 2752 nvlist_free(dummy);
2753 2753 if (err != -1)
2754 2754 return (err); /* special property already handled */
2755 2755 } else {
2756 2756 /*
2757 2757 * Only check this in the non-received case. We want to allow
2758 2758 * 'inherit -S' to revert non-inheritable properties like quota
2759 2759 * and reservation to the received or default values even though
2760 2760 * they are not considered inheritable.
2761 2761 */
2762 2762 if (prop != ZPROP_INVAL && !zfs_prop_inheritable(prop))
2763 2763 return (SET_ERROR(EINVAL));
2764 2764 }
2765 2765
2766 2766 /* property name has been validated by zfs_secpolicy_inherit_prop() */
2767 2767 return (dsl_prop_inherit(zc->zc_name, zc->zc_value, source));
2768 2768 }
2769 2769
2770 2770 static int
2771 2771 zfs_ioc_pool_set_props(zfs_cmd_t *zc)
2772 2772 {
2773 2773 nvlist_t *props;
2774 2774 spa_t *spa;
2775 2775 int error;
2776 2776 nvpair_t *pair;
2777 2777
2778 2778 if (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2779 2779 zc->zc_iflags, &props))
2780 2780 return (error);
2781 2781
2782 2782 /*
2783 2783 * If the only property is the configfile, then just do a spa_lookup()
2784 2784 * to handle the faulted case.
2785 2785 */
2786 2786 pair = nvlist_next_nvpair(props, NULL);
2787 2787 if (pair != NULL && strcmp(nvpair_name(pair),
2788 2788 zpool_prop_to_name(ZPOOL_PROP_CACHEFILE)) == 0 &&
2789 2789 nvlist_next_nvpair(props, pair) == NULL) {
2790 2790 mutex_enter(&spa_namespace_lock);
2791 2791 if ((spa = spa_lookup(zc->zc_name)) != NULL) {
2792 2792 spa_configfile_set(spa, props, B_FALSE);
2793 2793 spa_config_sync(spa, B_FALSE, B_TRUE);
2794 2794 }
2795 2795 mutex_exit(&spa_namespace_lock);
2796 2796 if (spa != NULL) {
2797 2797 nvlist_free(props);
2798 2798 return (0);
2799 2799 }
2800 2800 }
2801 2801
2802 2802 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) {
2803 2803 nvlist_free(props);
2804 2804 return (error);
2805 2805 }
2806 2806
2807 2807 error = spa_prop_set(spa, props);
2808 2808
2809 2809 nvlist_free(props);
2810 2810 spa_close(spa, FTAG);
2811 2811
2812 2812 return (error);
2813 2813 }
2814 2814
2815 2815 static int
2816 2816 zfs_ioc_pool_get_props(zfs_cmd_t *zc)
2817 2817 {
2818 2818 spa_t *spa;
2819 2819 int error;
2820 2820 nvlist_t *nvp = NULL;
2821 2821
2822 2822 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) {
2823 2823 /*
2824 2824 * If the pool is faulted, there may be properties we can still
2825 2825 * get (such as altroot and cachefile), so attempt to get them
2826 2826 * anyway.
2827 2827 */
2828 2828 mutex_enter(&spa_namespace_lock);
2829 2829 if ((spa = spa_lookup(zc->zc_name)) != NULL)
2830 2830 error = spa_prop_get(spa, &nvp);
2831 2831 mutex_exit(&spa_namespace_lock);
2832 2832 } else {
2833 2833 error = spa_prop_get(spa, &nvp);
2834 2834 spa_close(spa, FTAG);
2835 2835 }
2836 2836
2837 2837 if (error == 0 && zc->zc_nvlist_dst != NULL)
2838 2838 error = put_nvlist(zc, nvp);
2839 2839 else
2840 2840 error = SET_ERROR(EFAULT);
2841 2841
2842 2842 nvlist_free(nvp);
2843 2843 return (error);
2844 2844 }
2845 2845
2846 2846 /*
2847 2847 * inputs:
2848 2848 * zc_name name of filesystem
2849 2849 * zc_nvlist_src{_size} nvlist of delegated permissions
2850 2850 * zc_perm_action allow/unallow flag
2851 2851 *
2852 2852 * outputs: none
2853 2853 */
2854 2854 static int
2855 2855 zfs_ioc_set_fsacl(zfs_cmd_t *zc)
2856 2856 {
2857 2857 int error;
2858 2858 nvlist_t *fsaclnv = NULL;
2859 2859
2860 2860 if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2861 2861 zc->zc_iflags, &fsaclnv)) != 0)
2862 2862 return (error);
2863 2863
2864 2864 /*
2865 2865 * Verify nvlist is constructed correctly
2866 2866 */
2867 2867 if ((error = zfs_deleg_verify_nvlist(fsaclnv)) != 0) {
2868 2868 nvlist_free(fsaclnv);
2869 2869 return (SET_ERROR(EINVAL));
2870 2870 }
2871 2871
2872 2872 /*
2873 2873 * If we don't have PRIV_SYS_MOUNT, then validate
2874 2874 * that user is allowed to hand out each permission in
2875 2875 * the nvlist(s)
2876 2876 */
2877 2877
2878 2878 error = secpolicy_zfs(CRED());
2879 2879 if (error != 0) {
2880 2880 if (zc->zc_perm_action == B_FALSE) {
2881 2881 error = dsl_deleg_can_allow(zc->zc_name,
2882 2882 fsaclnv, CRED());
2883 2883 } else {
2884 2884 error = dsl_deleg_can_unallow(zc->zc_name,
2885 2885 fsaclnv, CRED());
2886 2886 }
2887 2887 }
2888 2888
2889 2889 if (error == 0)
2890 2890 error = dsl_deleg_set(zc->zc_name, fsaclnv, zc->zc_perm_action);
2891 2891
2892 2892 nvlist_free(fsaclnv);
2893 2893 return (error);
2894 2894 }
2895 2895
2896 2896 /*
2897 2897 * inputs:
2898 2898 * zc_name name of filesystem
2899 2899 *
2900 2900 * outputs:
2901 2901 * zc_nvlist_src{_size} nvlist of delegated permissions
2902 2902 */
2903 2903 static int
2904 2904 zfs_ioc_get_fsacl(zfs_cmd_t *zc)
2905 2905 {
2906 2906 nvlist_t *nvp;
2907 2907 int error;
2908 2908
2909 2909 if ((error = dsl_deleg_get(zc->zc_name, &nvp)) == 0) {
2910 2910 error = put_nvlist(zc, nvp);
2911 2911 nvlist_free(nvp);
2912 2912 }
2913 2913
2914 2914 return (error);
2915 2915 }
2916 2916
2917 2917 /*
2918 2918 * Search the vfs list for a specified resource. Returns a pointer to it
2919 2919 * or NULL if no suitable entry is found. The caller of this routine
2920 2920 * is responsible for releasing the returned vfs pointer.
2921 2921 */
2922 2922 static vfs_t *
2923 2923 zfs_get_vfs(const char *resource)
2924 2924 {
2925 2925 struct vfs *vfsp;
2926 2926 struct vfs *vfs_found = NULL;
2927 2927
2928 2928 vfs_list_read_lock();
2929 2929 vfsp = rootvfs;
2930 2930 do {
2931 2931 if (strcmp(refstr_value(vfsp->vfs_resource), resource) == 0) {
2932 2932 VFS_HOLD(vfsp);
2933 2933 vfs_found = vfsp;
2934 2934 break;
2935 2935 }
2936 2936 vfsp = vfsp->vfs_next;
2937 2937 } while (vfsp != rootvfs);
2938 2938 vfs_list_unlock();
2939 2939 return (vfs_found);
2940 2940 }
2941 2941
2942 2942 /* ARGSUSED */
2943 2943 static void
2944 2944 zfs_create_cb(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx)
2945 2945 {
2946 2946 zfs_creat_t *zct = arg;
2947 2947
2948 2948 zfs_create_fs(os, cr, zct->zct_zplprops, tx);
2949 2949 }
2950 2950
2951 2951 #define ZFS_PROP_UNDEFINED ((uint64_t)-1)
2952 2952
2953 2953 /*
2954 2954 * inputs:
2955 2955 * os parent objset pointer (NULL if root fs)
2956 2956 * fuids_ok fuids allowed in this version of the spa?
2957 2957 * sa_ok SAs allowed in this version of the spa?
2958 2958 * createprops list of properties requested by creator
2959 2959 *
2960 2960 * outputs:
2961 2961 * zplprops values for the zplprops we attach to the master node object
2962 2962 * is_ci true if requested file system will be purely case-insensitive
2963 2963 *
2964 2964 * Determine the settings for utf8only, normalization and
2965 2965 * casesensitivity. Specific values may have been requested by the
2966 2966 * creator and/or we can inherit values from the parent dataset. If
2967 2967 * the file system is of too early a vintage, a creator can not
2968 2968 * request settings for these properties, even if the requested
2969 2969 * setting is the default value. We don't actually want to create dsl
2970 2970 * properties for these, so remove them from the source nvlist after
2971 2971 * processing.
2972 2972 */
2973 2973 static int
2974 2974 zfs_fill_zplprops_impl(objset_t *os, uint64_t zplver,
2975 2975 boolean_t fuids_ok, boolean_t sa_ok, nvlist_t *createprops,
2976 2976 nvlist_t *zplprops, boolean_t *is_ci)
2977 2977 {
2978 2978 uint64_t sense = ZFS_PROP_UNDEFINED;
2979 2979 uint64_t norm = ZFS_PROP_UNDEFINED;
2980 2980 uint64_t u8 = ZFS_PROP_UNDEFINED;
2981 2981
2982 2982 ASSERT(zplprops != NULL);
2983 2983
2984 2984 /*
2985 2985 * Pull out creator prop choices, if any.
2986 2986 */
2987 2987 if (createprops) {
2988 2988 (void) nvlist_lookup_uint64(createprops,
2989 2989 zfs_prop_to_name(ZFS_PROP_VERSION), &zplver);
2990 2990 (void) nvlist_lookup_uint64(createprops,
2991 2991 zfs_prop_to_name(ZFS_PROP_NORMALIZE), &norm);
2992 2992 (void) nvlist_remove_all(createprops,
2993 2993 zfs_prop_to_name(ZFS_PROP_NORMALIZE));
2994 2994 (void) nvlist_lookup_uint64(createprops,
2995 2995 zfs_prop_to_name(ZFS_PROP_UTF8ONLY), &u8);
2996 2996 (void) nvlist_remove_all(createprops,
2997 2997 zfs_prop_to_name(ZFS_PROP_UTF8ONLY));
2998 2998 (void) nvlist_lookup_uint64(createprops,
2999 2999 zfs_prop_to_name(ZFS_PROP_CASE), &sense);
3000 3000 (void) nvlist_remove_all(createprops,
3001 3001 zfs_prop_to_name(ZFS_PROP_CASE));
3002 3002 }
3003 3003
3004 3004 /*
3005 3005 * If the zpl version requested is whacky or the file system
3006 3006 * or pool is version is too "young" to support normalization
3007 3007 * and the creator tried to set a value for one of the props,
3008 3008 * error out.
3009 3009 */
3010 3010 if ((zplver < ZPL_VERSION_INITIAL || zplver > ZPL_VERSION) ||
3011 3011 (zplver >= ZPL_VERSION_FUID && !fuids_ok) ||
3012 3012 (zplver >= ZPL_VERSION_SA && !sa_ok) ||
3013 3013 (zplver < ZPL_VERSION_NORMALIZATION &&
3014 3014 (norm != ZFS_PROP_UNDEFINED || u8 != ZFS_PROP_UNDEFINED ||
3015 3015 sense != ZFS_PROP_UNDEFINED)))
3016 3016 return (SET_ERROR(ENOTSUP));
3017 3017
3018 3018 /*
3019 3019 * Put the version in the zplprops
3020 3020 */
3021 3021 VERIFY(nvlist_add_uint64(zplprops,
3022 3022 zfs_prop_to_name(ZFS_PROP_VERSION), zplver) == 0);
3023 3023
3024 3024 if (norm == ZFS_PROP_UNDEFINED)
3025 3025 VERIFY(zfs_get_zplprop(os, ZFS_PROP_NORMALIZE, &norm) == 0);
3026 3026 VERIFY(nvlist_add_uint64(zplprops,
3027 3027 zfs_prop_to_name(ZFS_PROP_NORMALIZE), norm) == 0);
3028 3028
3029 3029 /*
3030 3030 * If we're normalizing, names must always be valid UTF-8 strings.
3031 3031 */
3032 3032 if (norm)
3033 3033 u8 = 1;
3034 3034 if (u8 == ZFS_PROP_UNDEFINED)
3035 3035 VERIFY(zfs_get_zplprop(os, ZFS_PROP_UTF8ONLY, &u8) == 0);
3036 3036 VERIFY(nvlist_add_uint64(zplprops,
3037 3037 zfs_prop_to_name(ZFS_PROP_UTF8ONLY), u8) == 0);
3038 3038
3039 3039 if (sense == ZFS_PROP_UNDEFINED)
3040 3040 VERIFY(zfs_get_zplprop(os, ZFS_PROP_CASE, &sense) == 0);
3041 3041 VERIFY(nvlist_add_uint64(zplprops,
3042 3042 zfs_prop_to_name(ZFS_PROP_CASE), sense) == 0);
3043 3043
3044 3044 if (is_ci)
3045 3045 *is_ci = (sense == ZFS_CASE_INSENSITIVE);
3046 3046
3047 3047 return (0);
3048 3048 }
3049 3049
3050 3050 static int
3051 3051 zfs_fill_zplprops(const char *dataset, nvlist_t *createprops,
3052 3052 nvlist_t *zplprops, boolean_t *is_ci)
3053 3053 {
3054 3054 boolean_t fuids_ok, sa_ok;
3055 3055 uint64_t zplver = ZPL_VERSION;
3056 3056 objset_t *os = NULL;
3057 3057 char parentname[MAXNAMELEN];
3058 3058 char *cp;
3059 3059 spa_t *spa;
3060 3060 uint64_t spa_vers;
3061 3061 int error;
3062 3062
3063 3063 (void) strlcpy(parentname, dataset, sizeof (parentname));
3064 3064 cp = strrchr(parentname, '/');
3065 3065 ASSERT(cp != NULL);
3066 3066 cp[0] = '\0';
3067 3067
3068 3068 if ((error = spa_open(dataset, &spa, FTAG)) != 0)
3069 3069 return (error);
3070 3070
3071 3071 spa_vers = spa_version(spa);
3072 3072 spa_close(spa, FTAG);
3073 3073
3074 3074 zplver = zfs_zpl_version_map(spa_vers);
3075 3075 fuids_ok = (zplver >= ZPL_VERSION_FUID);
3076 3076 sa_ok = (zplver >= ZPL_VERSION_SA);
3077 3077
3078 3078 /*
3079 3079 * Open parent object set so we can inherit zplprop values.
3080 3080 */
3081 3081 if ((error = dmu_objset_hold(parentname, FTAG, &os)) != 0)
3082 3082 return (error);
3083 3083
3084 3084 error = zfs_fill_zplprops_impl(os, zplver, fuids_ok, sa_ok, createprops,
3085 3085 zplprops, is_ci);
3086 3086 dmu_objset_rele(os, FTAG);
3087 3087 return (error);
3088 3088 }
3089 3089
3090 3090 static int
3091 3091 zfs_fill_zplprops_root(uint64_t spa_vers, nvlist_t *createprops,
3092 3092 nvlist_t *zplprops, boolean_t *is_ci)
3093 3093 {
3094 3094 boolean_t fuids_ok;
3095 3095 boolean_t sa_ok;
3096 3096 uint64_t zplver = ZPL_VERSION;
3097 3097 int error;
3098 3098
3099 3099 zplver = zfs_zpl_version_map(spa_vers);
3100 3100 fuids_ok = (zplver >= ZPL_VERSION_FUID);
3101 3101 sa_ok = (zplver >= ZPL_VERSION_SA);
3102 3102
3103 3103 error = zfs_fill_zplprops_impl(NULL, zplver, fuids_ok, sa_ok,
3104 3104 createprops, zplprops, is_ci);
3105 3105 return (error);
3106 3106 }
3107 3107
3108 3108 /*
3109 3109 * innvl: {
3110 3110 * "type" -> dmu_objset_type_t (int32)
3111 3111 * (optional) "props" -> { prop -> value }
3112 3112 * }
3113 3113 *
3114 3114 * outnvl: propname -> error code (int32)
3115 3115 */
3116 3116 static int
3117 3117 zfs_ioc_create(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3118 3118 {
3119 3119 int error = 0;
3120 3120 zfs_creat_t zct = { 0 };
3121 3121 nvlist_t *nvprops = NULL;
3122 3122 void (*cbfunc)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx);
3123 3123 int32_t type32;
3124 3124 dmu_objset_type_t type;
3125 3125 boolean_t is_insensitive = B_FALSE;
3126 3126
3127 3127 if (nvlist_lookup_int32(innvl, "type", &type32) != 0)
3128 3128 return (SET_ERROR(EINVAL));
3129 3129 type = type32;
3130 3130 (void) nvlist_lookup_nvlist(innvl, "props", &nvprops);
3131 3131
3132 3132 switch (type) {
3133 3133 case DMU_OST_ZFS:
3134 3134 cbfunc = zfs_create_cb;
3135 3135 break;
3136 3136
3137 3137 case DMU_OST_ZVOL:
3138 3138 cbfunc = zvol_create_cb;
3139 3139 break;
3140 3140
3141 3141 default:
3142 3142 cbfunc = NULL;
3143 3143 break;
3144 3144 }
3145 3145 if (strchr(fsname, '@') ||
3146 3146 strchr(fsname, '%'))
3147 3147 return (SET_ERROR(EINVAL));
3148 3148
3149 3149 zct.zct_props = nvprops;
3150 3150
3151 3151 if (cbfunc == NULL)
3152 3152 return (SET_ERROR(EINVAL));
3153 3153
3154 3154 if (type == DMU_OST_ZVOL) {
3155 3155 uint64_t volsize, volblocksize;
3156 3156
3157 3157 if (nvprops == NULL)
3158 3158 return (SET_ERROR(EINVAL));
3159 3159 if (nvlist_lookup_uint64(nvprops,
3160 3160 zfs_prop_to_name(ZFS_PROP_VOLSIZE), &volsize) != 0)
3161 3161 return (SET_ERROR(EINVAL));
3162 3162
3163 3163 if ((error = nvlist_lookup_uint64(nvprops,
3164 3164 zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE),
3165 3165 &volblocksize)) != 0 && error != ENOENT)
3166 3166 return (SET_ERROR(EINVAL));
3167 3167
3168 3168 if (error != 0)
3169 3169 volblocksize = zfs_prop_default_numeric(
3170 3170 ZFS_PROP_VOLBLOCKSIZE);
3171 3171
3172 3172 if ((error = zvol_check_volblocksize(
3173 3173 volblocksize)) != 0 ||
3174 3174 (error = zvol_check_volsize(volsize,
3175 3175 volblocksize)) != 0)
3176 3176 return (error);
3177 3177 } else if (type == DMU_OST_ZFS) {
3178 3178 int error;
3179 3179
3180 3180 /*
3181 3181 * We have to have normalization and
3182 3182 * case-folding flags correct when we do the
3183 3183 * file system creation, so go figure them out
3184 3184 * now.
3185 3185 */
3186 3186 VERIFY(nvlist_alloc(&zct.zct_zplprops,
3187 3187 NV_UNIQUE_NAME, KM_SLEEP) == 0);
3188 3188 error = zfs_fill_zplprops(fsname, nvprops,
3189 3189 zct.zct_zplprops, &is_insensitive);
3190 3190 if (error != 0) {
3191 3191 nvlist_free(zct.zct_zplprops);
3192 3192 return (error);
3193 3193 }
3194 3194 }
3195 3195
3196 3196 error = dmu_objset_create(fsname, type,
3197 3197 is_insensitive ? DS_FLAG_CI_DATASET : 0, cbfunc, &zct);
3198 3198 nvlist_free(zct.zct_zplprops);
3199 3199
3200 3200 /*
3201 3201 * It would be nice to do this atomically.
3202 3202 */
3203 3203 if (error == 0) {
3204 3204 error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL,
3205 3205 nvprops, outnvl);
3206 3206 if (error != 0)
3207 3207 (void) dsl_destroy_head(fsname);
3208 3208 }
3209 3209 return (error);
3210 3210 }
3211 3211
3212 3212 /*
3213 3213 * innvl: {
3214 3214 * "origin" -> name of origin snapshot
3215 3215 * (optional) "props" -> { prop -> value }
3216 3216 * }
3217 3217 *
3218 3218 * outnvl: propname -> error code (int32)
3219 3219 */
3220 3220 static int
3221 3221 zfs_ioc_clone(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3222 3222 {
3223 3223 int error = 0;
3224 3224 nvlist_t *nvprops = NULL;
3225 3225 char *origin_name;
3226 3226
3227 3227 if (nvlist_lookup_string(innvl, "origin", &origin_name) != 0)
3228 3228 return (SET_ERROR(EINVAL));
3229 3229 (void) nvlist_lookup_nvlist(innvl, "props", &nvprops);
3230 3230
3231 3231 if (strchr(fsname, '@') ||
3232 3232 strchr(fsname, '%'))
3233 3233 return (SET_ERROR(EINVAL));
3234 3234
3235 3235 if (dataset_namecheck(origin_name, NULL, NULL) != 0)
3236 3236 return (SET_ERROR(EINVAL));
3237 3237 error = dmu_objset_clone(fsname, origin_name);
3238 3238 if (error != 0)
3239 3239 return (error);
3240 3240
3241 3241 /*
3242 3242 * It would be nice to do this atomically.
3243 3243 */
3244 3244 if (error == 0) {
3245 3245 error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL,
3246 3246 nvprops, outnvl);
3247 3247 if (error != 0)
3248 3248 (void) dsl_destroy_head(fsname);
3249 3249 }
3250 3250 return (error);
3251 3251 }
3252 3252
3253 3253 /*
3254 3254 * innvl: {
3255 3255 * "snaps" -> { snapshot1, snapshot2 }
3256 3256 * (optional) "props" -> { prop -> value (string) }
3257 3257 * }
3258 3258 *
3259 3259 * outnvl: snapshot -> error code (int32)
3260 3260 */
3261 3261 static int
3262 3262 zfs_ioc_snapshot(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3263 3263 {
3264 3264 nvlist_t *snaps;
3265 3265 nvlist_t *props = NULL;
3266 3266 int error, poollen;
3267 3267 nvpair_t *pair;
3268 3268
3269 3269 (void) nvlist_lookup_nvlist(innvl, "props", &props);
3270 3270 if ((error = zfs_check_userprops(poolname, props)) != 0)
3271 3271 return (error);
3272 3272
3273 3273 if (!nvlist_empty(props) &&
3274 3274 zfs_earlier_version(poolname, SPA_VERSION_SNAP_PROPS))
3275 3275 return (SET_ERROR(ENOTSUP));
3276 3276
3277 3277 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
3278 3278 return (SET_ERROR(EINVAL));
3279 3279 poollen = strlen(poolname);
3280 3280 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
3281 3281 pair = nvlist_next_nvpair(snaps, pair)) {
3282 3282 const char *name = nvpair_name(pair);
3283 3283 const char *cp = strchr(name, '@');
3284 3284
3285 3285 /*
3286 3286 * The snap name must contain an @, and the part after it must
3287 3287 * contain only valid characters.
3288 3288 */
3289 3289 if (cp == NULL || snapshot_namecheck(cp + 1, NULL, NULL) != 0)
3290 3290 return (SET_ERROR(EINVAL));
3291 3291
3292 3292 /*
3293 3293 * The snap must be in the specified pool.
3294 3294 */
3295 3295 if (strncmp(name, poolname, poollen) != 0 ||
3296 3296 (name[poollen] != '/' && name[poollen] != '@'))
3297 3297 return (SET_ERROR(EXDEV));
3298 3298
3299 3299 /* This must be the only snap of this fs. */
3300 3300 for (nvpair_t *pair2 = nvlist_next_nvpair(snaps, pair);
3301 3301 pair2 != NULL; pair2 = nvlist_next_nvpair(snaps, pair2)) {
3302 3302 if (strncmp(name, nvpair_name(pair2), cp - name + 1)
3303 3303 == 0) {
3304 3304 return (SET_ERROR(EXDEV));
3305 3305 }
3306 3306 }
3307 3307 }
3308 3308
3309 3309 error = dsl_dataset_snapshot(snaps, props, outnvl);
3310 3310 return (error);
3311 3311 }
3312 3312
3313 3313 /*
3314 3314 * innvl: "message" -> string
3315 3315 */
3316 3316 /* ARGSUSED */
3317 3317 static int
3318 3318 zfs_ioc_log_history(const char *unused, nvlist_t *innvl, nvlist_t *outnvl)
3319 3319 {
3320 3320 char *message;
3321 3321 spa_t *spa;
3322 3322 int error;
3323 3323 char *poolname;
3324 3324
3325 3325 /*
3326 3326 * The poolname in the ioctl is not set, we get it from the TSD,
3327 3327 * which was set at the end of the last successful ioctl that allows
3328 3328 * logging. The secpolicy func already checked that it is set.
3329 3329 * Only one log ioctl is allowed after each successful ioctl, so
3330 3330 * we clear the TSD here.
3331 3331 */
3332 3332 poolname = tsd_get(zfs_allow_log_key);
3333 3333 (void) tsd_set(zfs_allow_log_key, NULL);
3334 3334 error = spa_open(poolname, &spa, FTAG);
3335 3335 strfree(poolname);
3336 3336 if (error != 0)
3337 3337 return (error);
3338 3338
3339 3339 if (nvlist_lookup_string(innvl, "message", &message) != 0) {
3340 3340 spa_close(spa, FTAG);
3341 3341 return (SET_ERROR(EINVAL));
3342 3342 }
3343 3343
3344 3344 if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) {
3345 3345 spa_close(spa, FTAG);
3346 3346 return (SET_ERROR(ENOTSUP));
3347 3347 }
3348 3348
3349 3349 error = spa_history_log(spa, message);
3350 3350 spa_close(spa, FTAG);
3351 3351 return (error);
3352 3352 }
3353 3353
3354 3354 /*
3355 3355 * The dp_config_rwlock must not be held when calling this, because the
3356 3356 * unmount may need to write out data.
3357 3357 *
3358 3358 * This function is best-effort. Callers must deal gracefully if it
3359 3359 * remains mounted (or is remounted after this call).
3360 3360 *
3361 3361 * Returns 0 if the argument is not a snapshot, or it is not currently a
3362 3362 * filesystem, or we were able to unmount it. Returns error code otherwise.
3363 3363 */
3364 3364 int
3365 3365 zfs_unmount_snap(const char *snapname)
3366 3366 {
3367 3367 vfs_t *vfsp;
3368 3368 zfsvfs_t *zfsvfs;
3369 3369 int err;
3370 3370
3371 3371 if (strchr(snapname, '@') == NULL)
3372 3372 return (0);
3373 3373
3374 3374 vfsp = zfs_get_vfs(snapname);
3375 3375 if (vfsp == NULL)
3376 3376 return (0);
3377 3377
3378 3378 zfsvfs = vfsp->vfs_data;
3379 3379 ASSERT(!dsl_pool_config_held(dmu_objset_pool(zfsvfs->z_os)));
3380 3380
3381 3381 err = vn_vfswlock(vfsp->vfs_vnodecovered);
3382 3382 VFS_RELE(vfsp);
3383 3383 if (err != 0)
3384 3384 return (SET_ERROR(err));
3385 3385
3386 3386 /*
3387 3387 * Always force the unmount for snapshots.
3388 3388 */
3389 3389 (void) dounmount(vfsp, MS_FORCE, kcred);
3390 3390 return (0);
3391 3391 }
3392 3392
3393 3393 /* ARGSUSED */
3394 3394 static int
3395 3395 zfs_unmount_snap_cb(const char *snapname, void *arg)
3396 3396 {
3397 3397 return (zfs_unmount_snap(snapname));
3398 3398 }
3399 3399
3400 3400 /*
3401 3401 * When a clone is destroyed, its origin may also need to be destroyed,
3402 3402 * in which case it must be unmounted. This routine will do that unmount
3403 3403 * if necessary.
3404 3404 */
3405 3405 void
3406 3406 zfs_destroy_unmount_origin(const char *fsname)
3407 3407 {
3408 3408 int error;
3409 3409 objset_t *os;
3410 3410 dsl_dataset_t *ds;
3411 3411
3412 3412 error = dmu_objset_hold(fsname, FTAG, &os);
3413 3413 if (error != 0)
3414 3414 return;
3415 3415 ds = dmu_objset_ds(os);
3416 3416 if (dsl_dir_is_clone(ds->ds_dir) && DS_IS_DEFER_DESTROY(ds->ds_prev)) {
3417 3417 char originname[MAXNAMELEN];
3418 3418 dsl_dataset_name(ds->ds_prev, originname);
3419 3419 dmu_objset_rele(os, FTAG);
3420 3420 (void) zfs_unmount_snap(originname);
3421 3421 } else {
3422 3422 dmu_objset_rele(os, FTAG);
3423 3423 }
3424 3424 }
3425 3425
3426 3426 /*
3427 3427 * innvl: {
3428 3428 * "snaps" -> { snapshot1, snapshot2 }
3429 3429 * (optional boolean) "defer"
3430 3430 * }
3431 3431 *
3432 3432 * outnvl: snapshot -> error code (int32)
3433 3433 *
3434 3434 */
3435 3435 static int
3436 3436 zfs_ioc_destroy_snaps(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3437 3437 {
3438 3438 int error, poollen;
3439 3439 nvlist_t *snaps;
3440 3440 nvpair_t *pair;
3441 3441 boolean_t defer;
3442 3442
3443 3443 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
3444 3444 return (SET_ERROR(EINVAL));
3445 3445 defer = nvlist_exists(innvl, "defer");
3446 3446
3447 3447 poollen = strlen(poolname);
3448 3448 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
3449 3449 pair = nvlist_next_nvpair(snaps, pair)) {
3450 3450 const char *name = nvpair_name(pair);
3451 3451
3452 3452 /*
3453 3453 * The snap must be in the specified pool.
3454 3454 */
3455 3455 if (strncmp(name, poolname, poollen) != 0 ||
3456 3456 (name[poollen] != '/' && name[poollen] != '@'))
3457 3457 return (SET_ERROR(EXDEV));
3458 3458
3459 3459 error = zfs_unmount_snap(name);
3460 3460 if (error != 0)
3461 3461 return (error);
3462 3462 }
3463 3463
3464 3464 return (dsl_destroy_snapshots_nvl(snaps, defer, outnvl));
3465 3465 }
3466 3466
3467 3467 /*
3468 3468 * inputs:
3469 3469 * zc_name name of dataset to destroy
3470 3470 * zc_objset_type type of objset
3471 3471 * zc_defer_destroy mark for deferred destroy
3472 3472 *
3473 3473 * outputs: none
3474 3474 */
3475 3475 static int
3476 3476 zfs_ioc_destroy(zfs_cmd_t *zc)
3477 3477 {
3478 3478 int err;
3479 3479
3480 3480 if (zc->zc_objset_type == DMU_OST_ZFS) {
3481 3481 err = zfs_unmount_snap(zc->zc_name);
3482 3482 if (err != 0)
3483 3483 return (err);
3484 3484 }
3485 3485
3486 3486 if (strchr(zc->zc_name, '@'))
3487 3487 err = dsl_destroy_snapshot(zc->zc_name, zc->zc_defer_destroy);
3488 3488 else
3489 3489 err = dsl_destroy_head(zc->zc_name);
3490 3490 if (zc->zc_objset_type == DMU_OST_ZVOL && err == 0)
3491 3491 (void) zvol_remove_minor(zc->zc_name);
3492 3492 return (err);
3493 3493 }
3494 3494
3495 3495 /*
3496 3496 * inputs:
3497 3497 * zc_name name of dataset to rollback (to most recent snapshot)
3498 3498 *
3499 3499 * outputs: none
3500 3500 */
3501 3501 static int
3502 3502 zfs_ioc_rollback(zfs_cmd_t *zc)
3503 3503 {
3504 3504 zfsvfs_t *zfsvfs;
3505 3505 int error;
3506 3506
3507 3507 if (getzfsvfs(zc->zc_name, &zfsvfs) == 0) {
3508 3508 error = zfs_suspend_fs(zfsvfs);
3509 3509 if (error == 0) {
3510 3510 int resume_err;
3511 3511
3512 3512 error = dsl_dataset_rollback(zc->zc_name);
3513 3513 resume_err = zfs_resume_fs(zfsvfs, zc->zc_name);
3514 3514 error = error ? error : resume_err;
3515 3515 }
3516 3516 VFS_RELE(zfsvfs->z_vfs);
3517 3517 } else {
3518 3518 error = dsl_dataset_rollback(zc->zc_name);
3519 3519 }
3520 3520 return (error);
3521 3521 }
3522 3522
3523 3523 static int
3524 3524 recursive_unmount(const char *fsname, void *arg)
3525 3525 {
3526 3526 const char *snapname = arg;
3527 3527 char fullname[MAXNAMELEN];
3528 3528
3529 3529 (void) snprintf(fullname, sizeof (fullname), "%s@%s", fsname, snapname);
3530 3530 return (zfs_unmount_snap(fullname));
3531 3531 }
3532 3532
3533 3533 /*
3534 3534 * inputs:
3535 3535 * zc_name old name of dataset
3536 3536 * zc_value new name of dataset
3537 3537 * zc_cookie recursive flag (only valid for snapshots)
3538 3538 *
3539 3539 * outputs: none
3540 3540 */
3541 3541 static int
3542 3542 zfs_ioc_rename(zfs_cmd_t *zc)
3543 3543 {
3544 3544 boolean_t recursive = zc->zc_cookie & 1;
|
↓ open down ↓ |
3544 lines elided |
↑ open up ↑ |
3545 3545 char *at;
3546 3546
3547 3547 zc->zc_value[sizeof (zc->zc_value) - 1] = '\0';
3548 3548 if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
3549 3549 strchr(zc->zc_value, '%'))
3550 3550 return (SET_ERROR(EINVAL));
3551 3551
3552 3552 at = strchr(zc->zc_name, '@');
3553 3553 if (at != NULL) {
3554 3554 /* snaps must be in same fs */
3555 + int error;
3556 +
3555 3557 if (strncmp(zc->zc_name, zc->zc_value, at - zc->zc_name + 1))
3556 3558 return (SET_ERROR(EXDEV));
3557 3559 *at = '\0';
3558 3560 if (zc->zc_objset_type == DMU_OST_ZFS) {
3559 - int error = dmu_objset_find(zc->zc_name,
3561 + error = dmu_objset_find(zc->zc_name,
3560 3562 recursive_unmount, at + 1,
3561 3563 recursive ? DS_FIND_CHILDREN : 0);
3562 - if (error != 0)
3564 + if (error != 0) {
3565 + *at = '@';
3563 3566 return (error);
3567 + }
3564 3568 }
3565 - return (dsl_dataset_rename_snapshot(zc->zc_name,
3566 - at + 1, strchr(zc->zc_value, '@') + 1, recursive));
3569 + error = dsl_dataset_rename_snapshot(zc->zc_name,
3570 + at + 1, strchr(zc->zc_value, '@') + 1, recursive);
3571 + *at = '@';
3572 +
3573 + return (error);
3567 3574 } else {
3568 3575 if (zc->zc_objset_type == DMU_OST_ZVOL)
3569 3576 (void) zvol_remove_minor(zc->zc_name);
3570 3577 return (dsl_dir_rename(zc->zc_name, zc->zc_value));
3571 3578 }
3572 3579 }
3573 3580
3574 3581 static int
3575 3582 zfs_check_settable(const char *dsname, nvpair_t *pair, cred_t *cr)
3576 3583 {
3577 3584 const char *propname = nvpair_name(pair);
3578 3585 boolean_t issnap = (strchr(dsname, '@') != NULL);
3579 3586 zfs_prop_t prop = zfs_name_to_prop(propname);
3580 3587 uint64_t intval;
3581 3588 int err;
3582 3589
3583 3590 if (prop == ZPROP_INVAL) {
3584 3591 if (zfs_prop_user(propname)) {
3585 3592 if (err = zfs_secpolicy_write_perms(dsname,
3586 3593 ZFS_DELEG_PERM_USERPROP, cr))
3587 3594 return (err);
3588 3595 return (0);
3589 3596 }
3590 3597
3591 3598 if (!issnap && zfs_prop_userquota(propname)) {
3592 3599 const char *perm = NULL;
3593 3600 const char *uq_prefix =
3594 3601 zfs_userquota_prop_prefixes[ZFS_PROP_USERQUOTA];
3595 3602 const char *gq_prefix =
3596 3603 zfs_userquota_prop_prefixes[ZFS_PROP_GROUPQUOTA];
3597 3604
3598 3605 if (strncmp(propname, uq_prefix,
3599 3606 strlen(uq_prefix)) == 0) {
3600 3607 perm = ZFS_DELEG_PERM_USERQUOTA;
3601 3608 } else if (strncmp(propname, gq_prefix,
3602 3609 strlen(gq_prefix)) == 0) {
3603 3610 perm = ZFS_DELEG_PERM_GROUPQUOTA;
3604 3611 } else {
3605 3612 /* USERUSED and GROUPUSED are read-only */
3606 3613 return (SET_ERROR(EINVAL));
3607 3614 }
3608 3615
3609 3616 if (err = zfs_secpolicy_write_perms(dsname, perm, cr))
3610 3617 return (err);
3611 3618 return (0);
3612 3619 }
3613 3620
3614 3621 return (SET_ERROR(EINVAL));
3615 3622 }
3616 3623
3617 3624 if (issnap)
3618 3625 return (SET_ERROR(EINVAL));
3619 3626
3620 3627 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
3621 3628 /*
3622 3629 * dsl_prop_get_all_impl() returns properties in this
3623 3630 * format.
3624 3631 */
3625 3632 nvlist_t *attrs;
3626 3633 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
3627 3634 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
3628 3635 &pair) == 0);
3629 3636 }
3630 3637
3631 3638 /*
3632 3639 * Check that this value is valid for this pool version
3633 3640 */
3634 3641 switch (prop) {
3635 3642 case ZFS_PROP_COMPRESSION:
3636 3643 /*
3637 3644 * If the user specified gzip compression, make sure
3638 3645 * the SPA supports it. We ignore any errors here since
3639 3646 * we'll catch them later.
3640 3647 */
3641 3648 if (nvpair_type(pair) == DATA_TYPE_UINT64 &&
3642 3649 nvpair_value_uint64(pair, &intval) == 0) {
3643 3650 if (intval >= ZIO_COMPRESS_GZIP_1 &&
3644 3651 intval <= ZIO_COMPRESS_GZIP_9 &&
3645 3652 zfs_earlier_version(dsname,
3646 3653 SPA_VERSION_GZIP_COMPRESSION)) {
3647 3654 return (SET_ERROR(ENOTSUP));
3648 3655 }
3649 3656
3650 3657 if (intval == ZIO_COMPRESS_ZLE &&
3651 3658 zfs_earlier_version(dsname,
3652 3659 SPA_VERSION_ZLE_COMPRESSION))
3653 3660 return (SET_ERROR(ENOTSUP));
3654 3661
3655 3662 if (intval == ZIO_COMPRESS_LZ4) {
3656 3663 zfeature_info_t *feature =
3657 3664 &spa_feature_table[
3658 3665 SPA_FEATURE_LZ4_COMPRESS];
3659 3666 spa_t *spa;
3660 3667
3661 3668 if ((err = spa_open(dsname, &spa, FTAG)) != 0)
3662 3669 return (err);
3663 3670
3664 3671 if (!spa_feature_is_enabled(spa, feature)) {
3665 3672 spa_close(spa, FTAG);
3666 3673 return (SET_ERROR(ENOTSUP));
3667 3674 }
3668 3675 spa_close(spa, FTAG);
3669 3676 }
3670 3677
3671 3678 /*
3672 3679 * If this is a bootable dataset then
3673 3680 * verify that the compression algorithm
3674 3681 * is supported for booting. We must return
3675 3682 * something other than ENOTSUP since it
3676 3683 * implies a downrev pool version.
3677 3684 */
3678 3685 if (zfs_is_bootfs(dsname) &&
3679 3686 !BOOTFS_COMPRESS_VALID(intval)) {
3680 3687 return (SET_ERROR(ERANGE));
3681 3688 }
3682 3689 }
3683 3690 break;
3684 3691
3685 3692 case ZFS_PROP_COPIES:
3686 3693 if (zfs_earlier_version(dsname, SPA_VERSION_DITTO_BLOCKS))
3687 3694 return (SET_ERROR(ENOTSUP));
3688 3695 break;
3689 3696
3690 3697 case ZFS_PROP_DEDUP:
3691 3698 if (zfs_earlier_version(dsname, SPA_VERSION_DEDUP))
3692 3699 return (SET_ERROR(ENOTSUP));
3693 3700 break;
3694 3701
3695 3702 case ZFS_PROP_SHARESMB:
3696 3703 if (zpl_earlier_version(dsname, ZPL_VERSION_FUID))
3697 3704 return (SET_ERROR(ENOTSUP));
3698 3705 break;
3699 3706
3700 3707 case ZFS_PROP_ACLINHERIT:
3701 3708 if (nvpair_type(pair) == DATA_TYPE_UINT64 &&
3702 3709 nvpair_value_uint64(pair, &intval) == 0) {
3703 3710 if (intval == ZFS_ACL_PASSTHROUGH_X &&
3704 3711 zfs_earlier_version(dsname,
3705 3712 SPA_VERSION_PASSTHROUGH_X))
3706 3713 return (SET_ERROR(ENOTSUP));
3707 3714 }
3708 3715 break;
3709 3716 }
3710 3717
3711 3718 return (zfs_secpolicy_setprop(dsname, prop, pair, CRED()));
3712 3719 }
3713 3720
3714 3721 /*
3715 3722 * Checks for a race condition to make sure we don't increment a feature flag
3716 3723 * multiple times.
3717 3724 */
3718 3725 static int
3719 3726 zfs_prop_activate_feature_check(void *arg, dmu_tx_t *tx)
3720 3727 {
3721 3728 spa_t *spa = dmu_tx_pool(tx)->dp_spa;
3722 3729 zfeature_info_t *feature = arg;
3723 3730
3724 3731 if (!spa_feature_is_active(spa, feature))
3725 3732 return (0);
3726 3733 else
3727 3734 return (SET_ERROR(EBUSY));
3728 3735 }
3729 3736
3730 3737 /*
3731 3738 * The callback invoked on feature activation in the sync task caused by
3732 3739 * zfs_prop_activate_feature.
3733 3740 */
3734 3741 static void
3735 3742 zfs_prop_activate_feature_sync(void *arg, dmu_tx_t *tx)
3736 3743 {
3737 3744 spa_t *spa = dmu_tx_pool(tx)->dp_spa;
3738 3745 zfeature_info_t *feature = arg;
3739 3746
3740 3747 spa_feature_incr(spa, feature, tx);
3741 3748 }
3742 3749
3743 3750 /*
3744 3751 * Activates a feature on a pool in response to a property setting. This
3745 3752 * creates a new sync task which modifies the pool to reflect the feature
3746 3753 * as being active.
3747 3754 */
3748 3755 static int
3749 3756 zfs_prop_activate_feature(spa_t *spa, zfeature_info_t *feature)
3750 3757 {
3751 3758 int err;
3752 3759
3753 3760 /* EBUSY here indicates that the feature is already active */
3754 3761 err = dsl_sync_task(spa_name(spa),
3755 3762 zfs_prop_activate_feature_check, zfs_prop_activate_feature_sync,
3756 3763 feature, 2);
3757 3764
3758 3765 if (err != 0 && err != EBUSY)
3759 3766 return (err);
3760 3767 else
3761 3768 return (0);
3762 3769 }
3763 3770
3764 3771 /*
3765 3772 * Removes properties from the given props list that fail permission checks
3766 3773 * needed to clear them and to restore them in case of a receive error. For each
3767 3774 * property, make sure we have both set and inherit permissions.
3768 3775 *
3769 3776 * Returns the first error encountered if any permission checks fail. If the
3770 3777 * caller provides a non-NULL errlist, it also gives the complete list of names
3771 3778 * of all the properties that failed a permission check along with the
3772 3779 * corresponding error numbers. The caller is responsible for freeing the
3773 3780 * returned errlist.
3774 3781 *
3775 3782 * If every property checks out successfully, zero is returned and the list
3776 3783 * pointed at by errlist is NULL.
3777 3784 */
3778 3785 static int
3779 3786 zfs_check_clearable(char *dataset, nvlist_t *props, nvlist_t **errlist)
3780 3787 {
3781 3788 zfs_cmd_t *zc;
3782 3789 nvpair_t *pair, *next_pair;
3783 3790 nvlist_t *errors;
3784 3791 int err, rv = 0;
3785 3792
3786 3793 if (props == NULL)
3787 3794 return (0);
3788 3795
3789 3796 VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0);
3790 3797
3791 3798 zc = kmem_alloc(sizeof (zfs_cmd_t), KM_SLEEP);
3792 3799 (void) strcpy(zc->zc_name, dataset);
3793 3800 pair = nvlist_next_nvpair(props, NULL);
3794 3801 while (pair != NULL) {
3795 3802 next_pair = nvlist_next_nvpair(props, pair);
3796 3803
3797 3804 (void) strcpy(zc->zc_value, nvpair_name(pair));
3798 3805 if ((err = zfs_check_settable(dataset, pair, CRED())) != 0 ||
3799 3806 (err = zfs_secpolicy_inherit_prop(zc, NULL, CRED())) != 0) {
3800 3807 VERIFY(nvlist_remove_nvpair(props, pair) == 0);
3801 3808 VERIFY(nvlist_add_int32(errors,
3802 3809 zc->zc_value, err) == 0);
3803 3810 }
3804 3811 pair = next_pair;
3805 3812 }
3806 3813 kmem_free(zc, sizeof (zfs_cmd_t));
3807 3814
3808 3815 if ((pair = nvlist_next_nvpair(errors, NULL)) == NULL) {
3809 3816 nvlist_free(errors);
3810 3817 errors = NULL;
3811 3818 } else {
3812 3819 VERIFY(nvpair_value_int32(pair, &rv) == 0);
3813 3820 }
3814 3821
3815 3822 if (errlist == NULL)
3816 3823 nvlist_free(errors);
3817 3824 else
3818 3825 *errlist = errors;
3819 3826
3820 3827 return (rv);
3821 3828 }
3822 3829
3823 3830 static boolean_t
3824 3831 propval_equals(nvpair_t *p1, nvpair_t *p2)
3825 3832 {
3826 3833 if (nvpair_type(p1) == DATA_TYPE_NVLIST) {
3827 3834 /* dsl_prop_get_all_impl() format */
3828 3835 nvlist_t *attrs;
3829 3836 VERIFY(nvpair_value_nvlist(p1, &attrs) == 0);
3830 3837 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
3831 3838 &p1) == 0);
3832 3839 }
3833 3840
3834 3841 if (nvpair_type(p2) == DATA_TYPE_NVLIST) {
3835 3842 nvlist_t *attrs;
3836 3843 VERIFY(nvpair_value_nvlist(p2, &attrs) == 0);
3837 3844 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
3838 3845 &p2) == 0);
3839 3846 }
3840 3847
3841 3848 if (nvpair_type(p1) != nvpair_type(p2))
3842 3849 return (B_FALSE);
3843 3850
3844 3851 if (nvpair_type(p1) == DATA_TYPE_STRING) {
3845 3852 char *valstr1, *valstr2;
3846 3853
3847 3854 VERIFY(nvpair_value_string(p1, (char **)&valstr1) == 0);
3848 3855 VERIFY(nvpair_value_string(p2, (char **)&valstr2) == 0);
3849 3856 return (strcmp(valstr1, valstr2) == 0);
3850 3857 } else {
3851 3858 uint64_t intval1, intval2;
3852 3859
3853 3860 VERIFY(nvpair_value_uint64(p1, &intval1) == 0);
3854 3861 VERIFY(nvpair_value_uint64(p2, &intval2) == 0);
3855 3862 return (intval1 == intval2);
3856 3863 }
3857 3864 }
3858 3865
3859 3866 /*
3860 3867 * Remove properties from props if they are not going to change (as determined
3861 3868 * by comparison with origprops). Remove them from origprops as well, since we
3862 3869 * do not need to clear or restore properties that won't change.
3863 3870 */
3864 3871 static void
3865 3872 props_reduce(nvlist_t *props, nvlist_t *origprops)
3866 3873 {
3867 3874 nvpair_t *pair, *next_pair;
3868 3875
3869 3876 if (origprops == NULL)
3870 3877 return; /* all props need to be received */
3871 3878
3872 3879 pair = nvlist_next_nvpair(props, NULL);
3873 3880 while (pair != NULL) {
3874 3881 const char *propname = nvpair_name(pair);
3875 3882 nvpair_t *match;
3876 3883
3877 3884 next_pair = nvlist_next_nvpair(props, pair);
3878 3885
3879 3886 if ((nvlist_lookup_nvpair(origprops, propname,
3880 3887 &match) != 0) || !propval_equals(pair, match))
3881 3888 goto next; /* need to set received value */
3882 3889
3883 3890 /* don't clear the existing received value */
3884 3891 (void) nvlist_remove_nvpair(origprops, match);
3885 3892 /* don't bother receiving the property */
3886 3893 (void) nvlist_remove_nvpair(props, pair);
3887 3894 next:
3888 3895 pair = next_pair;
3889 3896 }
3890 3897 }
3891 3898
3892 3899 #ifdef DEBUG
3893 3900 static boolean_t zfs_ioc_recv_inject_err;
3894 3901 #endif
3895 3902
3896 3903 /*
3897 3904 * inputs:
3898 3905 * zc_name name of containing filesystem
3899 3906 * zc_nvlist_src{_size} nvlist of properties to apply
3900 3907 * zc_value name of snapshot to create
3901 3908 * zc_string name of clone origin (if DRR_FLAG_CLONE)
3902 3909 * zc_cookie file descriptor to recv from
3903 3910 * zc_begin_record the BEGIN record of the stream (not byteswapped)
3904 3911 * zc_guid force flag
3905 3912 * zc_cleanup_fd cleanup-on-exit file descriptor
3906 3913 * zc_action_handle handle for this guid/ds mapping (or zero on first call)
3907 3914 *
3908 3915 * outputs:
3909 3916 * zc_cookie number of bytes read
3910 3917 * zc_nvlist_dst{_size} error for each unapplied received property
3911 3918 * zc_obj zprop_errflags_t
3912 3919 * zc_action_handle handle for this guid/ds mapping
3913 3920 */
3914 3921 static int
3915 3922 zfs_ioc_recv(zfs_cmd_t *zc)
3916 3923 {
3917 3924 file_t *fp;
3918 3925 dmu_recv_cookie_t drc;
3919 3926 boolean_t force = (boolean_t)zc->zc_guid;
3920 3927 int fd;
3921 3928 int error = 0;
3922 3929 int props_error = 0;
3923 3930 nvlist_t *errors;
3924 3931 offset_t off;
3925 3932 nvlist_t *props = NULL; /* sent properties */
3926 3933 nvlist_t *origprops = NULL; /* existing properties */
3927 3934 char *origin = NULL;
3928 3935 char *tosnap;
3929 3936 char tofs[ZFS_MAXNAMELEN];
3930 3937 boolean_t first_recvd_props = B_FALSE;
3931 3938
3932 3939 if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
3933 3940 strchr(zc->zc_value, '@') == NULL ||
3934 3941 strchr(zc->zc_value, '%'))
3935 3942 return (SET_ERROR(EINVAL));
3936 3943
3937 3944 (void) strcpy(tofs, zc->zc_value);
3938 3945 tosnap = strchr(tofs, '@');
3939 3946 *tosnap++ = '\0';
3940 3947
3941 3948 if (zc->zc_nvlist_src != NULL &&
3942 3949 (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
3943 3950 zc->zc_iflags, &props)) != 0)
3944 3951 return (error);
3945 3952
3946 3953 fd = zc->zc_cookie;
3947 3954 fp = getf(fd);
3948 3955 if (fp == NULL) {
3949 3956 nvlist_free(props);
3950 3957 return (SET_ERROR(EBADF));
3951 3958 }
3952 3959
3953 3960 VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0);
3954 3961
3955 3962 if (zc->zc_string[0])
3956 3963 origin = zc->zc_string;
3957 3964
3958 3965 error = dmu_recv_begin(tofs, tosnap,
3959 3966 &zc->zc_begin_record, force, origin, &drc);
3960 3967 if (error != 0)
3961 3968 goto out;
3962 3969
3963 3970 /*
3964 3971 * Set properties before we receive the stream so that they are applied
3965 3972 * to the new data. Note that we must call dmu_recv_stream() if
3966 3973 * dmu_recv_begin() succeeds.
3967 3974 */
3968 3975 if (props != NULL && !drc.drc_newfs) {
3969 3976 if (spa_version(dsl_dataset_get_spa(drc.drc_ds)) >=
3970 3977 SPA_VERSION_RECVD_PROPS &&
3971 3978 !dsl_prop_get_hasrecvd(tofs))
3972 3979 first_recvd_props = B_TRUE;
3973 3980
3974 3981 /*
3975 3982 * If new received properties are supplied, they are to
3976 3983 * completely replace the existing received properties, so stash
3977 3984 * away the existing ones.
3978 3985 */
3979 3986 if (dsl_prop_get_received(tofs, &origprops) == 0) {
3980 3987 nvlist_t *errlist = NULL;
3981 3988 /*
3982 3989 * Don't bother writing a property if its value won't
3983 3990 * change (and avoid the unnecessary security checks).
3984 3991 *
3985 3992 * The first receive after SPA_VERSION_RECVD_PROPS is a
3986 3993 * special case where we blow away all local properties
3987 3994 * regardless.
3988 3995 */
3989 3996 if (!first_recvd_props)
3990 3997 props_reduce(props, origprops);
3991 3998 if (zfs_check_clearable(tofs, origprops, &errlist) != 0)
3992 3999 (void) nvlist_merge(errors, errlist, 0);
3993 4000 nvlist_free(errlist);
3994 4001
3995 4002 if (clear_received_props(tofs, origprops,
3996 4003 first_recvd_props ? NULL : props) != 0)
3997 4004 zc->zc_obj |= ZPROP_ERR_NOCLEAR;
3998 4005 } else {
3999 4006 zc->zc_obj |= ZPROP_ERR_NOCLEAR;
4000 4007 }
4001 4008 }
4002 4009
4003 4010 if (props != NULL) {
4004 4011 props_error = dsl_prop_set_hasrecvd(tofs);
4005 4012
4006 4013 if (props_error == 0) {
4007 4014 (void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_RECEIVED,
4008 4015 props, errors);
4009 4016 }
4010 4017 }
4011 4018
4012 4019 if (zc->zc_nvlist_dst_size != 0 &&
4013 4020 (nvlist_smush(errors, zc->zc_nvlist_dst_size) != 0 ||
4014 4021 put_nvlist(zc, errors) != 0)) {
4015 4022 /*
4016 4023 * Caller made zc->zc_nvlist_dst less than the minimum expected
4017 4024 * size or supplied an invalid address.
4018 4025 */
4019 4026 props_error = SET_ERROR(EINVAL);
4020 4027 }
4021 4028
4022 4029 off = fp->f_offset;
4023 4030 error = dmu_recv_stream(&drc, fp->f_vnode, &off, zc->zc_cleanup_fd,
4024 4031 &zc->zc_action_handle);
4025 4032
4026 4033 if (error == 0) {
4027 4034 zfsvfs_t *zfsvfs = NULL;
4028 4035
4029 4036 if (getzfsvfs(tofs, &zfsvfs) == 0) {
4030 4037 /* online recv */
4031 4038 int end_err;
4032 4039
4033 4040 error = zfs_suspend_fs(zfsvfs);
4034 4041 /*
4035 4042 * If the suspend fails, then the recv_end will
4036 4043 * likely also fail, and clean up after itself.
4037 4044 */
4038 4045 end_err = dmu_recv_end(&drc);
4039 4046 if (error == 0)
4040 4047 error = zfs_resume_fs(zfsvfs, tofs);
4041 4048 error = error ? error : end_err;
4042 4049 VFS_RELE(zfsvfs->z_vfs);
4043 4050 } else {
4044 4051 error = dmu_recv_end(&drc);
4045 4052 }
4046 4053 }
4047 4054
4048 4055 zc->zc_cookie = off - fp->f_offset;
4049 4056 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
4050 4057 fp->f_offset = off;
4051 4058
4052 4059 #ifdef DEBUG
4053 4060 if (zfs_ioc_recv_inject_err) {
4054 4061 zfs_ioc_recv_inject_err = B_FALSE;
4055 4062 error = 1;
4056 4063 }
4057 4064 #endif
4058 4065 /*
4059 4066 * On error, restore the original props.
4060 4067 */
4061 4068 if (error != 0 && props != NULL && !drc.drc_newfs) {
4062 4069 if (clear_received_props(tofs, props, NULL) != 0) {
4063 4070 /*
4064 4071 * We failed to clear the received properties.
4065 4072 * Since we may have left a $recvd value on the
4066 4073 * system, we can't clear the $hasrecvd flag.
4067 4074 */
4068 4075 zc->zc_obj |= ZPROP_ERR_NORESTORE;
4069 4076 } else if (first_recvd_props) {
4070 4077 dsl_prop_unset_hasrecvd(tofs);
4071 4078 }
4072 4079
4073 4080 if (origprops == NULL && !drc.drc_newfs) {
4074 4081 /* We failed to stash the original properties. */
4075 4082 zc->zc_obj |= ZPROP_ERR_NORESTORE;
4076 4083 }
4077 4084
4078 4085 /*
4079 4086 * dsl_props_set() will not convert RECEIVED to LOCAL on or
4080 4087 * after SPA_VERSION_RECVD_PROPS, so we need to specify LOCAL
4081 4088 * explictly if we're restoring local properties cleared in the
4082 4089 * first new-style receive.
4083 4090 */
4084 4091 if (origprops != NULL &&
4085 4092 zfs_set_prop_nvlist(tofs, (first_recvd_props ?
4086 4093 ZPROP_SRC_LOCAL : ZPROP_SRC_RECEIVED),
4087 4094 origprops, NULL) != 0) {
4088 4095 /*
4089 4096 * We stashed the original properties but failed to
4090 4097 * restore them.
4091 4098 */
4092 4099 zc->zc_obj |= ZPROP_ERR_NORESTORE;
4093 4100 }
4094 4101 }
4095 4102 out:
4096 4103 nvlist_free(props);
4097 4104 nvlist_free(origprops);
4098 4105 nvlist_free(errors);
4099 4106 releasef(fd);
4100 4107
4101 4108 if (error == 0)
4102 4109 error = props_error;
4103 4110
4104 4111 return (error);
4105 4112 }
4106 4113
4107 4114 /*
4108 4115 * inputs:
4109 4116 * zc_name name of snapshot to send
4110 4117 * zc_cookie file descriptor to send stream to
4111 4118 * zc_obj fromorigin flag (mutually exclusive with zc_fromobj)
4112 4119 * zc_sendobj objsetid of snapshot to send
4113 4120 * zc_fromobj objsetid of incremental fromsnap (may be zero)
4114 4121 * zc_guid if set, estimate size of stream only. zc_cookie is ignored.
4115 4122 * output size in zc_objset_type.
4116 4123 *
4117 4124 * outputs: none
4118 4125 */
4119 4126 static int
4120 4127 zfs_ioc_send(zfs_cmd_t *zc)
4121 4128 {
4122 4129 int error;
4123 4130 offset_t off;
4124 4131 boolean_t estimate = (zc->zc_guid != 0);
4125 4132
4126 4133 if (zc->zc_obj != 0) {
4127 4134 dsl_pool_t *dp;
4128 4135 dsl_dataset_t *tosnap;
4129 4136
4130 4137 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
4131 4138 if (error != 0)
4132 4139 return (error);
4133 4140
4134 4141 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &tosnap);
4135 4142 if (error != 0) {
4136 4143 dsl_pool_rele(dp, FTAG);
4137 4144 return (error);
4138 4145 }
4139 4146
4140 4147 if (dsl_dir_is_clone(tosnap->ds_dir))
4141 4148 zc->zc_fromobj = tosnap->ds_dir->dd_phys->dd_origin_obj;
4142 4149 dsl_dataset_rele(tosnap, FTAG);
4143 4150 dsl_pool_rele(dp, FTAG);
4144 4151 }
4145 4152
4146 4153 if (estimate) {
4147 4154 dsl_pool_t *dp;
4148 4155 dsl_dataset_t *tosnap;
4149 4156 dsl_dataset_t *fromsnap = NULL;
4150 4157
4151 4158 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
4152 4159 if (error != 0)
4153 4160 return (error);
4154 4161
4155 4162 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &tosnap);
4156 4163 if (error != 0) {
4157 4164 dsl_pool_rele(dp, FTAG);
4158 4165 return (error);
4159 4166 }
4160 4167
4161 4168 if (zc->zc_fromobj != 0) {
4162 4169 error = dsl_dataset_hold_obj(dp, zc->zc_fromobj,
4163 4170 FTAG, &fromsnap);
4164 4171 if (error != 0) {
4165 4172 dsl_dataset_rele(tosnap, FTAG);
4166 4173 dsl_pool_rele(dp, FTAG);
4167 4174 return (error);
4168 4175 }
4169 4176 }
4170 4177
4171 4178 error = dmu_send_estimate(tosnap, fromsnap,
4172 4179 &zc->zc_objset_type);
4173 4180
4174 4181 if (fromsnap != NULL)
4175 4182 dsl_dataset_rele(fromsnap, FTAG);
4176 4183 dsl_dataset_rele(tosnap, FTAG);
4177 4184 dsl_pool_rele(dp, FTAG);
4178 4185 } else {
4179 4186 file_t *fp = getf(zc->zc_cookie);
4180 4187 if (fp == NULL)
4181 4188 return (SET_ERROR(EBADF));
4182 4189
4183 4190 off = fp->f_offset;
4184 4191 error = dmu_send_obj(zc->zc_name, zc->zc_sendobj,
4185 4192 zc->zc_fromobj, zc->zc_cookie, fp->f_vnode, &off);
4186 4193
4187 4194 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
4188 4195 fp->f_offset = off;
4189 4196 releasef(zc->zc_cookie);
4190 4197 }
4191 4198 return (error);
4192 4199 }
4193 4200
4194 4201 /*
4195 4202 * inputs:
4196 4203 * zc_name name of snapshot on which to report progress
4197 4204 * zc_cookie file descriptor of send stream
4198 4205 *
4199 4206 * outputs:
4200 4207 * zc_cookie number of bytes written in send stream thus far
4201 4208 */
4202 4209 static int
4203 4210 zfs_ioc_send_progress(zfs_cmd_t *zc)
4204 4211 {
4205 4212 dsl_pool_t *dp;
4206 4213 dsl_dataset_t *ds;
4207 4214 dmu_sendarg_t *dsp = NULL;
4208 4215 int error;
4209 4216
4210 4217 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
4211 4218 if (error != 0)
4212 4219 return (error);
4213 4220
4214 4221 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &ds);
4215 4222 if (error != 0) {
4216 4223 dsl_pool_rele(dp, FTAG);
4217 4224 return (error);
4218 4225 }
4219 4226
4220 4227 mutex_enter(&ds->ds_sendstream_lock);
4221 4228
4222 4229 /*
4223 4230 * Iterate over all the send streams currently active on this dataset.
4224 4231 * If there's one which matches the specified file descriptor _and_ the
4225 4232 * stream was started by the current process, return the progress of
4226 4233 * that stream.
4227 4234 */
4228 4235 for (dsp = list_head(&ds->ds_sendstreams); dsp != NULL;
4229 4236 dsp = list_next(&ds->ds_sendstreams, dsp)) {
4230 4237 if (dsp->dsa_outfd == zc->zc_cookie &&
4231 4238 dsp->dsa_proc == curproc)
4232 4239 break;
4233 4240 }
4234 4241
4235 4242 if (dsp != NULL)
4236 4243 zc->zc_cookie = *(dsp->dsa_off);
4237 4244 else
4238 4245 error = SET_ERROR(ENOENT);
4239 4246
4240 4247 mutex_exit(&ds->ds_sendstream_lock);
4241 4248 dsl_dataset_rele(ds, FTAG);
4242 4249 dsl_pool_rele(dp, FTAG);
4243 4250 return (error);
4244 4251 }
4245 4252
4246 4253 static int
4247 4254 zfs_ioc_inject_fault(zfs_cmd_t *zc)
4248 4255 {
4249 4256 int id, error;
4250 4257
4251 4258 error = zio_inject_fault(zc->zc_name, (int)zc->zc_guid, &id,
4252 4259 &zc->zc_inject_record);
4253 4260
4254 4261 if (error == 0)
4255 4262 zc->zc_guid = (uint64_t)id;
4256 4263
4257 4264 return (error);
4258 4265 }
4259 4266
4260 4267 static int
4261 4268 zfs_ioc_clear_fault(zfs_cmd_t *zc)
4262 4269 {
4263 4270 return (zio_clear_fault((int)zc->zc_guid));
4264 4271 }
4265 4272
4266 4273 static int
4267 4274 zfs_ioc_inject_list_next(zfs_cmd_t *zc)
4268 4275 {
4269 4276 int id = (int)zc->zc_guid;
4270 4277 int error;
4271 4278
4272 4279 error = zio_inject_list_next(&id, zc->zc_name, sizeof (zc->zc_name),
4273 4280 &zc->zc_inject_record);
4274 4281
4275 4282 zc->zc_guid = id;
4276 4283
4277 4284 return (error);
4278 4285 }
4279 4286
4280 4287 static int
4281 4288 zfs_ioc_error_log(zfs_cmd_t *zc)
4282 4289 {
4283 4290 spa_t *spa;
4284 4291 int error;
4285 4292 size_t count = (size_t)zc->zc_nvlist_dst_size;
4286 4293
4287 4294 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
4288 4295 return (error);
4289 4296
4290 4297 error = spa_get_errlog(spa, (void *)(uintptr_t)zc->zc_nvlist_dst,
4291 4298 &count);
4292 4299 if (error == 0)
4293 4300 zc->zc_nvlist_dst_size = count;
4294 4301 else
4295 4302 zc->zc_nvlist_dst_size = spa_get_errlog_size(spa);
4296 4303
4297 4304 spa_close(spa, FTAG);
4298 4305
4299 4306 return (error);
4300 4307 }
4301 4308
4302 4309 static int
4303 4310 zfs_ioc_clear(zfs_cmd_t *zc)
4304 4311 {
4305 4312 spa_t *spa;
4306 4313 vdev_t *vd;
4307 4314 int error;
4308 4315
4309 4316 /*
4310 4317 * On zpool clear we also fix up missing slogs
4311 4318 */
4312 4319 mutex_enter(&spa_namespace_lock);
4313 4320 spa = spa_lookup(zc->zc_name);
4314 4321 if (spa == NULL) {
4315 4322 mutex_exit(&spa_namespace_lock);
4316 4323 return (SET_ERROR(EIO));
4317 4324 }
4318 4325 if (spa_get_log_state(spa) == SPA_LOG_MISSING) {
4319 4326 /* we need to let spa_open/spa_load clear the chains */
4320 4327 spa_set_log_state(spa, SPA_LOG_CLEAR);
4321 4328 }
4322 4329 spa->spa_last_open_failed = 0;
4323 4330 mutex_exit(&spa_namespace_lock);
4324 4331
4325 4332 if (zc->zc_cookie & ZPOOL_NO_REWIND) {
4326 4333 error = spa_open(zc->zc_name, &spa, FTAG);
4327 4334 } else {
4328 4335 nvlist_t *policy;
4329 4336 nvlist_t *config = NULL;
4330 4337
4331 4338 if (zc->zc_nvlist_src == NULL)
4332 4339 return (SET_ERROR(EINVAL));
4333 4340
4334 4341 if ((error = get_nvlist(zc->zc_nvlist_src,
4335 4342 zc->zc_nvlist_src_size, zc->zc_iflags, &policy)) == 0) {
4336 4343 error = spa_open_rewind(zc->zc_name, &spa, FTAG,
4337 4344 policy, &config);
4338 4345 if (config != NULL) {
4339 4346 int err;
4340 4347
4341 4348 if ((err = put_nvlist(zc, config)) != 0)
4342 4349 error = err;
4343 4350 nvlist_free(config);
4344 4351 }
4345 4352 nvlist_free(policy);
4346 4353 }
4347 4354 }
4348 4355
4349 4356 if (error != 0)
4350 4357 return (error);
4351 4358
4352 4359 spa_vdev_state_enter(spa, SCL_NONE);
4353 4360
4354 4361 if (zc->zc_guid == 0) {
4355 4362 vd = NULL;
4356 4363 } else {
4357 4364 vd = spa_lookup_by_guid(spa, zc->zc_guid, B_TRUE);
4358 4365 if (vd == NULL) {
4359 4366 (void) spa_vdev_state_exit(spa, NULL, ENODEV);
4360 4367 spa_close(spa, FTAG);
4361 4368 return (SET_ERROR(ENODEV));
4362 4369 }
4363 4370 }
4364 4371
4365 4372 vdev_clear(spa, vd);
4366 4373
4367 4374 (void) spa_vdev_state_exit(spa, NULL, 0);
4368 4375
4369 4376 /*
4370 4377 * Resume any suspended I/Os.
4371 4378 */
4372 4379 if (zio_resume(spa) != 0)
4373 4380 error = SET_ERROR(EIO);
4374 4381
4375 4382 spa_close(spa, FTAG);
4376 4383
4377 4384 return (error);
4378 4385 }
4379 4386
4380 4387 static int
4381 4388 zfs_ioc_pool_reopen(zfs_cmd_t *zc)
4382 4389 {
4383 4390 spa_t *spa;
4384 4391 int error;
4385 4392
4386 4393 error = spa_open(zc->zc_name, &spa, FTAG);
4387 4394 if (error != 0)
4388 4395 return (error);
4389 4396
4390 4397 spa_vdev_state_enter(spa, SCL_NONE);
4391 4398
4392 4399 /*
4393 4400 * If a resilver is already in progress then set the
4394 4401 * spa_scrub_reopen flag to B_TRUE so that we don't restart
4395 4402 * the scan as a side effect of the reopen. Otherwise, let
4396 4403 * vdev_open() decided if a resilver is required.
4397 4404 */
4398 4405 spa->spa_scrub_reopen = dsl_scan_resilvering(spa->spa_dsl_pool);
4399 4406 vdev_reopen(spa->spa_root_vdev);
4400 4407 spa->spa_scrub_reopen = B_FALSE;
4401 4408
4402 4409 (void) spa_vdev_state_exit(spa, NULL, 0);
4403 4410 spa_close(spa, FTAG);
4404 4411 return (0);
4405 4412 }
4406 4413 /*
4407 4414 * inputs:
4408 4415 * zc_name name of filesystem
4409 4416 * zc_value name of origin snapshot
4410 4417 *
4411 4418 * outputs:
4412 4419 * zc_string name of conflicting snapshot, if there is one
4413 4420 */
4414 4421 static int
4415 4422 zfs_ioc_promote(zfs_cmd_t *zc)
4416 4423 {
4417 4424 char *cp;
4418 4425
4419 4426 /*
4420 4427 * We don't need to unmount *all* the origin fs's snapshots, but
4421 4428 * it's easier.
4422 4429 */
4423 4430 cp = strchr(zc->zc_value, '@');
4424 4431 if (cp)
4425 4432 *cp = '\0';
4426 4433 (void) dmu_objset_find(zc->zc_value,
4427 4434 zfs_unmount_snap_cb, NULL, DS_FIND_SNAPSHOTS);
4428 4435 return (dsl_dataset_promote(zc->zc_name, zc->zc_string));
4429 4436 }
4430 4437
4431 4438 /*
4432 4439 * Retrieve a single {user|group}{used|quota}@... property.
4433 4440 *
4434 4441 * inputs:
4435 4442 * zc_name name of filesystem
4436 4443 * zc_objset_type zfs_userquota_prop_t
4437 4444 * zc_value domain name (eg. "S-1-234-567-89")
4438 4445 * zc_guid RID/UID/GID
4439 4446 *
4440 4447 * outputs:
4441 4448 * zc_cookie property value
4442 4449 */
4443 4450 static int
4444 4451 zfs_ioc_userspace_one(zfs_cmd_t *zc)
4445 4452 {
4446 4453 zfsvfs_t *zfsvfs;
4447 4454 int error;
4448 4455
4449 4456 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
4450 4457 return (SET_ERROR(EINVAL));
4451 4458
4452 4459 error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE);
4453 4460 if (error != 0)
4454 4461 return (error);
4455 4462
4456 4463 error = zfs_userspace_one(zfsvfs,
4457 4464 zc->zc_objset_type, zc->zc_value, zc->zc_guid, &zc->zc_cookie);
4458 4465 zfsvfs_rele(zfsvfs, FTAG);
4459 4466
4460 4467 return (error);
4461 4468 }
4462 4469
4463 4470 /*
4464 4471 * inputs:
4465 4472 * zc_name name of filesystem
4466 4473 * zc_cookie zap cursor
4467 4474 * zc_objset_type zfs_userquota_prop_t
4468 4475 * zc_nvlist_dst[_size] buffer to fill (not really an nvlist)
4469 4476 *
4470 4477 * outputs:
4471 4478 * zc_nvlist_dst[_size] data buffer (array of zfs_useracct_t)
4472 4479 * zc_cookie zap cursor
4473 4480 */
4474 4481 static int
4475 4482 zfs_ioc_userspace_many(zfs_cmd_t *zc)
4476 4483 {
4477 4484 zfsvfs_t *zfsvfs;
4478 4485 int bufsize = zc->zc_nvlist_dst_size;
4479 4486
4480 4487 if (bufsize <= 0)
4481 4488 return (SET_ERROR(ENOMEM));
4482 4489
4483 4490 int error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE);
4484 4491 if (error != 0)
4485 4492 return (error);
4486 4493
4487 4494 void *buf = kmem_alloc(bufsize, KM_SLEEP);
4488 4495
4489 4496 error = zfs_userspace_many(zfsvfs, zc->zc_objset_type, &zc->zc_cookie,
4490 4497 buf, &zc->zc_nvlist_dst_size);
4491 4498
4492 4499 if (error == 0) {
4493 4500 error = xcopyout(buf,
4494 4501 (void *)(uintptr_t)zc->zc_nvlist_dst,
4495 4502 zc->zc_nvlist_dst_size);
4496 4503 }
4497 4504 kmem_free(buf, bufsize);
4498 4505 zfsvfs_rele(zfsvfs, FTAG);
4499 4506
4500 4507 return (error);
4501 4508 }
4502 4509
4503 4510 /*
4504 4511 * inputs:
4505 4512 * zc_name name of filesystem
4506 4513 *
4507 4514 * outputs:
4508 4515 * none
4509 4516 */
4510 4517 static int
4511 4518 zfs_ioc_userspace_upgrade(zfs_cmd_t *zc)
4512 4519 {
4513 4520 objset_t *os;
4514 4521 int error = 0;
4515 4522 zfsvfs_t *zfsvfs;
4516 4523
4517 4524 if (getzfsvfs(zc->zc_name, &zfsvfs) == 0) {
4518 4525 if (!dmu_objset_userused_enabled(zfsvfs->z_os)) {
4519 4526 /*
4520 4527 * If userused is not enabled, it may be because the
4521 4528 * objset needs to be closed & reopened (to grow the
4522 4529 * objset_phys_t). Suspend/resume the fs will do that.
4523 4530 */
4524 4531 error = zfs_suspend_fs(zfsvfs);
4525 4532 if (error == 0)
4526 4533 error = zfs_resume_fs(zfsvfs, zc->zc_name);
4527 4534 }
4528 4535 if (error == 0)
4529 4536 error = dmu_objset_userspace_upgrade(zfsvfs->z_os);
4530 4537 VFS_RELE(zfsvfs->z_vfs);
4531 4538 } else {
4532 4539 /* XXX kind of reading contents without owning */
4533 4540 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
4534 4541 if (error != 0)
4535 4542 return (error);
4536 4543
4537 4544 error = dmu_objset_userspace_upgrade(os);
4538 4545 dmu_objset_rele(os, FTAG);
4539 4546 }
4540 4547
4541 4548 return (error);
4542 4549 }
4543 4550
4544 4551 /*
4545 4552 * We don't want to have a hard dependency
4546 4553 * against some special symbols in sharefs
4547 4554 * nfs, and smbsrv. Determine them if needed when
4548 4555 * the first file system is shared.
4549 4556 * Neither sharefs, nfs or smbsrv are unloadable modules.
4550 4557 */
4551 4558 int (*znfsexport_fs)(void *arg);
4552 4559 int (*zshare_fs)(enum sharefs_sys_op, share_t *, uint32_t);
4553 4560 int (*zsmbexport_fs)(void *arg, boolean_t add_share);
4554 4561
4555 4562 int zfs_nfsshare_inited;
4556 4563 int zfs_smbshare_inited;
4557 4564
4558 4565 ddi_modhandle_t nfs_mod;
4559 4566 ddi_modhandle_t sharefs_mod;
4560 4567 ddi_modhandle_t smbsrv_mod;
4561 4568 kmutex_t zfs_share_lock;
4562 4569
4563 4570 static int
4564 4571 zfs_init_sharefs()
4565 4572 {
4566 4573 int error;
4567 4574
4568 4575 ASSERT(MUTEX_HELD(&zfs_share_lock));
4569 4576 /* Both NFS and SMB shares also require sharetab support. */
4570 4577 if (sharefs_mod == NULL && ((sharefs_mod =
4571 4578 ddi_modopen("fs/sharefs",
4572 4579 KRTLD_MODE_FIRST, &error)) == NULL)) {
4573 4580 return (SET_ERROR(ENOSYS));
4574 4581 }
4575 4582 if (zshare_fs == NULL && ((zshare_fs =
4576 4583 (int (*)(enum sharefs_sys_op, share_t *, uint32_t))
4577 4584 ddi_modsym(sharefs_mod, "sharefs_impl", &error)) == NULL)) {
4578 4585 return (SET_ERROR(ENOSYS));
4579 4586 }
4580 4587 return (0);
4581 4588 }
4582 4589
4583 4590 static int
4584 4591 zfs_ioc_share(zfs_cmd_t *zc)
4585 4592 {
4586 4593 int error;
4587 4594 int opcode;
4588 4595
4589 4596 switch (zc->zc_share.z_sharetype) {
4590 4597 case ZFS_SHARE_NFS:
4591 4598 case ZFS_UNSHARE_NFS:
4592 4599 if (zfs_nfsshare_inited == 0) {
4593 4600 mutex_enter(&zfs_share_lock);
4594 4601 if (nfs_mod == NULL && ((nfs_mod = ddi_modopen("fs/nfs",
4595 4602 KRTLD_MODE_FIRST, &error)) == NULL)) {
4596 4603 mutex_exit(&zfs_share_lock);
4597 4604 return (SET_ERROR(ENOSYS));
4598 4605 }
4599 4606 if (znfsexport_fs == NULL &&
4600 4607 ((znfsexport_fs = (int (*)(void *))
4601 4608 ddi_modsym(nfs_mod,
4602 4609 "nfs_export", &error)) == NULL)) {
4603 4610 mutex_exit(&zfs_share_lock);
4604 4611 return (SET_ERROR(ENOSYS));
4605 4612 }
4606 4613 error = zfs_init_sharefs();
4607 4614 if (error != 0) {
4608 4615 mutex_exit(&zfs_share_lock);
4609 4616 return (SET_ERROR(ENOSYS));
4610 4617 }
4611 4618 zfs_nfsshare_inited = 1;
4612 4619 mutex_exit(&zfs_share_lock);
4613 4620 }
4614 4621 break;
4615 4622 case ZFS_SHARE_SMB:
4616 4623 case ZFS_UNSHARE_SMB:
4617 4624 if (zfs_smbshare_inited == 0) {
4618 4625 mutex_enter(&zfs_share_lock);
4619 4626 if (smbsrv_mod == NULL && ((smbsrv_mod =
4620 4627 ddi_modopen("drv/smbsrv",
4621 4628 KRTLD_MODE_FIRST, &error)) == NULL)) {
4622 4629 mutex_exit(&zfs_share_lock);
4623 4630 return (SET_ERROR(ENOSYS));
4624 4631 }
4625 4632 if (zsmbexport_fs == NULL && ((zsmbexport_fs =
4626 4633 (int (*)(void *, boolean_t))ddi_modsym(smbsrv_mod,
4627 4634 "smb_server_share", &error)) == NULL)) {
4628 4635 mutex_exit(&zfs_share_lock);
4629 4636 return (SET_ERROR(ENOSYS));
4630 4637 }
4631 4638 error = zfs_init_sharefs();
4632 4639 if (error != 0) {
4633 4640 mutex_exit(&zfs_share_lock);
4634 4641 return (SET_ERROR(ENOSYS));
4635 4642 }
4636 4643 zfs_smbshare_inited = 1;
4637 4644 mutex_exit(&zfs_share_lock);
4638 4645 }
4639 4646 break;
4640 4647 default:
4641 4648 return (SET_ERROR(EINVAL));
4642 4649 }
4643 4650
4644 4651 switch (zc->zc_share.z_sharetype) {
4645 4652 case ZFS_SHARE_NFS:
4646 4653 case ZFS_UNSHARE_NFS:
4647 4654 if (error =
4648 4655 znfsexport_fs((void *)
4649 4656 (uintptr_t)zc->zc_share.z_exportdata))
4650 4657 return (error);
4651 4658 break;
4652 4659 case ZFS_SHARE_SMB:
4653 4660 case ZFS_UNSHARE_SMB:
4654 4661 if (error = zsmbexport_fs((void *)
4655 4662 (uintptr_t)zc->zc_share.z_exportdata,
4656 4663 zc->zc_share.z_sharetype == ZFS_SHARE_SMB ?
4657 4664 B_TRUE: B_FALSE)) {
4658 4665 return (error);
4659 4666 }
4660 4667 break;
4661 4668 }
4662 4669
4663 4670 opcode = (zc->zc_share.z_sharetype == ZFS_SHARE_NFS ||
4664 4671 zc->zc_share.z_sharetype == ZFS_SHARE_SMB) ?
4665 4672 SHAREFS_ADD : SHAREFS_REMOVE;
4666 4673
4667 4674 /*
4668 4675 * Add or remove share from sharetab
4669 4676 */
4670 4677 error = zshare_fs(opcode,
4671 4678 (void *)(uintptr_t)zc->zc_share.z_sharedata,
4672 4679 zc->zc_share.z_sharemax);
4673 4680
4674 4681 return (error);
4675 4682
4676 4683 }
4677 4684
4678 4685 ace_t full_access[] = {
4679 4686 {(uid_t)-1, ACE_ALL_PERMS, ACE_EVERYONE, 0}
4680 4687 };
4681 4688
4682 4689 /*
4683 4690 * inputs:
4684 4691 * zc_name name of containing filesystem
4685 4692 * zc_obj object # beyond which we want next in-use object #
4686 4693 *
4687 4694 * outputs:
4688 4695 * zc_obj next in-use object #
4689 4696 */
4690 4697 static int
4691 4698 zfs_ioc_next_obj(zfs_cmd_t *zc)
4692 4699 {
4693 4700 objset_t *os = NULL;
4694 4701 int error;
4695 4702
4696 4703 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
4697 4704 if (error != 0)
4698 4705 return (error);
4699 4706
4700 4707 error = dmu_object_next(os, &zc->zc_obj, B_FALSE,
4701 4708 os->os_dsl_dataset->ds_phys->ds_prev_snap_txg);
4702 4709
4703 4710 dmu_objset_rele(os, FTAG);
4704 4711 return (error);
4705 4712 }
4706 4713
4707 4714 /*
4708 4715 * inputs:
4709 4716 * zc_name name of filesystem
4710 4717 * zc_value prefix name for snapshot
4711 4718 * zc_cleanup_fd cleanup-on-exit file descriptor for calling process
4712 4719 *
4713 4720 * outputs:
4714 4721 * zc_value short name of new snapshot
4715 4722 */
4716 4723 static int
4717 4724 zfs_ioc_tmp_snapshot(zfs_cmd_t *zc)
4718 4725 {
4719 4726 char *snap_name;
4720 4727 char *hold_name;
4721 4728 int error;
4722 4729 minor_t minor;
4723 4730
4724 4731 error = zfs_onexit_fd_hold(zc->zc_cleanup_fd, &minor);
4725 4732 if (error != 0)
4726 4733 return (error);
4727 4734
4728 4735 snap_name = kmem_asprintf("%s-%016llx", zc->zc_value,
4729 4736 (u_longlong_t)ddi_get_lbolt64());
4730 4737 hold_name = kmem_asprintf("%%%s", zc->zc_value);
4731 4738
4732 4739 error = dsl_dataset_snapshot_tmp(zc->zc_name, snap_name, minor,
4733 4740 hold_name);
4734 4741 if (error == 0)
4735 4742 (void) strcpy(zc->zc_value, snap_name);
4736 4743 strfree(snap_name);
4737 4744 strfree(hold_name);
4738 4745 zfs_onexit_fd_rele(zc->zc_cleanup_fd);
4739 4746 return (error);
4740 4747 }
4741 4748
4742 4749 /*
4743 4750 * inputs:
4744 4751 * zc_name name of "to" snapshot
4745 4752 * zc_value name of "from" snapshot
4746 4753 * zc_cookie file descriptor to write diff data on
4747 4754 *
4748 4755 * outputs:
4749 4756 * dmu_diff_record_t's to the file descriptor
4750 4757 */
4751 4758 static int
4752 4759 zfs_ioc_diff(zfs_cmd_t *zc)
4753 4760 {
4754 4761 file_t *fp;
4755 4762 offset_t off;
4756 4763 int error;
4757 4764
4758 4765 fp = getf(zc->zc_cookie);
4759 4766 if (fp == NULL)
4760 4767 return (SET_ERROR(EBADF));
4761 4768
4762 4769 off = fp->f_offset;
4763 4770
4764 4771 error = dmu_diff(zc->zc_name, zc->zc_value, fp->f_vnode, &off);
4765 4772
4766 4773 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
4767 4774 fp->f_offset = off;
4768 4775 releasef(zc->zc_cookie);
4769 4776
4770 4777 return (error);
4771 4778 }
4772 4779
4773 4780 /*
4774 4781 * Remove all ACL files in shares dir
4775 4782 */
4776 4783 static int
4777 4784 zfs_smb_acl_purge(znode_t *dzp)
4778 4785 {
4779 4786 zap_cursor_t zc;
4780 4787 zap_attribute_t zap;
4781 4788 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
4782 4789 int error;
4783 4790
4784 4791 for (zap_cursor_init(&zc, zfsvfs->z_os, dzp->z_id);
4785 4792 (error = zap_cursor_retrieve(&zc, &zap)) == 0;
4786 4793 zap_cursor_advance(&zc)) {
4787 4794 if ((error = VOP_REMOVE(ZTOV(dzp), zap.za_name, kcred,
4788 4795 NULL, 0)) != 0)
4789 4796 break;
4790 4797 }
4791 4798 zap_cursor_fini(&zc);
4792 4799 return (error);
4793 4800 }
4794 4801
4795 4802 static int
4796 4803 zfs_ioc_smb_acl(zfs_cmd_t *zc)
4797 4804 {
4798 4805 vnode_t *vp;
4799 4806 znode_t *dzp;
4800 4807 vnode_t *resourcevp = NULL;
4801 4808 znode_t *sharedir;
4802 4809 zfsvfs_t *zfsvfs;
4803 4810 nvlist_t *nvlist;
4804 4811 char *src, *target;
4805 4812 vattr_t vattr;
4806 4813 vsecattr_t vsec;
4807 4814 int error = 0;
4808 4815
4809 4816 if ((error = lookupname(zc->zc_value, UIO_SYSSPACE,
4810 4817 NO_FOLLOW, NULL, &vp)) != 0)
4811 4818 return (error);
4812 4819
4813 4820 /* Now make sure mntpnt and dataset are ZFS */
4814 4821
4815 4822 if (vp->v_vfsp->vfs_fstype != zfsfstype ||
4816 4823 (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource),
4817 4824 zc->zc_name) != 0)) {
4818 4825 VN_RELE(vp);
4819 4826 return (SET_ERROR(EINVAL));
4820 4827 }
4821 4828
4822 4829 dzp = VTOZ(vp);
4823 4830 zfsvfs = dzp->z_zfsvfs;
4824 4831 ZFS_ENTER(zfsvfs);
4825 4832
4826 4833 /*
4827 4834 * Create share dir if its missing.
4828 4835 */
4829 4836 mutex_enter(&zfsvfs->z_lock);
4830 4837 if (zfsvfs->z_shares_dir == 0) {
4831 4838 dmu_tx_t *tx;
4832 4839
4833 4840 tx = dmu_tx_create(zfsvfs->z_os);
4834 4841 dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, TRUE,
4835 4842 ZFS_SHARES_DIR);
4836 4843 dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL);
4837 4844 error = dmu_tx_assign(tx, TXG_WAIT);
4838 4845 if (error != 0) {
4839 4846 dmu_tx_abort(tx);
4840 4847 } else {
4841 4848 error = zfs_create_share_dir(zfsvfs, tx);
4842 4849 dmu_tx_commit(tx);
4843 4850 }
4844 4851 if (error != 0) {
4845 4852 mutex_exit(&zfsvfs->z_lock);
4846 4853 VN_RELE(vp);
4847 4854 ZFS_EXIT(zfsvfs);
4848 4855 return (error);
4849 4856 }
4850 4857 }
4851 4858 mutex_exit(&zfsvfs->z_lock);
4852 4859
4853 4860 ASSERT(zfsvfs->z_shares_dir);
4854 4861 if ((error = zfs_zget(zfsvfs, zfsvfs->z_shares_dir, &sharedir)) != 0) {
4855 4862 VN_RELE(vp);
4856 4863 ZFS_EXIT(zfsvfs);
4857 4864 return (error);
4858 4865 }
4859 4866
4860 4867 switch (zc->zc_cookie) {
4861 4868 case ZFS_SMB_ACL_ADD:
4862 4869 vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE;
4863 4870 vattr.va_type = VREG;
4864 4871 vattr.va_mode = S_IFREG|0777;
4865 4872 vattr.va_uid = 0;
4866 4873 vattr.va_gid = 0;
4867 4874
4868 4875 vsec.vsa_mask = VSA_ACE;
4869 4876 vsec.vsa_aclentp = &full_access;
4870 4877 vsec.vsa_aclentsz = sizeof (full_access);
4871 4878 vsec.vsa_aclcnt = 1;
4872 4879
4873 4880 error = VOP_CREATE(ZTOV(sharedir), zc->zc_string,
4874 4881 &vattr, EXCL, 0, &resourcevp, kcred, 0, NULL, &vsec);
4875 4882 if (resourcevp)
4876 4883 VN_RELE(resourcevp);
4877 4884 break;
4878 4885
4879 4886 case ZFS_SMB_ACL_REMOVE:
4880 4887 error = VOP_REMOVE(ZTOV(sharedir), zc->zc_string, kcred,
4881 4888 NULL, 0);
4882 4889 break;
4883 4890
4884 4891 case ZFS_SMB_ACL_RENAME:
4885 4892 if ((error = get_nvlist(zc->zc_nvlist_src,
4886 4893 zc->zc_nvlist_src_size, zc->zc_iflags, &nvlist)) != 0) {
4887 4894 VN_RELE(vp);
4888 4895 ZFS_EXIT(zfsvfs);
4889 4896 return (error);
4890 4897 }
4891 4898 if (nvlist_lookup_string(nvlist, ZFS_SMB_ACL_SRC, &src) ||
4892 4899 nvlist_lookup_string(nvlist, ZFS_SMB_ACL_TARGET,
4893 4900 &target)) {
4894 4901 VN_RELE(vp);
4895 4902 VN_RELE(ZTOV(sharedir));
4896 4903 ZFS_EXIT(zfsvfs);
4897 4904 nvlist_free(nvlist);
4898 4905 return (error);
4899 4906 }
4900 4907 error = VOP_RENAME(ZTOV(sharedir), src, ZTOV(sharedir), target,
4901 4908 kcred, NULL, 0);
4902 4909 nvlist_free(nvlist);
4903 4910 break;
4904 4911
4905 4912 case ZFS_SMB_ACL_PURGE:
4906 4913 error = zfs_smb_acl_purge(sharedir);
4907 4914 break;
4908 4915
4909 4916 default:
4910 4917 error = SET_ERROR(EINVAL);
4911 4918 break;
4912 4919 }
4913 4920
4914 4921 VN_RELE(vp);
4915 4922 VN_RELE(ZTOV(sharedir));
4916 4923
4917 4924 ZFS_EXIT(zfsvfs);
4918 4925
4919 4926 return (error);
4920 4927 }
4921 4928
4922 4929 /*
4923 4930 * innvl: {
4924 4931 * "holds" -> { snapname -> holdname (string), ... }
4925 4932 * (optional) "cleanup_fd" -> fd (int32)
4926 4933 * }
4927 4934 *
4928 4935 * outnvl: {
4929 4936 * snapname -> error value (int32)
4930 4937 * ...
4931 4938 * }
4932 4939 */
4933 4940 /* ARGSUSED */
4934 4941 static int
4935 4942 zfs_ioc_hold(const char *pool, nvlist_t *args, nvlist_t *errlist)
4936 4943 {
4937 4944 nvlist_t *holds;
4938 4945 int cleanup_fd = -1;
4939 4946 int error;
4940 4947 minor_t minor = 0;
4941 4948
4942 4949 error = nvlist_lookup_nvlist(args, "holds", &holds);
4943 4950 if (error != 0)
4944 4951 return (SET_ERROR(EINVAL));
4945 4952
4946 4953 if (nvlist_lookup_int32(args, "cleanup_fd", &cleanup_fd) == 0) {
4947 4954 error = zfs_onexit_fd_hold(cleanup_fd, &minor);
4948 4955 if (error != 0)
4949 4956 return (error);
4950 4957 }
4951 4958
4952 4959 error = dsl_dataset_user_hold(holds, minor, errlist);
4953 4960 if (minor != 0)
4954 4961 zfs_onexit_fd_rele(cleanup_fd);
4955 4962 return (error);
4956 4963 }
4957 4964
4958 4965 /*
4959 4966 * innvl is not used.
4960 4967 *
4961 4968 * outnvl: {
4962 4969 * holdname -> time added (uint64 seconds since epoch)
4963 4970 * ...
4964 4971 * }
4965 4972 */
4966 4973 /* ARGSUSED */
4967 4974 static int
4968 4975 zfs_ioc_get_holds(const char *snapname, nvlist_t *args, nvlist_t *outnvl)
4969 4976 {
4970 4977 return (dsl_dataset_get_holds(snapname, outnvl));
4971 4978 }
4972 4979
4973 4980 /*
4974 4981 * innvl: {
4975 4982 * snapname -> { holdname, ... }
4976 4983 * ...
4977 4984 * }
4978 4985 *
4979 4986 * outnvl: {
4980 4987 * snapname -> error value (int32)
4981 4988 * ...
4982 4989 * }
4983 4990 */
4984 4991 /* ARGSUSED */
4985 4992 static int
4986 4993 zfs_ioc_release(const char *pool, nvlist_t *holds, nvlist_t *errlist)
4987 4994 {
4988 4995 return (dsl_dataset_user_release(holds, errlist));
4989 4996 }
4990 4997
4991 4998 /*
4992 4999 * inputs:
4993 5000 * zc_name name of new filesystem or snapshot
4994 5001 * zc_value full name of old snapshot
4995 5002 *
4996 5003 * outputs:
4997 5004 * zc_cookie space in bytes
4998 5005 * zc_objset_type compressed space in bytes
4999 5006 * zc_perm_action uncompressed space in bytes
5000 5007 */
5001 5008 static int
5002 5009 zfs_ioc_space_written(zfs_cmd_t *zc)
5003 5010 {
5004 5011 int error;
5005 5012 dsl_pool_t *dp;
5006 5013 dsl_dataset_t *new, *old;
5007 5014
5008 5015 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
5009 5016 if (error != 0)
5010 5017 return (error);
5011 5018 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &new);
5012 5019 if (error != 0) {
5013 5020 dsl_pool_rele(dp, FTAG);
5014 5021 return (error);
5015 5022 }
5016 5023 error = dsl_dataset_hold(dp, zc->zc_value, FTAG, &old);
5017 5024 if (error != 0) {
5018 5025 dsl_dataset_rele(new, FTAG);
5019 5026 dsl_pool_rele(dp, FTAG);
5020 5027 return (error);
5021 5028 }
5022 5029
5023 5030 error = dsl_dataset_space_written(old, new, &zc->zc_cookie,
5024 5031 &zc->zc_objset_type, &zc->zc_perm_action);
5025 5032 dsl_dataset_rele(old, FTAG);
5026 5033 dsl_dataset_rele(new, FTAG);
5027 5034 dsl_pool_rele(dp, FTAG);
5028 5035 return (error);
5029 5036 }
5030 5037
5031 5038 /*
5032 5039 * innvl: {
5033 5040 * "firstsnap" -> snapshot name
5034 5041 * }
5035 5042 *
5036 5043 * outnvl: {
5037 5044 * "used" -> space in bytes
5038 5045 * "compressed" -> compressed space in bytes
5039 5046 * "uncompressed" -> uncompressed space in bytes
5040 5047 * }
5041 5048 */
5042 5049 static int
5043 5050 zfs_ioc_space_snaps(const char *lastsnap, nvlist_t *innvl, nvlist_t *outnvl)
5044 5051 {
5045 5052 int error;
5046 5053 dsl_pool_t *dp;
5047 5054 dsl_dataset_t *new, *old;
5048 5055 char *firstsnap;
5049 5056 uint64_t used, comp, uncomp;
5050 5057
5051 5058 if (nvlist_lookup_string(innvl, "firstsnap", &firstsnap) != 0)
5052 5059 return (SET_ERROR(EINVAL));
5053 5060
5054 5061 error = dsl_pool_hold(lastsnap, FTAG, &dp);
5055 5062 if (error != 0)
5056 5063 return (error);
5057 5064
5058 5065 error = dsl_dataset_hold(dp, lastsnap, FTAG, &new);
5059 5066 if (error != 0) {
5060 5067 dsl_pool_rele(dp, FTAG);
5061 5068 return (error);
5062 5069 }
5063 5070 error = dsl_dataset_hold(dp, firstsnap, FTAG, &old);
5064 5071 if (error != 0) {
5065 5072 dsl_dataset_rele(new, FTAG);
5066 5073 dsl_pool_rele(dp, FTAG);
5067 5074 return (error);
5068 5075 }
5069 5076
5070 5077 error = dsl_dataset_space_wouldfree(old, new, &used, &comp, &uncomp);
5071 5078 dsl_dataset_rele(old, FTAG);
5072 5079 dsl_dataset_rele(new, FTAG);
5073 5080 dsl_pool_rele(dp, FTAG);
5074 5081 fnvlist_add_uint64(outnvl, "used", used);
5075 5082 fnvlist_add_uint64(outnvl, "compressed", comp);
5076 5083 fnvlist_add_uint64(outnvl, "uncompressed", uncomp);
5077 5084 return (error);
5078 5085 }
5079 5086
5080 5087 /*
5081 5088 * innvl: {
5082 5089 * "fd" -> file descriptor to write stream to (int32)
5083 5090 * (optional) "fromsnap" -> full snap name to send an incremental from
5084 5091 * }
5085 5092 *
5086 5093 * outnvl is unused
5087 5094 */
5088 5095 /* ARGSUSED */
5089 5096 static int
5090 5097 zfs_ioc_send_new(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl)
5091 5098 {
5092 5099 int error;
5093 5100 offset_t off;
5094 5101 char *fromname = NULL;
5095 5102 int fd;
5096 5103
5097 5104 error = nvlist_lookup_int32(innvl, "fd", &fd);
5098 5105 if (error != 0)
5099 5106 return (SET_ERROR(EINVAL));
5100 5107
5101 5108 (void) nvlist_lookup_string(innvl, "fromsnap", &fromname);
5102 5109
5103 5110 file_t *fp = getf(fd);
5104 5111 if (fp == NULL)
5105 5112 return (SET_ERROR(EBADF));
5106 5113
5107 5114 off = fp->f_offset;
5108 5115 error = dmu_send(snapname, fromname, fd, fp->f_vnode, &off);
5109 5116
5110 5117 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
5111 5118 fp->f_offset = off;
5112 5119 releasef(fd);
5113 5120 return (error);
5114 5121 }
5115 5122
5116 5123 /*
5117 5124 * Determine approximately how large a zfs send stream will be -- the number
5118 5125 * of bytes that will be written to the fd supplied to zfs_ioc_send_new().
5119 5126 *
5120 5127 * innvl: {
5121 5128 * (optional) "fromsnap" -> full snap name to send an incremental from
5122 5129 * }
5123 5130 *
5124 5131 * outnvl: {
5125 5132 * "space" -> bytes of space (uint64)
5126 5133 * }
5127 5134 */
5128 5135 static int
5129 5136 zfs_ioc_send_space(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl)
5130 5137 {
5131 5138 dsl_pool_t *dp;
5132 5139 dsl_dataset_t *fromsnap = NULL;
5133 5140 dsl_dataset_t *tosnap;
5134 5141 int error;
5135 5142 char *fromname;
5136 5143 uint64_t space;
5137 5144
5138 5145 error = dsl_pool_hold(snapname, FTAG, &dp);
5139 5146 if (error != 0)
5140 5147 return (error);
5141 5148
5142 5149 error = dsl_dataset_hold(dp, snapname, FTAG, &tosnap);
5143 5150 if (error != 0) {
5144 5151 dsl_pool_rele(dp, FTAG);
5145 5152 return (error);
5146 5153 }
5147 5154
5148 5155 error = nvlist_lookup_string(innvl, "fromsnap", &fromname);
5149 5156 if (error == 0) {
5150 5157 error = dsl_dataset_hold(dp, fromname, FTAG, &fromsnap);
5151 5158 if (error != 0) {
5152 5159 dsl_dataset_rele(tosnap, FTAG);
5153 5160 dsl_pool_rele(dp, FTAG);
5154 5161 return (error);
5155 5162 }
5156 5163 }
5157 5164
5158 5165 error = dmu_send_estimate(tosnap, fromsnap, &space);
5159 5166 fnvlist_add_uint64(outnvl, "space", space);
5160 5167
5161 5168 if (fromsnap != NULL)
5162 5169 dsl_dataset_rele(fromsnap, FTAG);
5163 5170 dsl_dataset_rele(tosnap, FTAG);
5164 5171 dsl_pool_rele(dp, FTAG);
5165 5172 return (error);
5166 5173 }
5167 5174
5168 5175
5169 5176 static zfs_ioc_vec_t zfs_ioc_vec[ZFS_IOC_LAST - ZFS_IOC_FIRST];
5170 5177
5171 5178 static void
5172 5179 zfs_ioctl_register_legacy(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5173 5180 zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck,
5174 5181 boolean_t log_history, zfs_ioc_poolcheck_t pool_check)
5175 5182 {
5176 5183 zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST];
5177 5184
5178 5185 ASSERT3U(ioc, >=, ZFS_IOC_FIRST);
5179 5186 ASSERT3U(ioc, <, ZFS_IOC_LAST);
5180 5187 ASSERT3P(vec->zvec_legacy_func, ==, NULL);
5181 5188 ASSERT3P(vec->zvec_func, ==, NULL);
5182 5189
5183 5190 vec->zvec_legacy_func = func;
5184 5191 vec->zvec_secpolicy = secpolicy;
5185 5192 vec->zvec_namecheck = namecheck;
5186 5193 vec->zvec_allow_log = log_history;
5187 5194 vec->zvec_pool_check = pool_check;
5188 5195 }
5189 5196
5190 5197 /*
5191 5198 * See the block comment at the beginning of this file for details on
5192 5199 * each argument to this function.
5193 5200 */
5194 5201 static void
5195 5202 zfs_ioctl_register(const char *name, zfs_ioc_t ioc, zfs_ioc_func_t *func,
5196 5203 zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck,
5197 5204 zfs_ioc_poolcheck_t pool_check, boolean_t smush_outnvlist,
5198 5205 boolean_t allow_log)
5199 5206 {
5200 5207 zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST];
5201 5208
5202 5209 ASSERT3U(ioc, >=, ZFS_IOC_FIRST);
5203 5210 ASSERT3U(ioc, <, ZFS_IOC_LAST);
5204 5211 ASSERT3P(vec->zvec_legacy_func, ==, NULL);
5205 5212 ASSERT3P(vec->zvec_func, ==, NULL);
5206 5213
5207 5214 /* if we are logging, the name must be valid */
5208 5215 ASSERT(!allow_log || namecheck != NO_NAME);
5209 5216
5210 5217 vec->zvec_name = name;
5211 5218 vec->zvec_func = func;
5212 5219 vec->zvec_secpolicy = secpolicy;
5213 5220 vec->zvec_namecheck = namecheck;
5214 5221 vec->zvec_pool_check = pool_check;
5215 5222 vec->zvec_smush_outnvlist = smush_outnvlist;
5216 5223 vec->zvec_allow_log = allow_log;
5217 5224 }
5218 5225
5219 5226 static void
5220 5227 zfs_ioctl_register_pool(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5221 5228 zfs_secpolicy_func_t *secpolicy, boolean_t log_history,
5222 5229 zfs_ioc_poolcheck_t pool_check)
5223 5230 {
5224 5231 zfs_ioctl_register_legacy(ioc, func, secpolicy,
5225 5232 POOL_NAME, log_history, pool_check);
5226 5233 }
5227 5234
5228 5235 static void
5229 5236 zfs_ioctl_register_dataset_nolog(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5230 5237 zfs_secpolicy_func_t *secpolicy, zfs_ioc_poolcheck_t pool_check)
5231 5238 {
5232 5239 zfs_ioctl_register_legacy(ioc, func, secpolicy,
5233 5240 DATASET_NAME, B_FALSE, pool_check);
5234 5241 }
5235 5242
5236 5243 static void
5237 5244 zfs_ioctl_register_pool_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func)
5238 5245 {
5239 5246 zfs_ioctl_register_legacy(ioc, func, zfs_secpolicy_config,
5240 5247 POOL_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5241 5248 }
5242 5249
5243 5250 static void
5244 5251 zfs_ioctl_register_pool_meta(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5245 5252 zfs_secpolicy_func_t *secpolicy)
5246 5253 {
5247 5254 zfs_ioctl_register_legacy(ioc, func, secpolicy,
5248 5255 NO_NAME, B_FALSE, POOL_CHECK_NONE);
5249 5256 }
5250 5257
5251 5258 static void
5252 5259 zfs_ioctl_register_dataset_read_secpolicy(zfs_ioc_t ioc,
5253 5260 zfs_ioc_legacy_func_t *func, zfs_secpolicy_func_t *secpolicy)
5254 5261 {
5255 5262 zfs_ioctl_register_legacy(ioc, func, secpolicy,
5256 5263 DATASET_NAME, B_FALSE, POOL_CHECK_SUSPENDED);
5257 5264 }
5258 5265
5259 5266 static void
5260 5267 zfs_ioctl_register_dataset_read(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func)
5261 5268 {
5262 5269 zfs_ioctl_register_dataset_read_secpolicy(ioc, func,
5263 5270 zfs_secpolicy_read);
5264 5271 }
5265 5272
5266 5273 static void
5267 5274 zfs_ioctl_register_dataset_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5268 5275 zfs_secpolicy_func_t *secpolicy)
5269 5276 {
5270 5277 zfs_ioctl_register_legacy(ioc, func, secpolicy,
5271 5278 DATASET_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5272 5279 }
5273 5280
5274 5281 static void
5275 5282 zfs_ioctl_init(void)
5276 5283 {
5277 5284 zfs_ioctl_register("snapshot", ZFS_IOC_SNAPSHOT,
5278 5285 zfs_ioc_snapshot, zfs_secpolicy_snapshot, POOL_NAME,
5279 5286 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5280 5287
5281 5288 zfs_ioctl_register("log_history", ZFS_IOC_LOG_HISTORY,
5282 5289 zfs_ioc_log_history, zfs_secpolicy_log_history, NO_NAME,
5283 5290 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_FALSE);
5284 5291
5285 5292 zfs_ioctl_register("space_snaps", ZFS_IOC_SPACE_SNAPS,
5286 5293 zfs_ioc_space_snaps, zfs_secpolicy_read, DATASET_NAME,
5287 5294 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5288 5295
5289 5296 zfs_ioctl_register("send", ZFS_IOC_SEND_NEW,
5290 5297 zfs_ioc_send_new, zfs_secpolicy_send_new, DATASET_NAME,
5291 5298 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5292 5299
5293 5300 zfs_ioctl_register("send_space", ZFS_IOC_SEND_SPACE,
5294 5301 zfs_ioc_send_space, zfs_secpolicy_read, DATASET_NAME,
5295 5302 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5296 5303
5297 5304 zfs_ioctl_register("create", ZFS_IOC_CREATE,
5298 5305 zfs_ioc_create, zfs_secpolicy_create_clone, DATASET_NAME,
5299 5306 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5300 5307
5301 5308 zfs_ioctl_register("clone", ZFS_IOC_CLONE,
5302 5309 zfs_ioc_clone, zfs_secpolicy_create_clone, DATASET_NAME,
5303 5310 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5304 5311
5305 5312 zfs_ioctl_register("destroy_snaps", ZFS_IOC_DESTROY_SNAPS,
5306 5313 zfs_ioc_destroy_snaps, zfs_secpolicy_destroy_snaps, POOL_NAME,
5307 5314 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5308 5315
5309 5316 zfs_ioctl_register("hold", ZFS_IOC_HOLD,
5310 5317 zfs_ioc_hold, zfs_secpolicy_hold, POOL_NAME,
5311 5318 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5312 5319 zfs_ioctl_register("release", ZFS_IOC_RELEASE,
5313 5320 zfs_ioc_release, zfs_secpolicy_release, POOL_NAME,
5314 5321 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5315 5322
5316 5323 zfs_ioctl_register("get_holds", ZFS_IOC_GET_HOLDS,
5317 5324 zfs_ioc_get_holds, zfs_secpolicy_read, DATASET_NAME,
5318 5325 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5319 5326
5320 5327 /* IOCTLS that use the legacy function signature */
5321 5328
5322 5329 zfs_ioctl_register_legacy(ZFS_IOC_POOL_FREEZE, zfs_ioc_pool_freeze,
5323 5330 zfs_secpolicy_config, NO_NAME, B_FALSE, POOL_CHECK_READONLY);
5324 5331
5325 5332 zfs_ioctl_register_pool(ZFS_IOC_POOL_CREATE, zfs_ioc_pool_create,
5326 5333 zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE);
5327 5334 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SCAN,
5328 5335 zfs_ioc_pool_scan);
5329 5336 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_UPGRADE,
5330 5337 zfs_ioc_pool_upgrade);
5331 5338 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ADD,
5332 5339 zfs_ioc_vdev_add);
5333 5340 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_REMOVE,
5334 5341 zfs_ioc_vdev_remove);
5335 5342 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SET_STATE,
5336 5343 zfs_ioc_vdev_set_state);
5337 5344 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ATTACH,
5338 5345 zfs_ioc_vdev_attach);
5339 5346 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_DETACH,
5340 5347 zfs_ioc_vdev_detach);
5341 5348 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETPATH,
5342 5349 zfs_ioc_vdev_setpath);
5343 5350 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETFRU,
5344 5351 zfs_ioc_vdev_setfru);
5345 5352 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SET_PROPS,
5346 5353 zfs_ioc_pool_set_props);
5347 5354 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SPLIT,
5348 5355 zfs_ioc_vdev_split);
5349 5356 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_REGUID,
5350 5357 zfs_ioc_pool_reguid);
5351 5358
5352 5359 zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_CONFIGS,
5353 5360 zfs_ioc_pool_configs, zfs_secpolicy_none);
5354 5361 zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_TRYIMPORT,
5355 5362 zfs_ioc_pool_tryimport, zfs_secpolicy_config);
5356 5363 zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_FAULT,
5357 5364 zfs_ioc_inject_fault, zfs_secpolicy_inject);
5358 5365 zfs_ioctl_register_pool_meta(ZFS_IOC_CLEAR_FAULT,
5359 5366 zfs_ioc_clear_fault, zfs_secpolicy_inject);
5360 5367 zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_LIST_NEXT,
5361 5368 zfs_ioc_inject_list_next, zfs_secpolicy_inject);
5362 5369
5363 5370 /*
5364 5371 * pool destroy, and export don't log the history as part of
5365 5372 * zfsdev_ioctl, but rather zfs_ioc_pool_export
5366 5373 * does the logging of those commands.
5367 5374 */
5368 5375 zfs_ioctl_register_pool(ZFS_IOC_POOL_DESTROY, zfs_ioc_pool_destroy,
5369 5376 zfs_secpolicy_config, B_FALSE, POOL_CHECK_NONE);
5370 5377 zfs_ioctl_register_pool(ZFS_IOC_POOL_EXPORT, zfs_ioc_pool_export,
5371 5378 zfs_secpolicy_config, B_FALSE, POOL_CHECK_NONE);
5372 5379
5373 5380 zfs_ioctl_register_pool(ZFS_IOC_POOL_STATS, zfs_ioc_pool_stats,
5374 5381 zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE);
5375 5382 zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_PROPS, zfs_ioc_pool_get_props,
5376 5383 zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE);
5377 5384
5378 5385 zfs_ioctl_register_pool(ZFS_IOC_ERROR_LOG, zfs_ioc_error_log,
5379 5386 zfs_secpolicy_inject, B_FALSE, POOL_CHECK_SUSPENDED);
5380 5387 zfs_ioctl_register_pool(ZFS_IOC_DSOBJ_TO_DSNAME,
5381 5388 zfs_ioc_dsobj_to_dsname,
5382 5389 zfs_secpolicy_diff, B_FALSE, POOL_CHECK_SUSPENDED);
5383 5390 zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_HISTORY,
5384 5391 zfs_ioc_pool_get_history,
5385 5392 zfs_secpolicy_config, B_FALSE, POOL_CHECK_SUSPENDED);
5386 5393
5387 5394 zfs_ioctl_register_pool(ZFS_IOC_POOL_IMPORT, zfs_ioc_pool_import,
5388 5395 zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE);
5389 5396
5390 5397 zfs_ioctl_register_pool(ZFS_IOC_CLEAR, zfs_ioc_clear,
5391 5398 zfs_secpolicy_config, B_TRUE, POOL_CHECK_SUSPENDED);
5392 5399 zfs_ioctl_register_pool(ZFS_IOC_POOL_REOPEN, zfs_ioc_pool_reopen,
5393 5400 zfs_secpolicy_config, B_TRUE, POOL_CHECK_SUSPENDED);
5394 5401
5395 5402 zfs_ioctl_register_dataset_read(ZFS_IOC_SPACE_WRITTEN,
5396 5403 zfs_ioc_space_written);
5397 5404 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_RECVD_PROPS,
5398 5405 zfs_ioc_objset_recvd_props);
5399 5406 zfs_ioctl_register_dataset_read(ZFS_IOC_NEXT_OBJ,
5400 5407 zfs_ioc_next_obj);
5401 5408 zfs_ioctl_register_dataset_read(ZFS_IOC_GET_FSACL,
5402 5409 zfs_ioc_get_fsacl);
5403 5410 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_STATS,
5404 5411 zfs_ioc_objset_stats);
5405 5412 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_ZPLPROPS,
5406 5413 zfs_ioc_objset_zplprops);
5407 5414 zfs_ioctl_register_dataset_read(ZFS_IOC_DATASET_LIST_NEXT,
5408 5415 zfs_ioc_dataset_list_next);
5409 5416 zfs_ioctl_register_dataset_read(ZFS_IOC_SNAPSHOT_LIST_NEXT,
5410 5417 zfs_ioc_snapshot_list_next);
5411 5418 zfs_ioctl_register_dataset_read(ZFS_IOC_SEND_PROGRESS,
5412 5419 zfs_ioc_send_progress);
5413 5420
5414 5421 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_DIFF,
5415 5422 zfs_ioc_diff, zfs_secpolicy_diff);
5416 5423 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_STATS,
5417 5424 zfs_ioc_obj_to_stats, zfs_secpolicy_diff);
5418 5425 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_PATH,
5419 5426 zfs_ioc_obj_to_path, zfs_secpolicy_diff);
5420 5427 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_ONE,
5421 5428 zfs_ioc_userspace_one, zfs_secpolicy_userspace_one);
5422 5429 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_MANY,
5423 5430 zfs_ioc_userspace_many, zfs_secpolicy_userspace_many);
5424 5431 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_SEND,
5425 5432 zfs_ioc_send, zfs_secpolicy_send);
5426 5433
5427 5434 zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_PROP, zfs_ioc_set_prop,
5428 5435 zfs_secpolicy_none);
5429 5436 zfs_ioctl_register_dataset_modify(ZFS_IOC_DESTROY, zfs_ioc_destroy,
5430 5437 zfs_secpolicy_destroy);
5431 5438 zfs_ioctl_register_dataset_modify(ZFS_IOC_ROLLBACK, zfs_ioc_rollback,
5432 5439 zfs_secpolicy_rollback);
5433 5440 zfs_ioctl_register_dataset_modify(ZFS_IOC_RENAME, zfs_ioc_rename,
5434 5441 zfs_secpolicy_rename);
5435 5442 zfs_ioctl_register_dataset_modify(ZFS_IOC_RECV, zfs_ioc_recv,
5436 5443 zfs_secpolicy_recv);
5437 5444 zfs_ioctl_register_dataset_modify(ZFS_IOC_PROMOTE, zfs_ioc_promote,
5438 5445 zfs_secpolicy_promote);
5439 5446 zfs_ioctl_register_dataset_modify(ZFS_IOC_INHERIT_PROP,
5440 5447 zfs_ioc_inherit_prop, zfs_secpolicy_inherit_prop);
5441 5448 zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_FSACL, zfs_ioc_set_fsacl,
5442 5449 zfs_secpolicy_set_fsacl);
5443 5450
5444 5451 zfs_ioctl_register_dataset_nolog(ZFS_IOC_SHARE, zfs_ioc_share,
5445 5452 zfs_secpolicy_share, POOL_CHECK_NONE);
5446 5453 zfs_ioctl_register_dataset_nolog(ZFS_IOC_SMB_ACL, zfs_ioc_smb_acl,
5447 5454 zfs_secpolicy_smb_acl, POOL_CHECK_NONE);
5448 5455 zfs_ioctl_register_dataset_nolog(ZFS_IOC_USERSPACE_UPGRADE,
5449 5456 zfs_ioc_userspace_upgrade, zfs_secpolicy_userspace_upgrade,
5450 5457 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5451 5458 zfs_ioctl_register_dataset_nolog(ZFS_IOC_TMP_SNAPSHOT,
5452 5459 zfs_ioc_tmp_snapshot, zfs_secpolicy_tmp_snapshot,
5453 5460 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5454 5461 }
5455 5462
5456 5463 int
5457 5464 pool_status_check(const char *name, zfs_ioc_namecheck_t type,
5458 5465 zfs_ioc_poolcheck_t check)
5459 5466 {
5460 5467 spa_t *spa;
5461 5468 int error;
5462 5469
5463 5470 ASSERT(type == POOL_NAME || type == DATASET_NAME);
5464 5471
5465 5472 if (check & POOL_CHECK_NONE)
5466 5473 return (0);
5467 5474
5468 5475 error = spa_open(name, &spa, FTAG);
5469 5476 if (error == 0) {
5470 5477 if ((check & POOL_CHECK_SUSPENDED) && spa_suspended(spa))
5471 5478 error = SET_ERROR(EAGAIN);
5472 5479 else if ((check & POOL_CHECK_READONLY) && !spa_writeable(spa))
5473 5480 error = SET_ERROR(EROFS);
5474 5481 spa_close(spa, FTAG);
5475 5482 }
5476 5483 return (error);
5477 5484 }
5478 5485
5479 5486 /*
5480 5487 * Find a free minor number.
5481 5488 */
5482 5489 minor_t
5483 5490 zfsdev_minor_alloc(void)
5484 5491 {
5485 5492 static minor_t last_minor;
5486 5493 minor_t m;
5487 5494
5488 5495 ASSERT(MUTEX_HELD(&zfsdev_state_lock));
5489 5496
5490 5497 for (m = last_minor + 1; m != last_minor; m++) {
5491 5498 if (m > ZFSDEV_MAX_MINOR)
5492 5499 m = 1;
5493 5500 if (ddi_get_soft_state(zfsdev_state, m) == NULL) {
5494 5501 last_minor = m;
5495 5502 return (m);
5496 5503 }
5497 5504 }
5498 5505
5499 5506 return (0);
5500 5507 }
5501 5508
5502 5509 static int
5503 5510 zfs_ctldev_init(dev_t *devp)
5504 5511 {
5505 5512 minor_t minor;
5506 5513 zfs_soft_state_t *zs;
5507 5514
5508 5515 ASSERT(MUTEX_HELD(&zfsdev_state_lock));
5509 5516 ASSERT(getminor(*devp) == 0);
5510 5517
5511 5518 minor = zfsdev_minor_alloc();
5512 5519 if (minor == 0)
5513 5520 return (SET_ERROR(ENXIO));
5514 5521
5515 5522 if (ddi_soft_state_zalloc(zfsdev_state, minor) != DDI_SUCCESS)
5516 5523 return (SET_ERROR(EAGAIN));
5517 5524
5518 5525 *devp = makedevice(getemajor(*devp), minor);
5519 5526
5520 5527 zs = ddi_get_soft_state(zfsdev_state, minor);
5521 5528 zs->zss_type = ZSST_CTLDEV;
5522 5529 zfs_onexit_init((zfs_onexit_t **)&zs->zss_data);
5523 5530
5524 5531 return (0);
5525 5532 }
5526 5533
5527 5534 static void
5528 5535 zfs_ctldev_destroy(zfs_onexit_t *zo, minor_t minor)
5529 5536 {
5530 5537 ASSERT(MUTEX_HELD(&zfsdev_state_lock));
5531 5538
5532 5539 zfs_onexit_destroy(zo);
5533 5540 ddi_soft_state_free(zfsdev_state, minor);
5534 5541 }
5535 5542
5536 5543 void *
5537 5544 zfsdev_get_soft_state(minor_t minor, enum zfs_soft_state_type which)
5538 5545 {
5539 5546 zfs_soft_state_t *zp;
5540 5547
5541 5548 zp = ddi_get_soft_state(zfsdev_state, minor);
5542 5549 if (zp == NULL || zp->zss_type != which)
5543 5550 return (NULL);
5544 5551
5545 5552 return (zp->zss_data);
5546 5553 }
5547 5554
5548 5555 static int
5549 5556 zfsdev_open(dev_t *devp, int flag, int otyp, cred_t *cr)
5550 5557 {
5551 5558 int error = 0;
5552 5559
5553 5560 if (getminor(*devp) != 0)
5554 5561 return (zvol_open(devp, flag, otyp, cr));
5555 5562
5556 5563 /* This is the control device. Allocate a new minor if requested. */
5557 5564 if (flag & FEXCL) {
5558 5565 mutex_enter(&zfsdev_state_lock);
5559 5566 error = zfs_ctldev_init(devp);
5560 5567 mutex_exit(&zfsdev_state_lock);
5561 5568 }
5562 5569
5563 5570 return (error);
5564 5571 }
5565 5572
5566 5573 static int
5567 5574 zfsdev_close(dev_t dev, int flag, int otyp, cred_t *cr)
5568 5575 {
5569 5576 zfs_onexit_t *zo;
5570 5577 minor_t minor = getminor(dev);
5571 5578
5572 5579 if (minor == 0)
5573 5580 return (0);
5574 5581
5575 5582 mutex_enter(&zfsdev_state_lock);
5576 5583 zo = zfsdev_get_soft_state(minor, ZSST_CTLDEV);
5577 5584 if (zo == NULL) {
5578 5585 mutex_exit(&zfsdev_state_lock);
5579 5586 return (zvol_close(dev, flag, otyp, cr));
5580 5587 }
5581 5588 zfs_ctldev_destroy(zo, minor);
5582 5589 mutex_exit(&zfsdev_state_lock);
5583 5590
5584 5591 return (0);
5585 5592 }
5586 5593
5587 5594 static int
5588 5595 zfsdev_ioctl(dev_t dev, int cmd, intptr_t arg, int flag, cred_t *cr, int *rvalp)
5589 5596 {
5590 5597 zfs_cmd_t *zc;
5591 5598 uint_t vecnum;
5592 5599 int error, rc, len;
5593 5600 minor_t minor = getminor(dev);
5594 5601 const zfs_ioc_vec_t *vec;
5595 5602 char *saved_poolname = NULL;
5596 5603 nvlist_t *innvl = NULL;
5597 5604
5598 5605 if (minor != 0 &&
5599 5606 zfsdev_get_soft_state(minor, ZSST_CTLDEV) == NULL)
5600 5607 return (zvol_ioctl(dev, cmd, arg, flag, cr, rvalp));
5601 5608
5602 5609 vecnum = cmd - ZFS_IOC_FIRST;
5603 5610 ASSERT3U(getmajor(dev), ==, ddi_driver_major(zfs_dip));
5604 5611
5605 5612 if (vecnum >= sizeof (zfs_ioc_vec) / sizeof (zfs_ioc_vec[0]))
5606 5613 return (SET_ERROR(EINVAL));
5607 5614 vec = &zfs_ioc_vec[vecnum];
5608 5615
5609 5616 zc = kmem_zalloc(sizeof (zfs_cmd_t), KM_SLEEP);
5610 5617
5611 5618 error = ddi_copyin((void *)arg, zc, sizeof (zfs_cmd_t), flag);
5612 5619 if (error != 0) {
5613 5620 error = SET_ERROR(EFAULT);
5614 5621 goto out;
5615 5622 }
5616 5623
5617 5624 zc->zc_iflags = flag & FKIOCTL;
5618 5625 if (zc->zc_nvlist_src_size != 0) {
5619 5626 error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
5620 5627 zc->zc_iflags, &innvl);
5621 5628 if (error != 0)
5622 5629 goto out;
5623 5630 }
5624 5631
5625 5632 /*
5626 5633 * Ensure that all pool/dataset names are valid before we pass down to
5627 5634 * the lower layers.
5628 5635 */
5629 5636 zc->zc_name[sizeof (zc->zc_name) - 1] = '\0';
5630 5637 switch (vec->zvec_namecheck) {
5631 5638 case POOL_NAME:
5632 5639 if (pool_namecheck(zc->zc_name, NULL, NULL) != 0)
5633 5640 error = SET_ERROR(EINVAL);
5634 5641 else
5635 5642 error = pool_status_check(zc->zc_name,
5636 5643 vec->zvec_namecheck, vec->zvec_pool_check);
5637 5644 break;
5638 5645
5639 5646 case DATASET_NAME:
5640 5647 if (dataset_namecheck(zc->zc_name, NULL, NULL) != 0)
5641 5648 error = SET_ERROR(EINVAL);
5642 5649 else
5643 5650 error = pool_status_check(zc->zc_name,
5644 5651 vec->zvec_namecheck, vec->zvec_pool_check);
5645 5652 break;
5646 5653
5647 5654 case NO_NAME:
5648 5655 break;
5649 5656 }
5650 5657
5651 5658
5652 5659 if (error == 0 && !(flag & FKIOCTL))
5653 5660 error = vec->zvec_secpolicy(zc, innvl, cr);
5654 5661
5655 5662 if (error != 0)
5656 5663 goto out;
5657 5664
5658 5665 /* legacy ioctls can modify zc_name */
5659 5666 len = strcspn(zc->zc_name, "/@") + 1;
5660 5667 saved_poolname = kmem_alloc(len, KM_SLEEP);
5661 5668 (void) strlcpy(saved_poolname, zc->zc_name, len);
5662 5669
5663 5670 if (vec->zvec_func != NULL) {
5664 5671 nvlist_t *outnvl;
5665 5672 int puterror = 0;
5666 5673 spa_t *spa;
5667 5674 nvlist_t *lognv = NULL;
5668 5675
5669 5676 ASSERT(vec->zvec_legacy_func == NULL);
5670 5677
5671 5678 /*
5672 5679 * Add the innvl to the lognv before calling the func,
5673 5680 * in case the func changes the innvl.
5674 5681 */
5675 5682 if (vec->zvec_allow_log) {
5676 5683 lognv = fnvlist_alloc();
5677 5684 fnvlist_add_string(lognv, ZPOOL_HIST_IOCTL,
5678 5685 vec->zvec_name);
5679 5686 if (!nvlist_empty(innvl)) {
5680 5687 fnvlist_add_nvlist(lognv, ZPOOL_HIST_INPUT_NVL,
5681 5688 innvl);
5682 5689 }
5683 5690 }
5684 5691
5685 5692 outnvl = fnvlist_alloc();
5686 5693 error = vec->zvec_func(zc->zc_name, innvl, outnvl);
5687 5694
5688 5695 if (error == 0 && vec->zvec_allow_log &&
5689 5696 spa_open(zc->zc_name, &spa, FTAG) == 0) {
5690 5697 if (!nvlist_empty(outnvl)) {
5691 5698 fnvlist_add_nvlist(lognv, ZPOOL_HIST_OUTPUT_NVL,
5692 5699 outnvl);
5693 5700 }
5694 5701 (void) spa_history_log_nvl(spa, lognv);
5695 5702 spa_close(spa, FTAG);
5696 5703 }
5697 5704 fnvlist_free(lognv);
5698 5705
5699 5706 if (!nvlist_empty(outnvl) || zc->zc_nvlist_dst_size != 0) {
5700 5707 int smusherror = 0;
5701 5708 if (vec->zvec_smush_outnvlist) {
5702 5709 smusherror = nvlist_smush(outnvl,
5703 5710 zc->zc_nvlist_dst_size);
5704 5711 }
5705 5712 if (smusherror == 0)
5706 5713 puterror = put_nvlist(zc, outnvl);
5707 5714 }
5708 5715
5709 5716 if (puterror != 0)
5710 5717 error = puterror;
5711 5718
5712 5719 nvlist_free(outnvl);
5713 5720 } else {
5714 5721 error = vec->zvec_legacy_func(zc);
5715 5722 }
5716 5723
5717 5724 out:
5718 5725 nvlist_free(innvl);
5719 5726 rc = ddi_copyout(zc, (void *)arg, sizeof (zfs_cmd_t), flag);
5720 5727 if (error == 0 && rc != 0)
5721 5728 error = SET_ERROR(EFAULT);
5722 5729 if (error == 0 && vec->zvec_allow_log) {
5723 5730 char *s = tsd_get(zfs_allow_log_key);
5724 5731 if (s != NULL)
5725 5732 strfree(s);
5726 5733 (void) tsd_set(zfs_allow_log_key, saved_poolname);
5727 5734 } else {
5728 5735 if (saved_poolname != NULL)
5729 5736 strfree(saved_poolname);
5730 5737 }
5731 5738
5732 5739 kmem_free(zc, sizeof (zfs_cmd_t));
5733 5740 return (error);
5734 5741 }
5735 5742
5736 5743 static int
5737 5744 zfs_attach(dev_info_t *dip, ddi_attach_cmd_t cmd)
5738 5745 {
5739 5746 if (cmd != DDI_ATTACH)
5740 5747 return (DDI_FAILURE);
5741 5748
5742 5749 if (ddi_create_minor_node(dip, "zfs", S_IFCHR, 0,
5743 5750 DDI_PSEUDO, 0) == DDI_FAILURE)
5744 5751 return (DDI_FAILURE);
5745 5752
5746 5753 zfs_dip = dip;
5747 5754
5748 5755 ddi_report_dev(dip);
5749 5756
5750 5757 return (DDI_SUCCESS);
5751 5758 }
5752 5759
5753 5760 static int
5754 5761 zfs_detach(dev_info_t *dip, ddi_detach_cmd_t cmd)
5755 5762 {
5756 5763 if (spa_busy() || zfs_busy() || zvol_busy())
5757 5764 return (DDI_FAILURE);
5758 5765
5759 5766 if (cmd != DDI_DETACH)
5760 5767 return (DDI_FAILURE);
5761 5768
5762 5769 zfs_dip = NULL;
5763 5770
5764 5771 ddi_prop_remove_all(dip);
5765 5772 ddi_remove_minor_node(dip, NULL);
5766 5773
5767 5774 return (DDI_SUCCESS);
5768 5775 }
5769 5776
5770 5777 /*ARGSUSED*/
5771 5778 static int
5772 5779 zfs_info(dev_info_t *dip, ddi_info_cmd_t infocmd, void *arg, void **result)
5773 5780 {
5774 5781 switch (infocmd) {
5775 5782 case DDI_INFO_DEVT2DEVINFO:
5776 5783 *result = zfs_dip;
5777 5784 return (DDI_SUCCESS);
5778 5785
5779 5786 case DDI_INFO_DEVT2INSTANCE:
5780 5787 *result = (void *)0;
5781 5788 return (DDI_SUCCESS);
5782 5789 }
5783 5790
5784 5791 return (DDI_FAILURE);
5785 5792 }
5786 5793
5787 5794 /*
5788 5795 * OK, so this is a little weird.
5789 5796 *
5790 5797 * /dev/zfs is the control node, i.e. minor 0.
5791 5798 * /dev/zvol/[r]dsk/pool/dataset are the zvols, minor > 0.
5792 5799 *
5793 5800 * /dev/zfs has basically nothing to do except serve up ioctls,
5794 5801 * so most of the standard driver entry points are in zvol.c.
5795 5802 */
5796 5803 static struct cb_ops zfs_cb_ops = {
5797 5804 zfsdev_open, /* open */
5798 5805 zfsdev_close, /* close */
5799 5806 zvol_strategy, /* strategy */
5800 5807 nodev, /* print */
5801 5808 zvol_dump, /* dump */
5802 5809 zvol_read, /* read */
5803 5810 zvol_write, /* write */
5804 5811 zfsdev_ioctl, /* ioctl */
5805 5812 nodev, /* devmap */
5806 5813 nodev, /* mmap */
5807 5814 nodev, /* segmap */
5808 5815 nochpoll, /* poll */
5809 5816 ddi_prop_op, /* prop_op */
5810 5817 NULL, /* streamtab */
5811 5818 D_NEW | D_MP | D_64BIT, /* Driver compatibility flag */
5812 5819 CB_REV, /* version */
5813 5820 nodev, /* async read */
5814 5821 nodev, /* async write */
5815 5822 };
5816 5823
5817 5824 static struct dev_ops zfs_dev_ops = {
5818 5825 DEVO_REV, /* version */
5819 5826 0, /* refcnt */
5820 5827 zfs_info, /* info */
5821 5828 nulldev, /* identify */
5822 5829 nulldev, /* probe */
5823 5830 zfs_attach, /* attach */
5824 5831 zfs_detach, /* detach */
5825 5832 nodev, /* reset */
5826 5833 &zfs_cb_ops, /* driver operations */
5827 5834 NULL, /* no bus operations */
5828 5835 NULL, /* power */
5829 5836 ddi_quiesce_not_needed, /* quiesce */
5830 5837 };
5831 5838
5832 5839 static struct modldrv zfs_modldrv = {
5833 5840 &mod_driverops,
5834 5841 "ZFS storage pool",
5835 5842 &zfs_dev_ops
5836 5843 };
5837 5844
5838 5845 static struct modlinkage modlinkage = {
5839 5846 MODREV_1,
5840 5847 (void *)&zfs_modlfs,
5841 5848 (void *)&zfs_modldrv,
5842 5849 NULL
5843 5850 };
5844 5851
5845 5852 static void
5846 5853 zfs_allow_log_destroy(void *arg)
5847 5854 {
5848 5855 char *poolname = arg;
5849 5856 strfree(poolname);
5850 5857 }
5851 5858
5852 5859 int
5853 5860 _init(void)
5854 5861 {
5855 5862 int error;
5856 5863
5857 5864 spa_init(FREAD | FWRITE);
5858 5865 zfs_init();
5859 5866 zvol_init();
5860 5867 zfs_ioctl_init();
5861 5868
5862 5869 if ((error = mod_install(&modlinkage)) != 0) {
5863 5870 zvol_fini();
5864 5871 zfs_fini();
5865 5872 spa_fini();
5866 5873 return (error);
5867 5874 }
5868 5875
5869 5876 tsd_create(&zfs_fsyncer_key, NULL);
5870 5877 tsd_create(&rrw_tsd_key, rrw_tsd_destroy);
5871 5878 tsd_create(&zfs_allow_log_key, zfs_allow_log_destroy);
5872 5879
5873 5880 error = ldi_ident_from_mod(&modlinkage, &zfs_li);
5874 5881 ASSERT(error == 0);
5875 5882 mutex_init(&zfs_share_lock, NULL, MUTEX_DEFAULT, NULL);
5876 5883
5877 5884 return (0);
5878 5885 }
5879 5886
5880 5887 int
5881 5888 _fini(void)
5882 5889 {
5883 5890 int error;
5884 5891
5885 5892 if (spa_busy() || zfs_busy() || zvol_busy() || zio_injection_enabled)
5886 5893 return (SET_ERROR(EBUSY));
5887 5894
5888 5895 if ((error = mod_remove(&modlinkage)) != 0)
5889 5896 return (error);
5890 5897
5891 5898 zvol_fini();
5892 5899 zfs_fini();
5893 5900 spa_fini();
5894 5901 if (zfs_nfsshare_inited)
5895 5902 (void) ddi_modclose(nfs_mod);
5896 5903 if (zfs_smbshare_inited)
5897 5904 (void) ddi_modclose(smbsrv_mod);
5898 5905 if (zfs_nfsshare_inited || zfs_smbshare_inited)
5899 5906 (void) ddi_modclose(sharefs_mod);
5900 5907
5901 5908 tsd_destroy(&zfs_fsyncer_key);
5902 5909 ldi_ident_release(zfs_li);
5903 5910 zfs_li = NULL;
5904 5911 mutex_destroy(&zfs_share_lock);
5905 5912
5906 5913 return (error);
5907 5914 }
5908 5915
5909 5916 int
5910 5917 _info(struct modinfo *modinfop)
5911 5918 {
5912 5919 return (mod_info(&modlinkage, modinfop));
5913 5920 }
|
↓ open down ↓ |
2337 lines elided |
↑ open up ↑ |
XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX