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