Print this page
4185 New hash algorithm support
| Split |
Close |
| Expand all |
| Collapse all |
--- old/usr/src/uts/common/fs/zfs/zfs_ioctl.c
+++ new/usr/src/uts/common/fs/zfs/zfs_ioctl.c
1 1 /*
2 2 * CDDL HEADER START
3 3 *
4 4 * The contents of this file are subject to the terms of the
5 5 * Common Development and Distribution License (the "License").
6 6 * You may not use this file except in compliance with the License.
7 7 *
8 8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9 9 * or http://www.opensolaris.org/os/licensing.
10 10 * See the License for the specific language governing permissions
11 11 * and limitations under the License.
12 12 *
13 13 * When distributing Covered Code, include this CDDL HEADER in each
14 14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15 15 * If applicable, add the following below this CDDL HEADER, with the
16 16 * fields enclosed by brackets "[]" replaced with your own identifying
17 17 * information: Portions Copyright [yyyy] [name of copyright owner]
18 18 *
19 19 * CDDL HEADER END
20 20 */
21 21
22 22 /*
23 23 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
24 24 * Portions Copyright 2011 Martin Matuska
25 25 * Copyright 2011 Nexenta Systems, Inc. All rights reserved.
26 26 * Copyright (c) 2012, Joyent, Inc. All rights reserved.
27 27 * Copyright (c) 2013 by Delphix. All rights reserved.
28 28 * Copyright (c) 2013 by Saso Kiselkov. All rights reserved.
29 29 * Copyright (c) 2013 Steven Hartland. All rights reserved.
30 30 */
31 31
32 32 /*
33 33 * ZFS ioctls.
34 34 *
35 35 * This file handles the ioctls to /dev/zfs, used for configuring ZFS storage
36 36 * pools and filesystems, e.g. with /sbin/zfs and /sbin/zpool.
37 37 *
38 38 * There are two ways that we handle ioctls: the legacy way where almost
39 39 * all of the logic is in the ioctl callback, and the new way where most
40 40 * of the marshalling is handled in the common entry point, zfsdev_ioctl().
41 41 *
42 42 * Non-legacy ioctls should be registered by calling
43 43 * zfs_ioctl_register() from zfs_ioctl_init(). The ioctl is invoked
44 44 * from userland by lzc_ioctl().
45 45 *
46 46 * The registration arguments are as follows:
47 47 *
48 48 * const char *name
49 49 * The name of the ioctl. This is used for history logging. If the
50 50 * ioctl returns successfully (the callback returns 0), and allow_log
51 51 * is true, then a history log entry will be recorded with the input &
52 52 * output nvlists. The log entry can be printed with "zpool history -i".
53 53 *
54 54 * zfs_ioc_t ioc
55 55 * The ioctl request number, which userland will pass to ioctl(2).
56 56 * The ioctl numbers can change from release to release, because
57 57 * the caller (libzfs) must be matched to the kernel.
58 58 *
59 59 * zfs_secpolicy_func_t *secpolicy
60 60 * This function will be called before the zfs_ioc_func_t, to
61 61 * determine if this operation is permitted. It should return EPERM
62 62 * on failure, and 0 on success. Checks include determining if the
63 63 * dataset is visible in this zone, and if the user has either all
64 64 * zfs privileges in the zone (SYS_MOUNT), or has been granted permission
65 65 * to do this operation on this dataset with "zfs allow".
66 66 *
67 67 * zfs_ioc_namecheck_t namecheck
68 68 * This specifies what to expect in the zfs_cmd_t:zc_name -- a pool
69 69 * name, a dataset name, or nothing. If the name is not well-formed,
70 70 * the ioctl will fail and the callback will not be called.
71 71 * Therefore, the callback can assume that the name is well-formed
72 72 * (e.g. is null-terminated, doesn't have more than one '@' character,
73 73 * doesn't have invalid characters).
74 74 *
75 75 * zfs_ioc_poolcheck_t pool_check
76 76 * This specifies requirements on the pool state. If the pool does
77 77 * not meet them (is suspended or is readonly), the ioctl will fail
78 78 * and the callback will not be called. If any checks are specified
79 79 * (i.e. it is not POOL_CHECK_NONE), namecheck must not be NO_NAME.
80 80 * Multiple checks can be or-ed together (e.g. POOL_CHECK_SUSPENDED |
81 81 * POOL_CHECK_READONLY).
82 82 *
83 83 * boolean_t smush_outnvlist
84 84 * If smush_outnvlist is true, then the output is presumed to be a
85 85 * list of errors, and it will be "smushed" down to fit into the
86 86 * caller's buffer, by removing some entries and replacing them with a
87 87 * single "N_MORE_ERRORS" entry indicating how many were removed. See
88 88 * nvlist_smush() for details. If smush_outnvlist is false, and the
89 89 * outnvlist does not fit into the userland-provided buffer, then the
90 90 * ioctl will fail with ENOMEM.
91 91 *
92 92 * zfs_ioc_func_t *func
93 93 * The callback function that will perform the operation.
94 94 *
95 95 * The callback should return 0 on success, or an error number on
96 96 * failure. If the function fails, the userland ioctl will return -1,
97 97 * and errno will be set to the callback's return value. The callback
98 98 * will be called with the following arguments:
99 99 *
100 100 * const char *name
101 101 * The name of the pool or dataset to operate on, from
102 102 * zfs_cmd_t:zc_name. The 'namecheck' argument specifies the
103 103 * expected type (pool, dataset, or none).
104 104 *
105 105 * nvlist_t *innvl
106 106 * The input nvlist, deserialized from zfs_cmd_t:zc_nvlist_src. Or
107 107 * NULL if no input nvlist was provided. Changes to this nvlist are
108 108 * ignored. If the input nvlist could not be deserialized, the
109 109 * ioctl will fail and the callback will not be called.
110 110 *
111 111 * nvlist_t *outnvl
112 112 * The output nvlist, initially empty. The callback can fill it in,
113 113 * and it will be returned to userland by serializing it into
114 114 * zfs_cmd_t:zc_nvlist_dst. If it is non-empty, and serialization
115 115 * fails (e.g. because the caller didn't supply a large enough
116 116 * buffer), then the overall ioctl will fail. See the
117 117 * 'smush_nvlist' argument above for additional behaviors.
118 118 *
119 119 * There are two typical uses of the output nvlist:
120 120 * - To return state, e.g. property values. In this case,
121 121 * smush_outnvlist should be false. If the buffer was not large
122 122 * enough, the caller will reallocate a larger buffer and try
123 123 * the ioctl again.
124 124 *
125 125 * - To return multiple errors from an ioctl which makes on-disk
126 126 * changes. In this case, smush_outnvlist should be true.
127 127 * Ioctls which make on-disk modifications should generally not
128 128 * use the outnvl if they succeed, because the caller can not
129 129 * distinguish between the operation failing, and
130 130 * deserialization failing.
131 131 */
132 132
133 133 #include <sys/types.h>
134 134 #include <sys/param.h>
135 135 #include <sys/errno.h>
136 136 #include <sys/uio.h>
137 137 #include <sys/buf.h>
138 138 #include <sys/modctl.h>
139 139 #include <sys/open.h>
140 140 #include <sys/file.h>
141 141 #include <sys/kmem.h>
142 142 #include <sys/conf.h>
143 143 #include <sys/cmn_err.h>
144 144 #include <sys/stat.h>
145 145 #include <sys/zfs_ioctl.h>
146 146 #include <sys/zfs_vfsops.h>
147 147 #include <sys/zfs_znode.h>
148 148 #include <sys/zap.h>
149 149 #include <sys/spa.h>
150 150 #include <sys/spa_impl.h>
151 151 #include <sys/vdev.h>
152 152 #include <sys/priv_impl.h>
153 153 #include <sys/dmu.h>
154 154 #include <sys/dsl_dir.h>
155 155 #include <sys/dsl_dataset.h>
156 156 #include <sys/dsl_prop.h>
157 157 #include <sys/dsl_deleg.h>
158 158 #include <sys/dmu_objset.h>
159 159 #include <sys/dmu_impl.h>
160 160 #include <sys/dmu_tx.h>
161 161 #include <sys/ddi.h>
162 162 #include <sys/sunddi.h>
163 163 #include <sys/sunldi.h>
164 164 #include <sys/policy.h>
165 165 #include <sys/zone.h>
166 166 #include <sys/nvpair.h>
167 167 #include <sys/pathname.h>
168 168 #include <sys/mount.h>
169 169 #include <sys/sdt.h>
170 170 #include <sys/fs/zfs.h>
171 171 #include <sys/zfs_ctldir.h>
|
↓ open down ↓ |
171 lines elided |
↑ open up ↑ |
172 172 #include <sys/zfs_dir.h>
173 173 #include <sys/zfs_onexit.h>
174 174 #include <sys/zvol.h>
175 175 #include <sys/dsl_scan.h>
176 176 #include <sharefs/share.h>
177 177 #include <sys/dmu_objset.h>
178 178 #include <sys/dmu_send.h>
179 179 #include <sys/dsl_destroy.h>
180 180 #include <sys/dsl_userhold.h>
181 181 #include <sys/zfeature.h>
182 +#include <sys/zio_checksum.h>
182 183
183 184 #include "zfs_namecheck.h"
184 185 #include "zfs_prop.h"
185 186 #include "zfs_deleg.h"
186 187 #include "zfs_comutil.h"
187 188
188 189 extern struct modlfs zfs_modlfs;
189 190
190 191 extern void zfs_init(void);
191 192 extern void zfs_fini(void);
192 193
193 194 ldi_ident_t zfs_li = NULL;
194 195 dev_info_t *zfs_dip;
195 196
196 197 uint_t zfs_fsyncer_key;
197 198 extern uint_t rrw_tsd_key;
198 199 static uint_t zfs_allow_log_key;
199 200
200 201 typedef int zfs_ioc_legacy_func_t(zfs_cmd_t *);
201 202 typedef int zfs_ioc_func_t(const char *, nvlist_t *, nvlist_t *);
202 203 typedef int zfs_secpolicy_func_t(zfs_cmd_t *, nvlist_t *, cred_t *);
203 204
204 205 typedef enum {
205 206 NO_NAME,
206 207 POOL_NAME,
207 208 DATASET_NAME
208 209 } zfs_ioc_namecheck_t;
209 210
210 211 typedef enum {
211 212 POOL_CHECK_NONE = 1 << 0,
212 213 POOL_CHECK_SUSPENDED = 1 << 1,
213 214 POOL_CHECK_READONLY = 1 << 2,
214 215 } zfs_ioc_poolcheck_t;
215 216
216 217 typedef struct zfs_ioc_vec {
217 218 zfs_ioc_legacy_func_t *zvec_legacy_func;
218 219 zfs_ioc_func_t *zvec_func;
219 220 zfs_secpolicy_func_t *zvec_secpolicy;
220 221 zfs_ioc_namecheck_t zvec_namecheck;
221 222 boolean_t zvec_allow_log;
222 223 zfs_ioc_poolcheck_t zvec_pool_check;
223 224 boolean_t zvec_smush_outnvlist;
224 225 const char *zvec_name;
225 226 } zfs_ioc_vec_t;
226 227
227 228 /* This array is indexed by zfs_userquota_prop_t */
228 229 static const char *userquota_perms[] = {
229 230 ZFS_DELEG_PERM_USERUSED,
230 231 ZFS_DELEG_PERM_USERQUOTA,
231 232 ZFS_DELEG_PERM_GROUPUSED,
232 233 ZFS_DELEG_PERM_GROUPQUOTA,
233 234 };
234 235
235 236 static int zfs_ioc_userspace_upgrade(zfs_cmd_t *zc);
236 237 static int zfs_check_settable(const char *name, nvpair_t *property,
237 238 cred_t *cr);
238 239 static int zfs_check_clearable(char *dataset, nvlist_t *props,
239 240 nvlist_t **errors);
240 241 static int zfs_fill_zplprops_root(uint64_t, nvlist_t *, nvlist_t *,
241 242 boolean_t *);
242 243 int zfs_set_prop_nvlist(const char *, zprop_source_t, nvlist_t *, nvlist_t *);
243 244 static int get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp);
244 245
245 246 static int zfs_prop_activate_feature(spa_t *spa, zfeature_info_t *feature);
246 247
247 248 /* _NOTE(PRINTFLIKE(4)) - this is printf-like, but lint is too whiney */
248 249 void
249 250 __dprintf(const char *file, const char *func, int line, const char *fmt, ...)
250 251 {
251 252 const char *newfile;
252 253 char buf[512];
253 254 va_list adx;
254 255
255 256 /*
256 257 * Get rid of annoying "../common/" prefix to filename.
257 258 */
258 259 newfile = strrchr(file, '/');
259 260 if (newfile != NULL) {
260 261 newfile = newfile + 1; /* Get rid of leading / */
261 262 } else {
262 263 newfile = file;
263 264 }
264 265
265 266 va_start(adx, fmt);
266 267 (void) vsnprintf(buf, sizeof (buf), fmt, adx);
267 268 va_end(adx);
268 269
269 270 /*
270 271 * To get this data, use the zfs-dprintf probe as so:
271 272 * dtrace -q -n 'zfs-dprintf \
272 273 * /stringof(arg0) == "dbuf.c"/ \
273 274 * {printf("%s: %s", stringof(arg1), stringof(arg3))}'
274 275 * arg0 = file name
275 276 * arg1 = function name
276 277 * arg2 = line number
277 278 * arg3 = message
278 279 */
279 280 DTRACE_PROBE4(zfs__dprintf,
280 281 char *, newfile, char *, func, int, line, char *, buf);
281 282 }
282 283
283 284 static void
284 285 history_str_free(char *buf)
285 286 {
286 287 kmem_free(buf, HIS_MAX_RECORD_LEN);
287 288 }
288 289
289 290 static char *
290 291 history_str_get(zfs_cmd_t *zc)
291 292 {
292 293 char *buf;
293 294
294 295 if (zc->zc_history == NULL)
295 296 return (NULL);
296 297
297 298 buf = kmem_alloc(HIS_MAX_RECORD_LEN, KM_SLEEP);
298 299 if (copyinstr((void *)(uintptr_t)zc->zc_history,
299 300 buf, HIS_MAX_RECORD_LEN, NULL) != 0) {
300 301 history_str_free(buf);
301 302 return (NULL);
302 303 }
303 304
304 305 buf[HIS_MAX_RECORD_LEN -1] = '\0';
305 306
306 307 return (buf);
307 308 }
308 309
309 310 /*
310 311 * Check to see if the named dataset is currently defined as bootable
311 312 */
312 313 static boolean_t
313 314 zfs_is_bootfs(const char *name)
314 315 {
315 316 objset_t *os;
316 317
317 318 if (dmu_objset_hold(name, FTAG, &os) == 0) {
318 319 boolean_t ret;
319 320 ret = (dmu_objset_id(os) == spa_bootfs(dmu_objset_spa(os)));
320 321 dmu_objset_rele(os, FTAG);
321 322 return (ret);
322 323 }
323 324 return (B_FALSE);
324 325 }
325 326
326 327 /*
327 328 * Return non-zero if the spa version is less than requested version.
328 329 */
329 330 static int
330 331 zfs_earlier_version(const char *name, int version)
331 332 {
332 333 spa_t *spa;
333 334
334 335 if (spa_open(name, &spa, FTAG) == 0) {
335 336 if (spa_version(spa) < version) {
336 337 spa_close(spa, FTAG);
337 338 return (1);
338 339 }
339 340 spa_close(spa, FTAG);
340 341 }
341 342 return (0);
342 343 }
343 344
344 345 /*
345 346 * Return TRUE if the ZPL version is less than requested version.
346 347 */
347 348 static boolean_t
348 349 zpl_earlier_version(const char *name, int version)
349 350 {
350 351 objset_t *os;
351 352 boolean_t rc = B_TRUE;
352 353
353 354 if (dmu_objset_hold(name, FTAG, &os) == 0) {
354 355 uint64_t zplversion;
355 356
356 357 if (dmu_objset_type(os) != DMU_OST_ZFS) {
357 358 dmu_objset_rele(os, FTAG);
358 359 return (B_TRUE);
359 360 }
360 361 /* XXX reading from non-owned objset */
361 362 if (zfs_get_zplprop(os, ZFS_PROP_VERSION, &zplversion) == 0)
362 363 rc = zplversion < version;
363 364 dmu_objset_rele(os, FTAG);
364 365 }
365 366 return (rc);
366 367 }
367 368
368 369 static void
369 370 zfs_log_history(zfs_cmd_t *zc)
370 371 {
371 372 spa_t *spa;
372 373 char *buf;
373 374
374 375 if ((buf = history_str_get(zc)) == NULL)
375 376 return;
376 377
377 378 if (spa_open(zc->zc_name, &spa, FTAG) == 0) {
378 379 if (spa_version(spa) >= SPA_VERSION_ZPOOL_HISTORY)
379 380 (void) spa_history_log(spa, buf);
380 381 spa_close(spa, FTAG);
381 382 }
382 383 history_str_free(buf);
383 384 }
384 385
385 386 /*
386 387 * Policy for top-level read operations (list pools). Requires no privileges,
387 388 * and can be used in the local zone, as there is no associated dataset.
388 389 */
389 390 /* ARGSUSED */
390 391 static int
391 392 zfs_secpolicy_none(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
392 393 {
393 394 return (0);
394 395 }
395 396
396 397 /*
397 398 * Policy for dataset read operations (list children, get statistics). Requires
398 399 * no privileges, but must be visible in the local zone.
399 400 */
400 401 /* ARGSUSED */
401 402 static int
402 403 zfs_secpolicy_read(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
403 404 {
404 405 if (INGLOBALZONE(curproc) ||
405 406 zone_dataset_visible(zc->zc_name, NULL))
406 407 return (0);
407 408
408 409 return (SET_ERROR(ENOENT));
409 410 }
410 411
411 412 static int
412 413 zfs_dozonecheck_impl(const char *dataset, uint64_t zoned, cred_t *cr)
413 414 {
414 415 int writable = 1;
415 416
416 417 /*
417 418 * The dataset must be visible by this zone -- check this first
418 419 * so they don't see EPERM on something they shouldn't know about.
419 420 */
420 421 if (!INGLOBALZONE(curproc) &&
421 422 !zone_dataset_visible(dataset, &writable))
422 423 return (SET_ERROR(ENOENT));
423 424
424 425 if (INGLOBALZONE(curproc)) {
425 426 /*
426 427 * If the fs is zoned, only root can access it from the
427 428 * global zone.
428 429 */
429 430 if (secpolicy_zfs(cr) && zoned)
430 431 return (SET_ERROR(EPERM));
431 432 } else {
432 433 /*
433 434 * If we are in a local zone, the 'zoned' property must be set.
434 435 */
435 436 if (!zoned)
436 437 return (SET_ERROR(EPERM));
437 438
438 439 /* must be writable by this zone */
439 440 if (!writable)
440 441 return (SET_ERROR(EPERM));
441 442 }
442 443 return (0);
443 444 }
444 445
445 446 static int
446 447 zfs_dozonecheck(const char *dataset, cred_t *cr)
447 448 {
448 449 uint64_t zoned;
449 450
450 451 if (dsl_prop_get_integer(dataset, "zoned", &zoned, NULL))
451 452 return (SET_ERROR(ENOENT));
452 453
453 454 return (zfs_dozonecheck_impl(dataset, zoned, cr));
454 455 }
455 456
456 457 static int
457 458 zfs_dozonecheck_ds(const char *dataset, dsl_dataset_t *ds, cred_t *cr)
458 459 {
459 460 uint64_t zoned;
460 461
461 462 if (dsl_prop_get_int_ds(ds, "zoned", &zoned))
462 463 return (SET_ERROR(ENOENT));
463 464
464 465 return (zfs_dozonecheck_impl(dataset, zoned, cr));
465 466 }
466 467
467 468 static int
468 469 zfs_secpolicy_write_perms_ds(const char *name, dsl_dataset_t *ds,
469 470 const char *perm, cred_t *cr)
470 471 {
471 472 int error;
472 473
473 474 error = zfs_dozonecheck_ds(name, ds, cr);
474 475 if (error == 0) {
475 476 error = secpolicy_zfs(cr);
476 477 if (error != 0)
477 478 error = dsl_deleg_access_impl(ds, perm, cr);
478 479 }
479 480 return (error);
480 481 }
481 482
482 483 static int
483 484 zfs_secpolicy_write_perms(const char *name, const char *perm, cred_t *cr)
484 485 {
485 486 int error;
486 487 dsl_dataset_t *ds;
487 488 dsl_pool_t *dp;
488 489
489 490 error = dsl_pool_hold(name, FTAG, &dp);
490 491 if (error != 0)
491 492 return (error);
492 493
493 494 error = dsl_dataset_hold(dp, name, FTAG, &ds);
494 495 if (error != 0) {
495 496 dsl_pool_rele(dp, FTAG);
496 497 return (error);
497 498 }
498 499
499 500 error = zfs_secpolicy_write_perms_ds(name, ds, perm, cr);
500 501
501 502 dsl_dataset_rele(ds, FTAG);
502 503 dsl_pool_rele(dp, FTAG);
503 504 return (error);
504 505 }
505 506
506 507 /*
507 508 * Policy for setting the security label property.
508 509 *
509 510 * Returns 0 for success, non-zero for access and other errors.
510 511 */
511 512 static int
512 513 zfs_set_slabel_policy(const char *name, char *strval, cred_t *cr)
513 514 {
514 515 char ds_hexsl[MAXNAMELEN];
515 516 bslabel_t ds_sl, new_sl;
516 517 boolean_t new_default = FALSE;
517 518 uint64_t zoned;
518 519 int needed_priv = -1;
519 520 int error;
520 521
521 522 /* First get the existing dataset label. */
522 523 error = dsl_prop_get(name, zfs_prop_to_name(ZFS_PROP_MLSLABEL),
523 524 1, sizeof (ds_hexsl), &ds_hexsl, NULL);
524 525 if (error != 0)
525 526 return (SET_ERROR(EPERM));
526 527
527 528 if (strcasecmp(strval, ZFS_MLSLABEL_DEFAULT) == 0)
528 529 new_default = TRUE;
529 530
530 531 /* The label must be translatable */
531 532 if (!new_default && (hexstr_to_label(strval, &new_sl) != 0))
532 533 return (SET_ERROR(EINVAL));
533 534
534 535 /*
535 536 * In a non-global zone, disallow attempts to set a label that
536 537 * doesn't match that of the zone; otherwise no other checks
537 538 * are needed.
538 539 */
539 540 if (!INGLOBALZONE(curproc)) {
540 541 if (new_default || !blequal(&new_sl, CR_SL(CRED())))
541 542 return (SET_ERROR(EPERM));
542 543 return (0);
543 544 }
544 545
545 546 /*
546 547 * For global-zone datasets (i.e., those whose zoned property is
547 548 * "off", verify that the specified new label is valid for the
548 549 * global zone.
549 550 */
550 551 if (dsl_prop_get_integer(name,
551 552 zfs_prop_to_name(ZFS_PROP_ZONED), &zoned, NULL))
552 553 return (SET_ERROR(EPERM));
553 554 if (!zoned) {
554 555 if (zfs_check_global_label(name, strval) != 0)
555 556 return (SET_ERROR(EPERM));
556 557 }
557 558
558 559 /*
559 560 * If the existing dataset label is nondefault, check if the
560 561 * dataset is mounted (label cannot be changed while mounted).
561 562 * Get the zfsvfs; if there isn't one, then the dataset isn't
562 563 * mounted (or isn't a dataset, doesn't exist, ...).
563 564 */
564 565 if (strcasecmp(ds_hexsl, ZFS_MLSLABEL_DEFAULT) != 0) {
565 566 objset_t *os;
566 567 static char *setsl_tag = "setsl_tag";
567 568
568 569 /*
569 570 * Try to own the dataset; abort if there is any error,
570 571 * (e.g., already mounted, in use, or other error).
571 572 */
572 573 error = dmu_objset_own(name, DMU_OST_ZFS, B_TRUE,
573 574 setsl_tag, &os);
574 575 if (error != 0)
575 576 return (SET_ERROR(EPERM));
576 577
577 578 dmu_objset_disown(os, setsl_tag);
578 579
579 580 if (new_default) {
580 581 needed_priv = PRIV_FILE_DOWNGRADE_SL;
581 582 goto out_check;
582 583 }
583 584
584 585 if (hexstr_to_label(strval, &new_sl) != 0)
585 586 return (SET_ERROR(EPERM));
586 587
587 588 if (blstrictdom(&ds_sl, &new_sl))
588 589 needed_priv = PRIV_FILE_DOWNGRADE_SL;
589 590 else if (blstrictdom(&new_sl, &ds_sl))
590 591 needed_priv = PRIV_FILE_UPGRADE_SL;
591 592 } else {
592 593 /* dataset currently has a default label */
593 594 if (!new_default)
594 595 needed_priv = PRIV_FILE_UPGRADE_SL;
595 596 }
596 597
597 598 out_check:
598 599 if (needed_priv != -1)
599 600 return (PRIV_POLICY(cr, needed_priv, B_FALSE, EPERM, NULL));
600 601 return (0);
601 602 }
602 603
603 604 static int
604 605 zfs_secpolicy_setprop(const char *dsname, zfs_prop_t prop, nvpair_t *propval,
605 606 cred_t *cr)
606 607 {
607 608 char *strval;
608 609
609 610 /*
610 611 * Check permissions for special properties.
611 612 */
612 613 switch (prop) {
613 614 case ZFS_PROP_ZONED:
614 615 /*
615 616 * Disallow setting of 'zoned' from within a local zone.
616 617 */
617 618 if (!INGLOBALZONE(curproc))
618 619 return (SET_ERROR(EPERM));
619 620 break;
620 621
621 622 case ZFS_PROP_QUOTA:
622 623 if (!INGLOBALZONE(curproc)) {
623 624 uint64_t zoned;
624 625 char setpoint[MAXNAMELEN];
625 626 /*
626 627 * Unprivileged users are allowed to modify the
627 628 * quota on things *under* (ie. contained by)
628 629 * the thing they own.
629 630 */
630 631 if (dsl_prop_get_integer(dsname, "zoned", &zoned,
631 632 setpoint))
632 633 return (SET_ERROR(EPERM));
633 634 if (!zoned || strlen(dsname) <= strlen(setpoint))
634 635 return (SET_ERROR(EPERM));
635 636 }
636 637 break;
637 638
638 639 case ZFS_PROP_MLSLABEL:
639 640 if (!is_system_labeled())
640 641 return (SET_ERROR(EPERM));
641 642
642 643 if (nvpair_value_string(propval, &strval) == 0) {
643 644 int err;
644 645
645 646 err = zfs_set_slabel_policy(dsname, strval, CRED());
646 647 if (err != 0)
647 648 return (err);
648 649 }
649 650 break;
650 651 }
651 652
652 653 return (zfs_secpolicy_write_perms(dsname, zfs_prop_to_name(prop), cr));
653 654 }
654 655
655 656 /* ARGSUSED */
656 657 static int
657 658 zfs_secpolicy_set_fsacl(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
658 659 {
659 660 int error;
660 661
661 662 error = zfs_dozonecheck(zc->zc_name, cr);
662 663 if (error != 0)
663 664 return (error);
664 665
665 666 /*
666 667 * permission to set permissions will be evaluated later in
667 668 * dsl_deleg_can_allow()
668 669 */
669 670 return (0);
670 671 }
671 672
672 673 /* ARGSUSED */
673 674 static int
674 675 zfs_secpolicy_rollback(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
675 676 {
676 677 return (zfs_secpolicy_write_perms(zc->zc_name,
677 678 ZFS_DELEG_PERM_ROLLBACK, cr));
678 679 }
679 680
680 681 /* ARGSUSED */
681 682 static int
682 683 zfs_secpolicy_send(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
683 684 {
684 685 dsl_pool_t *dp;
685 686 dsl_dataset_t *ds;
686 687 char *cp;
687 688 int error;
688 689
689 690 /*
690 691 * Generate the current snapshot name from the given objsetid, then
691 692 * use that name for the secpolicy/zone checks.
692 693 */
693 694 cp = strchr(zc->zc_name, '@');
694 695 if (cp == NULL)
695 696 return (SET_ERROR(EINVAL));
696 697 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
697 698 if (error != 0)
698 699 return (error);
699 700
700 701 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &ds);
701 702 if (error != 0) {
702 703 dsl_pool_rele(dp, FTAG);
703 704 return (error);
704 705 }
705 706
706 707 dsl_dataset_name(ds, zc->zc_name);
707 708
708 709 error = zfs_secpolicy_write_perms_ds(zc->zc_name, ds,
709 710 ZFS_DELEG_PERM_SEND, cr);
710 711 dsl_dataset_rele(ds, FTAG);
711 712 dsl_pool_rele(dp, FTAG);
712 713
713 714 return (error);
714 715 }
715 716
716 717 /* ARGSUSED */
717 718 static int
718 719 zfs_secpolicy_send_new(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
719 720 {
720 721 return (zfs_secpolicy_write_perms(zc->zc_name,
721 722 ZFS_DELEG_PERM_SEND, cr));
722 723 }
723 724
724 725 /* ARGSUSED */
725 726 static int
726 727 zfs_secpolicy_deleg_share(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
727 728 {
728 729 vnode_t *vp;
729 730 int error;
730 731
731 732 if ((error = lookupname(zc->zc_value, UIO_SYSSPACE,
732 733 NO_FOLLOW, NULL, &vp)) != 0)
733 734 return (error);
734 735
735 736 /* Now make sure mntpnt and dataset are ZFS */
736 737
737 738 if (vp->v_vfsp->vfs_fstype != zfsfstype ||
738 739 (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource),
739 740 zc->zc_name) != 0)) {
740 741 VN_RELE(vp);
741 742 return (SET_ERROR(EPERM));
742 743 }
743 744
744 745 VN_RELE(vp);
745 746 return (dsl_deleg_access(zc->zc_name,
746 747 ZFS_DELEG_PERM_SHARE, cr));
747 748 }
748 749
749 750 int
750 751 zfs_secpolicy_share(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
751 752 {
752 753 if (!INGLOBALZONE(curproc))
753 754 return (SET_ERROR(EPERM));
754 755
755 756 if (secpolicy_nfs(cr) == 0) {
756 757 return (0);
757 758 } else {
758 759 return (zfs_secpolicy_deleg_share(zc, innvl, cr));
759 760 }
760 761 }
761 762
762 763 int
763 764 zfs_secpolicy_smb_acl(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
764 765 {
765 766 if (!INGLOBALZONE(curproc))
766 767 return (SET_ERROR(EPERM));
767 768
768 769 if (secpolicy_smb(cr) == 0) {
769 770 return (0);
770 771 } else {
771 772 return (zfs_secpolicy_deleg_share(zc, innvl, cr));
772 773 }
773 774 }
774 775
775 776 static int
776 777 zfs_get_parent(const char *datasetname, char *parent, int parentsize)
777 778 {
778 779 char *cp;
779 780
780 781 /*
781 782 * Remove the @bla or /bla from the end of the name to get the parent.
782 783 */
783 784 (void) strncpy(parent, datasetname, parentsize);
784 785 cp = strrchr(parent, '@');
785 786 if (cp != NULL) {
786 787 cp[0] = '\0';
787 788 } else {
788 789 cp = strrchr(parent, '/');
789 790 if (cp == NULL)
790 791 return (SET_ERROR(ENOENT));
791 792 cp[0] = '\0';
792 793 }
793 794
794 795 return (0);
795 796 }
796 797
797 798 int
798 799 zfs_secpolicy_destroy_perms(const char *name, cred_t *cr)
799 800 {
800 801 int error;
801 802
802 803 if ((error = zfs_secpolicy_write_perms(name,
803 804 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
804 805 return (error);
805 806
806 807 return (zfs_secpolicy_write_perms(name, ZFS_DELEG_PERM_DESTROY, cr));
807 808 }
808 809
809 810 /* ARGSUSED */
810 811 static int
811 812 zfs_secpolicy_destroy(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
812 813 {
813 814 return (zfs_secpolicy_destroy_perms(zc->zc_name, cr));
814 815 }
815 816
816 817 /*
817 818 * Destroying snapshots with delegated permissions requires
818 819 * descendant mount and destroy permissions.
819 820 */
820 821 /* ARGSUSED */
821 822 static int
822 823 zfs_secpolicy_destroy_snaps(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
823 824 {
824 825 nvlist_t *snaps;
825 826 nvpair_t *pair, *nextpair;
826 827 int error = 0;
827 828
828 829 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
829 830 return (SET_ERROR(EINVAL));
830 831 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
831 832 pair = nextpair) {
832 833 dsl_pool_t *dp;
833 834 dsl_dataset_t *ds;
834 835
835 836 error = dsl_pool_hold(nvpair_name(pair), FTAG, &dp);
836 837 if (error != 0)
837 838 break;
838 839 nextpair = nvlist_next_nvpair(snaps, pair);
839 840 error = dsl_dataset_hold(dp, nvpair_name(pair), FTAG, &ds);
840 841 if (error == 0)
841 842 dsl_dataset_rele(ds, FTAG);
842 843 dsl_pool_rele(dp, FTAG);
843 844
844 845 if (error == 0) {
845 846 error = zfs_secpolicy_destroy_perms(nvpair_name(pair),
846 847 cr);
847 848 } else if (error == ENOENT) {
848 849 /*
849 850 * Ignore any snapshots that don't exist (we consider
850 851 * them "already destroyed"). Remove the name from the
851 852 * nvl here in case the snapshot is created between
852 853 * now and when we try to destroy it (in which case
853 854 * we don't want to destroy it since we haven't
854 855 * checked for permission).
855 856 */
856 857 fnvlist_remove_nvpair(snaps, pair);
857 858 error = 0;
858 859 }
859 860 if (error != 0)
860 861 break;
861 862 }
862 863
863 864 return (error);
864 865 }
865 866
866 867 int
867 868 zfs_secpolicy_rename_perms(const char *from, const char *to, cred_t *cr)
868 869 {
869 870 char parentname[MAXNAMELEN];
870 871 int error;
871 872
872 873 if ((error = zfs_secpolicy_write_perms(from,
873 874 ZFS_DELEG_PERM_RENAME, cr)) != 0)
874 875 return (error);
875 876
876 877 if ((error = zfs_secpolicy_write_perms(from,
877 878 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
878 879 return (error);
879 880
880 881 if ((error = zfs_get_parent(to, parentname,
881 882 sizeof (parentname))) != 0)
882 883 return (error);
883 884
884 885 if ((error = zfs_secpolicy_write_perms(parentname,
885 886 ZFS_DELEG_PERM_CREATE, cr)) != 0)
886 887 return (error);
887 888
888 889 if ((error = zfs_secpolicy_write_perms(parentname,
889 890 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
890 891 return (error);
891 892
892 893 return (error);
893 894 }
894 895
895 896 /* ARGSUSED */
896 897 static int
897 898 zfs_secpolicy_rename(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
898 899 {
899 900 return (zfs_secpolicy_rename_perms(zc->zc_name, zc->zc_value, cr));
900 901 }
901 902
902 903 /* ARGSUSED */
903 904 static int
904 905 zfs_secpolicy_promote(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
905 906 {
906 907 dsl_pool_t *dp;
907 908 dsl_dataset_t *clone;
908 909 int error;
909 910
910 911 error = zfs_secpolicy_write_perms(zc->zc_name,
911 912 ZFS_DELEG_PERM_PROMOTE, cr);
912 913 if (error != 0)
913 914 return (error);
914 915
915 916 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
916 917 if (error != 0)
917 918 return (error);
918 919
919 920 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &clone);
920 921
921 922 if (error == 0) {
922 923 char parentname[MAXNAMELEN];
923 924 dsl_dataset_t *origin = NULL;
924 925 dsl_dir_t *dd;
925 926 dd = clone->ds_dir;
926 927
927 928 error = dsl_dataset_hold_obj(dd->dd_pool,
928 929 dd->dd_phys->dd_origin_obj, FTAG, &origin);
929 930 if (error != 0) {
930 931 dsl_dataset_rele(clone, FTAG);
931 932 dsl_pool_rele(dp, FTAG);
932 933 return (error);
933 934 }
934 935
935 936 error = zfs_secpolicy_write_perms_ds(zc->zc_name, clone,
936 937 ZFS_DELEG_PERM_MOUNT, cr);
937 938
938 939 dsl_dataset_name(origin, parentname);
939 940 if (error == 0) {
940 941 error = zfs_secpolicy_write_perms_ds(parentname, origin,
941 942 ZFS_DELEG_PERM_PROMOTE, cr);
942 943 }
943 944 dsl_dataset_rele(clone, FTAG);
944 945 dsl_dataset_rele(origin, FTAG);
945 946 }
946 947 dsl_pool_rele(dp, FTAG);
947 948 return (error);
948 949 }
949 950
950 951 /* ARGSUSED */
951 952 static int
952 953 zfs_secpolicy_recv(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
953 954 {
954 955 int error;
955 956
956 957 if ((error = zfs_secpolicy_write_perms(zc->zc_name,
957 958 ZFS_DELEG_PERM_RECEIVE, cr)) != 0)
958 959 return (error);
959 960
960 961 if ((error = zfs_secpolicy_write_perms(zc->zc_name,
961 962 ZFS_DELEG_PERM_MOUNT, cr)) != 0)
962 963 return (error);
963 964
964 965 return (zfs_secpolicy_write_perms(zc->zc_name,
965 966 ZFS_DELEG_PERM_CREATE, cr));
966 967 }
967 968
968 969 int
969 970 zfs_secpolicy_snapshot_perms(const char *name, cred_t *cr)
970 971 {
971 972 return (zfs_secpolicy_write_perms(name,
972 973 ZFS_DELEG_PERM_SNAPSHOT, cr));
973 974 }
974 975
975 976 /*
976 977 * Check for permission to create each snapshot in the nvlist.
977 978 */
978 979 /* ARGSUSED */
979 980 static int
980 981 zfs_secpolicy_snapshot(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
981 982 {
982 983 nvlist_t *snaps;
983 984 int error = 0;
984 985 nvpair_t *pair;
985 986
986 987 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
987 988 return (SET_ERROR(EINVAL));
988 989 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
989 990 pair = nvlist_next_nvpair(snaps, pair)) {
990 991 char *name = nvpair_name(pair);
991 992 char *atp = strchr(name, '@');
992 993
993 994 if (atp == NULL) {
994 995 error = SET_ERROR(EINVAL);
995 996 break;
996 997 }
997 998 *atp = '\0';
998 999 error = zfs_secpolicy_snapshot_perms(name, cr);
999 1000 *atp = '@';
1000 1001 if (error != 0)
1001 1002 break;
1002 1003 }
1003 1004 return (error);
1004 1005 }
1005 1006
1006 1007 /* ARGSUSED */
1007 1008 static int
1008 1009 zfs_secpolicy_log_history(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1009 1010 {
1010 1011 /*
1011 1012 * Even root must have a proper TSD so that we know what pool
1012 1013 * to log to.
1013 1014 */
1014 1015 if (tsd_get(zfs_allow_log_key) == NULL)
1015 1016 return (SET_ERROR(EPERM));
1016 1017 return (0);
1017 1018 }
1018 1019
1019 1020 static int
1020 1021 zfs_secpolicy_create_clone(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1021 1022 {
1022 1023 char parentname[MAXNAMELEN];
1023 1024 int error;
1024 1025 char *origin;
1025 1026
1026 1027 if ((error = zfs_get_parent(zc->zc_name, parentname,
1027 1028 sizeof (parentname))) != 0)
1028 1029 return (error);
1029 1030
1030 1031 if (nvlist_lookup_string(innvl, "origin", &origin) == 0 &&
1031 1032 (error = zfs_secpolicy_write_perms(origin,
1032 1033 ZFS_DELEG_PERM_CLONE, cr)) != 0)
1033 1034 return (error);
1034 1035
1035 1036 if ((error = zfs_secpolicy_write_perms(parentname,
1036 1037 ZFS_DELEG_PERM_CREATE, cr)) != 0)
1037 1038 return (error);
1038 1039
1039 1040 return (zfs_secpolicy_write_perms(parentname,
1040 1041 ZFS_DELEG_PERM_MOUNT, cr));
1041 1042 }
1042 1043
1043 1044 /*
1044 1045 * Policy for pool operations - create/destroy pools, add vdevs, etc. Requires
1045 1046 * SYS_CONFIG privilege, which is not available in a local zone.
1046 1047 */
1047 1048 /* ARGSUSED */
1048 1049 static int
1049 1050 zfs_secpolicy_config(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1050 1051 {
1051 1052 if (secpolicy_sys_config(cr, B_FALSE) != 0)
1052 1053 return (SET_ERROR(EPERM));
1053 1054
1054 1055 return (0);
1055 1056 }
1056 1057
1057 1058 /*
1058 1059 * Policy for object to name lookups.
1059 1060 */
1060 1061 /* ARGSUSED */
1061 1062 static int
1062 1063 zfs_secpolicy_diff(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1063 1064 {
1064 1065 int error;
1065 1066
1066 1067 if ((error = secpolicy_sys_config(cr, B_FALSE)) == 0)
1067 1068 return (0);
1068 1069
1069 1070 error = zfs_secpolicy_write_perms(zc->zc_name, ZFS_DELEG_PERM_DIFF, cr);
1070 1071 return (error);
1071 1072 }
1072 1073
1073 1074 /*
1074 1075 * Policy for fault injection. Requires all privileges.
1075 1076 */
1076 1077 /* ARGSUSED */
1077 1078 static int
1078 1079 zfs_secpolicy_inject(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1079 1080 {
1080 1081 return (secpolicy_zinject(cr));
1081 1082 }
1082 1083
1083 1084 /* ARGSUSED */
1084 1085 static int
1085 1086 zfs_secpolicy_inherit_prop(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1086 1087 {
1087 1088 zfs_prop_t prop = zfs_name_to_prop(zc->zc_value);
1088 1089
1089 1090 if (prop == ZPROP_INVAL) {
1090 1091 if (!zfs_prop_user(zc->zc_value))
1091 1092 return (SET_ERROR(EINVAL));
1092 1093 return (zfs_secpolicy_write_perms(zc->zc_name,
1093 1094 ZFS_DELEG_PERM_USERPROP, cr));
1094 1095 } else {
1095 1096 return (zfs_secpolicy_setprop(zc->zc_name, prop,
1096 1097 NULL, cr));
1097 1098 }
1098 1099 }
1099 1100
1100 1101 static int
1101 1102 zfs_secpolicy_userspace_one(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1102 1103 {
1103 1104 int err = zfs_secpolicy_read(zc, innvl, cr);
1104 1105 if (err)
1105 1106 return (err);
1106 1107
1107 1108 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
1108 1109 return (SET_ERROR(EINVAL));
1109 1110
1110 1111 if (zc->zc_value[0] == 0) {
1111 1112 /*
1112 1113 * They are asking about a posix uid/gid. If it's
1113 1114 * themself, allow it.
1114 1115 */
1115 1116 if (zc->zc_objset_type == ZFS_PROP_USERUSED ||
1116 1117 zc->zc_objset_type == ZFS_PROP_USERQUOTA) {
1117 1118 if (zc->zc_guid == crgetuid(cr))
1118 1119 return (0);
1119 1120 } else {
1120 1121 if (groupmember(zc->zc_guid, cr))
1121 1122 return (0);
1122 1123 }
1123 1124 }
1124 1125
1125 1126 return (zfs_secpolicy_write_perms(zc->zc_name,
1126 1127 userquota_perms[zc->zc_objset_type], cr));
1127 1128 }
1128 1129
1129 1130 static int
1130 1131 zfs_secpolicy_userspace_many(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1131 1132 {
1132 1133 int err = zfs_secpolicy_read(zc, innvl, cr);
1133 1134 if (err)
1134 1135 return (err);
1135 1136
1136 1137 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
1137 1138 return (SET_ERROR(EINVAL));
1138 1139
1139 1140 return (zfs_secpolicy_write_perms(zc->zc_name,
1140 1141 userquota_perms[zc->zc_objset_type], cr));
1141 1142 }
1142 1143
1143 1144 /* ARGSUSED */
1144 1145 static int
1145 1146 zfs_secpolicy_userspace_upgrade(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1146 1147 {
1147 1148 return (zfs_secpolicy_setprop(zc->zc_name, ZFS_PROP_VERSION,
1148 1149 NULL, cr));
1149 1150 }
1150 1151
1151 1152 /* ARGSUSED */
1152 1153 static int
1153 1154 zfs_secpolicy_hold(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1154 1155 {
1155 1156 nvpair_t *pair;
1156 1157 nvlist_t *holds;
1157 1158 int error;
1158 1159
1159 1160 error = nvlist_lookup_nvlist(innvl, "holds", &holds);
1160 1161 if (error != 0)
1161 1162 return (SET_ERROR(EINVAL));
1162 1163
1163 1164 for (pair = nvlist_next_nvpair(holds, NULL); pair != NULL;
1164 1165 pair = nvlist_next_nvpair(holds, pair)) {
1165 1166 char fsname[MAXNAMELEN];
1166 1167 error = dmu_fsname(nvpair_name(pair), fsname);
1167 1168 if (error != 0)
1168 1169 return (error);
1169 1170 error = zfs_secpolicy_write_perms(fsname,
1170 1171 ZFS_DELEG_PERM_HOLD, cr);
1171 1172 if (error != 0)
1172 1173 return (error);
1173 1174 }
1174 1175 return (0);
1175 1176 }
1176 1177
1177 1178 /* ARGSUSED */
1178 1179 static int
1179 1180 zfs_secpolicy_release(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1180 1181 {
1181 1182 nvpair_t *pair;
1182 1183 int error;
1183 1184
1184 1185 for (pair = nvlist_next_nvpair(innvl, NULL); pair != NULL;
1185 1186 pair = nvlist_next_nvpair(innvl, pair)) {
1186 1187 char fsname[MAXNAMELEN];
1187 1188 error = dmu_fsname(nvpair_name(pair), fsname);
1188 1189 if (error != 0)
1189 1190 return (error);
1190 1191 error = zfs_secpolicy_write_perms(fsname,
1191 1192 ZFS_DELEG_PERM_RELEASE, cr);
1192 1193 if (error != 0)
1193 1194 return (error);
1194 1195 }
1195 1196 return (0);
1196 1197 }
1197 1198
1198 1199 /*
1199 1200 * Policy for allowing temporary snapshots to be taken or released
1200 1201 */
1201 1202 static int
1202 1203 zfs_secpolicy_tmp_snapshot(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1203 1204 {
1204 1205 /*
1205 1206 * A temporary snapshot is the same as a snapshot,
1206 1207 * hold, destroy and release all rolled into one.
1207 1208 * Delegated diff alone is sufficient that we allow this.
1208 1209 */
1209 1210 int error;
1210 1211
1211 1212 if ((error = zfs_secpolicy_write_perms(zc->zc_name,
1212 1213 ZFS_DELEG_PERM_DIFF, cr)) == 0)
1213 1214 return (0);
1214 1215
1215 1216 error = zfs_secpolicy_snapshot_perms(zc->zc_name, cr);
1216 1217 if (error == 0)
1217 1218 error = zfs_secpolicy_hold(zc, innvl, cr);
1218 1219 if (error == 0)
1219 1220 error = zfs_secpolicy_release(zc, innvl, cr);
1220 1221 if (error == 0)
1221 1222 error = zfs_secpolicy_destroy(zc, innvl, cr);
1222 1223 return (error);
1223 1224 }
1224 1225
1225 1226 /*
1226 1227 * Returns the nvlist as specified by the user in the zfs_cmd_t.
1227 1228 */
1228 1229 static int
1229 1230 get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp)
1230 1231 {
1231 1232 char *packed;
1232 1233 int error;
1233 1234 nvlist_t *list = NULL;
1234 1235
1235 1236 /*
1236 1237 * Read in and unpack the user-supplied nvlist.
1237 1238 */
1238 1239 if (size == 0)
1239 1240 return (SET_ERROR(EINVAL));
1240 1241
1241 1242 packed = kmem_alloc(size, KM_SLEEP);
1242 1243
1243 1244 if ((error = ddi_copyin((void *)(uintptr_t)nvl, packed, size,
1244 1245 iflag)) != 0) {
1245 1246 kmem_free(packed, size);
1246 1247 return (error);
1247 1248 }
1248 1249
1249 1250 if ((error = nvlist_unpack(packed, size, &list, 0)) != 0) {
1250 1251 kmem_free(packed, size);
1251 1252 return (error);
1252 1253 }
1253 1254
1254 1255 kmem_free(packed, size);
1255 1256
1256 1257 *nvp = list;
1257 1258 return (0);
1258 1259 }
1259 1260
1260 1261 /*
1261 1262 * Reduce the size of this nvlist until it can be serialized in 'max' bytes.
1262 1263 * Entries will be removed from the end of the nvlist, and one int32 entry
1263 1264 * named "N_MORE_ERRORS" will be added indicating how many entries were
1264 1265 * removed.
1265 1266 */
1266 1267 static int
1267 1268 nvlist_smush(nvlist_t *errors, size_t max)
1268 1269 {
1269 1270 size_t size;
1270 1271
1271 1272 size = fnvlist_size(errors);
1272 1273
1273 1274 if (size > max) {
1274 1275 nvpair_t *more_errors;
1275 1276 int n = 0;
1276 1277
1277 1278 if (max < 1024)
1278 1279 return (SET_ERROR(ENOMEM));
1279 1280
1280 1281 fnvlist_add_int32(errors, ZPROP_N_MORE_ERRORS, 0);
1281 1282 more_errors = nvlist_prev_nvpair(errors, NULL);
1282 1283
1283 1284 do {
1284 1285 nvpair_t *pair = nvlist_prev_nvpair(errors,
1285 1286 more_errors);
1286 1287 fnvlist_remove_nvpair(errors, pair);
1287 1288 n++;
1288 1289 size = fnvlist_size(errors);
1289 1290 } while (size > max);
1290 1291
1291 1292 fnvlist_remove_nvpair(errors, more_errors);
1292 1293 fnvlist_add_int32(errors, ZPROP_N_MORE_ERRORS, n);
1293 1294 ASSERT3U(fnvlist_size(errors), <=, max);
1294 1295 }
1295 1296
1296 1297 return (0);
1297 1298 }
1298 1299
1299 1300 static int
1300 1301 put_nvlist(zfs_cmd_t *zc, nvlist_t *nvl)
1301 1302 {
1302 1303 char *packed = NULL;
1303 1304 int error = 0;
1304 1305 size_t size;
1305 1306
1306 1307 size = fnvlist_size(nvl);
1307 1308
1308 1309 if (size > zc->zc_nvlist_dst_size) {
1309 1310 error = SET_ERROR(ENOMEM);
1310 1311 } else {
1311 1312 packed = fnvlist_pack(nvl, &size);
1312 1313 if (ddi_copyout(packed, (void *)(uintptr_t)zc->zc_nvlist_dst,
1313 1314 size, zc->zc_iflags) != 0)
1314 1315 error = SET_ERROR(EFAULT);
1315 1316 fnvlist_pack_free(packed, size);
1316 1317 }
1317 1318
1318 1319 zc->zc_nvlist_dst_size = size;
1319 1320 zc->zc_nvlist_dst_filled = B_TRUE;
1320 1321 return (error);
1321 1322 }
1322 1323
1323 1324 static int
1324 1325 getzfsvfs(const char *dsname, zfsvfs_t **zfvp)
1325 1326 {
1326 1327 objset_t *os;
1327 1328 int error;
1328 1329
1329 1330 error = dmu_objset_hold(dsname, FTAG, &os);
1330 1331 if (error != 0)
1331 1332 return (error);
1332 1333 if (dmu_objset_type(os) != DMU_OST_ZFS) {
1333 1334 dmu_objset_rele(os, FTAG);
1334 1335 return (SET_ERROR(EINVAL));
1335 1336 }
1336 1337
1337 1338 mutex_enter(&os->os_user_ptr_lock);
1338 1339 *zfvp = dmu_objset_get_user(os);
1339 1340 if (*zfvp) {
1340 1341 VFS_HOLD((*zfvp)->z_vfs);
1341 1342 } else {
1342 1343 error = SET_ERROR(ESRCH);
1343 1344 }
1344 1345 mutex_exit(&os->os_user_ptr_lock);
1345 1346 dmu_objset_rele(os, FTAG);
1346 1347 return (error);
1347 1348 }
1348 1349
1349 1350 /*
1350 1351 * Find a zfsvfs_t for a mounted filesystem, or create our own, in which
1351 1352 * case its z_vfs will be NULL, and it will be opened as the owner.
1352 1353 * If 'writer' is set, the z_teardown_lock will be held for RW_WRITER,
1353 1354 * which prevents all vnode ops from running.
1354 1355 */
1355 1356 static int
1356 1357 zfsvfs_hold(const char *name, void *tag, zfsvfs_t **zfvp, boolean_t writer)
1357 1358 {
1358 1359 int error = 0;
1359 1360
1360 1361 if (getzfsvfs(name, zfvp) != 0)
1361 1362 error = zfsvfs_create(name, zfvp);
1362 1363 if (error == 0) {
1363 1364 rrw_enter(&(*zfvp)->z_teardown_lock, (writer) ? RW_WRITER :
1364 1365 RW_READER, tag);
1365 1366 if ((*zfvp)->z_unmounted) {
1366 1367 /*
1367 1368 * XXX we could probably try again, since the unmounting
1368 1369 * thread should be just about to disassociate the
1369 1370 * objset from the zfsvfs.
1370 1371 */
1371 1372 rrw_exit(&(*zfvp)->z_teardown_lock, tag);
1372 1373 return (SET_ERROR(EBUSY));
1373 1374 }
1374 1375 }
1375 1376 return (error);
1376 1377 }
1377 1378
1378 1379 static void
1379 1380 zfsvfs_rele(zfsvfs_t *zfsvfs, void *tag)
1380 1381 {
1381 1382 rrw_exit(&zfsvfs->z_teardown_lock, tag);
1382 1383
1383 1384 if (zfsvfs->z_vfs) {
1384 1385 VFS_RELE(zfsvfs->z_vfs);
1385 1386 } else {
1386 1387 dmu_objset_disown(zfsvfs->z_os, zfsvfs);
1387 1388 zfsvfs_free(zfsvfs);
1388 1389 }
1389 1390 }
1390 1391
1391 1392 static int
1392 1393 zfs_ioc_pool_create(zfs_cmd_t *zc)
1393 1394 {
1394 1395 int error;
1395 1396 nvlist_t *config, *props = NULL;
1396 1397 nvlist_t *rootprops = NULL;
1397 1398 nvlist_t *zplprops = NULL;
1398 1399
1399 1400 if (error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1400 1401 zc->zc_iflags, &config))
1401 1402 return (error);
1402 1403
1403 1404 if (zc->zc_nvlist_src_size != 0 && (error =
1404 1405 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1405 1406 zc->zc_iflags, &props))) {
1406 1407 nvlist_free(config);
1407 1408 return (error);
1408 1409 }
1409 1410
1410 1411 if (props) {
1411 1412 nvlist_t *nvl = NULL;
1412 1413 uint64_t version = SPA_VERSION;
1413 1414
1414 1415 (void) nvlist_lookup_uint64(props,
1415 1416 zpool_prop_to_name(ZPOOL_PROP_VERSION), &version);
1416 1417 if (!SPA_VERSION_IS_SUPPORTED(version)) {
1417 1418 error = SET_ERROR(EINVAL);
1418 1419 goto pool_props_bad;
1419 1420 }
1420 1421 (void) nvlist_lookup_nvlist(props, ZPOOL_ROOTFS_PROPS, &nvl);
1421 1422 if (nvl) {
1422 1423 error = nvlist_dup(nvl, &rootprops, KM_SLEEP);
1423 1424 if (error != 0) {
1424 1425 nvlist_free(config);
1425 1426 nvlist_free(props);
1426 1427 return (error);
1427 1428 }
1428 1429 (void) nvlist_remove_all(props, ZPOOL_ROOTFS_PROPS);
1429 1430 }
1430 1431 VERIFY(nvlist_alloc(&zplprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
1431 1432 error = zfs_fill_zplprops_root(version, rootprops,
1432 1433 zplprops, NULL);
1433 1434 if (error != 0)
1434 1435 goto pool_props_bad;
1435 1436 }
1436 1437
1437 1438 error = spa_create(zc->zc_name, config, props, zplprops);
1438 1439
1439 1440 /*
1440 1441 * Set the remaining root properties
1441 1442 */
1442 1443 if (!error && (error = zfs_set_prop_nvlist(zc->zc_name,
1443 1444 ZPROP_SRC_LOCAL, rootprops, NULL)) != 0)
1444 1445 (void) spa_destroy(zc->zc_name);
1445 1446
1446 1447 pool_props_bad:
1447 1448 nvlist_free(rootprops);
1448 1449 nvlist_free(zplprops);
1449 1450 nvlist_free(config);
1450 1451 nvlist_free(props);
1451 1452
1452 1453 return (error);
1453 1454 }
1454 1455
1455 1456 static int
1456 1457 zfs_ioc_pool_destroy(zfs_cmd_t *zc)
1457 1458 {
1458 1459 int error;
1459 1460 zfs_log_history(zc);
1460 1461 error = spa_destroy(zc->zc_name);
1461 1462 if (error == 0)
1462 1463 zvol_remove_minors(zc->zc_name);
1463 1464 return (error);
1464 1465 }
1465 1466
1466 1467 static int
1467 1468 zfs_ioc_pool_import(zfs_cmd_t *zc)
1468 1469 {
1469 1470 nvlist_t *config, *props = NULL;
1470 1471 uint64_t guid;
1471 1472 int error;
1472 1473
1473 1474 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1474 1475 zc->zc_iflags, &config)) != 0)
1475 1476 return (error);
1476 1477
1477 1478 if (zc->zc_nvlist_src_size != 0 && (error =
1478 1479 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1479 1480 zc->zc_iflags, &props))) {
1480 1481 nvlist_free(config);
1481 1482 return (error);
1482 1483 }
1483 1484
1484 1485 if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, &guid) != 0 ||
1485 1486 guid != zc->zc_guid)
1486 1487 error = SET_ERROR(EINVAL);
1487 1488 else
1488 1489 error = spa_import(zc->zc_name, config, props, zc->zc_cookie);
1489 1490
1490 1491 if (zc->zc_nvlist_dst != 0) {
1491 1492 int err;
1492 1493
1493 1494 if ((err = put_nvlist(zc, config)) != 0)
1494 1495 error = err;
1495 1496 }
1496 1497
1497 1498 nvlist_free(config);
1498 1499
1499 1500 if (props)
1500 1501 nvlist_free(props);
1501 1502
1502 1503 return (error);
1503 1504 }
1504 1505
1505 1506 static int
1506 1507 zfs_ioc_pool_export(zfs_cmd_t *zc)
1507 1508 {
1508 1509 int error;
1509 1510 boolean_t force = (boolean_t)zc->zc_cookie;
1510 1511 boolean_t hardforce = (boolean_t)zc->zc_guid;
1511 1512
1512 1513 zfs_log_history(zc);
1513 1514 error = spa_export(zc->zc_name, NULL, force, hardforce);
1514 1515 if (error == 0)
1515 1516 zvol_remove_minors(zc->zc_name);
1516 1517 return (error);
1517 1518 }
1518 1519
1519 1520 static int
1520 1521 zfs_ioc_pool_configs(zfs_cmd_t *zc)
1521 1522 {
1522 1523 nvlist_t *configs;
1523 1524 int error;
1524 1525
1525 1526 if ((configs = spa_all_configs(&zc->zc_cookie)) == NULL)
1526 1527 return (SET_ERROR(EEXIST));
1527 1528
1528 1529 error = put_nvlist(zc, configs);
1529 1530
1530 1531 nvlist_free(configs);
1531 1532
1532 1533 return (error);
1533 1534 }
1534 1535
1535 1536 /*
1536 1537 * inputs:
1537 1538 * zc_name name of the pool
1538 1539 *
1539 1540 * outputs:
1540 1541 * zc_cookie real errno
1541 1542 * zc_nvlist_dst config nvlist
1542 1543 * zc_nvlist_dst_size size of config nvlist
1543 1544 */
1544 1545 static int
1545 1546 zfs_ioc_pool_stats(zfs_cmd_t *zc)
1546 1547 {
1547 1548 nvlist_t *config;
1548 1549 int error;
1549 1550 int ret = 0;
1550 1551
1551 1552 error = spa_get_stats(zc->zc_name, &config, zc->zc_value,
1552 1553 sizeof (zc->zc_value));
1553 1554
1554 1555 if (config != NULL) {
1555 1556 ret = put_nvlist(zc, config);
1556 1557 nvlist_free(config);
1557 1558
1558 1559 /*
1559 1560 * The config may be present even if 'error' is non-zero.
1560 1561 * In this case we return success, and preserve the real errno
1561 1562 * in 'zc_cookie'.
1562 1563 */
1563 1564 zc->zc_cookie = error;
1564 1565 } else {
1565 1566 ret = error;
1566 1567 }
1567 1568
1568 1569 return (ret);
1569 1570 }
1570 1571
1571 1572 /*
1572 1573 * Try to import the given pool, returning pool stats as appropriate so that
1573 1574 * user land knows which devices are available and overall pool health.
1574 1575 */
1575 1576 static int
1576 1577 zfs_ioc_pool_tryimport(zfs_cmd_t *zc)
1577 1578 {
1578 1579 nvlist_t *tryconfig, *config;
1579 1580 int error;
1580 1581
1581 1582 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1582 1583 zc->zc_iflags, &tryconfig)) != 0)
1583 1584 return (error);
1584 1585
1585 1586 config = spa_tryimport(tryconfig);
1586 1587
1587 1588 nvlist_free(tryconfig);
1588 1589
1589 1590 if (config == NULL)
1590 1591 return (SET_ERROR(EINVAL));
1591 1592
1592 1593 error = put_nvlist(zc, config);
1593 1594 nvlist_free(config);
1594 1595
1595 1596 return (error);
1596 1597 }
1597 1598
1598 1599 /*
1599 1600 * inputs:
1600 1601 * zc_name name of the pool
1601 1602 * zc_cookie scan func (pool_scan_func_t)
1602 1603 */
1603 1604 static int
1604 1605 zfs_ioc_pool_scan(zfs_cmd_t *zc)
1605 1606 {
1606 1607 spa_t *spa;
1607 1608 int error;
1608 1609
1609 1610 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1610 1611 return (error);
1611 1612
1612 1613 if (zc->zc_cookie == POOL_SCAN_NONE)
1613 1614 error = spa_scan_stop(spa);
1614 1615 else
1615 1616 error = spa_scan(spa, zc->zc_cookie);
1616 1617
1617 1618 spa_close(spa, FTAG);
1618 1619
1619 1620 return (error);
1620 1621 }
1621 1622
1622 1623 static int
1623 1624 zfs_ioc_pool_freeze(zfs_cmd_t *zc)
1624 1625 {
1625 1626 spa_t *spa;
1626 1627 int error;
1627 1628
1628 1629 error = spa_open(zc->zc_name, &spa, FTAG);
1629 1630 if (error == 0) {
1630 1631 spa_freeze(spa);
1631 1632 spa_close(spa, FTAG);
1632 1633 }
1633 1634 return (error);
1634 1635 }
1635 1636
1636 1637 static int
1637 1638 zfs_ioc_pool_upgrade(zfs_cmd_t *zc)
1638 1639 {
1639 1640 spa_t *spa;
1640 1641 int error;
1641 1642
1642 1643 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1643 1644 return (error);
1644 1645
1645 1646 if (zc->zc_cookie < spa_version(spa) ||
1646 1647 !SPA_VERSION_IS_SUPPORTED(zc->zc_cookie)) {
1647 1648 spa_close(spa, FTAG);
1648 1649 return (SET_ERROR(EINVAL));
1649 1650 }
1650 1651
1651 1652 spa_upgrade(spa, zc->zc_cookie);
1652 1653 spa_close(spa, FTAG);
1653 1654
1654 1655 return (error);
1655 1656 }
1656 1657
1657 1658 static int
1658 1659 zfs_ioc_pool_get_history(zfs_cmd_t *zc)
1659 1660 {
1660 1661 spa_t *spa;
1661 1662 char *hist_buf;
1662 1663 uint64_t size;
1663 1664 int error;
1664 1665
1665 1666 if ((size = zc->zc_history_len) == 0)
1666 1667 return (SET_ERROR(EINVAL));
1667 1668
1668 1669 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1669 1670 return (error);
1670 1671
1671 1672 if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) {
1672 1673 spa_close(spa, FTAG);
1673 1674 return (SET_ERROR(ENOTSUP));
1674 1675 }
1675 1676
1676 1677 hist_buf = kmem_alloc(size, KM_SLEEP);
1677 1678 if ((error = spa_history_get(spa, &zc->zc_history_offset,
1678 1679 &zc->zc_history_len, hist_buf)) == 0) {
1679 1680 error = ddi_copyout(hist_buf,
1680 1681 (void *)(uintptr_t)zc->zc_history,
1681 1682 zc->zc_history_len, zc->zc_iflags);
1682 1683 }
1683 1684
1684 1685 spa_close(spa, FTAG);
1685 1686 kmem_free(hist_buf, size);
1686 1687 return (error);
1687 1688 }
1688 1689
1689 1690 static int
1690 1691 zfs_ioc_pool_reguid(zfs_cmd_t *zc)
1691 1692 {
1692 1693 spa_t *spa;
1693 1694 int error;
1694 1695
1695 1696 error = spa_open(zc->zc_name, &spa, FTAG);
1696 1697 if (error == 0) {
1697 1698 error = spa_change_guid(spa);
1698 1699 spa_close(spa, FTAG);
1699 1700 }
1700 1701 return (error);
1701 1702 }
1702 1703
1703 1704 static int
1704 1705 zfs_ioc_dsobj_to_dsname(zfs_cmd_t *zc)
1705 1706 {
1706 1707 return (dsl_dsobj_to_dsname(zc->zc_name, zc->zc_obj, zc->zc_value));
1707 1708 }
1708 1709
1709 1710 /*
1710 1711 * inputs:
1711 1712 * zc_name name of filesystem
1712 1713 * zc_obj object to find
1713 1714 *
1714 1715 * outputs:
1715 1716 * zc_value name of object
1716 1717 */
1717 1718 static int
1718 1719 zfs_ioc_obj_to_path(zfs_cmd_t *zc)
1719 1720 {
1720 1721 objset_t *os;
1721 1722 int error;
1722 1723
1723 1724 /* XXX reading from objset not owned */
1724 1725 if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os)) != 0)
1725 1726 return (error);
1726 1727 if (dmu_objset_type(os) != DMU_OST_ZFS) {
1727 1728 dmu_objset_rele(os, FTAG);
1728 1729 return (SET_ERROR(EINVAL));
1729 1730 }
1730 1731 error = zfs_obj_to_path(os, zc->zc_obj, zc->zc_value,
1731 1732 sizeof (zc->zc_value));
1732 1733 dmu_objset_rele(os, FTAG);
1733 1734
1734 1735 return (error);
1735 1736 }
1736 1737
1737 1738 /*
1738 1739 * inputs:
1739 1740 * zc_name name of filesystem
1740 1741 * zc_obj object to find
1741 1742 *
1742 1743 * outputs:
1743 1744 * zc_stat stats on object
1744 1745 * zc_value path to object
1745 1746 */
1746 1747 static int
1747 1748 zfs_ioc_obj_to_stats(zfs_cmd_t *zc)
1748 1749 {
1749 1750 objset_t *os;
1750 1751 int error;
1751 1752
1752 1753 /* XXX reading from objset not owned */
1753 1754 if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os)) != 0)
1754 1755 return (error);
1755 1756 if (dmu_objset_type(os) != DMU_OST_ZFS) {
1756 1757 dmu_objset_rele(os, FTAG);
1757 1758 return (SET_ERROR(EINVAL));
1758 1759 }
1759 1760 error = zfs_obj_to_stats(os, zc->zc_obj, &zc->zc_stat, zc->zc_value,
1760 1761 sizeof (zc->zc_value));
1761 1762 dmu_objset_rele(os, FTAG);
1762 1763
1763 1764 return (error);
1764 1765 }
1765 1766
1766 1767 static int
1767 1768 zfs_ioc_vdev_add(zfs_cmd_t *zc)
1768 1769 {
1769 1770 spa_t *spa;
1770 1771 int error;
1771 1772 nvlist_t *config, **l2cache, **spares;
1772 1773 uint_t nl2cache = 0, nspares = 0;
1773 1774
1774 1775 error = spa_open(zc->zc_name, &spa, FTAG);
1775 1776 if (error != 0)
1776 1777 return (error);
1777 1778
1778 1779 error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1779 1780 zc->zc_iflags, &config);
1780 1781 (void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_L2CACHE,
1781 1782 &l2cache, &nl2cache);
1782 1783
1783 1784 (void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_SPARES,
1784 1785 &spares, &nspares);
1785 1786
1786 1787 /*
1787 1788 * A root pool with concatenated devices is not supported.
1788 1789 * Thus, can not add a device to a root pool.
1789 1790 *
1790 1791 * Intent log device can not be added to a rootpool because
1791 1792 * during mountroot, zil is replayed, a seperated log device
1792 1793 * can not be accessed during the mountroot time.
1793 1794 *
1794 1795 * l2cache and spare devices are ok to be added to a rootpool.
1795 1796 */
1796 1797 if (spa_bootfs(spa) != 0 && nl2cache == 0 && nspares == 0) {
1797 1798 nvlist_free(config);
1798 1799 spa_close(spa, FTAG);
1799 1800 return (SET_ERROR(EDOM));
1800 1801 }
1801 1802
1802 1803 if (error == 0) {
1803 1804 error = spa_vdev_add(spa, config);
1804 1805 nvlist_free(config);
1805 1806 }
1806 1807 spa_close(spa, FTAG);
1807 1808 return (error);
1808 1809 }
1809 1810
1810 1811 /*
1811 1812 * inputs:
1812 1813 * zc_name name of the pool
1813 1814 * zc_nvlist_conf nvlist of devices to remove
1814 1815 * zc_cookie to stop the remove?
1815 1816 */
1816 1817 static int
1817 1818 zfs_ioc_vdev_remove(zfs_cmd_t *zc)
1818 1819 {
1819 1820 spa_t *spa;
1820 1821 int error;
1821 1822
1822 1823 error = spa_open(zc->zc_name, &spa, FTAG);
1823 1824 if (error != 0)
1824 1825 return (error);
1825 1826 error = spa_vdev_remove(spa, zc->zc_guid, B_FALSE);
1826 1827 spa_close(spa, FTAG);
1827 1828 return (error);
1828 1829 }
1829 1830
1830 1831 static int
1831 1832 zfs_ioc_vdev_set_state(zfs_cmd_t *zc)
1832 1833 {
1833 1834 spa_t *spa;
1834 1835 int error;
1835 1836 vdev_state_t newstate = VDEV_STATE_UNKNOWN;
1836 1837
1837 1838 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1838 1839 return (error);
1839 1840 switch (zc->zc_cookie) {
1840 1841 case VDEV_STATE_ONLINE:
1841 1842 error = vdev_online(spa, zc->zc_guid, zc->zc_obj, &newstate);
1842 1843 break;
1843 1844
1844 1845 case VDEV_STATE_OFFLINE:
1845 1846 error = vdev_offline(spa, zc->zc_guid, zc->zc_obj);
1846 1847 break;
1847 1848
1848 1849 case VDEV_STATE_FAULTED:
1849 1850 if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED &&
1850 1851 zc->zc_obj != VDEV_AUX_EXTERNAL)
1851 1852 zc->zc_obj = VDEV_AUX_ERR_EXCEEDED;
1852 1853
1853 1854 error = vdev_fault(spa, zc->zc_guid, zc->zc_obj);
1854 1855 break;
1855 1856
1856 1857 case VDEV_STATE_DEGRADED:
1857 1858 if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED &&
1858 1859 zc->zc_obj != VDEV_AUX_EXTERNAL)
1859 1860 zc->zc_obj = VDEV_AUX_ERR_EXCEEDED;
1860 1861
1861 1862 error = vdev_degrade(spa, zc->zc_guid, zc->zc_obj);
1862 1863 break;
1863 1864
1864 1865 default:
1865 1866 error = SET_ERROR(EINVAL);
1866 1867 }
1867 1868 zc->zc_cookie = newstate;
1868 1869 spa_close(spa, FTAG);
1869 1870 return (error);
1870 1871 }
1871 1872
1872 1873 static int
1873 1874 zfs_ioc_vdev_attach(zfs_cmd_t *zc)
1874 1875 {
1875 1876 spa_t *spa;
1876 1877 int replacing = zc->zc_cookie;
1877 1878 nvlist_t *config;
1878 1879 int error;
1879 1880
1880 1881 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1881 1882 return (error);
1882 1883
1883 1884 if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1884 1885 zc->zc_iflags, &config)) == 0) {
1885 1886 error = spa_vdev_attach(spa, zc->zc_guid, config, replacing);
1886 1887 nvlist_free(config);
1887 1888 }
1888 1889
1889 1890 spa_close(spa, FTAG);
1890 1891 return (error);
1891 1892 }
1892 1893
1893 1894 static int
1894 1895 zfs_ioc_vdev_detach(zfs_cmd_t *zc)
1895 1896 {
1896 1897 spa_t *spa;
1897 1898 int error;
1898 1899
1899 1900 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1900 1901 return (error);
1901 1902
1902 1903 error = spa_vdev_detach(spa, zc->zc_guid, 0, B_FALSE);
1903 1904
1904 1905 spa_close(spa, FTAG);
1905 1906 return (error);
1906 1907 }
1907 1908
1908 1909 static int
1909 1910 zfs_ioc_vdev_split(zfs_cmd_t *zc)
1910 1911 {
1911 1912 spa_t *spa;
1912 1913 nvlist_t *config, *props = NULL;
1913 1914 int error;
1914 1915 boolean_t exp = !!(zc->zc_cookie & ZPOOL_EXPORT_AFTER_SPLIT);
1915 1916
1916 1917 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1917 1918 return (error);
1918 1919
1919 1920 if (error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1920 1921 zc->zc_iflags, &config)) {
1921 1922 spa_close(spa, FTAG);
1922 1923 return (error);
1923 1924 }
1924 1925
1925 1926 if (zc->zc_nvlist_src_size != 0 && (error =
1926 1927 get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1927 1928 zc->zc_iflags, &props))) {
1928 1929 spa_close(spa, FTAG);
1929 1930 nvlist_free(config);
1930 1931 return (error);
1931 1932 }
1932 1933
1933 1934 error = spa_vdev_split_mirror(spa, zc->zc_string, config, props, exp);
1934 1935
1935 1936 spa_close(spa, FTAG);
1936 1937
1937 1938 nvlist_free(config);
1938 1939 nvlist_free(props);
1939 1940
1940 1941 return (error);
1941 1942 }
1942 1943
1943 1944 static int
1944 1945 zfs_ioc_vdev_setpath(zfs_cmd_t *zc)
1945 1946 {
1946 1947 spa_t *spa;
1947 1948 char *path = zc->zc_value;
1948 1949 uint64_t guid = zc->zc_guid;
1949 1950 int error;
1950 1951
1951 1952 error = spa_open(zc->zc_name, &spa, FTAG);
1952 1953 if (error != 0)
1953 1954 return (error);
1954 1955
1955 1956 error = spa_vdev_setpath(spa, guid, path);
1956 1957 spa_close(spa, FTAG);
1957 1958 return (error);
1958 1959 }
1959 1960
1960 1961 static int
1961 1962 zfs_ioc_vdev_setfru(zfs_cmd_t *zc)
1962 1963 {
1963 1964 spa_t *spa;
1964 1965 char *fru = zc->zc_value;
1965 1966 uint64_t guid = zc->zc_guid;
1966 1967 int error;
1967 1968
1968 1969 error = spa_open(zc->zc_name, &spa, FTAG);
1969 1970 if (error != 0)
1970 1971 return (error);
1971 1972
1972 1973 error = spa_vdev_setfru(spa, guid, fru);
1973 1974 spa_close(spa, FTAG);
1974 1975 return (error);
1975 1976 }
1976 1977
1977 1978 static int
1978 1979 zfs_ioc_objset_stats_impl(zfs_cmd_t *zc, objset_t *os)
1979 1980 {
1980 1981 int error = 0;
1981 1982 nvlist_t *nv;
1982 1983
1983 1984 dmu_objset_fast_stat(os, &zc->zc_objset_stats);
1984 1985
1985 1986 if (zc->zc_nvlist_dst != 0 &&
1986 1987 (error = dsl_prop_get_all(os, &nv)) == 0) {
1987 1988 dmu_objset_stats(os, nv);
1988 1989 /*
1989 1990 * NB: zvol_get_stats() will read the objset contents,
1990 1991 * which we aren't supposed to do with a
1991 1992 * DS_MODE_USER hold, because it could be
1992 1993 * inconsistent. So this is a bit of a workaround...
1993 1994 * XXX reading with out owning
1994 1995 */
1995 1996 if (!zc->zc_objset_stats.dds_inconsistent &&
1996 1997 dmu_objset_type(os) == DMU_OST_ZVOL) {
1997 1998 error = zvol_get_stats(os, nv);
1998 1999 if (error == EIO)
1999 2000 return (error);
2000 2001 VERIFY0(error);
2001 2002 }
2002 2003 error = put_nvlist(zc, nv);
2003 2004 nvlist_free(nv);
2004 2005 }
2005 2006
2006 2007 return (error);
2007 2008 }
2008 2009
2009 2010 /*
2010 2011 * inputs:
2011 2012 * zc_name name of filesystem
2012 2013 * zc_nvlist_dst_size size of buffer for property nvlist
2013 2014 *
2014 2015 * outputs:
2015 2016 * zc_objset_stats stats
2016 2017 * zc_nvlist_dst property nvlist
2017 2018 * zc_nvlist_dst_size size of property nvlist
2018 2019 */
2019 2020 static int
2020 2021 zfs_ioc_objset_stats(zfs_cmd_t *zc)
2021 2022 {
2022 2023 objset_t *os;
2023 2024 int error;
2024 2025
2025 2026 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
2026 2027 if (error == 0) {
2027 2028 error = zfs_ioc_objset_stats_impl(zc, os);
2028 2029 dmu_objset_rele(os, FTAG);
2029 2030 }
2030 2031
2031 2032 return (error);
2032 2033 }
2033 2034
2034 2035 /*
2035 2036 * inputs:
2036 2037 * zc_name name of filesystem
2037 2038 * zc_nvlist_dst_size size of buffer for property nvlist
2038 2039 *
2039 2040 * outputs:
2040 2041 * zc_nvlist_dst received property nvlist
2041 2042 * zc_nvlist_dst_size size of received property nvlist
2042 2043 *
2043 2044 * Gets received properties (distinct from local properties on or after
2044 2045 * SPA_VERSION_RECVD_PROPS) for callers who want to differentiate received from
2045 2046 * local property values.
2046 2047 */
2047 2048 static int
2048 2049 zfs_ioc_objset_recvd_props(zfs_cmd_t *zc)
2049 2050 {
2050 2051 int error = 0;
2051 2052 nvlist_t *nv;
2052 2053
2053 2054 /*
2054 2055 * Without this check, we would return local property values if the
2055 2056 * caller has not already received properties on or after
2056 2057 * SPA_VERSION_RECVD_PROPS.
2057 2058 */
2058 2059 if (!dsl_prop_get_hasrecvd(zc->zc_name))
2059 2060 return (SET_ERROR(ENOTSUP));
2060 2061
2061 2062 if (zc->zc_nvlist_dst != 0 &&
2062 2063 (error = dsl_prop_get_received(zc->zc_name, &nv)) == 0) {
2063 2064 error = put_nvlist(zc, nv);
2064 2065 nvlist_free(nv);
2065 2066 }
2066 2067
2067 2068 return (error);
2068 2069 }
2069 2070
2070 2071 static int
2071 2072 nvl_add_zplprop(objset_t *os, nvlist_t *props, zfs_prop_t prop)
2072 2073 {
2073 2074 uint64_t value;
2074 2075 int error;
2075 2076
2076 2077 /*
2077 2078 * zfs_get_zplprop() will either find a value or give us
2078 2079 * the default value (if there is one).
2079 2080 */
2080 2081 if ((error = zfs_get_zplprop(os, prop, &value)) != 0)
2081 2082 return (error);
2082 2083 VERIFY(nvlist_add_uint64(props, zfs_prop_to_name(prop), value) == 0);
2083 2084 return (0);
2084 2085 }
2085 2086
2086 2087 /*
2087 2088 * inputs:
2088 2089 * zc_name name of filesystem
2089 2090 * zc_nvlist_dst_size size of buffer for zpl property nvlist
2090 2091 *
2091 2092 * outputs:
2092 2093 * zc_nvlist_dst zpl property nvlist
2093 2094 * zc_nvlist_dst_size size of zpl property nvlist
2094 2095 */
2095 2096 static int
2096 2097 zfs_ioc_objset_zplprops(zfs_cmd_t *zc)
2097 2098 {
2098 2099 objset_t *os;
2099 2100 int err;
2100 2101
2101 2102 /* XXX reading without owning */
2102 2103 if (err = dmu_objset_hold(zc->zc_name, FTAG, &os))
2103 2104 return (err);
2104 2105
2105 2106 dmu_objset_fast_stat(os, &zc->zc_objset_stats);
2106 2107
2107 2108 /*
2108 2109 * NB: nvl_add_zplprop() will read the objset contents,
2109 2110 * which we aren't supposed to do with a DS_MODE_USER
2110 2111 * hold, because it could be inconsistent.
2111 2112 */
2112 2113 if (zc->zc_nvlist_dst != NULL &&
2113 2114 !zc->zc_objset_stats.dds_inconsistent &&
2114 2115 dmu_objset_type(os) == DMU_OST_ZFS) {
2115 2116 nvlist_t *nv;
2116 2117
2117 2118 VERIFY(nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2118 2119 if ((err = nvl_add_zplprop(os, nv, ZFS_PROP_VERSION)) == 0 &&
2119 2120 (err = nvl_add_zplprop(os, nv, ZFS_PROP_NORMALIZE)) == 0 &&
2120 2121 (err = nvl_add_zplprop(os, nv, ZFS_PROP_UTF8ONLY)) == 0 &&
2121 2122 (err = nvl_add_zplprop(os, nv, ZFS_PROP_CASE)) == 0)
2122 2123 err = put_nvlist(zc, nv);
2123 2124 nvlist_free(nv);
2124 2125 } else {
2125 2126 err = SET_ERROR(ENOENT);
2126 2127 }
2127 2128 dmu_objset_rele(os, FTAG);
2128 2129 return (err);
2129 2130 }
2130 2131
2131 2132 static boolean_t
2132 2133 dataset_name_hidden(const char *name)
2133 2134 {
2134 2135 /*
2135 2136 * Skip over datasets that are not visible in this zone,
2136 2137 * internal datasets (which have a $ in their name), and
2137 2138 * temporary datasets (which have a % in their name).
2138 2139 */
2139 2140 if (strchr(name, '$') != NULL)
2140 2141 return (B_TRUE);
2141 2142 if (strchr(name, '%') != NULL)
2142 2143 return (B_TRUE);
2143 2144 if (!INGLOBALZONE(curproc) && !zone_dataset_visible(name, NULL))
2144 2145 return (B_TRUE);
2145 2146 return (B_FALSE);
2146 2147 }
2147 2148
2148 2149 /*
2149 2150 * inputs:
2150 2151 * zc_name name of filesystem
2151 2152 * zc_cookie zap cursor
2152 2153 * zc_nvlist_dst_size size of buffer for property nvlist
2153 2154 *
2154 2155 * outputs:
2155 2156 * zc_name name of next filesystem
2156 2157 * zc_cookie zap cursor
2157 2158 * zc_objset_stats stats
2158 2159 * zc_nvlist_dst property nvlist
2159 2160 * zc_nvlist_dst_size size of property nvlist
2160 2161 */
2161 2162 static int
2162 2163 zfs_ioc_dataset_list_next(zfs_cmd_t *zc)
2163 2164 {
2164 2165 objset_t *os;
2165 2166 int error;
2166 2167 char *p;
2167 2168 size_t orig_len = strlen(zc->zc_name);
2168 2169
2169 2170 top:
2170 2171 if (error = dmu_objset_hold(zc->zc_name, FTAG, &os)) {
2171 2172 if (error == ENOENT)
2172 2173 error = SET_ERROR(ESRCH);
2173 2174 return (error);
2174 2175 }
2175 2176
2176 2177 p = strrchr(zc->zc_name, '/');
2177 2178 if (p == NULL || p[1] != '\0')
2178 2179 (void) strlcat(zc->zc_name, "/", sizeof (zc->zc_name));
2179 2180 p = zc->zc_name + strlen(zc->zc_name);
2180 2181
2181 2182 do {
2182 2183 error = dmu_dir_list_next(os,
2183 2184 sizeof (zc->zc_name) - (p - zc->zc_name), p,
2184 2185 NULL, &zc->zc_cookie);
2185 2186 if (error == ENOENT)
2186 2187 error = SET_ERROR(ESRCH);
2187 2188 } while (error == 0 && dataset_name_hidden(zc->zc_name));
2188 2189 dmu_objset_rele(os, FTAG);
2189 2190
2190 2191 /*
2191 2192 * If it's an internal dataset (ie. with a '$' in its name),
2192 2193 * don't try to get stats for it, otherwise we'll return ENOENT.
2193 2194 */
2194 2195 if (error == 0 && strchr(zc->zc_name, '$') == NULL) {
2195 2196 error = zfs_ioc_objset_stats(zc); /* fill in the stats */
2196 2197 if (error == ENOENT) {
2197 2198 /* We lost a race with destroy, get the next one. */
2198 2199 zc->zc_name[orig_len] = '\0';
2199 2200 goto top;
2200 2201 }
2201 2202 }
2202 2203 return (error);
2203 2204 }
2204 2205
2205 2206 /*
2206 2207 * inputs:
2207 2208 * zc_name name of filesystem
2208 2209 * zc_cookie zap cursor
2209 2210 * zc_nvlist_dst_size size of buffer for property nvlist
2210 2211 *
2211 2212 * outputs:
2212 2213 * zc_name name of next snapshot
2213 2214 * zc_objset_stats stats
2214 2215 * zc_nvlist_dst property nvlist
2215 2216 * zc_nvlist_dst_size size of property nvlist
2216 2217 */
2217 2218 static int
2218 2219 zfs_ioc_snapshot_list_next(zfs_cmd_t *zc)
2219 2220 {
2220 2221 objset_t *os;
2221 2222 int error;
2222 2223
2223 2224 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
2224 2225 if (error != 0) {
2225 2226 return (error == ENOENT ? ESRCH : error);
2226 2227 }
2227 2228
2228 2229 /*
2229 2230 * A dataset name of maximum length cannot have any snapshots,
2230 2231 * so exit immediately.
2231 2232 */
2232 2233 if (strlcat(zc->zc_name, "@", sizeof (zc->zc_name)) >= MAXNAMELEN) {
2233 2234 dmu_objset_rele(os, FTAG);
2234 2235 return (SET_ERROR(ESRCH));
2235 2236 }
2236 2237
2237 2238 error = dmu_snapshot_list_next(os,
2238 2239 sizeof (zc->zc_name) - strlen(zc->zc_name),
2239 2240 zc->zc_name + strlen(zc->zc_name), &zc->zc_obj, &zc->zc_cookie,
2240 2241 NULL);
2241 2242
2242 2243 if (error == 0) {
2243 2244 dsl_dataset_t *ds;
2244 2245 dsl_pool_t *dp = os->os_dsl_dataset->ds_dir->dd_pool;
2245 2246
2246 2247 error = dsl_dataset_hold_obj(dp, zc->zc_obj, FTAG, &ds);
2247 2248 if (error == 0) {
2248 2249 objset_t *ossnap;
2249 2250
2250 2251 error = dmu_objset_from_ds(ds, &ossnap);
2251 2252 if (error == 0)
2252 2253 error = zfs_ioc_objset_stats_impl(zc, ossnap);
2253 2254 dsl_dataset_rele(ds, FTAG);
2254 2255 }
2255 2256 } else if (error == ENOENT) {
2256 2257 error = SET_ERROR(ESRCH);
2257 2258 }
2258 2259
2259 2260 dmu_objset_rele(os, FTAG);
2260 2261 /* if we failed, undo the @ that we tacked on to zc_name */
2261 2262 if (error != 0)
2262 2263 *strchr(zc->zc_name, '@') = '\0';
2263 2264 return (error);
2264 2265 }
2265 2266
2266 2267 static int
2267 2268 zfs_prop_set_userquota(const char *dsname, nvpair_t *pair)
2268 2269 {
2269 2270 const char *propname = nvpair_name(pair);
2270 2271 uint64_t *valary;
2271 2272 unsigned int vallen;
2272 2273 const char *domain;
2273 2274 char *dash;
2274 2275 zfs_userquota_prop_t type;
2275 2276 uint64_t rid;
2276 2277 uint64_t quota;
2277 2278 zfsvfs_t *zfsvfs;
2278 2279 int err;
2279 2280
2280 2281 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2281 2282 nvlist_t *attrs;
2282 2283 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
2283 2284 if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2284 2285 &pair) != 0)
2285 2286 return (SET_ERROR(EINVAL));
2286 2287 }
2287 2288
2288 2289 /*
2289 2290 * A correctly constructed propname is encoded as
2290 2291 * userquota@<rid>-<domain>.
2291 2292 */
2292 2293 if ((dash = strchr(propname, '-')) == NULL ||
2293 2294 nvpair_value_uint64_array(pair, &valary, &vallen) != 0 ||
2294 2295 vallen != 3)
2295 2296 return (SET_ERROR(EINVAL));
2296 2297
2297 2298 domain = dash + 1;
2298 2299 type = valary[0];
2299 2300 rid = valary[1];
2300 2301 quota = valary[2];
2301 2302
2302 2303 err = zfsvfs_hold(dsname, FTAG, &zfsvfs, B_FALSE);
2303 2304 if (err == 0) {
2304 2305 err = zfs_set_userquota(zfsvfs, type, domain, rid, quota);
2305 2306 zfsvfs_rele(zfsvfs, FTAG);
2306 2307 }
2307 2308
2308 2309 return (err);
2309 2310 }
2310 2311
2311 2312 /*
2312 2313 * If the named property is one that has a special function to set its value,
2313 2314 * return 0 on success and a positive error code on failure; otherwise if it is
2314 2315 * not one of the special properties handled by this function, return -1.
2315 2316 *
2316 2317 * XXX: It would be better for callers of the property interface if we handled
2317 2318 * these special cases in dsl_prop.c (in the dsl layer).
2318 2319 */
2319 2320 static int
2320 2321 zfs_prop_set_special(const char *dsname, zprop_source_t source,
2321 2322 nvpair_t *pair)
2322 2323 {
2323 2324 const char *propname = nvpair_name(pair);
2324 2325 zfs_prop_t prop = zfs_name_to_prop(propname);
2325 2326 uint64_t intval;
2326 2327 int err;
2327 2328
2328 2329 if (prop == ZPROP_INVAL) {
2329 2330 if (zfs_prop_userquota(propname))
2330 2331 return (zfs_prop_set_userquota(dsname, pair));
2331 2332 return (-1);
2332 2333 }
2333 2334
2334 2335 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2335 2336 nvlist_t *attrs;
2336 2337 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
2337 2338 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2338 2339 &pair) == 0);
2339 2340 }
2340 2341
2341 2342 if (zfs_prop_get_type(prop) == PROP_TYPE_STRING)
2342 2343 return (-1);
2343 2344
2344 2345 VERIFY(0 == nvpair_value_uint64(pair, &intval));
2345 2346
2346 2347 switch (prop) {
2347 2348 case ZFS_PROP_QUOTA:
2348 2349 err = dsl_dir_set_quota(dsname, source, intval);
2349 2350 break;
2350 2351 case ZFS_PROP_REFQUOTA:
2351 2352 err = dsl_dataset_set_refquota(dsname, source, intval);
2352 2353 break;
2353 2354 case ZFS_PROP_RESERVATION:
2354 2355 err = dsl_dir_set_reservation(dsname, source, intval);
2355 2356 break;
2356 2357 case ZFS_PROP_REFRESERVATION:
2357 2358 err = dsl_dataset_set_refreservation(dsname, source, intval);
2358 2359 break;
2359 2360 case ZFS_PROP_VOLSIZE:
2360 2361 err = zvol_set_volsize(dsname, intval);
2361 2362 break;
2362 2363 case ZFS_PROP_VERSION:
2363 2364 {
2364 2365 zfsvfs_t *zfsvfs;
2365 2366
2366 2367 if ((err = zfsvfs_hold(dsname, FTAG, &zfsvfs, B_TRUE)) != 0)
2367 2368 break;
2368 2369
2369 2370 err = zfs_set_version(zfsvfs, intval);
2370 2371 zfsvfs_rele(zfsvfs, FTAG);
2371 2372
2372 2373 if (err == 0 && intval >= ZPL_VERSION_USERSPACE) {
2373 2374 zfs_cmd_t *zc;
2374 2375
2375 2376 zc = kmem_zalloc(sizeof (zfs_cmd_t), KM_SLEEP);
2376 2377 (void) strcpy(zc->zc_name, dsname);
2377 2378 (void) zfs_ioc_userspace_upgrade(zc);
2378 2379 kmem_free(zc, sizeof (zfs_cmd_t));
2379 2380 }
2380 2381 break;
2381 2382 }
2382 2383 case ZFS_PROP_COMPRESSION:
2383 2384 {
2384 2385 if (intval == ZIO_COMPRESS_LZ4) {
2385 2386 zfeature_info_t *feature =
2386 2387 &spa_feature_table[SPA_FEATURE_LZ4_COMPRESS];
2387 2388 spa_t *spa;
2388 2389
2389 2390 if ((err = spa_open(dsname, &spa, FTAG)) != 0)
2390 2391 return (err);
2391 2392
2392 2393 /*
2393 2394 * Setting the LZ4 compression algorithm activates
2394 2395 * the feature.
2395 2396 */
2396 2397 if (!spa_feature_is_active(spa, feature)) {
2397 2398 if ((err = zfs_prop_activate_feature(spa,
2398 2399 feature)) != 0) {
2399 2400 spa_close(spa, FTAG);
2400 2401 return (err);
2401 2402 }
2402 2403 }
|
↓ open down ↓ |
2211 lines elided |
↑ open up ↑ |
2403 2404
2404 2405 spa_close(spa, FTAG);
2405 2406 }
2406 2407 /*
2407 2408 * We still want the default set action to be performed in the
2408 2409 * caller, we only performed zfeature settings here.
2409 2410 */
2410 2411 err = -1;
2411 2412 break;
2412 2413 }
2414 + case ZFS_PROP_CHECKSUM:
2415 + case ZFS_PROP_DEDUP:
2416 + {
2417 + zfeature_info_t *feature = NULL;
2418 + spa_t *spa;
2413 2419
2420 + if (intval == ZIO_CHECKSUM_SHA512)
2421 + feature = &spa_feature_table[SPA_FEATURE_SHA512];
2422 + else if (intval == ZIO_CHECKSUM_SKEIN)
2423 + feature = &spa_feature_table[SPA_FEATURE_SKEIN];
2424 + else if (intval == ZIO_CHECKSUM_EDONR)
2425 + feature = &spa_feature_table[SPA_FEATURE_EDONR];
2426 + if (feature == NULL) {
2427 + /* No features need to be activated for this cksum */
2428 + err = -1;
2429 + break;
2430 + }
2431 + if ((err = spa_open(dsname, &spa, FTAG)) != 0)
2432 + return (err);
2433 + if (!spa_feature_is_active(spa, feature)) {
2434 + /* Salted checksums must store the salt in the MOS */
2435 + if (zio_checksum_table[intval].ci_salted)
2436 + err = spa_activate_salted_cksum(spa, feature);
2437 + else
2438 + err = zfs_prop_activate_feature(spa, feature);
2439 + }
2440 + spa_close(spa, FTAG);
2441 + if (err == 0) {
2442 + /*
2443 + * We want the default set action to be performed in
2444 + * the caller, we only perform feature checks here.
2445 + */
2446 + err = -1;
2447 + }
2448 + break;
2449 + }
2450 +
2414 2451 default:
2415 2452 err = -1;
2416 2453 }
2417 2454
2418 2455 return (err);
2419 2456 }
2420 2457
2421 2458 /*
2422 2459 * This function is best effort. If it fails to set any of the given properties,
2423 2460 * it continues to set as many as it can and returns the last error
2424 2461 * encountered. If the caller provides a non-NULL errlist, it will be filled in
2425 2462 * with the list of names of all the properties that failed along with the
2426 2463 * corresponding error numbers.
2427 2464 *
2428 2465 * If every property is set successfully, zero is returned and errlist is not
2429 2466 * modified.
2430 2467 */
2431 2468 int
2432 2469 zfs_set_prop_nvlist(const char *dsname, zprop_source_t source, nvlist_t *nvl,
2433 2470 nvlist_t *errlist)
2434 2471 {
2435 2472 nvpair_t *pair;
2436 2473 nvpair_t *propval;
2437 2474 int rv = 0;
2438 2475 uint64_t intval;
2439 2476 char *strval;
2440 2477 nvlist_t *genericnvl = fnvlist_alloc();
2441 2478 nvlist_t *retrynvl = fnvlist_alloc();
2442 2479
2443 2480 retry:
2444 2481 pair = NULL;
2445 2482 while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) {
2446 2483 const char *propname = nvpair_name(pair);
2447 2484 zfs_prop_t prop = zfs_name_to_prop(propname);
2448 2485 int err = 0;
2449 2486
2450 2487 /* decode the property value */
2451 2488 propval = pair;
2452 2489 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2453 2490 nvlist_t *attrs;
2454 2491 attrs = fnvpair_value_nvlist(pair);
2455 2492 if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2456 2493 &propval) != 0)
2457 2494 err = SET_ERROR(EINVAL);
2458 2495 }
2459 2496
2460 2497 /* Validate value type */
2461 2498 if (err == 0 && prop == ZPROP_INVAL) {
2462 2499 if (zfs_prop_user(propname)) {
2463 2500 if (nvpair_type(propval) != DATA_TYPE_STRING)
2464 2501 err = SET_ERROR(EINVAL);
2465 2502 } else if (zfs_prop_userquota(propname)) {
2466 2503 if (nvpair_type(propval) !=
2467 2504 DATA_TYPE_UINT64_ARRAY)
2468 2505 err = SET_ERROR(EINVAL);
2469 2506 } else {
2470 2507 err = SET_ERROR(EINVAL);
2471 2508 }
2472 2509 } else if (err == 0) {
2473 2510 if (nvpair_type(propval) == DATA_TYPE_STRING) {
2474 2511 if (zfs_prop_get_type(prop) != PROP_TYPE_STRING)
2475 2512 err = SET_ERROR(EINVAL);
2476 2513 } else if (nvpair_type(propval) == DATA_TYPE_UINT64) {
2477 2514 const char *unused;
2478 2515
2479 2516 intval = fnvpair_value_uint64(propval);
2480 2517
2481 2518 switch (zfs_prop_get_type(prop)) {
2482 2519 case PROP_TYPE_NUMBER:
2483 2520 break;
2484 2521 case PROP_TYPE_STRING:
2485 2522 err = SET_ERROR(EINVAL);
2486 2523 break;
2487 2524 case PROP_TYPE_INDEX:
2488 2525 if (zfs_prop_index_to_string(prop,
2489 2526 intval, &unused) != 0)
2490 2527 err = SET_ERROR(EINVAL);
2491 2528 break;
2492 2529 default:
2493 2530 cmn_err(CE_PANIC,
2494 2531 "unknown property type");
2495 2532 }
2496 2533 } else {
2497 2534 err = SET_ERROR(EINVAL);
2498 2535 }
2499 2536 }
2500 2537
2501 2538 /* Validate permissions */
2502 2539 if (err == 0)
2503 2540 err = zfs_check_settable(dsname, pair, CRED());
2504 2541
2505 2542 if (err == 0) {
2506 2543 err = zfs_prop_set_special(dsname, source, pair);
2507 2544 if (err == -1) {
2508 2545 /*
2509 2546 * For better performance we build up a list of
2510 2547 * properties to set in a single transaction.
2511 2548 */
2512 2549 err = nvlist_add_nvpair(genericnvl, pair);
2513 2550 } else if (err != 0 && nvl != retrynvl) {
2514 2551 /*
2515 2552 * This may be a spurious error caused by
2516 2553 * receiving quota and reservation out of order.
2517 2554 * Try again in a second pass.
2518 2555 */
2519 2556 err = nvlist_add_nvpair(retrynvl, pair);
2520 2557 }
2521 2558 }
2522 2559
2523 2560 if (err != 0) {
2524 2561 if (errlist != NULL)
2525 2562 fnvlist_add_int32(errlist, propname, err);
2526 2563 rv = err;
2527 2564 }
2528 2565 }
2529 2566
2530 2567 if (nvl != retrynvl && !nvlist_empty(retrynvl)) {
2531 2568 nvl = retrynvl;
2532 2569 goto retry;
2533 2570 }
2534 2571
2535 2572 if (!nvlist_empty(genericnvl) &&
2536 2573 dsl_props_set(dsname, source, genericnvl) != 0) {
2537 2574 /*
2538 2575 * If this fails, we still want to set as many properties as we
2539 2576 * can, so try setting them individually.
2540 2577 */
2541 2578 pair = NULL;
2542 2579 while ((pair = nvlist_next_nvpair(genericnvl, pair)) != NULL) {
2543 2580 const char *propname = nvpair_name(pair);
2544 2581 int err = 0;
2545 2582
2546 2583 propval = pair;
2547 2584 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2548 2585 nvlist_t *attrs;
2549 2586 attrs = fnvpair_value_nvlist(pair);
2550 2587 propval = fnvlist_lookup_nvpair(attrs,
2551 2588 ZPROP_VALUE);
2552 2589 }
2553 2590
2554 2591 if (nvpair_type(propval) == DATA_TYPE_STRING) {
2555 2592 strval = fnvpair_value_string(propval);
2556 2593 err = dsl_prop_set_string(dsname, propname,
2557 2594 source, strval);
2558 2595 } else {
2559 2596 intval = fnvpair_value_uint64(propval);
2560 2597 err = dsl_prop_set_int(dsname, propname, source,
2561 2598 intval);
2562 2599 }
2563 2600
2564 2601 if (err != 0) {
2565 2602 if (errlist != NULL) {
2566 2603 fnvlist_add_int32(errlist, propname,
2567 2604 err);
2568 2605 }
2569 2606 rv = err;
2570 2607 }
2571 2608 }
2572 2609 }
2573 2610 nvlist_free(genericnvl);
2574 2611 nvlist_free(retrynvl);
2575 2612
2576 2613 return (rv);
2577 2614 }
2578 2615
2579 2616 /*
2580 2617 * Check that all the properties are valid user properties.
2581 2618 */
2582 2619 static int
2583 2620 zfs_check_userprops(const char *fsname, nvlist_t *nvl)
2584 2621 {
2585 2622 nvpair_t *pair = NULL;
2586 2623 int error = 0;
2587 2624
2588 2625 while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) {
2589 2626 const char *propname = nvpair_name(pair);
2590 2627 char *valstr;
2591 2628
2592 2629 if (!zfs_prop_user(propname) ||
2593 2630 nvpair_type(pair) != DATA_TYPE_STRING)
2594 2631 return (SET_ERROR(EINVAL));
2595 2632
2596 2633 if (error = zfs_secpolicy_write_perms(fsname,
2597 2634 ZFS_DELEG_PERM_USERPROP, CRED()))
2598 2635 return (error);
2599 2636
2600 2637 if (strlen(propname) >= ZAP_MAXNAMELEN)
2601 2638 return (SET_ERROR(ENAMETOOLONG));
2602 2639
2603 2640 VERIFY(nvpair_value_string(pair, &valstr) == 0);
2604 2641 if (strlen(valstr) >= ZAP_MAXVALUELEN)
2605 2642 return (E2BIG);
2606 2643 }
2607 2644 return (0);
2608 2645 }
2609 2646
2610 2647 static void
2611 2648 props_skip(nvlist_t *props, nvlist_t *skipped, nvlist_t **newprops)
2612 2649 {
2613 2650 nvpair_t *pair;
2614 2651
2615 2652 VERIFY(nvlist_alloc(newprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2616 2653
2617 2654 pair = NULL;
2618 2655 while ((pair = nvlist_next_nvpair(props, pair)) != NULL) {
2619 2656 if (nvlist_exists(skipped, nvpair_name(pair)))
2620 2657 continue;
2621 2658
2622 2659 VERIFY(nvlist_add_nvpair(*newprops, pair) == 0);
2623 2660 }
2624 2661 }
2625 2662
2626 2663 static int
2627 2664 clear_received_props(const char *dsname, nvlist_t *props,
2628 2665 nvlist_t *skipped)
2629 2666 {
2630 2667 int err = 0;
2631 2668 nvlist_t *cleared_props = NULL;
2632 2669 props_skip(props, skipped, &cleared_props);
2633 2670 if (!nvlist_empty(cleared_props)) {
2634 2671 /*
2635 2672 * Acts on local properties until the dataset has received
2636 2673 * properties at least once on or after SPA_VERSION_RECVD_PROPS.
2637 2674 */
2638 2675 zprop_source_t flags = (ZPROP_SRC_NONE |
2639 2676 (dsl_prop_get_hasrecvd(dsname) ? ZPROP_SRC_RECEIVED : 0));
2640 2677 err = zfs_set_prop_nvlist(dsname, flags, cleared_props, NULL);
2641 2678 }
2642 2679 nvlist_free(cleared_props);
2643 2680 return (err);
2644 2681 }
2645 2682
2646 2683 /*
2647 2684 * inputs:
2648 2685 * zc_name name of filesystem
2649 2686 * zc_value name of property to set
2650 2687 * zc_nvlist_src{_size} nvlist of properties to apply
2651 2688 * zc_cookie received properties flag
2652 2689 *
2653 2690 * outputs:
2654 2691 * zc_nvlist_dst{_size} error for each unapplied received property
2655 2692 */
2656 2693 static int
2657 2694 zfs_ioc_set_prop(zfs_cmd_t *zc)
2658 2695 {
2659 2696 nvlist_t *nvl;
2660 2697 boolean_t received = zc->zc_cookie;
2661 2698 zprop_source_t source = (received ? ZPROP_SRC_RECEIVED :
2662 2699 ZPROP_SRC_LOCAL);
2663 2700 nvlist_t *errors;
2664 2701 int error;
2665 2702
2666 2703 if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2667 2704 zc->zc_iflags, &nvl)) != 0)
2668 2705 return (error);
2669 2706
2670 2707 if (received) {
2671 2708 nvlist_t *origprops;
2672 2709
2673 2710 if (dsl_prop_get_received(zc->zc_name, &origprops) == 0) {
2674 2711 (void) clear_received_props(zc->zc_name,
2675 2712 origprops, nvl);
2676 2713 nvlist_free(origprops);
2677 2714 }
2678 2715
2679 2716 error = dsl_prop_set_hasrecvd(zc->zc_name);
2680 2717 }
2681 2718
2682 2719 errors = fnvlist_alloc();
2683 2720 if (error == 0)
2684 2721 error = zfs_set_prop_nvlist(zc->zc_name, source, nvl, errors);
2685 2722
2686 2723 if (zc->zc_nvlist_dst != NULL && errors != NULL) {
2687 2724 (void) put_nvlist(zc, errors);
2688 2725 }
2689 2726
2690 2727 nvlist_free(errors);
2691 2728 nvlist_free(nvl);
2692 2729 return (error);
2693 2730 }
2694 2731
2695 2732 /*
2696 2733 * inputs:
2697 2734 * zc_name name of filesystem
2698 2735 * zc_value name of property to inherit
2699 2736 * zc_cookie revert to received value if TRUE
2700 2737 *
2701 2738 * outputs: none
2702 2739 */
2703 2740 static int
2704 2741 zfs_ioc_inherit_prop(zfs_cmd_t *zc)
2705 2742 {
2706 2743 const char *propname = zc->zc_value;
2707 2744 zfs_prop_t prop = zfs_name_to_prop(propname);
2708 2745 boolean_t received = zc->zc_cookie;
2709 2746 zprop_source_t source = (received
2710 2747 ? ZPROP_SRC_NONE /* revert to received value, if any */
2711 2748 : ZPROP_SRC_INHERITED); /* explicitly inherit */
2712 2749
2713 2750 if (received) {
2714 2751 nvlist_t *dummy;
2715 2752 nvpair_t *pair;
2716 2753 zprop_type_t type;
2717 2754 int err;
2718 2755
2719 2756 /*
2720 2757 * zfs_prop_set_special() expects properties in the form of an
2721 2758 * nvpair with type info.
2722 2759 */
2723 2760 if (prop == ZPROP_INVAL) {
2724 2761 if (!zfs_prop_user(propname))
2725 2762 return (SET_ERROR(EINVAL));
2726 2763
2727 2764 type = PROP_TYPE_STRING;
2728 2765 } else if (prop == ZFS_PROP_VOLSIZE ||
2729 2766 prop == ZFS_PROP_VERSION) {
2730 2767 return (SET_ERROR(EINVAL));
2731 2768 } else {
2732 2769 type = zfs_prop_get_type(prop);
2733 2770 }
2734 2771
2735 2772 VERIFY(nvlist_alloc(&dummy, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2736 2773
2737 2774 switch (type) {
2738 2775 case PROP_TYPE_STRING:
2739 2776 VERIFY(0 == nvlist_add_string(dummy, propname, ""));
2740 2777 break;
2741 2778 case PROP_TYPE_NUMBER:
2742 2779 case PROP_TYPE_INDEX:
2743 2780 VERIFY(0 == nvlist_add_uint64(dummy, propname, 0));
2744 2781 break;
2745 2782 default:
2746 2783 nvlist_free(dummy);
2747 2784 return (SET_ERROR(EINVAL));
2748 2785 }
2749 2786
2750 2787 pair = nvlist_next_nvpair(dummy, NULL);
2751 2788 err = zfs_prop_set_special(zc->zc_name, source, pair);
2752 2789 nvlist_free(dummy);
2753 2790 if (err != -1)
2754 2791 return (err); /* special property already handled */
2755 2792 } else {
2756 2793 /*
2757 2794 * Only check this in the non-received case. We want to allow
2758 2795 * 'inherit -S' to revert non-inheritable properties like quota
2759 2796 * and reservation to the received or default values even though
2760 2797 * they are not considered inheritable.
2761 2798 */
2762 2799 if (prop != ZPROP_INVAL && !zfs_prop_inheritable(prop))
2763 2800 return (SET_ERROR(EINVAL));
2764 2801 }
2765 2802
2766 2803 /* property name has been validated by zfs_secpolicy_inherit_prop() */
2767 2804 return (dsl_prop_inherit(zc->zc_name, zc->zc_value, source));
2768 2805 }
2769 2806
2770 2807 static int
2771 2808 zfs_ioc_pool_set_props(zfs_cmd_t *zc)
2772 2809 {
2773 2810 nvlist_t *props;
2774 2811 spa_t *spa;
2775 2812 int error;
2776 2813 nvpair_t *pair;
2777 2814
2778 2815 if (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2779 2816 zc->zc_iflags, &props))
2780 2817 return (error);
2781 2818
2782 2819 /*
2783 2820 * If the only property is the configfile, then just do a spa_lookup()
2784 2821 * to handle the faulted case.
2785 2822 */
2786 2823 pair = nvlist_next_nvpair(props, NULL);
2787 2824 if (pair != NULL && strcmp(nvpair_name(pair),
2788 2825 zpool_prop_to_name(ZPOOL_PROP_CACHEFILE)) == 0 &&
2789 2826 nvlist_next_nvpair(props, pair) == NULL) {
2790 2827 mutex_enter(&spa_namespace_lock);
2791 2828 if ((spa = spa_lookup(zc->zc_name)) != NULL) {
2792 2829 spa_configfile_set(spa, props, B_FALSE);
2793 2830 spa_config_sync(spa, B_FALSE, B_TRUE);
2794 2831 }
2795 2832 mutex_exit(&spa_namespace_lock);
2796 2833 if (spa != NULL) {
2797 2834 nvlist_free(props);
2798 2835 return (0);
2799 2836 }
2800 2837 }
2801 2838
2802 2839 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) {
2803 2840 nvlist_free(props);
2804 2841 return (error);
2805 2842 }
2806 2843
2807 2844 error = spa_prop_set(spa, props);
2808 2845
2809 2846 nvlist_free(props);
2810 2847 spa_close(spa, FTAG);
2811 2848
2812 2849 return (error);
2813 2850 }
2814 2851
2815 2852 static int
2816 2853 zfs_ioc_pool_get_props(zfs_cmd_t *zc)
2817 2854 {
2818 2855 spa_t *spa;
2819 2856 int error;
2820 2857 nvlist_t *nvp = NULL;
2821 2858
2822 2859 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) {
2823 2860 /*
2824 2861 * If the pool is faulted, there may be properties we can still
2825 2862 * get (such as altroot and cachefile), so attempt to get them
2826 2863 * anyway.
2827 2864 */
2828 2865 mutex_enter(&spa_namespace_lock);
2829 2866 if ((spa = spa_lookup(zc->zc_name)) != NULL)
2830 2867 error = spa_prop_get(spa, &nvp);
2831 2868 mutex_exit(&spa_namespace_lock);
2832 2869 } else {
2833 2870 error = spa_prop_get(spa, &nvp);
2834 2871 spa_close(spa, FTAG);
2835 2872 }
2836 2873
2837 2874 if (error == 0 && zc->zc_nvlist_dst != NULL)
2838 2875 error = put_nvlist(zc, nvp);
2839 2876 else
2840 2877 error = SET_ERROR(EFAULT);
2841 2878
2842 2879 nvlist_free(nvp);
2843 2880 return (error);
2844 2881 }
2845 2882
2846 2883 /*
2847 2884 * inputs:
2848 2885 * zc_name name of filesystem
2849 2886 * zc_nvlist_src{_size} nvlist of delegated permissions
2850 2887 * zc_perm_action allow/unallow flag
2851 2888 *
2852 2889 * outputs: none
2853 2890 */
2854 2891 static int
2855 2892 zfs_ioc_set_fsacl(zfs_cmd_t *zc)
2856 2893 {
2857 2894 int error;
2858 2895 nvlist_t *fsaclnv = NULL;
2859 2896
2860 2897 if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2861 2898 zc->zc_iflags, &fsaclnv)) != 0)
2862 2899 return (error);
2863 2900
2864 2901 /*
2865 2902 * Verify nvlist is constructed correctly
2866 2903 */
2867 2904 if ((error = zfs_deleg_verify_nvlist(fsaclnv)) != 0) {
2868 2905 nvlist_free(fsaclnv);
2869 2906 return (SET_ERROR(EINVAL));
2870 2907 }
2871 2908
2872 2909 /*
2873 2910 * If we don't have PRIV_SYS_MOUNT, then validate
2874 2911 * that user is allowed to hand out each permission in
2875 2912 * the nvlist(s)
2876 2913 */
2877 2914
2878 2915 error = secpolicy_zfs(CRED());
2879 2916 if (error != 0) {
2880 2917 if (zc->zc_perm_action == B_FALSE) {
2881 2918 error = dsl_deleg_can_allow(zc->zc_name,
2882 2919 fsaclnv, CRED());
2883 2920 } else {
2884 2921 error = dsl_deleg_can_unallow(zc->zc_name,
2885 2922 fsaclnv, CRED());
2886 2923 }
2887 2924 }
2888 2925
2889 2926 if (error == 0)
2890 2927 error = dsl_deleg_set(zc->zc_name, fsaclnv, zc->zc_perm_action);
2891 2928
2892 2929 nvlist_free(fsaclnv);
2893 2930 return (error);
2894 2931 }
2895 2932
2896 2933 /*
2897 2934 * inputs:
2898 2935 * zc_name name of filesystem
2899 2936 *
2900 2937 * outputs:
2901 2938 * zc_nvlist_src{_size} nvlist of delegated permissions
2902 2939 */
2903 2940 static int
2904 2941 zfs_ioc_get_fsacl(zfs_cmd_t *zc)
2905 2942 {
2906 2943 nvlist_t *nvp;
2907 2944 int error;
2908 2945
2909 2946 if ((error = dsl_deleg_get(zc->zc_name, &nvp)) == 0) {
2910 2947 error = put_nvlist(zc, nvp);
2911 2948 nvlist_free(nvp);
2912 2949 }
2913 2950
2914 2951 return (error);
2915 2952 }
2916 2953
2917 2954 /*
2918 2955 * Search the vfs list for a specified resource. Returns a pointer to it
2919 2956 * or NULL if no suitable entry is found. The caller of this routine
2920 2957 * is responsible for releasing the returned vfs pointer.
2921 2958 */
2922 2959 static vfs_t *
2923 2960 zfs_get_vfs(const char *resource)
2924 2961 {
2925 2962 struct vfs *vfsp;
2926 2963 struct vfs *vfs_found = NULL;
2927 2964
2928 2965 vfs_list_read_lock();
2929 2966 vfsp = rootvfs;
2930 2967 do {
2931 2968 if (strcmp(refstr_value(vfsp->vfs_resource), resource) == 0) {
2932 2969 VFS_HOLD(vfsp);
2933 2970 vfs_found = vfsp;
2934 2971 break;
2935 2972 }
2936 2973 vfsp = vfsp->vfs_next;
2937 2974 } while (vfsp != rootvfs);
2938 2975 vfs_list_unlock();
2939 2976 return (vfs_found);
2940 2977 }
2941 2978
2942 2979 /* ARGSUSED */
2943 2980 static void
2944 2981 zfs_create_cb(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx)
2945 2982 {
2946 2983 zfs_creat_t *zct = arg;
2947 2984
2948 2985 zfs_create_fs(os, cr, zct->zct_zplprops, tx);
2949 2986 }
2950 2987
2951 2988 #define ZFS_PROP_UNDEFINED ((uint64_t)-1)
2952 2989
2953 2990 /*
2954 2991 * inputs:
2955 2992 * os parent objset pointer (NULL if root fs)
2956 2993 * fuids_ok fuids allowed in this version of the spa?
2957 2994 * sa_ok SAs allowed in this version of the spa?
2958 2995 * createprops list of properties requested by creator
2959 2996 *
2960 2997 * outputs:
2961 2998 * zplprops values for the zplprops we attach to the master node object
2962 2999 * is_ci true if requested file system will be purely case-insensitive
2963 3000 *
2964 3001 * Determine the settings for utf8only, normalization and
2965 3002 * casesensitivity. Specific values may have been requested by the
2966 3003 * creator and/or we can inherit values from the parent dataset. If
2967 3004 * the file system is of too early a vintage, a creator can not
2968 3005 * request settings for these properties, even if the requested
2969 3006 * setting is the default value. We don't actually want to create dsl
2970 3007 * properties for these, so remove them from the source nvlist after
2971 3008 * processing.
2972 3009 */
2973 3010 static int
2974 3011 zfs_fill_zplprops_impl(objset_t *os, uint64_t zplver,
2975 3012 boolean_t fuids_ok, boolean_t sa_ok, nvlist_t *createprops,
2976 3013 nvlist_t *zplprops, boolean_t *is_ci)
2977 3014 {
2978 3015 uint64_t sense = ZFS_PROP_UNDEFINED;
2979 3016 uint64_t norm = ZFS_PROP_UNDEFINED;
2980 3017 uint64_t u8 = ZFS_PROP_UNDEFINED;
2981 3018
2982 3019 ASSERT(zplprops != NULL);
2983 3020
2984 3021 /*
2985 3022 * Pull out creator prop choices, if any.
2986 3023 */
2987 3024 if (createprops) {
2988 3025 (void) nvlist_lookup_uint64(createprops,
2989 3026 zfs_prop_to_name(ZFS_PROP_VERSION), &zplver);
2990 3027 (void) nvlist_lookup_uint64(createprops,
2991 3028 zfs_prop_to_name(ZFS_PROP_NORMALIZE), &norm);
2992 3029 (void) nvlist_remove_all(createprops,
2993 3030 zfs_prop_to_name(ZFS_PROP_NORMALIZE));
2994 3031 (void) nvlist_lookup_uint64(createprops,
2995 3032 zfs_prop_to_name(ZFS_PROP_UTF8ONLY), &u8);
2996 3033 (void) nvlist_remove_all(createprops,
2997 3034 zfs_prop_to_name(ZFS_PROP_UTF8ONLY));
2998 3035 (void) nvlist_lookup_uint64(createprops,
2999 3036 zfs_prop_to_name(ZFS_PROP_CASE), &sense);
3000 3037 (void) nvlist_remove_all(createprops,
3001 3038 zfs_prop_to_name(ZFS_PROP_CASE));
3002 3039 }
3003 3040
3004 3041 /*
3005 3042 * If the zpl version requested is whacky or the file system
3006 3043 * or pool is version is too "young" to support normalization
3007 3044 * and the creator tried to set a value for one of the props,
3008 3045 * error out.
3009 3046 */
3010 3047 if ((zplver < ZPL_VERSION_INITIAL || zplver > ZPL_VERSION) ||
3011 3048 (zplver >= ZPL_VERSION_FUID && !fuids_ok) ||
3012 3049 (zplver >= ZPL_VERSION_SA && !sa_ok) ||
3013 3050 (zplver < ZPL_VERSION_NORMALIZATION &&
3014 3051 (norm != ZFS_PROP_UNDEFINED || u8 != ZFS_PROP_UNDEFINED ||
3015 3052 sense != ZFS_PROP_UNDEFINED)))
3016 3053 return (SET_ERROR(ENOTSUP));
3017 3054
3018 3055 /*
3019 3056 * Put the version in the zplprops
3020 3057 */
3021 3058 VERIFY(nvlist_add_uint64(zplprops,
3022 3059 zfs_prop_to_name(ZFS_PROP_VERSION), zplver) == 0);
3023 3060
3024 3061 if (norm == ZFS_PROP_UNDEFINED)
3025 3062 VERIFY(zfs_get_zplprop(os, ZFS_PROP_NORMALIZE, &norm) == 0);
3026 3063 VERIFY(nvlist_add_uint64(zplprops,
3027 3064 zfs_prop_to_name(ZFS_PROP_NORMALIZE), norm) == 0);
3028 3065
3029 3066 /*
3030 3067 * If we're normalizing, names must always be valid UTF-8 strings.
3031 3068 */
3032 3069 if (norm)
3033 3070 u8 = 1;
3034 3071 if (u8 == ZFS_PROP_UNDEFINED)
3035 3072 VERIFY(zfs_get_zplprop(os, ZFS_PROP_UTF8ONLY, &u8) == 0);
3036 3073 VERIFY(nvlist_add_uint64(zplprops,
3037 3074 zfs_prop_to_name(ZFS_PROP_UTF8ONLY), u8) == 0);
3038 3075
3039 3076 if (sense == ZFS_PROP_UNDEFINED)
3040 3077 VERIFY(zfs_get_zplprop(os, ZFS_PROP_CASE, &sense) == 0);
3041 3078 VERIFY(nvlist_add_uint64(zplprops,
3042 3079 zfs_prop_to_name(ZFS_PROP_CASE), sense) == 0);
3043 3080
3044 3081 if (is_ci)
3045 3082 *is_ci = (sense == ZFS_CASE_INSENSITIVE);
3046 3083
3047 3084 return (0);
3048 3085 }
3049 3086
3050 3087 static int
3051 3088 zfs_fill_zplprops(const char *dataset, nvlist_t *createprops,
3052 3089 nvlist_t *zplprops, boolean_t *is_ci)
3053 3090 {
3054 3091 boolean_t fuids_ok, sa_ok;
3055 3092 uint64_t zplver = ZPL_VERSION;
3056 3093 objset_t *os = NULL;
3057 3094 char parentname[MAXNAMELEN];
3058 3095 char *cp;
3059 3096 spa_t *spa;
3060 3097 uint64_t spa_vers;
3061 3098 int error;
3062 3099
3063 3100 (void) strlcpy(parentname, dataset, sizeof (parentname));
3064 3101 cp = strrchr(parentname, '/');
3065 3102 ASSERT(cp != NULL);
3066 3103 cp[0] = '\0';
3067 3104
3068 3105 if ((error = spa_open(dataset, &spa, FTAG)) != 0)
3069 3106 return (error);
3070 3107
3071 3108 spa_vers = spa_version(spa);
3072 3109 spa_close(spa, FTAG);
3073 3110
3074 3111 zplver = zfs_zpl_version_map(spa_vers);
3075 3112 fuids_ok = (zplver >= ZPL_VERSION_FUID);
3076 3113 sa_ok = (zplver >= ZPL_VERSION_SA);
3077 3114
3078 3115 /*
3079 3116 * Open parent object set so we can inherit zplprop values.
3080 3117 */
3081 3118 if ((error = dmu_objset_hold(parentname, FTAG, &os)) != 0)
3082 3119 return (error);
3083 3120
3084 3121 error = zfs_fill_zplprops_impl(os, zplver, fuids_ok, sa_ok, createprops,
3085 3122 zplprops, is_ci);
3086 3123 dmu_objset_rele(os, FTAG);
3087 3124 return (error);
3088 3125 }
3089 3126
3090 3127 static int
3091 3128 zfs_fill_zplprops_root(uint64_t spa_vers, nvlist_t *createprops,
3092 3129 nvlist_t *zplprops, boolean_t *is_ci)
3093 3130 {
3094 3131 boolean_t fuids_ok;
3095 3132 boolean_t sa_ok;
3096 3133 uint64_t zplver = ZPL_VERSION;
3097 3134 int error;
3098 3135
3099 3136 zplver = zfs_zpl_version_map(spa_vers);
3100 3137 fuids_ok = (zplver >= ZPL_VERSION_FUID);
3101 3138 sa_ok = (zplver >= ZPL_VERSION_SA);
3102 3139
3103 3140 error = zfs_fill_zplprops_impl(NULL, zplver, fuids_ok, sa_ok,
3104 3141 createprops, zplprops, is_ci);
3105 3142 return (error);
3106 3143 }
3107 3144
3108 3145 /*
3109 3146 * innvl: {
3110 3147 * "type" -> dmu_objset_type_t (int32)
3111 3148 * (optional) "props" -> { prop -> value }
3112 3149 * }
3113 3150 *
3114 3151 * outnvl: propname -> error code (int32)
3115 3152 */
3116 3153 static int
3117 3154 zfs_ioc_create(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3118 3155 {
3119 3156 int error = 0;
3120 3157 zfs_creat_t zct = { 0 };
3121 3158 nvlist_t *nvprops = NULL;
3122 3159 void (*cbfunc)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx);
3123 3160 int32_t type32;
3124 3161 dmu_objset_type_t type;
3125 3162 boolean_t is_insensitive = B_FALSE;
3126 3163
3127 3164 if (nvlist_lookup_int32(innvl, "type", &type32) != 0)
3128 3165 return (SET_ERROR(EINVAL));
3129 3166 type = type32;
3130 3167 (void) nvlist_lookup_nvlist(innvl, "props", &nvprops);
3131 3168
3132 3169 switch (type) {
3133 3170 case DMU_OST_ZFS:
3134 3171 cbfunc = zfs_create_cb;
3135 3172 break;
3136 3173
3137 3174 case DMU_OST_ZVOL:
3138 3175 cbfunc = zvol_create_cb;
3139 3176 break;
3140 3177
3141 3178 default:
3142 3179 cbfunc = NULL;
3143 3180 break;
3144 3181 }
3145 3182 if (strchr(fsname, '@') ||
3146 3183 strchr(fsname, '%'))
3147 3184 return (SET_ERROR(EINVAL));
3148 3185
3149 3186 zct.zct_props = nvprops;
3150 3187
3151 3188 if (cbfunc == NULL)
3152 3189 return (SET_ERROR(EINVAL));
3153 3190
3154 3191 if (type == DMU_OST_ZVOL) {
3155 3192 uint64_t volsize, volblocksize;
3156 3193
3157 3194 if (nvprops == NULL)
3158 3195 return (SET_ERROR(EINVAL));
3159 3196 if (nvlist_lookup_uint64(nvprops,
3160 3197 zfs_prop_to_name(ZFS_PROP_VOLSIZE), &volsize) != 0)
3161 3198 return (SET_ERROR(EINVAL));
3162 3199
3163 3200 if ((error = nvlist_lookup_uint64(nvprops,
3164 3201 zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE),
3165 3202 &volblocksize)) != 0 && error != ENOENT)
3166 3203 return (SET_ERROR(EINVAL));
3167 3204
3168 3205 if (error != 0)
3169 3206 volblocksize = zfs_prop_default_numeric(
3170 3207 ZFS_PROP_VOLBLOCKSIZE);
3171 3208
3172 3209 if ((error = zvol_check_volblocksize(
3173 3210 volblocksize)) != 0 ||
3174 3211 (error = zvol_check_volsize(volsize,
3175 3212 volblocksize)) != 0)
3176 3213 return (error);
3177 3214 } else if (type == DMU_OST_ZFS) {
3178 3215 int error;
3179 3216
3180 3217 /*
3181 3218 * We have to have normalization and
3182 3219 * case-folding flags correct when we do the
3183 3220 * file system creation, so go figure them out
3184 3221 * now.
3185 3222 */
3186 3223 VERIFY(nvlist_alloc(&zct.zct_zplprops,
3187 3224 NV_UNIQUE_NAME, KM_SLEEP) == 0);
3188 3225 error = zfs_fill_zplprops(fsname, nvprops,
3189 3226 zct.zct_zplprops, &is_insensitive);
3190 3227 if (error != 0) {
3191 3228 nvlist_free(zct.zct_zplprops);
3192 3229 return (error);
3193 3230 }
3194 3231 }
3195 3232
3196 3233 error = dmu_objset_create(fsname, type,
3197 3234 is_insensitive ? DS_FLAG_CI_DATASET : 0, cbfunc, &zct);
3198 3235 nvlist_free(zct.zct_zplprops);
3199 3236
3200 3237 /*
3201 3238 * It would be nice to do this atomically.
3202 3239 */
3203 3240 if (error == 0) {
3204 3241 error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL,
3205 3242 nvprops, outnvl);
3206 3243 if (error != 0)
3207 3244 (void) dsl_destroy_head(fsname);
3208 3245 }
3209 3246 return (error);
3210 3247 }
3211 3248
3212 3249 /*
3213 3250 * innvl: {
3214 3251 * "origin" -> name of origin snapshot
3215 3252 * (optional) "props" -> { prop -> value }
3216 3253 * }
3217 3254 *
3218 3255 * outnvl: propname -> error code (int32)
3219 3256 */
3220 3257 static int
3221 3258 zfs_ioc_clone(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3222 3259 {
3223 3260 int error = 0;
3224 3261 nvlist_t *nvprops = NULL;
3225 3262 char *origin_name;
3226 3263
3227 3264 if (nvlist_lookup_string(innvl, "origin", &origin_name) != 0)
3228 3265 return (SET_ERROR(EINVAL));
3229 3266 (void) nvlist_lookup_nvlist(innvl, "props", &nvprops);
3230 3267
3231 3268 if (strchr(fsname, '@') ||
3232 3269 strchr(fsname, '%'))
3233 3270 return (SET_ERROR(EINVAL));
3234 3271
3235 3272 if (dataset_namecheck(origin_name, NULL, NULL) != 0)
3236 3273 return (SET_ERROR(EINVAL));
3237 3274 error = dmu_objset_clone(fsname, origin_name);
3238 3275 if (error != 0)
3239 3276 return (error);
3240 3277
3241 3278 /*
3242 3279 * It would be nice to do this atomically.
3243 3280 */
3244 3281 if (error == 0) {
3245 3282 error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL,
3246 3283 nvprops, outnvl);
3247 3284 if (error != 0)
3248 3285 (void) dsl_destroy_head(fsname);
3249 3286 }
3250 3287 return (error);
3251 3288 }
3252 3289
3253 3290 /*
3254 3291 * innvl: {
3255 3292 * "snaps" -> { snapshot1, snapshot2 }
3256 3293 * (optional) "props" -> { prop -> value (string) }
3257 3294 * }
3258 3295 *
3259 3296 * outnvl: snapshot -> error code (int32)
3260 3297 */
3261 3298 static int
3262 3299 zfs_ioc_snapshot(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3263 3300 {
3264 3301 nvlist_t *snaps;
3265 3302 nvlist_t *props = NULL;
3266 3303 int error, poollen;
3267 3304 nvpair_t *pair;
3268 3305
3269 3306 (void) nvlist_lookup_nvlist(innvl, "props", &props);
3270 3307 if ((error = zfs_check_userprops(poolname, props)) != 0)
3271 3308 return (error);
3272 3309
3273 3310 if (!nvlist_empty(props) &&
3274 3311 zfs_earlier_version(poolname, SPA_VERSION_SNAP_PROPS))
3275 3312 return (SET_ERROR(ENOTSUP));
3276 3313
3277 3314 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
3278 3315 return (SET_ERROR(EINVAL));
3279 3316 poollen = strlen(poolname);
3280 3317 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
3281 3318 pair = nvlist_next_nvpair(snaps, pair)) {
3282 3319 const char *name = nvpair_name(pair);
3283 3320 const char *cp = strchr(name, '@');
3284 3321
3285 3322 /*
3286 3323 * The snap name must contain an @, and the part after it must
3287 3324 * contain only valid characters.
3288 3325 */
3289 3326 if (cp == NULL || snapshot_namecheck(cp + 1, NULL, NULL) != 0)
3290 3327 return (SET_ERROR(EINVAL));
3291 3328
3292 3329 /*
3293 3330 * The snap must be in the specified pool.
3294 3331 */
3295 3332 if (strncmp(name, poolname, poollen) != 0 ||
3296 3333 (name[poollen] != '/' && name[poollen] != '@'))
3297 3334 return (SET_ERROR(EXDEV));
3298 3335
3299 3336 /* This must be the only snap of this fs. */
3300 3337 for (nvpair_t *pair2 = nvlist_next_nvpair(snaps, pair);
3301 3338 pair2 != NULL; pair2 = nvlist_next_nvpair(snaps, pair2)) {
3302 3339 if (strncmp(name, nvpair_name(pair2), cp - name + 1)
3303 3340 == 0) {
3304 3341 return (SET_ERROR(EXDEV));
3305 3342 }
3306 3343 }
3307 3344 }
3308 3345
3309 3346 error = dsl_dataset_snapshot(snaps, props, outnvl);
3310 3347 return (error);
3311 3348 }
3312 3349
3313 3350 /*
3314 3351 * innvl: "message" -> string
3315 3352 */
3316 3353 /* ARGSUSED */
3317 3354 static int
3318 3355 zfs_ioc_log_history(const char *unused, nvlist_t *innvl, nvlist_t *outnvl)
3319 3356 {
3320 3357 char *message;
3321 3358 spa_t *spa;
3322 3359 int error;
3323 3360 char *poolname;
3324 3361
3325 3362 /*
3326 3363 * The poolname in the ioctl is not set, we get it from the TSD,
3327 3364 * which was set at the end of the last successful ioctl that allows
3328 3365 * logging. The secpolicy func already checked that it is set.
3329 3366 * Only one log ioctl is allowed after each successful ioctl, so
3330 3367 * we clear the TSD here.
3331 3368 */
3332 3369 poolname = tsd_get(zfs_allow_log_key);
3333 3370 (void) tsd_set(zfs_allow_log_key, NULL);
3334 3371 error = spa_open(poolname, &spa, FTAG);
3335 3372 strfree(poolname);
3336 3373 if (error != 0)
3337 3374 return (error);
3338 3375
3339 3376 if (nvlist_lookup_string(innvl, "message", &message) != 0) {
3340 3377 spa_close(spa, FTAG);
3341 3378 return (SET_ERROR(EINVAL));
3342 3379 }
3343 3380
3344 3381 if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) {
3345 3382 spa_close(spa, FTAG);
3346 3383 return (SET_ERROR(ENOTSUP));
3347 3384 }
3348 3385
3349 3386 error = spa_history_log(spa, message);
3350 3387 spa_close(spa, FTAG);
3351 3388 return (error);
3352 3389 }
3353 3390
3354 3391 /*
3355 3392 * The dp_config_rwlock must not be held when calling this, because the
3356 3393 * unmount may need to write out data.
3357 3394 *
3358 3395 * This function is best-effort. Callers must deal gracefully if it
3359 3396 * remains mounted (or is remounted after this call).
3360 3397 *
3361 3398 * Returns 0 if the argument is not a snapshot, or it is not currently a
3362 3399 * filesystem, or we were able to unmount it. Returns error code otherwise.
3363 3400 */
3364 3401 int
3365 3402 zfs_unmount_snap(const char *snapname)
3366 3403 {
3367 3404 vfs_t *vfsp;
3368 3405 zfsvfs_t *zfsvfs;
3369 3406 int err;
3370 3407
3371 3408 if (strchr(snapname, '@') == NULL)
3372 3409 return (0);
3373 3410
3374 3411 vfsp = zfs_get_vfs(snapname);
3375 3412 if (vfsp == NULL)
3376 3413 return (0);
3377 3414
3378 3415 zfsvfs = vfsp->vfs_data;
3379 3416 ASSERT(!dsl_pool_config_held(dmu_objset_pool(zfsvfs->z_os)));
3380 3417
3381 3418 err = vn_vfswlock(vfsp->vfs_vnodecovered);
3382 3419 VFS_RELE(vfsp);
3383 3420 if (err != 0)
3384 3421 return (SET_ERROR(err));
3385 3422
3386 3423 /*
3387 3424 * Always force the unmount for snapshots.
3388 3425 */
3389 3426 (void) dounmount(vfsp, MS_FORCE, kcred);
3390 3427 return (0);
3391 3428 }
3392 3429
3393 3430 /* ARGSUSED */
3394 3431 static int
3395 3432 zfs_unmount_snap_cb(const char *snapname, void *arg)
3396 3433 {
3397 3434 return (zfs_unmount_snap(snapname));
3398 3435 }
3399 3436
3400 3437 /*
3401 3438 * When a clone is destroyed, its origin may also need to be destroyed,
3402 3439 * in which case it must be unmounted. This routine will do that unmount
3403 3440 * if necessary.
3404 3441 */
3405 3442 void
3406 3443 zfs_destroy_unmount_origin(const char *fsname)
3407 3444 {
3408 3445 int error;
3409 3446 objset_t *os;
3410 3447 dsl_dataset_t *ds;
3411 3448
3412 3449 error = dmu_objset_hold(fsname, FTAG, &os);
3413 3450 if (error != 0)
3414 3451 return;
3415 3452 ds = dmu_objset_ds(os);
3416 3453 if (dsl_dir_is_clone(ds->ds_dir) && DS_IS_DEFER_DESTROY(ds->ds_prev)) {
3417 3454 char originname[MAXNAMELEN];
3418 3455 dsl_dataset_name(ds->ds_prev, originname);
3419 3456 dmu_objset_rele(os, FTAG);
3420 3457 (void) zfs_unmount_snap(originname);
3421 3458 } else {
3422 3459 dmu_objset_rele(os, FTAG);
3423 3460 }
3424 3461 }
3425 3462
3426 3463 /*
3427 3464 * innvl: {
3428 3465 * "snaps" -> { snapshot1, snapshot2 }
3429 3466 * (optional boolean) "defer"
3430 3467 * }
3431 3468 *
3432 3469 * outnvl: snapshot -> error code (int32)
3433 3470 *
3434 3471 */
3435 3472 static int
3436 3473 zfs_ioc_destroy_snaps(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3437 3474 {
3438 3475 int error, poollen;
3439 3476 nvlist_t *snaps;
3440 3477 nvpair_t *pair;
3441 3478 boolean_t defer;
3442 3479
3443 3480 if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
3444 3481 return (SET_ERROR(EINVAL));
3445 3482 defer = nvlist_exists(innvl, "defer");
3446 3483
3447 3484 poollen = strlen(poolname);
3448 3485 for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
3449 3486 pair = nvlist_next_nvpair(snaps, pair)) {
3450 3487 const char *name = nvpair_name(pair);
3451 3488
3452 3489 /*
3453 3490 * The snap must be in the specified pool.
3454 3491 */
3455 3492 if (strncmp(name, poolname, poollen) != 0 ||
3456 3493 (name[poollen] != '/' && name[poollen] != '@'))
3457 3494 return (SET_ERROR(EXDEV));
3458 3495
3459 3496 error = zfs_unmount_snap(name);
3460 3497 if (error != 0)
3461 3498 return (error);
3462 3499 }
3463 3500
3464 3501 return (dsl_destroy_snapshots_nvl(snaps, defer, outnvl));
3465 3502 }
3466 3503
3467 3504 /*
3468 3505 * inputs:
3469 3506 * zc_name name of dataset to destroy
3470 3507 * zc_objset_type type of objset
3471 3508 * zc_defer_destroy mark for deferred destroy
3472 3509 *
3473 3510 * outputs: none
3474 3511 */
3475 3512 static int
3476 3513 zfs_ioc_destroy(zfs_cmd_t *zc)
3477 3514 {
3478 3515 int err;
3479 3516
3480 3517 if (zc->zc_objset_type == DMU_OST_ZFS) {
3481 3518 err = zfs_unmount_snap(zc->zc_name);
3482 3519 if (err != 0)
3483 3520 return (err);
3484 3521 }
3485 3522
3486 3523 if (strchr(zc->zc_name, '@'))
3487 3524 err = dsl_destroy_snapshot(zc->zc_name, zc->zc_defer_destroy);
3488 3525 else
3489 3526 err = dsl_destroy_head(zc->zc_name);
3490 3527 if (zc->zc_objset_type == DMU_OST_ZVOL && err == 0)
3491 3528 (void) zvol_remove_minor(zc->zc_name);
3492 3529 return (err);
3493 3530 }
3494 3531
3495 3532 /*
3496 3533 * fsname is name of dataset to rollback (to most recent snapshot)
3497 3534 *
3498 3535 * innvl is not used.
3499 3536 *
3500 3537 * outnvl: "target" -> name of most recent snapshot
3501 3538 * }
3502 3539 */
3503 3540 /* ARGSUSED */
3504 3541 static int
3505 3542 zfs_ioc_rollback(const char *fsname, nvlist_t *args, nvlist_t *outnvl)
3506 3543 {
3507 3544 zfsvfs_t *zfsvfs;
3508 3545 int error;
3509 3546
3510 3547 if (getzfsvfs(fsname, &zfsvfs) == 0) {
3511 3548 error = zfs_suspend_fs(zfsvfs);
3512 3549 if (error == 0) {
3513 3550 int resume_err;
3514 3551
3515 3552 error = dsl_dataset_rollback(fsname, zfsvfs, outnvl);
3516 3553 resume_err = zfs_resume_fs(zfsvfs, fsname);
3517 3554 error = error ? error : resume_err;
3518 3555 }
3519 3556 VFS_RELE(zfsvfs->z_vfs);
3520 3557 } else {
3521 3558 error = dsl_dataset_rollback(fsname, NULL, outnvl);
3522 3559 }
3523 3560 return (error);
3524 3561 }
3525 3562
3526 3563 static int
3527 3564 recursive_unmount(const char *fsname, void *arg)
3528 3565 {
3529 3566 const char *snapname = arg;
3530 3567 char fullname[MAXNAMELEN];
3531 3568
3532 3569 (void) snprintf(fullname, sizeof (fullname), "%s@%s", fsname, snapname);
3533 3570 return (zfs_unmount_snap(fullname));
3534 3571 }
3535 3572
3536 3573 /*
3537 3574 * inputs:
3538 3575 * zc_name old name of dataset
3539 3576 * zc_value new name of dataset
3540 3577 * zc_cookie recursive flag (only valid for snapshots)
3541 3578 *
3542 3579 * outputs: none
3543 3580 */
3544 3581 static int
3545 3582 zfs_ioc_rename(zfs_cmd_t *zc)
3546 3583 {
3547 3584 boolean_t recursive = zc->zc_cookie & 1;
3548 3585 char *at;
3549 3586
3550 3587 zc->zc_value[sizeof (zc->zc_value) - 1] = '\0';
3551 3588 if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
3552 3589 strchr(zc->zc_value, '%'))
3553 3590 return (SET_ERROR(EINVAL));
3554 3591
3555 3592 at = strchr(zc->zc_name, '@');
3556 3593 if (at != NULL) {
3557 3594 /* snaps must be in same fs */
3558 3595 int error;
3559 3596
3560 3597 if (strncmp(zc->zc_name, zc->zc_value, at - zc->zc_name + 1))
3561 3598 return (SET_ERROR(EXDEV));
3562 3599 *at = '\0';
3563 3600 if (zc->zc_objset_type == DMU_OST_ZFS) {
3564 3601 error = dmu_objset_find(zc->zc_name,
3565 3602 recursive_unmount, at + 1,
3566 3603 recursive ? DS_FIND_CHILDREN : 0);
3567 3604 if (error != 0) {
3568 3605 *at = '@';
3569 3606 return (error);
3570 3607 }
3571 3608 }
3572 3609 error = dsl_dataset_rename_snapshot(zc->zc_name,
3573 3610 at + 1, strchr(zc->zc_value, '@') + 1, recursive);
3574 3611 *at = '@';
3575 3612
3576 3613 return (error);
3577 3614 } else {
3578 3615 if (zc->zc_objset_type == DMU_OST_ZVOL)
3579 3616 (void) zvol_remove_minor(zc->zc_name);
3580 3617 return (dsl_dir_rename(zc->zc_name, zc->zc_value));
3581 3618 }
3582 3619 }
3583 3620
3584 3621 static int
3585 3622 zfs_check_settable(const char *dsname, nvpair_t *pair, cred_t *cr)
3586 3623 {
3587 3624 const char *propname = nvpair_name(pair);
3588 3625 boolean_t issnap = (strchr(dsname, '@') != NULL);
3589 3626 zfs_prop_t prop = zfs_name_to_prop(propname);
3590 3627 uint64_t intval;
3591 3628 int err;
3592 3629
3593 3630 if (prop == ZPROP_INVAL) {
3594 3631 if (zfs_prop_user(propname)) {
3595 3632 if (err = zfs_secpolicy_write_perms(dsname,
3596 3633 ZFS_DELEG_PERM_USERPROP, cr))
3597 3634 return (err);
3598 3635 return (0);
3599 3636 }
3600 3637
3601 3638 if (!issnap && zfs_prop_userquota(propname)) {
3602 3639 const char *perm = NULL;
3603 3640 const char *uq_prefix =
3604 3641 zfs_userquota_prop_prefixes[ZFS_PROP_USERQUOTA];
3605 3642 const char *gq_prefix =
3606 3643 zfs_userquota_prop_prefixes[ZFS_PROP_GROUPQUOTA];
3607 3644
3608 3645 if (strncmp(propname, uq_prefix,
3609 3646 strlen(uq_prefix)) == 0) {
3610 3647 perm = ZFS_DELEG_PERM_USERQUOTA;
3611 3648 } else if (strncmp(propname, gq_prefix,
3612 3649 strlen(gq_prefix)) == 0) {
3613 3650 perm = ZFS_DELEG_PERM_GROUPQUOTA;
3614 3651 } else {
3615 3652 /* USERUSED and GROUPUSED are read-only */
3616 3653 return (SET_ERROR(EINVAL));
3617 3654 }
3618 3655
3619 3656 if (err = zfs_secpolicy_write_perms(dsname, perm, cr))
3620 3657 return (err);
3621 3658 return (0);
3622 3659 }
3623 3660
3624 3661 return (SET_ERROR(EINVAL));
3625 3662 }
3626 3663
3627 3664 if (issnap)
3628 3665 return (SET_ERROR(EINVAL));
3629 3666
3630 3667 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
3631 3668 /*
3632 3669 * dsl_prop_get_all_impl() returns properties in this
3633 3670 * format.
3634 3671 */
3635 3672 nvlist_t *attrs;
3636 3673 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
3637 3674 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
3638 3675 &pair) == 0);
3639 3676 }
3640 3677
3641 3678 /*
3642 3679 * Check that this value is valid for this pool version
3643 3680 */
3644 3681 switch (prop) {
3645 3682 case ZFS_PROP_COMPRESSION:
3646 3683 /*
3647 3684 * If the user specified gzip compression, make sure
3648 3685 * the SPA supports it. We ignore any errors here since
3649 3686 * we'll catch them later.
3650 3687 */
3651 3688 if (nvpair_type(pair) == DATA_TYPE_UINT64 &&
3652 3689 nvpair_value_uint64(pair, &intval) == 0) {
3653 3690 if (intval >= ZIO_COMPRESS_GZIP_1 &&
3654 3691 intval <= ZIO_COMPRESS_GZIP_9 &&
3655 3692 zfs_earlier_version(dsname,
3656 3693 SPA_VERSION_GZIP_COMPRESSION)) {
3657 3694 return (SET_ERROR(ENOTSUP));
3658 3695 }
3659 3696
3660 3697 if (intval == ZIO_COMPRESS_ZLE &&
3661 3698 zfs_earlier_version(dsname,
3662 3699 SPA_VERSION_ZLE_COMPRESSION))
3663 3700 return (SET_ERROR(ENOTSUP));
3664 3701
3665 3702 if (intval == ZIO_COMPRESS_LZ4) {
3666 3703 zfeature_info_t *feature =
3667 3704 &spa_feature_table[
3668 3705 SPA_FEATURE_LZ4_COMPRESS];
3669 3706 spa_t *spa;
3670 3707
3671 3708 if ((err = spa_open(dsname, &spa, FTAG)) != 0)
3672 3709 return (err);
3673 3710
3674 3711 if (!spa_feature_is_enabled(spa, feature)) {
3675 3712 spa_close(spa, FTAG);
3676 3713 return (SET_ERROR(ENOTSUP));
3677 3714 }
3678 3715 spa_close(spa, FTAG);
3679 3716 }
3680 3717
3681 3718 /*
3682 3719 * If this is a bootable dataset then
3683 3720 * verify that the compression algorithm
3684 3721 * is supported for booting. We must return
3685 3722 * something other than ENOTSUP since it
3686 3723 * implies a downrev pool version.
3687 3724 */
3688 3725 if (zfs_is_bootfs(dsname) &&
3689 3726 !BOOTFS_COMPRESS_VALID(intval)) {
|
↓ open down ↓ |
1266 lines elided |
↑ open up ↑ |
3690 3727 return (SET_ERROR(ERANGE));
3691 3728 }
3692 3729 }
3693 3730 break;
3694 3731
3695 3732 case ZFS_PROP_COPIES:
3696 3733 if (zfs_earlier_version(dsname, SPA_VERSION_DITTO_BLOCKS))
3697 3734 return (SET_ERROR(ENOTSUP));
3698 3735 break;
3699 3736
3700 - case ZFS_PROP_DEDUP:
3701 - if (zfs_earlier_version(dsname, SPA_VERSION_DEDUP))
3702 - return (SET_ERROR(ENOTSUP));
3703 - break;
3704 -
3705 3737 case ZFS_PROP_SHARESMB:
3706 3738 if (zpl_earlier_version(dsname, ZPL_VERSION_FUID))
3707 3739 return (SET_ERROR(ENOTSUP));
3708 3740 break;
3709 3741
3710 3742 case ZFS_PROP_ACLINHERIT:
3711 3743 if (nvpair_type(pair) == DATA_TYPE_UINT64 &&
3712 3744 nvpair_value_uint64(pair, &intval) == 0) {
3713 3745 if (intval == ZFS_ACL_PASSTHROUGH_X &&
3714 3746 zfs_earlier_version(dsname,
3715 3747 SPA_VERSION_PASSTHROUGH_X))
3716 3748 return (SET_ERROR(ENOTSUP));
3717 3749 }
3718 3750 break;
3751 +
3752 + case ZFS_PROP_CHECKSUM:
3753 + case ZFS_PROP_DEDUP: {
3754 + zfeature_info_t *feature = NULL;
3755 +
3756 + /* dedup feature version checks */
3757 + if (prop == ZFS_PROP_DEDUP &&
3758 + zfs_earlier_version(dsname, SPA_VERSION_DEDUP))
3759 + return (SET_ERROR(ENOTSUP));
3760 +
3761 + if (nvpair_value_uint64(pair, &intval) != 0)
3762 + return (SET_ERROR(EINVAL));
3763 +
3764 + /* check prop value is enabled in features */
3765 + if (intval == ZIO_CHECKSUM_SHA512)
3766 + feature = &spa_feature_table[SPA_FEATURE_SHA512];
3767 + else if (intval == ZIO_CHECKSUM_SKEIN)
3768 + feature = &spa_feature_table[SPA_FEATURE_SKEIN];
3769 + else if (intval == ZIO_CHECKSUM_EDONR)
3770 + feature = &spa_feature_table[SPA_FEATURE_EDONR];
3771 +
3772 + if (feature != NULL) {
3773 + spa_t *spa;
3774 +
3775 + if ((err = spa_open(dsname, &spa, FTAG)) != 0)
3776 + return (err);
3777 + /*
3778 + * Salted checksums are not supported on root pools.
3779 + */
3780 + if (spa_bootfs(spa) != 0 &&
3781 + intval < ZIO_CHECKSUM_FUNCTIONS &&
3782 + zio_checksum_table[intval].ci_salted != 0) {
3783 + spa_close(spa, FTAG);
3784 + return (SET_ERROR(ERANGE));
3785 + }
3786 + if (!spa_feature_is_enabled(spa, feature)) {
3787 + spa_close(spa, FTAG);
3788 + return (SET_ERROR(ENOTSUP));
3789 + }
3790 + spa_close(spa, FTAG);
3791 + }
3792 + break;
3719 3793 }
3794 + }
3720 3795
3721 3796 return (zfs_secpolicy_setprop(dsname, prop, pair, CRED()));
3722 3797 }
3723 3798
3724 3799 /*
3725 3800 * Checks for a race condition to make sure we don't increment a feature flag
3726 3801 * multiple times.
3727 3802 */
3728 3803 static int
3729 3804 zfs_prop_activate_feature_check(void *arg, dmu_tx_t *tx)
3730 3805 {
3731 3806 spa_t *spa = dmu_tx_pool(tx)->dp_spa;
3732 3807 zfeature_info_t *feature = arg;
3733 3808
3734 3809 if (!spa_feature_is_active(spa, feature))
3735 3810 return (0);
3736 3811 else
3737 3812 return (SET_ERROR(EBUSY));
3738 3813 }
3739 3814
3740 3815 /*
3741 3816 * The callback invoked on feature activation in the sync task caused by
3742 3817 * zfs_prop_activate_feature.
3743 3818 */
3744 3819 static void
3745 3820 zfs_prop_activate_feature_sync(void *arg, dmu_tx_t *tx)
3746 3821 {
3747 3822 spa_t *spa = dmu_tx_pool(tx)->dp_spa;
3748 3823 zfeature_info_t *feature = arg;
3749 3824
3750 3825 spa_feature_incr(spa, feature, tx);
3751 3826 }
3752 3827
3753 3828 /*
3754 3829 * Activates a feature on a pool in response to a property setting. This
3755 3830 * creates a new sync task which modifies the pool to reflect the feature
3756 3831 * as being active.
3757 3832 */
3758 3833 static int
3759 3834 zfs_prop_activate_feature(spa_t *spa, zfeature_info_t *feature)
3760 3835 {
3761 3836 int err;
3762 3837
3763 3838 /* EBUSY here indicates that the feature is already active */
3764 3839 err = dsl_sync_task(spa_name(spa),
3765 3840 zfs_prop_activate_feature_check, zfs_prop_activate_feature_sync,
3766 3841 feature, 2);
3767 3842
3768 3843 if (err != 0 && err != EBUSY)
3769 3844 return (err);
3770 3845 else
3771 3846 return (0);
3772 3847 }
3773 3848
3774 3849 /*
3775 3850 * Removes properties from the given props list that fail permission checks
3776 3851 * needed to clear them and to restore them in case of a receive error. For each
3777 3852 * property, make sure we have both set and inherit permissions.
3778 3853 *
3779 3854 * Returns the first error encountered if any permission checks fail. If the
3780 3855 * caller provides a non-NULL errlist, it also gives the complete list of names
3781 3856 * of all the properties that failed a permission check along with the
3782 3857 * corresponding error numbers. The caller is responsible for freeing the
3783 3858 * returned errlist.
3784 3859 *
3785 3860 * If every property checks out successfully, zero is returned and the list
3786 3861 * pointed at by errlist is NULL.
3787 3862 */
3788 3863 static int
3789 3864 zfs_check_clearable(char *dataset, nvlist_t *props, nvlist_t **errlist)
3790 3865 {
3791 3866 zfs_cmd_t *zc;
3792 3867 nvpair_t *pair, *next_pair;
3793 3868 nvlist_t *errors;
3794 3869 int err, rv = 0;
3795 3870
3796 3871 if (props == NULL)
3797 3872 return (0);
3798 3873
3799 3874 VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0);
3800 3875
3801 3876 zc = kmem_alloc(sizeof (zfs_cmd_t), KM_SLEEP);
3802 3877 (void) strcpy(zc->zc_name, dataset);
3803 3878 pair = nvlist_next_nvpair(props, NULL);
3804 3879 while (pair != NULL) {
3805 3880 next_pair = nvlist_next_nvpair(props, pair);
3806 3881
3807 3882 (void) strcpy(zc->zc_value, nvpair_name(pair));
3808 3883 if ((err = zfs_check_settable(dataset, pair, CRED())) != 0 ||
3809 3884 (err = zfs_secpolicy_inherit_prop(zc, NULL, CRED())) != 0) {
3810 3885 VERIFY(nvlist_remove_nvpair(props, pair) == 0);
3811 3886 VERIFY(nvlist_add_int32(errors,
3812 3887 zc->zc_value, err) == 0);
3813 3888 }
3814 3889 pair = next_pair;
3815 3890 }
3816 3891 kmem_free(zc, sizeof (zfs_cmd_t));
3817 3892
3818 3893 if ((pair = nvlist_next_nvpair(errors, NULL)) == NULL) {
3819 3894 nvlist_free(errors);
3820 3895 errors = NULL;
3821 3896 } else {
3822 3897 VERIFY(nvpair_value_int32(pair, &rv) == 0);
3823 3898 }
3824 3899
3825 3900 if (errlist == NULL)
3826 3901 nvlist_free(errors);
3827 3902 else
3828 3903 *errlist = errors;
3829 3904
3830 3905 return (rv);
3831 3906 }
3832 3907
3833 3908 static boolean_t
3834 3909 propval_equals(nvpair_t *p1, nvpair_t *p2)
3835 3910 {
3836 3911 if (nvpair_type(p1) == DATA_TYPE_NVLIST) {
3837 3912 /* dsl_prop_get_all_impl() format */
3838 3913 nvlist_t *attrs;
3839 3914 VERIFY(nvpair_value_nvlist(p1, &attrs) == 0);
3840 3915 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
3841 3916 &p1) == 0);
3842 3917 }
3843 3918
3844 3919 if (nvpair_type(p2) == DATA_TYPE_NVLIST) {
3845 3920 nvlist_t *attrs;
3846 3921 VERIFY(nvpair_value_nvlist(p2, &attrs) == 0);
3847 3922 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
3848 3923 &p2) == 0);
3849 3924 }
3850 3925
3851 3926 if (nvpair_type(p1) != nvpair_type(p2))
3852 3927 return (B_FALSE);
3853 3928
3854 3929 if (nvpair_type(p1) == DATA_TYPE_STRING) {
3855 3930 char *valstr1, *valstr2;
3856 3931
3857 3932 VERIFY(nvpair_value_string(p1, (char **)&valstr1) == 0);
3858 3933 VERIFY(nvpair_value_string(p2, (char **)&valstr2) == 0);
3859 3934 return (strcmp(valstr1, valstr2) == 0);
3860 3935 } else {
3861 3936 uint64_t intval1, intval2;
3862 3937
3863 3938 VERIFY(nvpair_value_uint64(p1, &intval1) == 0);
3864 3939 VERIFY(nvpair_value_uint64(p2, &intval2) == 0);
3865 3940 return (intval1 == intval2);
3866 3941 }
3867 3942 }
3868 3943
3869 3944 /*
3870 3945 * Remove properties from props if they are not going to change (as determined
3871 3946 * by comparison with origprops). Remove them from origprops as well, since we
3872 3947 * do not need to clear or restore properties that won't change.
3873 3948 */
3874 3949 static void
3875 3950 props_reduce(nvlist_t *props, nvlist_t *origprops)
3876 3951 {
3877 3952 nvpair_t *pair, *next_pair;
3878 3953
3879 3954 if (origprops == NULL)
3880 3955 return; /* all props need to be received */
3881 3956
3882 3957 pair = nvlist_next_nvpair(props, NULL);
3883 3958 while (pair != NULL) {
3884 3959 const char *propname = nvpair_name(pair);
3885 3960 nvpair_t *match;
3886 3961
3887 3962 next_pair = nvlist_next_nvpair(props, pair);
3888 3963
3889 3964 if ((nvlist_lookup_nvpair(origprops, propname,
3890 3965 &match) != 0) || !propval_equals(pair, match))
3891 3966 goto next; /* need to set received value */
3892 3967
3893 3968 /* don't clear the existing received value */
3894 3969 (void) nvlist_remove_nvpair(origprops, match);
3895 3970 /* don't bother receiving the property */
3896 3971 (void) nvlist_remove_nvpair(props, pair);
3897 3972 next:
3898 3973 pair = next_pair;
3899 3974 }
3900 3975 }
3901 3976
3902 3977 #ifdef DEBUG
3903 3978 static boolean_t zfs_ioc_recv_inject_err;
3904 3979 #endif
3905 3980
3906 3981 /*
3907 3982 * inputs:
3908 3983 * zc_name name of containing filesystem
3909 3984 * zc_nvlist_src{_size} nvlist of properties to apply
3910 3985 * zc_value name of snapshot to create
3911 3986 * zc_string name of clone origin (if DRR_FLAG_CLONE)
3912 3987 * zc_cookie file descriptor to recv from
3913 3988 * zc_begin_record the BEGIN record of the stream (not byteswapped)
3914 3989 * zc_guid force flag
3915 3990 * zc_cleanup_fd cleanup-on-exit file descriptor
3916 3991 * zc_action_handle handle for this guid/ds mapping (or zero on first call)
3917 3992 *
3918 3993 * outputs:
3919 3994 * zc_cookie number of bytes read
3920 3995 * zc_nvlist_dst{_size} error for each unapplied received property
3921 3996 * zc_obj zprop_errflags_t
3922 3997 * zc_action_handle handle for this guid/ds mapping
3923 3998 */
3924 3999 static int
3925 4000 zfs_ioc_recv(zfs_cmd_t *zc)
3926 4001 {
3927 4002 file_t *fp;
3928 4003 dmu_recv_cookie_t drc;
3929 4004 boolean_t force = (boolean_t)zc->zc_guid;
3930 4005 int fd;
3931 4006 int error = 0;
3932 4007 int props_error = 0;
3933 4008 nvlist_t *errors;
3934 4009 offset_t off;
3935 4010 nvlist_t *props = NULL; /* sent properties */
3936 4011 nvlist_t *origprops = NULL; /* existing properties */
3937 4012 char *origin = NULL;
3938 4013 char *tosnap;
3939 4014 char tofs[ZFS_MAXNAMELEN];
3940 4015 boolean_t first_recvd_props = B_FALSE;
3941 4016
3942 4017 if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
3943 4018 strchr(zc->zc_value, '@') == NULL ||
3944 4019 strchr(zc->zc_value, '%'))
3945 4020 return (SET_ERROR(EINVAL));
3946 4021
3947 4022 (void) strcpy(tofs, zc->zc_value);
3948 4023 tosnap = strchr(tofs, '@');
3949 4024 *tosnap++ = '\0';
3950 4025
3951 4026 if (zc->zc_nvlist_src != NULL &&
3952 4027 (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
3953 4028 zc->zc_iflags, &props)) != 0)
3954 4029 return (error);
3955 4030
3956 4031 fd = zc->zc_cookie;
3957 4032 fp = getf(fd);
3958 4033 if (fp == NULL) {
3959 4034 nvlist_free(props);
3960 4035 return (SET_ERROR(EBADF));
3961 4036 }
3962 4037
3963 4038 VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0);
3964 4039
3965 4040 if (zc->zc_string[0])
3966 4041 origin = zc->zc_string;
3967 4042
3968 4043 error = dmu_recv_begin(tofs, tosnap,
3969 4044 &zc->zc_begin_record, force, origin, &drc);
3970 4045 if (error != 0)
3971 4046 goto out;
3972 4047
3973 4048 /*
3974 4049 * Set properties before we receive the stream so that they are applied
3975 4050 * to the new data. Note that we must call dmu_recv_stream() if
3976 4051 * dmu_recv_begin() succeeds.
3977 4052 */
3978 4053 if (props != NULL && !drc.drc_newfs) {
3979 4054 if (spa_version(dsl_dataset_get_spa(drc.drc_ds)) >=
3980 4055 SPA_VERSION_RECVD_PROPS &&
3981 4056 !dsl_prop_get_hasrecvd(tofs))
3982 4057 first_recvd_props = B_TRUE;
3983 4058
3984 4059 /*
3985 4060 * If new received properties are supplied, they are to
3986 4061 * completely replace the existing received properties, so stash
3987 4062 * away the existing ones.
3988 4063 */
3989 4064 if (dsl_prop_get_received(tofs, &origprops) == 0) {
3990 4065 nvlist_t *errlist = NULL;
3991 4066 /*
3992 4067 * Don't bother writing a property if its value won't
3993 4068 * change (and avoid the unnecessary security checks).
3994 4069 *
3995 4070 * The first receive after SPA_VERSION_RECVD_PROPS is a
3996 4071 * special case where we blow away all local properties
3997 4072 * regardless.
3998 4073 */
3999 4074 if (!first_recvd_props)
4000 4075 props_reduce(props, origprops);
4001 4076 if (zfs_check_clearable(tofs, origprops, &errlist) != 0)
4002 4077 (void) nvlist_merge(errors, errlist, 0);
4003 4078 nvlist_free(errlist);
4004 4079
4005 4080 if (clear_received_props(tofs, origprops,
4006 4081 first_recvd_props ? NULL : props) != 0)
4007 4082 zc->zc_obj |= ZPROP_ERR_NOCLEAR;
4008 4083 } else {
4009 4084 zc->zc_obj |= ZPROP_ERR_NOCLEAR;
4010 4085 }
4011 4086 }
4012 4087
4013 4088 if (props != NULL) {
4014 4089 props_error = dsl_prop_set_hasrecvd(tofs);
4015 4090
4016 4091 if (props_error == 0) {
4017 4092 (void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_RECEIVED,
4018 4093 props, errors);
4019 4094 }
4020 4095 }
4021 4096
4022 4097 if (zc->zc_nvlist_dst_size != 0 &&
4023 4098 (nvlist_smush(errors, zc->zc_nvlist_dst_size) != 0 ||
4024 4099 put_nvlist(zc, errors) != 0)) {
4025 4100 /*
4026 4101 * Caller made zc->zc_nvlist_dst less than the minimum expected
4027 4102 * size or supplied an invalid address.
4028 4103 */
4029 4104 props_error = SET_ERROR(EINVAL);
4030 4105 }
4031 4106
4032 4107 off = fp->f_offset;
4033 4108 error = dmu_recv_stream(&drc, fp->f_vnode, &off, zc->zc_cleanup_fd,
4034 4109 &zc->zc_action_handle);
4035 4110
4036 4111 if (error == 0) {
4037 4112 zfsvfs_t *zfsvfs = NULL;
4038 4113
4039 4114 if (getzfsvfs(tofs, &zfsvfs) == 0) {
4040 4115 /* online recv */
4041 4116 int end_err;
4042 4117
4043 4118 error = zfs_suspend_fs(zfsvfs);
4044 4119 /*
4045 4120 * If the suspend fails, then the recv_end will
4046 4121 * likely also fail, and clean up after itself.
4047 4122 */
4048 4123 end_err = dmu_recv_end(&drc, zfsvfs);
4049 4124 if (error == 0)
4050 4125 error = zfs_resume_fs(zfsvfs, tofs);
4051 4126 error = error ? error : end_err;
4052 4127 VFS_RELE(zfsvfs->z_vfs);
4053 4128 } else {
4054 4129 error = dmu_recv_end(&drc, NULL);
4055 4130 }
4056 4131 }
4057 4132
4058 4133 zc->zc_cookie = off - fp->f_offset;
4059 4134 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
4060 4135 fp->f_offset = off;
4061 4136
4062 4137 #ifdef DEBUG
4063 4138 if (zfs_ioc_recv_inject_err) {
4064 4139 zfs_ioc_recv_inject_err = B_FALSE;
4065 4140 error = 1;
4066 4141 }
4067 4142 #endif
4068 4143 /*
4069 4144 * On error, restore the original props.
4070 4145 */
4071 4146 if (error != 0 && props != NULL && !drc.drc_newfs) {
4072 4147 if (clear_received_props(tofs, props, NULL) != 0) {
4073 4148 /*
4074 4149 * We failed to clear the received properties.
4075 4150 * Since we may have left a $recvd value on the
4076 4151 * system, we can't clear the $hasrecvd flag.
4077 4152 */
4078 4153 zc->zc_obj |= ZPROP_ERR_NORESTORE;
4079 4154 } else if (first_recvd_props) {
4080 4155 dsl_prop_unset_hasrecvd(tofs);
4081 4156 }
4082 4157
4083 4158 if (origprops == NULL && !drc.drc_newfs) {
4084 4159 /* We failed to stash the original properties. */
4085 4160 zc->zc_obj |= ZPROP_ERR_NORESTORE;
4086 4161 }
4087 4162
4088 4163 /*
4089 4164 * dsl_props_set() will not convert RECEIVED to LOCAL on or
4090 4165 * after SPA_VERSION_RECVD_PROPS, so we need to specify LOCAL
4091 4166 * explictly if we're restoring local properties cleared in the
4092 4167 * first new-style receive.
4093 4168 */
4094 4169 if (origprops != NULL &&
4095 4170 zfs_set_prop_nvlist(tofs, (first_recvd_props ?
4096 4171 ZPROP_SRC_LOCAL : ZPROP_SRC_RECEIVED),
4097 4172 origprops, NULL) != 0) {
4098 4173 /*
4099 4174 * We stashed the original properties but failed to
4100 4175 * restore them.
4101 4176 */
4102 4177 zc->zc_obj |= ZPROP_ERR_NORESTORE;
4103 4178 }
4104 4179 }
4105 4180 out:
4106 4181 nvlist_free(props);
4107 4182 nvlist_free(origprops);
4108 4183 nvlist_free(errors);
4109 4184 releasef(fd);
4110 4185
4111 4186 if (error == 0)
4112 4187 error = props_error;
4113 4188
4114 4189 return (error);
4115 4190 }
4116 4191
4117 4192 /*
4118 4193 * inputs:
4119 4194 * zc_name name of snapshot to send
4120 4195 * zc_cookie file descriptor to send stream to
4121 4196 * zc_obj fromorigin flag (mutually exclusive with zc_fromobj)
4122 4197 * zc_sendobj objsetid of snapshot to send
4123 4198 * zc_fromobj objsetid of incremental fromsnap (may be zero)
4124 4199 * zc_guid if set, estimate size of stream only. zc_cookie is ignored.
4125 4200 * output size in zc_objset_type.
4126 4201 *
4127 4202 * outputs: none
4128 4203 */
4129 4204 static int
4130 4205 zfs_ioc_send(zfs_cmd_t *zc)
4131 4206 {
4132 4207 int error;
4133 4208 offset_t off;
4134 4209 boolean_t estimate = (zc->zc_guid != 0);
4135 4210
4136 4211 if (zc->zc_obj != 0) {
4137 4212 dsl_pool_t *dp;
4138 4213 dsl_dataset_t *tosnap;
4139 4214
4140 4215 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
4141 4216 if (error != 0)
4142 4217 return (error);
4143 4218
4144 4219 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &tosnap);
4145 4220 if (error != 0) {
4146 4221 dsl_pool_rele(dp, FTAG);
4147 4222 return (error);
4148 4223 }
4149 4224
4150 4225 if (dsl_dir_is_clone(tosnap->ds_dir))
4151 4226 zc->zc_fromobj = tosnap->ds_dir->dd_phys->dd_origin_obj;
4152 4227 dsl_dataset_rele(tosnap, FTAG);
4153 4228 dsl_pool_rele(dp, FTAG);
4154 4229 }
4155 4230
4156 4231 if (estimate) {
4157 4232 dsl_pool_t *dp;
4158 4233 dsl_dataset_t *tosnap;
4159 4234 dsl_dataset_t *fromsnap = NULL;
4160 4235
4161 4236 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
4162 4237 if (error != 0)
4163 4238 return (error);
4164 4239
4165 4240 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &tosnap);
4166 4241 if (error != 0) {
4167 4242 dsl_pool_rele(dp, FTAG);
4168 4243 return (error);
4169 4244 }
4170 4245
4171 4246 if (zc->zc_fromobj != 0) {
4172 4247 error = dsl_dataset_hold_obj(dp, zc->zc_fromobj,
4173 4248 FTAG, &fromsnap);
4174 4249 if (error != 0) {
4175 4250 dsl_dataset_rele(tosnap, FTAG);
4176 4251 dsl_pool_rele(dp, FTAG);
4177 4252 return (error);
4178 4253 }
4179 4254 }
4180 4255
4181 4256 error = dmu_send_estimate(tosnap, fromsnap,
4182 4257 &zc->zc_objset_type);
4183 4258
4184 4259 if (fromsnap != NULL)
4185 4260 dsl_dataset_rele(fromsnap, FTAG);
4186 4261 dsl_dataset_rele(tosnap, FTAG);
4187 4262 dsl_pool_rele(dp, FTAG);
4188 4263 } else {
4189 4264 file_t *fp = getf(zc->zc_cookie);
4190 4265 if (fp == NULL)
4191 4266 return (SET_ERROR(EBADF));
4192 4267
4193 4268 off = fp->f_offset;
4194 4269 error = dmu_send_obj(zc->zc_name, zc->zc_sendobj,
4195 4270 zc->zc_fromobj, zc->zc_cookie, fp->f_vnode, &off);
4196 4271
4197 4272 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
4198 4273 fp->f_offset = off;
4199 4274 releasef(zc->zc_cookie);
4200 4275 }
4201 4276 return (error);
4202 4277 }
4203 4278
4204 4279 /*
4205 4280 * inputs:
4206 4281 * zc_name name of snapshot on which to report progress
4207 4282 * zc_cookie file descriptor of send stream
4208 4283 *
4209 4284 * outputs:
4210 4285 * zc_cookie number of bytes written in send stream thus far
4211 4286 */
4212 4287 static int
4213 4288 zfs_ioc_send_progress(zfs_cmd_t *zc)
4214 4289 {
4215 4290 dsl_pool_t *dp;
4216 4291 dsl_dataset_t *ds;
4217 4292 dmu_sendarg_t *dsp = NULL;
4218 4293 int error;
4219 4294
4220 4295 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
4221 4296 if (error != 0)
4222 4297 return (error);
4223 4298
4224 4299 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &ds);
4225 4300 if (error != 0) {
4226 4301 dsl_pool_rele(dp, FTAG);
4227 4302 return (error);
4228 4303 }
4229 4304
4230 4305 mutex_enter(&ds->ds_sendstream_lock);
4231 4306
4232 4307 /*
4233 4308 * Iterate over all the send streams currently active on this dataset.
4234 4309 * If there's one which matches the specified file descriptor _and_ the
4235 4310 * stream was started by the current process, return the progress of
4236 4311 * that stream.
4237 4312 */
4238 4313 for (dsp = list_head(&ds->ds_sendstreams); dsp != NULL;
4239 4314 dsp = list_next(&ds->ds_sendstreams, dsp)) {
4240 4315 if (dsp->dsa_outfd == zc->zc_cookie &&
4241 4316 dsp->dsa_proc == curproc)
4242 4317 break;
4243 4318 }
4244 4319
4245 4320 if (dsp != NULL)
4246 4321 zc->zc_cookie = *(dsp->dsa_off);
4247 4322 else
4248 4323 error = SET_ERROR(ENOENT);
4249 4324
4250 4325 mutex_exit(&ds->ds_sendstream_lock);
4251 4326 dsl_dataset_rele(ds, FTAG);
4252 4327 dsl_pool_rele(dp, FTAG);
4253 4328 return (error);
4254 4329 }
4255 4330
4256 4331 static int
4257 4332 zfs_ioc_inject_fault(zfs_cmd_t *zc)
4258 4333 {
4259 4334 int id, error;
4260 4335
4261 4336 error = zio_inject_fault(zc->zc_name, (int)zc->zc_guid, &id,
4262 4337 &zc->zc_inject_record);
4263 4338
4264 4339 if (error == 0)
4265 4340 zc->zc_guid = (uint64_t)id;
4266 4341
4267 4342 return (error);
4268 4343 }
4269 4344
4270 4345 static int
4271 4346 zfs_ioc_clear_fault(zfs_cmd_t *zc)
4272 4347 {
4273 4348 return (zio_clear_fault((int)zc->zc_guid));
4274 4349 }
4275 4350
4276 4351 static int
4277 4352 zfs_ioc_inject_list_next(zfs_cmd_t *zc)
4278 4353 {
4279 4354 int id = (int)zc->zc_guid;
4280 4355 int error;
4281 4356
4282 4357 error = zio_inject_list_next(&id, zc->zc_name, sizeof (zc->zc_name),
4283 4358 &zc->zc_inject_record);
4284 4359
4285 4360 zc->zc_guid = id;
4286 4361
4287 4362 return (error);
4288 4363 }
4289 4364
4290 4365 static int
4291 4366 zfs_ioc_error_log(zfs_cmd_t *zc)
4292 4367 {
4293 4368 spa_t *spa;
4294 4369 int error;
4295 4370 size_t count = (size_t)zc->zc_nvlist_dst_size;
4296 4371
4297 4372 if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
4298 4373 return (error);
4299 4374
4300 4375 error = spa_get_errlog(spa, (void *)(uintptr_t)zc->zc_nvlist_dst,
4301 4376 &count);
4302 4377 if (error == 0)
4303 4378 zc->zc_nvlist_dst_size = count;
4304 4379 else
4305 4380 zc->zc_nvlist_dst_size = spa_get_errlog_size(spa);
4306 4381
4307 4382 spa_close(spa, FTAG);
4308 4383
4309 4384 return (error);
4310 4385 }
4311 4386
4312 4387 static int
4313 4388 zfs_ioc_clear(zfs_cmd_t *zc)
4314 4389 {
4315 4390 spa_t *spa;
4316 4391 vdev_t *vd;
4317 4392 int error;
4318 4393
4319 4394 /*
4320 4395 * On zpool clear we also fix up missing slogs
4321 4396 */
4322 4397 mutex_enter(&spa_namespace_lock);
4323 4398 spa = spa_lookup(zc->zc_name);
4324 4399 if (spa == NULL) {
4325 4400 mutex_exit(&spa_namespace_lock);
4326 4401 return (SET_ERROR(EIO));
4327 4402 }
4328 4403 if (spa_get_log_state(spa) == SPA_LOG_MISSING) {
4329 4404 /* we need to let spa_open/spa_load clear the chains */
4330 4405 spa_set_log_state(spa, SPA_LOG_CLEAR);
4331 4406 }
4332 4407 spa->spa_last_open_failed = 0;
4333 4408 mutex_exit(&spa_namespace_lock);
4334 4409
4335 4410 if (zc->zc_cookie & ZPOOL_NO_REWIND) {
4336 4411 error = spa_open(zc->zc_name, &spa, FTAG);
4337 4412 } else {
4338 4413 nvlist_t *policy;
4339 4414 nvlist_t *config = NULL;
4340 4415
4341 4416 if (zc->zc_nvlist_src == NULL)
4342 4417 return (SET_ERROR(EINVAL));
4343 4418
4344 4419 if ((error = get_nvlist(zc->zc_nvlist_src,
4345 4420 zc->zc_nvlist_src_size, zc->zc_iflags, &policy)) == 0) {
4346 4421 error = spa_open_rewind(zc->zc_name, &spa, FTAG,
4347 4422 policy, &config);
4348 4423 if (config != NULL) {
4349 4424 int err;
4350 4425
4351 4426 if ((err = put_nvlist(zc, config)) != 0)
4352 4427 error = err;
4353 4428 nvlist_free(config);
4354 4429 }
4355 4430 nvlist_free(policy);
4356 4431 }
4357 4432 }
4358 4433
4359 4434 if (error != 0)
4360 4435 return (error);
4361 4436
4362 4437 spa_vdev_state_enter(spa, SCL_NONE);
4363 4438
4364 4439 if (zc->zc_guid == 0) {
4365 4440 vd = NULL;
4366 4441 } else {
4367 4442 vd = spa_lookup_by_guid(spa, zc->zc_guid, B_TRUE);
4368 4443 if (vd == NULL) {
4369 4444 (void) spa_vdev_state_exit(spa, NULL, ENODEV);
4370 4445 spa_close(spa, FTAG);
4371 4446 return (SET_ERROR(ENODEV));
4372 4447 }
4373 4448 }
4374 4449
4375 4450 vdev_clear(spa, vd);
4376 4451
4377 4452 (void) spa_vdev_state_exit(spa, NULL, 0);
4378 4453
4379 4454 /*
4380 4455 * Resume any suspended I/Os.
4381 4456 */
4382 4457 if (zio_resume(spa) != 0)
4383 4458 error = SET_ERROR(EIO);
4384 4459
4385 4460 spa_close(spa, FTAG);
4386 4461
4387 4462 return (error);
4388 4463 }
4389 4464
4390 4465 static int
4391 4466 zfs_ioc_pool_reopen(zfs_cmd_t *zc)
4392 4467 {
4393 4468 spa_t *spa;
4394 4469 int error;
4395 4470
4396 4471 error = spa_open(zc->zc_name, &spa, FTAG);
4397 4472 if (error != 0)
4398 4473 return (error);
4399 4474
4400 4475 spa_vdev_state_enter(spa, SCL_NONE);
4401 4476
4402 4477 /*
4403 4478 * If a resilver is already in progress then set the
4404 4479 * spa_scrub_reopen flag to B_TRUE so that we don't restart
4405 4480 * the scan as a side effect of the reopen. Otherwise, let
4406 4481 * vdev_open() decided if a resilver is required.
4407 4482 */
4408 4483 spa->spa_scrub_reopen = dsl_scan_resilvering(spa->spa_dsl_pool);
4409 4484 vdev_reopen(spa->spa_root_vdev);
4410 4485 spa->spa_scrub_reopen = B_FALSE;
4411 4486
4412 4487 (void) spa_vdev_state_exit(spa, NULL, 0);
4413 4488 spa_close(spa, FTAG);
4414 4489 return (0);
4415 4490 }
4416 4491 /*
4417 4492 * inputs:
4418 4493 * zc_name name of filesystem
4419 4494 * zc_value name of origin snapshot
4420 4495 *
4421 4496 * outputs:
4422 4497 * zc_string name of conflicting snapshot, if there is one
4423 4498 */
4424 4499 static int
4425 4500 zfs_ioc_promote(zfs_cmd_t *zc)
4426 4501 {
4427 4502 char *cp;
4428 4503
4429 4504 /*
4430 4505 * We don't need to unmount *all* the origin fs's snapshots, but
4431 4506 * it's easier.
4432 4507 */
4433 4508 cp = strchr(zc->zc_value, '@');
4434 4509 if (cp)
4435 4510 *cp = '\0';
4436 4511 (void) dmu_objset_find(zc->zc_value,
4437 4512 zfs_unmount_snap_cb, NULL, DS_FIND_SNAPSHOTS);
4438 4513 return (dsl_dataset_promote(zc->zc_name, zc->zc_string));
4439 4514 }
4440 4515
4441 4516 /*
4442 4517 * Retrieve a single {user|group}{used|quota}@... property.
4443 4518 *
4444 4519 * inputs:
4445 4520 * zc_name name of filesystem
4446 4521 * zc_objset_type zfs_userquota_prop_t
4447 4522 * zc_value domain name (eg. "S-1-234-567-89")
4448 4523 * zc_guid RID/UID/GID
4449 4524 *
4450 4525 * outputs:
4451 4526 * zc_cookie property value
4452 4527 */
4453 4528 static int
4454 4529 zfs_ioc_userspace_one(zfs_cmd_t *zc)
4455 4530 {
4456 4531 zfsvfs_t *zfsvfs;
4457 4532 int error;
4458 4533
4459 4534 if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
4460 4535 return (SET_ERROR(EINVAL));
4461 4536
4462 4537 error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE);
4463 4538 if (error != 0)
4464 4539 return (error);
4465 4540
4466 4541 error = zfs_userspace_one(zfsvfs,
4467 4542 zc->zc_objset_type, zc->zc_value, zc->zc_guid, &zc->zc_cookie);
4468 4543 zfsvfs_rele(zfsvfs, FTAG);
4469 4544
4470 4545 return (error);
4471 4546 }
4472 4547
4473 4548 /*
4474 4549 * inputs:
4475 4550 * zc_name name of filesystem
4476 4551 * zc_cookie zap cursor
4477 4552 * zc_objset_type zfs_userquota_prop_t
4478 4553 * zc_nvlist_dst[_size] buffer to fill (not really an nvlist)
4479 4554 *
4480 4555 * outputs:
4481 4556 * zc_nvlist_dst[_size] data buffer (array of zfs_useracct_t)
4482 4557 * zc_cookie zap cursor
4483 4558 */
4484 4559 static int
4485 4560 zfs_ioc_userspace_many(zfs_cmd_t *zc)
4486 4561 {
4487 4562 zfsvfs_t *zfsvfs;
4488 4563 int bufsize = zc->zc_nvlist_dst_size;
4489 4564
4490 4565 if (bufsize <= 0)
4491 4566 return (SET_ERROR(ENOMEM));
4492 4567
4493 4568 int error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE);
4494 4569 if (error != 0)
4495 4570 return (error);
4496 4571
4497 4572 void *buf = kmem_alloc(bufsize, KM_SLEEP);
4498 4573
4499 4574 error = zfs_userspace_many(zfsvfs, zc->zc_objset_type, &zc->zc_cookie,
4500 4575 buf, &zc->zc_nvlist_dst_size);
4501 4576
4502 4577 if (error == 0) {
4503 4578 error = xcopyout(buf,
4504 4579 (void *)(uintptr_t)zc->zc_nvlist_dst,
4505 4580 zc->zc_nvlist_dst_size);
4506 4581 }
4507 4582 kmem_free(buf, bufsize);
4508 4583 zfsvfs_rele(zfsvfs, FTAG);
4509 4584
4510 4585 return (error);
4511 4586 }
4512 4587
4513 4588 /*
4514 4589 * inputs:
4515 4590 * zc_name name of filesystem
4516 4591 *
4517 4592 * outputs:
4518 4593 * none
4519 4594 */
4520 4595 static int
4521 4596 zfs_ioc_userspace_upgrade(zfs_cmd_t *zc)
4522 4597 {
4523 4598 objset_t *os;
4524 4599 int error = 0;
4525 4600 zfsvfs_t *zfsvfs;
4526 4601
4527 4602 if (getzfsvfs(zc->zc_name, &zfsvfs) == 0) {
4528 4603 if (!dmu_objset_userused_enabled(zfsvfs->z_os)) {
4529 4604 /*
4530 4605 * If userused is not enabled, it may be because the
4531 4606 * objset needs to be closed & reopened (to grow the
4532 4607 * objset_phys_t). Suspend/resume the fs will do that.
4533 4608 */
4534 4609 error = zfs_suspend_fs(zfsvfs);
4535 4610 if (error == 0) {
4536 4611 dmu_objset_refresh_ownership(zfsvfs->z_os,
4537 4612 zfsvfs);
4538 4613 error = zfs_resume_fs(zfsvfs, zc->zc_name);
4539 4614 }
4540 4615 }
4541 4616 if (error == 0)
4542 4617 error = dmu_objset_userspace_upgrade(zfsvfs->z_os);
4543 4618 VFS_RELE(zfsvfs->z_vfs);
4544 4619 } else {
4545 4620 /* XXX kind of reading contents without owning */
4546 4621 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
4547 4622 if (error != 0)
4548 4623 return (error);
4549 4624
4550 4625 error = dmu_objset_userspace_upgrade(os);
4551 4626 dmu_objset_rele(os, FTAG);
4552 4627 }
4553 4628
4554 4629 return (error);
4555 4630 }
4556 4631
4557 4632 /*
4558 4633 * We don't want to have a hard dependency
4559 4634 * against some special symbols in sharefs
4560 4635 * nfs, and smbsrv. Determine them if needed when
4561 4636 * the first file system is shared.
4562 4637 * Neither sharefs, nfs or smbsrv are unloadable modules.
4563 4638 */
4564 4639 int (*znfsexport_fs)(void *arg);
4565 4640 int (*zshare_fs)(enum sharefs_sys_op, share_t *, uint32_t);
4566 4641 int (*zsmbexport_fs)(void *arg, boolean_t add_share);
4567 4642
4568 4643 int zfs_nfsshare_inited;
4569 4644 int zfs_smbshare_inited;
4570 4645
4571 4646 ddi_modhandle_t nfs_mod;
4572 4647 ddi_modhandle_t sharefs_mod;
4573 4648 ddi_modhandle_t smbsrv_mod;
4574 4649 kmutex_t zfs_share_lock;
4575 4650
4576 4651 static int
4577 4652 zfs_init_sharefs()
4578 4653 {
4579 4654 int error;
4580 4655
4581 4656 ASSERT(MUTEX_HELD(&zfs_share_lock));
4582 4657 /* Both NFS and SMB shares also require sharetab support. */
4583 4658 if (sharefs_mod == NULL && ((sharefs_mod =
4584 4659 ddi_modopen("fs/sharefs",
4585 4660 KRTLD_MODE_FIRST, &error)) == NULL)) {
4586 4661 return (SET_ERROR(ENOSYS));
4587 4662 }
4588 4663 if (zshare_fs == NULL && ((zshare_fs =
4589 4664 (int (*)(enum sharefs_sys_op, share_t *, uint32_t))
4590 4665 ddi_modsym(sharefs_mod, "sharefs_impl", &error)) == NULL)) {
4591 4666 return (SET_ERROR(ENOSYS));
4592 4667 }
4593 4668 return (0);
4594 4669 }
4595 4670
4596 4671 static int
4597 4672 zfs_ioc_share(zfs_cmd_t *zc)
4598 4673 {
4599 4674 int error;
4600 4675 int opcode;
4601 4676
4602 4677 switch (zc->zc_share.z_sharetype) {
4603 4678 case ZFS_SHARE_NFS:
4604 4679 case ZFS_UNSHARE_NFS:
4605 4680 if (zfs_nfsshare_inited == 0) {
4606 4681 mutex_enter(&zfs_share_lock);
4607 4682 if (nfs_mod == NULL && ((nfs_mod = ddi_modopen("fs/nfs",
4608 4683 KRTLD_MODE_FIRST, &error)) == NULL)) {
4609 4684 mutex_exit(&zfs_share_lock);
4610 4685 return (SET_ERROR(ENOSYS));
4611 4686 }
4612 4687 if (znfsexport_fs == NULL &&
4613 4688 ((znfsexport_fs = (int (*)(void *))
4614 4689 ddi_modsym(nfs_mod,
4615 4690 "nfs_export", &error)) == NULL)) {
4616 4691 mutex_exit(&zfs_share_lock);
4617 4692 return (SET_ERROR(ENOSYS));
4618 4693 }
4619 4694 error = zfs_init_sharefs();
4620 4695 if (error != 0) {
4621 4696 mutex_exit(&zfs_share_lock);
4622 4697 return (SET_ERROR(ENOSYS));
4623 4698 }
4624 4699 zfs_nfsshare_inited = 1;
4625 4700 mutex_exit(&zfs_share_lock);
4626 4701 }
4627 4702 break;
4628 4703 case ZFS_SHARE_SMB:
4629 4704 case ZFS_UNSHARE_SMB:
4630 4705 if (zfs_smbshare_inited == 0) {
4631 4706 mutex_enter(&zfs_share_lock);
4632 4707 if (smbsrv_mod == NULL && ((smbsrv_mod =
4633 4708 ddi_modopen("drv/smbsrv",
4634 4709 KRTLD_MODE_FIRST, &error)) == NULL)) {
4635 4710 mutex_exit(&zfs_share_lock);
4636 4711 return (SET_ERROR(ENOSYS));
4637 4712 }
4638 4713 if (zsmbexport_fs == NULL && ((zsmbexport_fs =
4639 4714 (int (*)(void *, boolean_t))ddi_modsym(smbsrv_mod,
4640 4715 "smb_server_share", &error)) == NULL)) {
4641 4716 mutex_exit(&zfs_share_lock);
4642 4717 return (SET_ERROR(ENOSYS));
4643 4718 }
4644 4719 error = zfs_init_sharefs();
4645 4720 if (error != 0) {
4646 4721 mutex_exit(&zfs_share_lock);
4647 4722 return (SET_ERROR(ENOSYS));
4648 4723 }
4649 4724 zfs_smbshare_inited = 1;
4650 4725 mutex_exit(&zfs_share_lock);
4651 4726 }
4652 4727 break;
4653 4728 default:
4654 4729 return (SET_ERROR(EINVAL));
4655 4730 }
4656 4731
4657 4732 switch (zc->zc_share.z_sharetype) {
4658 4733 case ZFS_SHARE_NFS:
4659 4734 case ZFS_UNSHARE_NFS:
4660 4735 if (error =
4661 4736 znfsexport_fs((void *)
4662 4737 (uintptr_t)zc->zc_share.z_exportdata))
4663 4738 return (error);
4664 4739 break;
4665 4740 case ZFS_SHARE_SMB:
4666 4741 case ZFS_UNSHARE_SMB:
4667 4742 if (error = zsmbexport_fs((void *)
4668 4743 (uintptr_t)zc->zc_share.z_exportdata,
4669 4744 zc->zc_share.z_sharetype == ZFS_SHARE_SMB ?
4670 4745 B_TRUE: B_FALSE)) {
4671 4746 return (error);
4672 4747 }
4673 4748 break;
4674 4749 }
4675 4750
4676 4751 opcode = (zc->zc_share.z_sharetype == ZFS_SHARE_NFS ||
4677 4752 zc->zc_share.z_sharetype == ZFS_SHARE_SMB) ?
4678 4753 SHAREFS_ADD : SHAREFS_REMOVE;
4679 4754
4680 4755 /*
4681 4756 * Add or remove share from sharetab
4682 4757 */
4683 4758 error = zshare_fs(opcode,
4684 4759 (void *)(uintptr_t)zc->zc_share.z_sharedata,
4685 4760 zc->zc_share.z_sharemax);
4686 4761
4687 4762 return (error);
4688 4763
4689 4764 }
4690 4765
4691 4766 ace_t full_access[] = {
4692 4767 {(uid_t)-1, ACE_ALL_PERMS, ACE_EVERYONE, 0}
4693 4768 };
4694 4769
4695 4770 /*
4696 4771 * inputs:
4697 4772 * zc_name name of containing filesystem
4698 4773 * zc_obj object # beyond which we want next in-use object #
4699 4774 *
4700 4775 * outputs:
4701 4776 * zc_obj next in-use object #
4702 4777 */
4703 4778 static int
4704 4779 zfs_ioc_next_obj(zfs_cmd_t *zc)
4705 4780 {
4706 4781 objset_t *os = NULL;
4707 4782 int error;
4708 4783
4709 4784 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
4710 4785 if (error != 0)
4711 4786 return (error);
4712 4787
4713 4788 error = dmu_object_next(os, &zc->zc_obj, B_FALSE,
4714 4789 os->os_dsl_dataset->ds_phys->ds_prev_snap_txg);
4715 4790
4716 4791 dmu_objset_rele(os, FTAG);
4717 4792 return (error);
4718 4793 }
4719 4794
4720 4795 /*
4721 4796 * inputs:
4722 4797 * zc_name name of filesystem
4723 4798 * zc_value prefix name for snapshot
4724 4799 * zc_cleanup_fd cleanup-on-exit file descriptor for calling process
4725 4800 *
4726 4801 * outputs:
4727 4802 * zc_value short name of new snapshot
4728 4803 */
4729 4804 static int
4730 4805 zfs_ioc_tmp_snapshot(zfs_cmd_t *zc)
4731 4806 {
4732 4807 char *snap_name;
4733 4808 char *hold_name;
4734 4809 int error;
4735 4810 minor_t minor;
4736 4811
4737 4812 error = zfs_onexit_fd_hold(zc->zc_cleanup_fd, &minor);
4738 4813 if (error != 0)
4739 4814 return (error);
4740 4815
4741 4816 snap_name = kmem_asprintf("%s-%016llx", zc->zc_value,
4742 4817 (u_longlong_t)ddi_get_lbolt64());
4743 4818 hold_name = kmem_asprintf("%%%s", zc->zc_value);
4744 4819
4745 4820 error = dsl_dataset_snapshot_tmp(zc->zc_name, snap_name, minor,
4746 4821 hold_name);
4747 4822 if (error == 0)
4748 4823 (void) strcpy(zc->zc_value, snap_name);
4749 4824 strfree(snap_name);
4750 4825 strfree(hold_name);
4751 4826 zfs_onexit_fd_rele(zc->zc_cleanup_fd);
4752 4827 return (error);
4753 4828 }
4754 4829
4755 4830 /*
4756 4831 * inputs:
4757 4832 * zc_name name of "to" snapshot
4758 4833 * zc_value name of "from" snapshot
4759 4834 * zc_cookie file descriptor to write diff data on
4760 4835 *
4761 4836 * outputs:
4762 4837 * dmu_diff_record_t's to the file descriptor
4763 4838 */
4764 4839 static int
4765 4840 zfs_ioc_diff(zfs_cmd_t *zc)
4766 4841 {
4767 4842 file_t *fp;
4768 4843 offset_t off;
4769 4844 int error;
4770 4845
4771 4846 fp = getf(zc->zc_cookie);
4772 4847 if (fp == NULL)
4773 4848 return (SET_ERROR(EBADF));
4774 4849
4775 4850 off = fp->f_offset;
4776 4851
4777 4852 error = dmu_diff(zc->zc_name, zc->zc_value, fp->f_vnode, &off);
4778 4853
4779 4854 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
4780 4855 fp->f_offset = off;
4781 4856 releasef(zc->zc_cookie);
4782 4857
4783 4858 return (error);
4784 4859 }
4785 4860
4786 4861 /*
4787 4862 * Remove all ACL files in shares dir
4788 4863 */
4789 4864 static int
4790 4865 zfs_smb_acl_purge(znode_t *dzp)
4791 4866 {
4792 4867 zap_cursor_t zc;
4793 4868 zap_attribute_t zap;
4794 4869 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
4795 4870 int error;
4796 4871
4797 4872 for (zap_cursor_init(&zc, zfsvfs->z_os, dzp->z_id);
4798 4873 (error = zap_cursor_retrieve(&zc, &zap)) == 0;
4799 4874 zap_cursor_advance(&zc)) {
4800 4875 if ((error = VOP_REMOVE(ZTOV(dzp), zap.za_name, kcred,
4801 4876 NULL, 0)) != 0)
4802 4877 break;
4803 4878 }
4804 4879 zap_cursor_fini(&zc);
4805 4880 return (error);
4806 4881 }
4807 4882
4808 4883 static int
4809 4884 zfs_ioc_smb_acl(zfs_cmd_t *zc)
4810 4885 {
4811 4886 vnode_t *vp;
4812 4887 znode_t *dzp;
4813 4888 vnode_t *resourcevp = NULL;
4814 4889 znode_t *sharedir;
4815 4890 zfsvfs_t *zfsvfs;
4816 4891 nvlist_t *nvlist;
4817 4892 char *src, *target;
4818 4893 vattr_t vattr;
4819 4894 vsecattr_t vsec;
4820 4895 int error = 0;
4821 4896
4822 4897 if ((error = lookupname(zc->zc_value, UIO_SYSSPACE,
4823 4898 NO_FOLLOW, NULL, &vp)) != 0)
4824 4899 return (error);
4825 4900
4826 4901 /* Now make sure mntpnt and dataset are ZFS */
4827 4902
4828 4903 if (vp->v_vfsp->vfs_fstype != zfsfstype ||
4829 4904 (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource),
4830 4905 zc->zc_name) != 0)) {
4831 4906 VN_RELE(vp);
4832 4907 return (SET_ERROR(EINVAL));
4833 4908 }
4834 4909
4835 4910 dzp = VTOZ(vp);
4836 4911 zfsvfs = dzp->z_zfsvfs;
4837 4912 ZFS_ENTER(zfsvfs);
4838 4913
4839 4914 /*
4840 4915 * Create share dir if its missing.
4841 4916 */
4842 4917 mutex_enter(&zfsvfs->z_lock);
4843 4918 if (zfsvfs->z_shares_dir == 0) {
4844 4919 dmu_tx_t *tx;
4845 4920
4846 4921 tx = dmu_tx_create(zfsvfs->z_os);
4847 4922 dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, TRUE,
4848 4923 ZFS_SHARES_DIR);
4849 4924 dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL);
4850 4925 error = dmu_tx_assign(tx, TXG_WAIT);
4851 4926 if (error != 0) {
4852 4927 dmu_tx_abort(tx);
4853 4928 } else {
4854 4929 error = zfs_create_share_dir(zfsvfs, tx);
4855 4930 dmu_tx_commit(tx);
4856 4931 }
4857 4932 if (error != 0) {
4858 4933 mutex_exit(&zfsvfs->z_lock);
4859 4934 VN_RELE(vp);
4860 4935 ZFS_EXIT(zfsvfs);
4861 4936 return (error);
4862 4937 }
4863 4938 }
4864 4939 mutex_exit(&zfsvfs->z_lock);
4865 4940
4866 4941 ASSERT(zfsvfs->z_shares_dir);
4867 4942 if ((error = zfs_zget(zfsvfs, zfsvfs->z_shares_dir, &sharedir)) != 0) {
4868 4943 VN_RELE(vp);
4869 4944 ZFS_EXIT(zfsvfs);
4870 4945 return (error);
4871 4946 }
4872 4947
4873 4948 switch (zc->zc_cookie) {
4874 4949 case ZFS_SMB_ACL_ADD:
4875 4950 vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE;
4876 4951 vattr.va_type = VREG;
4877 4952 vattr.va_mode = S_IFREG|0777;
4878 4953 vattr.va_uid = 0;
4879 4954 vattr.va_gid = 0;
4880 4955
4881 4956 vsec.vsa_mask = VSA_ACE;
4882 4957 vsec.vsa_aclentp = &full_access;
4883 4958 vsec.vsa_aclentsz = sizeof (full_access);
4884 4959 vsec.vsa_aclcnt = 1;
4885 4960
4886 4961 error = VOP_CREATE(ZTOV(sharedir), zc->zc_string,
4887 4962 &vattr, EXCL, 0, &resourcevp, kcred, 0, NULL, &vsec);
4888 4963 if (resourcevp)
4889 4964 VN_RELE(resourcevp);
4890 4965 break;
4891 4966
4892 4967 case ZFS_SMB_ACL_REMOVE:
4893 4968 error = VOP_REMOVE(ZTOV(sharedir), zc->zc_string, kcred,
4894 4969 NULL, 0);
4895 4970 break;
4896 4971
4897 4972 case ZFS_SMB_ACL_RENAME:
4898 4973 if ((error = get_nvlist(zc->zc_nvlist_src,
4899 4974 zc->zc_nvlist_src_size, zc->zc_iflags, &nvlist)) != 0) {
4900 4975 VN_RELE(vp);
4901 4976 ZFS_EXIT(zfsvfs);
4902 4977 return (error);
4903 4978 }
4904 4979 if (nvlist_lookup_string(nvlist, ZFS_SMB_ACL_SRC, &src) ||
4905 4980 nvlist_lookup_string(nvlist, ZFS_SMB_ACL_TARGET,
4906 4981 &target)) {
4907 4982 VN_RELE(vp);
4908 4983 VN_RELE(ZTOV(sharedir));
4909 4984 ZFS_EXIT(zfsvfs);
4910 4985 nvlist_free(nvlist);
4911 4986 return (error);
4912 4987 }
4913 4988 error = VOP_RENAME(ZTOV(sharedir), src, ZTOV(sharedir), target,
4914 4989 kcred, NULL, 0);
4915 4990 nvlist_free(nvlist);
4916 4991 break;
4917 4992
4918 4993 case ZFS_SMB_ACL_PURGE:
4919 4994 error = zfs_smb_acl_purge(sharedir);
4920 4995 break;
4921 4996
4922 4997 default:
4923 4998 error = SET_ERROR(EINVAL);
4924 4999 break;
4925 5000 }
4926 5001
4927 5002 VN_RELE(vp);
4928 5003 VN_RELE(ZTOV(sharedir));
4929 5004
4930 5005 ZFS_EXIT(zfsvfs);
4931 5006
4932 5007 return (error);
4933 5008 }
4934 5009
4935 5010 /*
4936 5011 * innvl: {
4937 5012 * "holds" -> { snapname -> holdname (string), ... }
4938 5013 * (optional) "cleanup_fd" -> fd (int32)
4939 5014 * }
4940 5015 *
4941 5016 * outnvl: {
4942 5017 * snapname -> error value (int32)
4943 5018 * ...
4944 5019 * }
4945 5020 */
4946 5021 /* ARGSUSED */
4947 5022 static int
4948 5023 zfs_ioc_hold(const char *pool, nvlist_t *args, nvlist_t *errlist)
4949 5024 {
4950 5025 nvlist_t *holds;
4951 5026 int cleanup_fd = -1;
4952 5027 int error;
4953 5028 minor_t minor = 0;
4954 5029
4955 5030 error = nvlist_lookup_nvlist(args, "holds", &holds);
4956 5031 if (error != 0)
4957 5032 return (SET_ERROR(EINVAL));
4958 5033
4959 5034 if (nvlist_lookup_int32(args, "cleanup_fd", &cleanup_fd) == 0) {
4960 5035 error = zfs_onexit_fd_hold(cleanup_fd, &minor);
4961 5036 if (error != 0)
4962 5037 return (error);
4963 5038 }
4964 5039
4965 5040 error = dsl_dataset_user_hold(holds, minor, errlist);
4966 5041 if (minor != 0)
4967 5042 zfs_onexit_fd_rele(cleanup_fd);
4968 5043 return (error);
4969 5044 }
4970 5045
4971 5046 /*
4972 5047 * innvl is not used.
4973 5048 *
4974 5049 * outnvl: {
4975 5050 * holdname -> time added (uint64 seconds since epoch)
4976 5051 * ...
4977 5052 * }
4978 5053 */
4979 5054 /* ARGSUSED */
4980 5055 static int
4981 5056 zfs_ioc_get_holds(const char *snapname, nvlist_t *args, nvlist_t *outnvl)
4982 5057 {
4983 5058 return (dsl_dataset_get_holds(snapname, outnvl));
4984 5059 }
4985 5060
4986 5061 /*
4987 5062 * innvl: {
4988 5063 * snapname -> { holdname, ... }
4989 5064 * ...
4990 5065 * }
4991 5066 *
4992 5067 * outnvl: {
4993 5068 * snapname -> error value (int32)
4994 5069 * ...
4995 5070 * }
4996 5071 */
4997 5072 /* ARGSUSED */
4998 5073 static int
4999 5074 zfs_ioc_release(const char *pool, nvlist_t *holds, nvlist_t *errlist)
5000 5075 {
5001 5076 return (dsl_dataset_user_release(holds, errlist));
5002 5077 }
5003 5078
5004 5079 /*
5005 5080 * inputs:
5006 5081 * zc_name name of new filesystem or snapshot
5007 5082 * zc_value full name of old snapshot
5008 5083 *
5009 5084 * outputs:
5010 5085 * zc_cookie space in bytes
5011 5086 * zc_objset_type compressed space in bytes
5012 5087 * zc_perm_action uncompressed space in bytes
5013 5088 */
5014 5089 static int
5015 5090 zfs_ioc_space_written(zfs_cmd_t *zc)
5016 5091 {
5017 5092 int error;
5018 5093 dsl_pool_t *dp;
5019 5094 dsl_dataset_t *new, *old;
5020 5095
5021 5096 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
5022 5097 if (error != 0)
5023 5098 return (error);
5024 5099 error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &new);
5025 5100 if (error != 0) {
5026 5101 dsl_pool_rele(dp, FTAG);
5027 5102 return (error);
5028 5103 }
5029 5104 error = dsl_dataset_hold(dp, zc->zc_value, FTAG, &old);
5030 5105 if (error != 0) {
5031 5106 dsl_dataset_rele(new, FTAG);
5032 5107 dsl_pool_rele(dp, FTAG);
5033 5108 return (error);
5034 5109 }
5035 5110
5036 5111 error = dsl_dataset_space_written(old, new, &zc->zc_cookie,
5037 5112 &zc->zc_objset_type, &zc->zc_perm_action);
5038 5113 dsl_dataset_rele(old, FTAG);
5039 5114 dsl_dataset_rele(new, FTAG);
5040 5115 dsl_pool_rele(dp, FTAG);
5041 5116 return (error);
5042 5117 }
5043 5118
5044 5119 /*
5045 5120 * innvl: {
5046 5121 * "firstsnap" -> snapshot name
5047 5122 * }
5048 5123 *
5049 5124 * outnvl: {
5050 5125 * "used" -> space in bytes
5051 5126 * "compressed" -> compressed space in bytes
5052 5127 * "uncompressed" -> uncompressed space in bytes
5053 5128 * }
5054 5129 */
5055 5130 static int
5056 5131 zfs_ioc_space_snaps(const char *lastsnap, nvlist_t *innvl, nvlist_t *outnvl)
5057 5132 {
5058 5133 int error;
5059 5134 dsl_pool_t *dp;
5060 5135 dsl_dataset_t *new, *old;
5061 5136 char *firstsnap;
5062 5137 uint64_t used, comp, uncomp;
5063 5138
5064 5139 if (nvlist_lookup_string(innvl, "firstsnap", &firstsnap) != 0)
5065 5140 return (SET_ERROR(EINVAL));
5066 5141
5067 5142 error = dsl_pool_hold(lastsnap, FTAG, &dp);
5068 5143 if (error != 0)
5069 5144 return (error);
5070 5145
5071 5146 error = dsl_dataset_hold(dp, lastsnap, FTAG, &new);
5072 5147 if (error != 0) {
5073 5148 dsl_pool_rele(dp, FTAG);
5074 5149 return (error);
5075 5150 }
5076 5151 error = dsl_dataset_hold(dp, firstsnap, FTAG, &old);
5077 5152 if (error != 0) {
5078 5153 dsl_dataset_rele(new, FTAG);
5079 5154 dsl_pool_rele(dp, FTAG);
5080 5155 return (error);
5081 5156 }
5082 5157
5083 5158 error = dsl_dataset_space_wouldfree(old, new, &used, &comp, &uncomp);
5084 5159 dsl_dataset_rele(old, FTAG);
5085 5160 dsl_dataset_rele(new, FTAG);
5086 5161 dsl_pool_rele(dp, FTAG);
5087 5162 fnvlist_add_uint64(outnvl, "used", used);
5088 5163 fnvlist_add_uint64(outnvl, "compressed", comp);
5089 5164 fnvlist_add_uint64(outnvl, "uncompressed", uncomp);
5090 5165 return (error);
5091 5166 }
5092 5167
5093 5168 /*
5094 5169 * innvl: {
5095 5170 * "fd" -> file descriptor to write stream to (int32)
5096 5171 * (optional) "fromsnap" -> full snap name to send an incremental from
5097 5172 * }
5098 5173 *
5099 5174 * outnvl is unused
5100 5175 */
5101 5176 /* ARGSUSED */
5102 5177 static int
5103 5178 zfs_ioc_send_new(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl)
5104 5179 {
5105 5180 int error;
5106 5181 offset_t off;
5107 5182 char *fromname = NULL;
5108 5183 int fd;
5109 5184
5110 5185 error = nvlist_lookup_int32(innvl, "fd", &fd);
5111 5186 if (error != 0)
5112 5187 return (SET_ERROR(EINVAL));
5113 5188
5114 5189 (void) nvlist_lookup_string(innvl, "fromsnap", &fromname);
5115 5190
5116 5191 file_t *fp = getf(fd);
5117 5192 if (fp == NULL)
5118 5193 return (SET_ERROR(EBADF));
5119 5194
5120 5195 off = fp->f_offset;
5121 5196 error = dmu_send(snapname, fromname, fd, fp->f_vnode, &off);
5122 5197
5123 5198 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
5124 5199 fp->f_offset = off;
5125 5200 releasef(fd);
5126 5201 return (error);
5127 5202 }
5128 5203
5129 5204 /*
5130 5205 * Determine approximately how large a zfs send stream will be -- the number
5131 5206 * of bytes that will be written to the fd supplied to zfs_ioc_send_new().
5132 5207 *
5133 5208 * innvl: {
5134 5209 * (optional) "fromsnap" -> full snap name to send an incremental from
5135 5210 * }
5136 5211 *
5137 5212 * outnvl: {
5138 5213 * "space" -> bytes of space (uint64)
5139 5214 * }
5140 5215 */
5141 5216 static int
5142 5217 zfs_ioc_send_space(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl)
5143 5218 {
5144 5219 dsl_pool_t *dp;
5145 5220 dsl_dataset_t *fromsnap = NULL;
5146 5221 dsl_dataset_t *tosnap;
5147 5222 int error;
5148 5223 char *fromname;
5149 5224 uint64_t space;
5150 5225
5151 5226 error = dsl_pool_hold(snapname, FTAG, &dp);
5152 5227 if (error != 0)
5153 5228 return (error);
5154 5229
5155 5230 error = dsl_dataset_hold(dp, snapname, FTAG, &tosnap);
5156 5231 if (error != 0) {
5157 5232 dsl_pool_rele(dp, FTAG);
5158 5233 return (error);
5159 5234 }
5160 5235
5161 5236 error = nvlist_lookup_string(innvl, "fromsnap", &fromname);
5162 5237 if (error == 0) {
5163 5238 error = dsl_dataset_hold(dp, fromname, FTAG, &fromsnap);
5164 5239 if (error != 0) {
5165 5240 dsl_dataset_rele(tosnap, FTAG);
5166 5241 dsl_pool_rele(dp, FTAG);
5167 5242 return (error);
5168 5243 }
5169 5244 }
5170 5245
5171 5246 error = dmu_send_estimate(tosnap, fromsnap, &space);
5172 5247 fnvlist_add_uint64(outnvl, "space", space);
5173 5248
5174 5249 if (fromsnap != NULL)
5175 5250 dsl_dataset_rele(fromsnap, FTAG);
5176 5251 dsl_dataset_rele(tosnap, FTAG);
5177 5252 dsl_pool_rele(dp, FTAG);
5178 5253 return (error);
5179 5254 }
5180 5255
5181 5256
5182 5257 static zfs_ioc_vec_t zfs_ioc_vec[ZFS_IOC_LAST - ZFS_IOC_FIRST];
5183 5258
5184 5259 static void
5185 5260 zfs_ioctl_register_legacy(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5186 5261 zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck,
5187 5262 boolean_t log_history, zfs_ioc_poolcheck_t pool_check)
5188 5263 {
5189 5264 zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST];
5190 5265
5191 5266 ASSERT3U(ioc, >=, ZFS_IOC_FIRST);
5192 5267 ASSERT3U(ioc, <, ZFS_IOC_LAST);
5193 5268 ASSERT3P(vec->zvec_legacy_func, ==, NULL);
5194 5269 ASSERT3P(vec->zvec_func, ==, NULL);
5195 5270
5196 5271 vec->zvec_legacy_func = func;
5197 5272 vec->zvec_secpolicy = secpolicy;
5198 5273 vec->zvec_namecheck = namecheck;
5199 5274 vec->zvec_allow_log = log_history;
5200 5275 vec->zvec_pool_check = pool_check;
5201 5276 }
5202 5277
5203 5278 /*
5204 5279 * See the block comment at the beginning of this file for details on
5205 5280 * each argument to this function.
5206 5281 */
5207 5282 static void
5208 5283 zfs_ioctl_register(const char *name, zfs_ioc_t ioc, zfs_ioc_func_t *func,
5209 5284 zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck,
5210 5285 zfs_ioc_poolcheck_t pool_check, boolean_t smush_outnvlist,
5211 5286 boolean_t allow_log)
5212 5287 {
5213 5288 zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST];
5214 5289
5215 5290 ASSERT3U(ioc, >=, ZFS_IOC_FIRST);
5216 5291 ASSERT3U(ioc, <, ZFS_IOC_LAST);
5217 5292 ASSERT3P(vec->zvec_legacy_func, ==, NULL);
5218 5293 ASSERT3P(vec->zvec_func, ==, NULL);
5219 5294
5220 5295 /* if we are logging, the name must be valid */
5221 5296 ASSERT(!allow_log || namecheck != NO_NAME);
5222 5297
5223 5298 vec->zvec_name = name;
5224 5299 vec->zvec_func = func;
5225 5300 vec->zvec_secpolicy = secpolicy;
5226 5301 vec->zvec_namecheck = namecheck;
5227 5302 vec->zvec_pool_check = pool_check;
5228 5303 vec->zvec_smush_outnvlist = smush_outnvlist;
5229 5304 vec->zvec_allow_log = allow_log;
5230 5305 }
5231 5306
5232 5307 static void
5233 5308 zfs_ioctl_register_pool(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5234 5309 zfs_secpolicy_func_t *secpolicy, boolean_t log_history,
5235 5310 zfs_ioc_poolcheck_t pool_check)
5236 5311 {
5237 5312 zfs_ioctl_register_legacy(ioc, func, secpolicy,
5238 5313 POOL_NAME, log_history, pool_check);
5239 5314 }
5240 5315
5241 5316 static void
5242 5317 zfs_ioctl_register_dataset_nolog(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5243 5318 zfs_secpolicy_func_t *secpolicy, zfs_ioc_poolcheck_t pool_check)
5244 5319 {
5245 5320 zfs_ioctl_register_legacy(ioc, func, secpolicy,
5246 5321 DATASET_NAME, B_FALSE, pool_check);
5247 5322 }
5248 5323
5249 5324 static void
5250 5325 zfs_ioctl_register_pool_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func)
5251 5326 {
5252 5327 zfs_ioctl_register_legacy(ioc, func, zfs_secpolicy_config,
5253 5328 POOL_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5254 5329 }
5255 5330
5256 5331 static void
5257 5332 zfs_ioctl_register_pool_meta(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5258 5333 zfs_secpolicy_func_t *secpolicy)
5259 5334 {
5260 5335 zfs_ioctl_register_legacy(ioc, func, secpolicy,
5261 5336 NO_NAME, B_FALSE, POOL_CHECK_NONE);
5262 5337 }
5263 5338
5264 5339 static void
5265 5340 zfs_ioctl_register_dataset_read_secpolicy(zfs_ioc_t ioc,
5266 5341 zfs_ioc_legacy_func_t *func, zfs_secpolicy_func_t *secpolicy)
5267 5342 {
5268 5343 zfs_ioctl_register_legacy(ioc, func, secpolicy,
5269 5344 DATASET_NAME, B_FALSE, POOL_CHECK_SUSPENDED);
5270 5345 }
5271 5346
5272 5347 static void
5273 5348 zfs_ioctl_register_dataset_read(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func)
5274 5349 {
5275 5350 zfs_ioctl_register_dataset_read_secpolicy(ioc, func,
5276 5351 zfs_secpolicy_read);
5277 5352 }
5278 5353
5279 5354 static void
5280 5355 zfs_ioctl_register_dataset_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5281 5356 zfs_secpolicy_func_t *secpolicy)
5282 5357 {
5283 5358 zfs_ioctl_register_legacy(ioc, func, secpolicy,
5284 5359 DATASET_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5285 5360 }
5286 5361
5287 5362 static void
5288 5363 zfs_ioctl_init(void)
5289 5364 {
5290 5365 zfs_ioctl_register("snapshot", ZFS_IOC_SNAPSHOT,
5291 5366 zfs_ioc_snapshot, zfs_secpolicy_snapshot, POOL_NAME,
5292 5367 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5293 5368
5294 5369 zfs_ioctl_register("log_history", ZFS_IOC_LOG_HISTORY,
5295 5370 zfs_ioc_log_history, zfs_secpolicy_log_history, NO_NAME,
5296 5371 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_FALSE);
5297 5372
5298 5373 zfs_ioctl_register("space_snaps", ZFS_IOC_SPACE_SNAPS,
5299 5374 zfs_ioc_space_snaps, zfs_secpolicy_read, DATASET_NAME,
5300 5375 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5301 5376
5302 5377 zfs_ioctl_register("send", ZFS_IOC_SEND_NEW,
5303 5378 zfs_ioc_send_new, zfs_secpolicy_send_new, DATASET_NAME,
5304 5379 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5305 5380
5306 5381 zfs_ioctl_register("send_space", ZFS_IOC_SEND_SPACE,
5307 5382 zfs_ioc_send_space, zfs_secpolicy_read, DATASET_NAME,
5308 5383 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5309 5384
5310 5385 zfs_ioctl_register("create", ZFS_IOC_CREATE,
5311 5386 zfs_ioc_create, zfs_secpolicy_create_clone, DATASET_NAME,
5312 5387 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5313 5388
5314 5389 zfs_ioctl_register("clone", ZFS_IOC_CLONE,
5315 5390 zfs_ioc_clone, zfs_secpolicy_create_clone, DATASET_NAME,
5316 5391 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5317 5392
5318 5393 zfs_ioctl_register("destroy_snaps", ZFS_IOC_DESTROY_SNAPS,
5319 5394 zfs_ioc_destroy_snaps, zfs_secpolicy_destroy_snaps, POOL_NAME,
5320 5395 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5321 5396
5322 5397 zfs_ioctl_register("hold", ZFS_IOC_HOLD,
5323 5398 zfs_ioc_hold, zfs_secpolicy_hold, POOL_NAME,
5324 5399 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5325 5400 zfs_ioctl_register("release", ZFS_IOC_RELEASE,
5326 5401 zfs_ioc_release, zfs_secpolicy_release, POOL_NAME,
5327 5402 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5328 5403
5329 5404 zfs_ioctl_register("get_holds", ZFS_IOC_GET_HOLDS,
5330 5405 zfs_ioc_get_holds, zfs_secpolicy_read, DATASET_NAME,
5331 5406 POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5332 5407
5333 5408 zfs_ioctl_register("rollback", ZFS_IOC_ROLLBACK,
5334 5409 zfs_ioc_rollback, zfs_secpolicy_rollback, DATASET_NAME,
5335 5410 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_TRUE);
5336 5411
5337 5412 /* IOCTLS that use the legacy function signature */
5338 5413
5339 5414 zfs_ioctl_register_legacy(ZFS_IOC_POOL_FREEZE, zfs_ioc_pool_freeze,
5340 5415 zfs_secpolicy_config, NO_NAME, B_FALSE, POOL_CHECK_READONLY);
5341 5416
5342 5417 zfs_ioctl_register_pool(ZFS_IOC_POOL_CREATE, zfs_ioc_pool_create,
5343 5418 zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE);
5344 5419 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SCAN,
5345 5420 zfs_ioc_pool_scan);
5346 5421 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_UPGRADE,
5347 5422 zfs_ioc_pool_upgrade);
5348 5423 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ADD,
5349 5424 zfs_ioc_vdev_add);
5350 5425 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_REMOVE,
5351 5426 zfs_ioc_vdev_remove);
5352 5427 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SET_STATE,
5353 5428 zfs_ioc_vdev_set_state);
5354 5429 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ATTACH,
5355 5430 zfs_ioc_vdev_attach);
5356 5431 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_DETACH,
5357 5432 zfs_ioc_vdev_detach);
5358 5433 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETPATH,
5359 5434 zfs_ioc_vdev_setpath);
5360 5435 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETFRU,
5361 5436 zfs_ioc_vdev_setfru);
5362 5437 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SET_PROPS,
5363 5438 zfs_ioc_pool_set_props);
5364 5439 zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SPLIT,
5365 5440 zfs_ioc_vdev_split);
5366 5441 zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_REGUID,
5367 5442 zfs_ioc_pool_reguid);
5368 5443
5369 5444 zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_CONFIGS,
5370 5445 zfs_ioc_pool_configs, zfs_secpolicy_none);
5371 5446 zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_TRYIMPORT,
5372 5447 zfs_ioc_pool_tryimport, zfs_secpolicy_config);
5373 5448 zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_FAULT,
5374 5449 zfs_ioc_inject_fault, zfs_secpolicy_inject);
5375 5450 zfs_ioctl_register_pool_meta(ZFS_IOC_CLEAR_FAULT,
5376 5451 zfs_ioc_clear_fault, zfs_secpolicy_inject);
5377 5452 zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_LIST_NEXT,
5378 5453 zfs_ioc_inject_list_next, zfs_secpolicy_inject);
5379 5454
5380 5455 /*
5381 5456 * pool destroy, and export don't log the history as part of
5382 5457 * zfsdev_ioctl, but rather zfs_ioc_pool_export
5383 5458 * does the logging of those commands.
5384 5459 */
5385 5460 zfs_ioctl_register_pool(ZFS_IOC_POOL_DESTROY, zfs_ioc_pool_destroy,
5386 5461 zfs_secpolicy_config, B_FALSE, POOL_CHECK_NONE);
5387 5462 zfs_ioctl_register_pool(ZFS_IOC_POOL_EXPORT, zfs_ioc_pool_export,
5388 5463 zfs_secpolicy_config, B_FALSE, POOL_CHECK_NONE);
5389 5464
5390 5465 zfs_ioctl_register_pool(ZFS_IOC_POOL_STATS, zfs_ioc_pool_stats,
5391 5466 zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE);
5392 5467 zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_PROPS, zfs_ioc_pool_get_props,
5393 5468 zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE);
5394 5469
5395 5470 zfs_ioctl_register_pool(ZFS_IOC_ERROR_LOG, zfs_ioc_error_log,
5396 5471 zfs_secpolicy_inject, B_FALSE, POOL_CHECK_SUSPENDED);
5397 5472 zfs_ioctl_register_pool(ZFS_IOC_DSOBJ_TO_DSNAME,
5398 5473 zfs_ioc_dsobj_to_dsname,
5399 5474 zfs_secpolicy_diff, B_FALSE, POOL_CHECK_SUSPENDED);
5400 5475 zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_HISTORY,
5401 5476 zfs_ioc_pool_get_history,
5402 5477 zfs_secpolicy_config, B_FALSE, POOL_CHECK_SUSPENDED);
5403 5478
5404 5479 zfs_ioctl_register_pool(ZFS_IOC_POOL_IMPORT, zfs_ioc_pool_import,
5405 5480 zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE);
5406 5481
5407 5482 zfs_ioctl_register_pool(ZFS_IOC_CLEAR, zfs_ioc_clear,
5408 5483 zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE);
5409 5484 zfs_ioctl_register_pool(ZFS_IOC_POOL_REOPEN, zfs_ioc_pool_reopen,
5410 5485 zfs_secpolicy_config, B_TRUE, POOL_CHECK_SUSPENDED);
5411 5486
5412 5487 zfs_ioctl_register_dataset_read(ZFS_IOC_SPACE_WRITTEN,
5413 5488 zfs_ioc_space_written);
5414 5489 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_RECVD_PROPS,
5415 5490 zfs_ioc_objset_recvd_props);
5416 5491 zfs_ioctl_register_dataset_read(ZFS_IOC_NEXT_OBJ,
5417 5492 zfs_ioc_next_obj);
5418 5493 zfs_ioctl_register_dataset_read(ZFS_IOC_GET_FSACL,
5419 5494 zfs_ioc_get_fsacl);
5420 5495 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_STATS,
5421 5496 zfs_ioc_objset_stats);
5422 5497 zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_ZPLPROPS,
5423 5498 zfs_ioc_objset_zplprops);
5424 5499 zfs_ioctl_register_dataset_read(ZFS_IOC_DATASET_LIST_NEXT,
5425 5500 zfs_ioc_dataset_list_next);
5426 5501 zfs_ioctl_register_dataset_read(ZFS_IOC_SNAPSHOT_LIST_NEXT,
5427 5502 zfs_ioc_snapshot_list_next);
5428 5503 zfs_ioctl_register_dataset_read(ZFS_IOC_SEND_PROGRESS,
5429 5504 zfs_ioc_send_progress);
5430 5505
5431 5506 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_DIFF,
5432 5507 zfs_ioc_diff, zfs_secpolicy_diff);
5433 5508 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_STATS,
5434 5509 zfs_ioc_obj_to_stats, zfs_secpolicy_diff);
5435 5510 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_PATH,
5436 5511 zfs_ioc_obj_to_path, zfs_secpolicy_diff);
5437 5512 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_ONE,
5438 5513 zfs_ioc_userspace_one, zfs_secpolicy_userspace_one);
5439 5514 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_MANY,
5440 5515 zfs_ioc_userspace_many, zfs_secpolicy_userspace_many);
5441 5516 zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_SEND,
5442 5517 zfs_ioc_send, zfs_secpolicy_send);
5443 5518
5444 5519 zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_PROP, zfs_ioc_set_prop,
5445 5520 zfs_secpolicy_none);
5446 5521 zfs_ioctl_register_dataset_modify(ZFS_IOC_DESTROY, zfs_ioc_destroy,
5447 5522 zfs_secpolicy_destroy);
5448 5523 zfs_ioctl_register_dataset_modify(ZFS_IOC_RENAME, zfs_ioc_rename,
5449 5524 zfs_secpolicy_rename);
5450 5525 zfs_ioctl_register_dataset_modify(ZFS_IOC_RECV, zfs_ioc_recv,
5451 5526 zfs_secpolicy_recv);
5452 5527 zfs_ioctl_register_dataset_modify(ZFS_IOC_PROMOTE, zfs_ioc_promote,
5453 5528 zfs_secpolicy_promote);
5454 5529 zfs_ioctl_register_dataset_modify(ZFS_IOC_INHERIT_PROP,
5455 5530 zfs_ioc_inherit_prop, zfs_secpolicy_inherit_prop);
5456 5531 zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_FSACL, zfs_ioc_set_fsacl,
5457 5532 zfs_secpolicy_set_fsacl);
5458 5533
5459 5534 zfs_ioctl_register_dataset_nolog(ZFS_IOC_SHARE, zfs_ioc_share,
5460 5535 zfs_secpolicy_share, POOL_CHECK_NONE);
5461 5536 zfs_ioctl_register_dataset_nolog(ZFS_IOC_SMB_ACL, zfs_ioc_smb_acl,
5462 5537 zfs_secpolicy_smb_acl, POOL_CHECK_NONE);
5463 5538 zfs_ioctl_register_dataset_nolog(ZFS_IOC_USERSPACE_UPGRADE,
5464 5539 zfs_ioc_userspace_upgrade, zfs_secpolicy_userspace_upgrade,
5465 5540 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5466 5541 zfs_ioctl_register_dataset_nolog(ZFS_IOC_TMP_SNAPSHOT,
5467 5542 zfs_ioc_tmp_snapshot, zfs_secpolicy_tmp_snapshot,
5468 5543 POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5469 5544 }
5470 5545
5471 5546 int
5472 5547 pool_status_check(const char *name, zfs_ioc_namecheck_t type,
5473 5548 zfs_ioc_poolcheck_t check)
5474 5549 {
5475 5550 spa_t *spa;
5476 5551 int error;
5477 5552
5478 5553 ASSERT(type == POOL_NAME || type == DATASET_NAME);
5479 5554
5480 5555 if (check & POOL_CHECK_NONE)
5481 5556 return (0);
5482 5557
5483 5558 error = spa_open(name, &spa, FTAG);
5484 5559 if (error == 0) {
5485 5560 if ((check & POOL_CHECK_SUSPENDED) && spa_suspended(spa))
5486 5561 error = SET_ERROR(EAGAIN);
5487 5562 else if ((check & POOL_CHECK_READONLY) && !spa_writeable(spa))
5488 5563 error = SET_ERROR(EROFS);
5489 5564 spa_close(spa, FTAG);
5490 5565 }
5491 5566 return (error);
5492 5567 }
5493 5568
5494 5569 /*
5495 5570 * Find a free minor number.
5496 5571 */
5497 5572 minor_t
5498 5573 zfsdev_minor_alloc(void)
5499 5574 {
5500 5575 static minor_t last_minor;
5501 5576 minor_t m;
5502 5577
5503 5578 ASSERT(MUTEX_HELD(&zfsdev_state_lock));
5504 5579
5505 5580 for (m = last_minor + 1; m != last_minor; m++) {
5506 5581 if (m > ZFSDEV_MAX_MINOR)
5507 5582 m = 1;
5508 5583 if (ddi_get_soft_state(zfsdev_state, m) == NULL) {
5509 5584 last_minor = m;
5510 5585 return (m);
5511 5586 }
5512 5587 }
5513 5588
5514 5589 return (0);
5515 5590 }
5516 5591
5517 5592 static int
5518 5593 zfs_ctldev_init(dev_t *devp)
5519 5594 {
5520 5595 minor_t minor;
5521 5596 zfs_soft_state_t *zs;
5522 5597
5523 5598 ASSERT(MUTEX_HELD(&zfsdev_state_lock));
5524 5599 ASSERT(getminor(*devp) == 0);
5525 5600
5526 5601 minor = zfsdev_minor_alloc();
5527 5602 if (minor == 0)
5528 5603 return (SET_ERROR(ENXIO));
5529 5604
5530 5605 if (ddi_soft_state_zalloc(zfsdev_state, minor) != DDI_SUCCESS)
5531 5606 return (SET_ERROR(EAGAIN));
5532 5607
5533 5608 *devp = makedevice(getemajor(*devp), minor);
5534 5609
5535 5610 zs = ddi_get_soft_state(zfsdev_state, minor);
5536 5611 zs->zss_type = ZSST_CTLDEV;
5537 5612 zfs_onexit_init((zfs_onexit_t **)&zs->zss_data);
5538 5613
5539 5614 return (0);
5540 5615 }
5541 5616
5542 5617 static void
5543 5618 zfs_ctldev_destroy(zfs_onexit_t *zo, minor_t minor)
5544 5619 {
5545 5620 ASSERT(MUTEX_HELD(&zfsdev_state_lock));
5546 5621
5547 5622 zfs_onexit_destroy(zo);
5548 5623 ddi_soft_state_free(zfsdev_state, minor);
5549 5624 }
5550 5625
5551 5626 void *
5552 5627 zfsdev_get_soft_state(minor_t minor, enum zfs_soft_state_type which)
5553 5628 {
5554 5629 zfs_soft_state_t *zp;
5555 5630
5556 5631 zp = ddi_get_soft_state(zfsdev_state, minor);
5557 5632 if (zp == NULL || zp->zss_type != which)
5558 5633 return (NULL);
5559 5634
5560 5635 return (zp->zss_data);
5561 5636 }
5562 5637
5563 5638 static int
5564 5639 zfsdev_open(dev_t *devp, int flag, int otyp, cred_t *cr)
5565 5640 {
5566 5641 int error = 0;
5567 5642
5568 5643 if (getminor(*devp) != 0)
5569 5644 return (zvol_open(devp, flag, otyp, cr));
5570 5645
5571 5646 /* This is the control device. Allocate a new minor if requested. */
5572 5647 if (flag & FEXCL) {
5573 5648 mutex_enter(&zfsdev_state_lock);
5574 5649 error = zfs_ctldev_init(devp);
5575 5650 mutex_exit(&zfsdev_state_lock);
5576 5651 }
5577 5652
5578 5653 return (error);
5579 5654 }
5580 5655
5581 5656 static int
5582 5657 zfsdev_close(dev_t dev, int flag, int otyp, cred_t *cr)
5583 5658 {
5584 5659 zfs_onexit_t *zo;
5585 5660 minor_t minor = getminor(dev);
5586 5661
5587 5662 if (minor == 0)
5588 5663 return (0);
5589 5664
5590 5665 mutex_enter(&zfsdev_state_lock);
5591 5666 zo = zfsdev_get_soft_state(minor, ZSST_CTLDEV);
5592 5667 if (zo == NULL) {
5593 5668 mutex_exit(&zfsdev_state_lock);
5594 5669 return (zvol_close(dev, flag, otyp, cr));
5595 5670 }
5596 5671 zfs_ctldev_destroy(zo, minor);
5597 5672 mutex_exit(&zfsdev_state_lock);
5598 5673
5599 5674 return (0);
5600 5675 }
5601 5676
5602 5677 static int
5603 5678 zfsdev_ioctl(dev_t dev, int cmd, intptr_t arg, int flag, cred_t *cr, int *rvalp)
5604 5679 {
5605 5680 zfs_cmd_t *zc;
5606 5681 uint_t vecnum;
5607 5682 int error, rc, len;
5608 5683 minor_t minor = getminor(dev);
5609 5684 const zfs_ioc_vec_t *vec;
5610 5685 char *saved_poolname = NULL;
5611 5686 nvlist_t *innvl = NULL;
5612 5687
5613 5688 if (minor != 0 &&
5614 5689 zfsdev_get_soft_state(minor, ZSST_CTLDEV) == NULL)
5615 5690 return (zvol_ioctl(dev, cmd, arg, flag, cr, rvalp));
5616 5691
5617 5692 vecnum = cmd - ZFS_IOC_FIRST;
5618 5693 ASSERT3U(getmajor(dev), ==, ddi_driver_major(zfs_dip));
5619 5694
5620 5695 if (vecnum >= sizeof (zfs_ioc_vec) / sizeof (zfs_ioc_vec[0]))
5621 5696 return (SET_ERROR(EINVAL));
5622 5697 vec = &zfs_ioc_vec[vecnum];
5623 5698
5624 5699 zc = kmem_zalloc(sizeof (zfs_cmd_t), KM_SLEEP);
5625 5700
5626 5701 error = ddi_copyin((void *)arg, zc, sizeof (zfs_cmd_t), flag);
5627 5702 if (error != 0) {
5628 5703 error = SET_ERROR(EFAULT);
5629 5704 goto out;
5630 5705 }
5631 5706
5632 5707 zc->zc_iflags = flag & FKIOCTL;
5633 5708 if (zc->zc_nvlist_src_size != 0) {
5634 5709 error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
5635 5710 zc->zc_iflags, &innvl);
5636 5711 if (error != 0)
5637 5712 goto out;
5638 5713 }
5639 5714
5640 5715 /*
5641 5716 * Ensure that all pool/dataset names are valid before we pass down to
5642 5717 * the lower layers.
5643 5718 */
5644 5719 zc->zc_name[sizeof (zc->zc_name) - 1] = '\0';
5645 5720 switch (vec->zvec_namecheck) {
5646 5721 case POOL_NAME:
5647 5722 if (pool_namecheck(zc->zc_name, NULL, NULL) != 0)
5648 5723 error = SET_ERROR(EINVAL);
5649 5724 else
5650 5725 error = pool_status_check(zc->zc_name,
5651 5726 vec->zvec_namecheck, vec->zvec_pool_check);
5652 5727 break;
5653 5728
5654 5729 case DATASET_NAME:
5655 5730 if (dataset_namecheck(zc->zc_name, NULL, NULL) != 0)
5656 5731 error = SET_ERROR(EINVAL);
5657 5732 else
5658 5733 error = pool_status_check(zc->zc_name,
5659 5734 vec->zvec_namecheck, vec->zvec_pool_check);
5660 5735 break;
5661 5736
5662 5737 case NO_NAME:
5663 5738 break;
5664 5739 }
5665 5740
5666 5741
5667 5742 if (error == 0 && !(flag & FKIOCTL))
5668 5743 error = vec->zvec_secpolicy(zc, innvl, cr);
5669 5744
5670 5745 if (error != 0)
5671 5746 goto out;
5672 5747
5673 5748 /* legacy ioctls can modify zc_name */
5674 5749 len = strcspn(zc->zc_name, "/@") + 1;
5675 5750 saved_poolname = kmem_alloc(len, KM_SLEEP);
5676 5751 (void) strlcpy(saved_poolname, zc->zc_name, len);
5677 5752
5678 5753 if (vec->zvec_func != NULL) {
5679 5754 nvlist_t *outnvl;
5680 5755 int puterror = 0;
5681 5756 spa_t *spa;
5682 5757 nvlist_t *lognv = NULL;
5683 5758
5684 5759 ASSERT(vec->zvec_legacy_func == NULL);
5685 5760
5686 5761 /*
5687 5762 * Add the innvl to the lognv before calling the func,
5688 5763 * in case the func changes the innvl.
5689 5764 */
5690 5765 if (vec->zvec_allow_log) {
5691 5766 lognv = fnvlist_alloc();
5692 5767 fnvlist_add_string(lognv, ZPOOL_HIST_IOCTL,
5693 5768 vec->zvec_name);
5694 5769 if (!nvlist_empty(innvl)) {
5695 5770 fnvlist_add_nvlist(lognv, ZPOOL_HIST_INPUT_NVL,
5696 5771 innvl);
5697 5772 }
5698 5773 }
5699 5774
5700 5775 outnvl = fnvlist_alloc();
5701 5776 error = vec->zvec_func(zc->zc_name, innvl, outnvl);
5702 5777
5703 5778 if (error == 0 && vec->zvec_allow_log &&
5704 5779 spa_open(zc->zc_name, &spa, FTAG) == 0) {
5705 5780 if (!nvlist_empty(outnvl)) {
5706 5781 fnvlist_add_nvlist(lognv, ZPOOL_HIST_OUTPUT_NVL,
5707 5782 outnvl);
5708 5783 }
5709 5784 (void) spa_history_log_nvl(spa, lognv);
5710 5785 spa_close(spa, FTAG);
5711 5786 }
5712 5787 fnvlist_free(lognv);
5713 5788
5714 5789 if (!nvlist_empty(outnvl) || zc->zc_nvlist_dst_size != 0) {
5715 5790 int smusherror = 0;
5716 5791 if (vec->zvec_smush_outnvlist) {
5717 5792 smusherror = nvlist_smush(outnvl,
5718 5793 zc->zc_nvlist_dst_size);
5719 5794 }
5720 5795 if (smusherror == 0)
5721 5796 puterror = put_nvlist(zc, outnvl);
5722 5797 }
5723 5798
5724 5799 if (puterror != 0)
5725 5800 error = puterror;
5726 5801
5727 5802 nvlist_free(outnvl);
5728 5803 } else {
5729 5804 error = vec->zvec_legacy_func(zc);
5730 5805 }
5731 5806
5732 5807 out:
5733 5808 nvlist_free(innvl);
5734 5809 rc = ddi_copyout(zc, (void *)arg, sizeof (zfs_cmd_t), flag);
5735 5810 if (error == 0 && rc != 0)
5736 5811 error = SET_ERROR(EFAULT);
5737 5812 if (error == 0 && vec->zvec_allow_log) {
5738 5813 char *s = tsd_get(zfs_allow_log_key);
5739 5814 if (s != NULL)
5740 5815 strfree(s);
5741 5816 (void) tsd_set(zfs_allow_log_key, saved_poolname);
5742 5817 } else {
5743 5818 if (saved_poolname != NULL)
5744 5819 strfree(saved_poolname);
5745 5820 }
5746 5821
5747 5822 kmem_free(zc, sizeof (zfs_cmd_t));
5748 5823 return (error);
5749 5824 }
5750 5825
5751 5826 static int
5752 5827 zfs_attach(dev_info_t *dip, ddi_attach_cmd_t cmd)
5753 5828 {
5754 5829 if (cmd != DDI_ATTACH)
5755 5830 return (DDI_FAILURE);
5756 5831
5757 5832 if (ddi_create_minor_node(dip, "zfs", S_IFCHR, 0,
5758 5833 DDI_PSEUDO, 0) == DDI_FAILURE)
5759 5834 return (DDI_FAILURE);
5760 5835
5761 5836 zfs_dip = dip;
5762 5837
5763 5838 ddi_report_dev(dip);
5764 5839
5765 5840 return (DDI_SUCCESS);
5766 5841 }
5767 5842
5768 5843 static int
5769 5844 zfs_detach(dev_info_t *dip, ddi_detach_cmd_t cmd)
5770 5845 {
5771 5846 if (spa_busy() || zfs_busy() || zvol_busy())
5772 5847 return (DDI_FAILURE);
5773 5848
5774 5849 if (cmd != DDI_DETACH)
5775 5850 return (DDI_FAILURE);
5776 5851
5777 5852 zfs_dip = NULL;
5778 5853
5779 5854 ddi_prop_remove_all(dip);
5780 5855 ddi_remove_minor_node(dip, NULL);
5781 5856
5782 5857 return (DDI_SUCCESS);
5783 5858 }
5784 5859
5785 5860 /*ARGSUSED*/
5786 5861 static int
5787 5862 zfs_info(dev_info_t *dip, ddi_info_cmd_t infocmd, void *arg, void **result)
5788 5863 {
5789 5864 switch (infocmd) {
5790 5865 case DDI_INFO_DEVT2DEVINFO:
5791 5866 *result = zfs_dip;
5792 5867 return (DDI_SUCCESS);
5793 5868
5794 5869 case DDI_INFO_DEVT2INSTANCE:
5795 5870 *result = (void *)0;
5796 5871 return (DDI_SUCCESS);
5797 5872 }
5798 5873
5799 5874 return (DDI_FAILURE);
5800 5875 }
5801 5876
5802 5877 /*
5803 5878 * OK, so this is a little weird.
5804 5879 *
5805 5880 * /dev/zfs is the control node, i.e. minor 0.
5806 5881 * /dev/zvol/[r]dsk/pool/dataset are the zvols, minor > 0.
5807 5882 *
5808 5883 * /dev/zfs has basically nothing to do except serve up ioctls,
5809 5884 * so most of the standard driver entry points are in zvol.c.
5810 5885 */
5811 5886 static struct cb_ops zfs_cb_ops = {
5812 5887 zfsdev_open, /* open */
5813 5888 zfsdev_close, /* close */
5814 5889 zvol_strategy, /* strategy */
5815 5890 nodev, /* print */
5816 5891 zvol_dump, /* dump */
5817 5892 zvol_read, /* read */
5818 5893 zvol_write, /* write */
5819 5894 zfsdev_ioctl, /* ioctl */
5820 5895 nodev, /* devmap */
5821 5896 nodev, /* mmap */
5822 5897 nodev, /* segmap */
5823 5898 nochpoll, /* poll */
5824 5899 ddi_prop_op, /* prop_op */
5825 5900 NULL, /* streamtab */
5826 5901 D_NEW | D_MP | D_64BIT, /* Driver compatibility flag */
5827 5902 CB_REV, /* version */
5828 5903 nodev, /* async read */
5829 5904 nodev, /* async write */
5830 5905 };
5831 5906
5832 5907 static struct dev_ops zfs_dev_ops = {
5833 5908 DEVO_REV, /* version */
5834 5909 0, /* refcnt */
5835 5910 zfs_info, /* info */
5836 5911 nulldev, /* identify */
5837 5912 nulldev, /* probe */
5838 5913 zfs_attach, /* attach */
5839 5914 zfs_detach, /* detach */
5840 5915 nodev, /* reset */
5841 5916 &zfs_cb_ops, /* driver operations */
5842 5917 NULL, /* no bus operations */
5843 5918 NULL, /* power */
5844 5919 ddi_quiesce_not_needed, /* quiesce */
5845 5920 };
5846 5921
5847 5922 static struct modldrv zfs_modldrv = {
5848 5923 &mod_driverops,
5849 5924 "ZFS storage pool",
5850 5925 &zfs_dev_ops
5851 5926 };
5852 5927
5853 5928 static struct modlinkage modlinkage = {
5854 5929 MODREV_1,
5855 5930 (void *)&zfs_modlfs,
5856 5931 (void *)&zfs_modldrv,
5857 5932 NULL
5858 5933 };
5859 5934
5860 5935 static void
5861 5936 zfs_allow_log_destroy(void *arg)
5862 5937 {
5863 5938 char *poolname = arg;
5864 5939 strfree(poolname);
5865 5940 }
5866 5941
5867 5942 int
5868 5943 _init(void)
5869 5944 {
5870 5945 int error;
5871 5946
5872 5947 spa_init(FREAD | FWRITE);
5873 5948 zfs_init();
5874 5949 zvol_init();
5875 5950 zfs_ioctl_init();
5876 5951
5877 5952 if ((error = mod_install(&modlinkage)) != 0) {
5878 5953 zvol_fini();
5879 5954 zfs_fini();
5880 5955 spa_fini();
5881 5956 return (error);
5882 5957 }
5883 5958
5884 5959 tsd_create(&zfs_fsyncer_key, NULL);
5885 5960 tsd_create(&rrw_tsd_key, rrw_tsd_destroy);
5886 5961 tsd_create(&zfs_allow_log_key, zfs_allow_log_destroy);
5887 5962
5888 5963 error = ldi_ident_from_mod(&modlinkage, &zfs_li);
5889 5964 ASSERT(error == 0);
5890 5965 mutex_init(&zfs_share_lock, NULL, MUTEX_DEFAULT, NULL);
5891 5966
5892 5967 return (0);
5893 5968 }
5894 5969
5895 5970 int
5896 5971 _fini(void)
5897 5972 {
5898 5973 int error;
5899 5974
5900 5975 if (spa_busy() || zfs_busy() || zvol_busy() || zio_injection_enabled)
5901 5976 return (SET_ERROR(EBUSY));
5902 5977
5903 5978 if ((error = mod_remove(&modlinkage)) != 0)
5904 5979 return (error);
5905 5980
5906 5981 zvol_fini();
5907 5982 zfs_fini();
5908 5983 spa_fini();
5909 5984 if (zfs_nfsshare_inited)
5910 5985 (void) ddi_modclose(nfs_mod);
5911 5986 if (zfs_smbshare_inited)
5912 5987 (void) ddi_modclose(smbsrv_mod);
5913 5988 if (zfs_nfsshare_inited || zfs_smbshare_inited)
5914 5989 (void) ddi_modclose(sharefs_mod);
5915 5990
5916 5991 tsd_destroy(&zfs_fsyncer_key);
5917 5992 ldi_ident_release(zfs_li);
5918 5993 zfs_li = NULL;
5919 5994 mutex_destroy(&zfs_share_lock);
5920 5995
5921 5996 return (error);
5922 5997 }
5923 5998
5924 5999 int
5925 6000 _info(struct modinfo *modinfop)
5926 6001 {
5927 6002 return (mod_info(&modlinkage, modinfop));
5928 6003 }
|
↓ open down ↓ |
2199 lines elided |
↑ open up ↑ |
XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX