?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
1949 crash during reguid causes stale config
[unleashed.git] / usr / src / uts / common / fs / zfs / zfs_acl.c
blob2b93fc8329f9a7a3cff99bf5e534c690f3f4af3a
1 /*
2 * CDDL HEADER START
3 *
4 * The contents of this file are subject to the terms of the
5 * Common Development and Distribution License (the "License").
6 * You may not use this file except in compliance with the License.
7 *
8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9 * or http://www.opensolaris.org/os/licensing.
10 * See the License for the specific language governing permissions
11 * and limitations under the License.
12 *
13 * When distributing Covered Code, include this CDDL HEADER in each
14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15 * If applicable, add the following below this CDDL HEADER, with the
16 * fields enclosed by brackets "[]" replaced with your own identifying
17 * information: Portions Copyright [yyyy] [name of copyright owner]
18 *
19 * CDDL HEADER END
20 */
21 /*
22 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
23 * Copyright 2011 Nexenta Systems, Inc. All rights reserved.
24 */
25
26 #include <sys/types.h>
27 #include <sys/param.h>
28 #include <sys/time.h>
29 #include <sys/systm.h>
30 #include <sys/sysmacros.h>
31 #include <sys/resource.h>
32 #include <sys/vfs.h>
33 #include <sys/vnode.h>
34 #include <sys/sid.h>
35 #include <sys/file.h>
36 #include <sys/stat.h>
37 #include <sys/kmem.h>
38 #include <sys/cmn_err.h>
39 #include <sys/errno.h>
40 #include <sys/unistd.h>
41 #include <sys/sdt.h>
42 #include <sys/fs/zfs.h>
43 #include <sys/mode.h>
44 #include <sys/policy.h>
45 #include <sys/zfs_znode.h>
46 #include <sys/zfs_fuid.h>
47 #include <sys/zfs_acl.h>
48 #include <sys/zfs_dir.h>
49 #include <sys/zfs_vfsops.h>
50 #include <sys/dmu.h>
51 #include <sys/dnode.h>
52 #include <sys/zap.h>
53 #include <sys/sa.h>
54 #include "fs/fs_subr.h"
55 #include <acl/acl_common.h>
56
57 #define ALLOW ACE_ACCESS_ALLOWED_ACE_TYPE
58 #define DENY ACE_ACCESS_DENIED_ACE_TYPE
59 #define MAX_ACE_TYPE ACE_SYSTEM_ALARM_CALLBACK_OBJECT_ACE_TYPE
60 #define MIN_ACE_TYPE ALLOW
61
62 #define OWNING_GROUP (ACE_GROUP|ACE_IDENTIFIER_GROUP)
63 #define EVERYONE_ALLOW_MASK (ACE_READ_ACL|ACE_READ_ATTRIBUTES | \
64 ACE_READ_NAMED_ATTRS|ACE_SYNCHRONIZE)
65 #define EVERYONE_DENY_MASK (ACE_WRITE_ACL|ACE_WRITE_OWNER | \
66 ACE_WRITE_ATTRIBUTES|ACE_WRITE_NAMED_ATTRS)
67 #define OWNER_ALLOW_MASK (ACE_WRITE_ACL | ACE_WRITE_OWNER | \
68 ACE_WRITE_ATTRIBUTES|ACE_WRITE_NAMED_ATTRS)
69
70 #define ZFS_CHECKED_MASKS (ACE_READ_ACL|ACE_READ_ATTRIBUTES|ACE_READ_DATA| \
71 ACE_READ_NAMED_ATTRS|ACE_WRITE_DATA|ACE_WRITE_ATTRIBUTES| \
72 ACE_WRITE_NAMED_ATTRS|ACE_APPEND_DATA|ACE_EXECUTE|ACE_WRITE_OWNER| \
73 ACE_WRITE_ACL|ACE_DELETE|ACE_DELETE_CHILD|ACE_SYNCHRONIZE)
74
75 #define WRITE_MASK_DATA (ACE_WRITE_DATA|ACE_APPEND_DATA|ACE_WRITE_NAMED_ATTRS)
76 #define WRITE_MASK_ATTRS (ACE_WRITE_ACL|ACE_WRITE_OWNER|ACE_WRITE_ATTRIBUTES| \
77 ACE_DELETE|ACE_DELETE_CHILD)
78 #define WRITE_MASK (WRITE_MASK_DATA|WRITE_MASK_ATTRS)
79
80 #define OGE_CLEAR (ACE_READ_DATA|ACE_LIST_DIRECTORY|ACE_WRITE_DATA| \
81 ACE_ADD_FILE|ACE_APPEND_DATA|ACE_ADD_SUBDIRECTORY|ACE_EXECUTE)
82
83 #define OKAY_MASK_BITS (ACE_READ_DATA|ACE_LIST_DIRECTORY|ACE_WRITE_DATA| \
84 ACE_ADD_FILE|ACE_APPEND_DATA|ACE_ADD_SUBDIRECTORY|ACE_EXECUTE)
85
86 #define ALL_INHERIT (ACE_FILE_INHERIT_ACE|ACE_DIRECTORY_INHERIT_ACE | \
87 ACE_NO_PROPAGATE_INHERIT_ACE|ACE_INHERIT_ONLY_ACE|ACE_INHERITED_ACE)
88
89 #define RESTRICTED_CLEAR (ACE_WRITE_ACL|ACE_WRITE_OWNER)
90
91 #define V4_ACL_WIDE_FLAGS (ZFS_ACL_AUTO_INHERIT|ZFS_ACL_DEFAULTED|\
92 ZFS_ACL_PROTECTED)
93
94 #define ZFS_ACL_WIDE_FLAGS (V4_ACL_WIDE_FLAGS|ZFS_ACL_TRIVIAL|ZFS_INHERIT_ACE|\
95 ZFS_ACL_OBJ_ACE)
96
97 #define ALL_MODE_EXECS (S_IXUSR | S_IXGRP | S_IXOTH)
98
99 static uint16_t
100 zfs_ace_v0_get_type(void *acep)
101 {
102 return (((zfs_oldace_t *)acep)->z_type);
103 }
104
105 static uint16_t
106 zfs_ace_v0_get_flags(void *acep)
107 {
108 return (((zfs_oldace_t *)acep)->z_flags);
109 }
110
111 static uint32_t
112 zfs_ace_v0_get_mask(void *acep)
113 {
114 return (((zfs_oldace_t *)acep)->z_access_mask);
115 }
116
117 static uint64_t
118 zfs_ace_v0_get_who(void *acep)
119 {
120 return (((zfs_oldace_t *)acep)->z_fuid);
121 }
122
123 static void
124 zfs_ace_v0_set_type(void *acep, uint16_t type)
125 {
126 ((zfs_oldace_t *)acep)->z_type = type;
127 }
128
129 static void
130 zfs_ace_v0_set_flags(void *acep, uint16_t flags)
131 {
132 ((zfs_oldace_t *)acep)->z_flags = flags;
133 }
134
135 static void
136 zfs_ace_v0_set_mask(void *acep, uint32_t mask)
137 {
138 ((zfs_oldace_t *)acep)->z_access_mask = mask;
139 }
140
141 static void
142 zfs_ace_v0_set_who(void *acep, uint64_t who)
143 {
144 ((zfs_oldace_t *)acep)->z_fuid = who;
145 }
146
147 /*ARGSUSED*/
148 static size_t
149 zfs_ace_v0_size(void *acep)
150 {
151 return (sizeof (zfs_oldace_t));
152 }
153
154 static size_t
155 zfs_ace_v0_abstract_size(void)
156 {
157 return (sizeof (zfs_oldace_t));
158 }
159
160 static int
161 zfs_ace_v0_mask_off(void)
162 {
163 return (offsetof(zfs_oldace_t, z_access_mask));
164 }
165
166 /*ARGSUSED*/
167 static int
168 zfs_ace_v0_data(void *acep, void **datap)
169 {
170 *datap = NULL;
171 return (0);
172 }
173
174 static acl_ops_t zfs_acl_v0_ops = {
175 zfs_ace_v0_get_mask,
176 zfs_ace_v0_set_mask,
177 zfs_ace_v0_get_flags,
178 zfs_ace_v0_set_flags,
179 zfs_ace_v0_get_type,
180 zfs_ace_v0_set_type,
181 zfs_ace_v0_get_who,
182 zfs_ace_v0_set_who,
183 zfs_ace_v0_size,
184 zfs_ace_v0_abstract_size,
185 zfs_ace_v0_mask_off,
186 zfs_ace_v0_data
187 };
188
189 static uint16_t
190 zfs_ace_fuid_get_type(void *acep)
191 {
192 return (((zfs_ace_hdr_t *)acep)->z_type);
193 }
194
195 static uint16_t
196 zfs_ace_fuid_get_flags(void *acep)
197 {
198 return (((zfs_ace_hdr_t *)acep)->z_flags);
199 }
200
201 static uint32_t
202 zfs_ace_fuid_get_mask(void *acep)
203 {
204 return (((zfs_ace_hdr_t *)acep)->z_access_mask);
205 }
206
207 static uint64_t
208 zfs_ace_fuid_get_who(void *args)
209 {
210 uint16_t entry_type;
211 zfs_ace_t *acep = args;
212
213 entry_type = acep->z_hdr.z_flags & ACE_TYPE_FLAGS;
214
215 if (entry_type == ACE_OWNER || entry_type == OWNING_GROUP ||
216 entry_type == ACE_EVERYONE)
217 return (-1);
218 return (((zfs_ace_t *)acep)->z_fuid);
219 }
220
221 static void
222 zfs_ace_fuid_set_type(void *acep, uint16_t type)
223 {
224 ((zfs_ace_hdr_t *)acep)->z_type = type;
225 }
226
227 static void
228 zfs_ace_fuid_set_flags(void *acep, uint16_t flags)
229 {
230 ((zfs_ace_hdr_t *)acep)->z_flags = flags;
231 }
232
233 static void
234 zfs_ace_fuid_set_mask(void *acep, uint32_t mask)
235 {
236 ((zfs_ace_hdr_t *)acep)->z_access_mask = mask;
237 }
238
239 static void
240 zfs_ace_fuid_set_who(void *arg, uint64_t who)
241 {
242 zfs_ace_t *acep = arg;
243
244 uint16_t entry_type = acep->z_hdr.z_flags & ACE_TYPE_FLAGS;
245
246 if (entry_type == ACE_OWNER || entry_type == OWNING_GROUP ||
247 entry_type == ACE_EVERYONE)
248 return;
249 acep->z_fuid = who;
250 }
251
252 static size_t
253 zfs_ace_fuid_size(void *acep)
254 {
255 zfs_ace_hdr_t *zacep = acep;
256 uint16_t entry_type;
257
258 switch (zacep->z_type) {
259 case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE:
260 case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE:
261 case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE:
262 case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE:
263 return (sizeof (zfs_object_ace_t));
264 case ALLOW:
265 case DENY:
266 entry_type =
267 (((zfs_ace_hdr_t *)acep)->z_flags & ACE_TYPE_FLAGS);
268 if (entry_type == ACE_OWNER ||
269 entry_type == OWNING_GROUP ||
270 entry_type == ACE_EVERYONE)
271 return (sizeof (zfs_ace_hdr_t));
272 /*FALLTHROUGH*/
273 default:
274 return (sizeof (zfs_ace_t));
275 }
276 }
277
278 static size_t
279 zfs_ace_fuid_abstract_size(void)
280 {
281 return (sizeof (zfs_ace_hdr_t));
282 }
283
284 static int
285 zfs_ace_fuid_mask_off(void)
286 {
287 return (offsetof(zfs_ace_hdr_t, z_access_mask));
288 }
289
290 static int
291 zfs_ace_fuid_data(void *acep, void **datap)
292 {
293 zfs_ace_t *zacep = acep;
294 zfs_object_ace_t *zobjp;
295
296 switch (zacep->z_hdr.z_type) {
297 case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE:
298 case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE:
299 case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE:
300 case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE:
301 zobjp = acep;
302 *datap = (caddr_t)zobjp + sizeof (zfs_ace_t);
303 return (sizeof (zfs_object_ace_t) - sizeof (zfs_ace_t));
304 default:
305 *datap = NULL;
306 return (0);
307 }
308 }
309
310 static acl_ops_t zfs_acl_fuid_ops = {
311 zfs_ace_fuid_get_mask,
312 zfs_ace_fuid_set_mask,
313 zfs_ace_fuid_get_flags,
314 zfs_ace_fuid_set_flags,
315 zfs_ace_fuid_get_type,
316 zfs_ace_fuid_set_type,
317 zfs_ace_fuid_get_who,
318 zfs_ace_fuid_set_who,
319 zfs_ace_fuid_size,
320 zfs_ace_fuid_abstract_size,
321 zfs_ace_fuid_mask_off,
322 zfs_ace_fuid_data
323 };
324
325 /*
326 * The following three functions are provided for compatibility with
327 * older ZPL version in order to determine if the file use to have
328 * an external ACL and what version of ACL previously existed on the
329 * file. Would really be nice to not need this, sigh.
330 */
331 uint64_t
332 zfs_external_acl(znode_t *zp)
333 {
334 zfs_acl_phys_t acl_phys;
335 int error;
336
337 if (zp->z_is_sa)
338 return (0);
339
340 /*
341 * Need to deal with a potential
342 * race where zfs_sa_upgrade could cause
343 * z_isa_sa to change.
344 *
345 * If the lookup fails then the state of z_is_sa should have
346 * changed.
347 */
348
349 if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_ZNODE_ACL(zp->z_zfsvfs),
350 &acl_phys, sizeof (acl_phys))) == 0)
351 return (acl_phys.z_acl_extern_obj);
352 else {
353 /*
354 * after upgrade the SA_ZPL_ZNODE_ACL should have been
355 * removed
356 */
357 VERIFY(zp->z_is_sa && error == ENOENT);
358 return (0);
359 }
360 }
361
362 /*
363 * Determine size of ACL in bytes
364 *
365 * This is more complicated than it should be since we have to deal
366 * with old external ACLs.
367 */
368 static int
369 zfs_acl_znode_info(znode_t *zp, int *aclsize, int *aclcount,
370 zfs_acl_phys_t *aclphys)
371 {
372 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
373 uint64_t acl_count;
374 int size;
375 int error;
376
377 ASSERT(MUTEX_HELD(&zp->z_acl_lock));
378 if (zp->z_is_sa) {
379 if ((error = sa_size(zp->z_sa_hdl, SA_ZPL_DACL_ACES(zfsvfs),
380 &size)) != 0)
381 return (error);
382 *aclsize = size;
383 if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_DACL_COUNT(zfsvfs),
384 &acl_count, sizeof (acl_count))) != 0)
385 return (error);
386 *aclcount = acl_count;
387 } else {
388 if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_ZNODE_ACL(zfsvfs),
389 aclphys, sizeof (*aclphys))) != 0)
390 return (error);
391
392 if (aclphys->z_acl_version == ZFS_ACL_VERSION_INITIAL) {
393 *aclsize = ZFS_ACL_SIZE(aclphys->z_acl_size);
394 *aclcount = aclphys->z_acl_size;
395 } else {
396 *aclsize = aclphys->z_acl_size;
397 *aclcount = aclphys->z_acl_count;
398 }
399 }
400 return (0);
401 }
402
403 int
404 zfs_znode_acl_version(znode_t *zp)
405 {
406 zfs_acl_phys_t acl_phys;
407
408 if (zp->z_is_sa)
409 return (ZFS_ACL_VERSION_FUID);
410 else {
411 int error;
412
413 /*
414 * Need to deal with a potential
415 * race where zfs_sa_upgrade could cause
416 * z_isa_sa to change.
417 *
418 * If the lookup fails then the state of z_is_sa should have
419 * changed.
420 */
421 if ((error = sa_lookup(zp->z_sa_hdl,
422 SA_ZPL_ZNODE_ACL(zp->z_zfsvfs),
423 &acl_phys, sizeof (acl_phys))) == 0)
424 return (acl_phys.z_acl_version);
425 else {
426 /*
427 * After upgrade SA_ZPL_ZNODE_ACL should have
428 * been removed.
429 */
430 VERIFY(zp->z_is_sa && error == ENOENT);
431 return (ZFS_ACL_VERSION_FUID);
432 }
433 }
434 }
435
436 static int
437 zfs_acl_version(int version)
438 {
439 if (version < ZPL_VERSION_FUID)
440 return (ZFS_ACL_VERSION_INITIAL);
441 else
442 return (ZFS_ACL_VERSION_FUID);
443 }
444
445 static int
446 zfs_acl_version_zp(znode_t *zp)
447 {
448 return (zfs_acl_version(zp->z_zfsvfs->z_version));
449 }
450
451 zfs_acl_t *
452 zfs_acl_alloc(int vers)
453 {
454 zfs_acl_t *aclp;
455
456 aclp = kmem_zalloc(sizeof (zfs_acl_t), KM_SLEEP);
457 list_create(&aclp->z_acl, sizeof (zfs_acl_node_t),
458 offsetof(zfs_acl_node_t, z_next));
459 aclp->z_version = vers;
460 if (vers == ZFS_ACL_VERSION_FUID)
461 aclp->z_ops = zfs_acl_fuid_ops;
462 else
463 aclp->z_ops = zfs_acl_v0_ops;
464 return (aclp);
465 }
466
467 zfs_acl_node_t *
468 zfs_acl_node_alloc(size_t bytes)
469 {
470 zfs_acl_node_t *aclnode;
471
472 aclnode = kmem_zalloc(sizeof (zfs_acl_node_t), KM_SLEEP);
473 if (bytes) {
474 aclnode->z_acldata = kmem_alloc(bytes, KM_SLEEP);
475 aclnode->z_allocdata = aclnode->z_acldata;
476 aclnode->z_allocsize = bytes;
477 aclnode->z_size = bytes;
478 }
479
480 return (aclnode);
481 }
482
483 static void
484 zfs_acl_node_free(zfs_acl_node_t *aclnode)
485 {
486 if (aclnode->z_allocsize)
487 kmem_free(aclnode->z_allocdata, aclnode->z_allocsize);
488 kmem_free(aclnode, sizeof (zfs_acl_node_t));
489 }
490
491 static void
492 zfs_acl_release_nodes(zfs_acl_t *aclp)
493 {
494 zfs_acl_node_t *aclnode;
495
496 while (aclnode = list_head(&aclp->z_acl)) {
497 list_remove(&aclp->z_acl, aclnode);
498 zfs_acl_node_free(aclnode);
499 }
500 aclp->z_acl_count = 0;
501 aclp->z_acl_bytes = 0;
502 }
503
504 void
505 zfs_acl_free(zfs_acl_t *aclp)
506 {
507 zfs_acl_release_nodes(aclp);
508 list_destroy(&aclp->z_acl);
509 kmem_free(aclp, sizeof (zfs_acl_t));
510 }
511
512 static boolean_t
513 zfs_acl_valid_ace_type(uint_t type, uint_t flags)
514 {
515 uint16_t entry_type;
516
517 switch (type) {
518 case ALLOW:
519 case DENY:
520 case ACE_SYSTEM_AUDIT_ACE_TYPE:
521 case ACE_SYSTEM_ALARM_ACE_TYPE:
522 entry_type = flags & ACE_TYPE_FLAGS;
523 return (entry_type == ACE_OWNER ||
524 entry_type == OWNING_GROUP ||
525 entry_type == ACE_EVERYONE || entry_type == 0 ||
526 entry_type == ACE_IDENTIFIER_GROUP);
527 default:
528 if (type >= MIN_ACE_TYPE && type <= MAX_ACE_TYPE)
529 return (B_TRUE);
530 }
531 return (B_FALSE);
532 }
533
534 static boolean_t
535 zfs_ace_valid(vtype_t obj_type, zfs_acl_t *aclp, uint16_t type, uint16_t iflags)
536 {
537 /*
538 * first check type of entry
539 */
540
541 if (!zfs_acl_valid_ace_type(type, iflags))
542 return (B_FALSE);
543
544 switch (type) {
545 case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE:
546 case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE:
547 case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE:
548 case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE:
549 if (aclp->z_version < ZFS_ACL_VERSION_FUID)
550 return (B_FALSE);
551 aclp->z_hints |= ZFS_ACL_OBJ_ACE;
552 }
553
554 /*
555 * next check inheritance level flags
556 */
557
558 if (obj_type == VDIR &&
559 (iflags & (ACE_FILE_INHERIT_ACE|ACE_DIRECTORY_INHERIT_ACE)))
560 aclp->z_hints |= ZFS_INHERIT_ACE;
561
562 if (iflags & (ACE_INHERIT_ONLY_ACE|ACE_NO_PROPAGATE_INHERIT_ACE)) {
563 if ((iflags & (ACE_FILE_INHERIT_ACE|
564 ACE_DIRECTORY_INHERIT_ACE)) == 0) {
565 return (B_FALSE);
566 }
567 }
568
569 return (B_TRUE);
570 }
571
572 static void *
573 zfs_acl_next_ace(zfs_acl_t *aclp, void *start, uint64_t *who,
574 uint32_t *access_mask, uint16_t *iflags, uint16_t *type)
575 {
576 zfs_acl_node_t *aclnode;
577
578 ASSERT(aclp);
579
580 if (start == NULL) {
581 aclnode = list_head(&aclp->z_acl);
582 if (aclnode == NULL)
583 return (NULL);
584
585 aclp->z_next_ace = aclnode->z_acldata;
586 aclp->z_curr_node = aclnode;
587 aclnode->z_ace_idx = 0;
588 }
589
590 aclnode = aclp->z_curr_node;
591
592 if (aclnode == NULL)
593 return (NULL);
594
595 if (aclnode->z_ace_idx >= aclnode->z_ace_count) {
596 aclnode = list_next(&aclp->z_acl, aclnode);
597 if (aclnode == NULL)
598 return (NULL);
599 else {
600 aclp->z_curr_node = aclnode;
601 aclnode->z_ace_idx = 0;
602 aclp->z_next_ace = aclnode->z_acldata;
603 }
604 }
605
606 if (aclnode->z_ace_idx < aclnode->z_ace_count) {
607 void *acep = aclp->z_next_ace;
608 size_t ace_size;
609
610 /*
611 * Make sure we don't overstep our bounds
612 */
613 ace_size = aclp->z_ops.ace_size(acep);
614
615 if (((caddr_t)acep + ace_size) >
616 ((caddr_t)aclnode->z_acldata + aclnode->z_size)) {
617 return (NULL);
618 }
619
620 *iflags = aclp->z_ops.ace_flags_get(acep);
621 *type = aclp->z_ops.ace_type_get(acep);
622 *access_mask = aclp->z_ops.ace_mask_get(acep);
623 *who = aclp->z_ops.ace_who_get(acep);
624 aclp->z_next_ace = (caddr_t)aclp->z_next_ace + ace_size;
625 aclnode->z_ace_idx++;
626
627 return ((void *)acep);
628 }
629 return (NULL);
630 }
631
632 /*ARGSUSED*/
633 static uint64_t
634 zfs_ace_walk(void *datap, uint64_t cookie, int aclcnt,
635 uint16_t *flags, uint16_t *type, uint32_t *mask)
636 {
637 zfs_acl_t *aclp = datap;
638 zfs_ace_hdr_t *acep = (zfs_ace_hdr_t *)(uintptr_t)cookie;
639 uint64_t who;
640
641 acep = zfs_acl_next_ace(aclp, acep, &who, mask,
642 flags, type);
643 return ((uint64_t)(uintptr_t)acep);
644 }
645
646 static zfs_acl_node_t *
647 zfs_acl_curr_node(zfs_acl_t *aclp)
648 {
649 ASSERT(aclp->z_curr_node);
650 return (aclp->z_curr_node);
651 }
652
653 /*
654 * Copy ACE to internal ZFS format.
655 * While processing the ACL each ACE will be validated for correctness.
656 * ACE FUIDs will be created later.
657 */
658 int
659 zfs_copy_ace_2_fuid(zfsvfs_t *zfsvfs, vtype_t obj_type, zfs_acl_t *aclp,
660 void *datap, zfs_ace_t *z_acl, uint64_t aclcnt, size_t *size,
661 zfs_fuid_info_t **fuidp, cred_t *cr)
662 {
663 int i;
664 uint16_t entry_type;
665 zfs_ace_t *aceptr = z_acl;
666 ace_t *acep = datap;
667 zfs_object_ace_t *zobjacep;
668 ace_object_t *aceobjp;
669
670 for (i = 0; i != aclcnt; i++) {
671 aceptr->z_hdr.z_access_mask = acep->a_access_mask;
672 aceptr->z_hdr.z_flags = acep->a_flags;
673 aceptr->z_hdr.z_type = acep->a_type;
674 entry_type = aceptr->z_hdr.z_flags & ACE_TYPE_FLAGS;
675 if (entry_type != ACE_OWNER && entry_type != OWNING_GROUP &&
676 entry_type != ACE_EVERYONE) {
677 aceptr->z_fuid = zfs_fuid_create(zfsvfs, acep->a_who,
678 cr, (entry_type == 0) ?
679 ZFS_ACE_USER : ZFS_ACE_GROUP, fuidp);
680 }
681
682 /*
683 * Make sure ACE is valid
684 */
685 if (zfs_ace_valid(obj_type, aclp, aceptr->z_hdr.z_type,
686 aceptr->z_hdr.z_flags) != B_TRUE)
687 return (EINVAL);
688
689 switch (acep->a_type) {
690 case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE:
691 case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE:
692 case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE:
693 case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE:
694 zobjacep = (zfs_object_ace_t *)aceptr;
695 aceobjp = (ace_object_t *)acep;
696
697 bcopy(aceobjp->a_obj_type, zobjacep->z_object_type,
698 sizeof (aceobjp->a_obj_type));
699 bcopy(aceobjp->a_inherit_obj_type,
700 zobjacep->z_inherit_type,
701 sizeof (aceobjp->a_inherit_obj_type));
702 acep = (ace_t *)((caddr_t)acep + sizeof (ace_object_t));
703 break;
704 default:
705 acep = (ace_t *)((caddr_t)acep + sizeof (ace_t));
706 }
707
708 aceptr = (zfs_ace_t *)((caddr_t)aceptr +
709 aclp->z_ops.ace_size(aceptr));
710 }
711
712 *size = (caddr_t)aceptr - (caddr_t)z_acl;
713
714 return (0);
715 }
716
717 /*
718 * Copy ZFS ACEs to fixed size ace_t layout
719 */
720 static void
721 zfs_copy_fuid_2_ace(zfsvfs_t *zfsvfs, zfs_acl_t *aclp, cred_t *cr,
722 void *datap, int filter)
723 {
724 uint64_t who;
725 uint32_t access_mask;
726 uint16_t iflags, type;
727 zfs_ace_hdr_t *zacep = NULL;
728 ace_t *acep = datap;
729 ace_object_t *objacep;
730 zfs_object_ace_t *zobjacep;
731 size_t ace_size;
732 uint16_t entry_type;
733
734 while (zacep = zfs_acl_next_ace(aclp, zacep,
735 &who, &access_mask, &iflags, &type)) {
736
737 switch (type) {
738 case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE:
739 case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE:
740 case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE:
741 case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE:
742 if (filter) {
743 continue;
744 }
745 zobjacep = (zfs_object_ace_t *)zacep;
746 objacep = (ace_object_t *)acep;
747 bcopy(zobjacep->z_object_type,
748 objacep->a_obj_type,
749 sizeof (zobjacep->z_object_type));
750 bcopy(zobjacep->z_inherit_type,
751 objacep->a_inherit_obj_type,
752 sizeof (zobjacep->z_inherit_type));
753 ace_size = sizeof (ace_object_t);
754 break;
755 default:
756 ace_size = sizeof (ace_t);
757 break;
758 }
759
760 entry_type = (iflags & ACE_TYPE_FLAGS);
761 if ((entry_type != ACE_OWNER &&
762 entry_type != OWNING_GROUP &&
763 entry_type != ACE_EVERYONE)) {
764 acep->a_who = zfs_fuid_map_id(zfsvfs, who,
765 cr, (entry_type & ACE_IDENTIFIER_GROUP) ?
766 ZFS_ACE_GROUP : ZFS_ACE_USER);
767 } else {
768 acep->a_who = (uid_t)(int64_t)who;
769 }
770 acep->a_access_mask = access_mask;
771 acep->a_flags = iflags;
772 acep->a_type = type;
773 acep = (ace_t *)((caddr_t)acep + ace_size);
774 }
775 }
776
777 static int
778 zfs_copy_ace_2_oldace(vtype_t obj_type, zfs_acl_t *aclp, ace_t *acep,
779 zfs_oldace_t *z_acl, int aclcnt, size_t *size)
780 {
781 int i;
782 zfs_oldace_t *aceptr = z_acl;
783
784 for (i = 0; i != aclcnt; i++, aceptr++) {
785 aceptr->z_access_mask = acep[i].a_access_mask;
786 aceptr->z_type = acep[i].a_type;
787 aceptr->z_flags = acep[i].a_flags;
788 aceptr->z_fuid = acep[i].a_who;
789 /*
790 * Make sure ACE is valid
791 */
792 if (zfs_ace_valid(obj_type, aclp, aceptr->z_type,
793 aceptr->z_flags) != B_TRUE)
794 return (EINVAL);
795 }
796 *size = (caddr_t)aceptr - (caddr_t)z_acl;
797 return (0);
798 }
799
800 /*
801 * convert old ACL format to new
802 */
803 void
804 zfs_acl_xform(znode_t *zp, zfs_acl_t *aclp, cred_t *cr)
805 {
806 zfs_oldace_t *oldaclp;
807 int i;
808 uint16_t type, iflags;
809 uint32_t access_mask;
810 uint64_t who;
811 void *cookie = NULL;
812 zfs_acl_node_t *newaclnode;
813
814 ASSERT(aclp->z_version == ZFS_ACL_VERSION_INITIAL);
815 /*
816 * First create the ACE in a contiguous piece of memory
817 * for zfs_copy_ace_2_fuid().
818 *
819 * We only convert an ACL once, so this won't happen
820 * everytime.
821 */
822 oldaclp = kmem_alloc(sizeof (zfs_oldace_t) * aclp->z_acl_count,
823 KM_SLEEP);
824 i = 0;
825 while (cookie = zfs_acl_next_ace(aclp, cookie, &who,
826 &access_mask, &iflags, &type)) {
827 oldaclp[i].z_flags = iflags;
828 oldaclp[i].z_type = type;
829 oldaclp[i].z_fuid = who;
830 oldaclp[i++].z_access_mask = access_mask;
831 }
832
833 newaclnode = zfs_acl_node_alloc(aclp->z_acl_count *
834 sizeof (zfs_object_ace_t));
835 aclp->z_ops = zfs_acl_fuid_ops;
836 VERIFY(zfs_copy_ace_2_fuid(zp->z_zfsvfs, ZTOV(zp)->v_type, aclp,
837 oldaclp, newaclnode->z_acldata, aclp->z_acl_count,
838 &newaclnode->z_size, NULL, cr) == 0);
839 newaclnode->z_ace_count = aclp->z_acl_count;
840 aclp->z_version = ZFS_ACL_VERSION;
841 kmem_free(oldaclp, aclp->z_acl_count * sizeof (zfs_oldace_t));
842
843 /*
844 * Release all previous ACL nodes
845 */
846
847 zfs_acl_release_nodes(aclp);
848
849 list_insert_head(&aclp->z_acl, newaclnode);
850
851 aclp->z_acl_bytes = newaclnode->z_size;
852 aclp->z_acl_count = newaclnode->z_ace_count;
853
854 }
855
856 /*
857 * Convert unix access mask to v4 access mask
858 */
859 static uint32_t
860 zfs_unix_to_v4(uint32_t access_mask)
861 {
862 uint32_t new_mask = 0;
863
864 if (access_mask & S_IXOTH)
865 new_mask |= ACE_EXECUTE;
866 if (access_mask & S_IWOTH)
867 new_mask |= ACE_WRITE_DATA;
868 if (access_mask & S_IROTH)
869 new_mask |= ACE_READ_DATA;
870 return (new_mask);
871 }
872
873 static void
874 zfs_set_ace(zfs_acl_t *aclp, void *acep, uint32_t access_mask,
875 uint16_t access_type, uint64_t fuid, uint16_t entry_type)
876 {
877 uint16_t type = entry_type & ACE_TYPE_FLAGS;
878
879 aclp->z_ops.ace_mask_set(acep, access_mask);
880 aclp->z_ops.ace_type_set(acep, access_type);
881 aclp->z_ops.ace_flags_set(acep, entry_type);
882 if ((type != ACE_OWNER && type != OWNING_GROUP &&
883 type != ACE_EVERYONE))
884 aclp->z_ops.ace_who_set(acep, fuid);
885 }
886
887 /*
888 * Determine mode of file based on ACL.
889 * Also, create FUIDs for any User/Group ACEs
890 */
891 uint64_t
892 zfs_mode_compute(uint64_t fmode, zfs_acl_t *aclp,
893 uint64_t *pflags, uint64_t fuid, uint64_t fgid)
894 {
895 int entry_type;
896 mode_t mode;
897 mode_t seen = 0;
898 zfs_ace_hdr_t *acep = NULL;
899 uint64_t who;
900 uint16_t iflags, type;
901 uint32_t access_mask;
902 boolean_t an_exec_denied = B_FALSE;
903
904 mode = (fmode & (S_IFMT | S_ISUID | S_ISGID | S_ISVTX));
905
906 while (acep = zfs_acl_next_ace(aclp, acep, &who,
907 &access_mask, &iflags, &type)) {
908
909 if (!zfs_acl_valid_ace_type(type, iflags))
910 continue;
911
912 entry_type = (iflags & ACE_TYPE_FLAGS);
913
914 /*
915 * Skip over owner@, group@ or everyone@ inherit only ACEs
916 */
917 if ((iflags & ACE_INHERIT_ONLY_ACE) &&
918 (entry_type == ACE_OWNER || entry_type == ACE_EVERYONE ||
919 entry_type == OWNING_GROUP))
920 continue;
921
922 if (entry_type == ACE_OWNER || (entry_type == 0 &&
923 who == fuid)) {
924 if ((access_mask & ACE_READ_DATA) &&
925 (!(seen & S_IRUSR))) {
926 seen |= S_IRUSR;
927 if (type == ALLOW) {
928 mode |= S_IRUSR;
929 }
930 }
931 if ((access_mask & ACE_WRITE_DATA) &&
932 (!(seen & S_IWUSR))) {
933 seen |= S_IWUSR;
934 if (type == ALLOW) {
935 mode |= S_IWUSR;
936 }
937 }
938 if ((access_mask & ACE_EXECUTE) &&
939 (!(seen & S_IXUSR))) {
940 seen |= S_IXUSR;
941 if (type == ALLOW) {
942 mode |= S_IXUSR;
943 }
944 }
945 } else if (entry_type == OWNING_GROUP ||
946 (entry_type == ACE_IDENTIFIER_GROUP && who == fgid)) {
947 if ((access_mask & ACE_READ_DATA) &&
948 (!(seen & S_IRGRP))) {
949 seen |= S_IRGRP;
950 if (type == ALLOW) {
951 mode |= S_IRGRP;
952 }
953 }
954 if ((access_mask & ACE_WRITE_DATA) &&
955 (!(seen & S_IWGRP))) {
956 seen |= S_IWGRP;
957 if (type == ALLOW) {
958 mode |= S_IWGRP;
959 }
960 }
961 if ((access_mask & ACE_EXECUTE) &&
962 (!(seen & S_IXGRP))) {
963 seen |= S_IXGRP;
964 if (type == ALLOW) {
965 mode |= S_IXGRP;
966 }
967 }
968 } else if (entry_type == ACE_EVERYONE) {
969 if ((access_mask & ACE_READ_DATA)) {
970 if (!(seen & S_IRUSR)) {
971 seen |= S_IRUSR;
972 if (type == ALLOW) {
973 mode |= S_IRUSR;
974 }
975 }
976 if (!(seen & S_IRGRP)) {
977 seen |= S_IRGRP;
978 if (type == ALLOW) {
979 mode |= S_IRGRP;
980 }
981 }
982 if (!(seen & S_IROTH)) {
983 seen |= S_IROTH;
984 if (type == ALLOW) {
985 mode |= S_IROTH;
986 }
987 }
988 }
989 if ((access_mask & ACE_WRITE_DATA)) {
990 if (!(seen & S_IWUSR)) {
991 seen |= S_IWUSR;
992 if (type == ALLOW) {
993 mode |= S_IWUSR;
994 }
995 }
996 if (!(seen & S_IWGRP)) {
997 seen |= S_IWGRP;
998 if (type == ALLOW) {
999 mode |= S_IWGRP;
1000 }
1001 }
1002 if (!(seen & S_IWOTH)) {
1003 seen |= S_IWOTH;
1004 if (type == ALLOW) {
1005 mode |= S_IWOTH;
1006 }
1007 }
1008 }
1009 if ((access_mask & ACE_EXECUTE)) {
1010 if (!(seen & S_IXUSR)) {
1011 seen |= S_IXUSR;
1012 if (type == ALLOW) {
1013 mode |= S_IXUSR;
1014 }
1015 }
1016 if (!(seen & S_IXGRP)) {
1017 seen |= S_IXGRP;
1018 if (type == ALLOW) {
1019 mode |= S_IXGRP;
1020 }
1021 }
1022 if (!(seen & S_IXOTH)) {
1023 seen |= S_IXOTH;
1024 if (type == ALLOW) {
1025 mode |= S_IXOTH;
1026 }
1027 }
1028 }
1029 } else {
1030 /*
1031 * Only care if this IDENTIFIER_GROUP or
1032 * USER ACE denies execute access to someone,
1033 * mode is not affected
1034 */
1035 if ((access_mask & ACE_EXECUTE) && type == DENY)
1036 an_exec_denied = B_TRUE;
1037 }
1038 }
1039
1040 /*
1041 * Failure to allow is effectively a deny, so execute permission
1042 * is denied if it was never mentioned or if we explicitly
1043 * weren't allowed it.
1044 */
1045 if (!an_exec_denied &&
1046 ((seen & ALL_MODE_EXECS) != ALL_MODE_EXECS ||
1047 (mode & ALL_MODE_EXECS) != ALL_MODE_EXECS))
1048 an_exec_denied = B_TRUE;
1049
1050 if (an_exec_denied)
1051 *pflags &= ~ZFS_NO_EXECS_DENIED;
1052 else
1053 *pflags |= ZFS_NO_EXECS_DENIED;
1054
1055 return (mode);
1056 }
1057
1058 /*
1059 * Read an external acl object. If the intent is to modify, always
1060 * create a new acl and leave any cached acl in place.
1061 */
1062 static int
1063 zfs_acl_node_read(znode_t *zp, boolean_t have_lock, zfs_acl_t **aclpp,
1064 boolean_t will_modify)
1065 {
1066 zfs_acl_t *aclp;
1067 int aclsize;
1068 int acl_count;
1069 zfs_acl_node_t *aclnode;
1070 zfs_acl_phys_t znode_acl;
1071 int version;
1072 int error;
1073 boolean_t drop_lock = B_FALSE;
1074
1075 ASSERT(MUTEX_HELD(&zp->z_acl_lock));
1076
1077 if (zp->z_acl_cached && !will_modify) {
1078 *aclpp = zp->z_acl_cached;
1079 return (0);
1080 }
1081
1082 /*
1083 * close race where znode could be upgrade while trying to
1084 * read the znode attributes.
1085 *
1086 * But this could only happen if the file isn't already an SA
1087 * znode
1088 */
1089 if (!zp->z_is_sa && !have_lock) {
1090 mutex_enter(&zp->z_lock);
1091 drop_lock = B_TRUE;
1092 }
1093 version = zfs_znode_acl_version(zp);
1094
1095 if ((error = zfs_acl_znode_info(zp, &aclsize,
1096 &acl_count, &znode_acl)) != 0) {
1097 goto done;
1098 }
1099
1100 aclp = zfs_acl_alloc(version);
1101
1102 aclp->z_acl_count = acl_count;
1103 aclp->z_acl_bytes = aclsize;
1104
1105 aclnode = zfs_acl_node_alloc(aclsize);
1106 aclnode->z_ace_count = aclp->z_acl_count;
1107 aclnode->z_size = aclsize;
1108
1109 if (!zp->z_is_sa) {
1110 if (znode_acl.z_acl_extern_obj) {
1111 error = dmu_read(zp->z_zfsvfs->z_os,
1112 znode_acl.z_acl_extern_obj, 0, aclnode->z_size,
1113 aclnode->z_acldata, DMU_READ_PREFETCH);
1114 } else {
1115 bcopy(znode_acl.z_ace_data, aclnode->z_acldata,
1116 aclnode->z_size);
1117 }
1118 } else {
1119 error = sa_lookup(zp->z_sa_hdl, SA_ZPL_DACL_ACES(zp->z_zfsvfs),
1120 aclnode->z_acldata, aclnode->z_size);
1121 }
1122
1123 if (error != 0) {
1124 zfs_acl_free(aclp);
1125 zfs_acl_node_free(aclnode);
1126 /* convert checksum errors into IO errors */
1127 if (error == ECKSUM)
1128 error = EIO;
1129 goto done;
1130 }
1131
1132 list_insert_head(&aclp->z_acl, aclnode);
1133
1134 *aclpp = aclp;
1135 if (!will_modify)
1136 zp->z_acl_cached = aclp;
1137 done:
1138 if (drop_lock)
1139 mutex_exit(&zp->z_lock);
1140 return (error);
1141 }
1142
1143 /*ARGSUSED*/
1144 void
1145 zfs_acl_data_locator(void **dataptr, uint32_t *length, uint32_t buflen,
1146 boolean_t start, void *userdata)
1147 {
1148 zfs_acl_locator_cb_t *cb = (zfs_acl_locator_cb_t *)userdata;
1149
1150 if (start) {
1151 cb->cb_acl_node = list_head(&cb->cb_aclp->z_acl);
1152 } else {
1153 cb->cb_acl_node = list_next(&cb->cb_aclp->z_acl,
1154 cb->cb_acl_node);
1155 }
1156 *dataptr = cb->cb_acl_node->z_acldata;
1157 *length = cb->cb_acl_node->z_size;
1158 }
1159
1160 int
1161 zfs_acl_chown_setattr(znode_t *zp)
1162 {
1163 int error;
1164 zfs_acl_t *aclp;
1165
1166 ASSERT(MUTEX_HELD(&zp->z_lock));
1167 ASSERT(MUTEX_HELD(&zp->z_acl_lock));
1168
1169 if ((error = zfs_acl_node_read(zp, B_TRUE, &aclp, B_FALSE)) == 0)
1170 zp->z_mode = zfs_mode_compute(zp->z_mode, aclp,
1171 &zp->z_pflags, zp->z_uid, zp->z_gid);
1172 return (error);
1173 }
1174
1175 /*
1176 * common code for setting ACLs.
1177 *
1178 * This function is called from zfs_mode_update, zfs_perm_init, and zfs_setacl.
1179 * zfs_setacl passes a non-NULL inherit pointer (ihp) to indicate that it's
1180 * already checked the acl and knows whether to inherit.
1181 */
1182 int
1183 zfs_aclset_common(znode_t *zp, zfs_acl_t *aclp, cred_t *cr, dmu_tx_t *tx)
1184 {
1185 int error;
1186 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1187 dmu_object_type_t otype;
1188 zfs_acl_locator_cb_t locate = { 0 };
1189 uint64_t mode;
1190 sa_bulk_attr_t bulk[5];
1191 uint64_t ctime[2];
1192 int count = 0;
1193
1194 mode = zp->z_mode;
1195
1196 mode = zfs_mode_compute(mode, aclp, &zp->z_pflags,
1197 zp->z_uid, zp->z_gid);
1198
1199 zp->z_mode = mode;
1200 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs), NULL,
1201 &mode, sizeof (mode));
1202 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL,
1203 &zp->z_pflags, sizeof (zp->z_pflags));
1204 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL,
1205 &ctime, sizeof (ctime));
1206
1207 if (zp->z_acl_cached) {
1208 zfs_acl_free(zp->z_acl_cached);
1209 zp->z_acl_cached = NULL;
1210 }
1211
1212 /*
1213 * Upgrade needed?
1214 */
1215 if (!zfsvfs->z_use_fuids) {
1216 otype = DMU_OT_OLDACL;
1217 } else {
1218 if ((aclp->z_version == ZFS_ACL_VERSION_INITIAL) &&
1219 (zfsvfs->z_version >= ZPL_VERSION_FUID))
1220 zfs_acl_xform(zp, aclp, cr);
1221 ASSERT(aclp->z_version >= ZFS_ACL_VERSION_FUID);
1222 otype = DMU_OT_ACL;
1223 }
1224
1225 /*
1226 * Arrgh, we have to handle old on disk format
1227 * as well as newer (preferred) SA format.
1228 */
1229
1230 if (zp->z_is_sa) { /* the easy case, just update the ACL attribute */
1231 locate.cb_aclp = aclp;
1232 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_DACL_ACES(zfsvfs),
1233 zfs_acl_data_locator, &locate, aclp->z_acl_bytes);
1234 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_DACL_COUNT(zfsvfs),
1235 NULL, &aclp->z_acl_count, sizeof (uint64_t));
1236 } else { /* Painful legacy way */
1237 zfs_acl_node_t *aclnode;
1238 uint64_t off = 0;
1239 zfs_acl_phys_t acl_phys;
1240 uint64_t aoid;
1241
1242 if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_ZNODE_ACL(zfsvfs),
1243 &acl_phys, sizeof (acl_phys))) != 0)
1244 return (error);
1245
1246 aoid = acl_phys.z_acl_extern_obj;
1247
1248 if (aclp->z_acl_bytes > ZFS_ACE_SPACE) {
1249 /*
1250 * If ACL was previously external and we are now
1251 * converting to new ACL format then release old
1252 * ACL object and create a new one.
1253 */
1254 if (aoid &&
1255 aclp->z_version != acl_phys.z_acl_version) {
1256 error = dmu_object_free(zfsvfs->z_os, aoid, tx);
1257 if (error)
1258 return (error);
1259 aoid = 0;
1260 }
1261 if (aoid == 0) {
1262 aoid = dmu_object_alloc(zfsvfs->z_os,
1263 otype, aclp->z_acl_bytes,
1264 otype == DMU_OT_ACL ?
1265 DMU_OT_SYSACL : DMU_OT_NONE,
1266 otype == DMU_OT_ACL ?
1267 DN_MAX_BONUSLEN : 0, tx);
1268 } else {
1269 (void) dmu_object_set_blocksize(zfsvfs->z_os,
1270 aoid, aclp->z_acl_bytes, 0, tx);
1271 }
1272 acl_phys.z_acl_extern_obj = aoid;
1273 for (aclnode = list_head(&aclp->z_acl); aclnode;
1274 aclnode = list_next(&aclp->z_acl, aclnode)) {
1275 if (aclnode->z_ace_count == 0)
1276 continue;
1277 dmu_write(zfsvfs->z_os, aoid, off,
1278 aclnode->z_size, aclnode->z_acldata, tx);
1279 off += aclnode->z_size;
1280 }
1281 } else {
1282 void *start = acl_phys.z_ace_data;
1283 /*
1284 * Migrating back embedded?
1285 */
1286 if (acl_phys.z_acl_extern_obj) {
1287 error = dmu_object_free(zfsvfs->z_os,
1288 acl_phys.z_acl_extern_obj, tx);
1289 if (error)
1290 return (error);
1291 acl_phys.z_acl_extern_obj = 0;
1292 }
1293
1294 for (aclnode = list_head(&aclp->z_acl); aclnode;
1295 aclnode = list_next(&aclp->z_acl, aclnode)) {
1296 if (aclnode->z_ace_count == 0)
1297 continue;
1298 bcopy(aclnode->z_acldata, start,
1299 aclnode->z_size);
1300 start = (caddr_t)start + aclnode->z_size;
1301 }
1302 }
1303 /*
1304 * If Old version then swap count/bytes to match old
1305 * layout of znode_acl_phys_t.
1306 */
1307 if (aclp->z_version == ZFS_ACL_VERSION_INITIAL) {
1308 acl_phys.z_acl_size = aclp->z_acl_count;
1309 acl_phys.z_acl_count = aclp->z_acl_bytes;
1310 } else {
1311 acl_phys.z_acl_size = aclp->z_acl_bytes;
1312 acl_phys.z_acl_count = aclp->z_acl_count;
1313 }
1314 acl_phys.z_acl_version = aclp->z_version;
1315
1316 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ZNODE_ACL(zfsvfs), NULL,
1317 &acl_phys, sizeof (acl_phys));
1318 }
1319
1320 /*
1321 * Replace ACL wide bits, but first clear them.
1322 */
1323 zp->z_pflags &= ~ZFS_ACL_WIDE_FLAGS;
1324
1325 zp->z_pflags |= aclp->z_hints;
1326
1327 if (ace_trivial_common(aclp, 0, zfs_ace_walk) == 0)
1328 zp->z_pflags |= ZFS_ACL_TRIVIAL;
1329
1330 zfs_tstamp_update_setup(zp, STATE_CHANGED, NULL, ctime, B_TRUE);
1331 return (sa_bulk_update(zp->z_sa_hdl, bulk, count, tx));
1332 }
1333
1334 static void
1335 zfs_acl_chmod(vtype_t vtype, uint64_t mode, boolean_t trim, zfs_acl_t *aclp)
1336 {
1337 void *acep = NULL;
1338 uint64_t who;
1339 int new_count, new_bytes;
1340 int ace_size;
1341 int entry_type;
1342 uint16_t iflags, type;
1343 uint32_t access_mask;
1344 zfs_acl_node_t *newnode;
1345 size_t abstract_size = aclp->z_ops.ace_abstract_size();
1346 void *zacep;
1347 boolean_t isdir;
1348 trivial_acl_t masks;
1349
1350 new_count = new_bytes = 0;
1351
1352 isdir = (vtype == VDIR);
1353
1354 acl_trivial_access_masks((mode_t)mode, isdir, &masks);
1355
1356 newnode = zfs_acl_node_alloc((abstract_size * 6) + aclp->z_acl_bytes);
1357
1358 zacep = newnode->z_acldata;
1359 if (masks.allow0) {
1360 zfs_set_ace(aclp, zacep, masks.allow0, ALLOW, -1, ACE_OWNER);
1361 zacep = (void *)((uintptr_t)zacep + abstract_size);
1362 new_count++;
1363 new_bytes += abstract_size;
1364 } if (masks.deny1) {
1365 zfs_set_ace(aclp, zacep, masks.deny1, DENY, -1, ACE_OWNER);
1366 zacep = (void *)((uintptr_t)zacep + abstract_size);
1367 new_count++;
1368 new_bytes += abstract_size;
1369 }
1370 if (masks.deny2) {
1371 zfs_set_ace(aclp, zacep, masks.deny2, DENY, -1, OWNING_GROUP);
1372 zacep = (void *)((uintptr_t)zacep + abstract_size);
1373 new_count++;
1374 new_bytes += abstract_size;
1375 }
1376
1377 while (acep = zfs_acl_next_ace(aclp, acep, &who, &access_mask,
1378 &iflags, &type)) {
1379 uint16_t inherit_flags;
1380
1381 entry_type = (iflags & ACE_TYPE_FLAGS);
1382 inherit_flags = (iflags & ALL_INHERIT);
1383
1384 if ((entry_type == ACE_OWNER || entry_type == ACE_EVERYONE ||
1385 (entry_type == OWNING_GROUP)) &&
1386 ((inherit_flags & ACE_INHERIT_ONLY_ACE) == 0)) {
1387 continue;
1388 }
1389
1390 /*
1391 * If this ACL has any inheritable ACEs, mark that in
1392 * the hints (which are later masked into the pflags)
1393 * so create knows to do inheritance.
1394 */
1395 if (isdir && (inherit_flags &
1396 (ACE_FILE_INHERIT_ACE|ACE_DIRECTORY_INHERIT_ACE)))
1397 aclp->z_hints |= ZFS_INHERIT_ACE;
1398
1399 if ((type != ALLOW && type != DENY) ||
1400 (inherit_flags & ACE_INHERIT_ONLY_ACE)) {
1401 switch (type) {
1402 case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE:
1403 case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE:
1404 case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE:
1405 case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE:
1406 aclp->z_hints |= ZFS_ACL_OBJ_ACE;
1407 break;
1408 }
1409 } else {
1410
1411 /*
1412 * Limit permissions to be no greater than
1413 * group permissions.
1414 * The "aclinherit" and "aclmode" properties
1415 * affect policy for create and chmod(2),
1416 * respectively.
1417 */
1418 if ((type == ALLOW) && trim)
1419 access_mask &= masks.group;
1420 }
1421 zfs_set_ace(aclp, zacep, access_mask, type, who, iflags);
1422 ace_size = aclp->z_ops.ace_size(acep);
1423 zacep = (void *)((uintptr_t)zacep + ace_size);
1424 new_count++;
1425 new_bytes += ace_size;
1426 }
1427 zfs_set_ace(aclp, zacep, masks.owner, 0, -1, ACE_OWNER);
1428 zacep = (void *)((uintptr_t)zacep + abstract_size);
1429 zfs_set_ace(aclp, zacep, masks.group, 0, -1, OWNING_GROUP);
1430 zacep = (void *)((uintptr_t)zacep + abstract_size);
1431 zfs_set_ace(aclp, zacep, masks.everyone, 0, -1, ACE_EVERYONE);
1432
1433 new_count += 3;
1434 new_bytes += abstract_size * 3;
1435 zfs_acl_release_nodes(aclp);
1436 aclp->z_acl_count = new_count;
1437 aclp->z_acl_bytes = new_bytes;
1438 newnode->z_ace_count = new_count;
1439 newnode->z_size = new_bytes;
1440 list_insert_tail(&aclp->z_acl, newnode);
1441 }
1442
1443 int
1444 zfs_acl_chmod_setattr(znode_t *zp, zfs_acl_t **aclp, uint64_t mode)
1445 {
1446 int error = 0;
1447
1448 mutex_enter(&zp->z_acl_lock);
1449 mutex_enter(&zp->z_lock);
1450 if (zp->z_zfsvfs->z_acl_mode == ZFS_ACL_DISCARD)
1451 *aclp = zfs_acl_alloc(zfs_acl_version_zp(zp));
1452 else
1453 error = zfs_acl_node_read(zp, B_TRUE, aclp, B_TRUE);
1454
1455 if (error == 0) {
1456 (*aclp)->z_hints = zp->z_pflags & V4_ACL_WIDE_FLAGS;
1457 zfs_acl_chmod(ZTOV(zp)->v_type, mode,
1458 (zp->z_zfsvfs->z_acl_mode == ZFS_ACL_GROUPMASK), *aclp);
1459 }
1460 mutex_exit(&zp->z_lock);
1461 mutex_exit(&zp->z_acl_lock);
1462
1463 return (error);
1464 }
1465
1466 /*
1467 * strip off write_owner and write_acl
1468 */
1469 static void
1470 zfs_restricted_update(zfsvfs_t *zfsvfs, zfs_acl_t *aclp, void *acep)
1471 {
1472 uint32_t mask = aclp->z_ops.ace_mask_get(acep);
1473
1474 if ((zfsvfs->z_acl_inherit == ZFS_ACL_RESTRICTED) &&
1475 (aclp->z_ops.ace_type_get(acep) == ALLOW)) {
1476 mask &= ~RESTRICTED_CLEAR;
1477 aclp->z_ops.ace_mask_set(acep, mask);
1478 }
1479 }
1480
1481 /*
1482 * Should ACE be inherited?
1483 */
1484 static int
1485 zfs_ace_can_use(vtype_t vtype, uint16_t acep_flags)
1486 {
1487 int iflags = (acep_flags & 0xf);
1488
1489 if ((vtype == VDIR) && (iflags & ACE_DIRECTORY_INHERIT_ACE))
1490 return (1);
1491 else if (iflags & ACE_FILE_INHERIT_ACE)
1492 return (!((vtype == VDIR) &&
1493 (iflags & ACE_NO_PROPAGATE_INHERIT_ACE)));
1494 return (0);
1495 }
1496
1497 /*
1498 * inherit inheritable ACEs from parent
1499 */
1500 static zfs_acl_t *
1501 zfs_acl_inherit(zfsvfs_t *zfsvfs, vtype_t vtype, zfs_acl_t *paclp,
1502 uint64_t mode, boolean_t *need_chmod)
1503 {
1504 void *pacep;
1505 void *acep;
1506 zfs_acl_node_t *aclnode;
1507 zfs_acl_t *aclp = NULL;
1508 uint64_t who;
1509 uint32_t access_mask;
1510 uint16_t iflags, newflags, type;
1511 size_t ace_size;
1512 void *data1, *data2;
1513 size_t data1sz, data2sz;
1514 boolean_t vdir = vtype == VDIR;
1515 boolean_t vreg = vtype == VREG;
1516 boolean_t passthrough, passthrough_x, noallow;
1517
1518 passthrough_x =
1519 zfsvfs->z_acl_inherit == ZFS_ACL_PASSTHROUGH_X;
1520 passthrough = passthrough_x ||
1521 zfsvfs->z_acl_inherit == ZFS_ACL_PASSTHROUGH;
1522 noallow =
1523 zfsvfs->z_acl_inherit == ZFS_ACL_NOALLOW;
1524
1525 *need_chmod = B_TRUE;
1526 pacep = NULL;
1527 aclp = zfs_acl_alloc(paclp->z_version);
1528 if (zfsvfs->z_acl_inherit == ZFS_ACL_DISCARD || vtype == VLNK)
1529 return (aclp);
1530 while (pacep = zfs_acl_next_ace(paclp, pacep, &who,
1531 &access_mask, &iflags, &type)) {
1532
1533 /*
1534 * don't inherit bogus ACEs
1535 */
1536 if (!zfs_acl_valid_ace_type(type, iflags))
1537 continue;
1538
1539 if (noallow && type == ALLOW)
1540 continue;
1541
1542 ace_size = aclp->z_ops.ace_size(pacep);
1543
1544 if (!zfs_ace_can_use(vtype, iflags))
1545 continue;
1546
1547 /*
1548 * If owner@, group@, or everyone@ inheritable
1549 * then zfs_acl_chmod() isn't needed.
1550 */
1551 if (passthrough &&
1552 ((iflags & (ACE_OWNER|ACE_EVERYONE)) ||
1553 ((iflags & OWNING_GROUP) ==
1554 OWNING_GROUP)) && (vreg || (vdir && (iflags &
1555 ACE_DIRECTORY_INHERIT_ACE)))) {
1556 *need_chmod = B_FALSE;
1557 }
1558
1559 if (!vdir && passthrough_x &&
1560 ((mode & (S_IXUSR | S_IXGRP | S_IXOTH)) == 0)) {
1561 access_mask &= ~ACE_EXECUTE;
1562 }
1563
1564 aclnode = zfs_acl_node_alloc(ace_size);
1565 list_insert_tail(&aclp->z_acl, aclnode);
1566 acep = aclnode->z_acldata;
1567
1568 zfs_set_ace(aclp, acep, access_mask, type,
1569 who, iflags|ACE_INHERITED_ACE);
1570
1571 /*
1572 * Copy special opaque data if any
1573 */
1574 if ((data1sz = paclp->z_ops.ace_data(pacep, &data1)) != 0) {
1575 VERIFY((data2sz = aclp->z_ops.ace_data(acep,
1576 &data2)) == data1sz);
1577 bcopy(data1, data2, data2sz);
1578 }
1579
1580 aclp->z_acl_count++;
1581 aclnode->z_ace_count++;
1582 aclp->z_acl_bytes += aclnode->z_size;
1583 newflags = aclp->z_ops.ace_flags_get(acep);
1584
1585 if (vdir)
1586 aclp->z_hints |= ZFS_INHERIT_ACE;
1587
1588 if ((iflags & ACE_NO_PROPAGATE_INHERIT_ACE) || !vdir) {
1589 newflags &= ~ALL_INHERIT;
1590 aclp->z_ops.ace_flags_set(acep,
1591 newflags|ACE_INHERITED_ACE);
1592 zfs_restricted_update(zfsvfs, aclp, acep);
1593 continue;
1594 }
1595
1596 ASSERT(vdir);
1597
1598 /*
1599 * If only FILE_INHERIT is set then turn on
1600 * inherit_only
1601 */
1602 if ((iflags & (ACE_FILE_INHERIT_ACE |
1603 ACE_DIRECTORY_INHERIT_ACE)) == ACE_FILE_INHERIT_ACE) {
1604 newflags |= ACE_INHERIT_ONLY_ACE;
1605 aclp->z_ops.ace_flags_set(acep,
1606 newflags|ACE_INHERITED_ACE);
1607 } else {
1608 newflags &= ~ACE_INHERIT_ONLY_ACE;
1609 aclp->z_ops.ace_flags_set(acep,
1610 newflags|ACE_INHERITED_ACE);
1611 }
1612 }
1613 return (aclp);
1614 }
1615
1616 /*
1617 * Create file system object initial permissions
1618 * including inheritable ACEs.
1619 */
1620 int
1621 zfs_acl_ids_create(znode_t *dzp, int flag, vattr_t *vap, cred_t *cr,
1622 vsecattr_t *vsecp, zfs_acl_ids_t *acl_ids)
1623 {
1624 int error;
1625 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
1626 zfs_acl_t *paclp;
1627 gid_t gid;
1628 boolean_t need_chmod = B_TRUE;
1629 boolean_t inherited = B_FALSE;
1630
1631 bzero(acl_ids, sizeof (zfs_acl_ids_t));
1632 acl_ids->z_mode = MAKEIMODE(vap->va_type, vap->va_mode);
1633
1634 if (vsecp)
1635 if ((error = zfs_vsec_2_aclp(zfsvfs, vap->va_type, vsecp, cr,
1636 &acl_ids->z_fuidp, &acl_ids->z_aclp)) != 0)
1637 return (error);
1638 /*
1639 * Determine uid and gid.
1640 */
1641 if ((flag & IS_ROOT_NODE) || zfsvfs->z_replay ||
1642 ((flag & IS_XATTR) && (vap->va_type == VDIR))) {
1643 acl_ids->z_fuid = zfs_fuid_create(zfsvfs,
1644 (uint64_t)vap->va_uid, cr,
1645 ZFS_OWNER, &acl_ids->z_fuidp);
1646 acl_ids->z_fgid = zfs_fuid_create(zfsvfs,
1647 (uint64_t)vap->va_gid, cr,
1648 ZFS_GROUP, &acl_ids->z_fuidp);
1649 gid = vap->va_gid;
1650 } else {
1651 acl_ids->z_fuid = zfs_fuid_create_cred(zfsvfs, ZFS_OWNER,
1652 cr, &acl_ids->z_fuidp);
1653 acl_ids->z_fgid = 0;
1654 if (vap->va_mask & AT_GID) {
1655 acl_ids->z_fgid = zfs_fuid_create(zfsvfs,
1656 (uint64_t)vap->va_gid,
1657 cr, ZFS_GROUP, &acl_ids->z_fuidp);
1658 gid = vap->va_gid;
1659 if (acl_ids->z_fgid != dzp->z_gid &&
1660 !groupmember(vap->va_gid, cr) &&
1661 secpolicy_vnode_create_gid(cr) != 0)
1662 acl_ids->z_fgid = 0;
1663 }
1664 if (acl_ids->z_fgid == 0) {
1665 if (dzp->z_mode & S_ISGID) {
1666 char *domain;
1667 uint32_t rid;
1668
1669 acl_ids->z_fgid = dzp->z_gid;
1670 gid = zfs_fuid_map_id(zfsvfs, acl_ids->z_fgid,
1671 cr, ZFS_GROUP);
1672
1673 if (zfsvfs->z_use_fuids &&
1674 IS_EPHEMERAL(acl_ids->z_fgid)) {
1675 domain = zfs_fuid_idx_domain(
1676 &zfsvfs->z_fuid_idx,
1677 FUID_INDEX(acl_ids->z_fgid));
1678 rid = FUID_RID(acl_ids->z_fgid);
1679 zfs_fuid_node_add(&acl_ids->z_fuidp,
1680 domain, rid,
1681 FUID_INDEX(acl_ids->z_fgid),
1682 acl_ids->z_fgid, ZFS_GROUP);
1683 }
1684 } else {
1685 acl_ids->z_fgid = zfs_fuid_create_cred(zfsvfs,
1686 ZFS_GROUP, cr, &acl_ids->z_fuidp);
1687 gid = crgetgid(cr);
1688 }
1689 }
1690 }
1691
1692 /*
1693 * If we're creating a directory, and the parent directory has the
1694 * set-GID bit set, set in on the new directory.
1695 * Otherwise, if the user is neither privileged nor a member of the
1696 * file's new group, clear the file's set-GID bit.
1697 */
1698
1699 if (!(flag & IS_ROOT_NODE) && (dzp->z_mode & S_ISGID) &&
1700 (vap->va_type == VDIR)) {
1701 acl_ids->z_mode |= S_ISGID;
1702 } else {
1703 if ((acl_ids->z_mode & S_ISGID) &&
1704 secpolicy_vnode_setids_setgids(cr, gid) != 0)
1705 acl_ids->z_mode &= ~S_ISGID;
1706 }
1707
1708 if (acl_ids->z_aclp == NULL) {
1709 mutex_enter(&dzp->z_acl_lock);
1710 mutex_enter(&dzp->z_lock);
1711 if (!(flag & IS_ROOT_NODE) &&
1712 (dzp->z_pflags & ZFS_INHERIT_ACE) &&
1713 !(dzp->z_pflags & ZFS_XATTR)) {
1714 VERIFY(0 == zfs_acl_node_read(dzp, B_TRUE,
1715 &paclp, B_FALSE));
1716 acl_ids->z_aclp = zfs_acl_inherit(zfsvfs,
1717 vap->va_type, paclp, acl_ids->z_mode, &need_chmod);
1718 inherited = B_TRUE;
1719 } else {
1720 acl_ids->z_aclp =
1721 zfs_acl_alloc(zfs_acl_version_zp(dzp));
1722 acl_ids->z_aclp->z_hints |= ZFS_ACL_TRIVIAL;
1723 }
1724 mutex_exit(&dzp->z_lock);
1725 mutex_exit(&dzp->z_acl_lock);
1726 if (need_chmod) {
1727 acl_ids->z_aclp->z_hints |= (vap->va_type == VDIR) ?
1728 ZFS_ACL_AUTO_INHERIT : 0;
1729 zfs_acl_chmod(vap->va_type, acl_ids->z_mode,
1730 (zfsvfs->z_acl_inherit == ZFS_ACL_RESTRICTED),
1731 acl_ids->z_aclp);
1732 }
1733 }
1734
1735 if (inherited || vsecp) {
1736 acl_ids->z_mode = zfs_mode_compute(acl_ids->z_mode,
1737 acl_ids->z_aclp, &acl_ids->z_aclp->z_hints,
1738 acl_ids->z_fuid, acl_ids->z_fgid);
1739 if (ace_trivial_common(acl_ids->z_aclp, 0, zfs_ace_walk) == 0)
1740 acl_ids->z_aclp->z_hints |= ZFS_ACL_TRIVIAL;
1741 }
1742
1743 return (0);
1744 }
1745
1746 /*
1747 * Free ACL and fuid_infop, but not the acl_ids structure
1748 */
1749 void
1750 zfs_acl_ids_free(zfs_acl_ids_t *acl_ids)
1751 {
1752 if (acl_ids->z_aclp)
1753 zfs_acl_free(acl_ids->z_aclp);
1754 if (acl_ids->z_fuidp)
1755 zfs_fuid_info_free(acl_ids->z_fuidp);
1756 acl_ids->z_aclp = NULL;
1757 acl_ids->z_fuidp = NULL;
1758 }
1759
1760 boolean_t
1761 zfs_acl_ids_overquota(zfsvfs_t *zfsvfs, zfs_acl_ids_t *acl_ids)
1762 {
1763 return (zfs_fuid_overquota(zfsvfs, B_FALSE, acl_ids->z_fuid) ||
1764 zfs_fuid_overquota(zfsvfs, B_TRUE, acl_ids->z_fgid));
1765 }
1766
1767 /*
1768 * Retrieve a files ACL
1769 */
1770 int
1771 zfs_getacl(znode_t *zp, vsecattr_t *vsecp, boolean_t skipaclchk, cred_t *cr)
1772 {
1773 zfs_acl_t *aclp;
1774 ulong_t mask;
1775 int error;
1776 int count = 0;
1777 int largeace = 0;
1778
1779 mask = vsecp->vsa_mask & (VSA_ACE | VSA_ACECNT |
1780 VSA_ACE_ACLFLAGS | VSA_ACE_ALLTYPES);
1781
1782 if (mask == 0)
1783 return (ENOSYS);
1784
1785 if (error = zfs_zaccess(zp, ACE_READ_ACL, 0, skipaclchk, cr))
1786 return (error);
1787
1788 mutex_enter(&zp->z_acl_lock);
1789
1790 error = zfs_acl_node_read(zp, B_FALSE, &aclp, B_FALSE);
1791 if (error != 0) {
1792 mutex_exit(&zp->z_acl_lock);
1793 return (error);
1794 }
1795
1796 /*
1797 * Scan ACL to determine number of ACEs
1798 */
1799 if ((zp->z_pflags & ZFS_ACL_OBJ_ACE) && !(mask & VSA_ACE_ALLTYPES)) {
1800 void *zacep = NULL;
1801 uint64_t who;
1802 uint32_t access_mask;
1803 uint16_t type, iflags;
1804
1805 while (zacep = zfs_acl_next_ace(aclp, zacep,
1806 &who, &access_mask, &iflags, &type)) {
1807 switch (type) {
1808 case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE:
1809 case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE:
1810 case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE:
1811 case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE:
1812 largeace++;
1813 continue;
1814 default:
1815 count++;
1816 }
1817 }
1818 vsecp->vsa_aclcnt = count;
1819 } else
1820 count = (int)aclp->z_acl_count;
1821
1822 if (mask & VSA_ACECNT) {
1823 vsecp->vsa_aclcnt = count;
1824 }
1825
1826 if (mask & VSA_ACE) {
1827 size_t aclsz;
1828
1829 aclsz = count * sizeof (ace_t) +
1830 sizeof (ace_object_t) * largeace;
1831
1832 vsecp->vsa_aclentp = kmem_alloc(aclsz, KM_SLEEP);
1833 vsecp->vsa_aclentsz = aclsz;
1834
1835 if (aclp->z_version == ZFS_ACL_VERSION_FUID)
1836 zfs_copy_fuid_2_ace(zp->z_zfsvfs, aclp, cr,
1837 vsecp->vsa_aclentp, !(mask & VSA_ACE_ALLTYPES));
1838 else {
1839 zfs_acl_node_t *aclnode;
1840 void *start = vsecp->vsa_aclentp;
1841
1842 for (aclnode = list_head(&aclp->z_acl); aclnode;
1843 aclnode = list_next(&aclp->z_acl, aclnode)) {
1844 bcopy(aclnode->z_acldata, start,
1845 aclnode->z_size);
1846 start = (caddr_t)start + aclnode->z_size;
1847 }
1848 ASSERT((caddr_t)start - (caddr_t)vsecp->vsa_aclentp ==
1849 aclp->z_acl_bytes);
1850 }
1851 }
1852 if (mask & VSA_ACE_ACLFLAGS) {
1853 vsecp->vsa_aclflags = 0;
1854 if (zp->z_pflags & ZFS_ACL_DEFAULTED)
1855 vsecp->vsa_aclflags |= ACL_DEFAULTED;
1856 if (zp->z_pflags & ZFS_ACL_PROTECTED)
1857 vsecp->vsa_aclflags |= ACL_PROTECTED;
1858 if (zp->z_pflags & ZFS_ACL_AUTO_INHERIT)
1859 vsecp->vsa_aclflags |= ACL_AUTO_INHERIT;
1860 }
1861
1862 mutex_exit(&zp->z_acl_lock);
1863
1864 return (0);
1865 }
1866
1867 int
1868 zfs_vsec_2_aclp(zfsvfs_t *zfsvfs, vtype_t obj_type,
1869 vsecattr_t *vsecp, cred_t *cr, zfs_fuid_info_t **fuidp, zfs_acl_t **zaclp)
1870 {
1871 zfs_acl_t *aclp;
1872 zfs_acl_node_t *aclnode;
1873 int aclcnt = vsecp->vsa_aclcnt;
1874 int error;
1875
1876 if (vsecp->vsa_aclcnt > MAX_ACL_ENTRIES || vsecp->vsa_aclcnt <= 0)
1877 return (EINVAL);
1878
1879 aclp = zfs_acl_alloc(zfs_acl_version(zfsvfs->z_version));
1880
1881 aclp->z_hints = 0;
1882 aclnode = zfs_acl_node_alloc(aclcnt * sizeof (zfs_object_ace_t));
1883 if (aclp->z_version == ZFS_ACL_VERSION_INITIAL) {
1884 if ((error = zfs_copy_ace_2_oldace(obj_type, aclp,
1885 (ace_t *)vsecp->vsa_aclentp, aclnode->z_acldata,
1886 aclcnt, &aclnode->z_size)) != 0) {
1887 zfs_acl_free(aclp);
1888 zfs_acl_node_free(aclnode);
1889 return (error);
1890 }
1891 } else {
1892 if ((error = zfs_copy_ace_2_fuid(zfsvfs, obj_type, aclp,
1893 vsecp->vsa_aclentp, aclnode->z_acldata, aclcnt,
1894 &aclnode->z_size, fuidp, cr)) != 0) {
1895 zfs_acl_free(aclp);
1896 zfs_acl_node_free(aclnode);
1897 return (error);
1898 }
1899 }
1900 aclp->z_acl_bytes = aclnode->z_size;
1901 aclnode->z_ace_count = aclcnt;
1902 aclp->z_acl_count = aclcnt;
1903 list_insert_head(&aclp->z_acl, aclnode);
1904
1905 /*
1906 * If flags are being set then add them to z_hints
1907 */
1908 if (vsecp->vsa_mask & VSA_ACE_ACLFLAGS) {
1909 if (vsecp->vsa_aclflags & ACL_PROTECTED)
1910 aclp->z_hints |= ZFS_ACL_PROTECTED;
1911 if (vsecp->vsa_aclflags & ACL_DEFAULTED)
1912 aclp->z_hints |= ZFS_ACL_DEFAULTED;
1913 if (vsecp->vsa_aclflags & ACL_AUTO_INHERIT)
1914 aclp->z_hints |= ZFS_ACL_AUTO_INHERIT;
1915 }
1916
1917 *zaclp = aclp;
1918
1919 return (0);
1920 }
1921
1922 /*
1923 * Set a files ACL
1924 */
1925 int
1926 zfs_setacl(znode_t *zp, vsecattr_t *vsecp, boolean_t skipaclchk, cred_t *cr)
1927 {
1928 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1929 zilog_t *zilog = zfsvfs->z_log;
1930 ulong_t mask = vsecp->vsa_mask & (VSA_ACE | VSA_ACECNT);
1931 dmu_tx_t *tx;
1932 int error;
1933 zfs_acl_t *aclp;
1934 zfs_fuid_info_t *fuidp = NULL;
1935 boolean_t fuid_dirtied;
1936 uint64_t acl_obj;
1937
1938 if (mask == 0)
1939 return (ENOSYS);
1940
1941 if (zp->z_pflags & ZFS_IMMUTABLE)
1942 return (EPERM);
1943
1944 if (error = zfs_zaccess(zp, ACE_WRITE_ACL, 0, skipaclchk, cr))
1945 return (error);
1946
1947 error = zfs_vsec_2_aclp(zfsvfs, ZTOV(zp)->v_type, vsecp, cr, &fuidp,
1948 &aclp);
1949 if (error)
1950 return (error);
1951
1952 /*
1953 * If ACL wide flags aren't being set then preserve any
1954 * existing flags.
1955 */
1956 if (!(vsecp->vsa_mask & VSA_ACE_ACLFLAGS)) {
1957 aclp->z_hints |=
1958 (zp->z_pflags & V4_ACL_WIDE_FLAGS);
1959 }
1960 top:
1961 mutex_enter(&zp->z_acl_lock);
1962 mutex_enter(&zp->z_lock);
1963
1964 tx = dmu_tx_create(zfsvfs->z_os);
1965
1966 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE);
1967
1968 fuid_dirtied = zfsvfs->z_fuid_dirty;
1969 if (fuid_dirtied)
1970 zfs_fuid_txhold(zfsvfs, tx);
1971
1972 /*
1973 * If old version and ACL won't fit in bonus and we aren't
1974 * upgrading then take out necessary DMU holds
1975 */
1976
1977 if ((acl_obj = zfs_external_acl(zp)) != 0) {
1978 if (zfsvfs->z_version >= ZPL_VERSION_FUID &&
1979 zfs_znode_acl_version(zp) <= ZFS_ACL_VERSION_INITIAL) {
1980 dmu_tx_hold_free(tx, acl_obj, 0,
1981 DMU_OBJECT_END);
1982 dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0,
1983 aclp->z_acl_bytes);
1984 } else {
1985 dmu_tx_hold_write(tx, acl_obj, 0, aclp->z_acl_bytes);
1986 }
1987 } else if (!zp->z_is_sa && aclp->z_acl_bytes > ZFS_ACE_SPACE) {
1988 dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, aclp->z_acl_bytes);
1989 }
1990
1991 zfs_sa_upgrade_txholds(tx, zp);
1992 error = dmu_tx_assign(tx, TXG_NOWAIT);
1993 if (error) {
1994 mutex_exit(&zp->z_acl_lock);
1995 mutex_exit(&zp->z_lock);
1996
1997 if (error == ERESTART) {
1998 dmu_tx_wait(tx);
1999 dmu_tx_abort(tx);
2000 goto top;
2001 }
2002 dmu_tx_abort(tx);
2003 zfs_acl_free(aclp);
2004 return (error);
2005 }
2006
2007 error = zfs_aclset_common(zp, aclp, cr, tx);
2008 ASSERT(error == 0);
2009 ASSERT(zp->z_acl_cached == NULL);
2010 zp->z_acl_cached = aclp;
2011
2012 if (fuid_dirtied)
2013 zfs_fuid_sync(zfsvfs, tx);
2014
2015 zfs_log_acl(zilog, tx, zp, vsecp, fuidp);
2016
2017 if (fuidp)
2018 zfs_fuid_info_free(fuidp);
2019 dmu_tx_commit(tx);
2020 done:
2021 mutex_exit(&zp->z_lock);
2022 mutex_exit(&zp->z_acl_lock);
2023
2024 return (error);
2025 }
2026
2027 /*
2028 * Check accesses of interest (AoI) against attributes of the dataset
2029 * such as read-only. Returns zero if no AoI conflict with dataset
2030 * attributes, otherwise an appropriate errno is returned.
2031 */
2032 static int
2033 zfs_zaccess_dataset_check(znode_t *zp, uint32_t v4_mode)
2034 {
2035 if ((v4_mode & WRITE_MASK) &&
2036 (zp->z_zfsvfs->z_vfs->vfs_flag & VFS_RDONLY) &&
2037 (!IS_DEVVP(ZTOV(zp)) ||
2038 (IS_DEVVP(ZTOV(zp)) && (v4_mode & WRITE_MASK_ATTRS)))) {
2039 return (EROFS);
2040 }
2041
2042 /*
2043 * Only check for READONLY on non-directories.
2044 */
2045 if ((v4_mode & WRITE_MASK_DATA) &&
2046 (((ZTOV(zp)->v_type != VDIR) &&
2047 (zp->z_pflags & (ZFS_READONLY | ZFS_IMMUTABLE))) ||
2048 (ZTOV(zp)->v_type == VDIR &&
2049 (zp->z_pflags & ZFS_IMMUTABLE)))) {
2050 return (EPERM);
2051 }
2052
2053 if ((v4_mode & (ACE_DELETE | ACE_DELETE_CHILD)) &&
2054 (zp->z_pflags & ZFS_NOUNLINK)) {
2055 return (EPERM);
2056 }
2057
2058 if (((v4_mode & (ACE_READ_DATA|ACE_EXECUTE)) &&
2059 (zp->z_pflags & ZFS_AV_QUARANTINED))) {
2060 return (EACCES);
2061 }
2062
2063 return (0);
2064 }
2065
2066 /*
2067 * The primary usage of this function is to loop through all of the
2068 * ACEs in the znode, determining what accesses of interest (AoI) to
2069 * the caller are allowed or denied. The AoI are expressed as bits in
2070 * the working_mode parameter. As each ACE is processed, bits covered
2071 * by that ACE are removed from the working_mode. This removal
2072 * facilitates two things. The first is that when the working mode is
2073 * empty (= 0), we know we've looked at all the AoI. The second is
2074 * that the ACE interpretation rules don't allow a later ACE to undo
2075 * something granted or denied by an earlier ACE. Removing the
2076 * discovered access or denial enforces this rule. At the end of
2077 * processing the ACEs, all AoI that were found to be denied are
2078 * placed into the working_mode, giving the caller a mask of denied
2079 * accesses. Returns:
2080 * 0 if all AoI granted
2081 * EACCESS if the denied mask is non-zero
2082 * other error if abnormal failure (e.g., IO error)
2083 *
2084 * A secondary usage of the function is to determine if any of the
2085 * AoI are granted. If an ACE grants any access in
2086 * the working_mode, we immediately short circuit out of the function.
2087 * This mode is chosen by setting anyaccess to B_TRUE. The
2088 * working_mode is not a denied access mask upon exit if the function
2089 * is used in this manner.
2090 */
2091 static int
2092 zfs_zaccess_aces_check(znode_t *zp, uint32_t *working_mode,
2093 boolean_t anyaccess, cred_t *cr)
2094 {
2095 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
2096 zfs_acl_t *aclp;
2097 int error;
2098 uid_t uid = crgetuid(cr);
2099 uint64_t who;
2100 uint16_t type, iflags;
2101 uint16_t entry_type;
2102 uint32_t access_mask;
2103 uint32_t deny_mask = 0;
2104 zfs_ace_hdr_t *acep = NULL;
2105 boolean_t checkit;
2106 uid_t gowner;
2107 uid_t fowner;
2108
2109 zfs_fuid_map_ids(zp, cr, &fowner, &gowner);
2110
2111 mutex_enter(&zp->z_acl_lock);
2112
2113 error = zfs_acl_node_read(zp, B_FALSE, &aclp, B_FALSE);
2114 if (error != 0) {
2115 mutex_exit(&zp->z_acl_lock);
2116 return (error);
2117 }
2118
2119 ASSERT(zp->z_acl_cached);
2120
2121 while (acep = zfs_acl_next_ace(aclp, acep, &who, &access_mask,
2122 &iflags, &type)) {
2123 uint32_t mask_matched;
2124
2125 if (!zfs_acl_valid_ace_type(type, iflags))
2126 continue;
2127
2128 if (ZTOV(zp)->v_type == VDIR && (iflags & ACE_INHERIT_ONLY_ACE))
2129 continue;
2130
2131 /* Skip ACE if it does not affect any AoI */
2132 mask_matched = (access_mask & *working_mode);
2133 if (!mask_matched)
2134 continue;
2135
2136 entry_type = (iflags & ACE_TYPE_FLAGS);
2137
2138 checkit = B_FALSE;
2139
2140 switch (entry_type) {
2141 case ACE_OWNER:
2142 if (uid == fowner)
2143 checkit = B_TRUE;
2144 break;
2145 case OWNING_GROUP:
2146 who = gowner;
2147 /*FALLTHROUGH*/
2148 case ACE_IDENTIFIER_GROUP:
2149 checkit = zfs_groupmember(zfsvfs, who, cr);
2150 break;
2151 case ACE_EVERYONE:
2152 checkit = B_TRUE;
2153 break;
2154
2155 /* USER Entry */
2156 default:
2157 if (entry_type == 0) {
2158 uid_t newid;
2159
2160 newid = zfs_fuid_map_id(zfsvfs, who, cr,
2161 ZFS_ACE_USER);
2162 if (newid != IDMAP_WK_CREATOR_OWNER_UID &&
2163 uid == newid)
2164 checkit = B_TRUE;
2165 break;
2166 } else {
2167 mutex_exit(&zp->z_acl_lock);
2168 return (EIO);
2169 }
2170 }
2171
2172 if (checkit) {
2173 if (type == DENY) {
2174 DTRACE_PROBE3(zfs__ace__denies,
2175 znode_t *, zp,
2176 zfs_ace_hdr_t *, acep,
2177 uint32_t, mask_matched);
2178 deny_mask |= mask_matched;
2179 } else {
2180 DTRACE_PROBE3(zfs__ace__allows,
2181 znode_t *, zp,
2182 zfs_ace_hdr_t *, acep,
2183 uint32_t, mask_matched);
2184 if (anyaccess) {
2185 mutex_exit(&zp->z_acl_lock);
2186 return (0);
2187 }
2188 }
2189 *working_mode &= ~mask_matched;
2190 }
2191
2192 /* Are we done? */
2193 if (*working_mode == 0)
2194 break;
2195 }
2196
2197 mutex_exit(&zp->z_acl_lock);
2198
2199 /* Put the found 'denies' back on the working mode */
2200 if (deny_mask) {
2201 *working_mode |= deny_mask;
2202 return (EACCES);
2203 } else if (*working_mode) {
2204 return (-1);
2205 }
2206
2207 return (0);
2208 }
2209
2210 /*
2211 * Return true if any access whatsoever granted, we don't actually
2212 * care what access is granted.
2213 */
2214 boolean_t
2215 zfs_has_access(znode_t *zp, cred_t *cr)
2216 {
2217 uint32_t have = ACE_ALL_PERMS;
2218
2219 if (zfs_zaccess_aces_check(zp, &have, B_TRUE, cr) != 0) {
2220 uid_t owner;
2221
2222 owner = zfs_fuid_map_id(zp->z_zfsvfs, zp->z_uid, cr, ZFS_OWNER);
2223 return (secpolicy_vnode_any_access(cr, ZTOV(zp), owner) == 0);
2224 }
2225 return (B_TRUE);
2226 }
2227
2228 static int
2229 zfs_zaccess_common(znode_t *zp, uint32_t v4_mode, uint32_t *working_mode,
2230 boolean_t *check_privs, boolean_t skipaclchk, cred_t *cr)
2231 {
2232 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
2233 int err;
2234
2235 *working_mode = v4_mode;
2236 *check_privs = B_TRUE;
2237
2238 /*
2239 * Short circuit empty requests
2240 */
2241 if (v4_mode == 0 || zfsvfs->z_replay) {
2242 *working_mode = 0;
2243 return (0);
2244 }
2245
2246 if ((err = zfs_zaccess_dataset_check(zp, v4_mode)) != 0) {
2247 *check_privs = B_FALSE;
2248 return (err);
2249 }
2250
2251 /*
2252 * The caller requested that the ACL check be skipped. This
2253 * would only happen if the caller checked VOP_ACCESS() with a
2254 * 32 bit ACE mask and already had the appropriate permissions.
2255 */
2256 if (skipaclchk) {
2257 *working_mode = 0;
2258 return (0);
2259 }
2260
2261 return (zfs_zaccess_aces_check(zp, working_mode, B_FALSE, cr));
2262 }
2263
2264 static int
2265 zfs_zaccess_append(znode_t *zp, uint32_t *working_mode, boolean_t *check_privs,
2266 cred_t *cr)
2267 {
2268 if (*working_mode != ACE_WRITE_DATA)
2269 return (EACCES);
2270
2271 return (zfs_zaccess_common(zp, ACE_APPEND_DATA, working_mode,
2272 check_privs, B_FALSE, cr));
2273 }
2274
2275 int
2276 zfs_fastaccesschk_execute(znode_t *zdp, cred_t *cr)
2277 {
2278 boolean_t owner = B_FALSE;
2279 boolean_t groupmbr = B_FALSE;
2280 boolean_t is_attr;
2281 uid_t uid = crgetuid(cr);
2282 int error;
2283
2284 if (zdp->z_pflags & ZFS_AV_QUARANTINED)
2285 return (EACCES);
2286
2287 is_attr = ((zdp->z_pflags & ZFS_XATTR) &&
2288 (ZTOV(zdp)->v_type == VDIR));
2289 if (is_attr)
2290 goto slow;
2291
2292
2293 mutex_enter(&zdp->z_acl_lock);
2294
2295 if (zdp->z_pflags & ZFS_NO_EXECS_DENIED) {
2296 mutex_exit(&zdp->z_acl_lock);
2297 return (0);
2298 }
2299
2300 if (FUID_INDEX(zdp->z_uid) != 0 || FUID_INDEX(zdp->z_gid) != 0) {
2301 mutex_exit(&zdp->z_acl_lock);
2302 goto slow;
2303 }
2304
2305 if (uid == zdp->z_uid) {
2306 owner = B_TRUE;
2307 if (zdp->z_mode & S_IXUSR) {
2308 mutex_exit(&zdp->z_acl_lock);
2309 return (0);
2310 } else {
2311 mutex_exit(&zdp->z_acl_lock);
2312 goto slow;
2313 }
2314 }
2315 if (groupmember(zdp->z_gid, cr)) {
2316 groupmbr = B_TRUE;
2317 if (zdp->z_mode & S_IXGRP) {
2318 mutex_exit(&zdp->z_acl_lock);
2319 return (0);
2320 } else {
2321 mutex_exit(&zdp->z_acl_lock);
2322 goto slow;
2323 }
2324 }
2325 if (!owner && !groupmbr) {
2326 if (zdp->z_mode & S_IXOTH) {
2327 mutex_exit(&zdp->z_acl_lock);
2328 return (0);
2329 }
2330 }
2331
2332 mutex_exit(&zdp->z_acl_lock);
2333
2334 slow:
2335 DTRACE_PROBE(zfs__fastpath__execute__access__miss);
2336 ZFS_ENTER(zdp->z_zfsvfs);
2337 error = zfs_zaccess(zdp, ACE_EXECUTE, 0, B_FALSE, cr);
2338 ZFS_EXIT(zdp->z_zfsvfs);
2339 return (error);
2340 }
2341
2342 /*
2343 * Determine whether Access should be granted/denied.
2344 * The least priv subsytem is always consulted as a basic privilege
2345 * can define any form of access.
2346 */
2347 int
2348 zfs_zaccess(znode_t *zp, int mode, int flags, boolean_t skipaclchk, cred_t *cr)
2349 {
2350 uint32_t working_mode;
2351 int error;
2352 int is_attr;
2353 boolean_t check_privs;
2354 znode_t *xzp;
2355 znode_t *check_zp = zp;
2356 mode_t needed_bits;
2357 uid_t owner;
2358
2359 is_attr = ((zp->z_pflags & ZFS_XATTR) && (ZTOV(zp)->v_type == VDIR));
2360
2361 /*
2362 * If attribute then validate against base file
2363 */
2364 if (is_attr) {
2365 uint64_t parent;
2366
2367 if ((error = sa_lookup(zp->z_sa_hdl,
2368 SA_ZPL_PARENT(zp->z_zfsvfs), &parent,
2369 sizeof (parent))) != 0)
2370 return (error);
2371
2372 if ((error = zfs_zget(zp->z_zfsvfs,
2373 parent, &xzp)) != 0) {
2374 return (error);
2375 }
2376
2377 check_zp = xzp;
2378
2379 /*
2380 * fixup mode to map to xattr perms
2381 */
2382
2383 if (mode & (ACE_WRITE_DATA|ACE_APPEND_DATA)) {
2384 mode &= ~(ACE_WRITE_DATA|ACE_APPEND_DATA);
2385 mode |= ACE_WRITE_NAMED_ATTRS;
2386 }
2387
2388 if (mode & (ACE_READ_DATA|ACE_EXECUTE)) {
2389 mode &= ~(ACE_READ_DATA|ACE_EXECUTE);
2390 mode |= ACE_READ_NAMED_ATTRS;
2391 }
2392 }
2393
2394 owner = zfs_fuid_map_id(zp->z_zfsvfs, zp->z_uid, cr, ZFS_OWNER);
2395 /*
2396 * Map the bits required to the standard vnode flags VREAD|VWRITE|VEXEC
2397 * in needed_bits. Map the bits mapped by working_mode (currently
2398 * missing) in missing_bits.
2399 * Call secpolicy_vnode_access2() with (needed_bits & ~checkmode),
2400 * needed_bits.
2401 */
2402 needed_bits = 0;
2403
2404 working_mode = mode;
2405 if ((working_mode & (ACE_READ_ACL|ACE_READ_ATTRIBUTES)) &&
2406 owner == crgetuid(cr))
2407 working_mode &= ~(ACE_READ_ACL|ACE_READ_ATTRIBUTES);
2408
2409 if (working_mode & (ACE_READ_DATA|ACE_READ_NAMED_ATTRS|
2410 ACE_READ_ACL|ACE_READ_ATTRIBUTES|ACE_SYNCHRONIZE))
2411 needed_bits |= VREAD;
2412 if (working_mode & (ACE_WRITE_DATA|ACE_WRITE_NAMED_ATTRS|
2413 ACE_APPEND_DATA|ACE_WRITE_ATTRIBUTES|ACE_SYNCHRONIZE))
2414 needed_bits |= VWRITE;
2415 if (working_mode & ACE_EXECUTE)
2416 needed_bits |= VEXEC;
2417
2418 if ((error = zfs_zaccess_common(check_zp, mode, &working_mode,
2419 &check_privs, skipaclchk, cr)) == 0) {
2420 if (is_attr)
2421 VN_RELE(ZTOV(xzp));
2422 return (secpolicy_vnode_access2(cr, ZTOV(zp), owner,
2423 needed_bits, needed_bits));
2424 }
2425
2426 if (error && !check_privs) {
2427 if (is_attr)
2428 VN_RELE(ZTOV(xzp));
2429 return (error);
2430 }
2431
2432 if (error && (flags & V_APPEND)) {
2433 error = zfs_zaccess_append(zp, &working_mode, &check_privs, cr);
2434 }
2435
2436 if (error && check_privs) {
2437 mode_t checkmode = 0;
2438
2439 /*
2440 * First check for implicit owner permission on
2441 * read_acl/read_attributes
2442 */
2443
2444 error = 0;
2445 ASSERT(working_mode != 0);
2446
2447 if ((working_mode & (ACE_READ_ACL|ACE_READ_ATTRIBUTES) &&
2448 owner == crgetuid(cr)))
2449 working_mode &= ~(ACE_READ_ACL|ACE_READ_ATTRIBUTES);
2450
2451 if (working_mode & (ACE_READ_DATA|ACE_READ_NAMED_ATTRS|
2452 ACE_READ_ACL|ACE_READ_ATTRIBUTES|ACE_SYNCHRONIZE))
2453 checkmode |= VREAD;
2454 if (working_mode & (ACE_WRITE_DATA|ACE_WRITE_NAMED_ATTRS|
2455 ACE_APPEND_DATA|ACE_WRITE_ATTRIBUTES|ACE_SYNCHRONIZE))
2456 checkmode |= VWRITE;
2457 if (working_mode & ACE_EXECUTE)
2458 checkmode |= VEXEC;
2459
2460 error = secpolicy_vnode_access2(cr, ZTOV(check_zp), owner,
2461 needed_bits & ~checkmode, needed_bits);
2462
2463 if (error == 0 && (working_mode & ACE_WRITE_OWNER))
2464 error = secpolicy_vnode_chown(cr, owner);
2465 if (error == 0 && (working_mode & ACE_WRITE_ACL))
2466 error = secpolicy_vnode_setdac(cr, owner);
2467
2468 if (error == 0 && (working_mode &
2469 (ACE_DELETE|ACE_DELETE_CHILD)))
2470 error = secpolicy_vnode_remove(cr);
2471
2472 if (error == 0 && (working_mode & ACE_SYNCHRONIZE)) {
2473 error = secpolicy_vnode_chown(cr, owner);
2474 }
2475 if (error == 0) {
2476 /*
2477 * See if any bits other than those already checked
2478 * for are still present. If so then return EACCES
2479 */
2480 if (working_mode & ~(ZFS_CHECKED_MASKS)) {
2481 error = EACCES;
2482 }
2483 }
2484 } else if (error == 0) {
2485 error = secpolicy_vnode_access2(cr, ZTOV(zp), owner,
2486 needed_bits, needed_bits);
2487 }
2488
2489
2490 if (is_attr)
2491 VN_RELE(ZTOV(xzp));
2492
2493 return (error);
2494 }
2495
2496 /*
2497 * Translate traditional unix VREAD/VWRITE/VEXEC mode into
2498 * native ACL format and call zfs_zaccess()
2499 */
2500 int
2501 zfs_zaccess_rwx(znode_t *zp, mode_t mode, int flags, cred_t *cr)
2502 {
2503 return (zfs_zaccess(zp, zfs_unix_to_v4(mode >> 6), flags, B_FALSE, cr));
2504 }
2505
2506 /*
2507 * Access function for secpolicy_vnode_setattr
2508 */
2509 int
2510 zfs_zaccess_unix(znode_t *zp, mode_t mode, cred_t *cr)
2511 {
2512 int v4_mode = zfs_unix_to_v4(mode >> 6);
2513
2514 return (zfs_zaccess(zp, v4_mode, 0, B_FALSE, cr));
2515 }
2516
2517 static int
2518 zfs_delete_final_check(znode_t *zp, znode_t *dzp,
2519 mode_t available_perms, cred_t *cr)
2520 {
2521 int error;
2522 uid_t downer;
2523
2524 downer = zfs_fuid_map_id(dzp->z_zfsvfs, dzp->z_uid, cr, ZFS_OWNER);
2525
2526 error = secpolicy_vnode_access2(cr, ZTOV(dzp),
2527 downer, available_perms, VWRITE|VEXEC);
2528
2529 if (error == 0)
2530 error = zfs_sticky_remove_access(dzp, zp, cr);
2531
2532 return (error);
2533 }
2534
2535 /*
2536 * Determine whether Access should be granted/deny, without
2537 * consulting least priv subsystem.
2538 *
2539 *
2540 * The following chart is the recommended NFSv4 enforcement for
2541 * ability to delete an object.
2542 *
2543 * -------------------------------------------------------
2544 * | Parent Dir | Target Object Permissions |
2545 * | permissions | |
2546 * -------------------------------------------------------
2547 * | | ACL Allows | ACL Denies| Delete |
2548 * | | Delete | Delete | unspecified|
2549 * -------------------------------------------------------
2550 * | ACL Allows | Permit | Permit | Permit |
2551 * | DELETE_CHILD | |
2552 * -------------------------------------------------------
2553 * | ACL Denies | Permit | Deny | Deny |
2554 * | DELETE_CHILD | | | |
2555 * -------------------------------------------------------
2556 * | ACL specifies | | | |
2557 * | only allow | Permit | Permit | Permit |
2558 * | write and | | | |
2559 * | execute | | | |
2560 * -------------------------------------------------------
2561 * | ACL denies | | | |
2562 * | write and | Permit | Deny | Deny |
2563 * | execute | | | |
2564 * -------------------------------------------------------
2565 * ^
2566 * |
2567 * No search privilege, can't even look up file?
2568 *
2569 */
2570 int
2571 zfs_zaccess_delete(znode_t *dzp, znode_t *zp, cred_t *cr)
2572 {
2573 uint32_t dzp_working_mode = 0;
2574 uint32_t zp_working_mode = 0;
2575 int dzp_error, zp_error;
2576 mode_t available_perms;
2577 boolean_t dzpcheck_privs = B_TRUE;
2578 boolean_t zpcheck_privs = B_TRUE;
2579
2580 /*
2581 * We want specific DELETE permissions to
2582 * take precedence over WRITE/EXECUTE. We don't
2583 * want an ACL such as this to mess us up.
2584 * user:joe:write_data:deny,user:joe:delete:allow
2585 *
2586 * However, deny permissions may ultimately be overridden
2587 * by secpolicy_vnode_access().
2588 *
2589 * We will ask for all of the necessary permissions and then
2590 * look at the working modes from the directory and target object
2591 * to determine what was found.
2592 */
2593
2594 if (zp->z_pflags & (ZFS_IMMUTABLE | ZFS_NOUNLINK))
2595 return (EPERM);
2596
2597 /*
2598 * First row
2599 * If the directory permissions allow the delete, we are done.
2600 */
2601 if ((dzp_error = zfs_zaccess_common(dzp, ACE_DELETE_CHILD,
2602 &dzp_working_mode, &dzpcheck_privs, B_FALSE, cr)) == 0)
2603 return (0);
2604
2605 /*
2606 * If target object has delete permission then we are done
2607 */
2608 if ((zp_error = zfs_zaccess_common(zp, ACE_DELETE, &zp_working_mode,
2609 &zpcheck_privs, B_FALSE, cr)) == 0)
2610 return (0);
2611
2612 ASSERT(dzp_error && zp_error);
2613
2614 if (!dzpcheck_privs)
2615 return (dzp_error);
2616 if (!zpcheck_privs)
2617 return (zp_error);
2618
2619 /*
2620 * Second row
2621 *
2622 * If directory returns EACCES then delete_child was denied
2623 * due to deny delete_child. In this case send the request through
2624 * secpolicy_vnode_remove(). We don't use zfs_delete_final_check()
2625 * since that *could* allow the delete based on write/execute permission
2626 * and we want delete permissions to override write/execute.
2627 */
2628
2629 if (dzp_error == EACCES)
2630 return (secpolicy_vnode_remove(cr));
2631
2632 /*
2633 * Third Row
2634 * only need to see if we have write/execute on directory.
2635 */
2636
2637 dzp_error = zfs_zaccess_common(dzp, ACE_EXECUTE|ACE_WRITE_DATA,
2638 &dzp_working_mode, &dzpcheck_privs, B_FALSE, cr);
2639
2640 if (dzp_error != 0 && !dzpcheck_privs)
2641 return (dzp_error);
2642
2643 /*
2644 * Fourth row
2645 */
2646
2647 available_perms = (dzp_working_mode & ACE_WRITE_DATA) ? 0 : VWRITE;
2648 available_perms |= (dzp_working_mode & ACE_EXECUTE) ? 0 : VEXEC;
2649
2650 return (zfs_delete_final_check(zp, dzp, available_perms, cr));
2651
2652 }
2653
2654 int
2655 zfs_zaccess_rename(znode_t *sdzp, znode_t *szp, znode_t *tdzp,
2656 znode_t *tzp, cred_t *cr)
2657 {
2658 int add_perm;
2659 int error;
2660
2661 if (szp->z_pflags & ZFS_AV_QUARANTINED)
2662 return (EACCES);
2663
2664 add_perm = (ZTOV(szp)->v_type == VDIR) ?
2665 ACE_ADD_SUBDIRECTORY : ACE_ADD_FILE;
2666
2667 /*
2668 * Rename permissions are combination of delete permission +
2669 * add file/subdir permission.
2670 */
2671
2672 /*
2673 * first make sure we do the delete portion.
2674 *
2675 * If that succeeds then check for add_file/add_subdir permissions
2676 */
2677
2678 if (error = zfs_zaccess_delete(sdzp, szp, cr))
2679 return (error);
2680
2681 /*
2682 * If we have a tzp, see if we can delete it?
2683 */
2684 if (tzp) {
2685 if (error = zfs_zaccess_delete(tdzp, tzp, cr))
2686 return (error);
2687 }
2688
2689 /*
2690 * Now check for add permissions
2691 */
2692 error = zfs_zaccess(tdzp, add_perm, 0, B_FALSE, cr);
2693
2694 return (error);
2695 }
Unleashed OS