]> git.proxmox.com Git - mirror_zfs.git/blob - module/zfs/zfs_acl.c
projects / mirror_zfs.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
Fix typo/etc in module/zfs/zfs_ctldir.c
[mirror_zfs.git] / module / zfs / zfs_acl.c
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 (c) 2013 by Delphix. All rights reserved.
24 */
25
26
27 #include <sys/types.h>
28 #include <sys/param.h>
29 #include <sys/time.h>
30 #include <sys/sysmacros.h>
31 #include <sys/vfs.h>
32 #include <sys/vnode.h>
33 #include <sys/sid.h>
34 #include <sys/file.h>
35 #include <sys/stat.h>
36 #include <sys/kmem.h>
37 #include <sys/cmn_err.h>
38 #include <sys/errno.h>
39 #include <sys/sdt.h>
40 #include <sys/fs/zfs.h>
41 #include <sys/mode.h>
42 #include <sys/policy.h>
43 #include <sys/zfs_znode.h>
44 #include <sys/zfs_fuid.h>
45 #include <sys/zfs_acl.h>
46 #include <sys/zfs_dir.h>
47 #include <sys/zfs_vfsops.h>
48 #include <sys/dmu.h>
49 #include <sys/dnode.h>
50 #include <sys/zap.h>
51 #include <sys/sa.h>
52 #include <sys/trace_acl.h>
53 #include <sys/zpl.h>
54
55 #define ALLOW ACE_ACCESS_ALLOWED_ACE_TYPE
56 #define DENY ACE_ACCESS_DENIED_ACE_TYPE
57 #define MAX_ACE_TYPE ACE_SYSTEM_ALARM_CALLBACK_OBJECT_ACE_TYPE
58 #define MIN_ACE_TYPE ALLOW
59
60 #define OWNING_GROUP (ACE_GROUP|ACE_IDENTIFIER_GROUP)
61 #define EVERYONE_ALLOW_MASK (ACE_READ_ACL|ACE_READ_ATTRIBUTES | \
62 ACE_READ_NAMED_ATTRS|ACE_SYNCHRONIZE)
63 #define EVERYONE_DENY_MASK (ACE_WRITE_ACL|ACE_WRITE_OWNER | \
64 ACE_WRITE_ATTRIBUTES|ACE_WRITE_NAMED_ATTRS)
65 #define OWNER_ALLOW_MASK (ACE_WRITE_ACL | ACE_WRITE_OWNER | \
66 ACE_WRITE_ATTRIBUTES|ACE_WRITE_NAMED_ATTRS)
67
68 #define ZFS_CHECKED_MASKS (ACE_READ_ACL|ACE_READ_ATTRIBUTES|ACE_READ_DATA| \
69 ACE_READ_NAMED_ATTRS|ACE_WRITE_DATA|ACE_WRITE_ATTRIBUTES| \
70 ACE_WRITE_NAMED_ATTRS|ACE_APPEND_DATA|ACE_EXECUTE|ACE_WRITE_OWNER| \
71 ACE_WRITE_ACL|ACE_DELETE|ACE_DELETE_CHILD|ACE_SYNCHRONIZE)
72
73 #define WRITE_MASK_DATA (ACE_WRITE_DATA|ACE_APPEND_DATA|ACE_WRITE_NAMED_ATTRS)
74 #define WRITE_MASK_ATTRS (ACE_WRITE_ACL|ACE_WRITE_OWNER|ACE_WRITE_ATTRIBUTES| \
75 ACE_DELETE|ACE_DELETE_CHILD)
76 #define WRITE_MASK (WRITE_MASK_DATA|WRITE_MASK_ATTRS)
77
78 #define OGE_CLEAR (ACE_READ_DATA|ACE_LIST_DIRECTORY|ACE_WRITE_DATA| \
79 ACE_ADD_FILE|ACE_APPEND_DATA|ACE_ADD_SUBDIRECTORY|ACE_EXECUTE)
80
81 #define OKAY_MASK_BITS (ACE_READ_DATA|ACE_LIST_DIRECTORY|ACE_WRITE_DATA| \
82 ACE_ADD_FILE|ACE_APPEND_DATA|ACE_ADD_SUBDIRECTORY|ACE_EXECUTE)
83
84 #define ALL_INHERIT (ACE_FILE_INHERIT_ACE|ACE_DIRECTORY_INHERIT_ACE | \
85 ACE_NO_PROPAGATE_INHERIT_ACE|ACE_INHERIT_ONLY_ACE|ACE_INHERITED_ACE)
86
87 #define RESTRICTED_CLEAR (ACE_WRITE_ACL|ACE_WRITE_OWNER)
88
89 #define V4_ACL_WIDE_FLAGS (ZFS_ACL_AUTO_INHERIT|ZFS_ACL_DEFAULTED|\
90 ZFS_ACL_PROTECTED)
91
92 #define ZFS_ACL_WIDE_FLAGS (V4_ACL_WIDE_FLAGS|ZFS_ACL_TRIVIAL|ZFS_INHERIT_ACE|\
93 ZFS_ACL_OBJ_ACE)
94
95 #define ALL_MODE_EXECS (S_IXUSR | S_IXGRP | S_IXOTH)
96
97 #define IDMAP_WK_CREATOR_OWNER_UID 2147483648U
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 .ace_mask_get = zfs_ace_v0_get_mask,
176 .ace_mask_set = zfs_ace_v0_set_mask,
177 .ace_flags_get = zfs_ace_v0_get_flags,
178 .ace_flags_set = zfs_ace_v0_set_flags,
179 .ace_type_get = zfs_ace_v0_get_type,
180 .ace_type_set = zfs_ace_v0_set_type,
181 .ace_who_get = zfs_ace_v0_get_who,
182 .ace_who_set = zfs_ace_v0_set_who,
183 .ace_size = zfs_ace_v0_size,
184 .ace_abstract_size = zfs_ace_v0_abstract_size,
185 .ace_mask_off = zfs_ace_v0_mask_off,
186 .ace_data = 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 .ace_mask_get = zfs_ace_fuid_get_mask,
312 .ace_mask_set = zfs_ace_fuid_set_mask,
313 .ace_flags_get = zfs_ace_fuid_get_flags,
314 .ace_flags_set = zfs_ace_fuid_set_flags,
315 .ace_type_get = zfs_ace_fuid_get_type,
316 .ace_type_set = zfs_ace_fuid_set_type,
317 .ace_who_get = zfs_ace_fuid_get_who,
318 .ace_who_set = zfs_ace_fuid_set_who,
319 .ace_size = zfs_ace_fuid_size,
320 .ace_abstract_size = zfs_ace_fuid_abstract_size,
321 .ace_mask_off = zfs_ace_fuid_mask_off,
322 .ace_data = 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(ZTOZSB(zp)),
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 = ZTOZSB(zp);
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(ZTOZSB(zp)),
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(ZTOZSB(zp)->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(umode_t obj_mode, 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 (S_ISDIR(obj_mode) &&
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 /*
647 * Copy ACE to internal ZFS format.
648 * While processing the ACL each ACE will be validated for correctness.
649 * ACE FUIDs will be created later.
650 */
651 int
652 zfs_copy_ace_2_fuid(zfsvfs_t *zfsvfs, umode_t obj_mode, zfs_acl_t *aclp,
653 void *datap, zfs_ace_t *z_acl, uint64_t aclcnt, size_t *size,
654 zfs_fuid_info_t **fuidp, cred_t *cr)
655 {
656 int i;
657 uint16_t entry_type;
658 zfs_ace_t *aceptr = z_acl;
659 ace_t *acep = datap;
660 zfs_object_ace_t *zobjacep;
661 ace_object_t *aceobjp;
662
663 for (i = 0; i != aclcnt; i++) {
664 aceptr->z_hdr.z_access_mask = acep->a_access_mask;
665 aceptr->z_hdr.z_flags = acep->a_flags;
666 aceptr->z_hdr.z_type = acep->a_type;
667 entry_type = aceptr->z_hdr.z_flags & ACE_TYPE_FLAGS;
668 if (entry_type != ACE_OWNER && entry_type != OWNING_GROUP &&
669 entry_type != ACE_EVERYONE) {
670 aceptr->z_fuid = zfs_fuid_create(zfsvfs, acep->a_who,
671 cr, (entry_type == 0) ?
672 ZFS_ACE_USER : ZFS_ACE_GROUP, fuidp);
673 }
674
675 /*
676 * Make sure ACE is valid
677 */
678 if (zfs_ace_valid(obj_mode, aclp, aceptr->z_hdr.z_type,
679 aceptr->z_hdr.z_flags) != B_TRUE)
680 return (SET_ERROR(EINVAL));
681
682 switch (acep->a_type) {
683 case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE:
684 case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE:
685 case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE:
686 case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE:
687 zobjacep = (zfs_object_ace_t *)aceptr;
688 aceobjp = (ace_object_t *)acep;
689
690 bcopy(aceobjp->a_obj_type, zobjacep->z_object_type,
691 sizeof (aceobjp->a_obj_type));
692 bcopy(aceobjp->a_inherit_obj_type,
693 zobjacep->z_inherit_type,
694 sizeof (aceobjp->a_inherit_obj_type));
695 acep = (ace_t *)((caddr_t)acep + sizeof (ace_object_t));
696 break;
697 default:
698 acep = (ace_t *)((caddr_t)acep + sizeof (ace_t));
699 }
700
701 aceptr = (zfs_ace_t *)((caddr_t)aceptr +
702 aclp->z_ops->ace_size(aceptr));
703 }
704
705 *size = (caddr_t)aceptr - (caddr_t)z_acl;
706
707 return (0);
708 }
709
710 /*
711 * Copy ZFS ACEs to fixed size ace_t layout
712 */
713 static void
714 zfs_copy_fuid_2_ace(zfsvfs_t *zfsvfs, zfs_acl_t *aclp, cred_t *cr,
715 void *datap, int filter)
716 {
717 uint64_t who;
718 uint32_t access_mask;
719 uint16_t iflags, type;
720 zfs_ace_hdr_t *zacep = NULL;
721 ace_t *acep = datap;
722 ace_object_t *objacep;
723 zfs_object_ace_t *zobjacep;
724 size_t ace_size;
725 uint16_t entry_type;
726
727 while ((zacep = zfs_acl_next_ace(aclp, zacep,
728 &who, &access_mask, &iflags, &type))) {
729
730 switch (type) {
731 case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE:
732 case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE:
733 case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE:
734 case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE:
735 if (filter) {
736 continue;
737 }
738 zobjacep = (zfs_object_ace_t *)zacep;
739 objacep = (ace_object_t *)acep;
740 bcopy(zobjacep->z_object_type,
741 objacep->a_obj_type,
742 sizeof (zobjacep->z_object_type));
743 bcopy(zobjacep->z_inherit_type,
744 objacep->a_inherit_obj_type,
745 sizeof (zobjacep->z_inherit_type));
746 ace_size = sizeof (ace_object_t);
747 break;
748 default:
749 ace_size = sizeof (ace_t);
750 break;
751 }
752
753 entry_type = (iflags & ACE_TYPE_FLAGS);
754 if ((entry_type != ACE_OWNER &&
755 entry_type != OWNING_GROUP &&
756 entry_type != ACE_EVERYONE)) {
757 acep->a_who = zfs_fuid_map_id(zfsvfs, who,
758 cr, (entry_type & ACE_IDENTIFIER_GROUP) ?
759 ZFS_ACE_GROUP : ZFS_ACE_USER);
760 } else {
761 acep->a_who = (uid_t)(int64_t)who;
762 }
763 acep->a_access_mask = access_mask;
764 acep->a_flags = iflags;
765 acep->a_type = type;
766 acep = (ace_t *)((caddr_t)acep + ace_size);
767 }
768 }
769
770 static int
771 zfs_copy_ace_2_oldace(umode_t obj_mode, zfs_acl_t *aclp, ace_t *acep,
772 zfs_oldace_t *z_acl, int aclcnt, size_t *size)
773 {
774 int i;
775 zfs_oldace_t *aceptr = z_acl;
776
777 for (i = 0; i != aclcnt; i++, aceptr++) {
778 aceptr->z_access_mask = acep[i].a_access_mask;
779 aceptr->z_type = acep[i].a_type;
780 aceptr->z_flags = acep[i].a_flags;
781 aceptr->z_fuid = acep[i].a_who;
782 /*
783 * Make sure ACE is valid
784 */
785 if (zfs_ace_valid(obj_mode, aclp, aceptr->z_type,
786 aceptr->z_flags) != B_TRUE)
787 return (SET_ERROR(EINVAL));
788 }
789 *size = (caddr_t)aceptr - (caddr_t)z_acl;
790 return (0);
791 }
792
793 /*
794 * convert old ACL format to new
795 */
796 void
797 zfs_acl_xform(znode_t *zp, zfs_acl_t *aclp, cred_t *cr)
798 {
799 zfs_oldace_t *oldaclp;
800 int i;
801 uint16_t type, iflags;
802 uint32_t access_mask;
803 uint64_t who;
804 void *cookie = NULL;
805 zfs_acl_node_t *newaclnode;
806
807 ASSERT(aclp->z_version == ZFS_ACL_VERSION_INITIAL);
808 /*
809 * First create the ACE in a contiguous piece of memory
810 * for zfs_copy_ace_2_fuid().
811 *
812 * We only convert an ACL once, so this won't happen
813 * everytime.
814 */
815 oldaclp = kmem_alloc(sizeof (zfs_oldace_t) * aclp->z_acl_count,
816 KM_SLEEP);
817 i = 0;
818 while ((cookie = zfs_acl_next_ace(aclp, cookie, &who,
819 &access_mask, &iflags, &type))) {
820 oldaclp[i].z_flags = iflags;
821 oldaclp[i].z_type = type;
822 oldaclp[i].z_fuid = who;
823 oldaclp[i++].z_access_mask = access_mask;
824 }
825
826 newaclnode = zfs_acl_node_alloc(aclp->z_acl_count *
827 sizeof (zfs_object_ace_t));
828 aclp->z_ops = &zfs_acl_fuid_ops;
829 VERIFY(zfs_copy_ace_2_fuid(ZTOZSB(zp), ZTOI(zp)->i_mode,
830 aclp, oldaclp, newaclnode->z_acldata, aclp->z_acl_count,
831 &newaclnode->z_size, NULL, cr) == 0);
832 newaclnode->z_ace_count = aclp->z_acl_count;
833 aclp->z_version = ZFS_ACL_VERSION;
834 kmem_free(oldaclp, aclp->z_acl_count * sizeof (zfs_oldace_t));
835
836 /*
837 * Release all previous ACL nodes
838 */
839
840 zfs_acl_release_nodes(aclp);
841
842 list_insert_head(&aclp->z_acl, newaclnode);
843
844 aclp->z_acl_bytes = newaclnode->z_size;
845 aclp->z_acl_count = newaclnode->z_ace_count;
846
847 }
848
849 /*
850 * Convert unix access mask to v4 access mask
851 */
852 static uint32_t
853 zfs_unix_to_v4(uint32_t access_mask)
854 {
855 uint32_t new_mask = 0;
856
857 if (access_mask & S_IXOTH)
858 new_mask |= ACE_EXECUTE;
859 if (access_mask & S_IWOTH)
860 new_mask |= ACE_WRITE_DATA;
861 if (access_mask & S_IROTH)
862 new_mask |= ACE_READ_DATA;
863 return (new_mask);
864 }
865
866 static void
867 zfs_set_ace(zfs_acl_t *aclp, void *acep, uint32_t access_mask,
868 uint16_t access_type, uint64_t fuid, uint16_t entry_type)
869 {
870 uint16_t type = entry_type & ACE_TYPE_FLAGS;
871
872 aclp->z_ops->ace_mask_set(acep, access_mask);
873 aclp->z_ops->ace_type_set(acep, access_type);
874 aclp->z_ops->ace_flags_set(acep, entry_type);
875 if ((type != ACE_OWNER && type != OWNING_GROUP &&
876 type != ACE_EVERYONE))
877 aclp->z_ops->ace_who_set(acep, fuid);
878 }
879
880 /*
881 * Determine mode of file based on ACL.
882 * Also, create FUIDs for any User/Group ACEs
883 */
884 uint64_t
885 zfs_mode_compute(uint64_t fmode, zfs_acl_t *aclp,
886 uint64_t *pflags, uint64_t fuid, uint64_t fgid)
887 {
888 int entry_type;
889 mode_t mode;
890 mode_t seen = 0;
891 zfs_ace_hdr_t *acep = NULL;
892 uint64_t who;
893 uint16_t iflags, type;
894 uint32_t access_mask;
895 boolean_t an_exec_denied = B_FALSE;
896
897 mode = (fmode & (S_IFMT | S_ISUID | S_ISGID | S_ISVTX));
898
899 while ((acep = zfs_acl_next_ace(aclp, acep, &who,
900 &access_mask, &iflags, &type))) {
901
902 if (!zfs_acl_valid_ace_type(type, iflags))
903 continue;
904
905 entry_type = (iflags & ACE_TYPE_FLAGS);
906
907 /*
908 * Skip over owner@, group@ or everyone@ inherit only ACEs
909 */
910 if ((iflags & ACE_INHERIT_ONLY_ACE) &&
911 (entry_type == ACE_OWNER || entry_type == ACE_EVERYONE ||
912 entry_type == OWNING_GROUP))
913 continue;
914
915 if (entry_type == ACE_OWNER || (entry_type == 0 &&
916 who == fuid)) {
917 if ((access_mask & ACE_READ_DATA) &&
918 (!(seen & S_IRUSR))) {
919 seen |= S_IRUSR;
920 if (type == ALLOW) {
921 mode |= S_IRUSR;
922 }
923 }
924 if ((access_mask & ACE_WRITE_DATA) &&
925 (!(seen & S_IWUSR))) {
926 seen |= S_IWUSR;
927 if (type == ALLOW) {
928 mode |= S_IWUSR;
929 }
930 }
931 if ((access_mask & ACE_EXECUTE) &&
932 (!(seen & S_IXUSR))) {
933 seen |= S_IXUSR;
934 if (type == ALLOW) {
935 mode |= S_IXUSR;
936 }
937 }
938 } else if (entry_type == OWNING_GROUP ||
939 (entry_type == ACE_IDENTIFIER_GROUP && who == fgid)) {
940 if ((access_mask & ACE_READ_DATA) &&
941 (!(seen & S_IRGRP))) {
942 seen |= S_IRGRP;
943 if (type == ALLOW) {
944 mode |= S_IRGRP;
945 }
946 }
947 if ((access_mask & ACE_WRITE_DATA) &&
948 (!(seen & S_IWGRP))) {
949 seen |= S_IWGRP;
950 if (type == ALLOW) {
951 mode |= S_IWGRP;
952 }
953 }
954 if ((access_mask & ACE_EXECUTE) &&
955 (!(seen & S_IXGRP))) {
956 seen |= S_IXGRP;
957 if (type == ALLOW) {
958 mode |= S_IXGRP;
959 }
960 }
961 } else if (entry_type == ACE_EVERYONE) {
962 if ((access_mask & ACE_READ_DATA)) {
963 if (!(seen & S_IRUSR)) {
964 seen |= S_IRUSR;
965 if (type == ALLOW) {
966 mode |= S_IRUSR;
967 }
968 }
969 if (!(seen & S_IRGRP)) {
970 seen |= S_IRGRP;
971 if (type == ALLOW) {
972 mode |= S_IRGRP;
973 }
974 }
975 if (!(seen & S_IROTH)) {
976 seen |= S_IROTH;
977 if (type == ALLOW) {
978 mode |= S_IROTH;
979 }
980 }
981 }
982 if ((access_mask & ACE_WRITE_DATA)) {
983 if (!(seen & S_IWUSR)) {
984 seen |= S_IWUSR;
985 if (type == ALLOW) {
986 mode |= S_IWUSR;
987 }
988 }
989 if (!(seen & S_IWGRP)) {
990 seen |= S_IWGRP;
991 if (type == ALLOW) {
992 mode |= S_IWGRP;
993 }
994 }
995 if (!(seen & S_IWOTH)) {
996 seen |= S_IWOTH;
997 if (type == ALLOW) {
998 mode |= S_IWOTH;
999 }
1000 }
1001 }
1002 if ((access_mask & ACE_EXECUTE)) {
1003 if (!(seen & S_IXUSR)) {
1004 seen |= S_IXUSR;
1005 if (type == ALLOW) {
1006 mode |= S_IXUSR;
1007 }
1008 }
1009 if (!(seen & S_IXGRP)) {
1010 seen |= S_IXGRP;
1011 if (type == ALLOW) {
1012 mode |= S_IXGRP;
1013 }
1014 }
1015 if (!(seen & S_IXOTH)) {
1016 seen |= S_IXOTH;
1017 if (type == ALLOW) {
1018 mode |= S_IXOTH;
1019 }
1020 }
1021 }
1022 } else {
1023 /*
1024 * Only care if this IDENTIFIER_GROUP or
1025 * USER ACE denies execute access to someone,
1026 * mode is not affected
1027 */
1028 if ((access_mask & ACE_EXECUTE) && type == DENY)
1029 an_exec_denied = B_TRUE;
1030 }
1031 }
1032
1033 /*
1034 * Failure to allow is effectively a deny, so execute permission
1035 * is denied if it was never mentioned or if we explicitly
1036 * weren't allowed it.
1037 */
1038 if (!an_exec_denied &&
1039 ((seen & ALL_MODE_EXECS) != ALL_MODE_EXECS ||
1040 (mode & ALL_MODE_EXECS) != ALL_MODE_EXECS))
1041 an_exec_denied = B_TRUE;
1042
1043 if (an_exec_denied)
1044 *pflags &= ~ZFS_NO_EXECS_DENIED;
1045 else
1046 *pflags |= ZFS_NO_EXECS_DENIED;
1047
1048 return (mode);
1049 }
1050
1051 /*
1052 * Read an external acl object. If the intent is to modify, always
1053 * create a new acl and leave any cached acl in place.
1054 */
1055 int
1056 zfs_acl_node_read(struct znode *zp, boolean_t have_lock, zfs_acl_t **aclpp,
1057 boolean_t will_modify)
1058 {
1059 zfs_acl_t *aclp;
1060 int aclsize = 0;
1061 int acl_count = 0;
1062 zfs_acl_node_t *aclnode;
1063 zfs_acl_phys_t znode_acl;
1064 int version;
1065 int error;
1066 boolean_t drop_lock = B_FALSE;
1067
1068 ASSERT(MUTEX_HELD(&zp->z_acl_lock));
1069
1070 if (zp->z_acl_cached && !will_modify) {
1071 *aclpp = zp->z_acl_cached;
1072 return (0);
1073 }
1074
1075 /*
1076 * close race where znode could be upgrade while trying to
1077 * read the znode attributes.
1078 *
1079 * But this could only happen if the file isn't already an SA
1080 * znode
1081 */
1082 if (!zp->z_is_sa && !have_lock) {
1083 mutex_enter(&zp->z_lock);
1084 drop_lock = B_TRUE;
1085 }
1086 version = zfs_znode_acl_version(zp);
1087
1088 if ((error = zfs_acl_znode_info(zp, &aclsize,
1089 &acl_count, &znode_acl)) != 0) {
1090 goto done;
1091 }
1092
1093 aclp = zfs_acl_alloc(version);
1094
1095 aclp->z_acl_count = acl_count;
1096 aclp->z_acl_bytes = aclsize;
1097
1098 aclnode = zfs_acl_node_alloc(aclsize);
1099 aclnode->z_ace_count = aclp->z_acl_count;
1100 aclnode->z_size = aclsize;
1101
1102 if (!zp->z_is_sa) {
1103 if (znode_acl.z_acl_extern_obj) {
1104 error = dmu_read(ZTOZSB(zp)->z_os,
1105 znode_acl.z_acl_extern_obj, 0, aclnode->z_size,
1106 aclnode->z_acldata, DMU_READ_PREFETCH);
1107 } else {
1108 bcopy(znode_acl.z_ace_data, aclnode->z_acldata,
1109 aclnode->z_size);
1110 }
1111 } else {
1112 error = sa_lookup(zp->z_sa_hdl, SA_ZPL_DACL_ACES(ZTOZSB(zp)),
1113 aclnode->z_acldata, aclnode->z_size);
1114 }
1115
1116 if (error != 0) {
1117 zfs_acl_free(aclp);
1118 zfs_acl_node_free(aclnode);
1119 /* convert checksum errors into IO errors */
1120 if (error == ECKSUM)
1121 error = SET_ERROR(EIO);
1122 goto done;
1123 }
1124
1125 list_insert_head(&aclp->z_acl, aclnode);
1126
1127 *aclpp = aclp;
1128 if (!will_modify)
1129 zp->z_acl_cached = aclp;
1130 done:
1131 if (drop_lock)
1132 mutex_exit(&zp->z_lock);
1133 return (error);
1134 }
1135
1136 /*ARGSUSED*/
1137 void
1138 zfs_acl_data_locator(void **dataptr, uint32_t *length, uint32_t buflen,
1139 boolean_t start, void *userdata)
1140 {
1141 zfs_acl_locator_cb_t *cb = (zfs_acl_locator_cb_t *)userdata;
1142
1143 if (start) {
1144 cb->cb_acl_node = list_head(&cb->cb_aclp->z_acl);
1145 } else {
1146 cb->cb_acl_node = list_next(&cb->cb_aclp->z_acl,
1147 cb->cb_acl_node);
1148 }
1149 *dataptr = cb->cb_acl_node->z_acldata;
1150 *length = cb->cb_acl_node->z_size;
1151 }
1152
1153 int
1154 zfs_acl_chown_setattr(znode_t *zp)
1155 {
1156 int error;
1157 zfs_acl_t *aclp;
1158
1159 if (ZTOZSB(zp)->z_acl_type == ZFS_ACLTYPE_POSIXACL)
1160 return (0);
1161
1162 ASSERT(MUTEX_HELD(&zp->z_lock));
1163 ASSERT(MUTEX_HELD(&zp->z_acl_lock));
1164
1165 error = zfs_acl_node_read(zp, B_TRUE, &aclp, B_FALSE);
1166 if (error == 0 && aclp->z_acl_count > 0)
1167 zp->z_mode = ZTOI(zp)->i_mode =
1168 zfs_mode_compute(zp->z_mode, aclp,
1169 &zp->z_pflags, KUID_TO_SUID(ZTOI(zp)->i_uid),
1170 KGID_TO_SGID(ZTOI(zp)->i_gid));
1171
1172 /*
1173 * Some ZFS implementations (ZEVO) create neither a ZNODE_ACL
1174 * nor a DACL_ACES SA in which case ENOENT is returned from
1175 * zfs_acl_node_read() when the SA can't be located.
1176 * Allow chown/chgrp to succeed in these cases rather than
1177 * returning an error that makes no sense in the context of
1178 * the caller.
1179 */
1180 if (error == ENOENT)
1181 return (0);
1182
1183 return (error);
1184 }
1185
1186 static void
1187 acl_trivial_access_masks(mode_t mode, uint32_t *allow0, uint32_t *deny1,
1188 uint32_t *deny2, uint32_t *owner, uint32_t *group, uint32_t *everyone)
1189 {
1190 *deny1 = *deny2 = *allow0 = *group = 0;
1191
1192 if (!(mode & S_IRUSR) && (mode & (S_IRGRP|S_IROTH)))
1193 *deny1 |= ACE_READ_DATA;
1194 if (!(mode & S_IWUSR) && (mode & (S_IWGRP|S_IWOTH)))
1195 *deny1 |= ACE_WRITE_DATA;
1196 if (!(mode & S_IXUSR) && (mode & (S_IXGRP|S_IXOTH)))
1197 *deny1 |= ACE_EXECUTE;
1198
1199 if (!(mode & S_IRGRP) && (mode & S_IROTH))
1200 *deny2 = ACE_READ_DATA;
1201 if (!(mode & S_IWGRP) && (mode & S_IWOTH))
1202 *deny2 |= ACE_WRITE_DATA;
1203 if (!(mode & S_IXGRP) && (mode & S_IXOTH))
1204 *deny2 |= ACE_EXECUTE;
1205
1206 if ((mode & S_IRUSR) && (!(mode & S_IRGRP) && (mode & S_IROTH)))
1207 *allow0 |= ACE_READ_DATA;
1208 if ((mode & S_IWUSR) && (!(mode & S_IWGRP) && (mode & S_IWOTH)))
1209 *allow0 |= ACE_WRITE_DATA;
1210 if ((mode & S_IXUSR) && (!(mode & S_IXGRP) && (mode & S_IXOTH)))
1211 *allow0 |= ACE_EXECUTE;
1212
1213 *owner = ACE_WRITE_ATTRIBUTES|ACE_WRITE_OWNER|ACE_WRITE_ACL|
1214 ACE_WRITE_NAMED_ATTRS|ACE_READ_ACL|ACE_READ_ATTRIBUTES|
1215 ACE_READ_NAMED_ATTRS|ACE_SYNCHRONIZE;
1216 if (mode & S_IRUSR)
1217 *owner |= ACE_READ_DATA;
1218 if (mode & S_IWUSR)
1219 *owner |= ACE_WRITE_DATA|ACE_APPEND_DATA;
1220 if (mode & S_IXUSR)
1221 *owner |= ACE_EXECUTE;
1222
1223 *group = ACE_READ_ACL|ACE_READ_ATTRIBUTES| ACE_READ_NAMED_ATTRS|
1224 ACE_SYNCHRONIZE;
1225 if (mode & S_IRGRP)
1226 *group |= ACE_READ_DATA;
1227 if (mode & S_IWGRP)
1228 *group |= ACE_WRITE_DATA|ACE_APPEND_DATA;
1229 if (mode & S_IXGRP)
1230 *group |= ACE_EXECUTE;
1231
1232 *everyone = ACE_READ_ACL|ACE_READ_ATTRIBUTES| ACE_READ_NAMED_ATTRS|
1233 ACE_SYNCHRONIZE;
1234 if (mode & S_IROTH)
1235 *everyone |= ACE_READ_DATA;
1236 if (mode & S_IWOTH)
1237 *everyone |= ACE_WRITE_DATA|ACE_APPEND_DATA;
1238 if (mode & S_IXOTH)
1239 *everyone |= ACE_EXECUTE;
1240 }
1241
1242 /*
1243 * ace_trivial:
1244 * determine whether an ace_t acl is trivial
1245 *
1246 * Trivialness implies that the acl is composed of only
1247 * owner, group, everyone entries. ACL can't
1248 * have read_acl denied, and write_owner/write_acl/write_attributes
1249 * can only be owner@ entry.
1250 */
1251 static int
1252 ace_trivial_common(void *acep, int aclcnt,
1253 uint64_t (*walk)(void *, uint64_t, int aclcnt,
1254 uint16_t *, uint16_t *, uint32_t *))
1255 {
1256 uint16_t flags;
1257 uint32_t mask;
1258 uint16_t type;
1259 uint64_t cookie = 0;
1260
1261 while ((cookie = walk(acep, cookie, aclcnt, &flags, &type, &mask))) {
1262 switch (flags & ACE_TYPE_FLAGS) {
1263 case ACE_OWNER:
1264 case ACE_GROUP|ACE_IDENTIFIER_GROUP:
1265 case ACE_EVERYONE:
1266 break;
1267 default:
1268 return (1);
1269 }
1270
1271 if (flags & (ACE_FILE_INHERIT_ACE|
1272 ACE_DIRECTORY_INHERIT_ACE|ACE_NO_PROPAGATE_INHERIT_ACE|
1273 ACE_INHERIT_ONLY_ACE))
1274 return (1);
1275
1276 /*
1277 * Special check for some special bits
1278 *
1279 * Don't allow anybody to deny reading basic
1280 * attributes or a files ACL.
1281 */
1282 if ((mask & (ACE_READ_ACL|ACE_READ_ATTRIBUTES)) &&
1283 (type == ACE_ACCESS_DENIED_ACE_TYPE))
1284 return (1);
1285
1286 /*
1287 * Delete permissions are never set by default
1288 */
1289 if (mask & (ACE_DELETE|ACE_DELETE_CHILD))
1290 return (1);
1291 /*
1292 * only allow owner@ to have
1293 * write_acl/write_owner/write_attributes/write_xattr/
1294 */
1295 if (type == ACE_ACCESS_ALLOWED_ACE_TYPE &&
1296 (!(flags & ACE_OWNER) && (mask &
1297 (ACE_WRITE_OWNER|ACE_WRITE_ACL| ACE_WRITE_ATTRIBUTES|
1298 ACE_WRITE_NAMED_ATTRS))))
1299 return (1);
1300
1301 }
1302
1303 return (0);
1304 }
1305
1306 /*
1307 * common code for setting ACLs.
1308 *
1309 * This function is called from zfs_mode_update, zfs_perm_init, and zfs_setacl.
1310 * zfs_setacl passes a non-NULL inherit pointer (ihp) to indicate that it's
1311 * already checked the acl and knows whether to inherit.
1312 */
1313 int
1314 zfs_aclset_common(znode_t *zp, zfs_acl_t *aclp, cred_t *cr, dmu_tx_t *tx)
1315 {
1316 int error;
1317 zfsvfs_t *zfsvfs = ZTOZSB(zp);
1318 dmu_object_type_t otype;
1319 zfs_acl_locator_cb_t locate = { 0 };
1320 uint64_t mode;
1321 sa_bulk_attr_t bulk[5];
1322 uint64_t ctime[2];
1323 int count = 0;
1324 zfs_acl_phys_t acl_phys;
1325
1326 mode = zp->z_mode;
1327
1328 mode = zfs_mode_compute(mode, aclp, &zp->z_pflags,
1329 KUID_TO_SUID(ZTOI(zp)->i_uid), KGID_TO_SGID(ZTOI(zp)->i_gid));
1330
1331 zp->z_mode = ZTOI(zp)->i_mode = mode;
1332 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs), NULL,
1333 &mode, sizeof (mode));
1334 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL,
1335 &zp->z_pflags, sizeof (zp->z_pflags));
1336 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL,
1337 &ctime, sizeof (ctime));
1338
1339 if (zp->z_acl_cached) {
1340 zfs_acl_free(zp->z_acl_cached);
1341 zp->z_acl_cached = NULL;
1342 }
1343
1344 /*
1345 * Upgrade needed?
1346 */
1347 if (!zfsvfs->z_use_fuids) {
1348 otype = DMU_OT_OLDACL;
1349 } else {
1350 if ((aclp->z_version == ZFS_ACL_VERSION_INITIAL) &&
1351 (zfsvfs->z_version >= ZPL_VERSION_FUID))
1352 zfs_acl_xform(zp, aclp, cr);
1353 ASSERT(aclp->z_version >= ZFS_ACL_VERSION_FUID);
1354 otype = DMU_OT_ACL;
1355 }
1356
1357 /*
1358 * Arrgh, we have to handle old on disk format
1359 * as well as newer (preferred) SA format.
1360 */
1361
1362 if (zp->z_is_sa) { /* the easy case, just update the ACL attribute */
1363 locate.cb_aclp = aclp;
1364 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_DACL_ACES(zfsvfs),
1365 zfs_acl_data_locator, &locate, aclp->z_acl_bytes);
1366 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_DACL_COUNT(zfsvfs),
1367 NULL, &aclp->z_acl_count, sizeof (uint64_t));
1368 } else { /* Painful legacy way */
1369 zfs_acl_node_t *aclnode;
1370 uint64_t off = 0;
1371 uint64_t aoid;
1372
1373 if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_ZNODE_ACL(zfsvfs),
1374 &acl_phys, sizeof (acl_phys))) != 0)
1375 return (error);
1376
1377 aoid = acl_phys.z_acl_extern_obj;
1378
1379 if (aclp->z_acl_bytes > ZFS_ACE_SPACE) {
1380 /*
1381 * If ACL was previously external and we are now
1382 * converting to new ACL format then release old
1383 * ACL object and create a new one.
1384 */
1385 if (aoid &&
1386 aclp->z_version != acl_phys.z_acl_version) {
1387 error = dmu_object_free(zfsvfs->z_os, aoid, tx);
1388 if (error)
1389 return (error);
1390 aoid = 0;
1391 }
1392 if (aoid == 0) {
1393 aoid = dmu_object_alloc(zfsvfs->z_os,
1394 otype, aclp->z_acl_bytes,
1395 otype == DMU_OT_ACL ?
1396 DMU_OT_SYSACL : DMU_OT_NONE,
1397 otype == DMU_OT_ACL ?
1398 DN_OLD_MAX_BONUSLEN : 0, tx);
1399 } else {
1400 (void) dmu_object_set_blocksize(zfsvfs->z_os,
1401 aoid, aclp->z_acl_bytes, 0, tx);
1402 }
1403 acl_phys.z_acl_extern_obj = aoid;
1404 for (aclnode = list_head(&aclp->z_acl); aclnode;
1405 aclnode = list_next(&aclp->z_acl, aclnode)) {
1406 if (aclnode->z_ace_count == 0)
1407 continue;
1408 dmu_write(zfsvfs->z_os, aoid, off,
1409 aclnode->z_size, aclnode->z_acldata, tx);
1410 off += aclnode->z_size;
1411 }
1412 } else {
1413 void *start = acl_phys.z_ace_data;
1414 /*
1415 * Migrating back embedded?
1416 */
1417 if (acl_phys.z_acl_extern_obj) {
1418 error = dmu_object_free(zfsvfs->z_os,
1419 acl_phys.z_acl_extern_obj, tx);
1420 if (error)
1421 return (error);
1422 acl_phys.z_acl_extern_obj = 0;
1423 }
1424
1425 for (aclnode = list_head(&aclp->z_acl); aclnode;
1426 aclnode = list_next(&aclp->z_acl, aclnode)) {
1427 if (aclnode->z_ace_count == 0)
1428 continue;
1429 bcopy(aclnode->z_acldata, start,
1430 aclnode->z_size);
1431 start = (caddr_t)start + aclnode->z_size;
1432 }
1433 }
1434 /*
1435 * If Old version then swap count/bytes to match old
1436 * layout of znode_acl_phys_t.
1437 */
1438 if (aclp->z_version == ZFS_ACL_VERSION_INITIAL) {
1439 acl_phys.z_acl_size = aclp->z_acl_count;
1440 acl_phys.z_acl_count = aclp->z_acl_bytes;
1441 } else {
1442 acl_phys.z_acl_size = aclp->z_acl_bytes;
1443 acl_phys.z_acl_count = aclp->z_acl_count;
1444 }
1445 acl_phys.z_acl_version = aclp->z_version;
1446
1447 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ZNODE_ACL(zfsvfs), NULL,
1448 &acl_phys, sizeof (acl_phys));
1449 }
1450
1451 /*
1452 * Replace ACL wide bits, but first clear them.
1453 */
1454 zp->z_pflags &= ~ZFS_ACL_WIDE_FLAGS;
1455
1456 zp->z_pflags |= aclp->z_hints;
1457
1458 if (ace_trivial_common(aclp, 0, zfs_ace_walk) == 0)
1459 zp->z_pflags |= ZFS_ACL_TRIVIAL;
1460
1461 zfs_tstamp_update_setup(zp, STATE_CHANGED, NULL, ctime);
1462 return (sa_bulk_update(zp->z_sa_hdl, bulk, count, tx));
1463 }
1464
1465 static void
1466 zfs_acl_chmod(zfsvfs_t *zfsvfs, uint64_t mode, zfs_acl_t *aclp)
1467 {
1468 void *acep = NULL;
1469 uint64_t who;
1470 int new_count, new_bytes;
1471 int ace_size;
1472 int entry_type;
1473 uint16_t iflags, type;
1474 uint32_t access_mask;
1475 zfs_acl_node_t *newnode;
1476 size_t abstract_size = aclp->z_ops->ace_abstract_size();
1477 void *zacep;
1478 uint32_t owner, group, everyone;
1479 uint32_t deny1, deny2, allow0;
1480
1481 new_count = new_bytes = 0;
1482
1483 acl_trivial_access_masks((mode_t)mode, &allow0, &deny1, &deny2,
1484 &owner, &group, &everyone);
1485
1486 newnode = zfs_acl_node_alloc((abstract_size * 6) + aclp->z_acl_bytes);
1487
1488 zacep = newnode->z_acldata;
1489 if (allow0) {
1490 zfs_set_ace(aclp, zacep, allow0, ALLOW, -1, ACE_OWNER);
1491 zacep = (void *)((uintptr_t)zacep + abstract_size);
1492 new_count++;
1493 new_bytes += abstract_size;
1494 }
1495 if (deny1) {
1496 zfs_set_ace(aclp, zacep, deny1, DENY, -1, ACE_OWNER);
1497 zacep = (void *)((uintptr_t)zacep + abstract_size);
1498 new_count++;
1499 new_bytes += abstract_size;
1500 }
1501 if (deny2) {
1502 zfs_set_ace(aclp, zacep, deny2, DENY, -1, OWNING_GROUP);
1503 zacep = (void *)((uintptr_t)zacep + abstract_size);
1504 new_count++;
1505 new_bytes += abstract_size;
1506 }
1507
1508 while ((acep = zfs_acl_next_ace(aclp, acep, &who, &access_mask,
1509 &iflags, &type))) {
1510 uint16_t inherit_flags;
1511
1512 entry_type = (iflags & ACE_TYPE_FLAGS);
1513 inherit_flags = (iflags & ALL_INHERIT);
1514
1515 if ((entry_type == ACE_OWNER || entry_type == ACE_EVERYONE ||
1516 (entry_type == OWNING_GROUP)) &&
1517 ((inherit_flags & ACE_INHERIT_ONLY_ACE) == 0)) {
1518 continue;
1519 }
1520
1521 if ((type != ALLOW && type != DENY) ||
1522 (inherit_flags & ACE_INHERIT_ONLY_ACE)) {
1523 if (inherit_flags)
1524 aclp->z_hints |= ZFS_INHERIT_ACE;
1525 switch (type) {
1526 case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE:
1527 case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE:
1528 case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE:
1529 case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE:
1530 aclp->z_hints |= ZFS_ACL_OBJ_ACE;
1531 break;
1532 }
1533 } else {
1534
1535 /*
1536 * Limit permissions to be no greater than
1537 * group permissions
1538 */
1539 if (zfsvfs->z_acl_inherit == ZFS_ACL_RESTRICTED) {
1540 if (!(mode & S_IRGRP))
1541 access_mask &= ~ACE_READ_DATA;
1542 if (!(mode & S_IWGRP))
1543 access_mask &=
1544 ~(ACE_WRITE_DATA|ACE_APPEND_DATA);
1545 if (!(mode & S_IXGRP))
1546 access_mask &= ~ACE_EXECUTE;
1547 access_mask &=
1548 ~(ACE_WRITE_OWNER|ACE_WRITE_ACL|
1549 ACE_WRITE_ATTRIBUTES|ACE_WRITE_NAMED_ATTRS);
1550 }
1551 }
1552 zfs_set_ace(aclp, zacep, access_mask, type, who, iflags);
1553 ace_size = aclp->z_ops->ace_size(acep);
1554 zacep = (void *)((uintptr_t)zacep + ace_size);
1555 new_count++;
1556 new_bytes += ace_size;
1557 }
1558 zfs_set_ace(aclp, zacep, owner, 0, -1, ACE_OWNER);
1559 zacep = (void *)((uintptr_t)zacep + abstract_size);
1560 zfs_set_ace(aclp, zacep, group, 0, -1, OWNING_GROUP);
1561 zacep = (void *)((uintptr_t)zacep + abstract_size);
1562 zfs_set_ace(aclp, zacep, everyone, 0, -1, ACE_EVERYONE);
1563
1564 new_count += 3;
1565 new_bytes += abstract_size * 3;
1566 zfs_acl_release_nodes(aclp);
1567 aclp->z_acl_count = new_count;
1568 aclp->z_acl_bytes = new_bytes;
1569 newnode->z_ace_count = new_count;
1570 newnode->z_size = new_bytes;
1571 list_insert_tail(&aclp->z_acl, newnode);
1572 }
1573
1574 void
1575 zfs_acl_chmod_setattr(znode_t *zp, zfs_acl_t **aclp, uint64_t mode)
1576 {
1577 mutex_enter(&zp->z_acl_lock);
1578 mutex_enter(&zp->z_lock);
1579 *aclp = zfs_acl_alloc(zfs_acl_version_zp(zp));
1580 (*aclp)->z_hints = zp->z_pflags & V4_ACL_WIDE_FLAGS;
1581 zfs_acl_chmod(ZTOZSB(zp), mode, *aclp);
1582 mutex_exit(&zp->z_lock);
1583 mutex_exit(&zp->z_acl_lock);
1584 ASSERT(*aclp);
1585 }
1586
1587 /*
1588 * strip off write_owner and write_acl
1589 */
1590 static void
1591 zfs_restricted_update(zfsvfs_t *zfsvfs, zfs_acl_t *aclp, void *acep)
1592 {
1593 uint32_t mask = aclp->z_ops->ace_mask_get(acep);
1594
1595 if ((zfsvfs->z_acl_inherit == ZFS_ACL_RESTRICTED) &&
1596 (aclp->z_ops->ace_type_get(acep) == ALLOW)) {
1597 mask &= ~RESTRICTED_CLEAR;
1598 aclp->z_ops->ace_mask_set(acep, mask);
1599 }
1600 }
1601
1602 /*
1603 * Should ACE be inherited?
1604 */
1605 static int
1606 zfs_ace_can_use(umode_t obj_mode, uint16_t acep_flags)
1607 {
1608 int iflags = (acep_flags & 0xf);
1609
1610 if (S_ISDIR(obj_mode) && (iflags & ACE_DIRECTORY_INHERIT_ACE))
1611 return (1);
1612 else if (iflags & ACE_FILE_INHERIT_ACE)
1613 return (!(S_ISDIR(obj_mode) &&
1614 (iflags & ACE_NO_PROPAGATE_INHERIT_ACE)));
1615 return (0);
1616 }
1617
1618 /*
1619 * inherit inheritable ACEs from parent
1620 */
1621 static zfs_acl_t *
1622 zfs_acl_inherit(zfsvfs_t *zfsvfs, umode_t obj_mode, zfs_acl_t *paclp,
1623 uint64_t mode, boolean_t *need_chmod)
1624 {
1625 void *pacep;
1626 void *acep;
1627 zfs_acl_node_t *aclnode;
1628 zfs_acl_t *aclp = NULL;
1629 uint64_t who;
1630 uint32_t access_mask;
1631 uint16_t iflags, newflags, type;
1632 size_t ace_size;
1633 void *data1, *data2;
1634 size_t data1sz, data2sz;
1635 boolean_t vdir = S_ISDIR(obj_mode);
1636 boolean_t vreg = S_ISREG(obj_mode);
1637 boolean_t passthrough, passthrough_x, noallow;
1638
1639 passthrough_x =
1640 zfsvfs->z_acl_inherit == ZFS_ACL_PASSTHROUGH_X;
1641 passthrough = passthrough_x ||
1642 zfsvfs->z_acl_inherit == ZFS_ACL_PASSTHROUGH;
1643 noallow =
1644 zfsvfs->z_acl_inherit == ZFS_ACL_NOALLOW;
1645
1646 *need_chmod = B_TRUE;
1647 pacep = NULL;
1648 aclp = zfs_acl_alloc(paclp->z_version);
1649 if (zfsvfs->z_acl_inherit == ZFS_ACL_DISCARD || S_ISLNK(obj_mode))
1650 return (aclp);
1651 while ((pacep = zfs_acl_next_ace(paclp, pacep, &who,
1652 &access_mask, &iflags, &type))) {
1653
1654 /*
1655 * don't inherit bogus ACEs
1656 */
1657 if (!zfs_acl_valid_ace_type(type, iflags))
1658 continue;
1659
1660 if (noallow && type == ALLOW)
1661 continue;
1662
1663 ace_size = aclp->z_ops->ace_size(pacep);
1664
1665 if (!zfs_ace_can_use(obj_mode, iflags))
1666 continue;
1667
1668 /*
1669 * If owner@, group@, or everyone@ inheritable
1670 * then zfs_acl_chmod() isn't needed.
1671 */
1672 if (passthrough &&
1673 ((iflags & (ACE_OWNER|ACE_EVERYONE)) ||
1674 ((iflags & OWNING_GROUP) ==
1675 OWNING_GROUP)) && (vreg || (vdir && (iflags &
1676 ACE_DIRECTORY_INHERIT_ACE)))) {
1677 *need_chmod = B_FALSE;
1678 }
1679
1680 if (!vdir && passthrough_x &&
1681 ((mode & (S_IXUSR | S_IXGRP | S_IXOTH)) == 0)) {
1682 access_mask &= ~ACE_EXECUTE;
1683 }
1684
1685 aclnode = zfs_acl_node_alloc(ace_size);
1686 list_insert_tail(&aclp->z_acl, aclnode);
1687 acep = aclnode->z_acldata;
1688
1689 zfs_set_ace(aclp, acep, access_mask, type,
1690 who, iflags|ACE_INHERITED_ACE);
1691
1692 /*
1693 * Copy special opaque data if any
1694 */
1695 if ((data1sz = paclp->z_ops->ace_data(pacep, &data1)) != 0) {
1696 VERIFY((data2sz = aclp->z_ops->ace_data(acep,
1697 &data2)) == data1sz);
1698 bcopy(data1, data2, data2sz);
1699 }
1700
1701 aclp->z_acl_count++;
1702 aclnode->z_ace_count++;
1703 aclp->z_acl_bytes += aclnode->z_size;
1704 newflags = aclp->z_ops->ace_flags_get(acep);
1705
1706 if (vdir)
1707 aclp->z_hints |= ZFS_INHERIT_ACE;
1708
1709 if ((iflags & ACE_NO_PROPAGATE_INHERIT_ACE) || !vdir) {
1710 newflags &= ~ALL_INHERIT;
1711 aclp->z_ops->ace_flags_set(acep,
1712 newflags|ACE_INHERITED_ACE);
1713 zfs_restricted_update(zfsvfs, aclp, acep);
1714 continue;
1715 }
1716
1717 ASSERT(vdir);
1718
1719 /*
1720 * If only FILE_INHERIT is set then turn on
1721 * inherit_only
1722 */
1723 if ((iflags & (ACE_FILE_INHERIT_ACE |
1724 ACE_DIRECTORY_INHERIT_ACE)) == ACE_FILE_INHERIT_ACE) {
1725 newflags |= ACE_INHERIT_ONLY_ACE;
1726 aclp->z_ops->ace_flags_set(acep,
1727 newflags|ACE_INHERITED_ACE);
1728 } else {
1729 newflags &= ~ACE_INHERIT_ONLY_ACE;
1730 aclp->z_ops->ace_flags_set(acep,
1731 newflags|ACE_INHERITED_ACE);
1732 }
1733 }
1734 return (aclp);
1735 }
1736
1737 /*
1738 * Create file system object initial permissions
1739 * including inheritable ACEs.
1740 */
1741 int
1742 zfs_acl_ids_create(znode_t *dzp, int flag, vattr_t *vap, cred_t *cr,
1743 vsecattr_t *vsecp, zfs_acl_ids_t *acl_ids)
1744 {
1745 int error;
1746 zfsvfs_t *zfsvfs = ZTOZSB(dzp);
1747 zfs_acl_t *paclp;
1748 gid_t gid = vap->va_gid;
1749 boolean_t need_chmod = B_TRUE;
1750 boolean_t inherited = B_FALSE;
1751
1752 bzero(acl_ids, sizeof (zfs_acl_ids_t));
1753 acl_ids->z_mode = vap->va_mode;
1754
1755 if (vsecp)
1756 if ((error = zfs_vsec_2_aclp(zfsvfs, vap->va_mode, vsecp,
1757 cr, &acl_ids->z_fuidp, &acl_ids->z_aclp)) != 0)
1758 return (error);
1759
1760 acl_ids->z_fuid = vap->va_uid;
1761 acl_ids->z_fgid = vap->va_gid;
1762 #ifdef HAVE_KSID
1763 /*
1764 * Determine uid and gid.
1765 */
1766 if ((flag & IS_ROOT_NODE) || zfsvfs->z_replay ||
1767 ((flag & IS_XATTR) && (S_ISDIR(vap->va_mode)))) {
1768 acl_ids->z_fuid = zfs_fuid_create(zfsvfs, (uint64_t)vap->va_uid,
1769 cr, ZFS_OWNER, &acl_ids->z_fuidp);
1770 acl_ids->z_fgid = zfs_fuid_create(zfsvfs, (uint64_t)vap->va_gid,
1771 cr, ZFS_GROUP, &acl_ids->z_fuidp);
1772 gid = vap->va_gid;
1773 } else {
1774 acl_ids->z_fuid = zfs_fuid_create_cred(zfsvfs, ZFS_OWNER,
1775 cr, &acl_ids->z_fuidp);
1776 acl_ids->z_fgid = 0;
1777 if (vap->va_mask & AT_GID) {
1778 acl_ids->z_fgid = zfs_fuid_create(zfsvfs,
1779 (uint64_t)vap->va_gid,
1780 cr, ZFS_GROUP, &acl_ids->z_fuidp);
1781 gid = vap->va_gid;
1782 if (acl_ids->z_fgid != KGID_TO_SGID(ZTOI(dzp)->i_gid) &&
1783 !groupmember(vap->va_gid, cr) &&
1784 secpolicy_vnode_create_gid(cr) != 0)
1785 acl_ids->z_fgid = 0;
1786 }
1787 if (acl_ids->z_fgid == 0) {
1788 if (dzp->z_mode & S_ISGID) {
1789 char *domain;
1790 uint32_t rid;
1791
1792 acl_ids->z_fgid = KGID_TO_SGID(
1793 ZTOI(dzp)->i_gid);
1794 gid = zfs_fuid_map_id(zfsvfs, acl_ids->z_fgid,
1795 cr, ZFS_GROUP);
1796
1797 if (zfsvfs->z_use_fuids &&
1798 IS_EPHEMERAL(acl_ids->z_fgid)) {
1799 domain = zfs_fuid_idx_domain(
1800 &zfsvfs->z_fuid_idx,
1801 FUID_INDEX(acl_ids->z_fgid));
1802 rid = FUID_RID(acl_ids->z_fgid);
1803 zfs_fuid_node_add(&acl_ids->z_fuidp,
1804 domain, rid,
1805 FUID_INDEX(acl_ids->z_fgid),
1806 acl_ids->z_fgid, ZFS_GROUP);
1807 }
1808 } else {
1809 acl_ids->z_fgid = zfs_fuid_create_cred(zfsvfs,
1810 ZFS_GROUP, cr, &acl_ids->z_fuidp);
1811 gid = crgetgid(cr);
1812 }
1813 }
1814 }
1815 #endif /* HAVE_KSID */
1816
1817 /*
1818 * If we're creating a directory, and the parent directory has the
1819 * set-GID bit set, set in on the new directory.
1820 * Otherwise, if the user is neither privileged nor a member of the
1821 * file's new group, clear the file's set-GID bit.
1822 */
1823
1824 if (!(flag & IS_ROOT_NODE) && (dzp->z_mode & S_ISGID) &&
1825 (S_ISDIR(vap->va_mode))) {
1826 acl_ids->z_mode |= S_ISGID;
1827 } else {
1828 if ((acl_ids->z_mode & S_ISGID) &&
1829 secpolicy_vnode_setids_setgids(cr, gid) != 0)
1830 acl_ids->z_mode &= ~S_ISGID;
1831 }
1832
1833 if (acl_ids->z_aclp == NULL) {
1834 mutex_enter(&dzp->z_acl_lock);
1835 mutex_enter(&dzp->z_lock);
1836 if (!(flag & IS_ROOT_NODE) && (S_ISDIR(ZTOI(dzp)->i_mode) &&
1837 (dzp->z_pflags & ZFS_INHERIT_ACE)) &&
1838 !(dzp->z_pflags & ZFS_XATTR)) {
1839 VERIFY(0 == zfs_acl_node_read(dzp, B_TRUE,
1840 &paclp, B_FALSE));
1841 acl_ids->z_aclp = zfs_acl_inherit(zfsvfs,
1842 vap->va_mode, paclp, acl_ids->z_mode, &need_chmod);
1843 inherited = B_TRUE;
1844 } else {
1845 acl_ids->z_aclp =
1846 zfs_acl_alloc(zfs_acl_version_zp(dzp));
1847 acl_ids->z_aclp->z_hints |= ZFS_ACL_TRIVIAL;
1848 }
1849 mutex_exit(&dzp->z_lock);
1850 mutex_exit(&dzp->z_acl_lock);
1851 if (need_chmod) {
1852 acl_ids->z_aclp->z_hints |= S_ISDIR(vap->va_mode) ?
1853 ZFS_ACL_AUTO_INHERIT : 0;
1854 zfs_acl_chmod(zfsvfs, acl_ids->z_mode, acl_ids->z_aclp);
1855 }
1856 }
1857
1858 if (inherited || vsecp) {
1859 acl_ids->z_mode = zfs_mode_compute(acl_ids->z_mode,
1860 acl_ids->z_aclp, &acl_ids->z_aclp->z_hints,
1861 acl_ids->z_fuid, acl_ids->z_fgid);
1862 if (ace_trivial_common(acl_ids->z_aclp, 0, zfs_ace_walk) == 0)
1863 acl_ids->z_aclp->z_hints |= ZFS_ACL_TRIVIAL;
1864 }
1865
1866 return (0);
1867 }
1868
1869 /*
1870 * Free ACL and fuid_infop, but not the acl_ids structure
1871 */
1872 void
1873 zfs_acl_ids_free(zfs_acl_ids_t *acl_ids)
1874 {
1875 if (acl_ids->z_aclp)
1876 zfs_acl_free(acl_ids->z_aclp);
1877 if (acl_ids->z_fuidp)
1878 zfs_fuid_info_free(acl_ids->z_fuidp);
1879 acl_ids->z_aclp = NULL;
1880 acl_ids->z_fuidp = NULL;
1881 }
1882
1883 boolean_t
1884 zfs_acl_ids_overquota(zfsvfs_t *zv, zfs_acl_ids_t *acl_ids, uint64_t projid)
1885 {
1886 return (zfs_id_overquota(zv, DMU_USERUSED_OBJECT, acl_ids->z_fuid) ||
1887 zfs_id_overquota(zv, DMU_GROUPUSED_OBJECT, acl_ids->z_fgid) ||
1888 (projid != ZFS_DEFAULT_PROJID && projid != ZFS_INVALID_PROJID &&
1889 zfs_id_overquota(zv, DMU_PROJECTUSED_OBJECT, projid)));
1890 }
1891
1892 /*
1893 * Retrieve a file's ACL
1894 */
1895 int
1896 zfs_getacl(znode_t *zp, vsecattr_t *vsecp, boolean_t skipaclchk, cred_t *cr)
1897 {
1898 zfs_acl_t *aclp;
1899 ulong_t mask;
1900 int error;
1901 int count = 0;
1902 int largeace = 0;
1903
1904 mask = vsecp->vsa_mask & (VSA_ACE | VSA_ACECNT |
1905 VSA_ACE_ACLFLAGS | VSA_ACE_ALLTYPES);
1906
1907 if (mask == 0)
1908 return (SET_ERROR(ENOSYS));
1909
1910 if ((error = zfs_zaccess(zp, ACE_READ_ACL, 0, skipaclchk, cr)))
1911 return (error);
1912
1913 mutex_enter(&zp->z_acl_lock);
1914
1915 error = zfs_acl_node_read(zp, B_FALSE, &aclp, B_FALSE);
1916 if (error != 0) {
1917 mutex_exit(&zp->z_acl_lock);
1918 return (error);
1919 }
1920
1921 /*
1922 * Scan ACL to determine number of ACEs
1923 */
1924 if ((zp->z_pflags & ZFS_ACL_OBJ_ACE) && !(mask & VSA_ACE_ALLTYPES)) {
1925 void *zacep = NULL;
1926 uint64_t who;
1927 uint32_t access_mask;
1928 uint16_t type, iflags;
1929
1930 while ((zacep = zfs_acl_next_ace(aclp, zacep,
1931 &who, &access_mask, &iflags, &type))) {
1932 switch (type) {
1933 case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE:
1934 case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE:
1935 case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE:
1936 case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE:
1937 largeace++;
1938 continue;
1939 default:
1940 count++;
1941 }
1942 }
1943 vsecp->vsa_aclcnt = count;
1944 } else
1945 count = (int)aclp->z_acl_count;
1946
1947 if (mask & VSA_ACECNT) {
1948 vsecp->vsa_aclcnt = count;
1949 }
1950
1951 if (mask & VSA_ACE) {
1952 size_t aclsz;
1953
1954 aclsz = count * sizeof (ace_t) +
1955 sizeof (ace_object_t) * largeace;
1956
1957 vsecp->vsa_aclentp = kmem_alloc(aclsz, KM_SLEEP);
1958 vsecp->vsa_aclentsz = aclsz;
1959
1960 if (aclp->z_version == ZFS_ACL_VERSION_FUID)
1961 zfs_copy_fuid_2_ace(ZTOZSB(zp), aclp, cr,
1962 vsecp->vsa_aclentp, !(mask & VSA_ACE_ALLTYPES));
1963 else {
1964 zfs_acl_node_t *aclnode;
1965 void *start = vsecp->vsa_aclentp;
1966
1967 for (aclnode = list_head(&aclp->z_acl); aclnode;
1968 aclnode = list_next(&aclp->z_acl, aclnode)) {
1969 bcopy(aclnode->z_acldata, start,
1970 aclnode->z_size);
1971 start = (caddr_t)start + aclnode->z_size;
1972 }
1973 ASSERT((caddr_t)start - (caddr_t)vsecp->vsa_aclentp ==
1974 aclp->z_acl_bytes);
1975 }
1976 }
1977 if (mask & VSA_ACE_ACLFLAGS) {
1978 vsecp->vsa_aclflags = 0;
1979 if (zp->z_pflags & ZFS_ACL_DEFAULTED)
1980 vsecp->vsa_aclflags |= ACL_DEFAULTED;
1981 if (zp->z_pflags & ZFS_ACL_PROTECTED)
1982 vsecp->vsa_aclflags |= ACL_PROTECTED;
1983 if (zp->z_pflags & ZFS_ACL_AUTO_INHERIT)
1984 vsecp->vsa_aclflags |= ACL_AUTO_INHERIT;
1985 }
1986
1987 mutex_exit(&zp->z_acl_lock);
1988
1989 return (0);
1990 }
1991
1992 int
1993 zfs_vsec_2_aclp(zfsvfs_t *zfsvfs, umode_t obj_mode,
1994 vsecattr_t *vsecp, cred_t *cr, zfs_fuid_info_t **fuidp, zfs_acl_t **zaclp)
1995 {
1996 zfs_acl_t *aclp;
1997 zfs_acl_node_t *aclnode;
1998 int aclcnt = vsecp->vsa_aclcnt;
1999 int error;
2000
2001 if (vsecp->vsa_aclcnt > MAX_ACL_ENTRIES || vsecp->vsa_aclcnt <= 0)
2002 return (SET_ERROR(EINVAL));
2003
2004 aclp = zfs_acl_alloc(zfs_acl_version(zfsvfs->z_version));
2005
2006 aclp->z_hints = 0;
2007 aclnode = zfs_acl_node_alloc(aclcnt * sizeof (zfs_object_ace_t));
2008 if (aclp->z_version == ZFS_ACL_VERSION_INITIAL) {
2009 if ((error = zfs_copy_ace_2_oldace(obj_mode, aclp,
2010 (ace_t *)vsecp->vsa_aclentp, aclnode->z_acldata,
2011 aclcnt, &aclnode->z_size)) != 0) {
2012 zfs_acl_free(aclp);
2013 zfs_acl_node_free(aclnode);
2014 return (error);
2015 }
2016 } else {
2017 if ((error = zfs_copy_ace_2_fuid(zfsvfs, obj_mode, aclp,
2018 vsecp->vsa_aclentp, aclnode->z_acldata, aclcnt,
2019 &aclnode->z_size, fuidp, cr)) != 0) {
2020 zfs_acl_free(aclp);
2021 zfs_acl_node_free(aclnode);
2022 return (error);
2023 }
2024 }
2025 aclp->z_acl_bytes = aclnode->z_size;
2026 aclnode->z_ace_count = aclcnt;
2027 aclp->z_acl_count = aclcnt;
2028 list_insert_head(&aclp->z_acl, aclnode);
2029
2030 /*
2031 * If flags are being set then add them to z_hints
2032 */
2033 if (vsecp->vsa_mask & VSA_ACE_ACLFLAGS) {
2034 if (vsecp->vsa_aclflags & ACL_PROTECTED)
2035 aclp->z_hints |= ZFS_ACL_PROTECTED;
2036 if (vsecp->vsa_aclflags & ACL_DEFAULTED)
2037 aclp->z_hints |= ZFS_ACL_DEFAULTED;
2038 if (vsecp->vsa_aclflags & ACL_AUTO_INHERIT)
2039 aclp->z_hints |= ZFS_ACL_AUTO_INHERIT;
2040 }
2041
2042 *zaclp = aclp;
2043
2044 return (0);
2045 }
2046
2047 /*
2048 * Set a file's ACL
2049 */
2050 int
2051 zfs_setacl(znode_t *zp, vsecattr_t *vsecp, boolean_t skipaclchk, cred_t *cr)
2052 {
2053 zfsvfs_t *zfsvfs = ZTOZSB(zp);
2054 zilog_t *zilog = zfsvfs->z_log;
2055 ulong_t mask = vsecp->vsa_mask & (VSA_ACE | VSA_ACECNT);
2056 dmu_tx_t *tx;
2057 int error;
2058 zfs_acl_t *aclp;
2059 zfs_fuid_info_t *fuidp = NULL;
2060 boolean_t fuid_dirtied;
2061 uint64_t acl_obj;
2062
2063 if (mask == 0)
2064 return (SET_ERROR(ENOSYS));
2065
2066 if (zp->z_pflags & ZFS_IMMUTABLE)
2067 return (SET_ERROR(EPERM));
2068
2069 if ((error = zfs_zaccess(zp, ACE_WRITE_ACL, 0, skipaclchk, cr)))
2070 return (error);
2071
2072 error = zfs_vsec_2_aclp(zfsvfs, ZTOI(zp)->i_mode, vsecp, cr, &fuidp,
2073 &aclp);
2074 if (error)
2075 return (error);
2076
2077 /*
2078 * If ACL wide flags aren't being set then preserve any
2079 * existing flags.
2080 */
2081 if (!(vsecp->vsa_mask & VSA_ACE_ACLFLAGS)) {
2082 aclp->z_hints |=
2083 (zp->z_pflags & V4_ACL_WIDE_FLAGS);
2084 }
2085 top:
2086 mutex_enter(&zp->z_acl_lock);
2087 mutex_enter(&zp->z_lock);
2088
2089 tx = dmu_tx_create(zfsvfs->z_os);
2090
2091 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE);
2092
2093 fuid_dirtied = zfsvfs->z_fuid_dirty;
2094 if (fuid_dirtied)
2095 zfs_fuid_txhold(zfsvfs, tx);
2096
2097 /*
2098 * If old version and ACL won't fit in bonus and we aren't
2099 * upgrading then take out necessary DMU holds
2100 */
2101
2102 if ((acl_obj = zfs_external_acl(zp)) != 0) {
2103 if (zfsvfs->z_version >= ZPL_VERSION_FUID &&
2104 zfs_znode_acl_version(zp) <= ZFS_ACL_VERSION_INITIAL) {
2105 dmu_tx_hold_free(tx, acl_obj, 0,
2106 DMU_OBJECT_END);
2107 dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0,
2108 aclp->z_acl_bytes);
2109 } else {
2110 dmu_tx_hold_write(tx, acl_obj, 0, aclp->z_acl_bytes);
2111 }
2112 } else if (!zp->z_is_sa && aclp->z_acl_bytes > ZFS_ACE_SPACE) {
2113 dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, aclp->z_acl_bytes);
2114 }
2115
2116 zfs_sa_upgrade_txholds(tx, zp);
2117 error = dmu_tx_assign(tx, TXG_NOWAIT);
2118 if (error) {
2119 mutex_exit(&zp->z_acl_lock);
2120 mutex_exit(&zp->z_lock);
2121
2122 if (error == ERESTART) {
2123 dmu_tx_wait(tx);
2124 dmu_tx_abort(tx);
2125 goto top;
2126 }
2127 dmu_tx_abort(tx);
2128 zfs_acl_free(aclp);
2129 return (error);
2130 }
2131
2132 error = zfs_aclset_common(zp, aclp, cr, tx);
2133 ASSERT(error == 0);
2134 ASSERT(zp->z_acl_cached == NULL);
2135 zp->z_acl_cached = aclp;
2136
2137 if (fuid_dirtied)
2138 zfs_fuid_sync(zfsvfs, tx);
2139
2140 zfs_log_acl(zilog, tx, zp, vsecp, fuidp);
2141
2142 if (fuidp)
2143 zfs_fuid_info_free(fuidp);
2144 dmu_tx_commit(tx);
2145
2146 mutex_exit(&zp->z_lock);
2147 mutex_exit(&zp->z_acl_lock);
2148
2149 return (error);
2150 }
2151
2152 /*
2153 * Check accesses of interest (AoI) against attributes of the dataset
2154 * such as read-only. Returns zero if no AoI conflict with dataset
2155 * attributes, otherwise an appropriate errno is returned.
2156 */
2157 static int
2158 zfs_zaccess_dataset_check(znode_t *zp, uint32_t v4_mode)
2159 {
2160 if ((v4_mode & WRITE_MASK) && (zfs_is_readonly(ZTOZSB(zp))) &&
2161 (!S_ISDEV(ZTOI(zp)->i_mode) ||
2162 (S_ISDEV(ZTOI(zp)->i_mode) && (v4_mode & WRITE_MASK_ATTRS)))) {
2163 return (SET_ERROR(EROFS));
2164 }
2165
2166 /*
2167 * Only check for READONLY on non-directories.
2168 */
2169 if ((v4_mode & WRITE_MASK_DATA) &&
2170 ((!S_ISDIR(ZTOI(zp)->i_mode) &&
2171 (zp->z_pflags & (ZFS_READONLY | ZFS_IMMUTABLE))) ||
2172 (S_ISDIR(ZTOI(zp)->i_mode) &&
2173 (zp->z_pflags & ZFS_IMMUTABLE)))) {
2174 return (SET_ERROR(EPERM));
2175 }
2176
2177 if ((v4_mode & (ACE_DELETE | ACE_DELETE_CHILD)) &&
2178 (zp->z_pflags & ZFS_NOUNLINK)) {
2179 return (SET_ERROR(EPERM));
2180 }
2181
2182 if (((v4_mode & (ACE_READ_DATA|ACE_EXECUTE)) &&
2183 (zp->z_pflags & ZFS_AV_QUARANTINED))) {
2184 return (SET_ERROR(EACCES));
2185 }
2186
2187 return (0);
2188 }
2189
2190 /*
2191 * The primary usage of this function is to loop through all of the
2192 * ACEs in the znode, determining what accesses of interest (AoI) to
2193 * the caller are allowed or denied. The AoI are expressed as bits in
2194 * the working_mode parameter. As each ACE is processed, bits covered
2195 * by that ACE are removed from the working_mode. This removal
2196 * facilitates two things. The first is that when the working mode is
2197 * empty (= 0), we know we've looked at all the AoI. The second is
2198 * that the ACE interpretation rules don't allow a later ACE to undo
2199 * something granted or denied by an earlier ACE. Removing the
2200 * discovered access or denial enforces this rule. At the end of
2201 * processing the ACEs, all AoI that were found to be denied are
2202 * placed into the working_mode, giving the caller a mask of denied
2203 * accesses. Returns:
2204 * 0 if all AoI granted
2205 * EACCES if the denied mask is non-zero
2206 * other error if abnormal failure (e.g., IO error)
2207 *
2208 * A secondary usage of the function is to determine if any of the
2209 * AoI are granted. If an ACE grants any access in
2210 * the working_mode, we immediately short circuit out of the function.
2211 * This mode is chosen by setting anyaccess to B_TRUE. The
2212 * working_mode is not a denied access mask upon exit if the function
2213 * is used in this manner.
2214 */
2215 static int
2216 zfs_zaccess_aces_check(znode_t *zp, uint32_t *working_mode,
2217 boolean_t anyaccess, cred_t *cr)
2218 {
2219 zfsvfs_t *zfsvfs = ZTOZSB(zp);
2220 zfs_acl_t *aclp;
2221 int error;
2222 uid_t uid = crgetuid(cr);
2223 uint64_t who;
2224 uint16_t type, iflags;
2225 uint16_t entry_type;
2226 uint32_t access_mask;
2227 uint32_t deny_mask = 0;
2228 zfs_ace_hdr_t *acep = NULL;
2229 boolean_t checkit;
2230 uid_t gowner;
2231 uid_t fowner;
2232
2233 zfs_fuid_map_ids(zp, cr, &fowner, &gowner);
2234
2235 mutex_enter(&zp->z_acl_lock);
2236
2237 error = zfs_acl_node_read(zp, B_FALSE, &aclp, B_FALSE);
2238 if (error != 0) {
2239 mutex_exit(&zp->z_acl_lock);
2240 return (error);
2241 }
2242
2243 ASSERT(zp->z_acl_cached);
2244
2245 while ((acep = zfs_acl_next_ace(aclp, acep, &who, &access_mask,
2246 &iflags, &type))) {
2247 uint32_t mask_matched;
2248
2249 if (!zfs_acl_valid_ace_type(type, iflags))
2250 continue;
2251
2252 if (S_ISDIR(ZTOI(zp)->i_mode) &&
2253 (iflags & ACE_INHERIT_ONLY_ACE))
2254 continue;
2255
2256 /* Skip ACE if it does not affect any AoI */
2257 mask_matched = (access_mask & *working_mode);
2258 if (!mask_matched)
2259 continue;
2260
2261 entry_type = (iflags & ACE_TYPE_FLAGS);
2262
2263 checkit = B_FALSE;
2264
2265 switch (entry_type) {
2266 case ACE_OWNER:
2267 if (uid == fowner)
2268 checkit = B_TRUE;
2269 break;
2270 case OWNING_GROUP:
2271 who = gowner;
2272 /*FALLTHROUGH*/
2273 case ACE_IDENTIFIER_GROUP:
2274 checkit = zfs_groupmember(zfsvfs, who, cr);
2275 break;
2276 case ACE_EVERYONE:
2277 checkit = B_TRUE;
2278 break;
2279
2280 /* USER Entry */
2281 default:
2282 if (entry_type == 0) {
2283 uid_t newid;
2284
2285 newid = zfs_fuid_map_id(zfsvfs, who, cr,
2286 ZFS_ACE_USER);
2287 if (newid != IDMAP_WK_CREATOR_OWNER_UID &&
2288 uid == newid)
2289 checkit = B_TRUE;
2290 break;
2291 } else {
2292 mutex_exit(&zp->z_acl_lock);
2293 return (SET_ERROR(EIO));
2294 }
2295 }
2296
2297 if (checkit) {
2298 if (type == DENY) {
2299 DTRACE_PROBE3(zfs__ace__denies,
2300 znode_t *, zp,
2301 zfs_ace_hdr_t *, acep,
2302 uint32_t, mask_matched);
2303 deny_mask |= mask_matched;
2304 } else {
2305 DTRACE_PROBE3(zfs__ace__allows,
2306 znode_t *, zp,
2307 zfs_ace_hdr_t *, acep,
2308 uint32_t, mask_matched);
2309 if (anyaccess) {
2310 mutex_exit(&zp->z_acl_lock);
2311 return (0);
2312 }
2313 }
2314 *working_mode &= ~mask_matched;
2315 }
2316
2317 /* Are we done? */
2318 if (*working_mode == 0)
2319 break;
2320 }
2321
2322 mutex_exit(&zp->z_acl_lock);
2323
2324 /* Put the found 'denies' back on the working mode */
2325 if (deny_mask) {
2326 *working_mode |= deny_mask;
2327 return (SET_ERROR(EACCES));
2328 } else if (*working_mode) {
2329 return (-1);
2330 }
2331
2332 return (0);
2333 }
2334
2335 /*
2336 * Return true if any access whatsoever granted, we don't actually
2337 * care what access is granted.
2338 */
2339 boolean_t
2340 zfs_has_access(znode_t *zp, cred_t *cr)
2341 {
2342 uint32_t have = ACE_ALL_PERMS;
2343
2344 if (zfs_zaccess_aces_check(zp, &have, B_TRUE, cr) != 0) {
2345 uid_t owner;
2346
2347 owner = zfs_fuid_map_id(ZTOZSB(zp),
2348 KUID_TO_SUID(ZTOI(zp)->i_uid), cr, ZFS_OWNER);
2349 return (secpolicy_vnode_any_access(cr, ZTOI(zp), owner) == 0);
2350 }
2351 return (B_TRUE);
2352 }
2353
2354 static int
2355 zfs_zaccess_common(znode_t *zp, uint32_t v4_mode, uint32_t *working_mode,
2356 boolean_t *check_privs, boolean_t skipaclchk, cred_t *cr)
2357 {
2358 zfsvfs_t *zfsvfs = ZTOZSB(zp);
2359 int err;
2360
2361 *working_mode = v4_mode;
2362 *check_privs = B_TRUE;
2363
2364 /*
2365 * Short circuit empty requests
2366 */
2367 if (v4_mode == 0 || zfsvfs->z_replay) {
2368 *working_mode = 0;
2369 return (0);
2370 }
2371
2372 if ((err = zfs_zaccess_dataset_check(zp, v4_mode)) != 0) {
2373 *check_privs = B_FALSE;
2374 return (err);
2375 }
2376
2377 /*
2378 * The caller requested that the ACL check be skipped. This
2379 * would only happen if the caller checked VOP_ACCESS() with a
2380 * 32 bit ACE mask and already had the appropriate permissions.
2381 */
2382 if (skipaclchk) {
2383 *working_mode = 0;
2384 return (0);
2385 }
2386
2387 return (zfs_zaccess_aces_check(zp, working_mode, B_FALSE, cr));
2388 }
2389
2390 static int
2391 zfs_zaccess_append(znode_t *zp, uint32_t *working_mode, boolean_t *check_privs,
2392 cred_t *cr)
2393 {
2394 if (*working_mode != ACE_WRITE_DATA)
2395 return (SET_ERROR(EACCES));
2396
2397 return (zfs_zaccess_common(zp, ACE_APPEND_DATA, working_mode,
2398 check_privs, B_FALSE, cr));
2399 }
2400
2401 int
2402 zfs_fastaccesschk_execute(znode_t *zdp, cred_t *cr)
2403 {
2404 boolean_t owner = B_FALSE;
2405 boolean_t groupmbr = B_FALSE;
2406 boolean_t is_attr;
2407 uid_t uid = crgetuid(cr);
2408 int error;
2409
2410 if (zdp->z_pflags & ZFS_AV_QUARANTINED)
2411 return (SET_ERROR(EACCES));
2412
2413 is_attr = ((zdp->z_pflags & ZFS_XATTR) &&
2414 (S_ISDIR(ZTOI(zdp)->i_mode)));
2415 if (is_attr)
2416 goto slow;
2417
2418
2419 mutex_enter(&zdp->z_acl_lock);
2420
2421 if (zdp->z_pflags & ZFS_NO_EXECS_DENIED) {
2422 mutex_exit(&zdp->z_acl_lock);
2423 return (0);
2424 }
2425
2426 if (KUID_TO_SUID(ZTOI(zdp)->i_uid) != 0 ||
2427 KGID_TO_SGID(ZTOI(zdp)->i_gid) != 0) {
2428 mutex_exit(&zdp->z_acl_lock);
2429 goto slow;
2430 }
2431
2432 if (uid == KUID_TO_SUID(ZTOI(zdp)->i_uid)) {
2433 owner = B_TRUE;
2434 if (zdp->z_mode & S_IXUSR) {
2435 mutex_exit(&zdp->z_acl_lock);
2436 return (0);
2437 } else {
2438 mutex_exit(&zdp->z_acl_lock);
2439 goto slow;
2440 }
2441 }
2442 if (groupmember(KGID_TO_SGID(ZTOI(zdp)->i_gid), cr)) {
2443 groupmbr = B_TRUE;
2444 if (zdp->z_mode & S_IXGRP) {
2445 mutex_exit(&zdp->z_acl_lock);
2446 return (0);
2447 } else {
2448 mutex_exit(&zdp->z_acl_lock);
2449 goto slow;
2450 }
2451 }
2452 if (!owner && !groupmbr) {
2453 if (zdp->z_mode & S_IXOTH) {
2454 mutex_exit(&zdp->z_acl_lock);
2455 return (0);
2456 }
2457 }
2458
2459 mutex_exit(&zdp->z_acl_lock);
2460
2461 slow:
2462 DTRACE_PROBE(zfs__fastpath__execute__access__miss);
2463 ZFS_ENTER(ZTOZSB(zdp));
2464 error = zfs_zaccess(zdp, ACE_EXECUTE, 0, B_FALSE, cr);
2465 ZFS_EXIT(ZTOZSB(zdp));
2466 return (error);
2467 }
2468
2469 /*
2470 * Determine whether Access should be granted/denied.
2471 *
2472 * The least priv subsystem is always consulted as a basic privilege
2473 * can define any form of access.
2474 */
2475 int
2476 zfs_zaccess(znode_t *zp, int mode, int flags, boolean_t skipaclchk, cred_t *cr)
2477 {
2478 uint32_t working_mode;
2479 int error;
2480 int is_attr;
2481 boolean_t check_privs;
2482 znode_t *xzp;
2483 znode_t *check_zp = zp;
2484 mode_t needed_bits;
2485 uid_t owner;
2486
2487 is_attr = ((zp->z_pflags & ZFS_XATTR) && S_ISDIR(ZTOI(zp)->i_mode));
2488
2489 /*
2490 * If attribute then validate against base file
2491 */
2492 if (is_attr) {
2493 if ((error = zfs_zget(ZTOZSB(zp),
2494 zp->z_xattr_parent, &xzp)) != 0) {
2495 return (error);
2496 }
2497
2498 check_zp = xzp;
2499
2500 /*
2501 * fixup mode to map to xattr perms
2502 */
2503
2504 if (mode & (ACE_WRITE_DATA|ACE_APPEND_DATA)) {
2505 mode &= ~(ACE_WRITE_DATA|ACE_APPEND_DATA);
2506 mode |= ACE_WRITE_NAMED_ATTRS;
2507 }
2508
2509 if (mode & (ACE_READ_DATA|ACE_EXECUTE)) {
2510 mode &= ~(ACE_READ_DATA|ACE_EXECUTE);
2511 mode |= ACE_READ_NAMED_ATTRS;
2512 }
2513 }
2514
2515 owner = zfs_fuid_map_id(ZTOZSB(zp), KUID_TO_SUID(ZTOI(zp)->i_uid),
2516 cr, ZFS_OWNER);
2517 /*
2518 * Map the bits required to the standard inode flags
2519 * S_IRUSR|S_IWUSR|S_IXUSR in the needed_bits. Map the bits
2520 * mapped by working_mode (currently missing) in missing_bits.
2521 * Call secpolicy_vnode_access2() with (needed_bits & ~checkmode),
2522 * needed_bits.
2523 */
2524 needed_bits = 0;
2525
2526 working_mode = mode;
2527 if ((working_mode & (ACE_READ_ACL|ACE_READ_ATTRIBUTES)) &&
2528 owner == crgetuid(cr))
2529 working_mode &= ~(ACE_READ_ACL|ACE_READ_ATTRIBUTES);
2530
2531 if (working_mode & (ACE_READ_DATA|ACE_READ_NAMED_ATTRS|
2532 ACE_READ_ACL|ACE_READ_ATTRIBUTES|ACE_SYNCHRONIZE))
2533 needed_bits |= S_IRUSR;
2534 if (working_mode & (ACE_WRITE_DATA|ACE_WRITE_NAMED_ATTRS|
2535 ACE_APPEND_DATA|ACE_WRITE_ATTRIBUTES|ACE_SYNCHRONIZE))
2536 needed_bits |= S_IWUSR;
2537 if (working_mode & ACE_EXECUTE)
2538 needed_bits |= S_IXUSR;
2539
2540 if ((error = zfs_zaccess_common(check_zp, mode, &working_mode,
2541 &check_privs, skipaclchk, cr)) == 0) {
2542 if (is_attr)
2543 iput(ZTOI(xzp));
2544 return (secpolicy_vnode_access2(cr, ZTOI(zp), owner,
2545 needed_bits, needed_bits));
2546 }
2547
2548 if (error && !check_privs) {
2549 if (is_attr)
2550 iput(ZTOI(xzp));
2551 return (error);
2552 }
2553
2554 if (error && (flags & V_APPEND)) {
2555 error = zfs_zaccess_append(zp, &working_mode, &check_privs, cr);
2556 }
2557
2558 if (error && check_privs) {
2559 mode_t checkmode = 0;
2560
2561 /*
2562 * First check for implicit owner permission on
2563 * read_acl/read_attributes
2564 */
2565
2566 error = 0;
2567 ASSERT(working_mode != 0);
2568
2569 if ((working_mode & (ACE_READ_ACL|ACE_READ_ATTRIBUTES) &&
2570 owner == crgetuid(cr)))
2571 working_mode &= ~(ACE_READ_ACL|ACE_READ_ATTRIBUTES);
2572
2573 if (working_mode & (ACE_READ_DATA|ACE_READ_NAMED_ATTRS|
2574 ACE_READ_ACL|ACE_READ_ATTRIBUTES|ACE_SYNCHRONIZE))
2575 checkmode |= S_IRUSR;
2576 if (working_mode & (ACE_WRITE_DATA|ACE_WRITE_NAMED_ATTRS|
2577 ACE_APPEND_DATA|ACE_WRITE_ATTRIBUTES|ACE_SYNCHRONIZE))
2578 checkmode |= S_IWUSR;
2579 if (working_mode & ACE_EXECUTE)
2580 checkmode |= S_IXUSR;
2581
2582 error = secpolicy_vnode_access2(cr, ZTOI(check_zp), owner,
2583 needed_bits & ~checkmode, needed_bits);
2584
2585 if (error == 0 && (working_mode & ACE_WRITE_OWNER))
2586 error = secpolicy_vnode_chown(cr, owner);
2587 if (error == 0 && (working_mode & ACE_WRITE_ACL))
2588 error = secpolicy_vnode_setdac(cr, owner);
2589
2590 if (error == 0 && (working_mode &
2591 (ACE_DELETE|ACE_DELETE_CHILD)))
2592 error = secpolicy_vnode_remove(cr);
2593
2594 if (error == 0 && (working_mode & ACE_SYNCHRONIZE)) {
2595 error = secpolicy_vnode_chown(cr, owner);
2596 }
2597 if (error == 0) {
2598 /*
2599 * See if any bits other than those already checked
2600 * for are still present. If so then return EACCES
2601 */
2602 if (working_mode & ~(ZFS_CHECKED_MASKS)) {
2603 error = SET_ERROR(EACCES);
2604 }
2605 }
2606 } else if (error == 0) {
2607 error = secpolicy_vnode_access2(cr, ZTOI(zp), owner,
2608 needed_bits, needed_bits);
2609 }
2610
2611 if (is_attr)
2612 iput(ZTOI(xzp));
2613
2614 return (error);
2615 }
2616
2617 /*
2618 * Translate traditional unix S_IRUSR/S_IWUSR/S_IXUSR mode into
2619 * native ACL format and call zfs_zaccess()
2620 */
2621 int
2622 zfs_zaccess_rwx(znode_t *zp, mode_t mode, int flags, cred_t *cr)
2623 {
2624 return (zfs_zaccess(zp, zfs_unix_to_v4(mode >> 6), flags, B_FALSE, cr));
2625 }
2626
2627 /*
2628 * Access function for secpolicy_vnode_setattr
2629 */
2630 int
2631 zfs_zaccess_unix(znode_t *zp, mode_t mode, cred_t *cr)
2632 {
2633 int v4_mode = zfs_unix_to_v4(mode >> 6);
2634
2635 return (zfs_zaccess(zp, v4_mode, 0, B_FALSE, cr));
2636 }
2637
2638 static int
2639 zfs_delete_final_check(znode_t *zp, znode_t *dzp,
2640 mode_t available_perms, cred_t *cr)
2641 {
2642 int error;
2643 uid_t downer;
2644
2645 downer = zfs_fuid_map_id(ZTOZSB(dzp), KUID_TO_SUID(ZTOI(dzp)->i_uid),
2646 cr, ZFS_OWNER);
2647
2648 error = secpolicy_vnode_access2(cr, ZTOI(dzp),
2649 downer, available_perms, S_IWUSR|S_IXUSR);
2650
2651 if (error == 0)
2652 error = zfs_sticky_remove_access(dzp, zp, cr);
2653
2654 return (error);
2655 }
2656
2657 /*
2658 * Determine whether Access should be granted/deny, without
2659 * consulting least priv subsystem.
2660 *
2661 * The following chart is the recommended NFSv4 enforcement for
2662 * ability to delete an object.
2663 *
2664 * -------------------------------------------------------
2665 * | Parent Dir | Target Object Permissions |
2666 * | permissions | |
2667 * -------------------------------------------------------
2668 * | | ACL Allows | ACL Denies| Delete |
2669 * | | Delete | Delete | unspecified|
2670 * -------------------------------------------------------
2671 * | ACL Allows | Permit | Permit | Permit |
2672 * | DELETE_CHILD | |
2673 * -------------------------------------------------------
2674 * | ACL Denies | Permit | Deny | Deny |
2675 * | DELETE_CHILD | | | |
2676 * -------------------------------------------------------
2677 * | ACL specifies | | | |
2678 * | only allow | Permit | Permit | Permit |
2679 * | write and | | | |
2680 * | execute | | | |
2681 * -------------------------------------------------------
2682 * | ACL denies | | | |
2683 * | write and | Permit | Deny | Deny |
2684 * | execute | | | |
2685 * -------------------------------------------------------
2686 * ^
2687 * |
2688 * No search privilege, can't even look up file?
2689 *
2690 */
2691 int
2692 zfs_zaccess_delete(znode_t *dzp, znode_t *zp, cred_t *cr)
2693 {
2694 uint32_t dzp_working_mode = 0;
2695 uint32_t zp_working_mode = 0;
2696 int dzp_error, zp_error;
2697 mode_t available_perms;
2698 boolean_t dzpcheck_privs = B_TRUE;
2699 boolean_t zpcheck_privs = B_TRUE;
2700
2701 /*
2702 * We want specific DELETE permissions to
2703 * take precedence over WRITE/EXECUTE. We don't
2704 * want an ACL such as this to mess us up.
2705 * user:joe:write_data:deny,user:joe:delete:allow
2706 *
2707 * However, deny permissions may ultimately be overridden
2708 * by secpolicy_vnode_access().
2709 *
2710 * We will ask for all of the necessary permissions and then
2711 * look at the working modes from the directory and target object
2712 * to determine what was found.
2713 */
2714
2715 if (zp->z_pflags & (ZFS_IMMUTABLE | ZFS_NOUNLINK))
2716 return (SET_ERROR(EPERM));
2717
2718 /*
2719 * First row
2720 * If the directory permissions allow the delete, we are done.
2721 */
2722 if ((dzp_error = zfs_zaccess_common(dzp, ACE_DELETE_CHILD,
2723 &dzp_working_mode, &dzpcheck_privs, B_FALSE, cr)) == 0)
2724 return (0);
2725
2726 /*
2727 * If target object has delete permission then we are done
2728 */
2729 if ((zp_error = zfs_zaccess_common(zp, ACE_DELETE, &zp_working_mode,
2730 &zpcheck_privs, B_FALSE, cr)) == 0)
2731 return (0);
2732
2733 ASSERT(dzp_error && zp_error);
2734
2735 if (!dzpcheck_privs)
2736 return (dzp_error);
2737 if (!zpcheck_privs)
2738 return (zp_error);
2739
2740 /*
2741 * Second row
2742 *
2743 * If directory returns EACCES then delete_child was denied
2744 * due to deny delete_child. In this case send the request through
2745 * secpolicy_vnode_remove(). We don't use zfs_delete_final_check()
2746 * since that *could* allow the delete based on write/execute permission
2747 * and we want delete permissions to override write/execute.
2748 */
2749
2750 if (dzp_error == EACCES)
2751 return (secpolicy_vnode_remove(cr));
2752
2753 /*
2754 * Third Row
2755 * only need to see if we have write/execute on directory.
2756 */
2757
2758 dzp_error = zfs_zaccess_common(dzp, ACE_EXECUTE|ACE_WRITE_DATA,
2759 &dzp_working_mode, &dzpcheck_privs, B_FALSE, cr);
2760
2761 if (dzp_error != 0 && !dzpcheck_privs)
2762 return (dzp_error);
2763
2764 /*
2765 * Fourth row
2766 */
2767
2768 available_perms = (dzp_working_mode & ACE_WRITE_DATA) ? 0 : S_IWUSR;
2769 available_perms |= (dzp_working_mode & ACE_EXECUTE) ? 0 : S_IXUSR;
2770
2771 return (zfs_delete_final_check(zp, dzp, available_perms, cr));
2772
2773 }
2774
2775 int
2776 zfs_zaccess_rename(znode_t *sdzp, znode_t *szp, znode_t *tdzp,
2777 znode_t *tzp, cred_t *cr)
2778 {
2779 int add_perm;
2780 int error;
2781
2782 if (szp->z_pflags & ZFS_AV_QUARANTINED)
2783 return (SET_ERROR(EACCES));
2784
2785 add_perm = S_ISDIR(ZTOI(szp)->i_mode) ?
2786 ACE_ADD_SUBDIRECTORY : ACE_ADD_FILE;
2787
2788 /*
2789 * Rename permissions are combination of delete permission +
2790 * add file/subdir permission.
2791 */
2792
2793 /*
2794 * first make sure we do the delete portion.
2795 *
2796 * If that succeeds then check for add_file/add_subdir permissions
2797 */
2798
2799 if ((error = zfs_zaccess_delete(sdzp, szp, cr)))
2800 return (error);
2801
2802 /*
2803 * If we have a tzp, see if we can delete it?
2804 */
2805 if (tzp) {
2806 if ((error = zfs_zaccess_delete(tdzp, tzp, cr)))
2807 return (error);
2808 }
2809
2810 /*
2811 * Now check for add permissions
2812 */
2813 error = zfs_zaccess(tdzp, add_perm, 0, B_FALSE, cr);
2814
2815 return (error);
2816 }
ZFS On Linux mirror for PVE