]> 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
Posix ACL Support
[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 */
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
56 #define ALLOW ACE_ACCESS_ALLOWED_ACE_TYPE
57 #define DENY ACE_ACCESS_DENIED_ACE_TYPE
58 #define MAX_ACE_TYPE ACE_SYSTEM_ALARM_CALLBACK_OBJECT_ACE_TYPE
59 #define MIN_ACE_TYPE ALLOW
60
61 #define OWNING_GROUP (ACE_GROUP|ACE_IDENTIFIER_GROUP)
62 #define EVERYONE_ALLOW_MASK (ACE_READ_ACL|ACE_READ_ATTRIBUTES | \
63 ACE_READ_NAMED_ATTRS|ACE_SYNCHRONIZE)
64 #define EVERYONE_DENY_MASK (ACE_WRITE_ACL|ACE_WRITE_OWNER | \
65 ACE_WRITE_ATTRIBUTES|ACE_WRITE_NAMED_ATTRS)
66 #define OWNER_ALLOW_MASK (ACE_WRITE_ACL | ACE_WRITE_OWNER | \
67 ACE_WRITE_ATTRIBUTES|ACE_WRITE_NAMED_ATTRS)
68
69 #define ZFS_CHECKED_MASKS (ACE_READ_ACL|ACE_READ_ATTRIBUTES|ACE_READ_DATA| \
70 ACE_READ_NAMED_ATTRS|ACE_WRITE_DATA|ACE_WRITE_ATTRIBUTES| \
71 ACE_WRITE_NAMED_ATTRS|ACE_APPEND_DATA|ACE_EXECUTE|ACE_WRITE_OWNER| \
72 ACE_WRITE_ACL|ACE_DELETE|ACE_DELETE_CHILD|ACE_SYNCHRONIZE)
73
74 #define WRITE_MASK_DATA (ACE_WRITE_DATA|ACE_APPEND_DATA|ACE_WRITE_NAMED_ATTRS)
75 #define WRITE_MASK_ATTRS (ACE_WRITE_ACL|ACE_WRITE_OWNER|ACE_WRITE_ATTRIBUTES| \
76 ACE_DELETE|ACE_DELETE_CHILD)
77 #define WRITE_MASK (WRITE_MASK_DATA|WRITE_MASK_ATTRS)
78
79 #define OGE_CLEAR (ACE_READ_DATA|ACE_LIST_DIRECTORY|ACE_WRITE_DATA| \
80 ACE_ADD_FILE|ACE_APPEND_DATA|ACE_ADD_SUBDIRECTORY|ACE_EXECUTE)
81
82 #define OKAY_MASK_BITS (ACE_READ_DATA|ACE_LIST_DIRECTORY|ACE_WRITE_DATA| \
83 ACE_ADD_FILE|ACE_APPEND_DATA|ACE_ADD_SUBDIRECTORY|ACE_EXECUTE)
84
85 #define ALL_INHERIT (ACE_FILE_INHERIT_ACE|ACE_DIRECTORY_INHERIT_ACE | \
86 ACE_NO_PROPAGATE_INHERIT_ACE|ACE_INHERIT_ONLY_ACE|ACE_INHERITED_ACE)
87
88 #define RESTRICTED_CLEAR (ACE_WRITE_ACL|ACE_WRITE_OWNER)
89
90 #define V4_ACL_WIDE_FLAGS (ZFS_ACL_AUTO_INHERIT|ZFS_ACL_DEFAULTED|\
91 ZFS_ACL_PROTECTED)
92
93 #define ZFS_ACL_WIDE_FLAGS (V4_ACL_WIDE_FLAGS|ZFS_ACL_TRIVIAL|ZFS_INHERIT_ACE|\
94 ZFS_ACL_OBJ_ACE)
95
96 #define ALL_MODE_EXECS (S_IXUSR | S_IXGRP | S_IXOTH)
97
98 static uint16_t
99 zfs_ace_v0_get_type(void *acep)
100 {
101 return (((zfs_oldace_t *)acep)->z_type);
102 }
103
104 static uint16_t
105 zfs_ace_v0_get_flags(void *acep)
106 {
107 return (((zfs_oldace_t *)acep)->z_flags);
108 }
109
110 static uint32_t
111 zfs_ace_v0_get_mask(void *acep)
112 {
113 return (((zfs_oldace_t *)acep)->z_access_mask);
114 }
115
116 static uint64_t
117 zfs_ace_v0_get_who(void *acep)
118 {
119 return (((zfs_oldace_t *)acep)->z_fuid);
120 }
121
122 static void
123 zfs_ace_v0_set_type(void *acep, uint16_t type)
124 {
125 ((zfs_oldace_t *)acep)->z_type = type;
126 }
127
128 static void
129 zfs_ace_v0_set_flags(void *acep, uint16_t flags)
130 {
131 ((zfs_oldace_t *)acep)->z_flags = flags;
132 }
133
134 static void
135 zfs_ace_v0_set_mask(void *acep, uint32_t mask)
136 {
137 ((zfs_oldace_t *)acep)->z_access_mask = mask;
138 }
139
140 static void
141 zfs_ace_v0_set_who(void *acep, uint64_t who)
142 {
143 ((zfs_oldace_t *)acep)->z_fuid = who;
144 }
145
146 /*ARGSUSED*/
147 static size_t
148 zfs_ace_v0_size(void *acep)
149 {
150 return (sizeof (zfs_oldace_t));
151 }
152
153 static size_t
154 zfs_ace_v0_abstract_size(void)
155 {
156 return (sizeof (zfs_oldace_t));
157 }
158
159 static int
160 zfs_ace_v0_mask_off(void)
161 {
162 return (offsetof(zfs_oldace_t, z_access_mask));
163 }
164
165 /*ARGSUSED*/
166 static int
167 zfs_ace_v0_data(void *acep, void **datap)
168 {
169 *datap = NULL;
170 return (0);
171 }
172
173 static acl_ops_t zfs_acl_v0_ops = {
174 zfs_ace_v0_get_mask,
175 zfs_ace_v0_set_mask,
176 zfs_ace_v0_get_flags,
177 zfs_ace_v0_set_flags,
178 zfs_ace_v0_get_type,
179 zfs_ace_v0_set_type,
180 zfs_ace_v0_get_who,
181 zfs_ace_v0_set_who,
182 zfs_ace_v0_size,
183 zfs_ace_v0_abstract_size,
184 zfs_ace_v0_mask_off,
185 zfs_ace_v0_data
186 };
187
188 static uint16_t
189 zfs_ace_fuid_get_type(void *acep)
190 {
191 return (((zfs_ace_hdr_t *)acep)->z_type);
192 }
193
194 static uint16_t
195 zfs_ace_fuid_get_flags(void *acep)
196 {
197 return (((zfs_ace_hdr_t *)acep)->z_flags);
198 }
199
200 static uint32_t
201 zfs_ace_fuid_get_mask(void *acep)
202 {
203 return (((zfs_ace_hdr_t *)acep)->z_access_mask);
204 }
205
206 static uint64_t
207 zfs_ace_fuid_get_who(void *args)
208 {
209 uint16_t entry_type;
210 zfs_ace_t *acep = args;
211
212 entry_type = acep->z_hdr.z_flags & ACE_TYPE_FLAGS;
213
214 if (entry_type == ACE_OWNER || entry_type == OWNING_GROUP ||
215 entry_type == ACE_EVERYONE)
216 return (-1);
217 return (((zfs_ace_t *)acep)->z_fuid);
218 }
219
220 static void
221 zfs_ace_fuid_set_type(void *acep, uint16_t type)
222 {
223 ((zfs_ace_hdr_t *)acep)->z_type = type;
224 }
225
226 static void
227 zfs_ace_fuid_set_flags(void *acep, uint16_t flags)
228 {
229 ((zfs_ace_hdr_t *)acep)->z_flags = flags;
230 }
231
232 static void
233 zfs_ace_fuid_set_mask(void *acep, uint32_t mask)
234 {
235 ((zfs_ace_hdr_t *)acep)->z_access_mask = mask;
236 }
237
238 static void
239 zfs_ace_fuid_set_who(void *arg, uint64_t who)
240 {
241 zfs_ace_t *acep = arg;
242
243 uint16_t entry_type = acep->z_hdr.z_flags & ACE_TYPE_FLAGS;
244
245 if (entry_type == ACE_OWNER || entry_type == OWNING_GROUP ||
246 entry_type == ACE_EVERYONE)
247 return;
248 acep->z_fuid = who;
249 }
250
251 static size_t
252 zfs_ace_fuid_size(void *acep)
253 {
254 zfs_ace_hdr_t *zacep = acep;
255 uint16_t entry_type;
256
257 switch (zacep->z_type) {
258 case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE:
259 case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE:
260 case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE:
261 case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE:
262 return (sizeof (zfs_object_ace_t));
263 case ALLOW:
264 case DENY:
265 entry_type =
266 (((zfs_ace_hdr_t *)acep)->z_flags & ACE_TYPE_FLAGS);
267 if (entry_type == ACE_OWNER ||
268 entry_type == OWNING_GROUP ||
269 entry_type == ACE_EVERYONE)
270 return (sizeof (zfs_ace_hdr_t));
271 /*FALLTHROUGH*/
272 default:
273 return (sizeof (zfs_ace_t));
274 }
275 }
276
277 static size_t
278 zfs_ace_fuid_abstract_size(void)
279 {
280 return (sizeof (zfs_ace_hdr_t));
281 }
282
283 static int
284 zfs_ace_fuid_mask_off(void)
285 {
286 return (offsetof(zfs_ace_hdr_t, z_access_mask));
287 }
288
289 static int
290 zfs_ace_fuid_data(void *acep, void **datap)
291 {
292 zfs_ace_t *zacep = acep;
293 zfs_object_ace_t *zobjp;
294
295 switch (zacep->z_hdr.z_type) {
296 case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE:
297 case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE:
298 case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE:
299 case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE:
300 zobjp = acep;
301 *datap = (caddr_t)zobjp + sizeof (zfs_ace_t);
302 return (sizeof (zfs_object_ace_t) - sizeof (zfs_ace_t));
303 default:
304 *datap = NULL;
305 return (0);
306 }
307 }
308
309 static acl_ops_t zfs_acl_fuid_ops = {
310 zfs_ace_fuid_get_mask,
311 zfs_ace_fuid_set_mask,
312 zfs_ace_fuid_get_flags,
313 zfs_ace_fuid_set_flags,
314 zfs_ace_fuid_get_type,
315 zfs_ace_fuid_set_type,
316 zfs_ace_fuid_get_who,
317 zfs_ace_fuid_set_who,
318 zfs_ace_fuid_size,
319 zfs_ace_fuid_abstract_size,
320 zfs_ace_fuid_mask_off,
321 zfs_ace_fuid_data
322 };
323
324 /*
325 * The following three functions are provided for compatibility with
326 * older ZPL version in order to determine if the file use to have
327 * an external ACL and what version of ACL previously existed on the
328 * file. Would really be nice to not need this, sigh.
329 */
330 uint64_t
331 zfs_external_acl(znode_t *zp)
332 {
333 zfs_acl_phys_t acl_phys;
334 int error;
335
336 if (zp->z_is_sa)
337 return (0);
338
339 /*
340 * Need to deal with a potential
341 * race where zfs_sa_upgrade could cause
342 * z_isa_sa to change.
343 *
344 * If the lookup fails then the state of z_is_sa should have
345 * changed.
346 */
347
348 if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_ZNODE_ACL(ZTOZSB(zp)),
349 &acl_phys, sizeof (acl_phys))) == 0)
350 return (acl_phys.z_acl_extern_obj);
351 else {
352 /*
353 * after upgrade the SA_ZPL_ZNODE_ACL should have been
354 * removed
355 */
356 VERIFY(zp->z_is_sa && error == ENOENT);
357 return (0);
358 }
359 }
360
361 /*
362 * Determine size of ACL in bytes
363 *
364 * This is more complicated than it should be since we have to deal
365 * with old external ACLs.
366 */
367 static int
368 zfs_acl_znode_info(znode_t *zp, int *aclsize, int *aclcount,
369 zfs_acl_phys_t *aclphys)
370 {
371 zfs_sb_t *zsb = ZTOZSB(zp);
372 uint64_t acl_count;
373 int size;
374 int error;
375
376 ASSERT(MUTEX_HELD(&zp->z_acl_lock));
377 if (zp->z_is_sa) {
378 if ((error = sa_size(zp->z_sa_hdl, SA_ZPL_DACL_ACES(zsb),
379 &size)) != 0)
380 return (error);
381 *aclsize = size;
382 if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_DACL_COUNT(zsb),
383 &acl_count, sizeof (acl_count))) != 0)
384 return (error);
385 *aclcount = acl_count;
386 } else {
387 if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_ZNODE_ACL(zsb),
388 aclphys, sizeof (*aclphys))) != 0)
389 return (error);
390
391 if (aclphys->z_acl_version == ZFS_ACL_VERSION_INITIAL) {
392 *aclsize = ZFS_ACL_SIZE(aclphys->z_acl_size);
393 *aclcount = aclphys->z_acl_size;
394 } else {
395 *aclsize = aclphys->z_acl_size;
396 *aclcount = aclphys->z_acl_count;
397 }
398 }
399 return (0);
400 }
401
402 int
403 zfs_znode_acl_version(znode_t *zp)
404 {
405 zfs_acl_phys_t acl_phys;
406
407 if (zp->z_is_sa)
408 return (ZFS_ACL_VERSION_FUID);
409 else {
410 int error;
411
412 /*
413 * Need to deal with a potential
414 * race where zfs_sa_upgrade could cause
415 * z_isa_sa to change.
416 *
417 * If the lookup fails then the state of z_is_sa should have
418 * changed.
419 */
420 if ((error = sa_lookup(zp->z_sa_hdl,
421 SA_ZPL_ZNODE_ACL(ZTOZSB(zp)),
422 &acl_phys, sizeof (acl_phys))) == 0)
423 return (acl_phys.z_acl_version);
424 else {
425 /*
426 * After upgrade SA_ZPL_ZNODE_ACL should have
427 * been removed.
428 */
429 VERIFY(zp->z_is_sa && error == ENOENT);
430 return (ZFS_ACL_VERSION_FUID);
431 }
432 }
433 }
434
435 static int
436 zfs_acl_version(int version)
437 {
438 if (version < ZPL_VERSION_FUID)
439 return (ZFS_ACL_VERSION_INITIAL);
440 else
441 return (ZFS_ACL_VERSION_FUID);
442 }
443
444 static int
445 zfs_acl_version_zp(znode_t *zp)
446 {
447 return (zfs_acl_version(ZTOZSB(zp)->z_version));
448 }
449
450 zfs_acl_t *
451 zfs_acl_alloc(int vers)
452 {
453 zfs_acl_t *aclp;
454
455 aclp = kmem_zalloc(sizeof (zfs_acl_t), KM_PUSHPAGE);
456 list_create(&aclp->z_acl, sizeof (zfs_acl_node_t),
457 offsetof(zfs_acl_node_t, z_next));
458 aclp->z_version = vers;
459 if (vers == ZFS_ACL_VERSION_FUID)
460 aclp->z_ops = &zfs_acl_fuid_ops;
461 else
462 aclp->z_ops = &zfs_acl_v0_ops;
463 return (aclp);
464 }
465
466 zfs_acl_node_t *
467 zfs_acl_node_alloc(size_t bytes)
468 {
469 zfs_acl_node_t *aclnode;
470
471 aclnode = kmem_zalloc(sizeof (zfs_acl_node_t), KM_PUSHPAGE);
472 if (bytes) {
473 aclnode->z_acldata = kmem_alloc(bytes, KM_PUSHPAGE);
474 aclnode->z_allocdata = aclnode->z_acldata;
475 aclnode->z_allocsize = bytes;
476 aclnode->z_size = bytes;
477 }
478
479 return (aclnode);
480 }
481
482 static void
483 zfs_acl_node_free(zfs_acl_node_t *aclnode)
484 {
485 if (aclnode->z_allocsize)
486 kmem_free(aclnode->z_allocdata, aclnode->z_allocsize);
487 kmem_free(aclnode, sizeof (zfs_acl_node_t));
488 }
489
490 static void
491 zfs_acl_release_nodes(zfs_acl_t *aclp)
492 {
493 zfs_acl_node_t *aclnode;
494
495 while ((aclnode = list_head(&aclp->z_acl))) {
496 list_remove(&aclp->z_acl, aclnode);
497 zfs_acl_node_free(aclnode);
498 }
499 aclp->z_acl_count = 0;
500 aclp->z_acl_bytes = 0;
501 }
502
503 void
504 zfs_acl_free(zfs_acl_t *aclp)
505 {
506 zfs_acl_release_nodes(aclp);
507 list_destroy(&aclp->z_acl);
508 kmem_free(aclp, sizeof (zfs_acl_t));
509 }
510
511 static boolean_t
512 zfs_acl_valid_ace_type(uint_t type, uint_t flags)
513 {
514 uint16_t entry_type;
515
516 switch (type) {
517 case ALLOW:
518 case DENY:
519 case ACE_SYSTEM_AUDIT_ACE_TYPE:
520 case ACE_SYSTEM_ALARM_ACE_TYPE:
521 entry_type = flags & ACE_TYPE_FLAGS;
522 return (entry_type == ACE_OWNER ||
523 entry_type == OWNING_GROUP ||
524 entry_type == ACE_EVERYONE || entry_type == 0 ||
525 entry_type == ACE_IDENTIFIER_GROUP);
526 default:
527 if (type >= MIN_ACE_TYPE && type <= MAX_ACE_TYPE)
528 return (B_TRUE);
529 }
530 return (B_FALSE);
531 }
532
533 static boolean_t
534 zfs_ace_valid(umode_t obj_mode, zfs_acl_t *aclp, uint16_t type, uint16_t iflags)
535 {
536 /*
537 * first check type of entry
538 */
539
540 if (!zfs_acl_valid_ace_type(type, iflags))
541 return (B_FALSE);
542
543 switch (type) {
544 case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE:
545 case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE:
546 case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE:
547 case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE:
548 if (aclp->z_version < ZFS_ACL_VERSION_FUID)
549 return (B_FALSE);
550 aclp->z_hints |= ZFS_ACL_OBJ_ACE;
551 }
552
553 /*
554 * next check inheritance level flags
555 */
556
557 if (S_ISDIR(obj_mode) &&
558 (iflags & (ACE_FILE_INHERIT_ACE|ACE_DIRECTORY_INHERIT_ACE)))
559 aclp->z_hints |= ZFS_INHERIT_ACE;
560
561 if (iflags & (ACE_INHERIT_ONLY_ACE|ACE_NO_PROPAGATE_INHERIT_ACE)) {
562 if ((iflags & (ACE_FILE_INHERIT_ACE|
563 ACE_DIRECTORY_INHERIT_ACE)) == 0) {
564 return (B_FALSE);
565 }
566 }
567
568 return (B_TRUE);
569 }
570
571 static void *
572 zfs_acl_next_ace(zfs_acl_t *aclp, void *start, uint64_t *who,
573 uint32_t *access_mask, uint16_t *iflags, uint16_t *type)
574 {
575 zfs_acl_node_t *aclnode;
576
577 ASSERT(aclp);
578
579 if (start == NULL) {
580 aclnode = list_head(&aclp->z_acl);
581 if (aclnode == NULL)
582 return (NULL);
583
584 aclp->z_next_ace = aclnode->z_acldata;
585 aclp->z_curr_node = aclnode;
586 aclnode->z_ace_idx = 0;
587 }
588
589 aclnode = aclp->z_curr_node;
590
591 if (aclnode == NULL)
592 return (NULL);
593
594 if (aclnode->z_ace_idx >= aclnode->z_ace_count) {
595 aclnode = list_next(&aclp->z_acl, aclnode);
596 if (aclnode == NULL)
597 return (NULL);
598 else {
599 aclp->z_curr_node = aclnode;
600 aclnode->z_ace_idx = 0;
601 aclp->z_next_ace = aclnode->z_acldata;
602 }
603 }
604
605 if (aclnode->z_ace_idx < aclnode->z_ace_count) {
606 void *acep = aclp->z_next_ace;
607 size_t ace_size;
608
609 /*
610 * Make sure we don't overstep our bounds
611 */
612 ace_size = aclp->z_ops->ace_size(acep);
613
614 if (((caddr_t)acep + ace_size) >
615 ((caddr_t)aclnode->z_acldata + aclnode->z_size)) {
616 return (NULL);
617 }
618
619 *iflags = aclp->z_ops->ace_flags_get(acep);
620 *type = aclp->z_ops->ace_type_get(acep);
621 *access_mask = aclp->z_ops->ace_mask_get(acep);
622 *who = aclp->z_ops->ace_who_get(acep);
623 aclp->z_next_ace = (caddr_t)aclp->z_next_ace + ace_size;
624 aclnode->z_ace_idx++;
625
626 return ((void *)acep);
627 }
628 return (NULL);
629 }
630
631 /*ARGSUSED*/
632 static uint64_t
633 zfs_ace_walk(void *datap, uint64_t cookie, int aclcnt,
634 uint16_t *flags, uint16_t *type, uint32_t *mask)
635 {
636 zfs_acl_t *aclp = datap;
637 zfs_ace_hdr_t *acep = (zfs_ace_hdr_t *)(uintptr_t)cookie;
638 uint64_t who;
639
640 acep = zfs_acl_next_ace(aclp, acep, &who, mask,
641 flags, type);
642 return ((uint64_t)(uintptr_t)acep);
643 }
644
645 /*
646 * Copy ACE to internal ZFS format.
647 * While processing the ACL each ACE will be validated for correctness.
648 * ACE FUIDs will be created later.
649 */
650 int
651 zfs_copy_ace_2_fuid(zfs_sb_t *zsb, umode_t obj_mode, zfs_acl_t *aclp,
652 void *datap, zfs_ace_t *z_acl, uint64_t aclcnt, size_t *size,
653 zfs_fuid_info_t **fuidp, cred_t *cr)
654 {
655 int i;
656 uint16_t entry_type;
657 zfs_ace_t *aceptr = z_acl;
658 ace_t *acep = datap;
659 zfs_object_ace_t *zobjacep;
660 ace_object_t *aceobjp;
661
662 for (i = 0; i != aclcnt; i++) {
663 aceptr->z_hdr.z_access_mask = acep->a_access_mask;
664 aceptr->z_hdr.z_flags = acep->a_flags;
665 aceptr->z_hdr.z_type = acep->a_type;
666 entry_type = aceptr->z_hdr.z_flags & ACE_TYPE_FLAGS;
667 if (entry_type != ACE_OWNER && entry_type != OWNING_GROUP &&
668 entry_type != ACE_EVERYONE) {
669 aceptr->z_fuid = zfs_fuid_create(zsb, acep->a_who,
670 cr, (entry_type == 0) ?
671 ZFS_ACE_USER : ZFS_ACE_GROUP, fuidp);
672 }
673
674 /*
675 * Make sure ACE is valid
676 */
677 if (zfs_ace_valid(obj_mode, aclp, aceptr->z_hdr.z_type,
678 aceptr->z_hdr.z_flags) != B_TRUE)
679 return (EINVAL);
680
681 switch (acep->a_type) {
682 case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE:
683 case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE:
684 case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE:
685 case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE:
686 zobjacep = (zfs_object_ace_t *)aceptr;
687 aceobjp = (ace_object_t *)acep;
688
689 bcopy(aceobjp->a_obj_type, zobjacep->z_object_type,
690 sizeof (aceobjp->a_obj_type));
691 bcopy(aceobjp->a_inherit_obj_type,
692 zobjacep->z_inherit_type,
693 sizeof (aceobjp->a_inherit_obj_type));
694 acep = (ace_t *)((caddr_t)acep + sizeof (ace_object_t));
695 break;
696 default:
697 acep = (ace_t *)((caddr_t)acep + sizeof (ace_t));
698 }
699
700 aceptr = (zfs_ace_t *)((caddr_t)aceptr +
701 aclp->z_ops->ace_size(aceptr));
702 }
703
704 *size = (caddr_t)aceptr - (caddr_t)z_acl;
705
706 return (0);
707 }
708
709 /*
710 * Copy ZFS ACEs to fixed size ace_t layout
711 */
712 static void
713 zfs_copy_fuid_2_ace(zfs_sb_t *zsb, zfs_acl_t *aclp, cred_t *cr,
714 void *datap, int filter)
715 {
716 uint64_t who;
717 uint32_t access_mask;
718 uint16_t iflags, type;
719 zfs_ace_hdr_t *zacep = NULL;
720 ace_t *acep = datap;
721 ace_object_t *objacep;
722 zfs_object_ace_t *zobjacep;
723 size_t ace_size;
724 uint16_t entry_type;
725
726 while ((zacep = zfs_acl_next_ace(aclp, zacep,
727 &who, &access_mask, &iflags, &type))) {
728
729 switch (type) {
730 case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE:
731 case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE:
732 case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE:
733 case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE:
734 if (filter) {
735 continue;
736 }
737 zobjacep = (zfs_object_ace_t *)zacep;
738 objacep = (ace_object_t *)acep;
739 bcopy(zobjacep->z_object_type,
740 objacep->a_obj_type,
741 sizeof (zobjacep->z_object_type));
742 bcopy(zobjacep->z_inherit_type,
743 objacep->a_inherit_obj_type,
744 sizeof (zobjacep->z_inherit_type));
745 ace_size = sizeof (ace_object_t);
746 break;
747 default:
748 ace_size = sizeof (ace_t);
749 break;
750 }
751
752 entry_type = (iflags & ACE_TYPE_FLAGS);
753 if ((entry_type != ACE_OWNER &&
754 entry_type != OWNING_GROUP &&
755 entry_type != ACE_EVERYONE)) {
756 acep->a_who = zfs_fuid_map_id(zsb, who,
757 cr, (entry_type & ACE_IDENTIFIER_GROUP) ?
758 ZFS_ACE_GROUP : ZFS_ACE_USER);
759 } else {
760 acep->a_who = (uid_t)(int64_t)who;
761 }
762 acep->a_access_mask = access_mask;
763 acep->a_flags = iflags;
764 acep->a_type = type;
765 acep = (ace_t *)((caddr_t)acep + ace_size);
766 }
767 }
768
769 static int
770 zfs_copy_ace_2_oldace(umode_t obj_mode, zfs_acl_t *aclp, ace_t *acep,
771 zfs_oldace_t *z_acl, int aclcnt, size_t *size)
772 {
773 int i;
774 zfs_oldace_t *aceptr = z_acl;
775
776 for (i = 0; i != aclcnt; i++, aceptr++) {
777 aceptr->z_access_mask = acep[i].a_access_mask;
778 aceptr->z_type = acep[i].a_type;
779 aceptr->z_flags = acep[i].a_flags;
780 aceptr->z_fuid = acep[i].a_who;
781 /*
782 * Make sure ACE is valid
783 */
784 if (zfs_ace_valid(obj_mode, aclp, aceptr->z_type,
785 aceptr->z_flags) != B_TRUE)
786 return (EINVAL);
787 }
788 *size = (caddr_t)aceptr - (caddr_t)z_acl;
789 return (0);
790 }
791
792 /*
793 * convert old ACL format to new
794 */
795 void
796 zfs_acl_xform(znode_t *zp, zfs_acl_t *aclp, cred_t *cr)
797 {
798 zfs_oldace_t *oldaclp;
799 int i;
800 uint16_t type, iflags;
801 uint32_t access_mask;
802 uint64_t who;
803 void *cookie = NULL;
804 zfs_acl_node_t *newaclnode;
805
806 ASSERT(aclp->z_version == ZFS_ACL_VERSION_INITIAL);
807 /*
808 * First create the ACE in a contiguous piece of memory
809 * for zfs_copy_ace_2_fuid().
810 *
811 * We only convert an ACL once, so this won't happen
812 * everytime.
813 */
814 oldaclp = kmem_alloc(sizeof (zfs_oldace_t) * aclp->z_acl_count,
815 KM_SLEEP);
816 i = 0;
817 while ((cookie = zfs_acl_next_ace(aclp, cookie, &who,
818 &access_mask, &iflags, &type))) {
819 oldaclp[i].z_flags = iflags;
820 oldaclp[i].z_type = type;
821 oldaclp[i].z_fuid = who;
822 oldaclp[i++].z_access_mask = access_mask;
823 }
824
825 newaclnode = zfs_acl_node_alloc(aclp->z_acl_count *
826 sizeof (zfs_object_ace_t));
827 aclp->z_ops = &zfs_acl_fuid_ops;
828 VERIFY(zfs_copy_ace_2_fuid(ZTOZSB(zp), ZTOI(zp)->i_mode,
829 aclp, oldaclp, newaclnode->z_acldata, aclp->z_acl_count,
830 &newaclnode->z_size, NULL, cr) == 0);
831 newaclnode->z_ace_count = aclp->z_acl_count;
832 aclp->z_version = ZFS_ACL_VERSION;
833 kmem_free(oldaclp, aclp->z_acl_count * sizeof (zfs_oldace_t));
834
835 /*
836 * Release all previous ACL nodes
837 */
838
839 zfs_acl_release_nodes(aclp);
840
841 list_insert_head(&aclp->z_acl, newaclnode);
842
843 aclp->z_acl_bytes = newaclnode->z_size;
844 aclp->z_acl_count = newaclnode->z_ace_count;
845
846 }
847
848 /*
849 * Convert unix access mask to v4 access mask
850 */
851 static uint32_t
852 zfs_unix_to_v4(uint32_t access_mask)
853 {
854 uint32_t new_mask = 0;
855
856 if (access_mask & S_IXOTH)
857 new_mask |= ACE_EXECUTE;
858 if (access_mask & S_IWOTH)
859 new_mask |= ACE_WRITE_DATA;
860 if (access_mask & S_IROTH)
861 new_mask |= ACE_READ_DATA;
862 return (new_mask);
863 }
864
865 static void
866 zfs_set_ace(zfs_acl_t *aclp, void *acep, uint32_t access_mask,
867 uint16_t access_type, uint64_t fuid, uint16_t entry_type)
868 {
869 uint16_t type = entry_type & ACE_TYPE_FLAGS;
870
871 aclp->z_ops->ace_mask_set(acep, access_mask);
872 aclp->z_ops->ace_type_set(acep, access_type);
873 aclp->z_ops->ace_flags_set(acep, entry_type);
874 if ((type != ACE_OWNER && type != OWNING_GROUP &&
875 type != ACE_EVERYONE))
876 aclp->z_ops->ace_who_set(acep, fuid);
877 }
878
879 /*
880 * Determine mode of file based on ACL.
881 * Also, create FUIDs for any User/Group ACEs
882 */
883 uint64_t
884 zfs_mode_compute(uint64_t fmode, zfs_acl_t *aclp,
885 uint64_t *pflags, uint64_t fuid, uint64_t fgid)
886 {
887 int entry_type;
888 mode_t mode;
889 mode_t seen = 0;
890 zfs_ace_hdr_t *acep = NULL;
891 uint64_t who;
892 uint16_t iflags, type;
893 uint32_t access_mask;
894 boolean_t an_exec_denied = B_FALSE;
895
896 mode = (fmode & (S_IFMT | S_ISUID | S_ISGID | S_ISVTX));
897
898 while ((acep = zfs_acl_next_ace(aclp, acep, &who,
899 &access_mask, &iflags, &type))) {
900
901 if (!zfs_acl_valid_ace_type(type, iflags))
902 continue;
903
904 entry_type = (iflags & ACE_TYPE_FLAGS);
905
906 /*
907 * Skip over owner@, group@ or everyone@ inherit only ACEs
908 */
909 if ((iflags & ACE_INHERIT_ONLY_ACE) &&
910 (entry_type == ACE_OWNER || entry_type == ACE_EVERYONE ||
911 entry_type == OWNING_GROUP))
912 continue;
913
914 if (entry_type == ACE_OWNER || (entry_type == 0 &&
915 who == fuid)) {
916 if ((access_mask & ACE_READ_DATA) &&
917 (!(seen & S_IRUSR))) {
918 seen |= S_IRUSR;
919 if (type == ALLOW) {
920 mode |= S_IRUSR;
921 }
922 }
923 if ((access_mask & ACE_WRITE_DATA) &&
924 (!(seen & S_IWUSR))) {
925 seen |= S_IWUSR;
926 if (type == ALLOW) {
927 mode |= S_IWUSR;
928 }
929 }
930 if ((access_mask & ACE_EXECUTE) &&
931 (!(seen & S_IXUSR))) {
932 seen |= S_IXUSR;
933 if (type == ALLOW) {
934 mode |= S_IXUSR;
935 }
936 }
937 } else if (entry_type == OWNING_GROUP ||
938 (entry_type == ACE_IDENTIFIER_GROUP && who == fgid)) {
939 if ((access_mask & ACE_READ_DATA) &&
940 (!(seen & S_IRGRP))) {
941 seen |= S_IRGRP;
942 if (type == ALLOW) {
943 mode |= S_IRGRP;
944 }
945 }
946 if ((access_mask & ACE_WRITE_DATA) &&
947 (!(seen & S_IWGRP))) {
948 seen |= S_IWGRP;
949 if (type == ALLOW) {
950 mode |= S_IWGRP;
951 }
952 }
953 if ((access_mask & ACE_EXECUTE) &&
954 (!(seen & S_IXGRP))) {
955 seen |= S_IXGRP;
956 if (type == ALLOW) {
957 mode |= S_IXGRP;
958 }
959 }
960 } else if (entry_type == ACE_EVERYONE) {
961 if ((access_mask & ACE_READ_DATA)) {
962 if (!(seen & S_IRUSR)) {
963 seen |= S_IRUSR;
964 if (type == ALLOW) {
965 mode |= S_IRUSR;
966 }
967 }
968 if (!(seen & S_IRGRP)) {
969 seen |= S_IRGRP;
970 if (type == ALLOW) {
971 mode |= S_IRGRP;
972 }
973 }
974 if (!(seen & S_IROTH)) {
975 seen |= S_IROTH;
976 if (type == ALLOW) {
977 mode |= S_IROTH;
978 }
979 }
980 }
981 if ((access_mask & ACE_WRITE_DATA)) {
982 if (!(seen & S_IWUSR)) {
983 seen |= S_IWUSR;
984 if (type == ALLOW) {
985 mode |= S_IWUSR;
986 }
987 }
988 if (!(seen & S_IWGRP)) {
989 seen |= S_IWGRP;
990 if (type == ALLOW) {
991 mode |= S_IWGRP;
992 }
993 }
994 if (!(seen & S_IWOTH)) {
995 seen |= S_IWOTH;
996 if (type == ALLOW) {
997 mode |= S_IWOTH;
998 }
999 }
1000 }
1001 if ((access_mask & ACE_EXECUTE)) {
1002 if (!(seen & S_IXUSR)) {
1003 seen |= S_IXUSR;
1004 if (type == ALLOW) {
1005 mode |= S_IXUSR;
1006 }
1007 }
1008 if (!(seen & S_IXGRP)) {
1009 seen |= S_IXGRP;
1010 if (type == ALLOW) {
1011 mode |= S_IXGRP;
1012 }
1013 }
1014 if (!(seen & S_IXOTH)) {
1015 seen |= S_IXOTH;
1016 if (type == ALLOW) {
1017 mode |= S_IXOTH;
1018 }
1019 }
1020 }
1021 } else {
1022 /*
1023 * Only care if this IDENTIFIER_GROUP or
1024 * USER ACE denies execute access to someone,
1025 * mode is not affected
1026 */
1027 if ((access_mask & ACE_EXECUTE) && type == DENY)
1028 an_exec_denied = B_TRUE;
1029 }
1030 }
1031
1032 /*
1033 * Failure to allow is effectively a deny, so execute permission
1034 * is denied if it was never mentioned or if we explicitly
1035 * weren't allowed it.
1036 */
1037 if (!an_exec_denied &&
1038 ((seen & ALL_MODE_EXECS) != ALL_MODE_EXECS ||
1039 (mode & ALL_MODE_EXECS) != ALL_MODE_EXECS))
1040 an_exec_denied = B_TRUE;
1041
1042 if (an_exec_denied)
1043 *pflags &= ~ZFS_NO_EXECS_DENIED;
1044 else
1045 *pflags |= ZFS_NO_EXECS_DENIED;
1046
1047 return (mode);
1048 }
1049
1050 /*
1051 * Read an external acl object. If the intent is to modify, always
1052 * create a new acl and leave any cached acl in place.
1053 */
1054 static int
1055 zfs_acl_node_read(znode_t *zp, boolean_t have_lock, zfs_acl_t **aclpp,
1056 boolean_t will_modify)
1057 {
1058 zfs_acl_t *aclp;
1059 int aclsize = 0;
1060 int acl_count = 0;
1061 zfs_acl_node_t *aclnode;
1062 zfs_acl_phys_t znode_acl;
1063 int version;
1064 int error;
1065 boolean_t drop_lock = B_FALSE;
1066
1067 ASSERT(MUTEX_HELD(&zp->z_acl_lock));
1068
1069 if (zp->z_acl_cached && !will_modify) {
1070 *aclpp = zp->z_acl_cached;
1071 return (0);
1072 }
1073
1074 /*
1075 * close race where znode could be upgrade while trying to
1076 * read the znode attributes.
1077 *
1078 * But this could only happen if the file isn't already an SA
1079 * znode
1080 */
1081 if (!zp->z_is_sa && !have_lock) {
1082 mutex_enter(&zp->z_lock);
1083 drop_lock = B_TRUE;
1084 }
1085 version = zfs_znode_acl_version(zp);
1086
1087 if ((error = zfs_acl_znode_info(zp, &aclsize,
1088 &acl_count, &znode_acl)) != 0) {
1089 goto done;
1090 }
1091
1092 aclp = zfs_acl_alloc(version);
1093
1094 aclp->z_acl_count = acl_count;
1095 aclp->z_acl_bytes = aclsize;
1096
1097 aclnode = zfs_acl_node_alloc(aclsize);
1098 aclnode->z_ace_count = aclp->z_acl_count;
1099 aclnode->z_size = aclsize;
1100
1101 if (!zp->z_is_sa) {
1102 if (znode_acl.z_acl_extern_obj) {
1103 error = dmu_read(ZTOZSB(zp)->z_os,
1104 znode_acl.z_acl_extern_obj, 0, aclnode->z_size,
1105 aclnode->z_acldata, DMU_READ_PREFETCH);
1106 } else {
1107 bcopy(znode_acl.z_ace_data, aclnode->z_acldata,
1108 aclnode->z_size);
1109 }
1110 } else {
1111 error = sa_lookup(zp->z_sa_hdl, SA_ZPL_DACL_ACES(ZTOZSB(zp)),
1112 aclnode->z_acldata, aclnode->z_size);
1113 }
1114
1115 if (error != 0) {
1116 zfs_acl_free(aclp);
1117 zfs_acl_node_free(aclnode);
1118 /* convert checksum errors into IO errors */
1119 if (error == ECKSUM)
1120 error = EIO;
1121 goto done;
1122 }
1123
1124 list_insert_head(&aclp->z_acl, aclnode);
1125
1126 *aclpp = aclp;
1127 if (!will_modify)
1128 zp->z_acl_cached = aclp;
1129 done:
1130 if (drop_lock)
1131 mutex_exit(&zp->z_lock);
1132 return (error);
1133 }
1134
1135 /*ARGSUSED*/
1136 void
1137 zfs_acl_data_locator(void **dataptr, uint32_t *length, uint32_t buflen,
1138 boolean_t start, void *userdata)
1139 {
1140 zfs_acl_locator_cb_t *cb = (zfs_acl_locator_cb_t *)userdata;
1141
1142 if (start) {
1143 cb->cb_acl_node = list_head(&cb->cb_aclp->z_acl);
1144 } else {
1145 cb->cb_acl_node = list_next(&cb->cb_aclp->z_acl,
1146 cb->cb_acl_node);
1147 }
1148 *dataptr = cb->cb_acl_node->z_acldata;
1149 *length = cb->cb_acl_node->z_size;
1150 }
1151
1152 int
1153 zfs_acl_chown_setattr(znode_t *zp)
1154 {
1155 int error;
1156 zfs_acl_t *aclp;
1157
1158 if (ZTOZSB(zp)->z_acl_type == ZFS_ACLTYPE_POSIXACL)
1159 return 0;
1160
1161 ASSERT(MUTEX_HELD(&zp->z_lock));
1162 ASSERT(MUTEX_HELD(&zp->z_acl_lock));
1163
1164 if ((error = zfs_acl_node_read(zp, B_TRUE, &aclp, B_FALSE)) == 0)
1165 zp->z_mode = zfs_mode_compute(zp->z_mode, aclp,
1166 &zp->z_pflags, zp->z_uid, zp->z_gid);
1167 return (error);
1168 }
1169
1170 static void
1171 acl_trivial_access_masks(mode_t mode, uint32_t *allow0, uint32_t *deny1,
1172 uint32_t *deny2, uint32_t *owner, uint32_t *group, uint32_t *everyone)
1173 {
1174 *deny1 = *deny2 = *allow0 = *group = 0;
1175
1176 if (!(mode & S_IRUSR) && (mode & (S_IRGRP|S_IROTH)))
1177 *deny1 |= ACE_READ_DATA;
1178 if (!(mode & S_IWUSR) && (mode & (S_IWGRP|S_IWOTH)))
1179 *deny1 |= ACE_WRITE_DATA;
1180 if (!(mode & S_IXUSR) && (mode & (S_IXGRP|S_IXOTH)))
1181 *deny1 |= ACE_EXECUTE;
1182
1183 if (!(mode & S_IRGRP) && (mode & S_IROTH))
1184 *deny2 = ACE_READ_DATA;
1185 if (!(mode & S_IWGRP) && (mode & S_IWOTH))
1186 *deny2 |= ACE_WRITE_DATA;
1187 if (!(mode & S_IXGRP) && (mode & S_IXOTH))
1188 *deny2 |= ACE_EXECUTE;
1189
1190 if ((mode & S_IRUSR) && (!(mode & S_IRGRP) && (mode & S_IROTH)))
1191 *allow0 |= ACE_READ_DATA;
1192 if ((mode & S_IWUSR) && (!(mode & S_IWGRP) && (mode & S_IWOTH)))
1193 *allow0 |= ACE_WRITE_DATA;
1194 if ((mode & S_IXUSR) && (!(mode & S_IXGRP) && (mode & S_IXOTH)))
1195 *allow0 |= ACE_EXECUTE;
1196
1197 *owner = ACE_WRITE_ATTRIBUTES|ACE_WRITE_OWNER|ACE_WRITE_ACL|
1198 ACE_WRITE_NAMED_ATTRS|ACE_READ_ACL|ACE_READ_ATTRIBUTES|
1199 ACE_READ_NAMED_ATTRS|ACE_SYNCHRONIZE;
1200 if (mode & S_IRUSR)
1201 *owner |= ACE_READ_DATA;
1202 if (mode & S_IWUSR)
1203 *owner |= ACE_WRITE_DATA|ACE_APPEND_DATA;
1204 if (mode & S_IXUSR)
1205 *owner |= ACE_EXECUTE;
1206
1207 *group = ACE_READ_ACL|ACE_READ_ATTRIBUTES| ACE_READ_NAMED_ATTRS|
1208 ACE_SYNCHRONIZE;
1209 if (mode & S_IRGRP)
1210 *group |= ACE_READ_DATA;
1211 if (mode & S_IWGRP)
1212 *group |= ACE_WRITE_DATA|ACE_APPEND_DATA;
1213 if (mode & S_IXGRP)
1214 *group |= ACE_EXECUTE;
1215
1216 *everyone = ACE_READ_ACL|ACE_READ_ATTRIBUTES| ACE_READ_NAMED_ATTRS|
1217 ACE_SYNCHRONIZE;
1218 if (mode & S_IROTH)
1219 *everyone |= ACE_READ_DATA;
1220 if (mode & S_IWOTH)
1221 *everyone |= ACE_WRITE_DATA|ACE_APPEND_DATA;
1222 if (mode & S_IXOTH)
1223 *everyone |= ACE_EXECUTE;
1224 }
1225
1226 /*
1227 * ace_trivial:
1228 * determine whether an ace_t acl is trivial
1229 *
1230 * Trivialness implies that the acl is composed of only
1231 * owner, group, everyone entries. ACL can't
1232 * have read_acl denied, and write_owner/write_acl/write_attributes
1233 * can only be owner@ entry.
1234 */
1235 static int
1236 ace_trivial_common(void *acep, int aclcnt,
1237 uint64_t (*walk)(void *, uint64_t, int aclcnt,
1238 uint16_t *, uint16_t *, uint32_t *))
1239 {
1240 uint16_t flags;
1241 uint32_t mask;
1242 uint16_t type;
1243 uint64_t cookie = 0;
1244
1245 while ((cookie = walk(acep, cookie, aclcnt, &flags, &type, &mask))) {
1246 switch (flags & ACE_TYPE_FLAGS) {
1247 case ACE_OWNER:
1248 case ACE_GROUP|ACE_IDENTIFIER_GROUP:
1249 case ACE_EVERYONE:
1250 break;
1251 default:
1252 return (1);
1253 }
1254
1255 if (flags & (ACE_FILE_INHERIT_ACE|
1256 ACE_DIRECTORY_INHERIT_ACE|ACE_NO_PROPAGATE_INHERIT_ACE|
1257 ACE_INHERIT_ONLY_ACE))
1258 return (1);
1259
1260 /*
1261 * Special check for some special bits
1262 *
1263 * Don't allow anybody to deny reading basic
1264 * attributes or a files ACL.
1265 */
1266 if ((mask & (ACE_READ_ACL|ACE_READ_ATTRIBUTES)) &&
1267 (type == ACE_ACCESS_DENIED_ACE_TYPE))
1268 return (1);
1269
1270 /*
1271 * Delete permissions are never set by default
1272 */
1273 if (mask & (ACE_DELETE|ACE_DELETE_CHILD))
1274 return (1);
1275 /*
1276 * only allow owner@ to have
1277 * write_acl/write_owner/write_attributes/write_xattr/
1278 */
1279 if (type == ACE_ACCESS_ALLOWED_ACE_TYPE &&
1280 (!(flags & ACE_OWNER) && (mask &
1281 (ACE_WRITE_OWNER|ACE_WRITE_ACL| ACE_WRITE_ATTRIBUTES|
1282 ACE_WRITE_NAMED_ATTRS))))
1283 return (1);
1284
1285 }
1286
1287 return (0);
1288 }
1289
1290 /*
1291 * common code for setting ACLs.
1292 *
1293 * This function is called from zfs_mode_update, zfs_perm_init, and zfs_setacl.
1294 * zfs_setacl passes a non-NULL inherit pointer (ihp) to indicate that it's
1295 * already checked the acl and knows whether to inherit.
1296 */
1297 int
1298 zfs_aclset_common(znode_t *zp, zfs_acl_t *aclp, cred_t *cr, dmu_tx_t *tx)
1299 {
1300 int error;
1301 zfs_sb_t *zsb = ZTOZSB(zp);
1302 dmu_object_type_t otype;
1303 zfs_acl_locator_cb_t locate = { 0 };
1304 uint64_t mode;
1305 sa_bulk_attr_t bulk[5];
1306 uint64_t ctime[2];
1307 int count = 0;
1308
1309 mode = zp->z_mode;
1310
1311 mode = zfs_mode_compute(mode, aclp, &zp->z_pflags,
1312 zp->z_uid, zp->z_gid);
1313
1314 zp->z_mode = mode;
1315 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zsb), NULL,
1316 &mode, sizeof (mode));
1317 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zsb), NULL,
1318 &zp->z_pflags, sizeof (zp->z_pflags));
1319 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zsb), NULL,
1320 &ctime, sizeof (ctime));
1321
1322 if (zp->z_acl_cached) {
1323 zfs_acl_free(zp->z_acl_cached);
1324 zp->z_acl_cached = NULL;
1325 }
1326
1327 /*
1328 * Upgrade needed?
1329 */
1330 if (!zsb->z_use_fuids) {
1331 otype = DMU_OT_OLDACL;
1332 } else {
1333 if ((aclp->z_version == ZFS_ACL_VERSION_INITIAL) &&
1334 (zsb->z_version >= ZPL_VERSION_FUID))
1335 zfs_acl_xform(zp, aclp, cr);
1336 ASSERT(aclp->z_version >= ZFS_ACL_VERSION_FUID);
1337 otype = DMU_OT_ACL;
1338 }
1339
1340 /*
1341 * Arrgh, we have to handle old on disk format
1342 * as well as newer (preferred) SA format.
1343 */
1344
1345 if (zp->z_is_sa) { /* the easy case, just update the ACL attribute */
1346 locate.cb_aclp = aclp;
1347 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_DACL_ACES(zsb),
1348 zfs_acl_data_locator, &locate, aclp->z_acl_bytes);
1349 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_DACL_COUNT(zsb),
1350 NULL, &aclp->z_acl_count, sizeof (uint64_t));
1351 } else { /* Painful legacy way */
1352 zfs_acl_node_t *aclnode;
1353 uint64_t off = 0;
1354 zfs_acl_phys_t acl_phys;
1355 uint64_t aoid;
1356
1357 if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_ZNODE_ACL(zsb),
1358 &acl_phys, sizeof (acl_phys))) != 0)
1359 return (error);
1360
1361 aoid = acl_phys.z_acl_extern_obj;
1362
1363 if (aclp->z_acl_bytes > ZFS_ACE_SPACE) {
1364 /*
1365 * If ACL was previously external and we are now
1366 * converting to new ACL format then release old
1367 * ACL object and create a new one.
1368 */
1369 if (aoid &&
1370 aclp->z_version != acl_phys.z_acl_version) {
1371 error = dmu_object_free(zsb->z_os, aoid, tx);
1372 if (error)
1373 return (error);
1374 aoid = 0;
1375 }
1376 if (aoid == 0) {
1377 aoid = dmu_object_alloc(zsb->z_os,
1378 otype, aclp->z_acl_bytes,
1379 otype == DMU_OT_ACL ?
1380 DMU_OT_SYSACL : DMU_OT_NONE,
1381 otype == DMU_OT_ACL ?
1382 DN_MAX_BONUSLEN : 0, tx);
1383 } else {
1384 (void) dmu_object_set_blocksize(zsb->z_os,
1385 aoid, aclp->z_acl_bytes, 0, tx);
1386 }
1387 acl_phys.z_acl_extern_obj = aoid;
1388 for (aclnode = list_head(&aclp->z_acl); aclnode;
1389 aclnode = list_next(&aclp->z_acl, aclnode)) {
1390 if (aclnode->z_ace_count == 0)
1391 continue;
1392 dmu_write(zsb->z_os, aoid, off,
1393 aclnode->z_size, aclnode->z_acldata, tx);
1394 off += aclnode->z_size;
1395 }
1396 } else {
1397 void *start = acl_phys.z_ace_data;
1398 /*
1399 * Migrating back embedded?
1400 */
1401 if (acl_phys.z_acl_extern_obj) {
1402 error = dmu_object_free(zsb->z_os,
1403 acl_phys.z_acl_extern_obj, tx);
1404 if (error)
1405 return (error);
1406 acl_phys.z_acl_extern_obj = 0;
1407 }
1408
1409 for (aclnode = list_head(&aclp->z_acl); aclnode;
1410 aclnode = list_next(&aclp->z_acl, aclnode)) {
1411 if (aclnode->z_ace_count == 0)
1412 continue;
1413 bcopy(aclnode->z_acldata, start,
1414 aclnode->z_size);
1415 start = (caddr_t)start + aclnode->z_size;
1416 }
1417 }
1418 /*
1419 * If Old version then swap count/bytes to match old
1420 * layout of znode_acl_phys_t.
1421 */
1422 if (aclp->z_version == ZFS_ACL_VERSION_INITIAL) {
1423 acl_phys.z_acl_size = aclp->z_acl_count;
1424 acl_phys.z_acl_count = aclp->z_acl_bytes;
1425 } else {
1426 acl_phys.z_acl_size = aclp->z_acl_bytes;
1427 acl_phys.z_acl_count = aclp->z_acl_count;
1428 }
1429 acl_phys.z_acl_version = aclp->z_version;
1430
1431 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ZNODE_ACL(zsb), NULL,
1432 &acl_phys, sizeof (acl_phys));
1433 }
1434
1435 /*
1436 * Replace ACL wide bits, but first clear them.
1437 */
1438 zp->z_pflags &= ~ZFS_ACL_WIDE_FLAGS;
1439
1440 zp->z_pflags |= aclp->z_hints;
1441
1442 if (ace_trivial_common(aclp, 0, zfs_ace_walk) == 0)
1443 zp->z_pflags |= ZFS_ACL_TRIVIAL;
1444
1445 zfs_tstamp_update_setup(zp, STATE_CHANGED, NULL, ctime, B_TRUE);
1446 return (sa_bulk_update(zp->z_sa_hdl, bulk, count, tx));
1447 }
1448
1449 static void
1450 zfs_acl_chmod(zfs_sb_t *zsb, uint64_t mode, zfs_acl_t *aclp)
1451 {
1452 void *acep = NULL;
1453 uint64_t who;
1454 int new_count, new_bytes;
1455 int ace_size;
1456 int entry_type;
1457 uint16_t iflags, type;
1458 uint32_t access_mask;
1459 zfs_acl_node_t *newnode;
1460 size_t abstract_size = aclp->z_ops->ace_abstract_size();
1461 void *zacep;
1462 uint32_t owner, group, everyone;
1463 uint32_t deny1, deny2, allow0;
1464
1465 new_count = new_bytes = 0;
1466
1467 acl_trivial_access_masks((mode_t)mode, &allow0, &deny1, &deny2,
1468 &owner, &group, &everyone);
1469
1470 newnode = zfs_acl_node_alloc((abstract_size * 6) + aclp->z_acl_bytes);
1471
1472 zacep = newnode->z_acldata;
1473 if (allow0) {
1474 zfs_set_ace(aclp, zacep, allow0, ALLOW, -1, ACE_OWNER);
1475 zacep = (void *)((uintptr_t)zacep + abstract_size);
1476 new_count++;
1477 new_bytes += abstract_size;
1478 } if (deny1) {
1479 zfs_set_ace(aclp, zacep, deny1, DENY, -1, ACE_OWNER);
1480 zacep = (void *)((uintptr_t)zacep + abstract_size);
1481 new_count++;
1482 new_bytes += abstract_size;
1483 }
1484 if (deny2) {
1485 zfs_set_ace(aclp, zacep, deny2, DENY, -1, OWNING_GROUP);
1486 zacep = (void *)((uintptr_t)zacep + abstract_size);
1487 new_count++;
1488 new_bytes += abstract_size;
1489 }
1490
1491 while ((acep = zfs_acl_next_ace(aclp, acep, &who, &access_mask,
1492 &iflags, &type))) {
1493 uint16_t inherit_flags;
1494
1495 entry_type = (iflags & ACE_TYPE_FLAGS);
1496 inherit_flags = (iflags & ALL_INHERIT);
1497
1498 if ((entry_type == ACE_OWNER || entry_type == ACE_EVERYONE ||
1499 (entry_type == OWNING_GROUP)) &&
1500 ((inherit_flags & ACE_INHERIT_ONLY_ACE) == 0)) {
1501 continue;
1502 }
1503
1504 if ((type != ALLOW && type != DENY) ||
1505 (inherit_flags & ACE_INHERIT_ONLY_ACE)) {
1506 if (inherit_flags)
1507 aclp->z_hints |= ZFS_INHERIT_ACE;
1508 switch (type) {
1509 case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE:
1510 case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE:
1511 case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE:
1512 case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE:
1513 aclp->z_hints |= ZFS_ACL_OBJ_ACE;
1514 break;
1515 }
1516 } else {
1517
1518 /*
1519 * Limit permissions to be no greater than
1520 * group permissions
1521 */
1522 if (zsb->z_acl_inherit == ZFS_ACL_RESTRICTED) {
1523 if (!(mode & S_IRGRP))
1524 access_mask &= ~ACE_READ_DATA;
1525 if (!(mode & S_IWGRP))
1526 access_mask &=
1527 ~(ACE_WRITE_DATA|ACE_APPEND_DATA);
1528 if (!(mode & S_IXGRP))
1529 access_mask &= ~ACE_EXECUTE;
1530 access_mask &=
1531 ~(ACE_WRITE_OWNER|ACE_WRITE_ACL|
1532 ACE_WRITE_ATTRIBUTES|ACE_WRITE_NAMED_ATTRS);
1533 }
1534 }
1535 zfs_set_ace(aclp, zacep, access_mask, type, who, iflags);
1536 ace_size = aclp->z_ops->ace_size(acep);
1537 zacep = (void *)((uintptr_t)zacep + ace_size);
1538 new_count++;
1539 new_bytes += ace_size;
1540 }
1541 zfs_set_ace(aclp, zacep, owner, 0, -1, ACE_OWNER);
1542 zacep = (void *)((uintptr_t)zacep + abstract_size);
1543 zfs_set_ace(aclp, zacep, group, 0, -1, OWNING_GROUP);
1544 zacep = (void *)((uintptr_t)zacep + abstract_size);
1545 zfs_set_ace(aclp, zacep, everyone, 0, -1, ACE_EVERYONE);
1546
1547 new_count += 3;
1548 new_bytes += abstract_size * 3;
1549 zfs_acl_release_nodes(aclp);
1550 aclp->z_acl_count = new_count;
1551 aclp->z_acl_bytes = new_bytes;
1552 newnode->z_ace_count = new_count;
1553 newnode->z_size = new_bytes;
1554 list_insert_tail(&aclp->z_acl, newnode);
1555 }
1556
1557 void
1558 zfs_acl_chmod_setattr(znode_t *zp, zfs_acl_t **aclp, uint64_t mode)
1559 {
1560 mutex_enter(&zp->z_acl_lock);
1561 mutex_enter(&zp->z_lock);
1562 *aclp = zfs_acl_alloc(zfs_acl_version_zp(zp));
1563 (*aclp)->z_hints = zp->z_pflags & V4_ACL_WIDE_FLAGS;
1564 zfs_acl_chmod(ZTOZSB(zp), mode, *aclp);
1565 mutex_exit(&zp->z_lock);
1566 mutex_exit(&zp->z_acl_lock);
1567 ASSERT(*aclp);
1568 }
1569
1570 /*
1571 * strip off write_owner and write_acl
1572 */
1573 static void
1574 zfs_restricted_update(zfs_sb_t *zsb, zfs_acl_t *aclp, void *acep)
1575 {
1576 uint32_t mask = aclp->z_ops->ace_mask_get(acep);
1577
1578 if ((zsb->z_acl_inherit == ZFS_ACL_RESTRICTED) &&
1579 (aclp->z_ops->ace_type_get(acep) == ALLOW)) {
1580 mask &= ~RESTRICTED_CLEAR;
1581 aclp->z_ops->ace_mask_set(acep, mask);
1582 }
1583 }
1584
1585 /*
1586 * Should ACE be inherited?
1587 */
1588 static int
1589 zfs_ace_can_use(umode_t obj_mode, uint16_t acep_flags)
1590 {
1591 int iflags = (acep_flags & 0xf);
1592
1593 if (S_ISDIR(obj_mode) && (iflags & ACE_DIRECTORY_INHERIT_ACE))
1594 return (1);
1595 else if (iflags & ACE_FILE_INHERIT_ACE)
1596 return (!(S_ISDIR(obj_mode) &&
1597 (iflags & ACE_NO_PROPAGATE_INHERIT_ACE)));
1598 return (0);
1599 }
1600
1601 /*
1602 * inherit inheritable ACEs from parent
1603 */
1604 static zfs_acl_t *
1605 zfs_acl_inherit(zfs_sb_t *zsb, umode_t obj_mode, zfs_acl_t *paclp,
1606 uint64_t mode, boolean_t *need_chmod)
1607 {
1608 void *pacep;
1609 void *acep;
1610 zfs_acl_node_t *aclnode;
1611 zfs_acl_t *aclp = NULL;
1612 uint64_t who;
1613 uint32_t access_mask;
1614 uint16_t iflags, newflags, type;
1615 size_t ace_size;
1616 void *data1, *data2;
1617 size_t data1sz, data2sz;
1618 boolean_t vdir = S_ISDIR(obj_mode);
1619 boolean_t vreg = S_ISREG(obj_mode);
1620 boolean_t passthrough, passthrough_x, noallow;
1621
1622 passthrough_x =
1623 zsb->z_acl_inherit == ZFS_ACL_PASSTHROUGH_X;
1624 passthrough = passthrough_x ||
1625 zsb->z_acl_inherit == ZFS_ACL_PASSTHROUGH;
1626 noallow =
1627 zsb->z_acl_inherit == ZFS_ACL_NOALLOW;
1628
1629 *need_chmod = B_TRUE;
1630 pacep = NULL;
1631 aclp = zfs_acl_alloc(paclp->z_version);
1632 if (zsb->z_acl_inherit == ZFS_ACL_DISCARD || S_ISLNK(obj_mode))
1633 return (aclp);
1634 while ((pacep = zfs_acl_next_ace(paclp, pacep, &who,
1635 &access_mask, &iflags, &type))) {
1636
1637 /*
1638 * don't inherit bogus ACEs
1639 */
1640 if (!zfs_acl_valid_ace_type(type, iflags))
1641 continue;
1642
1643 if (noallow && type == ALLOW)
1644 continue;
1645
1646 ace_size = aclp->z_ops->ace_size(pacep);
1647
1648 if (!zfs_ace_can_use(obj_mode, iflags))
1649 continue;
1650
1651 /*
1652 * If owner@, group@, or everyone@ inheritable
1653 * then zfs_acl_chmod() isn't needed.
1654 */
1655 if (passthrough &&
1656 ((iflags & (ACE_OWNER|ACE_EVERYONE)) ||
1657 ((iflags & OWNING_GROUP) ==
1658 OWNING_GROUP)) && (vreg || (vdir && (iflags &
1659 ACE_DIRECTORY_INHERIT_ACE)))) {
1660 *need_chmod = B_FALSE;
1661 }
1662
1663 if (!vdir && passthrough_x &&
1664 ((mode & (S_IXUSR | S_IXGRP | S_IXOTH)) == 0)) {
1665 access_mask &= ~ACE_EXECUTE;
1666 }
1667
1668 aclnode = zfs_acl_node_alloc(ace_size);
1669 list_insert_tail(&aclp->z_acl, aclnode);
1670 acep = aclnode->z_acldata;
1671
1672 zfs_set_ace(aclp, acep, access_mask, type,
1673 who, iflags|ACE_INHERITED_ACE);
1674
1675 /*
1676 * Copy special opaque data if any
1677 */
1678 if ((data1sz = paclp->z_ops->ace_data(pacep, &data1)) != 0) {
1679 VERIFY((data2sz = aclp->z_ops->ace_data(acep,
1680 &data2)) == data1sz);
1681 bcopy(data1, data2, data2sz);
1682 }
1683
1684 aclp->z_acl_count++;
1685 aclnode->z_ace_count++;
1686 aclp->z_acl_bytes += aclnode->z_size;
1687 newflags = aclp->z_ops->ace_flags_get(acep);
1688
1689 if (vdir)
1690 aclp->z_hints |= ZFS_INHERIT_ACE;
1691
1692 if ((iflags & ACE_NO_PROPAGATE_INHERIT_ACE) || !vdir) {
1693 newflags &= ~ALL_INHERIT;
1694 aclp->z_ops->ace_flags_set(acep,
1695 newflags|ACE_INHERITED_ACE);
1696 zfs_restricted_update(zsb, aclp, acep);
1697 continue;
1698 }
1699
1700 ASSERT(vdir);
1701
1702 /*
1703 * If only FILE_INHERIT is set then turn on
1704 * inherit_only
1705 */
1706 if ((iflags & (ACE_FILE_INHERIT_ACE |
1707 ACE_DIRECTORY_INHERIT_ACE)) == ACE_FILE_INHERIT_ACE) {
1708 newflags |= ACE_INHERIT_ONLY_ACE;
1709 aclp->z_ops->ace_flags_set(acep,
1710 newflags|ACE_INHERITED_ACE);
1711 } else {
1712 newflags &= ~ACE_INHERIT_ONLY_ACE;
1713 aclp->z_ops->ace_flags_set(acep,
1714 newflags|ACE_INHERITED_ACE);
1715 }
1716 }
1717 return (aclp);
1718 }
1719
1720 /*
1721 * Create file system object initial permissions
1722 * including inheritable ACEs.
1723 */
1724 int
1725 zfs_acl_ids_create(znode_t *dzp, int flag, vattr_t *vap, cred_t *cr,
1726 vsecattr_t *vsecp, zfs_acl_ids_t *acl_ids)
1727 {
1728 int error;
1729 zfs_sb_t *zsb = ZTOZSB(dzp);
1730 zfs_acl_t *paclp;
1731 #ifdef HAVE_KSID
1732 gid_t gid;
1733 #endif /* HAVE_KSID */
1734 boolean_t need_chmod = B_TRUE;
1735 boolean_t inherited = B_FALSE;
1736
1737 bzero(acl_ids, sizeof (zfs_acl_ids_t));
1738 acl_ids->z_mode = vap->va_mode;
1739
1740 if (vsecp)
1741 if ((error = zfs_vsec_2_aclp(zsb, vap->va_mode, vsecp,
1742 cr, &acl_ids->z_fuidp, &acl_ids->z_aclp)) != 0)
1743 return (error);
1744
1745 acl_ids->z_fuid = vap->va_uid;
1746 acl_ids->z_fgid = vap->va_gid;
1747 #ifdef HAVE_KSID
1748 /*
1749 * Determine uid and gid.
1750 */
1751 if ((flag & IS_ROOT_NODE) || zsb->z_replay ||
1752 ((flag & IS_XATTR) && (S_ISDIR(vap->va_mode)))) {
1753 acl_ids->z_fuid = zfs_fuid_create(zsb, (uint64_t)vap->va_uid,
1754 cr, ZFS_OWNER, &acl_ids->z_fuidp);
1755 acl_ids->z_fgid = zfs_fuid_create(zsb, (uint64_t)vap->va_gid,
1756 cr, ZFS_GROUP, &acl_ids->z_fuidp);
1757 gid = vap->va_gid;
1758 } else {
1759 acl_ids->z_fuid = zfs_fuid_create_cred(zsb, ZFS_OWNER,
1760 cr, &acl_ids->z_fuidp);
1761 acl_ids->z_fgid = 0;
1762 if (vap->va_mask & AT_GID) {
1763 acl_ids->z_fgid = zfs_fuid_create(zsb,
1764 (uint64_t)vap->va_gid,
1765 cr, ZFS_GROUP, &acl_ids->z_fuidp);
1766 gid = vap->va_gid;
1767 if (acl_ids->z_fgid != dzp->z_gid &&
1768 !groupmember(vap->va_gid, cr) &&
1769 secpolicy_vnode_create_gid(cr) != 0)
1770 acl_ids->z_fgid = 0;
1771 }
1772 if (acl_ids->z_fgid == 0) {
1773 if (dzp->z_mode & S_ISGID) {
1774 char *domain;
1775 uint32_t rid;
1776
1777 acl_ids->z_fgid = dzp->z_gid;
1778 gid = zfs_fuid_map_id(zsb, acl_ids->z_fgid,
1779 cr, ZFS_GROUP);
1780
1781 if (zsb->z_use_fuids &&
1782 IS_EPHEMERAL(acl_ids->z_fgid)) {
1783 domain = zfs_fuid_idx_domain(
1784 &zsb->z_fuid_idx,
1785 FUID_INDEX(acl_ids->z_fgid));
1786 rid = FUID_RID(acl_ids->z_fgid);
1787 zfs_fuid_node_add(&acl_ids->z_fuidp,
1788 domain, rid,
1789 FUID_INDEX(acl_ids->z_fgid),
1790 acl_ids->z_fgid, ZFS_GROUP);
1791 }
1792 } else {
1793 acl_ids->z_fgid = zfs_fuid_create_cred(zsb,
1794 ZFS_GROUP, cr, &acl_ids->z_fuidp);
1795 gid = crgetgid(cr);
1796 }
1797 }
1798 }
1799 #endif /* HAVE_KSID */
1800
1801 /*
1802 * If we're creating a directory, and the parent directory has the
1803 * set-GID bit set, set in on the new directory.
1804 * Otherwise, if the user is neither privileged nor a member of the
1805 * file's new group, clear the file's set-GID bit.
1806 */
1807
1808 if (!(flag & IS_ROOT_NODE) && (dzp->z_mode & S_ISGID) &&
1809 (S_ISDIR(vap->va_mode))) {
1810 acl_ids->z_mode |= S_ISGID;
1811 } else {
1812 if ((acl_ids->z_mode & S_ISGID) &&
1813 secpolicy_vnode_setids_setgids(cr, gid) != 0)
1814 acl_ids->z_mode &= ~S_ISGID;
1815 }
1816
1817 if (acl_ids->z_aclp == NULL) {
1818 mutex_enter(&dzp->z_acl_lock);
1819 mutex_enter(&dzp->z_lock);
1820 if (!(flag & IS_ROOT_NODE) && (S_ISDIR(ZTOI(dzp)->i_mode) &&
1821 (dzp->z_pflags & ZFS_INHERIT_ACE)) &&
1822 !(dzp->z_pflags & ZFS_XATTR)) {
1823 VERIFY(0 == zfs_acl_node_read(dzp, B_TRUE,
1824 &paclp, B_FALSE));
1825 acl_ids->z_aclp = zfs_acl_inherit(zsb,
1826 vap->va_mode, paclp, acl_ids->z_mode, &need_chmod);
1827 inherited = B_TRUE;
1828 } else {
1829 acl_ids->z_aclp =
1830 zfs_acl_alloc(zfs_acl_version_zp(dzp));
1831 acl_ids->z_aclp->z_hints |= ZFS_ACL_TRIVIAL;
1832 }
1833 mutex_exit(&dzp->z_lock);
1834 mutex_exit(&dzp->z_acl_lock);
1835 if (need_chmod) {
1836 acl_ids->z_aclp->z_hints |= S_ISDIR(vap->va_mode) ?
1837 ZFS_ACL_AUTO_INHERIT : 0;
1838 zfs_acl_chmod(zsb, acl_ids->z_mode, acl_ids->z_aclp);
1839 }
1840 }
1841
1842 if (inherited || vsecp) {
1843 acl_ids->z_mode = zfs_mode_compute(acl_ids->z_mode,
1844 acl_ids->z_aclp, &acl_ids->z_aclp->z_hints,
1845 acl_ids->z_fuid, acl_ids->z_fgid);
1846 if (ace_trivial_common(acl_ids->z_aclp, 0, zfs_ace_walk) == 0)
1847 acl_ids->z_aclp->z_hints |= ZFS_ACL_TRIVIAL;
1848 }
1849
1850 return (0);
1851 }
1852
1853 /*
1854 * Free ACL and fuid_infop, but not the acl_ids structure
1855 */
1856 void
1857 zfs_acl_ids_free(zfs_acl_ids_t *acl_ids)
1858 {
1859 if (acl_ids->z_aclp)
1860 zfs_acl_free(acl_ids->z_aclp);
1861 if (acl_ids->z_fuidp)
1862 zfs_fuid_info_free(acl_ids->z_fuidp);
1863 acl_ids->z_aclp = NULL;
1864 acl_ids->z_fuidp = NULL;
1865 }
1866
1867 boolean_t
1868 zfs_acl_ids_overquota(zfs_sb_t *zsb, zfs_acl_ids_t *acl_ids)
1869 {
1870 return (zfs_fuid_overquota(zsb, B_FALSE, acl_ids->z_fuid) ||
1871 zfs_fuid_overquota(zsb, B_TRUE, acl_ids->z_fgid));
1872 }
1873
1874 /*
1875 * Retrieve a files ACL
1876 */
1877 int
1878 zfs_getacl(znode_t *zp, vsecattr_t *vsecp, boolean_t skipaclchk, cred_t *cr)
1879 {
1880 zfs_acl_t *aclp;
1881 ulong_t mask;
1882 int error;
1883 int count = 0;
1884 int largeace = 0;
1885
1886 mask = vsecp->vsa_mask & (VSA_ACE | VSA_ACECNT |
1887 VSA_ACE_ACLFLAGS | VSA_ACE_ALLTYPES);
1888
1889 if (mask == 0)
1890 return (ENOSYS);
1891
1892 if ((error = zfs_zaccess(zp, ACE_READ_ACL, 0, skipaclchk, cr)))
1893 return (error);
1894
1895 mutex_enter(&zp->z_acl_lock);
1896
1897 error = zfs_acl_node_read(zp, B_FALSE, &aclp, B_FALSE);
1898 if (error != 0) {
1899 mutex_exit(&zp->z_acl_lock);
1900 return (error);
1901 }
1902
1903 /*
1904 * Scan ACL to determine number of ACEs
1905 */
1906 if ((zp->z_pflags & ZFS_ACL_OBJ_ACE) && !(mask & VSA_ACE_ALLTYPES)) {
1907 void *zacep = NULL;
1908 uint64_t who;
1909 uint32_t access_mask;
1910 uint16_t type, iflags;
1911
1912 while ((zacep = zfs_acl_next_ace(aclp, zacep,
1913 &who, &access_mask, &iflags, &type))) {
1914 switch (type) {
1915 case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE:
1916 case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE:
1917 case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE:
1918 case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE:
1919 largeace++;
1920 continue;
1921 default:
1922 count++;
1923 }
1924 }
1925 vsecp->vsa_aclcnt = count;
1926 } else
1927 count = (int)aclp->z_acl_count;
1928
1929 if (mask & VSA_ACECNT) {
1930 vsecp->vsa_aclcnt = count;
1931 }
1932
1933 if (mask & VSA_ACE) {
1934 size_t aclsz;
1935
1936 aclsz = count * sizeof (ace_t) +
1937 sizeof (ace_object_t) * largeace;
1938
1939 vsecp->vsa_aclentp = kmem_alloc(aclsz, KM_SLEEP);
1940 vsecp->vsa_aclentsz = aclsz;
1941
1942 if (aclp->z_version == ZFS_ACL_VERSION_FUID)
1943 zfs_copy_fuid_2_ace(ZTOZSB(zp), aclp, cr,
1944 vsecp->vsa_aclentp, !(mask & VSA_ACE_ALLTYPES));
1945 else {
1946 zfs_acl_node_t *aclnode;
1947 void *start = vsecp->vsa_aclentp;
1948
1949 for (aclnode = list_head(&aclp->z_acl); aclnode;
1950 aclnode = list_next(&aclp->z_acl, aclnode)) {
1951 bcopy(aclnode->z_acldata, start,
1952 aclnode->z_size);
1953 start = (caddr_t)start + aclnode->z_size;
1954 }
1955 ASSERT((caddr_t)start - (caddr_t)vsecp->vsa_aclentp ==
1956 aclp->z_acl_bytes);
1957 }
1958 }
1959 if (mask & VSA_ACE_ACLFLAGS) {
1960 vsecp->vsa_aclflags = 0;
1961 if (zp->z_pflags & ZFS_ACL_DEFAULTED)
1962 vsecp->vsa_aclflags |= ACL_DEFAULTED;
1963 if (zp->z_pflags & ZFS_ACL_PROTECTED)
1964 vsecp->vsa_aclflags |= ACL_PROTECTED;
1965 if (zp->z_pflags & ZFS_ACL_AUTO_INHERIT)
1966 vsecp->vsa_aclflags |= ACL_AUTO_INHERIT;
1967 }
1968
1969 mutex_exit(&zp->z_acl_lock);
1970
1971 return (0);
1972 }
1973
1974 int
1975 zfs_vsec_2_aclp(zfs_sb_t *zsb, umode_t obj_mode,
1976 vsecattr_t *vsecp, cred_t *cr, zfs_fuid_info_t **fuidp, zfs_acl_t **zaclp)
1977 {
1978 zfs_acl_t *aclp;
1979 zfs_acl_node_t *aclnode;
1980 int aclcnt = vsecp->vsa_aclcnt;
1981 int error;
1982
1983 if (vsecp->vsa_aclcnt > MAX_ACL_ENTRIES || vsecp->vsa_aclcnt <= 0)
1984 return (EINVAL);
1985
1986 aclp = zfs_acl_alloc(zfs_acl_version(zsb->z_version));
1987
1988 aclp->z_hints = 0;
1989 aclnode = zfs_acl_node_alloc(aclcnt * sizeof (zfs_object_ace_t));
1990 if (aclp->z_version == ZFS_ACL_VERSION_INITIAL) {
1991 if ((error = zfs_copy_ace_2_oldace(obj_mode, aclp,
1992 (ace_t *)vsecp->vsa_aclentp, aclnode->z_acldata,
1993 aclcnt, &aclnode->z_size)) != 0) {
1994 zfs_acl_free(aclp);
1995 zfs_acl_node_free(aclnode);
1996 return (error);
1997 }
1998 } else {
1999 if ((error = zfs_copy_ace_2_fuid(zsb, obj_mode, aclp,
2000 vsecp->vsa_aclentp, aclnode->z_acldata, aclcnt,
2001 &aclnode->z_size, fuidp, cr)) != 0) {
2002 zfs_acl_free(aclp);
2003 zfs_acl_node_free(aclnode);
2004 return (error);
2005 }
2006 }
2007 aclp->z_acl_bytes = aclnode->z_size;
2008 aclnode->z_ace_count = aclcnt;
2009 aclp->z_acl_count = aclcnt;
2010 list_insert_head(&aclp->z_acl, aclnode);
2011
2012 /*
2013 * If flags are being set then add them to z_hints
2014 */
2015 if (vsecp->vsa_mask & VSA_ACE_ACLFLAGS) {
2016 if (vsecp->vsa_aclflags & ACL_PROTECTED)
2017 aclp->z_hints |= ZFS_ACL_PROTECTED;
2018 if (vsecp->vsa_aclflags & ACL_DEFAULTED)
2019 aclp->z_hints |= ZFS_ACL_DEFAULTED;
2020 if (vsecp->vsa_aclflags & ACL_AUTO_INHERIT)
2021 aclp->z_hints |= ZFS_ACL_AUTO_INHERIT;
2022 }
2023
2024 *zaclp = aclp;
2025
2026 return (0);
2027 }
2028
2029 /*
2030 * Set a files ACL
2031 */
2032 int
2033 zfs_setacl(znode_t *zp, vsecattr_t *vsecp, boolean_t skipaclchk, cred_t *cr)
2034 {
2035 zfs_sb_t *zsb = ZTOZSB(zp);
2036 zilog_t *zilog = zsb->z_log;
2037 ulong_t mask = vsecp->vsa_mask & (VSA_ACE | VSA_ACECNT);
2038 dmu_tx_t *tx;
2039 int error;
2040 zfs_acl_t *aclp;
2041 zfs_fuid_info_t *fuidp = NULL;
2042 boolean_t fuid_dirtied;
2043 uint64_t acl_obj;
2044
2045 if (mask == 0)
2046 return (ENOSYS);
2047
2048 if (zp->z_pflags & ZFS_IMMUTABLE)
2049 return (EPERM);
2050
2051 if ((error = zfs_zaccess(zp, ACE_WRITE_ACL, 0, skipaclchk, cr)))
2052 return (error);
2053
2054 error = zfs_vsec_2_aclp(zsb, ZTOI(zp)->i_mode, vsecp, cr, &fuidp,
2055 &aclp);
2056 if (error)
2057 return (error);
2058
2059 /*
2060 * If ACL wide flags aren't being set then preserve any
2061 * existing flags.
2062 */
2063 if (!(vsecp->vsa_mask & VSA_ACE_ACLFLAGS)) {
2064 aclp->z_hints |=
2065 (zp->z_pflags & V4_ACL_WIDE_FLAGS);
2066 }
2067 top:
2068 mutex_enter(&zp->z_acl_lock);
2069 mutex_enter(&zp->z_lock);
2070
2071 tx = dmu_tx_create(zsb->z_os);
2072
2073 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE);
2074
2075 fuid_dirtied = zsb->z_fuid_dirty;
2076 if (fuid_dirtied)
2077 zfs_fuid_txhold(zsb, tx);
2078
2079 /*
2080 * If old version and ACL won't fit in bonus and we aren't
2081 * upgrading then take out necessary DMU holds
2082 */
2083
2084 if ((acl_obj = zfs_external_acl(zp)) != 0) {
2085 if (zsb->z_version >= ZPL_VERSION_FUID &&
2086 zfs_znode_acl_version(zp) <= ZFS_ACL_VERSION_INITIAL) {
2087 dmu_tx_hold_free(tx, acl_obj, 0,
2088 DMU_OBJECT_END);
2089 dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0,
2090 aclp->z_acl_bytes);
2091 } else {
2092 dmu_tx_hold_write(tx, acl_obj, 0, aclp->z_acl_bytes);
2093 }
2094 } else if (!zp->z_is_sa && aclp->z_acl_bytes > ZFS_ACE_SPACE) {
2095 dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, aclp->z_acl_bytes);
2096 }
2097
2098 zfs_sa_upgrade_txholds(tx, zp);
2099 error = dmu_tx_assign(tx, TXG_NOWAIT);
2100 if (error) {
2101 mutex_exit(&zp->z_acl_lock);
2102 mutex_exit(&zp->z_lock);
2103
2104 if (error == ERESTART) {
2105 dmu_tx_wait(tx);
2106 dmu_tx_abort(tx);
2107 goto top;
2108 }
2109 dmu_tx_abort(tx);
2110 zfs_acl_free(aclp);
2111 return (error);
2112 }
2113
2114 error = zfs_aclset_common(zp, aclp, cr, tx);
2115 ASSERT(error == 0);
2116 ASSERT(zp->z_acl_cached == NULL);
2117 zp->z_acl_cached = aclp;
2118
2119 if (fuid_dirtied)
2120 zfs_fuid_sync(zsb, tx);
2121
2122 zfs_log_acl(zilog, tx, zp, vsecp, fuidp);
2123
2124 if (fuidp)
2125 zfs_fuid_info_free(fuidp);
2126 dmu_tx_commit(tx);
2127
2128 mutex_exit(&zp->z_lock);
2129 mutex_exit(&zp->z_acl_lock);
2130
2131 return (error);
2132 }
2133
2134 /*
2135 * Check accesses of interest (AoI) against attributes of the dataset
2136 * such as read-only. Returns zero if no AoI conflict with dataset
2137 * attributes, otherwise an appropriate errno is returned.
2138 */
2139 static int
2140 zfs_zaccess_dataset_check(znode_t *zp, uint32_t v4_mode)
2141 {
2142 if ((v4_mode & WRITE_MASK) && (zfs_is_readonly(ZTOZSB(zp))) &&
2143 (!S_ISDEV(ZTOI(zp)->i_mode) ||
2144 (S_ISDEV(ZTOI(zp)->i_mode) && (v4_mode & WRITE_MASK_ATTRS)))) {
2145 return (EROFS);
2146 }
2147
2148 /*
2149 * Only check for READONLY on non-directories.
2150 */
2151 if ((v4_mode & WRITE_MASK_DATA) &&
2152 ((!S_ISDIR(ZTOI(zp)->i_mode) &&
2153 (zp->z_pflags & (ZFS_READONLY | ZFS_IMMUTABLE))) ||
2154 (S_ISDIR(ZTOI(zp)->i_mode) &&
2155 (zp->z_pflags & ZFS_IMMUTABLE)))) {
2156 return (EPERM);
2157 }
2158
2159 if ((v4_mode & (ACE_DELETE | ACE_DELETE_CHILD)) &&
2160 (zp->z_pflags & ZFS_NOUNLINK)) {
2161 return (EPERM);
2162 }
2163
2164 if (((v4_mode & (ACE_READ_DATA|ACE_EXECUTE)) &&
2165 (zp->z_pflags & ZFS_AV_QUARANTINED))) {
2166 return (EACCES);
2167 }
2168
2169 return (0);
2170 }
2171
2172 /*
2173 * The primary usage of this function is to loop through all of the
2174 * ACEs in the znode, determining what accesses of interest (AoI) to
2175 * the caller are allowed or denied. The AoI are expressed as bits in
2176 * the working_mode parameter. As each ACE is processed, bits covered
2177 * by that ACE are removed from the working_mode. This removal
2178 * facilitates two things. The first is that when the working mode is
2179 * empty (= 0), we know we've looked at all the AoI. The second is
2180 * that the ACE interpretation rules don't allow a later ACE to undo
2181 * something granted or denied by an earlier ACE. Removing the
2182 * discovered access or denial enforces this rule. At the end of
2183 * processing the ACEs, all AoI that were found to be denied are
2184 * placed into the working_mode, giving the caller a mask of denied
2185 * accesses. Returns:
2186 * 0 if all AoI granted
2187 * EACCESS if the denied mask is non-zero
2188 * other error if abnormal failure (e.g., IO error)
2189 *
2190 * A secondary usage of the function is to determine if any of the
2191 * AoI are granted. If an ACE grants any access in
2192 * the working_mode, we immediately short circuit out of the function.
2193 * This mode is chosen by setting anyaccess to B_TRUE. The
2194 * working_mode is not a denied access mask upon exit if the function
2195 * is used in this manner.
2196 */
2197 static int
2198 zfs_zaccess_aces_check(znode_t *zp, uint32_t *working_mode,
2199 boolean_t anyaccess, cred_t *cr)
2200 {
2201 zfs_sb_t *zsb = ZTOZSB(zp);
2202 zfs_acl_t *aclp;
2203 int error;
2204 uid_t uid = crgetuid(cr);
2205 uint64_t who;
2206 uint16_t type, iflags;
2207 uint16_t entry_type;
2208 uint32_t access_mask;
2209 uint32_t deny_mask = 0;
2210 zfs_ace_hdr_t *acep = NULL;
2211 boolean_t checkit;
2212 uid_t gowner;
2213 uid_t fowner;
2214
2215 zfs_fuid_map_ids(zp, cr, &fowner, &gowner);
2216
2217 mutex_enter(&zp->z_acl_lock);
2218
2219 error = zfs_acl_node_read(zp, B_FALSE, &aclp, B_FALSE);
2220 if (error != 0) {
2221 mutex_exit(&zp->z_acl_lock);
2222 return (error);
2223 }
2224
2225 ASSERT(zp->z_acl_cached);
2226
2227 while ((acep = zfs_acl_next_ace(aclp, acep, &who, &access_mask,
2228 &iflags, &type))) {
2229 uint32_t mask_matched;
2230
2231 if (!zfs_acl_valid_ace_type(type, iflags))
2232 continue;
2233
2234 if (S_ISDIR(ZTOI(zp)->i_mode) &&
2235 (iflags & ACE_INHERIT_ONLY_ACE))
2236 continue;
2237
2238 /* Skip ACE if it does not affect any AoI */
2239 mask_matched = (access_mask & *working_mode);
2240 if (!mask_matched)
2241 continue;
2242
2243 entry_type = (iflags & ACE_TYPE_FLAGS);
2244
2245 checkit = B_FALSE;
2246
2247 switch (entry_type) {
2248 case ACE_OWNER:
2249 if (uid == fowner)
2250 checkit = B_TRUE;
2251 break;
2252 case OWNING_GROUP:
2253 who = gowner;
2254 /*FALLTHROUGH*/
2255 case ACE_IDENTIFIER_GROUP:
2256 checkit = zfs_groupmember(zsb, who, cr);
2257 break;
2258 case ACE_EVERYONE:
2259 checkit = B_TRUE;
2260 break;
2261
2262 /* USER Entry */
2263 default:
2264 if (entry_type == 0) {
2265 uid_t newid;
2266
2267 newid = zfs_fuid_map_id(zsb, who, cr,
2268 ZFS_ACE_USER);
2269 if (newid != IDMAP_WK_CREATOR_OWNER_UID &&
2270 uid == newid)
2271 checkit = B_TRUE;
2272 break;
2273 } else {
2274 mutex_exit(&zp->z_acl_lock);
2275 return (EIO);
2276 }
2277 }
2278
2279 if (checkit) {
2280 if (type == DENY) {
2281 DTRACE_PROBE3(zfs__ace__denies,
2282 znode_t *, zp,
2283 zfs_ace_hdr_t *, acep,
2284 uint32_t, mask_matched);
2285 deny_mask |= mask_matched;
2286 } else {
2287 DTRACE_PROBE3(zfs__ace__allows,
2288 znode_t *, zp,
2289 zfs_ace_hdr_t *, acep,
2290 uint32_t, mask_matched);
2291 if (anyaccess) {
2292 mutex_exit(&zp->z_acl_lock);
2293 return (0);
2294 }
2295 }
2296 *working_mode &= ~mask_matched;
2297 }
2298
2299 /* Are we done? */
2300 if (*working_mode == 0)
2301 break;
2302 }
2303
2304 mutex_exit(&zp->z_acl_lock);
2305
2306 /* Put the found 'denies' back on the working mode */
2307 if (deny_mask) {
2308 *working_mode |= deny_mask;
2309 return (EACCES);
2310 } else if (*working_mode) {
2311 return (-1);
2312 }
2313
2314 return (0);
2315 }
2316
2317 /*
2318 * Return true if any access whatsoever granted, we don't actually
2319 * care what access is granted.
2320 */
2321 boolean_t
2322 zfs_has_access(znode_t *zp, cred_t *cr)
2323 {
2324 uint32_t have = ACE_ALL_PERMS;
2325
2326 if (zfs_zaccess_aces_check(zp, &have, B_TRUE, cr) != 0) {
2327 uid_t owner;
2328
2329 owner = zfs_fuid_map_id(ZTOZSB(zp), zp->z_uid, cr, ZFS_OWNER);
2330 return (secpolicy_vnode_any_access(cr, ZTOI(zp), owner) == 0);
2331 }
2332 return (B_TRUE);
2333 }
2334
2335 static int
2336 zfs_zaccess_common(znode_t *zp, uint32_t v4_mode, uint32_t *working_mode,
2337 boolean_t *check_privs, boolean_t skipaclchk, cred_t *cr)
2338 {
2339 zfs_sb_t *zsb = ZTOZSB(zp);
2340 int err;
2341
2342 *working_mode = v4_mode;
2343 *check_privs = B_TRUE;
2344
2345 /*
2346 * Short circuit empty requests
2347 */
2348 if (v4_mode == 0 || zsb->z_replay) {
2349 *working_mode = 0;
2350 return (0);
2351 }
2352
2353 if ((err = zfs_zaccess_dataset_check(zp, v4_mode)) != 0) {
2354 *check_privs = B_FALSE;
2355 return (err);
2356 }
2357
2358 /*
2359 * The caller requested that the ACL check be skipped. This
2360 * would only happen if the caller checked VOP_ACCESS() with a
2361 * 32 bit ACE mask and already had the appropriate permissions.
2362 */
2363 if (skipaclchk) {
2364 *working_mode = 0;
2365 return (0);
2366 }
2367
2368 return (zfs_zaccess_aces_check(zp, working_mode, B_FALSE, cr));
2369 }
2370
2371 static int
2372 zfs_zaccess_append(znode_t *zp, uint32_t *working_mode, boolean_t *check_privs,
2373 cred_t *cr)
2374 {
2375 if (*working_mode != ACE_WRITE_DATA)
2376 return (EACCES);
2377
2378 return (zfs_zaccess_common(zp, ACE_APPEND_DATA, working_mode,
2379 check_privs, B_FALSE, cr));
2380 }
2381
2382 int
2383 zfs_fastaccesschk_execute(znode_t *zdp, cred_t *cr)
2384 {
2385 boolean_t owner = B_FALSE;
2386 boolean_t groupmbr = B_FALSE;
2387 boolean_t is_attr;
2388 uid_t uid = crgetuid(cr);
2389 int error;
2390
2391 if (zdp->z_pflags & ZFS_AV_QUARANTINED)
2392 return (EACCES);
2393
2394 is_attr = ((zdp->z_pflags & ZFS_XATTR) &&
2395 (S_ISDIR(ZTOI(zdp)->i_mode)));
2396 if (is_attr)
2397 goto slow;
2398
2399
2400 mutex_enter(&zdp->z_acl_lock);
2401
2402 if (zdp->z_pflags & ZFS_NO_EXECS_DENIED) {
2403 mutex_exit(&zdp->z_acl_lock);
2404 return (0);
2405 }
2406
2407 if (FUID_INDEX(zdp->z_uid) != 0 || FUID_INDEX(zdp->z_gid) != 0) {
2408 mutex_exit(&zdp->z_acl_lock);
2409 goto slow;
2410 }
2411
2412 if (uid == zdp->z_uid) {
2413 owner = B_TRUE;
2414 if (zdp->z_mode & S_IXUSR) {
2415 mutex_exit(&zdp->z_acl_lock);
2416 return (0);
2417 } else {
2418 mutex_exit(&zdp->z_acl_lock);
2419 goto slow;
2420 }
2421 }
2422 if (groupmember(zdp->z_gid, cr)) {
2423 groupmbr = B_TRUE;
2424 if (zdp->z_mode & S_IXGRP) {
2425 mutex_exit(&zdp->z_acl_lock);
2426 return (0);
2427 } else {
2428 mutex_exit(&zdp->z_acl_lock);
2429 goto slow;
2430 }
2431 }
2432 if (!owner && !groupmbr) {
2433 if (zdp->z_mode & S_IXOTH) {
2434 mutex_exit(&zdp->z_acl_lock);
2435 return (0);
2436 }
2437 }
2438
2439 mutex_exit(&zdp->z_acl_lock);
2440
2441 slow:
2442 DTRACE_PROBE(zfs__fastpath__execute__access__miss);
2443 ZFS_ENTER(ZTOZSB(zdp));
2444 error = zfs_zaccess(zdp, ACE_EXECUTE, 0, B_FALSE, cr);
2445 ZFS_EXIT(ZTOZSB(zdp));
2446 return (error);
2447 }
2448
2449 /*
2450 * Determine whether Access should be granted/denied.
2451 * The least priv subsytem is always consulted as a basic privilege
2452 * can define any form of access.
2453 */
2454 int
2455 zfs_zaccess(znode_t *zp, int mode, int flags, boolean_t skipaclchk, cred_t *cr)
2456 {
2457 uint32_t working_mode;
2458 int error;
2459 boolean_t check_privs;
2460 znode_t *check_zp = zp;
2461 mode_t needed_bits;
2462 uid_t owner;
2463
2464 /*
2465 * If attribute then validate against base file
2466 */
2467 if ((zp->z_pflags & ZFS_XATTR) && S_ISDIR(ZTOI(zp)->i_mode)) {
2468 uint64_t parent;
2469
2470 rw_enter(&zp->z_xattr_lock, RW_READER);
2471 if (zp->z_xattr_parent) {
2472 check_zp = zp->z_xattr_parent;
2473 rw_exit(&zp->z_xattr_lock);
2474
2475 /*
2476 * Verify a lookup yields the same znode.
2477 */
2478 ASSERT3S(sa_lookup(zp->z_sa_hdl, SA_ZPL_PARENT(
2479 ZTOZSB(zp)), &parent, sizeof (parent)), ==, 0);
2480 ASSERT3U(check_zp->z_id, ==, parent);
2481 } else {
2482 rw_exit(&zp->z_xattr_lock);
2483
2484 error = sa_lookup(zp->z_sa_hdl, SA_ZPL_PARENT(
2485 ZTOZSB(zp)), &parent, sizeof (parent));
2486 if (error)
2487 return (error);
2488
2489 /*
2490 * Cache the lookup on the parent file znode as
2491 * zp->z_xattr_parent and hold a reference. This
2492 * effectively pins the parent in memory until all
2493 * child xattr znodes have been destroyed and
2494 * release their references in zfs_inode_destroy().
2495 */
2496 error = zfs_zget(ZTOZSB(zp), parent, &check_zp);
2497 if (error)
2498 return (error);
2499
2500 rw_enter(&zp->z_xattr_lock, RW_WRITER);
2501 if (zp->z_xattr_parent == NULL)
2502 zp->z_xattr_parent = check_zp;
2503 rw_exit(&zp->z_xattr_lock);
2504 }
2505
2506 /*
2507 * fixup mode to map to xattr perms
2508 */
2509
2510 if (mode & (ACE_WRITE_DATA|ACE_APPEND_DATA)) {
2511 mode &= ~(ACE_WRITE_DATA|ACE_APPEND_DATA);
2512 mode |= ACE_WRITE_NAMED_ATTRS;
2513 }
2514
2515 if (mode & (ACE_READ_DATA|ACE_EXECUTE)) {
2516 mode &= ~(ACE_READ_DATA|ACE_EXECUTE);
2517 mode |= ACE_READ_NAMED_ATTRS;
2518 }
2519 }
2520
2521 owner = zfs_fuid_map_id(ZTOZSB(zp), zp->z_uid, cr, ZFS_OWNER);
2522 /*
2523 * Map the bits required to the standard inode flags
2524 * S_IRUSR|S_IWUSR|S_IXUSR in the needed_bits. Map the bits
2525 * mapped by working_mode (currently missing) in missing_bits.
2526 * Call secpolicy_vnode_access2() with (needed_bits & ~checkmode),
2527 * needed_bits.
2528 */
2529 needed_bits = 0;
2530
2531 working_mode = mode;
2532 if ((working_mode & (ACE_READ_ACL|ACE_READ_ATTRIBUTES)) &&
2533 owner == crgetuid(cr))
2534 working_mode &= ~(ACE_READ_ACL|ACE_READ_ATTRIBUTES);
2535
2536 if (working_mode & (ACE_READ_DATA|ACE_READ_NAMED_ATTRS|
2537 ACE_READ_ACL|ACE_READ_ATTRIBUTES|ACE_SYNCHRONIZE))
2538 needed_bits |= S_IRUSR;
2539 if (working_mode & (ACE_WRITE_DATA|ACE_WRITE_NAMED_ATTRS|
2540 ACE_APPEND_DATA|ACE_WRITE_ATTRIBUTES|ACE_SYNCHRONIZE))
2541 needed_bits |= S_IWUSR;
2542 if (working_mode & ACE_EXECUTE)
2543 needed_bits |= S_IXUSR;
2544
2545 if ((error = zfs_zaccess_common(check_zp, mode, &working_mode,
2546 &check_privs, skipaclchk, cr)) == 0) {
2547 return (secpolicy_vnode_access2(cr, ZTOI(zp), owner,
2548 needed_bits, needed_bits));
2549 }
2550
2551 if (error && !check_privs) {
2552 return (error);
2553 }
2554
2555 if (error && (flags & V_APPEND)) {
2556 error = zfs_zaccess_append(zp, &working_mode, &check_privs, cr);
2557 }
2558
2559 if (error && check_privs) {
2560 mode_t checkmode = 0;
2561
2562 /*
2563 * First check for implicit owner permission on
2564 * read_acl/read_attributes
2565 */
2566
2567 error = 0;
2568 ASSERT(working_mode != 0);
2569
2570 if ((working_mode & (ACE_READ_ACL|ACE_READ_ATTRIBUTES) &&
2571 owner == crgetuid(cr)))
2572 working_mode &= ~(ACE_READ_ACL|ACE_READ_ATTRIBUTES);
2573
2574 if (working_mode & (ACE_READ_DATA|ACE_READ_NAMED_ATTRS|
2575 ACE_READ_ACL|ACE_READ_ATTRIBUTES|ACE_SYNCHRONIZE))
2576 checkmode |= S_IRUSR;
2577 if (working_mode & (ACE_WRITE_DATA|ACE_WRITE_NAMED_ATTRS|
2578 ACE_APPEND_DATA|ACE_WRITE_ATTRIBUTES|ACE_SYNCHRONIZE))
2579 checkmode |= S_IWUSR;
2580 if (working_mode & ACE_EXECUTE)
2581 checkmode |= S_IXUSR;
2582
2583 error = secpolicy_vnode_access2(cr, ZTOI(check_zp), owner,
2584 needed_bits & ~checkmode, needed_bits);
2585
2586 if (error == 0 && (working_mode & ACE_WRITE_OWNER))
2587 error = secpolicy_vnode_chown(cr, owner);
2588 if (error == 0 && (working_mode & ACE_WRITE_ACL))
2589 error = secpolicy_vnode_setdac(cr, owner);
2590
2591 if (error == 0 && (working_mode &
2592 (ACE_DELETE|ACE_DELETE_CHILD)))
2593 error = secpolicy_vnode_remove(cr);
2594
2595 if (error == 0 && (working_mode & ACE_SYNCHRONIZE)) {
2596 error = secpolicy_vnode_chown(cr, owner);
2597 }
2598 if (error == 0) {
2599 /*
2600 * See if any bits other than those already checked
2601 * for are still present. If so then return EACCES
2602 */
2603 if (working_mode & ~(ZFS_CHECKED_MASKS)) {
2604 error = EACCES;
2605 }
2606 }
2607 } else if (error == 0) {
2608 error = secpolicy_vnode_access2(cr, ZTOI(zp), owner,
2609 needed_bits, needed_bits);
2610 }
2611
2612 return (error);
2613 }
2614
2615 /*
2616 * Translate traditional unix S_IRUSR/S_IWUSR/S_IXUSR mode into
2617 * native ACL format and call zfs_zaccess()
2618 */
2619 int
2620 zfs_zaccess_rwx(znode_t *zp, mode_t mode, int flags, cred_t *cr)
2621 {
2622 return (zfs_zaccess(zp, zfs_unix_to_v4(mode >> 6), flags, B_FALSE, cr));
2623 }
2624
2625 /*
2626 * Access function for secpolicy_vnode_setattr
2627 */
2628 int
2629 zfs_zaccess_unix(znode_t *zp, mode_t mode, cred_t *cr)
2630 {
2631 int v4_mode = zfs_unix_to_v4(mode >> 6);
2632
2633 return (zfs_zaccess(zp, v4_mode, 0, B_FALSE, cr));
2634 }
2635
2636 static int
2637 zfs_delete_final_check(znode_t *zp, znode_t *dzp,
2638 mode_t available_perms, cred_t *cr)
2639 {
2640 int error;
2641 uid_t downer;
2642
2643 downer = zfs_fuid_map_id(ZTOZSB(dzp), dzp->z_uid, cr, ZFS_OWNER);
2644
2645 error = secpolicy_vnode_access2(cr, ZTOI(dzp),
2646 downer, available_perms, S_IWUSR|S_IXUSR);
2647
2648 if (error == 0)
2649 error = zfs_sticky_remove_access(dzp, zp, cr);
2650
2651 return (error);
2652 }
2653
2654 /*
2655 * Determine whether Access should be granted/deny, without
2656 * consulting least priv subsystem.
2657 *
2658 *
2659 * The following chart is the recommended NFSv4 enforcement for
2660 * ability to delete an object.
2661 *
2662 * -------------------------------------------------------
2663 * | Parent Dir | Target Object Permissions |
2664 * | permissions | |
2665 * -------------------------------------------------------
2666 * | | ACL Allows | ACL Denies| Delete |
2667 * | | Delete | Delete | unspecified|
2668 * -------------------------------------------------------
2669 * | ACL Allows | Permit | Permit | Permit |
2670 * | DELETE_CHILD | |
2671 * -------------------------------------------------------
2672 * | ACL Denies | Permit | Deny | Deny |
2673 * | DELETE_CHILD | | | |
2674 * -------------------------------------------------------
2675 * | ACL specifies | | | |
2676 * | only allow | Permit | Permit | Permit |
2677 * | write and | | | |
2678 * | execute | | | |
2679 * -------------------------------------------------------
2680 * | ACL denies | | | |
2681 * | write and | Permit | Deny | Deny |
2682 * | execute | | | |
2683 * -------------------------------------------------------
2684 * ^
2685 * |
2686 * No search privilege, can't even look up file?
2687 *
2688 */
2689 int
2690 zfs_zaccess_delete(znode_t *dzp, znode_t *zp, cred_t *cr)
2691 {
2692 uint32_t dzp_working_mode = 0;
2693 uint32_t zp_working_mode = 0;
2694 int dzp_error, zp_error;
2695 mode_t available_perms;
2696 boolean_t dzpcheck_privs = B_TRUE;
2697 boolean_t zpcheck_privs = B_TRUE;
2698
2699 /*
2700 * We want specific DELETE permissions to
2701 * take precedence over WRITE/EXECUTE. We don't
2702 * want an ACL such as this to mess us up.
2703 * user:joe:write_data:deny,user:joe:delete:allow
2704 *
2705 * However, deny permissions may ultimately be overridden
2706 * by secpolicy_vnode_access().
2707 *
2708 * We will ask for all of the necessary permissions and then
2709 * look at the working modes from the directory and target object
2710 * to determine what was found.
2711 */
2712
2713 if (zp->z_pflags & (ZFS_IMMUTABLE | ZFS_NOUNLINK))
2714 return (EPERM);
2715
2716 /*
2717 * First row
2718 * If the directory permissions allow the delete, we are done.
2719 */
2720 if ((dzp_error = zfs_zaccess_common(dzp, ACE_DELETE_CHILD,
2721 &dzp_working_mode, &dzpcheck_privs, B_FALSE, cr)) == 0)
2722 return (0);
2723
2724 /*
2725 * If target object has delete permission then we are done
2726 */
2727 if ((zp_error = zfs_zaccess_common(zp, ACE_DELETE, &zp_working_mode,
2728 &zpcheck_privs, B_FALSE, cr)) == 0)
2729 return (0);
2730
2731 ASSERT(dzp_error && zp_error);
2732
2733 if (!dzpcheck_privs)
2734 return (dzp_error);
2735 if (!zpcheck_privs)
2736 return (zp_error);
2737
2738 /*
2739 * Second row
2740 *
2741 * If directory returns EACCES then delete_child was denied
2742 * due to deny delete_child. In this case send the request through
2743 * secpolicy_vnode_remove(). We don't use zfs_delete_final_check()
2744 * since that *could* allow the delete based on write/execute permission
2745 * and we want delete permissions to override write/execute.
2746 */
2747
2748 if (dzp_error == EACCES)
2749 return (secpolicy_vnode_remove(cr));
2750
2751 /*
2752 * Third Row
2753 * only need to see if we have write/execute on directory.
2754 */
2755
2756 dzp_error = zfs_zaccess_common(dzp, ACE_EXECUTE|ACE_WRITE_DATA,
2757 &dzp_working_mode, &dzpcheck_privs, B_FALSE, cr);
2758
2759 if (dzp_error != 0 && !dzpcheck_privs)
2760 return (dzp_error);
2761
2762 /*
2763 * Fourth row
2764 */
2765
2766 available_perms = (dzp_working_mode & ACE_WRITE_DATA) ? 0 : S_IWUSR;
2767 available_perms |= (dzp_working_mode & ACE_EXECUTE) ? 0 : S_IXUSR;
2768
2769 return (zfs_delete_final_check(zp, dzp, available_perms, cr));
2770
2771 }
2772
2773 int
2774 zfs_zaccess_rename(znode_t *sdzp, znode_t *szp, znode_t *tdzp,
2775 znode_t *tzp, cred_t *cr)
2776 {
2777 int add_perm;
2778 int error;
2779
2780 if (szp->z_pflags & ZFS_AV_QUARANTINED)
2781 return (EACCES);
2782
2783 add_perm = S_ISDIR(ZTOI(szp)->i_mode) ?
2784 ACE_ADD_SUBDIRECTORY : ACE_ADD_FILE;
2785
2786 /*
2787 * Rename permissions are combination of delete permission +
2788 * add file/subdir permission.
2789 */
2790
2791 /*
2792 * first make sure we do the delete portion.
2793 *
2794 * If that succeeds then check for add_file/add_subdir permissions
2795 */
2796
2797 if ((error = zfs_zaccess_delete(sdzp, szp, cr)))
2798 return (error);
2799
2800 /*
2801 * If we have a tzp, see if we can delete it?
2802 */
2803 if (tzp) {
2804 if ((error = zfs_zaccess_delete(tdzp, tzp, cr)))
2805 return (error);
2806 }
2807
2808 /*
2809 * Now check for add permissions
2810 */
2811 error = zfs_zaccess(tdzp, add_perm, 0, B_FALSE, cr);
2812
2813 return (error);
2814 }
ZFS On Linux mirror for PVE