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.
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.
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]
22 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
23 * Copyright 2011 Nexenta Systems, Inc. All rights reserved.
24 * Copyright (c) 2013 by Delphix. All rights reserved.
27 #include <sys/types.h>
28 #include <sys/param.h>
30 #include <sys/systm.h>
31 #include <sys/sysmacros.h>
32 #include <sys/resource.h>
34 #include <sys/vnode.h>
38 #include <sys/cmn_err.h>
39 #include <sys/errno.h>
40 #include <sys/unistd.h>
42 #include <sys/fs/zfs.h>
43 #include <sys/policy.h>
44 #include <sys/zfs_znode.h>
45 #include <sys/zfs_fuid.h>
46 #include <sys/zfs_acl.h>
47 #include <sys/zfs_dir.h>
48 #include <sys/zfs_vfsops.h>
50 #include <sys/dnode.h>
53 #include <acl/acl_common.h>
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
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)
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)
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)
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)
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)
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)
87 #define RESTRICTED_CLEAR (ACE_WRITE_ACL|ACE_WRITE_OWNER)
89 #define V4_ACL_WIDE_FLAGS (ZFS_ACL_AUTO_INHERIT|ZFS_ACL_DEFAULTED|\
92 #define ZFS_ACL_WIDE_FLAGS (V4_ACL_WIDE_FLAGS|ZFS_ACL_TRIVIAL|ZFS_INHERIT_ACE|\
95 #define ALL_MODE_EXECS (S_IXUSR | S_IXGRP | S_IXOTH)
98 zfs_ace_v0_get_type(void *acep)
100 return (((zfs_oldace_t *)acep)->z_type);
104 zfs_ace_v0_get_flags(void *acep)
106 return (((zfs_oldace_t *)acep)->z_flags);
110 zfs_ace_v0_get_mask(void *acep)
112 return (((zfs_oldace_t *)acep)->z_access_mask);
116 zfs_ace_v0_get_who(void *acep)
118 return (((zfs_oldace_t *)acep)->z_fuid);
122 zfs_ace_v0_set_type(void *acep, uint16_t type)
124 ((zfs_oldace_t *)acep)->z_type = type;
128 zfs_ace_v0_set_flags(void *acep, uint16_t flags)
130 ((zfs_oldace_t *)acep)->z_flags = flags;
134 zfs_ace_v0_set_mask(void *acep, uint32_t mask)
136 ((zfs_oldace_t *)acep)->z_access_mask = mask;
140 zfs_ace_v0_set_who(void *acep, uint64_t who)
142 ((zfs_oldace_t *)acep)->z_fuid = who;
147 zfs_ace_v0_size(void *acep)
149 return (sizeof (zfs_oldace_t));
153 zfs_ace_v0_abstract_size(void)
155 return (sizeof (zfs_oldace_t));
159 zfs_ace_v0_mask_off(void)
161 return (offsetof(zfs_oldace_t, z_access_mask));
166 zfs_ace_v0_data(void *acep, void **datap)
172 static acl_ops_t zfs_acl_v0_ops = {
175 zfs_ace_v0_get_flags,
176 zfs_ace_v0_set_flags,
182 zfs_ace_v0_abstract_size,
188 zfs_ace_fuid_get_type(void *acep)
190 return (((zfs_ace_hdr_t *)acep)->z_type);
194 zfs_ace_fuid_get_flags(void *acep)
196 return (((zfs_ace_hdr_t *)acep)->z_flags);
200 zfs_ace_fuid_get_mask(void *acep)
202 return (((zfs_ace_hdr_t *)acep)->z_access_mask);
206 zfs_ace_fuid_get_who(void *args)
209 zfs_ace_t *acep = args;
211 entry_type = acep->z_hdr.z_flags & ACE_TYPE_FLAGS;
213 if (entry_type == ACE_OWNER || entry_type == OWNING_GROUP ||
214 entry_type == ACE_EVERYONE)
216 return (((zfs_ace_t *)acep)->z_fuid);
220 zfs_ace_fuid_set_type(void *acep, uint16_t type)
222 ((zfs_ace_hdr_t *)acep)->z_type = type;
226 zfs_ace_fuid_set_flags(void *acep, uint16_t flags)
228 ((zfs_ace_hdr_t *)acep)->z_flags = flags;
232 zfs_ace_fuid_set_mask(void *acep, uint32_t mask)
234 ((zfs_ace_hdr_t *)acep)->z_access_mask = mask;
238 zfs_ace_fuid_set_who(void *arg, uint64_t who)
240 zfs_ace_t *acep = arg;
242 uint16_t entry_type = acep->z_hdr.z_flags & ACE_TYPE_FLAGS;
244 if (entry_type == ACE_OWNER || entry_type == OWNING_GROUP ||
245 entry_type == ACE_EVERYONE)
251 zfs_ace_fuid_size(void *acep)
253 zfs_ace_hdr_t *zacep = acep;
256 switch (zacep->z_type) {
257 case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE:
258 case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE:
259 case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE:
260 case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE:
261 return (sizeof (zfs_object_ace_t));
265 (((zfs_ace_hdr_t *)acep)->z_flags & ACE_TYPE_FLAGS);
266 if (entry_type == ACE_OWNER ||
267 entry_type == OWNING_GROUP ||
268 entry_type == ACE_EVERYONE)
269 return (sizeof (zfs_ace_hdr_t));
272 return (sizeof (zfs_ace_t));
277 zfs_ace_fuid_abstract_size(void)
279 return (sizeof (zfs_ace_hdr_t));
283 zfs_ace_fuid_mask_off(void)
285 return (offsetof(zfs_ace_hdr_t, z_access_mask));
289 zfs_ace_fuid_data(void *acep, void **datap)
291 zfs_ace_t *zacep = acep;
292 zfs_object_ace_t *zobjp;
294 switch (zacep->z_hdr.z_type) {
295 case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE:
296 case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE:
297 case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE:
298 case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE:
300 *datap = (caddr_t)zobjp + sizeof (zfs_ace_t);
301 return (sizeof (zfs_object_ace_t) - sizeof (zfs_ace_t));
308 static acl_ops_t zfs_acl_fuid_ops = {
309 zfs_ace_fuid_get_mask,
310 zfs_ace_fuid_set_mask,
311 zfs_ace_fuid_get_flags,
312 zfs_ace_fuid_set_flags,
313 zfs_ace_fuid_get_type,
314 zfs_ace_fuid_set_type,
315 zfs_ace_fuid_get_who,
316 zfs_ace_fuid_set_who,
318 zfs_ace_fuid_abstract_size,
319 zfs_ace_fuid_mask_off,
324 * The following three functions are provided for compatibility with
325 * older ZPL version in order to determine if the file use to have
326 * an external ACL and what version of ACL previously existed on the
327 * file. Would really be nice to not need this, sigh.
330 zfs_external_acl(znode_t *zp)
332 zfs_acl_phys_t acl_phys;
339 * Need to deal with a potential
340 * race where zfs_sa_upgrade could cause
341 * z_isa_sa to change.
343 * If the lookup fails then the state of z_is_sa should have
347 if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_ZNODE_ACL(zp->z_zfsvfs),
348 &acl_phys, sizeof (acl_phys))) == 0)
349 return (acl_phys.z_acl_extern_obj);
352 * after upgrade the SA_ZPL_ZNODE_ACL should have been
355 VERIFY(zp->z_is_sa && error == ENOENT);
361 * Determine size of ACL in bytes
363 * This is more complicated than it should be since we have to deal
364 * with old external ACLs.
367 zfs_acl_znode_info(znode_t *zp, int *aclsize, int *aclcount,
368 zfs_acl_phys_t *aclphys)
370 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
375 ASSERT(MUTEX_HELD(&zp->z_acl_lock));
377 if ((error = sa_size(zp->z_sa_hdl, SA_ZPL_DACL_ACES(zfsvfs),
381 if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_DACL_COUNT(zfsvfs),
382 &acl_count, sizeof (acl_count))) != 0)
384 *aclcount = acl_count;
386 if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_ZNODE_ACL(zfsvfs),
387 aclphys, sizeof (*aclphys))) != 0)
390 if (aclphys->z_acl_version == ZFS_ACL_VERSION_INITIAL) {
391 *aclsize = ZFS_ACL_SIZE(aclphys->z_acl_size);
392 *aclcount = aclphys->z_acl_size;
394 *aclsize = aclphys->z_acl_size;
395 *aclcount = aclphys->z_acl_count;
402 zfs_znode_acl_version(znode_t *zp)
404 zfs_acl_phys_t acl_phys;
407 return (ZFS_ACL_VERSION_FUID);
412 * Need to deal with a potential
413 * race where zfs_sa_upgrade could cause
414 * z_isa_sa to change.
416 * If the lookup fails then the state of z_is_sa should have
419 if ((error = sa_lookup(zp->z_sa_hdl,
420 SA_ZPL_ZNODE_ACL(zp->z_zfsvfs),
421 &acl_phys, sizeof (acl_phys))) == 0)
422 return (acl_phys.z_acl_version);
425 * After upgrade SA_ZPL_ZNODE_ACL should have
428 VERIFY(zp->z_is_sa && error == ENOENT);
429 return (ZFS_ACL_VERSION_FUID);
435 zfs_acl_version(int version)
437 if (version < ZPL_VERSION_FUID)
438 return (ZFS_ACL_VERSION_INITIAL);
440 return (ZFS_ACL_VERSION_FUID);
444 zfs_acl_version_zp(znode_t *zp)
446 return (zfs_acl_version(zp->z_zfsvfs->z_version));
450 zfs_acl_alloc(int vers)
454 aclp = kmem_zalloc(sizeof (zfs_acl_t), KM_SLEEP);
455 list_create(&aclp->z_acl, sizeof (zfs_acl_node_t),
456 offsetof(zfs_acl_node_t, z_next));
457 aclp->z_version = vers;
458 if (vers == ZFS_ACL_VERSION_FUID)
459 aclp->z_ops = zfs_acl_fuid_ops;
461 aclp->z_ops = zfs_acl_v0_ops;
466 zfs_acl_node_alloc(size_t bytes)
468 zfs_acl_node_t *aclnode;
470 aclnode = kmem_zalloc(sizeof (zfs_acl_node_t), KM_SLEEP);
472 aclnode->z_acldata = kmem_alloc(bytes, KM_SLEEP);
473 aclnode->z_allocdata = aclnode->z_acldata;
474 aclnode->z_allocsize = bytes;
475 aclnode->z_size = bytes;
482 zfs_acl_node_free(zfs_acl_node_t *aclnode)
484 if (aclnode->z_allocsize)
485 kmem_free(aclnode->z_allocdata, aclnode->z_allocsize);
486 kmem_free(aclnode, sizeof (zfs_acl_node_t));
490 zfs_acl_release_nodes(zfs_acl_t *aclp)
492 zfs_acl_node_t *aclnode;
494 while (aclnode = list_head(&aclp->z_acl)) {
495 list_remove(&aclp->z_acl, aclnode);
496 zfs_acl_node_free(aclnode);
498 aclp->z_acl_count = 0;
499 aclp->z_acl_bytes = 0;
503 zfs_acl_free(zfs_acl_t *aclp)
505 zfs_acl_release_nodes(aclp);
506 list_destroy(&aclp->z_acl);
507 kmem_free(aclp, sizeof (zfs_acl_t));
511 zfs_acl_valid_ace_type(uint_t type, uint_t flags)
518 case ACE_SYSTEM_AUDIT_ACE_TYPE:
519 case ACE_SYSTEM_ALARM_ACE_TYPE:
520 entry_type = flags & ACE_TYPE_FLAGS;
521 return (entry_type == ACE_OWNER ||
522 entry_type == OWNING_GROUP ||
523 entry_type == ACE_EVERYONE || entry_type == 0 ||
524 entry_type == ACE_IDENTIFIER_GROUP);
526 if (type >= MIN_ACE_TYPE && type <= MAX_ACE_TYPE)
533 zfs_ace_valid(vtype_t obj_type, zfs_acl_t *aclp, uint16_t type, uint16_t iflags)
536 * first check type of entry
539 if (!zfs_acl_valid_ace_type(type, iflags))
543 case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE:
544 case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE:
545 case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE:
546 case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE:
547 if (aclp->z_version < ZFS_ACL_VERSION_FUID)
549 aclp->z_hints |= ZFS_ACL_OBJ_ACE;
553 * next check inheritance level flags
556 if (obj_type == VDIR &&
557 (iflags & (ACE_FILE_INHERIT_ACE|ACE_DIRECTORY_INHERIT_ACE)))
558 aclp->z_hints |= ZFS_INHERIT_ACE;
560 if (iflags & (ACE_INHERIT_ONLY_ACE|ACE_NO_PROPAGATE_INHERIT_ACE)) {
561 if ((iflags & (ACE_FILE_INHERIT_ACE|
562 ACE_DIRECTORY_INHERIT_ACE)) == 0) {
571 zfs_acl_next_ace(zfs_acl_t *aclp, void *start, uint64_t *who,
572 uint32_t *access_mask, uint16_t *iflags, uint16_t *type)
574 zfs_acl_node_t *aclnode;
579 aclnode = list_head(&aclp->z_acl);
583 aclp->z_next_ace = aclnode->z_acldata;
584 aclp->z_curr_node = aclnode;
585 aclnode->z_ace_idx = 0;
588 aclnode = aclp->z_curr_node;
593 if (aclnode->z_ace_idx >= aclnode->z_ace_count) {
594 aclnode = list_next(&aclp->z_acl, aclnode);
598 aclp->z_curr_node = aclnode;
599 aclnode->z_ace_idx = 0;
600 aclp->z_next_ace = aclnode->z_acldata;
604 if (aclnode->z_ace_idx < aclnode->z_ace_count) {
605 void *acep = aclp->z_next_ace;
609 * Make sure we don't overstep our bounds
611 ace_size = aclp->z_ops.ace_size(acep);
613 if (((caddr_t)acep + ace_size) >
614 ((caddr_t)aclnode->z_acldata + aclnode->z_size)) {
618 *iflags = aclp->z_ops.ace_flags_get(acep);
619 *type = aclp->z_ops.ace_type_get(acep);
620 *access_mask = aclp->z_ops.ace_mask_get(acep);
621 *who = aclp->z_ops.ace_who_get(acep);
622 aclp->z_next_ace = (caddr_t)aclp->z_next_ace + ace_size;
623 aclnode->z_ace_idx++;
625 return ((void *)acep);
632 zfs_ace_walk(void *datap, uint64_t cookie, int aclcnt,
633 uint16_t *flags, uint16_t *type, uint32_t *mask)
635 zfs_acl_t *aclp = datap;
636 zfs_ace_hdr_t *acep = (zfs_ace_hdr_t *)(uintptr_t)cookie;
639 acep = zfs_acl_next_ace(aclp, acep, &who, mask,
641 return ((uint64_t)(uintptr_t)acep);
644 static zfs_acl_node_t *
645 zfs_acl_curr_node(zfs_acl_t *aclp)
647 ASSERT(aclp->z_curr_node);
648 return (aclp->z_curr_node);
652 * Copy ACE to internal ZFS format.
653 * While processing the ACL each ACE will be validated for correctness.
654 * ACE FUIDs will be created later.
657 zfs_copy_ace_2_fuid(zfsvfs_t *zfsvfs, vtype_t obj_type, zfs_acl_t *aclp,
658 void *datap, zfs_ace_t *z_acl, uint64_t aclcnt, size_t *size,
659 zfs_fuid_info_t **fuidp, cred_t *cr)
663 zfs_ace_t *aceptr = z_acl;
665 zfs_object_ace_t *zobjacep;
666 ace_object_t *aceobjp;
668 for (i = 0; i != aclcnt; i++) {
669 aceptr->z_hdr.z_access_mask = acep->a_access_mask;
670 aceptr->z_hdr.z_flags = acep->a_flags;
671 aceptr->z_hdr.z_type = acep->a_type;
672 entry_type = aceptr->z_hdr.z_flags & ACE_TYPE_FLAGS;
673 if (entry_type != ACE_OWNER && entry_type != OWNING_GROUP &&
674 entry_type != ACE_EVERYONE) {
675 aceptr->z_fuid = zfs_fuid_create(zfsvfs, acep->a_who,
676 cr, (entry_type == 0) ?
677 ZFS_ACE_USER : ZFS_ACE_GROUP, fuidp);
681 * Make sure ACE is valid
683 if (zfs_ace_valid(obj_type, aclp, aceptr->z_hdr.z_type,
684 aceptr->z_hdr.z_flags) != B_TRUE)
685 return (SET_ERROR(EINVAL));
687 switch (acep->a_type) {
688 case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE:
689 case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE:
690 case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE:
691 case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE:
692 zobjacep = (zfs_object_ace_t *)aceptr;
693 aceobjp = (ace_object_t *)acep;
695 bcopy(aceobjp->a_obj_type, zobjacep->z_object_type,
696 sizeof (aceobjp->a_obj_type));
697 bcopy(aceobjp->a_inherit_obj_type,
698 zobjacep->z_inherit_type,
699 sizeof (aceobjp->a_inherit_obj_type));
700 acep = (ace_t *)((caddr_t)acep + sizeof (ace_object_t));
703 acep = (ace_t *)((caddr_t)acep + sizeof (ace_t));
706 aceptr = (zfs_ace_t *)((caddr_t)aceptr +
707 aclp->z_ops.ace_size(aceptr));
710 *size = (caddr_t)aceptr - (caddr_t)z_acl;
716 * Copy ZFS ACEs to fixed size ace_t layout
719 zfs_copy_fuid_2_ace(zfsvfs_t *zfsvfs, zfs_acl_t *aclp, cred_t *cr,
720 void *datap, int filter)
723 uint32_t access_mask;
724 uint16_t iflags, type;
725 zfs_ace_hdr_t *zacep = NULL;
727 ace_object_t *objacep;
728 zfs_object_ace_t *zobjacep;
732 while (zacep = zfs_acl_next_ace(aclp, zacep,
733 &who, &access_mask, &iflags, &type)) {
736 case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE:
737 case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE:
738 case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE:
739 case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE:
743 zobjacep = (zfs_object_ace_t *)zacep;
744 objacep = (ace_object_t *)acep;
745 bcopy(zobjacep->z_object_type,
747 sizeof (zobjacep->z_object_type));
748 bcopy(zobjacep->z_inherit_type,
749 objacep->a_inherit_obj_type,
750 sizeof (zobjacep->z_inherit_type));
751 ace_size = sizeof (ace_object_t);
754 ace_size = sizeof (ace_t);
758 entry_type = (iflags & ACE_TYPE_FLAGS);
759 if ((entry_type != ACE_OWNER &&
760 entry_type != OWNING_GROUP &&
761 entry_type != ACE_EVERYONE)) {
762 acep->a_who = zfs_fuid_map_id(zfsvfs, who,
763 cr, (entry_type & ACE_IDENTIFIER_GROUP) ?
764 ZFS_ACE_GROUP : ZFS_ACE_USER);
766 acep->a_who = (uid_t)(int64_t)who;
768 acep->a_access_mask = access_mask;
769 acep->a_flags = iflags;
771 acep = (ace_t *)((caddr_t)acep + ace_size);
776 zfs_copy_ace_2_oldace(vtype_t obj_type, zfs_acl_t *aclp, ace_t *acep,
777 zfs_oldace_t *z_acl, int aclcnt, size_t *size)
780 zfs_oldace_t *aceptr = z_acl;
782 for (i = 0; i != aclcnt; i++, aceptr++) {
783 aceptr->z_access_mask = acep[i].a_access_mask;
784 aceptr->z_type = acep[i].a_type;
785 aceptr->z_flags = acep[i].a_flags;
786 aceptr->z_fuid = acep[i].a_who;
788 * Make sure ACE is valid
790 if (zfs_ace_valid(obj_type, aclp, aceptr->z_type,
791 aceptr->z_flags) != B_TRUE)
792 return (SET_ERROR(EINVAL));
794 *size = (caddr_t)aceptr - (caddr_t)z_acl;
799 * convert old ACL format to new
802 zfs_acl_xform(znode_t *zp, zfs_acl_t *aclp, cred_t *cr)
804 zfs_oldace_t *oldaclp;
806 uint16_t type, iflags;
807 uint32_t access_mask;
810 zfs_acl_node_t *newaclnode;
812 ASSERT(aclp->z_version == ZFS_ACL_VERSION_INITIAL);
814 * First create the ACE in a contiguous piece of memory
815 * for zfs_copy_ace_2_fuid().
817 * We only convert an ACL once, so this won't happen
820 oldaclp = kmem_alloc(sizeof (zfs_oldace_t) * aclp->z_acl_count,
823 while (cookie = zfs_acl_next_ace(aclp, cookie, &who,
824 &access_mask, &iflags, &type)) {
825 oldaclp[i].z_flags = iflags;
826 oldaclp[i].z_type = type;
827 oldaclp[i].z_fuid = who;
828 oldaclp[i++].z_access_mask = access_mask;
831 newaclnode = zfs_acl_node_alloc(aclp->z_acl_count *
832 sizeof (zfs_object_ace_t));
833 aclp->z_ops = zfs_acl_fuid_ops;
834 VERIFY(zfs_copy_ace_2_fuid(zp->z_zfsvfs, ZTOV(zp)->v_type, aclp,
835 oldaclp, newaclnode->z_acldata, aclp->z_acl_count,
836 &newaclnode->z_size, NULL, cr) == 0);
837 newaclnode->z_ace_count = aclp->z_acl_count;
838 aclp->z_version = ZFS_ACL_VERSION;
839 kmem_free(oldaclp, aclp->z_acl_count * sizeof (zfs_oldace_t));
842 * Release all previous ACL nodes
845 zfs_acl_release_nodes(aclp);
847 list_insert_head(&aclp->z_acl, newaclnode);
849 aclp->z_acl_bytes = newaclnode->z_size;
850 aclp->z_acl_count = newaclnode->z_ace_count;
855 * Convert unix access mask to v4 access mask
858 zfs_unix_to_v4(uint32_t access_mask)
860 uint32_t new_mask = 0;
862 if (access_mask & S_IXOTH)
863 new_mask |= ACE_EXECUTE;
864 if (access_mask & S_IWOTH)
865 new_mask |= ACE_WRITE_DATA;
866 if (access_mask & S_IROTH)
867 new_mask |= ACE_READ_DATA;
872 zfs_set_ace(zfs_acl_t *aclp, void *acep, uint32_t access_mask,
873 uint16_t access_type, uint64_t fuid, uint16_t entry_type)
875 uint16_t type = entry_type & ACE_TYPE_FLAGS;
877 aclp->z_ops.ace_mask_set(acep, access_mask);
878 aclp->z_ops.ace_type_set(acep, access_type);
879 aclp->z_ops.ace_flags_set(acep, entry_type);
880 if ((type != ACE_OWNER && type != OWNING_GROUP &&
881 type != ACE_EVERYONE))
882 aclp->z_ops.ace_who_set(acep, fuid);
886 * Determine mode of file based on ACL.
889 zfs_mode_compute(uint64_t fmode, zfs_acl_t *aclp,
890 uint64_t *pflags, uint64_t fuid, uint64_t fgid)
895 zfs_ace_hdr_t *acep = NULL;
897 uint16_t iflags, type;
898 uint32_t access_mask;
899 boolean_t an_exec_denied = B_FALSE;
901 mode = (fmode & (S_IFMT | S_ISUID | S_ISGID | S_ISVTX));
903 while (acep = zfs_acl_next_ace(aclp, acep, &who,
904 &access_mask, &iflags, &type)) {
906 if (!zfs_acl_valid_ace_type(type, iflags))
909 entry_type = (iflags & ACE_TYPE_FLAGS);
912 * Skip over any inherit_only ACEs
914 if (iflags & ACE_INHERIT_ONLY_ACE)
917 if (entry_type == ACE_OWNER || (entry_type == 0 &&
919 if ((access_mask & ACE_READ_DATA) &&
920 (!(seen & S_IRUSR))) {
926 if ((access_mask & ACE_WRITE_DATA) &&
927 (!(seen & S_IWUSR))) {
933 if ((access_mask & ACE_EXECUTE) &&
934 (!(seen & S_IXUSR))) {
940 } else if (entry_type == OWNING_GROUP ||
941 (entry_type == ACE_IDENTIFIER_GROUP && who == fgid)) {
942 if ((access_mask & ACE_READ_DATA) &&
943 (!(seen & S_IRGRP))) {
949 if ((access_mask & ACE_WRITE_DATA) &&
950 (!(seen & S_IWGRP))) {
956 if ((access_mask & ACE_EXECUTE) &&
957 (!(seen & S_IXGRP))) {
963 } else if (entry_type == ACE_EVERYONE) {
964 if ((access_mask & ACE_READ_DATA)) {
965 if (!(seen & S_IRUSR)) {
971 if (!(seen & S_IRGRP)) {
977 if (!(seen & S_IROTH)) {
984 if ((access_mask & ACE_WRITE_DATA)) {
985 if (!(seen & S_IWUSR)) {
991 if (!(seen & S_IWGRP)) {
997 if (!(seen & S_IWOTH)) {
1004 if ((access_mask & ACE_EXECUTE)) {
1005 if (!(seen & S_IXUSR)) {
1007 if (type == ALLOW) {
1011 if (!(seen & S_IXGRP)) {
1013 if (type == ALLOW) {
1017 if (!(seen & S_IXOTH)) {
1019 if (type == ALLOW) {
1026 * Only care if this IDENTIFIER_GROUP or
1027 * USER ACE denies execute access to someone,
1028 * mode is not affected
1030 if ((access_mask & ACE_EXECUTE) && type == DENY)
1031 an_exec_denied = B_TRUE;
1036 * Failure to allow is effectively a deny, so execute permission
1037 * is denied if it was never mentioned or if we explicitly
1038 * weren't allowed it.
1040 if (!an_exec_denied &&
1041 ((seen & ALL_MODE_EXECS) != ALL_MODE_EXECS ||
1042 (mode & ALL_MODE_EXECS) != ALL_MODE_EXECS))
1043 an_exec_denied = B_TRUE;
1046 *pflags &= ~ZFS_NO_EXECS_DENIED;
1048 *pflags |= ZFS_NO_EXECS_DENIED;
1054 * Read an external acl object. If the intent is to modify, always
1055 * create a new acl and leave any cached acl in place.
1058 zfs_acl_node_read(znode_t *zp, zfs_acl_t **aclpp, boolean_t will_modify)
1063 zfs_acl_node_t *aclnode;
1064 zfs_acl_phys_t znode_acl;
1068 ASSERT(MUTEX_HELD(&zp->z_acl_lock));
1069 ASSERT_VOP_LOCKED(ZTOV(zp), __func__);
1071 if (zp->z_acl_cached && !will_modify) {
1072 *aclpp = zp->z_acl_cached;
1076 version = zfs_znode_acl_version(zp);
1078 if ((error = zfs_acl_znode_info(zp, &aclsize,
1079 &acl_count, &znode_acl)) != 0) {
1083 aclp = zfs_acl_alloc(version);
1085 aclp->z_acl_count = acl_count;
1086 aclp->z_acl_bytes = aclsize;
1088 aclnode = zfs_acl_node_alloc(aclsize);
1089 aclnode->z_ace_count = aclp->z_acl_count;
1090 aclnode->z_size = aclsize;
1093 if (znode_acl.z_acl_extern_obj) {
1094 error = dmu_read(zp->z_zfsvfs->z_os,
1095 znode_acl.z_acl_extern_obj, 0, aclnode->z_size,
1096 aclnode->z_acldata, DMU_READ_PREFETCH);
1098 bcopy(znode_acl.z_ace_data, aclnode->z_acldata,
1102 error = sa_lookup(zp->z_sa_hdl, SA_ZPL_DACL_ACES(zp->z_zfsvfs),
1103 aclnode->z_acldata, aclnode->z_size);
1108 zfs_acl_node_free(aclnode);
1109 /* convert checksum errors into IO errors */
1110 if (error == ECKSUM)
1111 error = SET_ERROR(EIO);
1115 list_insert_head(&aclp->z_acl, aclnode);
1119 zp->z_acl_cached = aclp;
1126 zfs_acl_data_locator(void **dataptr, uint32_t *length, uint32_t buflen,
1127 boolean_t start, void *userdata)
1129 zfs_acl_locator_cb_t *cb = (zfs_acl_locator_cb_t *)userdata;
1132 cb->cb_acl_node = list_head(&cb->cb_aclp->z_acl);
1134 cb->cb_acl_node = list_next(&cb->cb_aclp->z_acl,
1137 *dataptr = cb->cb_acl_node->z_acldata;
1138 *length = cb->cb_acl_node->z_size;
1142 zfs_acl_chown_setattr(znode_t *zp)
1147 ASSERT_VOP_ELOCKED(ZTOV(zp), __func__);
1148 ASSERT(MUTEX_HELD(&zp->z_acl_lock));
1150 if ((error = zfs_acl_node_read(zp, &aclp, B_FALSE)) == 0)
1151 zp->z_mode = zfs_mode_compute(zp->z_mode, aclp,
1152 &zp->z_pflags, zp->z_uid, zp->z_gid);
1157 * common code for setting ACLs.
1159 * This function is called from zfs_mode_update, zfs_perm_init, and zfs_setacl.
1160 * zfs_setacl passes a non-NULL inherit pointer (ihp) to indicate that it's
1161 * already checked the acl and knows whether to inherit.
1164 zfs_aclset_common(znode_t *zp, zfs_acl_t *aclp, cred_t *cr, dmu_tx_t *tx)
1167 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1168 dmu_object_type_t otype;
1169 zfs_acl_locator_cb_t locate = { 0 };
1171 sa_bulk_attr_t bulk[5];
1177 mode = zfs_mode_compute(mode, aclp, &zp->z_pflags,
1178 zp->z_uid, zp->z_gid);
1181 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs), NULL,
1182 &mode, sizeof (mode));
1183 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL,
1184 &zp->z_pflags, sizeof (zp->z_pflags));
1185 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL,
1186 &ctime, sizeof (ctime));
1188 if (zp->z_acl_cached) {
1189 zfs_acl_free(zp->z_acl_cached);
1190 zp->z_acl_cached = NULL;
1196 if (!zfsvfs->z_use_fuids) {
1197 otype = DMU_OT_OLDACL;
1199 if ((aclp->z_version == ZFS_ACL_VERSION_INITIAL) &&
1200 (zfsvfs->z_version >= ZPL_VERSION_FUID))
1201 zfs_acl_xform(zp, aclp, cr);
1202 ASSERT(aclp->z_version >= ZFS_ACL_VERSION_FUID);
1207 * Arrgh, we have to handle old on disk format
1208 * as well as newer (preferred) SA format.
1211 if (zp->z_is_sa) { /* the easy case, just update the ACL attribute */
1212 locate.cb_aclp = aclp;
1213 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_DACL_ACES(zfsvfs),
1214 zfs_acl_data_locator, &locate, aclp->z_acl_bytes);
1215 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_DACL_COUNT(zfsvfs),
1216 NULL, &aclp->z_acl_count, sizeof (uint64_t));
1217 } else { /* Painful legacy way */
1218 zfs_acl_node_t *aclnode;
1220 zfs_acl_phys_t acl_phys;
1223 if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_ZNODE_ACL(zfsvfs),
1224 &acl_phys, sizeof (acl_phys))) != 0)
1227 aoid = acl_phys.z_acl_extern_obj;
1229 if (aclp->z_acl_bytes > ZFS_ACE_SPACE) {
1231 * If ACL was previously external and we are now
1232 * converting to new ACL format then release old
1233 * ACL object and create a new one.
1236 aclp->z_version != acl_phys.z_acl_version) {
1237 error = dmu_object_free(zfsvfs->z_os, aoid, tx);
1243 aoid = dmu_object_alloc(zfsvfs->z_os,
1244 otype, aclp->z_acl_bytes,
1245 otype == DMU_OT_ACL ?
1246 DMU_OT_SYSACL : DMU_OT_NONE,
1247 otype == DMU_OT_ACL ?
1248 DN_MAX_BONUSLEN : 0, tx);
1250 (void) dmu_object_set_blocksize(zfsvfs->z_os,
1251 aoid, aclp->z_acl_bytes, 0, tx);
1253 acl_phys.z_acl_extern_obj = aoid;
1254 for (aclnode = list_head(&aclp->z_acl); aclnode;
1255 aclnode = list_next(&aclp->z_acl, aclnode)) {
1256 if (aclnode->z_ace_count == 0)
1258 dmu_write(zfsvfs->z_os, aoid, off,
1259 aclnode->z_size, aclnode->z_acldata, tx);
1260 off += aclnode->z_size;
1263 void *start = acl_phys.z_ace_data;
1265 * Migrating back embedded?
1267 if (acl_phys.z_acl_extern_obj) {
1268 error = dmu_object_free(zfsvfs->z_os,
1269 acl_phys.z_acl_extern_obj, tx);
1272 acl_phys.z_acl_extern_obj = 0;
1275 for (aclnode = list_head(&aclp->z_acl); aclnode;
1276 aclnode = list_next(&aclp->z_acl, aclnode)) {
1277 if (aclnode->z_ace_count == 0)
1279 bcopy(aclnode->z_acldata, start,
1281 start = (caddr_t)start + aclnode->z_size;
1285 * If Old version then swap count/bytes to match old
1286 * layout of znode_acl_phys_t.
1288 if (aclp->z_version == ZFS_ACL_VERSION_INITIAL) {
1289 acl_phys.z_acl_size = aclp->z_acl_count;
1290 acl_phys.z_acl_count = aclp->z_acl_bytes;
1292 acl_phys.z_acl_size = aclp->z_acl_bytes;
1293 acl_phys.z_acl_count = aclp->z_acl_count;
1295 acl_phys.z_acl_version = aclp->z_version;
1297 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ZNODE_ACL(zfsvfs), NULL,
1298 &acl_phys, sizeof (acl_phys));
1302 * Replace ACL wide bits, but first clear them.
1304 zp->z_pflags &= ~ZFS_ACL_WIDE_FLAGS;
1306 zp->z_pflags |= aclp->z_hints;
1308 if (ace_trivial_common(aclp, 0, zfs_ace_walk) == 0)
1309 zp->z_pflags |= ZFS_ACL_TRIVIAL;
1311 zfs_tstamp_update_setup(zp, STATE_CHANGED, NULL, ctime, B_TRUE);
1312 return (sa_bulk_update(zp->z_sa_hdl, bulk, count, tx));
1316 zfs_acl_chmod(vtype_t vtype, uint64_t mode, boolean_t split, boolean_t trim,
1321 int new_count, new_bytes;
1324 uint16_t iflags, type;
1325 uint32_t access_mask;
1326 zfs_acl_node_t *newnode;
1327 size_t abstract_size = aclp->z_ops.ace_abstract_size();
1330 trivial_acl_t masks;
1332 new_count = new_bytes = 0;
1334 isdir = (vtype == VDIR);
1336 acl_trivial_access_masks((mode_t)mode, isdir, &masks);
1338 newnode = zfs_acl_node_alloc((abstract_size * 6) + aclp->z_acl_bytes);
1340 zacep = newnode->z_acldata;
1342 zfs_set_ace(aclp, zacep, masks.allow0, ALLOW, -1, ACE_OWNER);
1343 zacep = (void *)((uintptr_t)zacep + abstract_size);
1345 new_bytes += abstract_size;
1348 zfs_set_ace(aclp, zacep, masks.deny1, DENY, -1, ACE_OWNER);
1349 zacep = (void *)((uintptr_t)zacep + abstract_size);
1351 new_bytes += abstract_size;
1354 zfs_set_ace(aclp, zacep, masks.deny2, DENY, -1, OWNING_GROUP);
1355 zacep = (void *)((uintptr_t)zacep + abstract_size);
1357 new_bytes += abstract_size;
1360 while (acep = zfs_acl_next_ace(aclp, acep, &who, &access_mask,
1362 entry_type = (iflags & ACE_TYPE_FLAGS);
1364 * ACEs used to represent the file mode may be divided
1365 * into an equivalent pair of inherit-only and regular
1366 * ACEs, if they are inheritable.
1367 * Skip regular ACEs, which are replaced by the new mode.
1369 if (split && (entry_type == ACE_OWNER ||
1370 entry_type == OWNING_GROUP ||
1371 entry_type == ACE_EVERYONE)) {
1372 if (!isdir || !(iflags &
1373 (ACE_FILE_INHERIT_ACE|ACE_DIRECTORY_INHERIT_ACE)))
1376 * We preserve owner@, group@, or @everyone
1377 * permissions, if they are inheritable, by
1378 * copying them to inherit_only ACEs. This
1379 * prevents inheritable permissions from being
1380 * altered along with the file mode.
1382 iflags |= ACE_INHERIT_ONLY_ACE;
1386 * If this ACL has any inheritable ACEs, mark that in
1387 * the hints (which are later masked into the pflags)
1388 * so create knows to do inheritance.
1390 if (isdir && (iflags &
1391 (ACE_FILE_INHERIT_ACE|ACE_DIRECTORY_INHERIT_ACE)))
1392 aclp->z_hints |= ZFS_INHERIT_ACE;
1394 if ((type != ALLOW && type != DENY) ||
1395 (iflags & ACE_INHERIT_ONLY_ACE)) {
1397 case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE:
1398 case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE:
1399 case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE:
1400 case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE:
1401 aclp->z_hints |= ZFS_ACL_OBJ_ACE;
1406 * Limit permissions granted by ACEs to be no greater
1407 * than permissions of the requested group mode.
1408 * Applies when the "aclmode" property is set to
1411 if ((type == ALLOW) && trim)
1412 access_mask &= masks.group;
1414 zfs_set_ace(aclp, zacep, access_mask, type, who, iflags);
1415 ace_size = aclp->z_ops.ace_size(acep);
1416 zacep = (void *)((uintptr_t)zacep + ace_size);
1418 new_bytes += ace_size;
1420 zfs_set_ace(aclp, zacep, masks.owner, ALLOW, -1, ACE_OWNER);
1421 zacep = (void *)((uintptr_t)zacep + abstract_size);
1422 zfs_set_ace(aclp, zacep, masks.group, ALLOW, -1, OWNING_GROUP);
1423 zacep = (void *)((uintptr_t)zacep + abstract_size);
1424 zfs_set_ace(aclp, zacep, masks.everyone, ALLOW, -1, ACE_EVERYONE);
1427 new_bytes += abstract_size * 3;
1428 zfs_acl_release_nodes(aclp);
1429 aclp->z_acl_count = new_count;
1430 aclp->z_acl_bytes = new_bytes;
1431 newnode->z_ace_count = new_count;
1432 newnode->z_size = new_bytes;
1433 list_insert_tail(&aclp->z_acl, newnode);
1437 zfs_acl_chmod_setattr(znode_t *zp, zfs_acl_t **aclp, uint64_t mode)
1441 mutex_enter(&zp->z_acl_lock);
1442 ASSERT_VOP_ELOCKED(ZTOV(zp), __func__);
1443 if (zp->z_zfsvfs->z_acl_mode == ZFS_ACL_DISCARD)
1444 *aclp = zfs_acl_alloc(zfs_acl_version_zp(zp));
1446 error = zfs_acl_node_read(zp, aclp, B_TRUE);
1449 (*aclp)->z_hints = zp->z_pflags & V4_ACL_WIDE_FLAGS;
1450 zfs_acl_chmod(ZTOV(zp)->v_type, mode, B_TRUE,
1451 (zp->z_zfsvfs->z_acl_mode == ZFS_ACL_GROUPMASK), *aclp);
1453 mutex_exit(&zp->z_acl_lock);
1459 * Should ACE be inherited?
1462 zfs_ace_can_use(vtype_t vtype, uint16_t acep_flags)
1464 int iflags = (acep_flags & 0xf);
1466 if ((vtype == VDIR) && (iflags & ACE_DIRECTORY_INHERIT_ACE))
1468 else if (iflags & ACE_FILE_INHERIT_ACE)
1469 return (!((vtype == VDIR) &&
1470 (iflags & ACE_NO_PROPAGATE_INHERIT_ACE)));
1475 * inherit inheritable ACEs from parent
1478 zfs_acl_inherit(zfsvfs_t *zfsvfs, vtype_t vtype, zfs_acl_t *paclp,
1483 zfs_acl_node_t *aclnode;
1484 zfs_acl_t *aclp = NULL;
1486 uint32_t access_mask;
1487 uint16_t iflags, newflags, type;
1489 void *data1, *data2;
1490 size_t data1sz, data2sz;
1492 boolean_t isdir = (vtype == VDIR);
1494 aclp = zfs_acl_alloc(paclp->z_version);
1495 aclinherit = zfsvfs->z_acl_inherit;
1496 if (aclinherit == ZFS_ACL_DISCARD || vtype == VLNK)
1499 while (pacep = zfs_acl_next_ace(paclp, pacep, &who,
1500 &access_mask, &iflags, &type)) {
1503 * don't inherit bogus ACEs
1505 if (!zfs_acl_valid_ace_type(type, iflags))
1509 * Check if ACE is inheritable by this vnode
1511 if ((aclinherit == ZFS_ACL_NOALLOW && type == ALLOW) ||
1512 !zfs_ace_can_use(vtype, iflags))
1516 * Strip inherited execute permission from file if
1519 if (aclinherit == ZFS_ACL_PASSTHROUGH_X && type == ALLOW &&
1520 !isdir && ((mode & (S_IXUSR|S_IXGRP|S_IXOTH)) == 0)) {
1521 access_mask &= ~ACE_EXECUTE;
1525 * Strip write_acl and write_owner from permissions
1526 * when inheriting an ACE
1528 if (aclinherit == ZFS_ACL_RESTRICTED && type == ALLOW) {
1529 access_mask &= ~RESTRICTED_CLEAR;
1532 ace_size = aclp->z_ops.ace_size(pacep);
1533 aclnode = zfs_acl_node_alloc(ace_size);
1534 list_insert_tail(&aclp->z_acl, aclnode);
1535 acep = aclnode->z_acldata;
1537 zfs_set_ace(aclp, acep, access_mask, type,
1538 who, iflags|ACE_INHERITED_ACE);
1541 * Copy special opaque data if any
1543 if ((data1sz = paclp->z_ops.ace_data(pacep, &data1)) != 0) {
1544 VERIFY((data2sz = aclp->z_ops.ace_data(acep,
1545 &data2)) == data1sz);
1546 bcopy(data1, data2, data2sz);
1549 aclp->z_acl_count++;
1550 aclnode->z_ace_count++;
1551 aclp->z_acl_bytes += aclnode->z_size;
1552 newflags = aclp->z_ops.ace_flags_get(acep);
1555 * If ACE is not to be inherited further, or if the vnode is
1556 * not a directory, remove all inheritance flags
1558 if (!isdir || (iflags & ACE_NO_PROPAGATE_INHERIT_ACE)) {
1559 newflags &= ~ALL_INHERIT;
1560 aclp->z_ops.ace_flags_set(acep,
1561 newflags|ACE_INHERITED_ACE);
1566 * This directory has an inheritable ACE
1568 aclp->z_hints |= ZFS_INHERIT_ACE;
1571 * If only FILE_INHERIT is set then turn on
1574 if ((iflags & (ACE_FILE_INHERIT_ACE |
1575 ACE_DIRECTORY_INHERIT_ACE)) == ACE_FILE_INHERIT_ACE) {
1576 newflags |= ACE_INHERIT_ONLY_ACE;
1577 aclp->z_ops.ace_flags_set(acep,
1578 newflags|ACE_INHERITED_ACE);
1580 newflags &= ~ACE_INHERIT_ONLY_ACE;
1581 aclp->z_ops.ace_flags_set(acep,
1582 newflags|ACE_INHERITED_ACE);
1590 * Create file system object initial permissions
1591 * including inheritable ACEs.
1592 * Also, create FUIDs for owner and group.
1595 zfs_acl_ids_create(znode_t *dzp, int flag, vattr_t *vap, cred_t *cr,
1596 vsecattr_t *vsecp, zfs_acl_ids_t *acl_ids)
1599 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
1602 boolean_t trim = B_FALSE;
1603 boolean_t inherited = B_FALSE;
1605 ASSERT_VOP_ELOCKED(ZTOV(dzp), __func__);
1606 bzero(acl_ids, sizeof (zfs_acl_ids_t));
1607 acl_ids->z_mode = MAKEIMODE(vap->va_type, vap->va_mode);
1610 if ((error = zfs_vsec_2_aclp(zfsvfs, vap->va_type, vsecp, cr,
1611 &acl_ids->z_fuidp, &acl_ids->z_aclp)) != 0)
1614 * Determine uid and gid.
1616 if ((flag & IS_ROOT_NODE) || zfsvfs->z_replay ||
1617 ((flag & IS_XATTR) && (vap->va_type == VDIR))) {
1618 acl_ids->z_fuid = zfs_fuid_create(zfsvfs,
1619 (uint64_t)vap->va_uid, cr,
1620 ZFS_OWNER, &acl_ids->z_fuidp);
1621 acl_ids->z_fgid = zfs_fuid_create(zfsvfs,
1622 (uint64_t)vap->va_gid, cr,
1623 ZFS_GROUP, &acl_ids->z_fuidp);
1626 acl_ids->z_fuid = zfs_fuid_create_cred(zfsvfs, ZFS_OWNER,
1627 cr, &acl_ids->z_fuidp);
1628 acl_ids->z_fgid = 0;
1629 if (vap->va_mask & AT_GID) {
1630 acl_ids->z_fgid = zfs_fuid_create(zfsvfs,
1631 (uint64_t)vap->va_gid,
1632 cr, ZFS_GROUP, &acl_ids->z_fuidp);
1634 if (acl_ids->z_fgid != dzp->z_gid &&
1635 !groupmember(vap->va_gid, cr) &&
1636 secpolicy_vnode_create_gid(cr) != 0)
1637 acl_ids->z_fgid = 0;
1639 if (acl_ids->z_fgid == 0) {
1640 if (dzp->z_mode & S_ISGID) {
1644 acl_ids->z_fgid = dzp->z_gid;
1645 gid = zfs_fuid_map_id(zfsvfs, acl_ids->z_fgid,
1648 if (zfsvfs->z_use_fuids &&
1649 IS_EPHEMERAL(acl_ids->z_fgid)) {
1650 domain = zfs_fuid_idx_domain(
1651 &zfsvfs->z_fuid_idx,
1652 FUID_INDEX(acl_ids->z_fgid));
1653 rid = FUID_RID(acl_ids->z_fgid);
1654 zfs_fuid_node_add(&acl_ids->z_fuidp,
1656 FUID_INDEX(acl_ids->z_fgid),
1657 acl_ids->z_fgid, ZFS_GROUP);
1660 acl_ids->z_fgid = zfs_fuid_create_cred(zfsvfs,
1661 ZFS_GROUP, cr, &acl_ids->z_fuidp);
1662 #ifdef __FreeBSD_kernel__
1663 gid = acl_ids->z_fgid = dzp->z_gid;
1672 * If we're creating a directory, and the parent directory has the
1673 * set-GID bit set, set in on the new directory.
1674 * Otherwise, if the user is neither privileged nor a member of the
1675 * file's new group, clear the file's set-GID bit.
1678 if (!(flag & IS_ROOT_NODE) && (dzp->z_mode & S_ISGID) &&
1679 (vap->va_type == VDIR)) {
1680 acl_ids->z_mode |= S_ISGID;
1682 if ((acl_ids->z_mode & S_ISGID) &&
1683 secpolicy_vnode_setids_setgids(ZTOV(dzp), cr, gid) != 0)
1684 acl_ids->z_mode &= ~S_ISGID;
1687 if (acl_ids->z_aclp == NULL) {
1688 mutex_enter(&dzp->z_acl_lock);
1689 if (!(flag & IS_ROOT_NODE) &&
1690 (dzp->z_pflags & ZFS_INHERIT_ACE) &&
1691 !(dzp->z_pflags & ZFS_XATTR)) {
1692 VERIFY(0 == zfs_acl_node_read(dzp, &paclp, B_FALSE));
1693 acl_ids->z_aclp = zfs_acl_inherit(zfsvfs,
1694 vap->va_type, paclp, acl_ids->z_mode);
1698 zfs_acl_alloc(zfs_acl_version_zp(dzp));
1699 acl_ids->z_aclp->z_hints |= ZFS_ACL_TRIVIAL;
1701 mutex_exit(&dzp->z_acl_lock);
1703 if (vap->va_type == VDIR)
1704 acl_ids->z_aclp->z_hints |= ZFS_ACL_AUTO_INHERIT;
1706 if (zfsvfs->z_acl_mode == ZFS_ACL_GROUPMASK &&
1707 zfsvfs->z_acl_inherit != ZFS_ACL_PASSTHROUGH &&
1708 zfsvfs->z_acl_inherit != ZFS_ACL_PASSTHROUGH_X)
1710 zfs_acl_chmod(vap->va_type, acl_ids->z_mode, B_FALSE, trim,
1714 if (inherited || vsecp) {
1715 acl_ids->z_mode = zfs_mode_compute(acl_ids->z_mode,
1716 acl_ids->z_aclp, &acl_ids->z_aclp->z_hints,
1717 acl_ids->z_fuid, acl_ids->z_fgid);
1718 if (ace_trivial_common(acl_ids->z_aclp, 0, zfs_ace_walk) == 0)
1719 acl_ids->z_aclp->z_hints |= ZFS_ACL_TRIVIAL;
1726 * Free ACL and fuid_infop, but not the acl_ids structure
1729 zfs_acl_ids_free(zfs_acl_ids_t *acl_ids)
1731 if (acl_ids->z_aclp)
1732 zfs_acl_free(acl_ids->z_aclp);
1733 if (acl_ids->z_fuidp)
1734 zfs_fuid_info_free(acl_ids->z_fuidp);
1735 acl_ids->z_aclp = NULL;
1736 acl_ids->z_fuidp = NULL;
1740 zfs_acl_ids_overquota(zfsvfs_t *zfsvfs, zfs_acl_ids_t *acl_ids)
1742 return (zfs_fuid_overquota(zfsvfs, B_FALSE, acl_ids->z_fuid) ||
1743 zfs_fuid_overquota(zfsvfs, B_TRUE, acl_ids->z_fgid));
1747 * Retrieve a file's ACL
1750 zfs_getacl(znode_t *zp, vsecattr_t *vsecp, boolean_t skipaclchk, cred_t *cr)
1758 mask = vsecp->vsa_mask & (VSA_ACE | VSA_ACECNT |
1759 VSA_ACE_ACLFLAGS | VSA_ACE_ALLTYPES);
1762 return (SET_ERROR(ENOSYS));
1764 if (error = zfs_zaccess(zp, ACE_READ_ACL, 0, skipaclchk, cr))
1767 mutex_enter(&zp->z_acl_lock);
1769 ASSERT_VOP_LOCKED(ZTOV(zp), __func__);
1770 error = zfs_acl_node_read(zp, &aclp, B_FALSE);
1772 mutex_exit(&zp->z_acl_lock);
1777 * Scan ACL to determine number of ACEs
1779 if ((zp->z_pflags & ZFS_ACL_OBJ_ACE) && !(mask & VSA_ACE_ALLTYPES)) {
1782 uint32_t access_mask;
1783 uint16_t type, iflags;
1785 while (zacep = zfs_acl_next_ace(aclp, zacep,
1786 &who, &access_mask, &iflags, &type)) {
1788 case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE:
1789 case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE:
1790 case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE:
1791 case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE:
1798 vsecp->vsa_aclcnt = count;
1800 count = (int)aclp->z_acl_count;
1802 if (mask & VSA_ACECNT) {
1803 vsecp->vsa_aclcnt = count;
1806 if (mask & VSA_ACE) {
1809 aclsz = count * sizeof (ace_t) +
1810 sizeof (ace_object_t) * largeace;
1812 vsecp->vsa_aclentp = kmem_alloc(aclsz, KM_SLEEP);
1813 vsecp->vsa_aclentsz = aclsz;
1815 if (aclp->z_version == ZFS_ACL_VERSION_FUID)
1816 zfs_copy_fuid_2_ace(zp->z_zfsvfs, aclp, cr,
1817 vsecp->vsa_aclentp, !(mask & VSA_ACE_ALLTYPES));
1819 zfs_acl_node_t *aclnode;
1820 void *start = vsecp->vsa_aclentp;
1822 for (aclnode = list_head(&aclp->z_acl); aclnode;
1823 aclnode = list_next(&aclp->z_acl, aclnode)) {
1824 bcopy(aclnode->z_acldata, start,
1826 start = (caddr_t)start + aclnode->z_size;
1828 ASSERT((caddr_t)start - (caddr_t)vsecp->vsa_aclentp ==
1832 if (mask & VSA_ACE_ACLFLAGS) {
1833 vsecp->vsa_aclflags = 0;
1834 if (zp->z_pflags & ZFS_ACL_DEFAULTED)
1835 vsecp->vsa_aclflags |= ACL_DEFAULTED;
1836 if (zp->z_pflags & ZFS_ACL_PROTECTED)
1837 vsecp->vsa_aclflags |= ACL_PROTECTED;
1838 if (zp->z_pflags & ZFS_ACL_AUTO_INHERIT)
1839 vsecp->vsa_aclflags |= ACL_AUTO_INHERIT;
1842 mutex_exit(&zp->z_acl_lock);
1848 zfs_vsec_2_aclp(zfsvfs_t *zfsvfs, vtype_t obj_type,
1849 vsecattr_t *vsecp, cred_t *cr, zfs_fuid_info_t **fuidp, zfs_acl_t **zaclp)
1852 zfs_acl_node_t *aclnode;
1853 int aclcnt = vsecp->vsa_aclcnt;
1856 if (vsecp->vsa_aclcnt > MAX_ACL_ENTRIES || vsecp->vsa_aclcnt <= 0)
1857 return (SET_ERROR(EINVAL));
1859 aclp = zfs_acl_alloc(zfs_acl_version(zfsvfs->z_version));
1862 aclnode = zfs_acl_node_alloc(aclcnt * sizeof (zfs_object_ace_t));
1863 if (aclp->z_version == ZFS_ACL_VERSION_INITIAL) {
1864 if ((error = zfs_copy_ace_2_oldace(obj_type, aclp,
1865 (ace_t *)vsecp->vsa_aclentp, aclnode->z_acldata,
1866 aclcnt, &aclnode->z_size)) != 0) {
1868 zfs_acl_node_free(aclnode);
1872 if ((error = zfs_copy_ace_2_fuid(zfsvfs, obj_type, aclp,
1873 vsecp->vsa_aclentp, aclnode->z_acldata, aclcnt,
1874 &aclnode->z_size, fuidp, cr)) != 0) {
1876 zfs_acl_node_free(aclnode);
1880 aclp->z_acl_bytes = aclnode->z_size;
1881 aclnode->z_ace_count = aclcnt;
1882 aclp->z_acl_count = aclcnt;
1883 list_insert_head(&aclp->z_acl, aclnode);
1886 * If flags are being set then add them to z_hints
1888 if (vsecp->vsa_mask & VSA_ACE_ACLFLAGS) {
1889 if (vsecp->vsa_aclflags & ACL_PROTECTED)
1890 aclp->z_hints |= ZFS_ACL_PROTECTED;
1891 if (vsecp->vsa_aclflags & ACL_DEFAULTED)
1892 aclp->z_hints |= ZFS_ACL_DEFAULTED;
1893 if (vsecp->vsa_aclflags & ACL_AUTO_INHERIT)
1894 aclp->z_hints |= ZFS_ACL_AUTO_INHERIT;
1906 zfs_setacl(znode_t *zp, vsecattr_t *vsecp, boolean_t skipaclchk, cred_t *cr)
1908 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1909 zilog_t *zilog = zfsvfs->z_log;
1910 ulong_t mask = vsecp->vsa_mask & (VSA_ACE | VSA_ACECNT);
1914 zfs_fuid_info_t *fuidp = NULL;
1915 boolean_t fuid_dirtied;
1918 ASSERT_VOP_ELOCKED(ZTOV(zp), __func__);
1920 return (SET_ERROR(ENOSYS));
1922 if (zp->z_pflags & ZFS_IMMUTABLE)
1923 return (SET_ERROR(EPERM));
1925 if (error = zfs_zaccess(zp, ACE_WRITE_ACL, 0, skipaclchk, cr))
1928 error = zfs_vsec_2_aclp(zfsvfs, ZTOV(zp)->v_type, vsecp, cr, &fuidp,
1934 * If ACL wide flags aren't being set then preserve any
1937 if (!(vsecp->vsa_mask & VSA_ACE_ACLFLAGS)) {
1939 (zp->z_pflags & V4_ACL_WIDE_FLAGS);
1942 mutex_enter(&zp->z_acl_lock);
1944 tx = dmu_tx_create(zfsvfs->z_os);
1946 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE);
1948 fuid_dirtied = zfsvfs->z_fuid_dirty;
1950 zfs_fuid_txhold(zfsvfs, tx);
1953 * If old version and ACL won't fit in bonus and we aren't
1954 * upgrading then take out necessary DMU holds
1957 if ((acl_obj = zfs_external_acl(zp)) != 0) {
1958 if (zfsvfs->z_version >= ZPL_VERSION_FUID &&
1959 zfs_znode_acl_version(zp) <= ZFS_ACL_VERSION_INITIAL) {
1960 dmu_tx_hold_free(tx, acl_obj, 0,
1962 dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0,
1965 dmu_tx_hold_write(tx, acl_obj, 0, aclp->z_acl_bytes);
1967 } else if (!zp->z_is_sa && aclp->z_acl_bytes > ZFS_ACE_SPACE) {
1968 dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, aclp->z_acl_bytes);
1971 zfs_sa_upgrade_txholds(tx, zp);
1972 error = dmu_tx_assign(tx, TXG_NOWAIT);
1974 mutex_exit(&zp->z_acl_lock);
1976 if (error == ERESTART) {
1986 error = zfs_aclset_common(zp, aclp, cr, tx);
1988 ASSERT(zp->z_acl_cached == NULL);
1989 zp->z_acl_cached = aclp;
1992 zfs_fuid_sync(zfsvfs, tx);
1994 zfs_log_acl(zilog, tx, zp, vsecp, fuidp);
1997 zfs_fuid_info_free(fuidp);
1999 mutex_exit(&zp->z_acl_lock);
2005 * Check accesses of interest (AoI) against attributes of the dataset
2006 * such as read-only. Returns zero if no AoI conflict with dataset
2007 * attributes, otherwise an appropriate errno is returned.
2010 zfs_zaccess_dataset_check(znode_t *zp, uint32_t v4_mode)
2012 if ((v4_mode & WRITE_MASK) &&
2013 (zp->z_zfsvfs->z_vfs->vfs_flag & VFS_RDONLY) &&
2014 (!IS_DEVVP(ZTOV(zp)) ||
2015 (IS_DEVVP(ZTOV(zp)) && (v4_mode & WRITE_MASK_ATTRS)))) {
2016 return (SET_ERROR(EROFS));
2020 * Only check for READONLY on non-directories.
2022 if ((v4_mode & WRITE_MASK_DATA) &&
2023 (((ZTOV(zp)->v_type != VDIR) &&
2024 (zp->z_pflags & (ZFS_READONLY | ZFS_IMMUTABLE))) ||
2025 (ZTOV(zp)->v_type == VDIR &&
2026 (zp->z_pflags & ZFS_IMMUTABLE)))) {
2027 return (SET_ERROR(EPERM));
2031 if ((v4_mode & (ACE_DELETE | ACE_DELETE_CHILD)) &&
2032 (zp->z_pflags & ZFS_NOUNLINK)) {
2033 return (SET_ERROR(EPERM));
2037 * In FreeBSD we allow to modify directory's content is ZFS_NOUNLINK
2038 * (sunlnk) is set. We just don't allow directory removal, which is
2039 * handled in zfs_zaccess_delete().
2041 if ((v4_mode & ACE_DELETE) &&
2042 (zp->z_pflags & ZFS_NOUNLINK)) {
2047 if (((v4_mode & (ACE_READ_DATA|ACE_EXECUTE)) &&
2048 (zp->z_pflags & ZFS_AV_QUARANTINED))) {
2049 return (SET_ERROR(EACCES));
2056 * The primary usage of this function is to loop through all of the
2057 * ACEs in the znode, determining what accesses of interest (AoI) to
2058 * the caller are allowed or denied. The AoI are expressed as bits in
2059 * the working_mode parameter. As each ACE is processed, bits covered
2060 * by that ACE are removed from the working_mode. This removal
2061 * facilitates two things. The first is that when the working mode is
2062 * empty (= 0), we know we've looked at all the AoI. The second is
2063 * that the ACE interpretation rules don't allow a later ACE to undo
2064 * something granted or denied by an earlier ACE. Removing the
2065 * discovered access or denial enforces this rule. At the end of
2066 * processing the ACEs, all AoI that were found to be denied are
2067 * placed into the working_mode, giving the caller a mask of denied
2068 * accesses. Returns:
2069 * 0 if all AoI granted
2070 * EACCESS if the denied mask is non-zero
2071 * other error if abnormal failure (e.g., IO error)
2073 * A secondary usage of the function is to determine if any of the
2074 * AoI are granted. If an ACE grants any access in
2075 * the working_mode, we immediately short circuit out of the function.
2076 * This mode is chosen by setting anyaccess to B_TRUE. The
2077 * working_mode is not a denied access mask upon exit if the function
2078 * is used in this manner.
2081 zfs_zaccess_aces_check(znode_t *zp, uint32_t *working_mode,
2082 boolean_t anyaccess, cred_t *cr)
2084 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
2087 uid_t uid = crgetuid(cr);
2089 uint16_t type, iflags;
2090 uint16_t entry_type;
2091 uint32_t access_mask;
2092 uint32_t deny_mask = 0;
2093 zfs_ace_hdr_t *acep = NULL;
2098 zfs_fuid_map_ids(zp, cr, &fowner, &gowner);
2100 mutex_enter(&zp->z_acl_lock);
2102 ASSERT_VOP_LOCKED(ZTOV(zp), __func__);
2103 error = zfs_acl_node_read(zp, &aclp, B_FALSE);
2105 mutex_exit(&zp->z_acl_lock);
2109 ASSERT(zp->z_acl_cached);
2111 while (acep = zfs_acl_next_ace(aclp, acep, &who, &access_mask,
2113 uint32_t mask_matched;
2115 if (!zfs_acl_valid_ace_type(type, iflags))
2118 if (ZTOV(zp)->v_type == VDIR && (iflags & ACE_INHERIT_ONLY_ACE))
2121 /* Skip ACE if it does not affect any AoI */
2122 mask_matched = (access_mask & *working_mode);
2126 entry_type = (iflags & ACE_TYPE_FLAGS);
2130 switch (entry_type) {
2138 case ACE_IDENTIFIER_GROUP:
2139 checkit = zfs_groupmember(zfsvfs, who, cr);
2147 if (entry_type == 0) {
2150 newid = zfs_fuid_map_id(zfsvfs, who, cr,
2152 if (newid != IDMAP_WK_CREATOR_OWNER_UID &&
2157 mutex_exit(&zp->z_acl_lock);
2158 return (SET_ERROR(EIO));
2164 DTRACE_PROBE3(zfs__ace__denies,
2166 zfs_ace_hdr_t *, acep,
2167 uint32_t, mask_matched);
2168 deny_mask |= mask_matched;
2170 DTRACE_PROBE3(zfs__ace__allows,
2172 zfs_ace_hdr_t *, acep,
2173 uint32_t, mask_matched);
2175 mutex_exit(&zp->z_acl_lock);
2179 *working_mode &= ~mask_matched;
2183 if (*working_mode == 0)
2187 mutex_exit(&zp->z_acl_lock);
2189 /* Put the found 'denies' back on the working mode */
2191 *working_mode |= deny_mask;
2192 return (SET_ERROR(EACCES));
2193 } else if (*working_mode) {
2201 * Return true if any access whatsoever granted, we don't actually
2202 * care what access is granted.
2205 zfs_has_access(znode_t *zp, cred_t *cr)
2207 uint32_t have = ACE_ALL_PERMS;
2209 if (zfs_zaccess_aces_check(zp, &have, B_TRUE, cr) != 0) {
2212 owner = zfs_fuid_map_id(zp->z_zfsvfs, zp->z_uid, cr, ZFS_OWNER);
2213 return (secpolicy_vnode_any_access(cr, ZTOV(zp), owner) == 0);
2219 zfs_zaccess_common(znode_t *zp, uint32_t v4_mode, uint32_t *working_mode,
2220 boolean_t *check_privs, boolean_t skipaclchk, cred_t *cr)
2222 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
2225 *working_mode = v4_mode;
2226 *check_privs = B_TRUE;
2229 * Short circuit empty requests
2231 if (v4_mode == 0 || zfsvfs->z_replay) {
2236 if ((err = zfs_zaccess_dataset_check(zp, v4_mode)) != 0) {
2237 *check_privs = B_FALSE;
2242 * The caller requested that the ACL check be skipped. This
2243 * would only happen if the caller checked VOP_ACCESS() with a
2244 * 32 bit ACE mask and already had the appropriate permissions.
2251 return (zfs_zaccess_aces_check(zp, working_mode, B_FALSE, cr));
2255 zfs_zaccess_append(znode_t *zp, uint32_t *working_mode, boolean_t *check_privs,
2258 if (*working_mode != ACE_WRITE_DATA)
2259 return (SET_ERROR(EACCES));
2261 return (zfs_zaccess_common(zp, ACE_APPEND_DATA, working_mode,
2262 check_privs, B_FALSE, cr));
2266 zfs_fastaccesschk_execute(znode_t *zdp, cred_t *cr)
2268 boolean_t owner = B_FALSE;
2269 boolean_t groupmbr = B_FALSE;
2271 uid_t uid = crgetuid(cr);
2274 if (zdp->z_pflags & ZFS_AV_QUARANTINED)
2275 return (SET_ERROR(EACCES));
2277 is_attr = ((zdp->z_pflags & ZFS_XATTR) &&
2278 (ZTOV(zdp)->v_type == VDIR));
2283 mutex_enter(&zdp->z_acl_lock);
2285 if (zdp->z_pflags & ZFS_NO_EXECS_DENIED) {
2286 mutex_exit(&zdp->z_acl_lock);
2290 if (FUID_INDEX(zdp->z_uid) != 0 || FUID_INDEX(zdp->z_gid) != 0) {
2291 mutex_exit(&zdp->z_acl_lock);
2295 if (uid == zdp->z_uid) {
2297 if (zdp->z_mode & S_IXUSR) {
2298 mutex_exit(&zdp->z_acl_lock);
2301 mutex_exit(&zdp->z_acl_lock);
2305 if (groupmember(zdp->z_gid, cr)) {
2307 if (zdp->z_mode & S_IXGRP) {
2308 mutex_exit(&zdp->z_acl_lock);
2311 mutex_exit(&zdp->z_acl_lock);
2315 if (!owner && !groupmbr) {
2316 if (zdp->z_mode & S_IXOTH) {
2317 mutex_exit(&zdp->z_acl_lock);
2322 mutex_exit(&zdp->z_acl_lock);
2325 DTRACE_PROBE(zfs__fastpath__execute__access__miss);
2326 ZFS_ENTER(zdp->z_zfsvfs);
2327 error = zfs_zaccess(zdp, ACE_EXECUTE, 0, B_FALSE, cr);
2328 ZFS_EXIT(zdp->z_zfsvfs);
2333 * Determine whether Access should be granted/denied.
2335 * The least priv subsytem is always consulted as a basic privilege
2336 * can define any form of access.
2339 zfs_zaccess(znode_t *zp, int mode, int flags, boolean_t skipaclchk, cred_t *cr)
2341 uint32_t working_mode;
2344 boolean_t check_privs;
2346 znode_t *check_zp = zp;
2350 is_attr = ((zp->z_pflags & ZFS_XATTR) && (ZTOV(zp)->v_type == VDIR));
2352 #ifdef __FreeBSD_kernel__
2354 * In FreeBSD, we don't care about permissions of individual ADS.
2355 * Note that not checking them is not just an optimization - without
2356 * this shortcut, EA operations may bogusly fail with EACCES.
2358 if (zp->z_pflags & ZFS_XATTR)
2362 * If attribute then validate against base file
2367 if ((error = sa_lookup(zp->z_sa_hdl,
2368 SA_ZPL_PARENT(zp->z_zfsvfs), &parent,
2369 sizeof (parent))) != 0)
2372 if ((error = zfs_zget(zp->z_zfsvfs,
2373 parent, &xzp)) != 0) {
2380 * fixup mode to map to xattr perms
2383 if (mode & (ACE_WRITE_DATA|ACE_APPEND_DATA)) {
2384 mode &= ~(ACE_WRITE_DATA|ACE_APPEND_DATA);
2385 mode |= ACE_WRITE_NAMED_ATTRS;
2388 if (mode & (ACE_READ_DATA|ACE_EXECUTE)) {
2389 mode &= ~(ACE_READ_DATA|ACE_EXECUTE);
2390 mode |= ACE_READ_NAMED_ATTRS;
2395 owner = zfs_fuid_map_id(zp->z_zfsvfs, zp->z_uid, cr, ZFS_OWNER);
2397 * Map the bits required to the standard vnode flags VREAD|VWRITE|VEXEC
2398 * in needed_bits. Map the bits mapped by working_mode (currently
2399 * missing) in missing_bits.
2400 * Call secpolicy_vnode_access2() with (needed_bits & ~checkmode),
2405 working_mode = mode;
2406 if ((working_mode & (ACE_READ_ACL|ACE_READ_ATTRIBUTES)) &&
2407 owner == crgetuid(cr))
2408 working_mode &= ~(ACE_READ_ACL|ACE_READ_ATTRIBUTES);
2410 if (working_mode & (ACE_READ_DATA|ACE_READ_NAMED_ATTRS|
2411 ACE_READ_ACL|ACE_READ_ATTRIBUTES|ACE_SYNCHRONIZE))
2412 needed_bits |= VREAD;
2413 if (working_mode & (ACE_WRITE_DATA|ACE_WRITE_NAMED_ATTRS|
2414 ACE_APPEND_DATA|ACE_WRITE_ATTRIBUTES|ACE_SYNCHRONIZE))
2415 needed_bits |= VWRITE;
2416 if (working_mode & ACE_EXECUTE)
2417 needed_bits |= VEXEC;
2419 if ((error = zfs_zaccess_common(check_zp, mode, &working_mode,
2420 &check_privs, skipaclchk, cr)) == 0) {
2423 return (secpolicy_vnode_access2(cr, ZTOV(zp), owner,
2424 needed_bits, needed_bits));
2427 if (error && !check_privs) {
2433 if (error && (flags & V_APPEND)) {
2434 error = zfs_zaccess_append(zp, &working_mode, &check_privs, cr);
2437 if (error && check_privs) {
2438 mode_t checkmode = 0;
2441 * First check for implicit owner permission on
2442 * read_acl/read_attributes
2446 ASSERT(working_mode != 0);
2448 if ((working_mode & (ACE_READ_ACL|ACE_READ_ATTRIBUTES) &&
2449 owner == crgetuid(cr)))
2450 working_mode &= ~(ACE_READ_ACL|ACE_READ_ATTRIBUTES);
2452 if (working_mode & (ACE_READ_DATA|ACE_READ_NAMED_ATTRS|
2453 ACE_READ_ACL|ACE_READ_ATTRIBUTES|ACE_SYNCHRONIZE))
2455 if (working_mode & (ACE_WRITE_DATA|ACE_WRITE_NAMED_ATTRS|
2456 ACE_APPEND_DATA|ACE_WRITE_ATTRIBUTES|ACE_SYNCHRONIZE))
2457 checkmode |= VWRITE;
2458 if (working_mode & ACE_EXECUTE)
2461 error = secpolicy_vnode_access2(cr, ZTOV(check_zp), owner,
2462 needed_bits & ~checkmode, needed_bits);
2464 if (error == 0 && (working_mode & ACE_WRITE_OWNER))
2465 error = secpolicy_vnode_chown(ZTOV(check_zp), cr, owner);
2466 if (error == 0 && (working_mode & ACE_WRITE_ACL))
2467 error = secpolicy_vnode_setdac(ZTOV(check_zp), cr, owner);
2469 if (error == 0 && (working_mode &
2470 (ACE_DELETE|ACE_DELETE_CHILD)))
2471 error = secpolicy_vnode_remove(ZTOV(check_zp), cr);
2473 if (error == 0 && (working_mode & ACE_SYNCHRONIZE)) {
2474 error = secpolicy_vnode_chown(ZTOV(check_zp), cr, owner);
2478 * See if any bits other than those already checked
2479 * for are still present. If so then return EACCES
2481 if (working_mode & ~(ZFS_CHECKED_MASKS)) {
2482 error = SET_ERROR(EACCES);
2485 } else if (error == 0) {
2486 error = secpolicy_vnode_access2(cr, ZTOV(zp), owner,
2487 needed_bits, needed_bits);
2498 * Translate traditional unix VREAD/VWRITE/VEXEC mode into
2499 * native ACL format and call zfs_zaccess()
2502 zfs_zaccess_rwx(znode_t *zp, mode_t mode, int flags, cred_t *cr)
2504 return (zfs_zaccess(zp, zfs_unix_to_v4(mode >> 6), flags, B_FALSE, cr));
2508 * Access function for secpolicy_vnode_setattr
2511 zfs_zaccess_unix(znode_t *zp, mode_t mode, cred_t *cr)
2513 int v4_mode = zfs_unix_to_v4(mode >> 6);
2515 return (zfs_zaccess(zp, v4_mode, 0, B_FALSE, cr));
2519 zfs_delete_final_check(znode_t *zp, znode_t *dzp,
2520 mode_t available_perms, cred_t *cr)
2525 downer = zfs_fuid_map_id(dzp->z_zfsvfs, dzp->z_uid, cr, ZFS_OWNER);
2527 error = secpolicy_vnode_access2(cr, ZTOV(dzp),
2528 downer, available_perms, VWRITE|VEXEC);
2531 error = zfs_sticky_remove_access(dzp, zp, cr);
2537 * Determine whether Access should be granted/deny, without
2538 * consulting least priv subsystem.
2540 * The following chart is the recommended NFSv4 enforcement for
2541 * ability to delete an object.
2543 * -------------------------------------------------------
2544 * | Parent Dir | Target Object Permissions |
2546 * -------------------------------------------------------
2547 * | | ACL Allows | ACL Denies| Delete |
2548 * | | Delete | Delete | unspecified|
2549 * -------------------------------------------------------
2550 * | ACL Allows | Permit | Permit | Permit |
2551 * | DELETE_CHILD | |
2552 * -------------------------------------------------------
2553 * | ACL Denies | Permit | Deny | Deny |
2554 * | DELETE_CHILD | | | |
2555 * -------------------------------------------------------
2556 * | ACL specifies | | | |
2557 * | only allow | Permit | Permit | Permit |
2558 * | write and | | | |
2560 * -------------------------------------------------------
2561 * | ACL denies | | | |
2562 * | write and | Permit | Deny | Deny |
2564 * -------------------------------------------------------
2567 * No search privilege, can't even look up file?
2571 zfs_zaccess_delete(znode_t *dzp, znode_t *zp, cred_t *cr)
2573 uint32_t dzp_working_mode = 0;
2574 uint32_t zp_working_mode = 0;
2575 int dzp_error, zp_error;
2576 mode_t available_perms;
2577 boolean_t dzpcheck_privs = B_TRUE;
2578 boolean_t zpcheck_privs = B_TRUE;
2581 * We want specific DELETE permissions to
2582 * take precedence over WRITE/EXECUTE. We don't
2583 * want an ACL such as this to mess us up.
2584 * user:joe:write_data:deny,user:joe:delete:allow
2586 * However, deny permissions may ultimately be overridden
2587 * by secpolicy_vnode_access().
2589 * We will ask for all of the necessary permissions and then
2590 * look at the working modes from the directory and target object
2591 * to determine what was found.
2594 if (zp->z_pflags & (ZFS_IMMUTABLE | ZFS_NOUNLINK))
2595 return (SET_ERROR(EPERM));
2599 * If the directory permissions allow the delete, we are done.
2601 if ((dzp_error = zfs_zaccess_common(dzp, ACE_DELETE_CHILD,
2602 &dzp_working_mode, &dzpcheck_privs, B_FALSE, cr)) == 0)
2606 * If target object has delete permission then we are done
2608 if ((zp_error = zfs_zaccess_common(zp, ACE_DELETE, &zp_working_mode,
2609 &zpcheck_privs, B_FALSE, cr)) == 0)
2612 ASSERT(dzp_error && zp_error);
2614 if (!dzpcheck_privs)
2622 * If directory returns EACCES then delete_child was denied
2623 * due to deny delete_child. In this case send the request through
2624 * secpolicy_vnode_remove(). We don't use zfs_delete_final_check()
2625 * since that *could* allow the delete based on write/execute permission
2626 * and we want delete permissions to override write/execute.
2629 if (dzp_error == EACCES)
2630 return (secpolicy_vnode_remove(ZTOV(dzp), cr)); /* XXXPJD: s/dzp/zp/ ? */
2634 * only need to see if we have write/execute on directory.
2637 dzp_error = zfs_zaccess_common(dzp, ACE_EXECUTE|ACE_WRITE_DATA,
2638 &dzp_working_mode, &dzpcheck_privs, B_FALSE, cr);
2640 if (dzp_error != 0 && !dzpcheck_privs)
2647 available_perms = (dzp_working_mode & ACE_WRITE_DATA) ? 0 : VWRITE;
2648 available_perms |= (dzp_working_mode & ACE_EXECUTE) ? 0 : VEXEC;
2650 return (zfs_delete_final_check(zp, dzp, available_perms, cr));
2655 zfs_zaccess_rename(znode_t *sdzp, znode_t *szp, znode_t *tdzp,
2656 znode_t *tzp, cred_t *cr)
2661 if (szp->z_pflags & ZFS_AV_QUARANTINED)
2662 return (SET_ERROR(EACCES));
2664 add_perm = (ZTOV(szp)->v_type == VDIR) ?
2665 ACE_ADD_SUBDIRECTORY : ACE_ADD_FILE;
2668 * Rename permissions are combination of delete permission +
2669 * add file/subdir permission.
2671 * BSD operating systems also require write permission
2672 * on the directory being moved from one parent directory
2675 if (ZTOV(szp)->v_type == VDIR && ZTOV(sdzp) != ZTOV(tdzp)) {
2676 if (error = zfs_zaccess(szp, ACE_WRITE_DATA, 0, B_FALSE, cr))
2681 * first make sure we do the delete portion.
2683 * If that succeeds then check for add_file/add_subdir permissions
2686 if (error = zfs_zaccess_delete(sdzp, szp, cr))
2690 * If we have a tzp, see if we can delete it?
2693 if (error = zfs_zaccess_delete(tdzp, tzp, cr))
2698 * Now check for add permissions
2700 error = zfs_zaccess(tdzp, add_perm, 0, B_FALSE, cr);