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.
887 * Also, create FUIDs for any User/Group ACEs
890 zfs_mode_compute(uint64_t fmode, zfs_acl_t *aclp,
891 uint64_t *pflags, uint64_t fuid, uint64_t fgid)
896 zfs_ace_hdr_t *acep = NULL;
898 uint16_t iflags, type;
899 uint32_t access_mask;
900 boolean_t an_exec_denied = B_FALSE;
902 mode = (fmode & (S_IFMT | S_ISUID | S_ISGID | S_ISVTX));
904 while (acep = zfs_acl_next_ace(aclp, acep, &who,
905 &access_mask, &iflags, &type)) {
907 if (!zfs_acl_valid_ace_type(type, iflags))
910 entry_type = (iflags & ACE_TYPE_FLAGS);
913 * Skip over owner@, group@ or everyone@ inherit only ACEs
915 if ((iflags & ACE_INHERIT_ONLY_ACE) &&
916 (entry_type == ACE_OWNER || entry_type == ACE_EVERYONE ||
917 entry_type == OWNING_GROUP))
920 if (entry_type == ACE_OWNER || (entry_type == 0 &&
922 if ((access_mask & ACE_READ_DATA) &&
923 (!(seen & S_IRUSR))) {
929 if ((access_mask & ACE_WRITE_DATA) &&
930 (!(seen & S_IWUSR))) {
936 if ((access_mask & ACE_EXECUTE) &&
937 (!(seen & S_IXUSR))) {
943 } else if (entry_type == OWNING_GROUP ||
944 (entry_type == ACE_IDENTIFIER_GROUP && who == fgid)) {
945 if ((access_mask & ACE_READ_DATA) &&
946 (!(seen & S_IRGRP))) {
952 if ((access_mask & ACE_WRITE_DATA) &&
953 (!(seen & S_IWGRP))) {
959 if ((access_mask & ACE_EXECUTE) &&
960 (!(seen & S_IXGRP))) {
966 } else if (entry_type == ACE_EVERYONE) {
967 if ((access_mask & ACE_READ_DATA)) {
968 if (!(seen & S_IRUSR)) {
974 if (!(seen & S_IRGRP)) {
980 if (!(seen & S_IROTH)) {
987 if ((access_mask & ACE_WRITE_DATA)) {
988 if (!(seen & S_IWUSR)) {
994 if (!(seen & S_IWGRP)) {
1000 if (!(seen & S_IWOTH)) {
1002 if (type == ALLOW) {
1007 if ((access_mask & ACE_EXECUTE)) {
1008 if (!(seen & S_IXUSR)) {
1010 if (type == ALLOW) {
1014 if (!(seen & S_IXGRP)) {
1016 if (type == ALLOW) {
1020 if (!(seen & S_IXOTH)) {
1022 if (type == ALLOW) {
1029 * Only care if this IDENTIFIER_GROUP or
1030 * USER ACE denies execute access to someone,
1031 * mode is not affected
1033 if ((access_mask & ACE_EXECUTE) && type == DENY)
1034 an_exec_denied = B_TRUE;
1039 * Failure to allow is effectively a deny, so execute permission
1040 * is denied if it was never mentioned or if we explicitly
1041 * weren't allowed it.
1043 if (!an_exec_denied &&
1044 ((seen & ALL_MODE_EXECS) != ALL_MODE_EXECS ||
1045 (mode & ALL_MODE_EXECS) != ALL_MODE_EXECS))
1046 an_exec_denied = B_TRUE;
1049 *pflags &= ~ZFS_NO_EXECS_DENIED;
1051 *pflags |= ZFS_NO_EXECS_DENIED;
1057 * Read an external acl object. If the intent is to modify, always
1058 * create a new acl and leave any cached acl in place.
1061 zfs_acl_node_read(znode_t *zp, boolean_t have_lock, zfs_acl_t **aclpp,
1062 boolean_t will_modify)
1067 zfs_acl_node_t *aclnode;
1068 zfs_acl_phys_t znode_acl;
1071 boolean_t drop_lock = B_FALSE;
1073 ASSERT(MUTEX_HELD(&zp->z_acl_lock));
1075 if (zp->z_acl_cached && !will_modify) {
1076 *aclpp = zp->z_acl_cached;
1081 * close race where znode could be upgrade while trying to
1082 * read the znode attributes.
1084 * But this could only happen if the file isn't already an SA
1087 if (!zp->z_is_sa && !have_lock) {
1088 mutex_enter(&zp->z_lock);
1091 version = zfs_znode_acl_version(zp);
1093 if ((error = zfs_acl_znode_info(zp, &aclsize,
1094 &acl_count, &znode_acl)) != 0) {
1098 aclp = zfs_acl_alloc(version);
1100 aclp->z_acl_count = acl_count;
1101 aclp->z_acl_bytes = aclsize;
1103 aclnode = zfs_acl_node_alloc(aclsize);
1104 aclnode->z_ace_count = aclp->z_acl_count;
1105 aclnode->z_size = aclsize;
1108 if (znode_acl.z_acl_extern_obj) {
1109 error = dmu_read(zp->z_zfsvfs->z_os,
1110 znode_acl.z_acl_extern_obj, 0, aclnode->z_size,
1111 aclnode->z_acldata, DMU_READ_PREFETCH);
1113 bcopy(znode_acl.z_ace_data, aclnode->z_acldata,
1117 error = sa_lookup(zp->z_sa_hdl, SA_ZPL_DACL_ACES(zp->z_zfsvfs),
1118 aclnode->z_acldata, aclnode->z_size);
1123 zfs_acl_node_free(aclnode);
1124 /* convert checksum errors into IO errors */
1125 if (error == ECKSUM)
1126 error = SET_ERROR(EIO);
1130 list_insert_head(&aclp->z_acl, aclnode);
1134 zp->z_acl_cached = aclp;
1137 mutex_exit(&zp->z_lock);
1143 zfs_acl_data_locator(void **dataptr, uint32_t *length, uint32_t buflen,
1144 boolean_t start, void *userdata)
1146 zfs_acl_locator_cb_t *cb = (zfs_acl_locator_cb_t *)userdata;
1149 cb->cb_acl_node = list_head(&cb->cb_aclp->z_acl);
1151 cb->cb_acl_node = list_next(&cb->cb_aclp->z_acl,
1154 *dataptr = cb->cb_acl_node->z_acldata;
1155 *length = cb->cb_acl_node->z_size;
1159 zfs_acl_chown_setattr(znode_t *zp)
1164 ASSERT(MUTEX_HELD(&zp->z_lock));
1165 ASSERT(MUTEX_HELD(&zp->z_acl_lock));
1167 if ((error = zfs_acl_node_read(zp, B_TRUE, &aclp, B_FALSE)) == 0)
1168 zp->z_mode = zfs_mode_compute(zp->z_mode, aclp,
1169 &zp->z_pflags, zp->z_uid, zp->z_gid);
1174 * common code for setting ACLs.
1176 * This function is called from zfs_mode_update, zfs_perm_init, and zfs_setacl.
1177 * zfs_setacl passes a non-NULL inherit pointer (ihp) to indicate that it's
1178 * already checked the acl and knows whether to inherit.
1181 zfs_aclset_common(znode_t *zp, zfs_acl_t *aclp, cred_t *cr, dmu_tx_t *tx)
1184 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1185 dmu_object_type_t otype;
1186 zfs_acl_locator_cb_t locate = { 0 };
1188 sa_bulk_attr_t bulk[5];
1194 mode = zfs_mode_compute(mode, aclp, &zp->z_pflags,
1195 zp->z_uid, zp->z_gid);
1198 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs), NULL,
1199 &mode, sizeof (mode));
1200 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL,
1201 &zp->z_pflags, sizeof (zp->z_pflags));
1202 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL,
1203 &ctime, sizeof (ctime));
1205 if (zp->z_acl_cached) {
1206 zfs_acl_free(zp->z_acl_cached);
1207 zp->z_acl_cached = NULL;
1213 if (!zfsvfs->z_use_fuids) {
1214 otype = DMU_OT_OLDACL;
1216 if ((aclp->z_version == ZFS_ACL_VERSION_INITIAL) &&
1217 (zfsvfs->z_version >= ZPL_VERSION_FUID))
1218 zfs_acl_xform(zp, aclp, cr);
1219 ASSERT(aclp->z_version >= ZFS_ACL_VERSION_FUID);
1224 * Arrgh, we have to handle old on disk format
1225 * as well as newer (preferred) SA format.
1228 if (zp->z_is_sa) { /* the easy case, just update the ACL attribute */
1229 locate.cb_aclp = aclp;
1230 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_DACL_ACES(zfsvfs),
1231 zfs_acl_data_locator, &locate, aclp->z_acl_bytes);
1232 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_DACL_COUNT(zfsvfs),
1233 NULL, &aclp->z_acl_count, sizeof (uint64_t));
1234 } else { /* Painful legacy way */
1235 zfs_acl_node_t *aclnode;
1237 zfs_acl_phys_t acl_phys;
1240 if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_ZNODE_ACL(zfsvfs),
1241 &acl_phys, sizeof (acl_phys))) != 0)
1244 aoid = acl_phys.z_acl_extern_obj;
1246 if (aclp->z_acl_bytes > ZFS_ACE_SPACE) {
1248 * If ACL was previously external and we are now
1249 * converting to new ACL format then release old
1250 * ACL object and create a new one.
1253 aclp->z_version != acl_phys.z_acl_version) {
1254 error = dmu_object_free(zfsvfs->z_os, aoid, tx);
1260 aoid = dmu_object_alloc(zfsvfs->z_os,
1261 otype, aclp->z_acl_bytes,
1262 otype == DMU_OT_ACL ?
1263 DMU_OT_SYSACL : DMU_OT_NONE,
1264 otype == DMU_OT_ACL ?
1265 DN_MAX_BONUSLEN : 0, tx);
1267 (void) dmu_object_set_blocksize(zfsvfs->z_os,
1268 aoid, aclp->z_acl_bytes, 0, tx);
1270 acl_phys.z_acl_extern_obj = aoid;
1271 for (aclnode = list_head(&aclp->z_acl); aclnode;
1272 aclnode = list_next(&aclp->z_acl, aclnode)) {
1273 if (aclnode->z_ace_count == 0)
1275 dmu_write(zfsvfs->z_os, aoid, off,
1276 aclnode->z_size, aclnode->z_acldata, tx);
1277 off += aclnode->z_size;
1280 void *start = acl_phys.z_ace_data;
1282 * Migrating back embedded?
1284 if (acl_phys.z_acl_extern_obj) {
1285 error = dmu_object_free(zfsvfs->z_os,
1286 acl_phys.z_acl_extern_obj, tx);
1289 acl_phys.z_acl_extern_obj = 0;
1292 for (aclnode = list_head(&aclp->z_acl); aclnode;
1293 aclnode = list_next(&aclp->z_acl, aclnode)) {
1294 if (aclnode->z_ace_count == 0)
1296 bcopy(aclnode->z_acldata, start,
1298 start = (caddr_t)start + aclnode->z_size;
1302 * If Old version then swap count/bytes to match old
1303 * layout of znode_acl_phys_t.
1305 if (aclp->z_version == ZFS_ACL_VERSION_INITIAL) {
1306 acl_phys.z_acl_size = aclp->z_acl_count;
1307 acl_phys.z_acl_count = aclp->z_acl_bytes;
1309 acl_phys.z_acl_size = aclp->z_acl_bytes;
1310 acl_phys.z_acl_count = aclp->z_acl_count;
1312 acl_phys.z_acl_version = aclp->z_version;
1314 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ZNODE_ACL(zfsvfs), NULL,
1315 &acl_phys, sizeof (acl_phys));
1319 * Replace ACL wide bits, but first clear them.
1321 zp->z_pflags &= ~ZFS_ACL_WIDE_FLAGS;
1323 zp->z_pflags |= aclp->z_hints;
1325 if (ace_trivial_common(aclp, 0, zfs_ace_walk) == 0)
1326 zp->z_pflags |= ZFS_ACL_TRIVIAL;
1328 zfs_tstamp_update_setup(zp, STATE_CHANGED, NULL, ctime, B_TRUE);
1329 return (sa_bulk_update(zp->z_sa_hdl, bulk, count, tx));
1333 zfs_acl_chmod(vtype_t vtype, uint64_t mode, boolean_t trim, zfs_acl_t *aclp)
1337 int new_count, new_bytes;
1340 uint16_t iflags, type;
1341 uint32_t access_mask;
1342 zfs_acl_node_t *newnode;
1343 size_t abstract_size = aclp->z_ops.ace_abstract_size();
1346 trivial_acl_t masks;
1348 new_count = new_bytes = 0;
1350 isdir = (vtype == VDIR);
1352 acl_trivial_access_masks((mode_t)mode, isdir, &masks);
1354 newnode = zfs_acl_node_alloc((abstract_size * 6) + aclp->z_acl_bytes);
1356 zacep = newnode->z_acldata;
1358 zfs_set_ace(aclp, zacep, masks.allow0, ALLOW, -1, ACE_OWNER);
1359 zacep = (void *)((uintptr_t)zacep + abstract_size);
1361 new_bytes += abstract_size;
1364 zfs_set_ace(aclp, zacep, masks.deny1, DENY, -1, ACE_OWNER);
1365 zacep = (void *)((uintptr_t)zacep + abstract_size);
1367 new_bytes += abstract_size;
1370 zfs_set_ace(aclp, zacep, masks.deny2, DENY, -1, OWNING_GROUP);
1371 zacep = (void *)((uintptr_t)zacep + abstract_size);
1373 new_bytes += abstract_size;
1376 while (acep = zfs_acl_next_ace(aclp, acep, &who, &access_mask,
1378 uint16_t inherit_flags;
1380 entry_type = (iflags & ACE_TYPE_FLAGS);
1381 inherit_flags = (iflags & ALL_INHERIT);
1383 if ((entry_type == ACE_OWNER || entry_type == ACE_EVERYONE ||
1384 (entry_type == OWNING_GROUP)) &&
1385 ((inherit_flags & ACE_INHERIT_ONLY_ACE) == 0)) {
1390 * If this ACL has any inheritable ACEs, mark that in
1391 * the hints (which are later masked into the pflags)
1392 * so create knows to do inheritance.
1394 if (isdir && (inherit_flags &
1395 (ACE_FILE_INHERIT_ACE|ACE_DIRECTORY_INHERIT_ACE)))
1396 aclp->z_hints |= ZFS_INHERIT_ACE;
1398 if ((type != ALLOW && type != DENY) ||
1399 (inherit_flags & ACE_INHERIT_ONLY_ACE)) {
1401 case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE:
1402 case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE:
1403 case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE:
1404 case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE:
1405 aclp->z_hints |= ZFS_ACL_OBJ_ACE;
1411 * Limit permissions to be no greater than
1412 * group permissions.
1413 * The "aclinherit" and "aclmode" properties
1414 * affect policy for create and chmod(2),
1417 if ((type == ALLOW) && trim)
1418 access_mask &= masks.group;
1420 zfs_set_ace(aclp, zacep, access_mask, type, who, iflags);
1421 ace_size = aclp->z_ops.ace_size(acep);
1422 zacep = (void *)((uintptr_t)zacep + ace_size);
1424 new_bytes += ace_size;
1426 zfs_set_ace(aclp, zacep, masks.owner, 0, -1, ACE_OWNER);
1427 zacep = (void *)((uintptr_t)zacep + abstract_size);
1428 zfs_set_ace(aclp, zacep, masks.group, 0, -1, OWNING_GROUP);
1429 zacep = (void *)((uintptr_t)zacep + abstract_size);
1430 zfs_set_ace(aclp, zacep, masks.everyone, 0, -1, ACE_EVERYONE);
1433 new_bytes += abstract_size * 3;
1434 zfs_acl_release_nodes(aclp);
1435 aclp->z_acl_count = new_count;
1436 aclp->z_acl_bytes = new_bytes;
1437 newnode->z_ace_count = new_count;
1438 newnode->z_size = new_bytes;
1439 list_insert_tail(&aclp->z_acl, newnode);
1443 zfs_acl_chmod_setattr(znode_t *zp, zfs_acl_t **aclp, uint64_t mode)
1447 mutex_enter(&zp->z_acl_lock);
1448 mutex_enter(&zp->z_lock);
1449 if (zp->z_zfsvfs->z_acl_mode == ZFS_ACL_DISCARD)
1450 *aclp = zfs_acl_alloc(zfs_acl_version_zp(zp));
1452 error = zfs_acl_node_read(zp, B_TRUE, aclp, B_TRUE);
1455 (*aclp)->z_hints = zp->z_pflags & V4_ACL_WIDE_FLAGS;
1456 zfs_acl_chmod(ZTOV(zp)->v_type, mode,
1457 (zp->z_zfsvfs->z_acl_mode == ZFS_ACL_GROUPMASK), *aclp);
1459 mutex_exit(&zp->z_lock);
1460 mutex_exit(&zp->z_acl_lock);
1466 * strip off write_owner and write_acl
1469 zfs_restricted_update(zfsvfs_t *zfsvfs, zfs_acl_t *aclp, void *acep)
1471 uint32_t mask = aclp->z_ops.ace_mask_get(acep);
1473 if ((zfsvfs->z_acl_inherit == ZFS_ACL_RESTRICTED) &&
1474 (aclp->z_ops.ace_type_get(acep) == ALLOW)) {
1475 mask &= ~RESTRICTED_CLEAR;
1476 aclp->z_ops.ace_mask_set(acep, mask);
1481 * Should ACE be inherited?
1484 zfs_ace_can_use(vtype_t vtype, uint16_t acep_flags)
1486 int iflags = (acep_flags & 0xf);
1488 if ((vtype == VDIR) && (iflags & ACE_DIRECTORY_INHERIT_ACE))
1490 else if (iflags & ACE_FILE_INHERIT_ACE)
1491 return (!((vtype == VDIR) &&
1492 (iflags & ACE_NO_PROPAGATE_INHERIT_ACE)));
1497 * inherit inheritable ACEs from parent
1500 zfs_acl_inherit(zfsvfs_t *zfsvfs, vtype_t vtype, zfs_acl_t *paclp,
1501 uint64_t mode, boolean_t *need_chmod)
1505 zfs_acl_node_t *aclnode;
1506 zfs_acl_t *aclp = NULL;
1508 uint32_t access_mask;
1509 uint16_t iflags, newflags, type;
1511 void *data1, *data2;
1512 size_t data1sz, data2sz;
1513 boolean_t vdir = vtype == VDIR;
1514 boolean_t vreg = vtype == VREG;
1515 boolean_t passthrough, passthrough_x, noallow;
1518 zfsvfs->z_acl_inherit == ZFS_ACL_PASSTHROUGH_X;
1519 passthrough = passthrough_x ||
1520 zfsvfs->z_acl_inherit == ZFS_ACL_PASSTHROUGH;
1522 zfsvfs->z_acl_inherit == ZFS_ACL_NOALLOW;
1524 *need_chmod = B_TRUE;
1526 aclp = zfs_acl_alloc(paclp->z_version);
1527 if (zfsvfs->z_acl_inherit == ZFS_ACL_DISCARD || vtype == VLNK)
1529 while (pacep = zfs_acl_next_ace(paclp, pacep, &who,
1530 &access_mask, &iflags, &type)) {
1533 * don't inherit bogus ACEs
1535 if (!zfs_acl_valid_ace_type(type, iflags))
1538 if (noallow && type == ALLOW)
1541 ace_size = aclp->z_ops.ace_size(pacep);
1543 if (!zfs_ace_can_use(vtype, iflags))
1547 * If owner@, group@, or everyone@ inheritable
1548 * then zfs_acl_chmod() isn't needed.
1551 ((iflags & (ACE_OWNER|ACE_EVERYONE)) ||
1552 ((iflags & OWNING_GROUP) ==
1553 OWNING_GROUP)) && (vreg || (vdir && (iflags &
1554 ACE_DIRECTORY_INHERIT_ACE)))) {
1555 *need_chmod = B_FALSE;
1558 if (!vdir && passthrough_x &&
1559 ((mode & (S_IXUSR | S_IXGRP | S_IXOTH)) == 0)) {
1560 access_mask &= ~ACE_EXECUTE;
1563 aclnode = zfs_acl_node_alloc(ace_size);
1564 list_insert_tail(&aclp->z_acl, aclnode);
1565 acep = aclnode->z_acldata;
1567 zfs_set_ace(aclp, acep, access_mask, type,
1568 who, iflags|ACE_INHERITED_ACE);
1571 * Copy special opaque data if any
1573 if ((data1sz = paclp->z_ops.ace_data(pacep, &data1)) != 0) {
1574 VERIFY((data2sz = aclp->z_ops.ace_data(acep,
1575 &data2)) == data1sz);
1576 bcopy(data1, data2, data2sz);
1579 aclp->z_acl_count++;
1580 aclnode->z_ace_count++;
1581 aclp->z_acl_bytes += aclnode->z_size;
1582 newflags = aclp->z_ops.ace_flags_get(acep);
1585 aclp->z_hints |= ZFS_INHERIT_ACE;
1587 if ((iflags & ACE_NO_PROPAGATE_INHERIT_ACE) || !vdir) {
1588 newflags &= ~ALL_INHERIT;
1589 aclp->z_ops.ace_flags_set(acep,
1590 newflags|ACE_INHERITED_ACE);
1591 zfs_restricted_update(zfsvfs, aclp, acep);
1598 * If only FILE_INHERIT is set then turn on
1601 if ((iflags & (ACE_FILE_INHERIT_ACE |
1602 ACE_DIRECTORY_INHERIT_ACE)) == ACE_FILE_INHERIT_ACE) {
1603 newflags |= ACE_INHERIT_ONLY_ACE;
1604 aclp->z_ops.ace_flags_set(acep,
1605 newflags|ACE_INHERITED_ACE);
1607 newflags &= ~ACE_INHERIT_ONLY_ACE;
1608 aclp->z_ops.ace_flags_set(acep,
1609 newflags|ACE_INHERITED_ACE);
1616 * Create file system object initial permissions
1617 * including inheritable ACEs.
1620 zfs_acl_ids_create(znode_t *dzp, int flag, vattr_t *vap, cred_t *cr,
1621 vsecattr_t *vsecp, zfs_acl_ids_t *acl_ids)
1624 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
1627 boolean_t need_chmod = B_TRUE;
1628 boolean_t inherited = B_FALSE;
1630 bzero(acl_ids, sizeof (zfs_acl_ids_t));
1631 acl_ids->z_mode = MAKEIMODE(vap->va_type, vap->va_mode);
1634 if ((error = zfs_vsec_2_aclp(zfsvfs, vap->va_type, vsecp, cr,
1635 &acl_ids->z_fuidp, &acl_ids->z_aclp)) != 0)
1638 * Determine uid and gid.
1640 if ((flag & IS_ROOT_NODE) || zfsvfs->z_replay ||
1641 ((flag & IS_XATTR) && (vap->va_type == VDIR))) {
1642 acl_ids->z_fuid = zfs_fuid_create(zfsvfs,
1643 (uint64_t)vap->va_uid, cr,
1644 ZFS_OWNER, &acl_ids->z_fuidp);
1645 acl_ids->z_fgid = zfs_fuid_create(zfsvfs,
1646 (uint64_t)vap->va_gid, cr,
1647 ZFS_GROUP, &acl_ids->z_fuidp);
1650 acl_ids->z_fuid = zfs_fuid_create_cred(zfsvfs, ZFS_OWNER,
1651 cr, &acl_ids->z_fuidp);
1652 acl_ids->z_fgid = 0;
1653 if (vap->va_mask & AT_GID) {
1654 acl_ids->z_fgid = zfs_fuid_create(zfsvfs,
1655 (uint64_t)vap->va_gid,
1656 cr, ZFS_GROUP, &acl_ids->z_fuidp);
1658 if (acl_ids->z_fgid != dzp->z_gid &&
1659 !groupmember(vap->va_gid, cr) &&
1660 secpolicy_vnode_create_gid(cr) != 0)
1661 acl_ids->z_fgid = 0;
1663 if (acl_ids->z_fgid == 0) {
1664 if (dzp->z_mode & S_ISGID) {
1668 acl_ids->z_fgid = dzp->z_gid;
1669 gid = zfs_fuid_map_id(zfsvfs, acl_ids->z_fgid,
1672 if (zfsvfs->z_use_fuids &&
1673 IS_EPHEMERAL(acl_ids->z_fgid)) {
1674 domain = zfs_fuid_idx_domain(
1675 &zfsvfs->z_fuid_idx,
1676 FUID_INDEX(acl_ids->z_fgid));
1677 rid = FUID_RID(acl_ids->z_fgid);
1678 zfs_fuid_node_add(&acl_ids->z_fuidp,
1680 FUID_INDEX(acl_ids->z_fgid),
1681 acl_ids->z_fgid, ZFS_GROUP);
1684 acl_ids->z_fgid = zfs_fuid_create_cred(zfsvfs,
1685 ZFS_GROUP, cr, &acl_ids->z_fuidp);
1687 gid = acl_ids->z_fgid = dzp->z_gid;
1696 * If we're creating a directory, and the parent directory has the
1697 * set-GID bit set, set in on the new directory.
1698 * Otherwise, if the user is neither privileged nor a member of the
1699 * file's new group, clear the file's set-GID bit.
1702 if (!(flag & IS_ROOT_NODE) && (dzp->z_mode & S_ISGID) &&
1703 (vap->va_type == VDIR)) {
1704 acl_ids->z_mode |= S_ISGID;
1706 if ((acl_ids->z_mode & S_ISGID) &&
1707 secpolicy_vnode_setids_setgids(ZTOV(dzp), cr, gid) != 0)
1708 acl_ids->z_mode &= ~S_ISGID;
1711 if (acl_ids->z_aclp == NULL) {
1712 mutex_enter(&dzp->z_acl_lock);
1713 mutex_enter(&dzp->z_lock);
1714 if (!(flag & IS_ROOT_NODE) &&
1715 (dzp->z_pflags & ZFS_INHERIT_ACE) &&
1716 !(dzp->z_pflags & ZFS_XATTR)) {
1717 VERIFY(0 == zfs_acl_node_read(dzp, B_TRUE,
1719 acl_ids->z_aclp = zfs_acl_inherit(zfsvfs,
1720 vap->va_type, paclp, acl_ids->z_mode, &need_chmod);
1724 zfs_acl_alloc(zfs_acl_version_zp(dzp));
1725 acl_ids->z_aclp->z_hints |= ZFS_ACL_TRIVIAL;
1727 mutex_exit(&dzp->z_lock);
1728 mutex_exit(&dzp->z_acl_lock);
1730 acl_ids->z_aclp->z_hints |= (vap->va_type == VDIR) ?
1731 ZFS_ACL_AUTO_INHERIT : 0;
1732 zfs_acl_chmod(vap->va_type, acl_ids->z_mode,
1733 (zfsvfs->z_acl_inherit == ZFS_ACL_RESTRICTED),
1738 if (inherited || vsecp) {
1739 acl_ids->z_mode = zfs_mode_compute(acl_ids->z_mode,
1740 acl_ids->z_aclp, &acl_ids->z_aclp->z_hints,
1741 acl_ids->z_fuid, acl_ids->z_fgid);
1742 if (ace_trivial_common(acl_ids->z_aclp, 0, zfs_ace_walk) == 0)
1743 acl_ids->z_aclp->z_hints |= ZFS_ACL_TRIVIAL;
1750 * Free ACL and fuid_infop, but not the acl_ids structure
1753 zfs_acl_ids_free(zfs_acl_ids_t *acl_ids)
1755 if (acl_ids->z_aclp)
1756 zfs_acl_free(acl_ids->z_aclp);
1757 if (acl_ids->z_fuidp)
1758 zfs_fuid_info_free(acl_ids->z_fuidp);
1759 acl_ids->z_aclp = NULL;
1760 acl_ids->z_fuidp = NULL;
1764 zfs_acl_ids_overquota(zfsvfs_t *zfsvfs, zfs_acl_ids_t *acl_ids)
1766 return (zfs_fuid_overquota(zfsvfs, B_FALSE, acl_ids->z_fuid) ||
1767 zfs_fuid_overquota(zfsvfs, B_TRUE, acl_ids->z_fgid));
1771 * Retrieve a file's ACL
1774 zfs_getacl(znode_t *zp, vsecattr_t *vsecp, boolean_t skipaclchk, cred_t *cr)
1782 mask = vsecp->vsa_mask & (VSA_ACE | VSA_ACECNT |
1783 VSA_ACE_ACLFLAGS | VSA_ACE_ALLTYPES);
1786 return (SET_ERROR(ENOSYS));
1788 if (error = zfs_zaccess(zp, ACE_READ_ACL, 0, skipaclchk, cr))
1791 mutex_enter(&zp->z_acl_lock);
1793 error = zfs_acl_node_read(zp, B_FALSE, &aclp, B_FALSE);
1795 mutex_exit(&zp->z_acl_lock);
1800 * Scan ACL to determine number of ACEs
1802 if ((zp->z_pflags & ZFS_ACL_OBJ_ACE) && !(mask & VSA_ACE_ALLTYPES)) {
1805 uint32_t access_mask;
1806 uint16_t type, iflags;
1808 while (zacep = zfs_acl_next_ace(aclp, zacep,
1809 &who, &access_mask, &iflags, &type)) {
1811 case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE:
1812 case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE:
1813 case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE:
1814 case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE:
1821 vsecp->vsa_aclcnt = count;
1823 count = (int)aclp->z_acl_count;
1825 if (mask & VSA_ACECNT) {
1826 vsecp->vsa_aclcnt = count;
1829 if (mask & VSA_ACE) {
1832 aclsz = count * sizeof (ace_t) +
1833 sizeof (ace_object_t) * largeace;
1835 vsecp->vsa_aclentp = kmem_alloc(aclsz, KM_SLEEP);
1836 vsecp->vsa_aclentsz = aclsz;
1838 if (aclp->z_version == ZFS_ACL_VERSION_FUID)
1839 zfs_copy_fuid_2_ace(zp->z_zfsvfs, aclp, cr,
1840 vsecp->vsa_aclentp, !(mask & VSA_ACE_ALLTYPES));
1842 zfs_acl_node_t *aclnode;
1843 void *start = vsecp->vsa_aclentp;
1845 for (aclnode = list_head(&aclp->z_acl); aclnode;
1846 aclnode = list_next(&aclp->z_acl, aclnode)) {
1847 bcopy(aclnode->z_acldata, start,
1849 start = (caddr_t)start + aclnode->z_size;
1851 ASSERT((caddr_t)start - (caddr_t)vsecp->vsa_aclentp ==
1855 if (mask & VSA_ACE_ACLFLAGS) {
1856 vsecp->vsa_aclflags = 0;
1857 if (zp->z_pflags & ZFS_ACL_DEFAULTED)
1858 vsecp->vsa_aclflags |= ACL_DEFAULTED;
1859 if (zp->z_pflags & ZFS_ACL_PROTECTED)
1860 vsecp->vsa_aclflags |= ACL_PROTECTED;
1861 if (zp->z_pflags & ZFS_ACL_AUTO_INHERIT)
1862 vsecp->vsa_aclflags |= ACL_AUTO_INHERIT;
1865 mutex_exit(&zp->z_acl_lock);
1871 zfs_vsec_2_aclp(zfsvfs_t *zfsvfs, vtype_t obj_type,
1872 vsecattr_t *vsecp, cred_t *cr, zfs_fuid_info_t **fuidp, zfs_acl_t **zaclp)
1875 zfs_acl_node_t *aclnode;
1876 int aclcnt = vsecp->vsa_aclcnt;
1879 if (vsecp->vsa_aclcnt > MAX_ACL_ENTRIES || vsecp->vsa_aclcnt <= 0)
1880 return (SET_ERROR(EINVAL));
1882 aclp = zfs_acl_alloc(zfs_acl_version(zfsvfs->z_version));
1885 aclnode = zfs_acl_node_alloc(aclcnt * sizeof (zfs_object_ace_t));
1886 if (aclp->z_version == ZFS_ACL_VERSION_INITIAL) {
1887 if ((error = zfs_copy_ace_2_oldace(obj_type, aclp,
1888 (ace_t *)vsecp->vsa_aclentp, aclnode->z_acldata,
1889 aclcnt, &aclnode->z_size)) != 0) {
1891 zfs_acl_node_free(aclnode);
1895 if ((error = zfs_copy_ace_2_fuid(zfsvfs, obj_type, aclp,
1896 vsecp->vsa_aclentp, aclnode->z_acldata, aclcnt,
1897 &aclnode->z_size, fuidp, cr)) != 0) {
1899 zfs_acl_node_free(aclnode);
1903 aclp->z_acl_bytes = aclnode->z_size;
1904 aclnode->z_ace_count = aclcnt;
1905 aclp->z_acl_count = aclcnt;
1906 list_insert_head(&aclp->z_acl, aclnode);
1909 * If flags are being set then add them to z_hints
1911 if (vsecp->vsa_mask & VSA_ACE_ACLFLAGS) {
1912 if (vsecp->vsa_aclflags & ACL_PROTECTED)
1913 aclp->z_hints |= ZFS_ACL_PROTECTED;
1914 if (vsecp->vsa_aclflags & ACL_DEFAULTED)
1915 aclp->z_hints |= ZFS_ACL_DEFAULTED;
1916 if (vsecp->vsa_aclflags & ACL_AUTO_INHERIT)
1917 aclp->z_hints |= ZFS_ACL_AUTO_INHERIT;
1929 zfs_setacl(znode_t *zp, vsecattr_t *vsecp, boolean_t skipaclchk, cred_t *cr)
1931 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1932 zilog_t *zilog = zfsvfs->z_log;
1933 ulong_t mask = vsecp->vsa_mask & (VSA_ACE | VSA_ACECNT);
1937 zfs_fuid_info_t *fuidp = NULL;
1938 boolean_t fuid_dirtied;
1942 return (SET_ERROR(ENOSYS));
1944 if (zp->z_pflags & ZFS_IMMUTABLE)
1945 return (SET_ERROR(EPERM));
1947 if (error = zfs_zaccess(zp, ACE_WRITE_ACL, 0, skipaclchk, cr))
1950 error = zfs_vsec_2_aclp(zfsvfs, ZTOV(zp)->v_type, vsecp, cr, &fuidp,
1956 * If ACL wide flags aren't being set then preserve any
1959 if (!(vsecp->vsa_mask & VSA_ACE_ACLFLAGS)) {
1961 (zp->z_pflags & V4_ACL_WIDE_FLAGS);
1964 mutex_enter(&zp->z_acl_lock);
1965 mutex_enter(&zp->z_lock);
1967 tx = dmu_tx_create(zfsvfs->z_os);
1969 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE);
1971 fuid_dirtied = zfsvfs->z_fuid_dirty;
1973 zfs_fuid_txhold(zfsvfs, tx);
1976 * If old version and ACL won't fit in bonus and we aren't
1977 * upgrading then take out necessary DMU holds
1980 if ((acl_obj = zfs_external_acl(zp)) != 0) {
1981 if (zfsvfs->z_version >= ZPL_VERSION_FUID &&
1982 zfs_znode_acl_version(zp) <= ZFS_ACL_VERSION_INITIAL) {
1983 dmu_tx_hold_free(tx, acl_obj, 0,
1985 dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0,
1988 dmu_tx_hold_write(tx, acl_obj, 0, aclp->z_acl_bytes);
1990 } else if (!zp->z_is_sa && aclp->z_acl_bytes > ZFS_ACE_SPACE) {
1991 dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, aclp->z_acl_bytes);
1994 zfs_sa_upgrade_txholds(tx, zp);
1995 error = dmu_tx_assign(tx, TXG_NOWAIT);
1997 mutex_exit(&zp->z_acl_lock);
1998 mutex_exit(&zp->z_lock);
2000 if (error == ERESTART) {
2010 error = zfs_aclset_common(zp, aclp, cr, tx);
2012 ASSERT(zp->z_acl_cached == NULL);
2013 zp->z_acl_cached = aclp;
2016 zfs_fuid_sync(zfsvfs, tx);
2018 zfs_log_acl(zilog, tx, zp, vsecp, fuidp);
2021 zfs_fuid_info_free(fuidp);
2024 mutex_exit(&zp->z_lock);
2025 mutex_exit(&zp->z_acl_lock);
2031 * Check accesses of interest (AoI) against attributes of the dataset
2032 * such as read-only. Returns zero if no AoI conflict with dataset
2033 * attributes, otherwise an appropriate errno is returned.
2036 zfs_zaccess_dataset_check(znode_t *zp, uint32_t v4_mode)
2038 if ((v4_mode & WRITE_MASK) &&
2039 (zp->z_zfsvfs->z_vfs->vfs_flag & VFS_RDONLY) &&
2040 (!IS_DEVVP(ZTOV(zp)) ||
2041 (IS_DEVVP(ZTOV(zp)) && (v4_mode & WRITE_MASK_ATTRS)))) {
2042 return (SET_ERROR(EROFS));
2046 * Only check for READONLY on non-directories.
2048 if ((v4_mode & WRITE_MASK_DATA) &&
2049 (((ZTOV(zp)->v_type != VDIR) &&
2050 (zp->z_pflags & (ZFS_READONLY | ZFS_IMMUTABLE))) ||
2051 (ZTOV(zp)->v_type == VDIR &&
2052 (zp->z_pflags & ZFS_IMMUTABLE)))) {
2053 return (SET_ERROR(EPERM));
2057 if ((v4_mode & (ACE_DELETE | ACE_DELETE_CHILD)) &&
2058 (zp->z_pflags & ZFS_NOUNLINK)) {
2059 return (SET_ERROR(EPERM));
2063 * In FreeBSD we allow to modify directory's content is ZFS_NOUNLINK
2064 * (sunlnk) is set. We just don't allow directory removal, which is
2065 * handled in zfs_zaccess_delete().
2067 if ((v4_mode & ACE_DELETE) &&
2068 (zp->z_pflags & ZFS_NOUNLINK)) {
2073 if (((v4_mode & (ACE_READ_DATA|ACE_EXECUTE)) &&
2074 (zp->z_pflags & ZFS_AV_QUARANTINED))) {
2075 return (SET_ERROR(EACCES));
2082 * The primary usage of this function is to loop through all of the
2083 * ACEs in the znode, determining what accesses of interest (AoI) to
2084 * the caller are allowed or denied. The AoI are expressed as bits in
2085 * the working_mode parameter. As each ACE is processed, bits covered
2086 * by that ACE are removed from the working_mode. This removal
2087 * facilitates two things. The first is that when the working mode is
2088 * empty (= 0), we know we've looked at all the AoI. The second is
2089 * that the ACE interpretation rules don't allow a later ACE to undo
2090 * something granted or denied by an earlier ACE. Removing the
2091 * discovered access or denial enforces this rule. At the end of
2092 * processing the ACEs, all AoI that were found to be denied are
2093 * placed into the working_mode, giving the caller a mask of denied
2094 * accesses. Returns:
2095 * 0 if all AoI granted
2096 * EACCESS if the denied mask is non-zero
2097 * other error if abnormal failure (e.g., IO error)
2099 * A secondary usage of the function is to determine if any of the
2100 * AoI are granted. If an ACE grants any access in
2101 * the working_mode, we immediately short circuit out of the function.
2102 * This mode is chosen by setting anyaccess to B_TRUE. The
2103 * working_mode is not a denied access mask upon exit if the function
2104 * is used in this manner.
2107 zfs_zaccess_aces_check(znode_t *zp, uint32_t *working_mode,
2108 boolean_t anyaccess, cred_t *cr)
2110 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
2113 uid_t uid = crgetuid(cr);
2115 uint16_t type, iflags;
2116 uint16_t entry_type;
2117 uint32_t access_mask;
2118 uint32_t deny_mask = 0;
2119 zfs_ace_hdr_t *acep = NULL;
2124 zfs_fuid_map_ids(zp, cr, &fowner, &gowner);
2126 mutex_enter(&zp->z_acl_lock);
2128 error = zfs_acl_node_read(zp, B_FALSE, &aclp, B_FALSE);
2130 mutex_exit(&zp->z_acl_lock);
2134 ASSERT(zp->z_acl_cached);
2136 while (acep = zfs_acl_next_ace(aclp, acep, &who, &access_mask,
2138 uint32_t mask_matched;
2140 if (!zfs_acl_valid_ace_type(type, iflags))
2143 if (ZTOV(zp)->v_type == VDIR && (iflags & ACE_INHERIT_ONLY_ACE))
2146 /* Skip ACE if it does not affect any AoI */
2147 mask_matched = (access_mask & *working_mode);
2151 entry_type = (iflags & ACE_TYPE_FLAGS);
2155 switch (entry_type) {
2163 case ACE_IDENTIFIER_GROUP:
2164 checkit = zfs_groupmember(zfsvfs, who, cr);
2172 if (entry_type == 0) {
2175 newid = zfs_fuid_map_id(zfsvfs, who, cr,
2177 if (newid != IDMAP_WK_CREATOR_OWNER_UID &&
2182 mutex_exit(&zp->z_acl_lock);
2183 return (SET_ERROR(EIO));
2189 DTRACE_PROBE3(zfs__ace__denies,
2191 zfs_ace_hdr_t *, acep,
2192 uint32_t, mask_matched);
2193 deny_mask |= mask_matched;
2195 DTRACE_PROBE3(zfs__ace__allows,
2197 zfs_ace_hdr_t *, acep,
2198 uint32_t, mask_matched);
2200 mutex_exit(&zp->z_acl_lock);
2204 *working_mode &= ~mask_matched;
2208 if (*working_mode == 0)
2212 mutex_exit(&zp->z_acl_lock);
2214 /* Put the found 'denies' back on the working mode */
2216 *working_mode |= deny_mask;
2217 return (SET_ERROR(EACCES));
2218 } else if (*working_mode) {
2226 * Return true if any access whatsoever granted, we don't actually
2227 * care what access is granted.
2230 zfs_has_access(znode_t *zp, cred_t *cr)
2232 uint32_t have = ACE_ALL_PERMS;
2234 if (zfs_zaccess_aces_check(zp, &have, B_TRUE, cr) != 0) {
2237 owner = zfs_fuid_map_id(zp->z_zfsvfs, zp->z_uid, cr, ZFS_OWNER);
2238 return (secpolicy_vnode_any_access(cr, ZTOV(zp), owner) == 0);
2244 zfs_zaccess_common(znode_t *zp, uint32_t v4_mode, uint32_t *working_mode,
2245 boolean_t *check_privs, boolean_t skipaclchk, cred_t *cr)
2247 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
2250 *working_mode = v4_mode;
2251 *check_privs = B_TRUE;
2254 * Short circuit empty requests
2256 if (v4_mode == 0 || zfsvfs->z_replay) {
2261 if ((err = zfs_zaccess_dataset_check(zp, v4_mode)) != 0) {
2262 *check_privs = B_FALSE;
2267 * The caller requested that the ACL check be skipped. This
2268 * would only happen if the caller checked VOP_ACCESS() with a
2269 * 32 bit ACE mask and already had the appropriate permissions.
2276 return (zfs_zaccess_aces_check(zp, working_mode, B_FALSE, cr));
2280 zfs_zaccess_append(znode_t *zp, uint32_t *working_mode, boolean_t *check_privs,
2283 if (*working_mode != ACE_WRITE_DATA)
2284 return (SET_ERROR(EACCES));
2286 return (zfs_zaccess_common(zp, ACE_APPEND_DATA, working_mode,
2287 check_privs, B_FALSE, cr));
2291 zfs_fastaccesschk_execute(znode_t *zdp, cred_t *cr)
2293 boolean_t owner = B_FALSE;
2294 boolean_t groupmbr = B_FALSE;
2296 uid_t uid = crgetuid(cr);
2299 if (zdp->z_pflags & ZFS_AV_QUARANTINED)
2300 return (SET_ERROR(EACCES));
2302 is_attr = ((zdp->z_pflags & ZFS_XATTR) &&
2303 (ZTOV(zdp)->v_type == VDIR));
2308 mutex_enter(&zdp->z_acl_lock);
2310 if (zdp->z_pflags & ZFS_NO_EXECS_DENIED) {
2311 mutex_exit(&zdp->z_acl_lock);
2315 if (FUID_INDEX(zdp->z_uid) != 0 || FUID_INDEX(zdp->z_gid) != 0) {
2316 mutex_exit(&zdp->z_acl_lock);
2320 if (uid == zdp->z_uid) {
2322 if (zdp->z_mode & S_IXUSR) {
2323 mutex_exit(&zdp->z_acl_lock);
2326 mutex_exit(&zdp->z_acl_lock);
2330 if (groupmember(zdp->z_gid, cr)) {
2332 if (zdp->z_mode & S_IXGRP) {
2333 mutex_exit(&zdp->z_acl_lock);
2336 mutex_exit(&zdp->z_acl_lock);
2340 if (!owner && !groupmbr) {
2341 if (zdp->z_mode & S_IXOTH) {
2342 mutex_exit(&zdp->z_acl_lock);
2347 mutex_exit(&zdp->z_acl_lock);
2350 DTRACE_PROBE(zfs__fastpath__execute__access__miss);
2351 ZFS_ENTER(zdp->z_zfsvfs);
2352 error = zfs_zaccess(zdp, ACE_EXECUTE, 0, B_FALSE, cr);
2353 ZFS_EXIT(zdp->z_zfsvfs);
2358 * Determine whether Access should be granted/denied.
2360 * The least priv subsytem is always consulted as a basic privilege
2361 * can define any form of access.
2364 zfs_zaccess(znode_t *zp, int mode, int flags, boolean_t skipaclchk, cred_t *cr)
2366 uint32_t working_mode;
2369 boolean_t check_privs;
2371 znode_t *check_zp = zp;
2375 is_attr = ((zp->z_pflags & ZFS_XATTR) && (ZTOV(zp)->v_type == VDIR));
2379 * In FreeBSD, we don't care about permissions of individual ADS.
2380 * Note that not checking them is not just an optimization - without
2381 * this shortcut, EA operations may bogusly fail with EACCES.
2383 if (zp->z_pflags & ZFS_XATTR)
2387 * If attribute then validate against base file
2392 if ((error = sa_lookup(zp->z_sa_hdl,
2393 SA_ZPL_PARENT(zp->z_zfsvfs), &parent,
2394 sizeof (parent))) != 0)
2397 if ((error = zfs_zget(zp->z_zfsvfs,
2398 parent, &xzp)) != 0) {
2405 * fixup mode to map to xattr perms
2408 if (mode & (ACE_WRITE_DATA|ACE_APPEND_DATA)) {
2409 mode &= ~(ACE_WRITE_DATA|ACE_APPEND_DATA);
2410 mode |= ACE_WRITE_NAMED_ATTRS;
2413 if (mode & (ACE_READ_DATA|ACE_EXECUTE)) {
2414 mode &= ~(ACE_READ_DATA|ACE_EXECUTE);
2415 mode |= ACE_READ_NAMED_ATTRS;
2420 owner = zfs_fuid_map_id(zp->z_zfsvfs, zp->z_uid, cr, ZFS_OWNER);
2422 * Map the bits required to the standard vnode flags VREAD|VWRITE|VEXEC
2423 * in needed_bits. Map the bits mapped by working_mode (currently
2424 * missing) in missing_bits.
2425 * Call secpolicy_vnode_access2() with (needed_bits & ~checkmode),
2430 working_mode = mode;
2431 if ((working_mode & (ACE_READ_ACL|ACE_READ_ATTRIBUTES)) &&
2432 owner == crgetuid(cr))
2433 working_mode &= ~(ACE_READ_ACL|ACE_READ_ATTRIBUTES);
2435 if (working_mode & (ACE_READ_DATA|ACE_READ_NAMED_ATTRS|
2436 ACE_READ_ACL|ACE_READ_ATTRIBUTES|ACE_SYNCHRONIZE))
2437 needed_bits |= VREAD;
2438 if (working_mode & (ACE_WRITE_DATA|ACE_WRITE_NAMED_ATTRS|
2439 ACE_APPEND_DATA|ACE_WRITE_ATTRIBUTES|ACE_SYNCHRONIZE))
2440 needed_bits |= VWRITE;
2441 if (working_mode & ACE_EXECUTE)
2442 needed_bits |= VEXEC;
2444 if ((error = zfs_zaccess_common(check_zp, mode, &working_mode,
2445 &check_privs, skipaclchk, cr)) == 0) {
2448 return (secpolicy_vnode_access2(cr, ZTOV(zp), owner,
2449 needed_bits, needed_bits));
2452 if (error && !check_privs) {
2458 if (error && (flags & V_APPEND)) {
2459 error = zfs_zaccess_append(zp, &working_mode, &check_privs, cr);
2462 if (error && check_privs) {
2463 mode_t checkmode = 0;
2466 * First check for implicit owner permission on
2467 * read_acl/read_attributes
2471 ASSERT(working_mode != 0);
2473 if ((working_mode & (ACE_READ_ACL|ACE_READ_ATTRIBUTES) &&
2474 owner == crgetuid(cr)))
2475 working_mode &= ~(ACE_READ_ACL|ACE_READ_ATTRIBUTES);
2477 if (working_mode & (ACE_READ_DATA|ACE_READ_NAMED_ATTRS|
2478 ACE_READ_ACL|ACE_READ_ATTRIBUTES|ACE_SYNCHRONIZE))
2480 if (working_mode & (ACE_WRITE_DATA|ACE_WRITE_NAMED_ATTRS|
2481 ACE_APPEND_DATA|ACE_WRITE_ATTRIBUTES|ACE_SYNCHRONIZE))
2482 checkmode |= VWRITE;
2483 if (working_mode & ACE_EXECUTE)
2486 error = secpolicy_vnode_access2(cr, ZTOV(check_zp), owner,
2487 needed_bits & ~checkmode, needed_bits);
2489 if (error == 0 && (working_mode & ACE_WRITE_OWNER))
2490 error = secpolicy_vnode_chown(ZTOV(check_zp), cr, owner);
2491 if (error == 0 && (working_mode & ACE_WRITE_ACL))
2492 error = secpolicy_vnode_setdac(ZTOV(check_zp), cr, owner);
2494 if (error == 0 && (working_mode &
2495 (ACE_DELETE|ACE_DELETE_CHILD)))
2496 error = secpolicy_vnode_remove(ZTOV(check_zp), cr);
2498 if (error == 0 && (working_mode & ACE_SYNCHRONIZE)) {
2499 error = secpolicy_vnode_chown(ZTOV(check_zp), cr, owner);
2503 * See if any bits other than those already checked
2504 * for are still present. If so then return EACCES
2506 if (working_mode & ~(ZFS_CHECKED_MASKS)) {
2507 error = SET_ERROR(EACCES);
2510 } else if (error == 0) {
2511 error = secpolicy_vnode_access2(cr, ZTOV(zp), owner,
2512 needed_bits, needed_bits);
2523 * Translate traditional unix VREAD/VWRITE/VEXEC mode into
2524 * native ACL format and call zfs_zaccess()
2527 zfs_zaccess_rwx(znode_t *zp, mode_t mode, int flags, cred_t *cr)
2529 return (zfs_zaccess(zp, zfs_unix_to_v4(mode >> 6), flags, B_FALSE, cr));
2533 * Access function for secpolicy_vnode_setattr
2536 zfs_zaccess_unix(znode_t *zp, mode_t mode, cred_t *cr)
2538 int v4_mode = zfs_unix_to_v4(mode >> 6);
2540 return (zfs_zaccess(zp, v4_mode, 0, B_FALSE, cr));
2544 zfs_delete_final_check(znode_t *zp, znode_t *dzp,
2545 mode_t available_perms, cred_t *cr)
2550 downer = zfs_fuid_map_id(dzp->z_zfsvfs, dzp->z_uid, cr, ZFS_OWNER);
2552 error = secpolicy_vnode_access2(cr, ZTOV(dzp),
2553 downer, available_perms, VWRITE|VEXEC);
2556 error = zfs_sticky_remove_access(dzp, zp, cr);
2562 * Determine whether Access should be granted/deny, without
2563 * consulting least priv subsystem.
2565 * The following chart is the recommended NFSv4 enforcement for
2566 * ability to delete an object.
2568 * -------------------------------------------------------
2569 * | Parent Dir | Target Object Permissions |
2571 * -------------------------------------------------------
2572 * | | ACL Allows | ACL Denies| Delete |
2573 * | | Delete | Delete | unspecified|
2574 * -------------------------------------------------------
2575 * | ACL Allows | Permit | Permit | Permit |
2576 * | DELETE_CHILD | |
2577 * -------------------------------------------------------
2578 * | ACL Denies | Permit | Deny | Deny |
2579 * | DELETE_CHILD | | | |
2580 * -------------------------------------------------------
2581 * | ACL specifies | | | |
2582 * | only allow | Permit | Permit | Permit |
2583 * | write and | | | |
2585 * -------------------------------------------------------
2586 * | ACL denies | | | |
2587 * | write and | Permit | Deny | Deny |
2589 * -------------------------------------------------------
2592 * No search privilege, can't even look up file?
2596 zfs_zaccess_delete(znode_t *dzp, znode_t *zp, cred_t *cr)
2598 uint32_t dzp_working_mode = 0;
2599 uint32_t zp_working_mode = 0;
2600 int dzp_error, zp_error;
2601 mode_t available_perms;
2602 boolean_t dzpcheck_privs = B_TRUE;
2603 boolean_t zpcheck_privs = B_TRUE;
2606 * We want specific DELETE permissions to
2607 * take precedence over WRITE/EXECUTE. We don't
2608 * want an ACL such as this to mess us up.
2609 * user:joe:write_data:deny,user:joe:delete:allow
2611 * However, deny permissions may ultimately be overridden
2612 * by secpolicy_vnode_access().
2614 * We will ask for all of the necessary permissions and then
2615 * look at the working modes from the directory and target object
2616 * to determine what was found.
2619 if (zp->z_pflags & (ZFS_IMMUTABLE | ZFS_NOUNLINK))
2620 return (SET_ERROR(EPERM));
2624 * If the directory permissions allow the delete, we are done.
2626 if ((dzp_error = zfs_zaccess_common(dzp, ACE_DELETE_CHILD,
2627 &dzp_working_mode, &dzpcheck_privs, B_FALSE, cr)) == 0)
2631 * If target object has delete permission then we are done
2633 if ((zp_error = zfs_zaccess_common(zp, ACE_DELETE, &zp_working_mode,
2634 &zpcheck_privs, B_FALSE, cr)) == 0)
2637 ASSERT(dzp_error && zp_error);
2639 if (!dzpcheck_privs)
2647 * If directory returns EACCES then delete_child was denied
2648 * due to deny delete_child. In this case send the request through
2649 * secpolicy_vnode_remove(). We don't use zfs_delete_final_check()
2650 * since that *could* allow the delete based on write/execute permission
2651 * and we want delete permissions to override write/execute.
2654 if (dzp_error == EACCES)
2655 return (secpolicy_vnode_remove(ZTOV(dzp), cr)); /* XXXPJD: s/dzp/zp/ ? */
2659 * only need to see if we have write/execute on directory.
2662 dzp_error = zfs_zaccess_common(dzp, ACE_EXECUTE|ACE_WRITE_DATA,
2663 &dzp_working_mode, &dzpcheck_privs, B_FALSE, cr);
2665 if (dzp_error != 0 && !dzpcheck_privs)
2672 available_perms = (dzp_working_mode & ACE_WRITE_DATA) ? 0 : VWRITE;
2673 available_perms |= (dzp_working_mode & ACE_EXECUTE) ? 0 : VEXEC;
2675 return (zfs_delete_final_check(zp, dzp, available_perms, cr));
2680 zfs_zaccess_rename(znode_t *sdzp, znode_t *szp, znode_t *tdzp,
2681 znode_t *tzp, cred_t *cr)
2686 if (szp->z_pflags & ZFS_AV_QUARANTINED)
2687 return (SET_ERROR(EACCES));
2689 add_perm = (ZTOV(szp)->v_type == VDIR) ?
2690 ACE_ADD_SUBDIRECTORY : ACE_ADD_FILE;
2693 * Rename permissions are combination of delete permission +
2694 * add file/subdir permission.
2696 * BSD operating systems also require write permission
2697 * on the directory being moved from one parent directory
2700 if (ZTOV(szp)->v_type == VDIR && ZTOV(sdzp) != ZTOV(tdzp)) {
2701 if (error = zfs_zaccess(szp, ACE_WRITE_DATA, 0, B_FALSE, cr))
2706 * first make sure we do the delete portion.
2708 * If that succeeds then check for add_file/add_subdir permissions
2711 if (error = zfs_zaccess_delete(sdzp, szp, cr))
2715 * If we have a tzp, see if we can delete it?
2718 if (error = zfs_zaccess_delete(tdzp, tzp, cr))
2723 * Now check for add permissions
2725 error = zfs_zaccess(tdzp, add_perm, 0, B_FALSE, cr);