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 https://opensource.org/licenses/CDDL-1.0.
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 (c) 2013 by Delphix. All rights reserved.
24 * Copyright 2017 Nexenta Systems, Inc. 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_quota.h>
49 #include <sys/zfs_vfsops.h>
51 #include <sys/dnode.h>
54 #include <acl/acl_common.h>
57 #define ALLOW ACE_ACCESS_ALLOWED_ACE_TYPE
58 #define DENY ACE_ACCESS_DENIED_ACE_TYPE
59 #define MAX_ACE_TYPE ACE_SYSTEM_ALARM_CALLBACK_OBJECT_ACE_TYPE
60 #define MIN_ACE_TYPE ALLOW
62 #define OWNING_GROUP (ACE_GROUP|ACE_IDENTIFIER_GROUP)
63 #define EVERYONE_ALLOW_MASK (ACE_READ_ACL|ACE_READ_ATTRIBUTES | \
64 ACE_READ_NAMED_ATTRS|ACE_SYNCHRONIZE)
65 #define EVERYONE_DENY_MASK (ACE_WRITE_ACL|ACE_WRITE_OWNER | \
66 ACE_WRITE_ATTRIBUTES|ACE_WRITE_NAMED_ATTRS)
67 #define OWNER_ALLOW_MASK (ACE_WRITE_ACL | ACE_WRITE_OWNER | \
68 ACE_WRITE_ATTRIBUTES|ACE_WRITE_NAMED_ATTRS)
70 #define ZFS_CHECKED_MASKS (ACE_READ_ACL|ACE_READ_ATTRIBUTES|ACE_READ_DATA| \
71 ACE_READ_NAMED_ATTRS|ACE_WRITE_DATA|ACE_WRITE_ATTRIBUTES| \
72 ACE_WRITE_NAMED_ATTRS|ACE_APPEND_DATA|ACE_EXECUTE|ACE_WRITE_OWNER| \
73 ACE_WRITE_ACL|ACE_DELETE|ACE_DELETE_CHILD|ACE_SYNCHRONIZE)
75 #define WRITE_MASK_DATA (ACE_WRITE_DATA|ACE_APPEND_DATA|ACE_WRITE_NAMED_ATTRS)
76 #define WRITE_MASK_ATTRS (ACE_WRITE_ACL|ACE_WRITE_OWNER|ACE_WRITE_ATTRIBUTES| \
77 ACE_DELETE|ACE_DELETE_CHILD)
78 #define WRITE_MASK (WRITE_MASK_DATA|WRITE_MASK_ATTRS)
80 #define OGE_CLEAR (ACE_READ_DATA|ACE_LIST_DIRECTORY|ACE_WRITE_DATA| \
81 ACE_ADD_FILE|ACE_APPEND_DATA|ACE_ADD_SUBDIRECTORY|ACE_EXECUTE)
83 #define OKAY_MASK_BITS (ACE_READ_DATA|ACE_LIST_DIRECTORY|ACE_WRITE_DATA| \
84 ACE_ADD_FILE|ACE_APPEND_DATA|ACE_ADD_SUBDIRECTORY|ACE_EXECUTE)
86 #define ALL_INHERIT (ACE_FILE_INHERIT_ACE|ACE_DIRECTORY_INHERIT_ACE | \
87 ACE_NO_PROPAGATE_INHERIT_ACE|ACE_INHERIT_ONLY_ACE|ACE_INHERITED_ACE)
89 #define RESTRICTED_CLEAR (ACE_WRITE_ACL|ACE_WRITE_OWNER)
91 #define V4_ACL_WIDE_FLAGS (ZFS_ACL_AUTO_INHERIT|ZFS_ACL_DEFAULTED|\
94 #define ZFS_ACL_WIDE_FLAGS (V4_ACL_WIDE_FLAGS|ZFS_ACL_TRIVIAL|ZFS_INHERIT_ACE|\
97 #define ALL_MODE_EXECS (S_IXUSR | S_IXGRP | S_IXOTH)
100 zfs_ace_v0_get_type(void *acep)
102 return (((zfs_oldace_t *)acep)->z_type);
106 zfs_ace_v0_get_flags(void *acep)
108 return (((zfs_oldace_t *)acep)->z_flags);
112 zfs_ace_v0_get_mask(void *acep)
114 return (((zfs_oldace_t *)acep)->z_access_mask);
118 zfs_ace_v0_get_who(void *acep)
120 return (((zfs_oldace_t *)acep)->z_fuid);
124 zfs_ace_v0_set_type(void *acep, uint16_t type)
126 ((zfs_oldace_t *)acep)->z_type = type;
130 zfs_ace_v0_set_flags(void *acep, uint16_t flags)
132 ((zfs_oldace_t *)acep)->z_flags = flags;
136 zfs_ace_v0_set_mask(void *acep, uint32_t mask)
138 ((zfs_oldace_t *)acep)->z_access_mask = mask;
142 zfs_ace_v0_set_who(void *acep, uint64_t who)
144 ((zfs_oldace_t *)acep)->z_fuid = who;
148 zfs_ace_v0_size(void *acep)
151 return (sizeof (zfs_oldace_t));
155 zfs_ace_v0_abstract_size(void)
157 return (sizeof (zfs_oldace_t));
161 zfs_ace_v0_mask_off(void)
163 return (offsetof(zfs_oldace_t, z_access_mask));
167 zfs_ace_v0_data(void *acep, void **datap)
174 static const acl_ops_t zfs_acl_v0_ops = {
177 zfs_ace_v0_get_flags,
178 zfs_ace_v0_set_flags,
184 zfs_ace_v0_abstract_size,
190 zfs_ace_fuid_get_type(void *acep)
192 return (((zfs_ace_hdr_t *)acep)->z_type);
196 zfs_ace_fuid_get_flags(void *acep)
198 return (((zfs_ace_hdr_t *)acep)->z_flags);
202 zfs_ace_fuid_get_mask(void *acep)
204 return (((zfs_ace_hdr_t *)acep)->z_access_mask);
208 zfs_ace_fuid_get_who(void *args)
211 zfs_ace_t *acep = args;
213 entry_type = acep->z_hdr.z_flags & ACE_TYPE_FLAGS;
215 if (entry_type == ACE_OWNER || entry_type == OWNING_GROUP ||
216 entry_type == ACE_EVERYONE)
218 return (((zfs_ace_t *)acep)->z_fuid);
222 zfs_ace_fuid_set_type(void *acep, uint16_t type)
224 ((zfs_ace_hdr_t *)acep)->z_type = type;
228 zfs_ace_fuid_set_flags(void *acep, uint16_t flags)
230 ((zfs_ace_hdr_t *)acep)->z_flags = flags;
234 zfs_ace_fuid_set_mask(void *acep, uint32_t mask)
236 ((zfs_ace_hdr_t *)acep)->z_access_mask = mask;
240 zfs_ace_fuid_set_who(void *arg, uint64_t who)
242 zfs_ace_t *acep = arg;
244 uint16_t entry_type = acep->z_hdr.z_flags & ACE_TYPE_FLAGS;
246 if (entry_type == ACE_OWNER || entry_type == OWNING_GROUP ||
247 entry_type == ACE_EVERYONE)
253 zfs_ace_fuid_size(void *acep)
255 zfs_ace_hdr_t *zacep = acep;
258 switch (zacep->z_type) {
259 case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE:
260 case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE:
261 case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE:
262 case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE:
263 return (sizeof (zfs_object_ace_t));
267 (((zfs_ace_hdr_t *)acep)->z_flags & ACE_TYPE_FLAGS);
268 if (entry_type == ACE_OWNER ||
269 entry_type == OWNING_GROUP ||
270 entry_type == ACE_EVERYONE)
271 return (sizeof (zfs_ace_hdr_t));
274 return (sizeof (zfs_ace_t));
279 zfs_ace_fuid_abstract_size(void)
281 return (sizeof (zfs_ace_hdr_t));
285 zfs_ace_fuid_mask_off(void)
287 return (offsetof(zfs_ace_hdr_t, z_access_mask));
291 zfs_ace_fuid_data(void *acep, void **datap)
293 zfs_ace_t *zacep = acep;
294 zfs_object_ace_t *zobjp;
296 switch (zacep->z_hdr.z_type) {
297 case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE:
298 case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE:
299 case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE:
300 case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE:
302 *datap = (caddr_t)zobjp + sizeof (zfs_ace_t);
303 return (sizeof (zfs_object_ace_t) - sizeof (zfs_ace_t));
310 static const acl_ops_t zfs_acl_fuid_ops = {
311 zfs_ace_fuid_get_mask,
312 zfs_ace_fuid_set_mask,
313 zfs_ace_fuid_get_flags,
314 zfs_ace_fuid_set_flags,
315 zfs_ace_fuid_get_type,
316 zfs_ace_fuid_set_type,
317 zfs_ace_fuid_get_who,
318 zfs_ace_fuid_set_who,
320 zfs_ace_fuid_abstract_size,
321 zfs_ace_fuid_mask_off,
326 * The following three functions are provided for compatibility with
327 * older ZPL version in order to determine if the file use to have
328 * an external ACL and what version of ACL previously existed on the
329 * file. Would really be nice to not need this, sigh.
332 zfs_external_acl(znode_t *zp)
334 zfs_acl_phys_t acl_phys;
341 * Need to deal with a potential
342 * race where zfs_sa_upgrade could cause
343 * z_isa_sa to change.
345 * If the lookup fails then the state of z_is_sa should have
349 if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_ZNODE_ACL(zp->z_zfsvfs),
350 &acl_phys, sizeof (acl_phys))) == 0)
351 return (acl_phys.z_acl_extern_obj);
354 * after upgrade the SA_ZPL_ZNODE_ACL should have been
358 VERIFY3S(error, ==, ENOENT);
364 * Determine size of ACL in bytes
366 * This is more complicated than it should be since we have to deal
367 * with old external ACLs.
370 zfs_acl_znode_info(znode_t *zp, int *aclsize, int *aclcount,
371 zfs_acl_phys_t *aclphys)
373 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
378 ASSERT(MUTEX_HELD(&zp->z_acl_lock));
380 if ((error = sa_size(zp->z_sa_hdl, SA_ZPL_DACL_ACES(zfsvfs),
384 if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_DACL_COUNT(zfsvfs),
385 &acl_count, sizeof (acl_count))) != 0)
387 *aclcount = acl_count;
389 if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_ZNODE_ACL(zfsvfs),
390 aclphys, sizeof (*aclphys))) != 0)
393 if (aclphys->z_acl_version == ZFS_ACL_VERSION_INITIAL) {
394 *aclsize = ZFS_ACL_SIZE(aclphys->z_acl_size);
395 *aclcount = aclphys->z_acl_size;
397 *aclsize = aclphys->z_acl_size;
398 *aclcount = aclphys->z_acl_count;
405 zfs_znode_acl_version(znode_t *zp)
407 zfs_acl_phys_t acl_phys;
410 return (ZFS_ACL_VERSION_FUID);
415 * Need to deal with a potential
416 * race where zfs_sa_upgrade could cause
417 * z_isa_sa to change.
419 * If the lookup fails then the state of z_is_sa should have
422 if ((error = sa_lookup(zp->z_sa_hdl,
423 SA_ZPL_ZNODE_ACL(zp->z_zfsvfs),
424 &acl_phys, sizeof (acl_phys))) == 0)
425 return (acl_phys.z_acl_version);
428 * After upgrade SA_ZPL_ZNODE_ACL should have
432 VERIFY3S(error, ==, ENOENT);
433 return (ZFS_ACL_VERSION_FUID);
439 zfs_acl_version(int version)
441 if (version < ZPL_VERSION_FUID)
442 return (ZFS_ACL_VERSION_INITIAL);
444 return (ZFS_ACL_VERSION_FUID);
448 zfs_acl_version_zp(znode_t *zp)
450 return (zfs_acl_version(zp->z_zfsvfs->z_version));
454 zfs_acl_alloc(int vers)
458 aclp = kmem_zalloc(sizeof (zfs_acl_t), KM_SLEEP);
459 list_create(&aclp->z_acl, sizeof (zfs_acl_node_t),
460 offsetof(zfs_acl_node_t, z_next));
461 aclp->z_version = vers;
462 if (vers == ZFS_ACL_VERSION_FUID)
463 aclp->z_ops = &zfs_acl_fuid_ops;
465 aclp->z_ops = &zfs_acl_v0_ops;
470 zfs_acl_node_alloc(size_t bytes)
472 zfs_acl_node_t *aclnode;
474 aclnode = kmem_zalloc(sizeof (zfs_acl_node_t), KM_SLEEP);
476 aclnode->z_acldata = kmem_alloc(bytes, KM_SLEEP);
477 aclnode->z_allocdata = aclnode->z_acldata;
478 aclnode->z_allocsize = bytes;
479 aclnode->z_size = bytes;
486 zfs_acl_node_free(zfs_acl_node_t *aclnode)
488 if (aclnode->z_allocsize)
489 kmem_free(aclnode->z_allocdata, aclnode->z_allocsize);
490 kmem_free(aclnode, sizeof (zfs_acl_node_t));
494 zfs_acl_release_nodes(zfs_acl_t *aclp)
496 zfs_acl_node_t *aclnode;
498 while ((aclnode = list_head(&aclp->z_acl))) {
499 list_remove(&aclp->z_acl, aclnode);
500 zfs_acl_node_free(aclnode);
502 aclp->z_acl_count = 0;
503 aclp->z_acl_bytes = 0;
507 zfs_acl_free(zfs_acl_t *aclp)
509 zfs_acl_release_nodes(aclp);
510 list_destroy(&aclp->z_acl);
511 kmem_free(aclp, sizeof (zfs_acl_t));
515 zfs_acl_valid_ace_type(uint_t type, uint_t flags)
522 case ACE_SYSTEM_AUDIT_ACE_TYPE:
523 case ACE_SYSTEM_ALARM_ACE_TYPE:
524 entry_type = flags & ACE_TYPE_FLAGS;
525 return (entry_type == ACE_OWNER ||
526 entry_type == OWNING_GROUP ||
527 entry_type == ACE_EVERYONE || entry_type == 0 ||
528 entry_type == ACE_IDENTIFIER_GROUP);
530 if (type <= MAX_ACE_TYPE)
537 zfs_ace_valid(vtype_t obj_type, zfs_acl_t *aclp, uint16_t type, uint16_t iflags)
540 * first check type of entry
543 if (!zfs_acl_valid_ace_type(type, iflags))
547 case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE:
548 case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE:
549 case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE:
550 case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE:
551 if (aclp->z_version < ZFS_ACL_VERSION_FUID)
553 aclp->z_hints |= ZFS_ACL_OBJ_ACE;
557 * next check inheritance level flags
560 if (obj_type == VDIR &&
561 (iflags & (ACE_FILE_INHERIT_ACE|ACE_DIRECTORY_INHERIT_ACE)))
562 aclp->z_hints |= ZFS_INHERIT_ACE;
564 if (iflags & (ACE_INHERIT_ONLY_ACE|ACE_NO_PROPAGATE_INHERIT_ACE)) {
565 if ((iflags & (ACE_FILE_INHERIT_ACE|
566 ACE_DIRECTORY_INHERIT_ACE)) == 0) {
575 zfs_acl_next_ace(zfs_acl_t *aclp, void *start, uint64_t *who,
576 uint32_t *access_mask, uint16_t *iflags, uint16_t *type)
578 zfs_acl_node_t *aclnode;
580 ASSERT3P(aclp, !=, NULL);
583 aclnode = list_head(&aclp->z_acl);
587 aclp->z_next_ace = aclnode->z_acldata;
588 aclp->z_curr_node = aclnode;
589 aclnode->z_ace_idx = 0;
592 aclnode = aclp->z_curr_node;
597 if (aclnode->z_ace_idx >= aclnode->z_ace_count) {
598 aclnode = list_next(&aclp->z_acl, aclnode);
602 aclp->z_curr_node = aclnode;
603 aclnode->z_ace_idx = 0;
604 aclp->z_next_ace = aclnode->z_acldata;
608 if (aclnode->z_ace_idx < aclnode->z_ace_count) {
609 void *acep = aclp->z_next_ace;
613 * Make sure we don't overstep our bounds
615 ace_size = aclp->z_ops->ace_size(acep);
617 if (((caddr_t)acep + ace_size) >
618 ((caddr_t)aclnode->z_acldata + aclnode->z_size)) {
622 *iflags = aclp->z_ops->ace_flags_get(acep);
623 *type = aclp->z_ops->ace_type_get(acep);
624 *access_mask = aclp->z_ops->ace_mask_get(acep);
625 *who = aclp->z_ops->ace_who_get(acep);
626 aclp->z_next_ace = (caddr_t)aclp->z_next_ace + ace_size;
627 aclnode->z_ace_idx++;
629 return ((void *)acep);
635 zfs_ace_walk(void *datap, uint64_t cookie, int aclcnt,
636 uint16_t *flags, uint16_t *type, uint32_t *mask)
639 zfs_acl_t *aclp = datap;
640 zfs_ace_hdr_t *acep = (zfs_ace_hdr_t *)(uintptr_t)cookie;
643 acep = zfs_acl_next_ace(aclp, acep, &who, mask,
645 return ((uint64_t)(uintptr_t)acep);
649 * Copy ACE to internal ZFS format.
650 * While processing the ACL each ACE will be validated for correctness.
651 * ACE FUIDs will be created later.
654 zfs_copy_ace_2_fuid(zfsvfs_t *zfsvfs, vtype_t obj_type, zfs_acl_t *aclp,
655 void *datap, zfs_ace_t *z_acl, uint64_t aclcnt, size_t *size,
656 zfs_fuid_info_t **fuidp, cred_t *cr)
660 zfs_ace_t *aceptr = z_acl;
662 zfs_object_ace_t *zobjacep;
663 ace_object_t *aceobjp;
665 for (i = 0; i != aclcnt; i++) {
666 aceptr->z_hdr.z_access_mask = acep->a_access_mask;
667 aceptr->z_hdr.z_flags = acep->a_flags;
668 aceptr->z_hdr.z_type = acep->a_type;
669 entry_type = aceptr->z_hdr.z_flags & ACE_TYPE_FLAGS;
670 if (entry_type != ACE_OWNER && entry_type != OWNING_GROUP &&
671 entry_type != ACE_EVERYONE) {
672 aceptr->z_fuid = zfs_fuid_create(zfsvfs, acep->a_who,
673 cr, (entry_type == 0) ?
674 ZFS_ACE_USER : ZFS_ACE_GROUP, fuidp);
678 * Make sure ACE is valid
680 if (zfs_ace_valid(obj_type, aclp, aceptr->z_hdr.z_type,
681 aceptr->z_hdr.z_flags) != B_TRUE)
682 return (SET_ERROR(EINVAL));
684 switch (acep->a_type) {
685 case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE:
686 case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE:
687 case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE:
688 case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE:
689 zobjacep = (zfs_object_ace_t *)aceptr;
690 aceobjp = (ace_object_t *)acep;
692 memcpy(zobjacep->z_object_type, aceobjp->a_obj_type,
693 sizeof (aceobjp->a_obj_type));
694 memcpy(zobjacep->z_inherit_type,
695 aceobjp->a_inherit_obj_type,
696 sizeof (aceobjp->a_inherit_obj_type));
697 acep = (ace_t *)((caddr_t)acep + sizeof (ace_object_t));
700 acep = (ace_t *)((caddr_t)acep + sizeof (ace_t));
703 aceptr = (zfs_ace_t *)((caddr_t)aceptr +
704 aclp->z_ops->ace_size(aceptr));
707 *size = (caddr_t)aceptr - (caddr_t)z_acl;
713 * Copy ZFS ACEs to fixed size ace_t layout
716 zfs_copy_fuid_2_ace(zfsvfs_t *zfsvfs, zfs_acl_t *aclp, cred_t *cr,
717 void *datap, int filter)
720 uint32_t access_mask;
721 uint16_t iflags, type;
722 zfs_ace_hdr_t *zacep = NULL;
724 ace_object_t *objacep;
725 zfs_object_ace_t *zobjacep;
729 while ((zacep = zfs_acl_next_ace(aclp, zacep,
730 &who, &access_mask, &iflags, &type))) {
733 case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE:
734 case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE:
735 case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE:
736 case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE:
740 zobjacep = (zfs_object_ace_t *)zacep;
741 objacep = (ace_object_t *)acep;
742 memcpy(objacep->a_obj_type,
743 zobjacep->z_object_type,
744 sizeof (zobjacep->z_object_type));
745 memcpy(objacep->a_inherit_obj_type,
746 zobjacep->z_inherit_type,
747 sizeof (zobjacep->z_inherit_type));
748 ace_size = sizeof (ace_object_t);
751 ace_size = sizeof (ace_t);
755 entry_type = (iflags & ACE_TYPE_FLAGS);
756 if ((entry_type != ACE_OWNER &&
757 entry_type != OWNING_GROUP &&
758 entry_type != ACE_EVERYONE)) {
759 acep->a_who = zfs_fuid_map_id(zfsvfs, who,
760 cr, (entry_type & ACE_IDENTIFIER_GROUP) ?
761 ZFS_ACE_GROUP : ZFS_ACE_USER);
763 acep->a_who = (uid_t)(int64_t)who;
765 acep->a_access_mask = access_mask;
766 acep->a_flags = iflags;
768 acep = (ace_t *)((caddr_t)acep + ace_size);
773 zfs_copy_ace_2_oldace(vtype_t obj_type, zfs_acl_t *aclp, ace_t *acep,
774 zfs_oldace_t *z_acl, int aclcnt, size_t *size)
777 zfs_oldace_t *aceptr = z_acl;
779 for (i = 0; i != aclcnt; i++, aceptr++) {
780 aceptr->z_access_mask = acep[i].a_access_mask;
781 aceptr->z_type = acep[i].a_type;
782 aceptr->z_flags = acep[i].a_flags;
783 aceptr->z_fuid = acep[i].a_who;
785 * Make sure ACE is valid
787 if (zfs_ace_valid(obj_type, aclp, aceptr->z_type,
788 aceptr->z_flags) != B_TRUE)
789 return (SET_ERROR(EINVAL));
791 *size = (caddr_t)aceptr - (caddr_t)z_acl;
796 * convert old ACL format to new
799 zfs_acl_xform(znode_t *zp, zfs_acl_t *aclp, cred_t *cr)
801 zfs_oldace_t *oldaclp;
803 uint16_t type, iflags;
804 uint32_t access_mask;
807 zfs_acl_node_t *newaclnode;
809 ASSERT3U(aclp->z_version, ==, ZFS_ACL_VERSION_INITIAL);
811 * First create the ACE in a contiguous piece of memory
812 * for zfs_copy_ace_2_fuid().
814 * We only convert an ACL once, so this won't happen
817 oldaclp = kmem_alloc(sizeof (zfs_oldace_t) * aclp->z_acl_count,
820 while ((cookie = zfs_acl_next_ace(aclp, cookie, &who,
821 &access_mask, &iflags, &type))) {
822 oldaclp[i].z_flags = iflags;
823 oldaclp[i].z_type = type;
824 oldaclp[i].z_fuid = who;
825 oldaclp[i++].z_access_mask = access_mask;
828 newaclnode = zfs_acl_node_alloc(aclp->z_acl_count *
829 sizeof (zfs_object_ace_t));
830 aclp->z_ops = &zfs_acl_fuid_ops;
831 VERIFY0(zfs_copy_ace_2_fuid(zp->z_zfsvfs, ZTOV(zp)->v_type, aclp,
832 oldaclp, newaclnode->z_acldata, aclp->z_acl_count,
833 &newaclnode->z_size, NULL, cr));
834 newaclnode->z_ace_count = aclp->z_acl_count;
835 aclp->z_version = ZFS_ACL_VERSION;
836 kmem_free(oldaclp, aclp->z_acl_count * sizeof (zfs_oldace_t));
839 * Release all previous ACL nodes
842 zfs_acl_release_nodes(aclp);
844 list_insert_head(&aclp->z_acl, newaclnode);
846 aclp->z_acl_bytes = newaclnode->z_size;
847 aclp->z_acl_count = newaclnode->z_ace_count;
852 * Convert unix access mask to v4 access mask
855 zfs_unix_to_v4(uint32_t access_mask)
857 uint32_t new_mask = 0;
859 if (access_mask & S_IXOTH)
860 new_mask |= ACE_EXECUTE;
861 if (access_mask & S_IWOTH)
862 new_mask |= ACE_WRITE_DATA;
863 if (access_mask & S_IROTH)
864 new_mask |= ACE_READ_DATA;
869 zfs_set_ace(zfs_acl_t *aclp, void *acep, uint32_t access_mask,
870 uint16_t access_type, uint64_t fuid, uint16_t entry_type)
872 uint16_t type = entry_type & ACE_TYPE_FLAGS;
874 aclp->z_ops->ace_mask_set(acep, access_mask);
875 aclp->z_ops->ace_type_set(acep, access_type);
876 aclp->z_ops->ace_flags_set(acep, entry_type);
877 if ((type != ACE_OWNER && type != OWNING_GROUP &&
878 type != ACE_EVERYONE))
879 aclp->z_ops->ace_who_set(acep, fuid);
883 * Determine mode of file based on ACL.
886 zfs_mode_compute(uint64_t fmode, zfs_acl_t *aclp,
887 uint64_t *pflags, uint64_t fuid, uint64_t fgid)
892 zfs_ace_hdr_t *acep = NULL;
894 uint16_t iflags, type;
895 uint32_t access_mask;
896 boolean_t an_exec_denied = B_FALSE;
898 mode = (fmode & (S_IFMT | S_ISUID | S_ISGID | S_ISVTX));
900 while ((acep = zfs_acl_next_ace(aclp, acep, &who,
901 &access_mask, &iflags, &type))) {
903 if (!zfs_acl_valid_ace_type(type, iflags))
906 entry_type = (iflags & ACE_TYPE_FLAGS);
909 * Skip over any inherit_only ACEs
911 if (iflags & ACE_INHERIT_ONLY_ACE)
914 if (entry_type == ACE_OWNER || (entry_type == 0 &&
916 if ((access_mask & ACE_READ_DATA) &&
917 (!(seen & S_IRUSR))) {
923 if ((access_mask & ACE_WRITE_DATA) &&
924 (!(seen & S_IWUSR))) {
930 if ((access_mask & ACE_EXECUTE) &&
931 (!(seen & S_IXUSR))) {
937 } else if (entry_type == OWNING_GROUP ||
938 (entry_type == ACE_IDENTIFIER_GROUP && who == fgid)) {
939 if ((access_mask & ACE_READ_DATA) &&
940 (!(seen & S_IRGRP))) {
946 if ((access_mask & ACE_WRITE_DATA) &&
947 (!(seen & S_IWGRP))) {
953 if ((access_mask & ACE_EXECUTE) &&
954 (!(seen & S_IXGRP))) {
960 } else if (entry_type == ACE_EVERYONE) {
961 if ((access_mask & ACE_READ_DATA)) {
962 if (!(seen & S_IRUSR)) {
968 if (!(seen & S_IRGRP)) {
974 if (!(seen & S_IROTH)) {
981 if ((access_mask & ACE_WRITE_DATA)) {
982 if (!(seen & S_IWUSR)) {
988 if (!(seen & S_IWGRP)) {
994 if (!(seen & S_IWOTH)) {
1001 if ((access_mask & ACE_EXECUTE)) {
1002 if (!(seen & S_IXUSR)) {
1004 if (type == ALLOW) {
1008 if (!(seen & S_IXGRP)) {
1010 if (type == ALLOW) {
1014 if (!(seen & S_IXOTH)) {
1016 if (type == ALLOW) {
1023 * Only care if this IDENTIFIER_GROUP or
1024 * USER ACE denies execute access to someone,
1025 * mode is not affected
1027 if ((access_mask & ACE_EXECUTE) && type == DENY)
1028 an_exec_denied = B_TRUE;
1033 * Failure to allow is effectively a deny, so execute permission
1034 * is denied if it was never mentioned or if we explicitly
1035 * weren't allowed it.
1037 if (!an_exec_denied &&
1038 ((seen & ALL_MODE_EXECS) != ALL_MODE_EXECS ||
1039 (mode & ALL_MODE_EXECS) != ALL_MODE_EXECS))
1040 an_exec_denied = B_TRUE;
1043 *pflags &= ~ZFS_NO_EXECS_DENIED;
1045 *pflags |= ZFS_NO_EXECS_DENIED;
1051 * Read an external acl object. If the intent is to modify, always
1052 * create a new acl and leave any cached acl in place.
1055 zfs_acl_node_read(znode_t *zp, boolean_t have_lock, zfs_acl_t **aclpp,
1056 boolean_t will_modify)
1061 zfs_acl_node_t *aclnode;
1062 zfs_acl_phys_t znode_acl;
1066 ASSERT(MUTEX_HELD(&zp->z_acl_lock));
1067 if (zp->z_zfsvfs->z_replay == B_FALSE)
1068 ASSERT_VOP_LOCKED(ZTOV(zp), __func__);
1070 if (zp->z_acl_cached && !will_modify) {
1071 *aclpp = zp->z_acl_cached;
1075 version = zfs_znode_acl_version(zp);
1077 if ((error = zfs_acl_znode_info(zp, &aclsize,
1078 &acl_count, &znode_acl)) != 0) {
1082 aclp = zfs_acl_alloc(version);
1084 aclp->z_acl_count = acl_count;
1085 aclp->z_acl_bytes = aclsize;
1087 aclnode = zfs_acl_node_alloc(aclsize);
1088 aclnode->z_ace_count = aclp->z_acl_count;
1089 aclnode->z_size = aclsize;
1092 if (znode_acl.z_acl_extern_obj) {
1093 error = dmu_read(zp->z_zfsvfs->z_os,
1094 znode_acl.z_acl_extern_obj, 0, aclnode->z_size,
1095 aclnode->z_acldata, DMU_READ_PREFETCH);
1097 memcpy(aclnode->z_acldata, znode_acl.z_ace_data,
1101 error = sa_lookup(zp->z_sa_hdl, SA_ZPL_DACL_ACES(zp->z_zfsvfs),
1102 aclnode->z_acldata, aclnode->z_size);
1107 zfs_acl_node_free(aclnode);
1108 /* convert checksum errors into IO errors */
1109 if (error == ECKSUM)
1110 error = SET_ERROR(EIO);
1114 list_insert_head(&aclp->z_acl, aclnode);
1118 zp->z_acl_cached = aclp;
1124 zfs_acl_data_locator(void **dataptr, uint32_t *length, uint32_t buflen,
1125 boolean_t start, void *userdata)
1128 zfs_acl_locator_cb_t *cb = (zfs_acl_locator_cb_t *)userdata;
1131 cb->cb_acl_node = list_head(&cb->cb_aclp->z_acl);
1133 cb->cb_acl_node = list_next(&cb->cb_aclp->z_acl,
1136 ASSERT3P(cb->cb_acl_node, !=, NULL);
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 if (zp->z_zfsvfs->z_replay == B_FALSE) {
1148 ASSERT_VOP_ELOCKED(ZTOV(zp), __func__);
1149 ASSERT_VOP_IN_SEQC(ZTOV(zp));
1151 ASSERT(MUTEX_HELD(&zp->z_acl_lock));
1153 if ((error = zfs_acl_node_read(zp, B_TRUE, &aclp, B_FALSE)) == 0)
1154 zp->z_mode = zfs_mode_compute(zp->z_mode, aclp,
1155 &zp->z_pflags, zp->z_uid, zp->z_gid);
1160 * common code for setting ACLs.
1162 * This function is called from zfs_mode_update, zfs_perm_init, and zfs_setacl.
1163 * zfs_setacl passes a non-NULL inherit pointer (ihp) to indicate that it's
1164 * already checked the acl and knows whether to inherit.
1167 zfs_aclset_common(znode_t *zp, zfs_acl_t *aclp, cred_t *cr, dmu_tx_t *tx)
1170 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1171 dmu_object_type_t otype;
1172 zfs_acl_locator_cb_t locate = { 0 };
1174 sa_bulk_attr_t bulk[5];
1177 zfs_acl_phys_t acl_phys;
1179 if (zp->z_zfsvfs->z_replay == B_FALSE) {
1180 ASSERT_VOP_IN_SEQC(ZTOV(zp));
1185 mode = zfs_mode_compute(mode, aclp, &zp->z_pflags,
1186 zp->z_uid, zp->z_gid);
1189 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs), NULL,
1190 &mode, sizeof (mode));
1191 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL,
1192 &zp->z_pflags, sizeof (zp->z_pflags));
1193 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL,
1194 &ctime, sizeof (ctime));
1196 if (zp->z_acl_cached) {
1197 zfs_acl_free(zp->z_acl_cached);
1198 zp->z_acl_cached = NULL;
1204 if (!zfsvfs->z_use_fuids) {
1205 otype = DMU_OT_OLDACL;
1207 if ((aclp->z_version == ZFS_ACL_VERSION_INITIAL) &&
1208 (zfsvfs->z_version >= ZPL_VERSION_FUID))
1209 zfs_acl_xform(zp, aclp, cr);
1210 ASSERT3U(aclp->z_version, >=, ZFS_ACL_VERSION_FUID);
1215 * Arrgh, we have to handle old on disk format
1216 * as well as newer (preferred) SA format.
1219 if (zp->z_is_sa) { /* the easy case, just update the ACL attribute */
1220 locate.cb_aclp = aclp;
1221 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_DACL_ACES(zfsvfs),
1222 zfs_acl_data_locator, &locate, aclp->z_acl_bytes);
1223 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_DACL_COUNT(zfsvfs),
1224 NULL, &aclp->z_acl_count, sizeof (uint64_t));
1225 } else { /* Painful legacy way */
1226 zfs_acl_node_t *aclnode;
1230 if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_ZNODE_ACL(zfsvfs),
1231 &acl_phys, sizeof (acl_phys))) != 0)
1234 aoid = acl_phys.z_acl_extern_obj;
1236 if (aclp->z_acl_bytes > ZFS_ACE_SPACE) {
1238 * If ACL was previously external and we are now
1239 * converting to new ACL format then release old
1240 * ACL object and create a new one.
1243 aclp->z_version != acl_phys.z_acl_version) {
1244 error = dmu_object_free(zfsvfs->z_os, aoid, tx);
1250 aoid = dmu_object_alloc(zfsvfs->z_os,
1251 otype, aclp->z_acl_bytes,
1252 otype == DMU_OT_ACL ?
1253 DMU_OT_SYSACL : DMU_OT_NONE,
1254 otype == DMU_OT_ACL ?
1255 DN_OLD_MAX_BONUSLEN : 0, tx);
1257 (void) dmu_object_set_blocksize(zfsvfs->z_os,
1258 aoid, aclp->z_acl_bytes, 0, tx);
1260 acl_phys.z_acl_extern_obj = aoid;
1261 for (aclnode = list_head(&aclp->z_acl); aclnode;
1262 aclnode = list_next(&aclp->z_acl, aclnode)) {
1263 if (aclnode->z_ace_count == 0)
1265 dmu_write(zfsvfs->z_os, aoid, off,
1266 aclnode->z_size, aclnode->z_acldata, tx);
1267 off += aclnode->z_size;
1270 void *start = acl_phys.z_ace_data;
1272 * Migrating back embedded?
1274 if (acl_phys.z_acl_extern_obj) {
1275 error = dmu_object_free(zfsvfs->z_os,
1276 acl_phys.z_acl_extern_obj, tx);
1279 acl_phys.z_acl_extern_obj = 0;
1282 for (aclnode = list_head(&aclp->z_acl); aclnode;
1283 aclnode = list_next(&aclp->z_acl, aclnode)) {
1284 if (aclnode->z_ace_count == 0)
1286 memcpy(start, aclnode->z_acldata,
1288 start = (caddr_t)start + aclnode->z_size;
1292 * If Old version then swap count/bytes to match old
1293 * layout of znode_acl_phys_t.
1295 if (aclp->z_version == ZFS_ACL_VERSION_INITIAL) {
1296 acl_phys.z_acl_size = aclp->z_acl_count;
1297 acl_phys.z_acl_count = aclp->z_acl_bytes;
1299 acl_phys.z_acl_size = aclp->z_acl_bytes;
1300 acl_phys.z_acl_count = aclp->z_acl_count;
1302 acl_phys.z_acl_version = aclp->z_version;
1304 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ZNODE_ACL(zfsvfs), NULL,
1305 &acl_phys, sizeof (acl_phys));
1309 * Replace ACL wide bits, but first clear them.
1311 zp->z_pflags &= ~ZFS_ACL_WIDE_FLAGS;
1313 zp->z_pflags |= aclp->z_hints;
1315 if (ace_trivial_common(aclp, 0, zfs_ace_walk) == 0)
1316 zp->z_pflags |= ZFS_ACL_TRIVIAL;
1318 zfs_tstamp_update_setup(zp, STATE_CHANGED, NULL, ctime);
1319 return (sa_bulk_update(zp->z_sa_hdl, bulk, count, tx));
1323 zfs_acl_chmod(vtype_t vtype, uint64_t mode, boolean_t split, boolean_t trim,
1328 int new_count, new_bytes;
1331 uint16_t iflags, type;
1332 uint32_t access_mask;
1333 zfs_acl_node_t *newnode;
1334 size_t abstract_size = aclp->z_ops->ace_abstract_size();
1337 trivial_acl_t masks;
1339 new_count = new_bytes = 0;
1341 isdir = (vtype == VDIR);
1343 acl_trivial_access_masks((mode_t)mode, isdir, &masks);
1345 newnode = zfs_acl_node_alloc((abstract_size * 6) + aclp->z_acl_bytes);
1347 zacep = newnode->z_acldata;
1349 zfs_set_ace(aclp, zacep, masks.allow0, ALLOW, -1, ACE_OWNER);
1350 zacep = (void *)((uintptr_t)zacep + abstract_size);
1352 new_bytes += abstract_size;
1355 zfs_set_ace(aclp, zacep, masks.deny1, DENY, -1, ACE_OWNER);
1356 zacep = (void *)((uintptr_t)zacep + abstract_size);
1358 new_bytes += abstract_size;
1361 zfs_set_ace(aclp, zacep, masks.deny2, DENY, -1, OWNING_GROUP);
1362 zacep = (void *)((uintptr_t)zacep + abstract_size);
1364 new_bytes += abstract_size;
1367 while ((acep = zfs_acl_next_ace(aclp, acep, &who, &access_mask,
1369 entry_type = (iflags & ACE_TYPE_FLAGS);
1371 * ACEs used to represent the file mode may be divided
1372 * into an equivalent pair of inherit-only and regular
1373 * ACEs, if they are inheritable.
1374 * Skip regular ACEs, which are replaced by the new mode.
1376 if (split && (entry_type == ACE_OWNER ||
1377 entry_type == OWNING_GROUP ||
1378 entry_type == ACE_EVERYONE)) {
1379 if (!isdir || !(iflags &
1380 (ACE_FILE_INHERIT_ACE|ACE_DIRECTORY_INHERIT_ACE)))
1383 * We preserve owner@, group@, or @everyone
1384 * permissions, if they are inheritable, by
1385 * copying them to inherit_only ACEs. This
1386 * prevents inheritable permissions from being
1387 * altered along with the file mode.
1389 iflags |= ACE_INHERIT_ONLY_ACE;
1393 * If this ACL has any inheritable ACEs, mark that in
1394 * the hints (which are later masked into the pflags)
1395 * so create knows to do inheritance.
1397 if (isdir && (iflags &
1398 (ACE_FILE_INHERIT_ACE|ACE_DIRECTORY_INHERIT_ACE)))
1399 aclp->z_hints |= ZFS_INHERIT_ACE;
1401 if ((type != ALLOW && type != DENY) ||
1402 (iflags & ACE_INHERIT_ONLY_ACE)) {
1404 case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE:
1405 case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE:
1406 case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE:
1407 case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE:
1408 aclp->z_hints |= ZFS_ACL_OBJ_ACE;
1413 * Limit permissions granted by ACEs to be no greater
1414 * than permissions of the requested group mode.
1415 * Applies when the "aclmode" property is set to
1418 if ((type == ALLOW) && trim)
1419 access_mask &= masks.group;
1421 zfs_set_ace(aclp, zacep, access_mask, type, who, iflags);
1422 ace_size = aclp->z_ops->ace_size(acep);
1423 zacep = (void *)((uintptr_t)zacep + ace_size);
1425 new_bytes += ace_size;
1427 zfs_set_ace(aclp, zacep, masks.owner, ALLOW, -1, ACE_OWNER);
1428 zacep = (void *)((uintptr_t)zacep + abstract_size);
1429 zfs_set_ace(aclp, zacep, masks.group, ALLOW, -1, OWNING_GROUP);
1430 zacep = (void *)((uintptr_t)zacep + abstract_size);
1431 zfs_set_ace(aclp, zacep, masks.everyone, ALLOW, -1, ACE_EVERYONE);
1434 new_bytes += abstract_size * 3;
1435 zfs_acl_release_nodes(aclp);
1436 aclp->z_acl_count = new_count;
1437 aclp->z_acl_bytes = new_bytes;
1438 newnode->z_ace_count = new_count;
1439 newnode->z_size = new_bytes;
1440 list_insert_tail(&aclp->z_acl, newnode);
1444 zfs_acl_chmod_setattr(znode_t *zp, zfs_acl_t **aclp, uint64_t mode)
1448 mutex_enter(&zp->z_acl_lock);
1449 if (zp->z_zfsvfs->z_replay == B_FALSE)
1450 ASSERT_VOP_ELOCKED(ZTOV(zp), __func__);
1451 if (zp->z_zfsvfs->z_acl_mode == ZFS_ACL_DISCARD)
1452 *aclp = zfs_acl_alloc(zfs_acl_version_zp(zp));
1454 error = zfs_acl_node_read(zp, B_TRUE, aclp, B_TRUE);
1457 (*aclp)->z_hints = zp->z_pflags & V4_ACL_WIDE_FLAGS;
1458 zfs_acl_chmod(ZTOV(zp)->v_type, mode, B_TRUE,
1459 (zp->z_zfsvfs->z_acl_mode == ZFS_ACL_GROUPMASK), *aclp);
1461 mutex_exit(&zp->z_acl_lock);
1467 * Should ACE be inherited?
1470 zfs_ace_can_use(vtype_t vtype, uint16_t acep_flags)
1472 int iflags = (acep_flags & 0xf);
1474 if ((vtype == VDIR) && (iflags & ACE_DIRECTORY_INHERIT_ACE))
1476 else if (iflags & ACE_FILE_INHERIT_ACE)
1477 return (!((vtype == VDIR) &&
1478 (iflags & ACE_NO_PROPAGATE_INHERIT_ACE)));
1483 * inherit inheritable ACEs from parent
1486 zfs_acl_inherit(zfsvfs_t *zfsvfs, vtype_t vtype, zfs_acl_t *paclp,
1487 uint64_t mode, boolean_t *need_chmod)
1491 zfs_acl_node_t *aclnode;
1492 zfs_acl_t *aclp = NULL;
1494 uint32_t access_mask;
1495 uint16_t iflags, newflags, type;
1497 void *data1, *data2;
1498 size_t data1sz, data2sz;
1500 boolean_t isdir = (vtype == VDIR);
1501 boolean_t isreg = (vtype == VREG);
1503 *need_chmod = B_TRUE;
1505 aclp = zfs_acl_alloc(paclp->z_version);
1506 aclinherit = zfsvfs->z_acl_inherit;
1507 if (aclinherit == ZFS_ACL_DISCARD || vtype == VLNK)
1510 while ((pacep = zfs_acl_next_ace(paclp, pacep, &who,
1511 &access_mask, &iflags, &type))) {
1514 * don't inherit bogus ACEs
1516 if (!zfs_acl_valid_ace_type(type, iflags))
1520 * Check if ACE is inheritable by this vnode
1522 if ((aclinherit == ZFS_ACL_NOALLOW && type == ALLOW) ||
1523 !zfs_ace_can_use(vtype, iflags))
1527 * If owner@, group@, or everyone@ inheritable
1528 * then zfs_acl_chmod() isn't needed.
1530 if ((aclinherit == ZFS_ACL_PASSTHROUGH ||
1531 aclinherit == ZFS_ACL_PASSTHROUGH_X) &&
1532 ((iflags & (ACE_OWNER|ACE_EVERYONE)) ||
1533 ((iflags & OWNING_GROUP) == OWNING_GROUP)) &&
1534 (isreg || (isdir && (iflags & ACE_DIRECTORY_INHERIT_ACE))))
1535 *need_chmod = B_FALSE;
1538 * Strip inherited execute permission from file if
1541 if (aclinherit == ZFS_ACL_PASSTHROUGH_X && type == ALLOW &&
1542 !isdir && ((mode & (S_IXUSR|S_IXGRP|S_IXOTH)) == 0)) {
1543 access_mask &= ~ACE_EXECUTE;
1547 * Strip write_acl and write_owner from permissions
1548 * when inheriting an ACE
1550 if (aclinherit == ZFS_ACL_RESTRICTED && type == ALLOW) {
1551 access_mask &= ~RESTRICTED_CLEAR;
1554 ace_size = aclp->z_ops->ace_size(pacep);
1555 aclnode = zfs_acl_node_alloc(ace_size);
1556 list_insert_tail(&aclp->z_acl, aclnode);
1557 acep = aclnode->z_acldata;
1559 zfs_set_ace(aclp, acep, access_mask, type,
1560 who, iflags|ACE_INHERITED_ACE);
1563 * Copy special opaque data if any
1565 if ((data1sz = paclp->z_ops->ace_data(pacep, &data1)) != 0) {
1566 data2sz = aclp->z_ops->ace_data(acep, &data2);
1567 VERIFY3U(data2sz, ==, data1sz);
1568 memcpy(data2, data1, data2sz);
1571 aclp->z_acl_count++;
1572 aclnode->z_ace_count++;
1573 aclp->z_acl_bytes += aclnode->z_size;
1574 newflags = aclp->z_ops->ace_flags_get(acep);
1577 * If ACE is not to be inherited further, or if the vnode is
1578 * not a directory, remove all inheritance flags
1580 if (!isdir || (iflags & ACE_NO_PROPAGATE_INHERIT_ACE)) {
1581 newflags &= ~ALL_INHERIT;
1582 aclp->z_ops->ace_flags_set(acep,
1583 newflags|ACE_INHERITED_ACE);
1588 * This directory has an inheritable ACE
1590 aclp->z_hints |= ZFS_INHERIT_ACE;
1593 * If only FILE_INHERIT is set then turn on
1596 if ((iflags & (ACE_FILE_INHERIT_ACE |
1597 ACE_DIRECTORY_INHERIT_ACE)) == ACE_FILE_INHERIT_ACE) {
1598 newflags |= ACE_INHERIT_ONLY_ACE;
1599 aclp->z_ops->ace_flags_set(acep,
1600 newflags|ACE_INHERITED_ACE);
1602 newflags &= ~ACE_INHERIT_ONLY_ACE;
1603 aclp->z_ops->ace_flags_set(acep,
1604 newflags|ACE_INHERITED_ACE);
1607 if (zfsvfs->z_acl_mode == ZFS_ACL_RESTRICTED &&
1608 aclp->z_acl_count != 0) {
1609 *need_chmod = B_FALSE;
1616 * Create file system object initial permissions
1617 * including inheritable ACEs.
1618 * Also, create FUIDs for owner and group.
1621 zfs_acl_ids_create(znode_t *dzp, int flag, vattr_t *vap, cred_t *cr,
1622 vsecattr_t *vsecp, zfs_acl_ids_t *acl_ids, zuserns_t *mnt_ns)
1625 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
1628 boolean_t need_chmod = B_TRUE;
1629 boolean_t trim = B_FALSE;
1630 boolean_t inherited = B_FALSE;
1632 if ((flag & IS_ROOT_NODE) == 0) {
1633 if (zfsvfs->z_replay == B_FALSE)
1634 ASSERT_VOP_ELOCKED(ZTOV(dzp), __func__);
1636 ASSERT3P(dzp->z_vnode, ==, NULL);
1637 memset(acl_ids, 0, sizeof (zfs_acl_ids_t));
1638 acl_ids->z_mode = MAKEIMODE(vap->va_type, vap->va_mode);
1641 if ((error = zfs_vsec_2_aclp(zfsvfs, vap->va_type, vsecp, cr,
1642 &acl_ids->z_fuidp, &acl_ids->z_aclp)) != 0)
1645 * Determine uid and gid.
1647 if ((flag & IS_ROOT_NODE) || zfsvfs->z_replay ||
1648 ((flag & IS_XATTR) && (vap->va_type == VDIR))) {
1649 acl_ids->z_fuid = zfs_fuid_create(zfsvfs,
1650 (uint64_t)vap->va_uid, cr,
1651 ZFS_OWNER, &acl_ids->z_fuidp);
1652 acl_ids->z_fgid = zfs_fuid_create(zfsvfs,
1653 (uint64_t)vap->va_gid, cr,
1654 ZFS_GROUP, &acl_ids->z_fuidp);
1657 uid_t id = crgetuid(cr);
1658 if (IS_EPHEMERAL(id))
1660 acl_ids->z_fuid = (uint64_t)id;
1661 acl_ids->z_fgid = 0;
1662 if (vap->va_mask & AT_GID) {
1663 acl_ids->z_fgid = zfs_fuid_create(zfsvfs,
1664 (uint64_t)vap->va_gid,
1665 cr, ZFS_GROUP, &acl_ids->z_fuidp);
1667 if (acl_ids->z_fgid != dzp->z_gid &&
1668 !groupmember(vap->va_gid, cr) &&
1669 secpolicy_vnode_create_gid(cr) != 0)
1670 acl_ids->z_fgid = 0;
1672 if (acl_ids->z_fgid == 0) {
1676 acl_ids->z_fgid = dzp->z_gid;
1677 gid = zfs_fuid_map_id(zfsvfs, acl_ids->z_fgid,
1680 if (zfsvfs->z_use_fuids &&
1681 IS_EPHEMERAL(acl_ids->z_fgid)) {
1683 zfs_fuid_idx_domain(&zfsvfs->z_fuid_idx,
1684 FUID_INDEX(acl_ids->z_fgid));
1685 rid = FUID_RID(acl_ids->z_fgid);
1686 zfs_fuid_node_add(&acl_ids->z_fuidp,
1687 domain, rid, FUID_INDEX(acl_ids->z_fgid),
1688 acl_ids->z_fgid, ZFS_GROUP);
1694 * If we're creating a directory, and the parent directory has the
1695 * set-GID bit set, set in on the new directory.
1696 * Otherwise, if the user is neither privileged nor a member of the
1697 * file's new group, clear the file's set-GID bit.
1700 if (!(flag & IS_ROOT_NODE) && (dzp->z_mode & S_ISGID) &&
1701 (vap->va_type == VDIR)) {
1702 acl_ids->z_mode |= S_ISGID;
1704 if ((acl_ids->z_mode & S_ISGID) &&
1705 secpolicy_vnode_setids_setgids(ZTOV(dzp), cr, gid) != 0)
1706 acl_ids->z_mode &= ~S_ISGID;
1709 if (acl_ids->z_aclp == NULL) {
1710 mutex_enter(&dzp->z_acl_lock);
1711 if (!(flag & IS_ROOT_NODE) &&
1712 (dzp->z_pflags & ZFS_INHERIT_ACE) &&
1713 !(dzp->z_pflags & ZFS_XATTR)) {
1714 VERIFY0(zfs_acl_node_read(dzp, B_TRUE,
1716 acl_ids->z_aclp = zfs_acl_inherit(zfsvfs,
1717 vap->va_type, paclp, acl_ids->z_mode, &need_chmod);
1721 zfs_acl_alloc(zfs_acl_version_zp(dzp));
1722 acl_ids->z_aclp->z_hints |= ZFS_ACL_TRIVIAL;
1724 mutex_exit(&dzp->z_acl_lock);
1727 if (vap->va_type == VDIR)
1728 acl_ids->z_aclp->z_hints |=
1729 ZFS_ACL_AUTO_INHERIT;
1731 if (zfsvfs->z_acl_mode == ZFS_ACL_GROUPMASK &&
1732 zfsvfs->z_acl_inherit != ZFS_ACL_PASSTHROUGH &&
1733 zfsvfs->z_acl_inherit != ZFS_ACL_PASSTHROUGH_X)
1735 zfs_acl_chmod(vap->va_type, acl_ids->z_mode, B_FALSE,
1736 trim, acl_ids->z_aclp);
1740 if (inherited || vsecp) {
1741 acl_ids->z_mode = zfs_mode_compute(acl_ids->z_mode,
1742 acl_ids->z_aclp, &acl_ids->z_aclp->z_hints,
1743 acl_ids->z_fuid, acl_ids->z_fgid);
1744 if (ace_trivial_common(acl_ids->z_aclp, 0, zfs_ace_walk) == 0)
1745 acl_ids->z_aclp->z_hints |= ZFS_ACL_TRIVIAL;
1752 * Free ACL and fuid_infop, but not the acl_ids structure
1755 zfs_acl_ids_free(zfs_acl_ids_t *acl_ids)
1757 if (acl_ids->z_aclp)
1758 zfs_acl_free(acl_ids->z_aclp);
1759 if (acl_ids->z_fuidp)
1760 zfs_fuid_info_free(acl_ids->z_fuidp);
1761 acl_ids->z_aclp = NULL;
1762 acl_ids->z_fuidp = NULL;
1766 zfs_acl_ids_overquota(zfsvfs_t *zv, zfs_acl_ids_t *acl_ids, uint64_t projid)
1768 return (zfs_id_overquota(zv, DMU_USERUSED_OBJECT, acl_ids->z_fuid) ||
1769 zfs_id_overquota(zv, DMU_GROUPUSED_OBJECT, acl_ids->z_fgid) ||
1770 (projid != ZFS_DEFAULT_PROJID && projid != ZFS_INVALID_PROJID &&
1771 zfs_id_overquota(zv, DMU_PROJECTUSED_OBJECT, projid)));
1775 * Retrieve a file's ACL
1778 zfs_getacl(znode_t *zp, vsecattr_t *vsecp, boolean_t skipaclchk, cred_t *cr)
1786 mask = vsecp->vsa_mask & (VSA_ACE | VSA_ACECNT |
1787 VSA_ACE_ACLFLAGS | VSA_ACE_ALLTYPES);
1790 return (SET_ERROR(ENOSYS));
1792 if ((error = zfs_zaccess(zp, ACE_READ_ACL, 0, skipaclchk, cr, NULL)))
1795 mutex_enter(&zp->z_acl_lock);
1797 if (zp->z_zfsvfs->z_replay == B_FALSE)
1798 ASSERT_VOP_LOCKED(ZTOV(zp), __func__);
1799 error = zfs_acl_node_read(zp, B_TRUE, &aclp, B_FALSE);
1801 mutex_exit(&zp->z_acl_lock);
1806 * Scan ACL to determine number of ACEs
1808 if ((zp->z_pflags & ZFS_ACL_OBJ_ACE) && !(mask & VSA_ACE_ALLTYPES)) {
1811 uint32_t access_mask;
1812 uint16_t type, iflags;
1814 while ((zacep = zfs_acl_next_ace(aclp, zacep,
1815 &who, &access_mask, &iflags, &type))) {
1817 case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE:
1818 case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE:
1819 case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE:
1820 case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE:
1827 vsecp->vsa_aclcnt = count;
1829 count = (int)aclp->z_acl_count;
1831 if (mask & VSA_ACECNT) {
1832 vsecp->vsa_aclcnt = count;
1835 if (mask & VSA_ACE) {
1838 aclsz = count * sizeof (ace_t) +
1839 sizeof (ace_object_t) * largeace;
1841 vsecp->vsa_aclentp = kmem_alloc(aclsz, KM_SLEEP);
1842 vsecp->vsa_aclentsz = aclsz;
1844 if (aclp->z_version == ZFS_ACL_VERSION_FUID)
1845 zfs_copy_fuid_2_ace(zp->z_zfsvfs, aclp, cr,
1846 vsecp->vsa_aclentp, !(mask & VSA_ACE_ALLTYPES));
1848 zfs_acl_node_t *aclnode;
1849 void *start = vsecp->vsa_aclentp;
1851 for (aclnode = list_head(&aclp->z_acl); aclnode;
1852 aclnode = list_next(&aclp->z_acl, aclnode)) {
1853 memcpy(start, aclnode->z_acldata,
1855 start = (caddr_t)start + aclnode->z_size;
1857 ASSERT3U((caddr_t)start - (caddr_t)vsecp->vsa_aclentp,
1858 ==, aclp->z_acl_bytes);
1861 if (mask & VSA_ACE_ACLFLAGS) {
1862 vsecp->vsa_aclflags = 0;
1863 if (zp->z_pflags & ZFS_ACL_DEFAULTED)
1864 vsecp->vsa_aclflags |= ACL_DEFAULTED;
1865 if (zp->z_pflags & ZFS_ACL_PROTECTED)
1866 vsecp->vsa_aclflags |= ACL_PROTECTED;
1867 if (zp->z_pflags & ZFS_ACL_AUTO_INHERIT)
1868 vsecp->vsa_aclflags |= ACL_AUTO_INHERIT;
1871 mutex_exit(&zp->z_acl_lock);
1877 zfs_vsec_2_aclp(zfsvfs_t *zfsvfs, umode_t obj_type,
1878 vsecattr_t *vsecp, cred_t *cr, zfs_fuid_info_t **fuidp, zfs_acl_t **zaclp)
1881 zfs_acl_node_t *aclnode;
1882 int aclcnt = vsecp->vsa_aclcnt;
1885 if (vsecp->vsa_aclcnt > MAX_ACL_ENTRIES || vsecp->vsa_aclcnt <= 0)
1886 return (SET_ERROR(EINVAL));
1888 aclp = zfs_acl_alloc(zfs_acl_version(zfsvfs->z_version));
1891 aclnode = zfs_acl_node_alloc(aclcnt * sizeof (zfs_object_ace_t));
1892 if (aclp->z_version == ZFS_ACL_VERSION_INITIAL) {
1893 if ((error = zfs_copy_ace_2_oldace(obj_type, aclp,
1894 (ace_t *)vsecp->vsa_aclentp, aclnode->z_acldata,
1895 aclcnt, &aclnode->z_size)) != 0) {
1897 zfs_acl_node_free(aclnode);
1901 if ((error = zfs_copy_ace_2_fuid(zfsvfs, obj_type, aclp,
1902 vsecp->vsa_aclentp, aclnode->z_acldata, aclcnt,
1903 &aclnode->z_size, fuidp, cr)) != 0) {
1905 zfs_acl_node_free(aclnode);
1909 aclp->z_acl_bytes = aclnode->z_size;
1910 aclnode->z_ace_count = aclcnt;
1911 aclp->z_acl_count = aclcnt;
1912 list_insert_head(&aclp->z_acl, aclnode);
1915 * If flags are being set then add them to z_hints
1917 if (vsecp->vsa_mask & VSA_ACE_ACLFLAGS) {
1918 if (vsecp->vsa_aclflags & ACL_PROTECTED)
1919 aclp->z_hints |= ZFS_ACL_PROTECTED;
1920 if (vsecp->vsa_aclflags & ACL_DEFAULTED)
1921 aclp->z_hints |= ZFS_ACL_DEFAULTED;
1922 if (vsecp->vsa_aclflags & ACL_AUTO_INHERIT)
1923 aclp->z_hints |= ZFS_ACL_AUTO_INHERIT;
1935 zfs_setacl(znode_t *zp, vsecattr_t *vsecp, boolean_t skipaclchk, cred_t *cr)
1937 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1938 zilog_t *zilog = zfsvfs->z_log;
1939 ulong_t mask = vsecp->vsa_mask & (VSA_ACE | VSA_ACECNT);
1943 zfs_fuid_info_t *fuidp = NULL;
1944 boolean_t fuid_dirtied;
1947 if (zp->z_zfsvfs->z_replay == B_FALSE)
1948 ASSERT_VOP_ELOCKED(ZTOV(zp), __func__);
1950 return (SET_ERROR(ENOSYS));
1952 if (zp->z_pflags & ZFS_IMMUTABLE)
1953 return (SET_ERROR(EPERM));
1955 if ((error = zfs_zaccess(zp, ACE_WRITE_ACL, 0, skipaclchk, cr, NULL)))
1958 error = zfs_vsec_2_aclp(zfsvfs, ZTOV(zp)->v_type, vsecp, cr, &fuidp,
1964 * If ACL wide flags aren't being set then preserve any
1967 if (!(vsecp->vsa_mask & VSA_ACE_ACLFLAGS)) {
1969 (zp->z_pflags & V4_ACL_WIDE_FLAGS);
1972 mutex_enter(&zp->z_acl_lock);
1974 tx = dmu_tx_create(zfsvfs->z_os);
1976 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE);
1978 fuid_dirtied = zfsvfs->z_fuid_dirty;
1980 zfs_fuid_txhold(zfsvfs, tx);
1983 * If old version and ACL won't fit in bonus and we aren't
1984 * upgrading then take out necessary DMU holds
1987 if ((acl_obj = zfs_external_acl(zp)) != 0) {
1988 if (zfsvfs->z_version >= ZPL_VERSION_FUID &&
1989 zfs_znode_acl_version(zp) <= ZFS_ACL_VERSION_INITIAL) {
1990 dmu_tx_hold_free(tx, acl_obj, 0,
1992 dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0,
1995 dmu_tx_hold_write(tx, acl_obj, 0, aclp->z_acl_bytes);
1997 } else if (!zp->z_is_sa && aclp->z_acl_bytes > ZFS_ACE_SPACE) {
1998 dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, aclp->z_acl_bytes);
2001 zfs_sa_upgrade_txholds(tx, zp);
2002 error = dmu_tx_assign(tx, TXG_NOWAIT);
2004 mutex_exit(&zp->z_acl_lock);
2006 if (error == ERESTART) {
2016 error = zfs_aclset_common(zp, aclp, cr, tx);
2018 ASSERT3P(zp->z_acl_cached, ==, NULL);
2019 zp->z_acl_cached = aclp;
2022 zfs_fuid_sync(zfsvfs, tx);
2024 zfs_log_acl(zilog, tx, zp, vsecp, fuidp);
2027 zfs_fuid_info_free(fuidp);
2029 mutex_exit(&zp->z_acl_lock);
2035 * Check accesses of interest (AoI) against attributes of the dataset
2036 * such as read-only. Returns zero if no AoI conflict with dataset
2037 * attributes, otherwise an appropriate errno is returned.
2040 zfs_zaccess_dataset_check(znode_t *zp, uint32_t v4_mode)
2042 if ((v4_mode & WRITE_MASK) &&
2043 (zp->z_zfsvfs->z_vfs->vfs_flag & VFS_RDONLY) &&
2044 (!IS_DEVVP(ZTOV(zp)) ||
2045 (IS_DEVVP(ZTOV(zp)) && (v4_mode & WRITE_MASK_ATTRS)))) {
2046 return (SET_ERROR(EROFS));
2050 * Intentionally allow ZFS_READONLY through here.
2051 * See zfs_zaccess_common().
2053 if ((v4_mode & WRITE_MASK_DATA) &&
2054 (zp->z_pflags & ZFS_IMMUTABLE)) {
2055 return (SET_ERROR(EPERM));
2059 * In FreeBSD we allow to modify directory's content is ZFS_NOUNLINK
2060 * (sunlnk) is set. We just don't allow directory removal, which is
2061 * handled in zfs_zaccess_delete().
2063 if ((v4_mode & ACE_DELETE) &&
2064 (zp->z_pflags & ZFS_NOUNLINK)) {
2068 if (((v4_mode & (ACE_READ_DATA|ACE_EXECUTE)) &&
2069 (zp->z_pflags & ZFS_AV_QUARANTINED))) {
2070 return (SET_ERROR(EACCES));
2077 * The primary usage of this function is to loop through all of the
2078 * ACEs in the znode, determining what accesses of interest (AoI) to
2079 * the caller are allowed or denied. The AoI are expressed as bits in
2080 * the working_mode parameter. As each ACE is processed, bits covered
2081 * by that ACE are removed from the working_mode. This removal
2082 * facilitates two things. The first is that when the working mode is
2083 * empty (= 0), we know we've looked at all the AoI. The second is
2084 * that the ACE interpretation rules don't allow a later ACE to undo
2085 * something granted or denied by an earlier ACE. Removing the
2086 * discovered access or denial enforces this rule. At the end of
2087 * processing the ACEs, all AoI that were found to be denied are
2088 * placed into the working_mode, giving the caller a mask of denied
2089 * accesses. Returns:
2090 * 0 if all AoI granted
2091 * EACCESS if the denied mask is non-zero
2092 * other error if abnormal failure (e.g., IO error)
2094 * A secondary usage of the function is to determine if any of the
2095 * AoI are granted. If an ACE grants any access in
2096 * the working_mode, we immediately short circuit out of the function.
2097 * This mode is chosen by setting anyaccess to B_TRUE. The
2098 * working_mode is not a denied access mask upon exit if the function
2099 * is used in this manner.
2102 zfs_zaccess_aces_check(znode_t *zp, uint32_t *working_mode,
2103 boolean_t anyaccess, cred_t *cr)
2105 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
2108 uid_t uid = crgetuid(cr);
2110 uint16_t type, iflags;
2111 uint16_t entry_type;
2112 uint32_t access_mask;
2113 uint32_t deny_mask = 0;
2114 zfs_ace_hdr_t *acep = NULL;
2119 zfs_fuid_map_ids(zp, cr, &fowner, &gowner);
2121 mutex_enter(&zp->z_acl_lock);
2123 if (zp->z_zfsvfs->z_replay == B_FALSE)
2124 ASSERT_VOP_LOCKED(ZTOV(zp), __func__);
2125 error = zfs_acl_node_read(zp, B_TRUE, &aclp, B_FALSE);
2127 mutex_exit(&zp->z_acl_lock);
2131 ASSERT3P(zp->z_acl_cached, !=, NULL);
2133 while ((acep = zfs_acl_next_ace(aclp, acep, &who, &access_mask,
2135 uint32_t mask_matched;
2137 if (!zfs_acl_valid_ace_type(type, iflags))
2140 if (ZTOV(zp)->v_type == VDIR && (iflags & ACE_INHERIT_ONLY_ACE))
2143 /* Skip ACE if it does not affect any AoI */
2144 mask_matched = (access_mask & *working_mode);
2148 entry_type = (iflags & ACE_TYPE_FLAGS);
2152 switch (entry_type) {
2160 case ACE_IDENTIFIER_GROUP:
2161 checkit = zfs_groupmember(zfsvfs, who, cr);
2169 if (entry_type == 0) {
2172 newid = zfs_fuid_map_id(zfsvfs, who, cr,
2174 if (newid != UID_NOBODY &&
2179 mutex_exit(&zp->z_acl_lock);
2180 return (SET_ERROR(EIO));
2186 DTRACE_PROBE3(zfs__ace__denies,
2188 zfs_ace_hdr_t *, acep,
2189 uint32_t, mask_matched);
2190 deny_mask |= mask_matched;
2192 DTRACE_PROBE3(zfs__ace__allows,
2194 zfs_ace_hdr_t *, acep,
2195 uint32_t, mask_matched);
2197 mutex_exit(&zp->z_acl_lock);
2201 *working_mode &= ~mask_matched;
2205 if (*working_mode == 0)
2209 mutex_exit(&zp->z_acl_lock);
2211 /* Put the found 'denies' back on the working mode */
2213 *working_mode |= deny_mask;
2214 return (SET_ERROR(EACCES));
2215 } else if (*working_mode) {
2223 * Return true if any access whatsoever granted, we don't actually
2224 * care what access is granted.
2227 zfs_has_access(znode_t *zp, cred_t *cr)
2229 uint32_t have = ACE_ALL_PERMS;
2231 if (zfs_zaccess_aces_check(zp, &have, B_TRUE, cr) != 0) {
2234 owner = zfs_fuid_map_id(zp->z_zfsvfs, zp->z_uid, cr, ZFS_OWNER);
2235 return (secpolicy_vnode_any_access(cr, ZTOV(zp), owner) == 0);
2241 zfs_zaccess_common(znode_t *zp, uint32_t v4_mode, uint32_t *working_mode,
2242 boolean_t *check_privs, boolean_t skipaclchk, cred_t *cr)
2244 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
2247 *working_mode = v4_mode;
2248 *check_privs = B_TRUE;
2251 * Short circuit empty requests
2253 if (v4_mode == 0 || zfsvfs->z_replay) {
2258 if ((err = zfs_zaccess_dataset_check(zp, v4_mode)) != 0) {
2259 *check_privs = B_FALSE;
2264 * The caller requested that the ACL check be skipped. This
2265 * would only happen if the caller checked VOP_ACCESS() with a
2266 * 32 bit ACE mask and already had the appropriate permissions.
2274 * Note: ZFS_READONLY represents the "DOS R/O" attribute.
2275 * When that flag is set, we should behave as if write access
2276 * were not granted by anything in the ACL. In particular:
2277 * We _must_ allow writes after opening the file r/w, then
2278 * setting the DOS R/O attribute, and writing some more.
2279 * (Similar to how you can write after fchmod(fd, 0444).)
2281 * Therefore ZFS_READONLY is ignored in the dataset check
2282 * above, and checked here as if part of the ACL check.
2283 * Also note: DOS R/O is ignored for directories.
2285 if ((v4_mode & WRITE_MASK_DATA) &&
2286 (ZTOV(zp)->v_type != VDIR) &&
2287 (zp->z_pflags & ZFS_READONLY)) {
2288 return (SET_ERROR(EPERM));
2291 return (zfs_zaccess_aces_check(zp, working_mode, B_FALSE, cr));
2295 zfs_zaccess_append(znode_t *zp, uint32_t *working_mode, boolean_t *check_privs,
2298 if (*working_mode != ACE_WRITE_DATA)
2299 return (SET_ERROR(EACCES));
2301 return (zfs_zaccess_common(zp, ACE_APPEND_DATA, working_mode,
2302 check_privs, B_FALSE, cr));
2306 * Check if VEXEC is allowed.
2308 * This routine is based on zfs_fastaccesschk_execute which has slowpath
2309 * calling zfs_zaccess. This would be incorrect on FreeBSD (see
2310 * zfs_freebsd_access for the difference). Thus this variant let's the
2311 * caller handle the slowpath (if necessary).
2313 * On top of that we perform a lockless check for ZFS_NO_EXECS_DENIED.
2315 * Safe access to znode_t is provided by the vnode lock.
2318 zfs_fastaccesschk_execute(znode_t *zdp, cred_t *cr)
2322 if (zdp->z_pflags & ZFS_AV_QUARANTINED)
2325 is_attr = ((zdp->z_pflags & ZFS_XATTR) &&
2326 (ZTOV(zdp)->v_type == VDIR));
2330 if (zdp->z_pflags & ZFS_NO_EXECS_DENIED)
2338 * Determine whether Access should be granted/denied.
2340 * The least priv subsystem is always consulted as a basic privilege
2341 * can define any form of access.
2344 zfs_zaccess(znode_t *zp, int mode, int flags, boolean_t skipaclchk, cred_t *cr,
2347 uint32_t working_mode;
2350 boolean_t check_privs;
2351 znode_t *xzp = NULL;
2352 znode_t *check_zp = zp;
2356 is_attr = ((zp->z_pflags & ZFS_XATTR) && (ZTOV(zp)->v_type == VDIR));
2359 * In FreeBSD, we don't care about permissions of individual ADS.
2360 * Note that not checking them is not just an optimization - without
2361 * this shortcut, EA operations may bogusly fail with EACCES.
2363 if (zp->z_pflags & ZFS_XATTR)
2366 owner = zfs_fuid_map_id(zp->z_zfsvfs, zp->z_uid, cr, ZFS_OWNER);
2369 * Map the bits required to the standard vnode flags VREAD|VWRITE|VEXEC
2370 * in needed_bits. Map the bits mapped by working_mode (currently
2371 * missing) in missing_bits.
2372 * Call secpolicy_vnode_access2() with (needed_bits & ~checkmode),
2377 working_mode = mode;
2378 if ((working_mode & (ACE_READ_ACL|ACE_READ_ATTRIBUTES)) &&
2379 owner == crgetuid(cr))
2380 working_mode &= ~(ACE_READ_ACL|ACE_READ_ATTRIBUTES);
2382 if (working_mode & (ACE_READ_DATA|ACE_READ_NAMED_ATTRS|
2383 ACE_READ_ACL|ACE_READ_ATTRIBUTES|ACE_SYNCHRONIZE))
2384 needed_bits |= VREAD;
2385 if (working_mode & (ACE_WRITE_DATA|ACE_WRITE_NAMED_ATTRS|
2386 ACE_APPEND_DATA|ACE_WRITE_ATTRIBUTES|ACE_SYNCHRONIZE))
2387 needed_bits |= VWRITE;
2388 if (working_mode & ACE_EXECUTE)
2389 needed_bits |= VEXEC;
2391 if ((error = zfs_zaccess_common(check_zp, mode, &working_mode,
2392 &check_privs, skipaclchk, cr)) == 0) {
2395 return (secpolicy_vnode_access2(cr, ZTOV(zp), owner,
2396 needed_bits, needed_bits));
2399 if (error && !check_privs) {
2405 if (error && (flags & V_APPEND)) {
2406 error = zfs_zaccess_append(zp, &working_mode, &check_privs, cr);
2409 if (error && check_privs) {
2410 mode_t checkmode = 0;
2411 vnode_t *check_vp = ZTOV(check_zp);
2414 * First check for implicit owner permission on
2415 * read_acl/read_attributes
2419 ASSERT3U(working_mode, !=, 0);
2421 if ((working_mode & (ACE_READ_ACL|ACE_READ_ATTRIBUTES) &&
2422 owner == crgetuid(cr)))
2423 working_mode &= ~(ACE_READ_ACL|ACE_READ_ATTRIBUTES);
2425 if (working_mode & (ACE_READ_DATA|ACE_READ_NAMED_ATTRS|
2426 ACE_READ_ACL|ACE_READ_ATTRIBUTES|ACE_SYNCHRONIZE))
2428 if (working_mode & (ACE_WRITE_DATA|ACE_WRITE_NAMED_ATTRS|
2429 ACE_APPEND_DATA|ACE_WRITE_ATTRIBUTES|ACE_SYNCHRONIZE))
2430 checkmode |= VWRITE;
2431 if (working_mode & ACE_EXECUTE)
2434 error = secpolicy_vnode_access2(cr, check_vp, owner,
2435 needed_bits & ~checkmode, needed_bits);
2437 if (error == 0 && (working_mode & ACE_WRITE_OWNER))
2438 error = secpolicy_vnode_chown(check_vp, cr, owner);
2439 if (error == 0 && (working_mode & ACE_WRITE_ACL))
2440 error = secpolicy_vnode_setdac(check_vp, cr, owner);
2442 if (error == 0 && (working_mode &
2443 (ACE_DELETE|ACE_DELETE_CHILD)))
2444 error = secpolicy_vnode_remove(check_vp, cr);
2446 if (error == 0 && (working_mode & ACE_SYNCHRONIZE)) {
2447 error = secpolicy_vnode_chown(check_vp, cr, owner);
2451 * See if any bits other than those already checked
2452 * for are still present. If so then return EACCES
2454 if (working_mode & ~(ZFS_CHECKED_MASKS)) {
2455 error = SET_ERROR(EACCES);
2458 } else if (error == 0) {
2459 error = secpolicy_vnode_access2(cr, ZTOV(zp), owner,
2460 needed_bits, needed_bits);
2471 * Translate traditional unix VREAD/VWRITE/VEXEC mode into
2472 * NFSv4-style ZFS ACL format and call zfs_zaccess()
2475 zfs_zaccess_rwx(znode_t *zp, mode_t mode, int flags, cred_t *cr,
2478 return (zfs_zaccess(zp, zfs_unix_to_v4(mode >> 6), flags, B_FALSE, cr,
2483 * Access function for secpolicy_vnode_setattr
2486 zfs_zaccess_unix(znode_t *zp, mode_t mode, cred_t *cr)
2488 int v4_mode = zfs_unix_to_v4(mode >> 6);
2490 return (zfs_zaccess(zp, v4_mode, 0, B_FALSE, cr, NULL));
2494 zfs_delete_final_check(znode_t *zp, znode_t *dzp,
2495 mode_t available_perms, cred_t *cr)
2500 downer = zfs_fuid_map_id(dzp->z_zfsvfs, dzp->z_uid, cr, ZFS_OWNER);
2502 error = secpolicy_vnode_access2(cr, ZTOV(dzp),
2503 downer, available_perms, VWRITE|VEXEC);
2506 error = zfs_sticky_remove_access(dzp, zp, cr);
2512 * Determine whether Access should be granted/deny, without
2513 * consulting least priv subsystem.
2515 * The following chart is the recommended NFSv4 enforcement for
2516 * ability to delete an object.
2518 * -------------------------------------------------------
2519 * | Parent Dir | Target Object Permissions |
2521 * -------------------------------------------------------
2522 * | | ACL Allows | ACL Denies| Delete |
2523 * | | Delete | Delete | unspecified|
2524 * -------------------------------------------------------
2525 * | ACL Allows | Permit | Permit | Permit |
2526 * | DELETE_CHILD | |
2527 * -------------------------------------------------------
2528 * | ACL Denies | Permit | Deny | Deny |
2529 * | DELETE_CHILD | | | |
2530 * -------------------------------------------------------
2531 * | ACL specifies | | | |
2532 * | only allow | Permit | Permit | Permit |
2533 * | write and | | | |
2535 * -------------------------------------------------------
2536 * | ACL denies | | | |
2537 * | write and | Permit | Deny | Deny |
2539 * -------------------------------------------------------
2542 * No search privilege, can't even look up file?
2546 zfs_zaccess_delete(znode_t *dzp, znode_t *zp, cred_t *cr, zuserns_t *mnt_ns)
2548 uint32_t dzp_working_mode = 0;
2549 uint32_t zp_working_mode = 0;
2550 int dzp_error, zp_error;
2551 mode_t available_perms;
2552 boolean_t dzpcheck_privs = B_TRUE;
2553 boolean_t zpcheck_privs = B_TRUE;
2556 * We want specific DELETE permissions to
2557 * take precedence over WRITE/EXECUTE. We don't
2558 * want an ACL such as this to mess us up.
2559 * user:joe:write_data:deny,user:joe:delete:allow
2561 * However, deny permissions may ultimately be overridden
2562 * by secpolicy_vnode_access().
2564 * We will ask for all of the necessary permissions and then
2565 * look at the working modes from the directory and target object
2566 * to determine what was found.
2569 if (zp->z_pflags & (ZFS_IMMUTABLE | ZFS_NOUNLINK))
2570 return (SET_ERROR(EPERM));
2574 * If the directory permissions allow the delete, we are done.
2576 if ((dzp_error = zfs_zaccess_common(dzp, ACE_DELETE_CHILD,
2577 &dzp_working_mode, &dzpcheck_privs, B_FALSE, cr)) == 0)
2581 * If target object has delete permission then we are done
2583 if ((zp_error = zfs_zaccess_common(zp, ACE_DELETE, &zp_working_mode,
2584 &zpcheck_privs, B_FALSE, cr)) == 0)
2590 if (!dzpcheck_privs)
2598 * If directory returns EACCES then delete_child was denied
2599 * due to deny delete_child. In this case send the request through
2600 * secpolicy_vnode_remove(). We don't use zfs_delete_final_check()
2601 * since that *could* allow the delete based on write/execute permission
2602 * and we want delete permissions to override write/execute.
2605 if (dzp_error == EACCES) {
2606 /* XXXPJD: s/dzp/zp/ ? */
2607 return (secpolicy_vnode_remove(ZTOV(dzp), cr));
2611 * only need to see if we have write/execute on directory.
2614 dzp_error = zfs_zaccess_common(dzp, ACE_EXECUTE|ACE_WRITE_DATA,
2615 &dzp_working_mode, &dzpcheck_privs, B_FALSE, cr);
2617 if (dzp_error != 0 && !dzpcheck_privs)
2624 available_perms = (dzp_working_mode & ACE_WRITE_DATA) ? 0 : VWRITE;
2625 available_perms |= (dzp_working_mode & ACE_EXECUTE) ? 0 : VEXEC;
2627 return (zfs_delete_final_check(zp, dzp, available_perms, cr));
2632 zfs_zaccess_rename(znode_t *sdzp, znode_t *szp, znode_t *tdzp,
2633 znode_t *tzp, cred_t *cr, zuserns_t *mnt_ns)
2638 if (szp->z_pflags & ZFS_AV_QUARANTINED)
2639 return (SET_ERROR(EACCES));
2641 add_perm = (ZTOV(szp)->v_type == VDIR) ?
2642 ACE_ADD_SUBDIRECTORY : ACE_ADD_FILE;
2645 * Rename permissions are combination of delete permission +
2646 * add file/subdir permission.
2648 * BSD operating systems also require write permission
2649 * on the directory being moved from one parent directory
2652 if (ZTOV(szp)->v_type == VDIR && ZTOV(sdzp) != ZTOV(tdzp)) {
2653 if ((error = zfs_zaccess(szp, ACE_WRITE_DATA, 0, B_FALSE, cr,
2659 * first make sure we do the delete portion.
2661 * If that succeeds then check for add_file/add_subdir permissions
2664 if ((error = zfs_zaccess_delete(sdzp, szp, cr, mnt_ns)))
2668 * If we have a tzp, see if we can delete it?
2670 if (tzp && (error = zfs_zaccess_delete(tdzp, tzp, cr, mnt_ns)))
2674 * Now check for add permissions
2676 error = zfs_zaccess(tdzp, add_perm, 0, B_FALSE, cr, mnt_ns);