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.
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>
39 #include <sys/cmn_err.h>
40 #include <sys/errno.h>
41 #include <sys/unistd.h>
43 #include <sys/fs/zfs.h>
45 #include <sys/policy.h>
46 #include <sys/zfs_znode.h>
47 #include <sys/zfs_fuid.h>
48 #include <sys/zfs_acl.h>
49 #include <sys/zfs_dir.h>
50 #include <sys/zfs_vfsops.h>
52 #include <sys/dnode.h>
55 #include "fs/fs_subr.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;
149 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));
168 zfs_ace_v0_data(void *acep, void **datap)
174 static 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 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
357 VERIFY(zp->z_is_sa && error == ENOENT);
363 * Determine size of ACL in bytes
365 * This is more complicated than it should be since we have to deal
366 * with old external ACLs.
369 zfs_acl_znode_info(znode_t *zp, int *aclsize, int *aclcount,
370 zfs_acl_phys_t *aclphys)
372 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
377 ASSERT(MUTEX_HELD(&zp->z_acl_lock));
379 if ((error = sa_size(zp->z_sa_hdl, SA_ZPL_DACL_ACES(zfsvfs),
383 if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_DACL_COUNT(zfsvfs),
384 &acl_count, sizeof (acl_count))) != 0)
386 *aclcount = acl_count;
388 if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_ZNODE_ACL(zfsvfs),
389 aclphys, sizeof (*aclphys))) != 0)
392 if (aclphys->z_acl_version == ZFS_ACL_VERSION_INITIAL) {
393 *aclsize = ZFS_ACL_SIZE(aclphys->z_acl_size);
394 *aclcount = aclphys->z_acl_size;
396 *aclsize = aclphys->z_acl_size;
397 *aclcount = aclphys->z_acl_count;
404 zfs_znode_acl_version(znode_t *zp)
406 zfs_acl_phys_t acl_phys;
409 return (ZFS_ACL_VERSION_FUID);
414 * Need to deal with a potential
415 * race where zfs_sa_upgrade could cause
416 * z_isa_sa to change.
418 * If the lookup fails then the state of z_is_sa should have
421 if ((error = sa_lookup(zp->z_sa_hdl,
422 SA_ZPL_ZNODE_ACL(zp->z_zfsvfs),
423 &acl_phys, sizeof (acl_phys))) == 0)
424 return (acl_phys.z_acl_version);
427 * After upgrade SA_ZPL_ZNODE_ACL should have
430 VERIFY(zp->z_is_sa && error == ENOENT);
431 return (ZFS_ACL_VERSION_FUID);
437 zfs_acl_version(int version)
439 if (version < ZPL_VERSION_FUID)
440 return (ZFS_ACL_VERSION_INITIAL);
442 return (ZFS_ACL_VERSION_FUID);
446 zfs_acl_version_zp(znode_t *zp)
448 return (zfs_acl_version(zp->z_zfsvfs->z_version));
452 zfs_acl_alloc(int vers)
456 aclp = kmem_zalloc(sizeof (zfs_acl_t), KM_SLEEP);
457 list_create(&aclp->z_acl, sizeof (zfs_acl_node_t),
458 offsetof(zfs_acl_node_t, z_next));
459 aclp->z_version = vers;
460 if (vers == ZFS_ACL_VERSION_FUID)
461 aclp->z_ops = zfs_acl_fuid_ops;
463 aclp->z_ops = zfs_acl_v0_ops;
468 zfs_acl_node_alloc(size_t bytes)
470 zfs_acl_node_t *aclnode;
472 aclnode = kmem_zalloc(sizeof (zfs_acl_node_t), KM_SLEEP);
474 aclnode->z_acldata = kmem_alloc(bytes, KM_SLEEP);
475 aclnode->z_allocdata = aclnode->z_acldata;
476 aclnode->z_allocsize = bytes;
477 aclnode->z_size = bytes;
484 zfs_acl_node_free(zfs_acl_node_t *aclnode)
486 if (aclnode->z_allocsize)
487 kmem_free(aclnode->z_allocdata, aclnode->z_allocsize);
488 kmem_free(aclnode, sizeof (zfs_acl_node_t));
492 zfs_acl_release_nodes(zfs_acl_t *aclp)
494 zfs_acl_node_t *aclnode;
496 while ((aclnode = list_head(&aclp->z_acl))) {
497 list_remove(&aclp->z_acl, aclnode);
498 zfs_acl_node_free(aclnode);
500 aclp->z_acl_count = 0;
501 aclp->z_acl_bytes = 0;
505 zfs_acl_free(zfs_acl_t *aclp)
507 zfs_acl_release_nodes(aclp);
508 list_destroy(&aclp->z_acl);
509 kmem_free(aclp, sizeof (zfs_acl_t));
513 zfs_acl_valid_ace_type(uint_t type, uint_t flags)
520 case ACE_SYSTEM_AUDIT_ACE_TYPE:
521 case ACE_SYSTEM_ALARM_ACE_TYPE:
522 entry_type = flags & ACE_TYPE_FLAGS;
523 return (entry_type == ACE_OWNER ||
524 entry_type == OWNING_GROUP ||
525 entry_type == ACE_EVERYONE || entry_type == 0 ||
526 entry_type == ACE_IDENTIFIER_GROUP);
528 if (type >= MIN_ACE_TYPE && type <= MAX_ACE_TYPE)
535 zfs_ace_valid(vtype_t obj_type, zfs_acl_t *aclp, uint16_t type, uint16_t iflags)
538 * first check type of entry
541 if (!zfs_acl_valid_ace_type(type, iflags))
545 case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE:
546 case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE:
547 case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE:
548 case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE:
549 if (aclp->z_version < ZFS_ACL_VERSION_FUID)
551 aclp->z_hints |= ZFS_ACL_OBJ_ACE;
555 * next check inheritance level flags
558 if (obj_type == VDIR &&
559 (iflags & (ACE_FILE_INHERIT_ACE|ACE_DIRECTORY_INHERIT_ACE)))
560 aclp->z_hints |= ZFS_INHERIT_ACE;
562 if (iflags & (ACE_INHERIT_ONLY_ACE|ACE_NO_PROPAGATE_INHERIT_ACE)) {
563 if ((iflags & (ACE_FILE_INHERIT_ACE|
564 ACE_DIRECTORY_INHERIT_ACE)) == 0) {
573 zfs_acl_next_ace(zfs_acl_t *aclp, void *start, uint64_t *who,
574 uint32_t *access_mask, uint16_t *iflags, uint16_t *type)
576 zfs_acl_node_t *aclnode;
581 aclnode = list_head(&aclp->z_acl);
585 aclp->z_next_ace = aclnode->z_acldata;
586 aclp->z_curr_node = aclnode;
587 aclnode->z_ace_idx = 0;
590 aclnode = aclp->z_curr_node;
595 if (aclnode->z_ace_idx >= aclnode->z_ace_count) {
596 aclnode = list_next(&aclp->z_acl, aclnode);
600 aclp->z_curr_node = aclnode;
601 aclnode->z_ace_idx = 0;
602 aclp->z_next_ace = aclnode->z_acldata;
606 if (aclnode->z_ace_idx < aclnode->z_ace_count) {
607 void *acep = aclp->z_next_ace;
611 * Make sure we don't overstep our bounds
613 ace_size = aclp->z_ops.ace_size(acep);
615 if (((caddr_t)acep + ace_size) >
616 ((caddr_t)aclnode->z_acldata + aclnode->z_size)) {
620 *iflags = aclp->z_ops.ace_flags_get(acep);
621 *type = aclp->z_ops.ace_type_get(acep);
622 *access_mask = aclp->z_ops.ace_mask_get(acep);
623 *who = aclp->z_ops.ace_who_get(acep);
624 aclp->z_next_ace = (caddr_t)aclp->z_next_ace + ace_size;
625 aclnode->z_ace_idx++;
627 return ((void *)acep);
634 zfs_ace_walk(void *datap, uint64_t cookie, int aclcnt,
635 uint16_t *flags, uint16_t *type, uint32_t *mask)
637 zfs_acl_t *aclp = datap;
638 zfs_ace_hdr_t *acep = (zfs_ace_hdr_t *)(uintptr_t)cookie;
641 acep = zfs_acl_next_ace(aclp, acep, &who, mask,
643 return ((uint64_t)(uintptr_t)acep);
647 * Copy ACE to internal ZFS format.
648 * While processing the ACL each ACE will be validated for correctness.
649 * ACE FUIDs will be created later.
652 zfs_copy_ace_2_fuid(zfsvfs_t *zfsvfs, vtype_t obj_type, zfs_acl_t *aclp,
653 void *datap, zfs_ace_t *z_acl, uint64_t aclcnt, size_t *size,
654 zfs_fuid_info_t **fuidp, cred_t *cr)
658 zfs_ace_t *aceptr = z_acl;
660 zfs_object_ace_t *zobjacep;
661 ace_object_t *aceobjp;
663 for (i = 0; i != aclcnt; i++) {
664 aceptr->z_hdr.z_access_mask = acep->a_access_mask;
665 aceptr->z_hdr.z_flags = acep->a_flags;
666 aceptr->z_hdr.z_type = acep->a_type;
667 entry_type = aceptr->z_hdr.z_flags & ACE_TYPE_FLAGS;
668 if (entry_type != ACE_OWNER && entry_type != OWNING_GROUP &&
669 entry_type != ACE_EVERYONE) {
670 aceptr->z_fuid = zfs_fuid_create(zfsvfs, acep->a_who,
671 cr, (entry_type == 0) ?
672 ZFS_ACE_USER : ZFS_ACE_GROUP, fuidp);
676 * Make sure ACE is valid
678 if (zfs_ace_valid(obj_type, aclp, aceptr->z_hdr.z_type,
679 aceptr->z_hdr.z_flags) != B_TRUE)
682 switch (acep->a_type) {
683 case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE:
684 case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE:
685 case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE:
686 case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE:
687 zobjacep = (zfs_object_ace_t *)aceptr;
688 aceobjp = (ace_object_t *)acep;
690 bcopy(aceobjp->a_obj_type, zobjacep->z_object_type,
691 sizeof (aceobjp->a_obj_type));
692 bcopy(aceobjp->a_inherit_obj_type,
693 zobjacep->z_inherit_type,
694 sizeof (aceobjp->a_inherit_obj_type));
695 acep = (ace_t *)((caddr_t)acep + sizeof (ace_object_t));
698 acep = (ace_t *)((caddr_t)acep + sizeof (ace_t));
701 aceptr = (zfs_ace_t *)((caddr_t)aceptr +
702 aclp->z_ops.ace_size(aceptr));
705 *size = (caddr_t)aceptr - (caddr_t)z_acl;
711 * Copy ZFS ACEs to fixed size ace_t layout
714 zfs_copy_fuid_2_ace(zfsvfs_t *zfsvfs, zfs_acl_t *aclp, cred_t *cr,
715 void *datap, int filter)
718 uint32_t access_mask;
719 uint16_t iflags, type;
720 zfs_ace_hdr_t *zacep = NULL;
722 ace_object_t *objacep;
723 zfs_object_ace_t *zobjacep;
727 while ((zacep = zfs_acl_next_ace(aclp, zacep,
728 &who, &access_mask, &iflags, &type))) {
731 case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE:
732 case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE:
733 case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE:
734 case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE:
738 zobjacep = (zfs_object_ace_t *)zacep;
739 objacep = (ace_object_t *)acep;
740 bcopy(zobjacep->z_object_type,
742 sizeof (zobjacep->z_object_type));
743 bcopy(zobjacep->z_inherit_type,
744 objacep->a_inherit_obj_type,
745 sizeof (zobjacep->z_inherit_type));
746 ace_size = sizeof (ace_object_t);
749 ace_size = sizeof (ace_t);
753 entry_type = (iflags & ACE_TYPE_FLAGS);
754 if ((entry_type != ACE_OWNER &&
755 entry_type != OWNING_GROUP &&
756 entry_type != ACE_EVERYONE)) {
757 acep->a_who = zfs_fuid_map_id(zfsvfs, who,
758 cr, (entry_type & ACE_IDENTIFIER_GROUP) ?
759 ZFS_ACE_GROUP : ZFS_ACE_USER);
761 acep->a_who = (uid_t)(int64_t)who;
763 acep->a_access_mask = access_mask;
764 acep->a_flags = iflags;
766 acep = (ace_t *)((caddr_t)acep + ace_size);
771 zfs_copy_ace_2_oldace(vtype_t obj_type, zfs_acl_t *aclp, ace_t *acep,
772 zfs_oldace_t *z_acl, int aclcnt, size_t *size)
775 zfs_oldace_t *aceptr = z_acl;
777 for (i = 0; i != aclcnt; i++, aceptr++) {
778 aceptr->z_access_mask = acep[i].a_access_mask;
779 aceptr->z_type = acep[i].a_type;
780 aceptr->z_flags = acep[i].a_flags;
781 aceptr->z_fuid = acep[i].a_who;
783 * Make sure ACE is valid
785 if (zfs_ace_valid(obj_type, aclp, aceptr->z_type,
786 aceptr->z_flags) != B_TRUE)
789 *size = (caddr_t)aceptr - (caddr_t)z_acl;
794 * convert old ACL format to new
797 zfs_acl_xform(znode_t *zp, zfs_acl_t *aclp, cred_t *cr)
799 zfs_oldace_t *oldaclp;
801 uint16_t type, iflags;
802 uint32_t access_mask;
805 zfs_acl_node_t *newaclnode;
807 ASSERT(aclp->z_version == ZFS_ACL_VERSION_INITIAL);
809 * First create the ACE in a contiguous piece of memory
810 * for zfs_copy_ace_2_fuid().
812 * We only convert an ACL once, so this won't happen
815 oldaclp = kmem_alloc(sizeof (zfs_oldace_t) * aclp->z_acl_count,
818 while ((cookie = zfs_acl_next_ace(aclp, cookie, &who,
819 &access_mask, &iflags, &type))) {
820 oldaclp[i].z_flags = iflags;
821 oldaclp[i].z_type = type;
822 oldaclp[i].z_fuid = who;
823 oldaclp[i++].z_access_mask = access_mask;
826 newaclnode = zfs_acl_node_alloc(aclp->z_acl_count *
827 sizeof (zfs_object_ace_t));
828 aclp->z_ops = zfs_acl_fuid_ops;
829 VERIFY(zfs_copy_ace_2_fuid(zp->z_zfsvfs, ZTOV(zp)->v_type, aclp,
830 oldaclp, newaclnode->z_acldata, aclp->z_acl_count,
831 &newaclnode->z_size, NULL, cr) == 0);
832 newaclnode->z_ace_count = aclp->z_acl_count;
833 aclp->z_version = ZFS_ACL_VERSION;
834 kmem_free(oldaclp, aclp->z_acl_count * sizeof (zfs_oldace_t));
837 * Release all previous ACL nodes
840 zfs_acl_release_nodes(aclp);
842 list_insert_head(&aclp->z_acl, newaclnode);
844 aclp->z_acl_bytes = newaclnode->z_size;
845 aclp->z_acl_count = newaclnode->z_ace_count;
850 * Convert unix access mask to v4 access mask
853 zfs_unix_to_v4(uint32_t access_mask)
855 uint32_t new_mask = 0;
857 if (access_mask & S_IXOTH)
858 new_mask |= ACE_EXECUTE;
859 if (access_mask & S_IWOTH)
860 new_mask |= ACE_WRITE_DATA;
861 if (access_mask & S_IROTH)
862 new_mask |= ACE_READ_DATA;
867 zfs_set_ace(zfs_acl_t *aclp, void *acep, uint32_t access_mask,
868 uint16_t access_type, uint64_t fuid, uint16_t entry_type)
870 uint16_t type = entry_type & ACE_TYPE_FLAGS;
872 aclp->z_ops.ace_mask_set(acep, access_mask);
873 aclp->z_ops.ace_type_set(acep, access_type);
874 aclp->z_ops.ace_flags_set(acep, entry_type);
875 if ((type != ACE_OWNER && type != OWNING_GROUP &&
876 type != ACE_EVERYONE))
877 aclp->z_ops.ace_who_set(acep, fuid);
881 * Determine mode of file based on ACL.
882 * Also, create FUIDs for any User/Group ACEs
885 zfs_mode_compute(uint64_t fmode, zfs_acl_t *aclp,
886 uint64_t *pflags, uint64_t fuid, uint64_t fgid)
891 zfs_ace_hdr_t *acep = NULL;
893 uint16_t iflags, type;
894 uint32_t access_mask;
895 boolean_t an_exec_denied = B_FALSE;
897 mode = (fmode & (S_IFMT | S_ISUID | S_ISGID | S_ISVTX));
899 while ((acep = zfs_acl_next_ace(aclp, acep, &who,
900 &access_mask, &iflags, &type))) {
902 if (!zfs_acl_valid_ace_type(type, iflags))
905 entry_type = (iflags & ACE_TYPE_FLAGS);
908 * Skip over owner@, group@ or everyone@ inherit only ACEs
910 if ((iflags & ACE_INHERIT_ONLY_ACE) &&
911 (entry_type == ACE_OWNER || entry_type == ACE_EVERYONE ||
912 entry_type == OWNING_GROUP))
915 if (entry_type == ACE_OWNER || (entry_type == 0 &&
917 if ((access_mask & ACE_READ_DATA) &&
918 (!(seen & S_IRUSR))) {
924 if ((access_mask & ACE_WRITE_DATA) &&
925 (!(seen & S_IWUSR))) {
931 if ((access_mask & ACE_EXECUTE) &&
932 (!(seen & S_IXUSR))) {
938 } else if (entry_type == OWNING_GROUP ||
939 (entry_type == ACE_IDENTIFIER_GROUP && who == fgid)) {
940 if ((access_mask & ACE_READ_DATA) &&
941 (!(seen & S_IRGRP))) {
947 if ((access_mask & ACE_WRITE_DATA) &&
948 (!(seen & S_IWGRP))) {
954 if ((access_mask & ACE_EXECUTE) &&
955 (!(seen & S_IXGRP))) {
961 } else if (entry_type == ACE_EVERYONE) {
962 if ((access_mask & ACE_READ_DATA)) {
963 if (!(seen & S_IRUSR)) {
969 if (!(seen & S_IRGRP)) {
975 if (!(seen & S_IROTH)) {
982 if ((access_mask & ACE_WRITE_DATA)) {
983 if (!(seen & S_IWUSR)) {
989 if (!(seen & S_IWGRP)) {
995 if (!(seen & S_IWOTH)) {
1002 if ((access_mask & ACE_EXECUTE)) {
1003 if (!(seen & S_IXUSR)) {
1005 if (type == ALLOW) {
1009 if (!(seen & S_IXGRP)) {
1011 if (type == ALLOW) {
1015 if (!(seen & S_IXOTH)) {
1017 if (type == ALLOW) {
1024 * Only care if this IDENTIFIER_GROUP or
1025 * USER ACE denies execute access to someone,
1026 * mode is not affected
1028 if ((access_mask & ACE_EXECUTE) && type == DENY)
1029 an_exec_denied = B_TRUE;
1034 * Failure to allow is effectively a deny, so execute permission
1035 * is denied if it was never mentioned or if we explicitly
1036 * weren't allowed it.
1038 if (!an_exec_denied &&
1039 ((seen & ALL_MODE_EXECS) != ALL_MODE_EXECS ||
1040 (mode & ALL_MODE_EXECS) != ALL_MODE_EXECS))
1041 an_exec_denied = B_TRUE;
1044 *pflags &= ~ZFS_NO_EXECS_DENIED;
1046 *pflags |= ZFS_NO_EXECS_DENIED;
1052 * Read an external acl object. If the intent is to modify, always
1053 * create a new acl and leave any cached acl in place.
1056 zfs_acl_node_read(znode_t *zp, boolean_t have_lock, zfs_acl_t **aclpp,
1057 boolean_t will_modify)
1062 zfs_acl_node_t *aclnode;
1063 zfs_acl_phys_t znode_acl;
1066 boolean_t drop_lock = B_FALSE;
1068 ASSERT(MUTEX_HELD(&zp->z_acl_lock));
1070 if (zp->z_acl_cached && !will_modify) {
1071 *aclpp = zp->z_acl_cached;
1076 * close race where znode could be upgrade while trying to
1077 * read the znode attributes.
1079 * But this could only happen if the file isn't already an SA
1082 if (!zp->z_is_sa && !have_lock) {
1083 mutex_enter(&zp->z_lock);
1086 version = zfs_znode_acl_version(zp);
1088 if ((error = zfs_acl_znode_info(zp, &aclsize,
1089 &acl_count, &znode_acl)) != 0) {
1093 aclp = zfs_acl_alloc(version);
1095 aclp->z_acl_count = acl_count;
1096 aclp->z_acl_bytes = aclsize;
1098 aclnode = zfs_acl_node_alloc(aclsize);
1099 aclnode->z_ace_count = aclp->z_acl_count;
1100 aclnode->z_size = aclsize;
1103 if (znode_acl.z_acl_extern_obj) {
1104 error = dmu_read(zp->z_zfsvfs->z_os,
1105 znode_acl.z_acl_extern_obj, 0, aclnode->z_size,
1106 aclnode->z_acldata, DMU_READ_PREFETCH);
1108 bcopy(znode_acl.z_ace_data, aclnode->z_acldata,
1112 error = sa_lookup(zp->z_sa_hdl, SA_ZPL_DACL_ACES(zp->z_zfsvfs),
1113 aclnode->z_acldata, aclnode->z_size);
1118 zfs_acl_node_free(aclnode);
1119 /* convert checksum errors into IO errors */
1120 if (error == ECKSUM)
1125 list_insert_head(&aclp->z_acl, aclnode);
1129 zp->z_acl_cached = aclp;
1132 mutex_exit(&zp->z_lock);
1138 zfs_acl_data_locator(void **dataptr, uint32_t *length, uint32_t buflen,
1139 boolean_t start, void *userdata)
1141 zfs_acl_locator_cb_t *cb = (zfs_acl_locator_cb_t *)userdata;
1144 cb->cb_acl_node = list_head(&cb->cb_aclp->z_acl);
1146 cb->cb_acl_node = list_next(&cb->cb_aclp->z_acl,
1149 *dataptr = cb->cb_acl_node->z_acldata;
1150 *length = cb->cb_acl_node->z_size;
1154 zfs_acl_chown_setattr(znode_t *zp)
1159 ASSERT(MUTEX_HELD(&zp->z_lock));
1160 ASSERT(MUTEX_HELD(&zp->z_acl_lock));
1162 if ((error = zfs_acl_node_read(zp, B_TRUE, &aclp, B_FALSE)) == 0)
1163 zp->z_mode = zfs_mode_compute(zp->z_mode, aclp,
1164 &zp->z_pflags, zp->z_uid, zp->z_gid);
1169 acl_trivial_access_masks(mode_t mode, uint32_t *allow0, uint32_t *deny1,
1170 uint32_t *deny2, uint32_t *owner, uint32_t *group, uint32_t *everyone)
1172 *deny1 = *deny2 = *allow0 = *group = 0;
1174 if (!(mode & S_IRUSR) && (mode & (S_IRGRP|S_IROTH)))
1175 *deny1 |= ACE_READ_DATA;
1176 if (!(mode & S_IWUSR) && (mode & (S_IWGRP|S_IWOTH)))
1177 *deny1 |= ACE_WRITE_DATA;
1178 if (!(mode & S_IXUSR) && (mode & (S_IXGRP|S_IXOTH)))
1179 *deny1 |= ACE_EXECUTE;
1181 if (!(mode & S_IRGRP) && (mode & S_IROTH))
1182 *deny2 = ACE_READ_DATA;
1183 if (!(mode & S_IWGRP) && (mode & S_IWOTH))
1184 *deny2 |= ACE_WRITE_DATA;
1185 if (!(mode & S_IXGRP) && (mode & S_IXOTH))
1186 *deny2 |= ACE_EXECUTE;
1188 if ((mode & S_IRUSR) && (!(mode & S_IRGRP) && (mode & S_IROTH)))
1189 *allow0 |= ACE_READ_DATA;
1190 if ((mode & S_IWUSR) && (!(mode & S_IWGRP) && (mode & S_IWOTH)))
1191 *allow0 |= ACE_WRITE_DATA;
1192 if ((mode & S_IXUSR) && (!(mode & S_IXGRP) && (mode & S_IXOTH)))
1193 *allow0 |= ACE_EXECUTE;
1195 *owner = ACE_WRITE_ATTRIBUTES|ACE_WRITE_OWNER|ACE_WRITE_ACL|
1196 ACE_WRITE_NAMED_ATTRS|ACE_READ_ACL|ACE_READ_ATTRIBUTES|
1197 ACE_READ_NAMED_ATTRS|ACE_SYNCHRONIZE;
1199 *owner |= ACE_READ_DATA;
1201 *owner |= ACE_WRITE_DATA|ACE_APPEND_DATA;
1203 *owner |= ACE_EXECUTE;
1205 *group = ACE_READ_ACL|ACE_READ_ATTRIBUTES| ACE_READ_NAMED_ATTRS|
1208 *group |= ACE_READ_DATA;
1210 *group |= ACE_WRITE_DATA|ACE_APPEND_DATA;
1212 *group |= ACE_EXECUTE;
1214 *everyone = ACE_READ_ACL|ACE_READ_ATTRIBUTES| ACE_READ_NAMED_ATTRS|
1217 *everyone |= ACE_READ_DATA;
1219 *everyone |= ACE_WRITE_DATA|ACE_APPEND_DATA;
1221 *everyone |= ACE_EXECUTE;
1226 * determine whether an ace_t acl is trivial
1228 * Trivialness implies that the acl is composed of only
1229 * owner, group, everyone entries. ACL can't
1230 * have read_acl denied, and write_owner/write_acl/write_attributes
1231 * can only be owner@ entry.
1234 ace_trivial_common(void *acep, int aclcnt,
1235 uint64_t (*walk)(void *, uint64_t, int aclcnt,
1236 uint16_t *, uint16_t *, uint32_t *))
1241 uint64_t cookie = 0;
1243 while ((cookie = walk(acep, cookie, aclcnt, &flags, &type, &mask))) {
1244 switch (flags & ACE_TYPE_FLAGS) {
1246 case ACE_GROUP|ACE_IDENTIFIER_GROUP:
1253 if (flags & (ACE_FILE_INHERIT_ACE|
1254 ACE_DIRECTORY_INHERIT_ACE|ACE_NO_PROPAGATE_INHERIT_ACE|
1255 ACE_INHERIT_ONLY_ACE))
1259 * Special check for some special bits
1261 * Don't allow anybody to deny reading basic
1262 * attributes or a files ACL.
1264 if ((mask & (ACE_READ_ACL|ACE_READ_ATTRIBUTES)) &&
1265 (type == ACE_ACCESS_DENIED_ACE_TYPE))
1269 * Delete permissions are never set by default
1271 if (mask & (ACE_DELETE|ACE_DELETE_CHILD))
1274 * only allow owner@ to have
1275 * write_acl/write_owner/write_attributes/write_xattr/
1277 if (type == ACE_ACCESS_ALLOWED_ACE_TYPE &&
1278 (!(flags & ACE_OWNER) && (mask &
1279 (ACE_WRITE_OWNER|ACE_WRITE_ACL| ACE_WRITE_ATTRIBUTES|
1280 ACE_WRITE_NAMED_ATTRS))))
1289 * common code for setting ACLs.
1291 * This function is called from zfs_mode_update, zfs_perm_init, and zfs_setacl.
1292 * zfs_setacl passes a non-NULL inherit pointer (ihp) to indicate that it's
1293 * already checked the acl and knows whether to inherit.
1296 zfs_aclset_common(znode_t *zp, zfs_acl_t *aclp, cred_t *cr, dmu_tx_t *tx)
1299 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1300 dmu_object_type_t otype;
1301 zfs_acl_locator_cb_t locate = { 0 };
1303 sa_bulk_attr_t bulk[5];
1309 mode = zfs_mode_compute(mode, aclp, &zp->z_pflags,
1310 zp->z_uid, zp->z_gid);
1313 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs), NULL,
1314 &mode, sizeof (mode));
1315 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL,
1316 &zp->z_pflags, sizeof (zp->z_pflags));
1317 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL,
1318 &ctime, sizeof (ctime));
1320 if (zp->z_acl_cached) {
1321 zfs_acl_free(zp->z_acl_cached);
1322 zp->z_acl_cached = NULL;
1328 if (!zfsvfs->z_use_fuids) {
1329 otype = DMU_OT_OLDACL;
1331 if ((aclp->z_version == ZFS_ACL_VERSION_INITIAL) &&
1332 (zfsvfs->z_version >= ZPL_VERSION_FUID))
1333 zfs_acl_xform(zp, aclp, cr);
1334 ASSERT(aclp->z_version >= ZFS_ACL_VERSION_FUID);
1339 * Arrgh, we have to handle old on disk format
1340 * as well as newer (preferred) SA format.
1343 if (zp->z_is_sa) { /* the easy case, just update the ACL attribute */
1344 locate.cb_aclp = aclp;
1345 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_DACL_ACES(zfsvfs),
1346 zfs_acl_data_locator, &locate, aclp->z_acl_bytes);
1347 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_DACL_COUNT(zfsvfs),
1348 NULL, &aclp->z_acl_count, sizeof (uint64_t));
1349 } else { /* Painful legacy way */
1350 zfs_acl_node_t *aclnode;
1352 zfs_acl_phys_t acl_phys;
1355 if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_ZNODE_ACL(zfsvfs),
1356 &acl_phys, sizeof (acl_phys))) != 0)
1359 aoid = acl_phys.z_acl_extern_obj;
1361 if (aclp->z_acl_bytes > ZFS_ACE_SPACE) {
1363 * If ACL was previously external and we are now
1364 * converting to new ACL format then release old
1365 * ACL object and create a new one.
1368 aclp->z_version != acl_phys.z_acl_version) {
1369 error = dmu_object_free(zfsvfs->z_os, aoid, tx);
1375 aoid = dmu_object_alloc(zfsvfs->z_os,
1376 otype, aclp->z_acl_bytes,
1377 otype == DMU_OT_ACL ?
1378 DMU_OT_SYSACL : DMU_OT_NONE,
1379 otype == DMU_OT_ACL ?
1380 DN_MAX_BONUSLEN : 0, tx);
1382 (void) dmu_object_set_blocksize(zfsvfs->z_os,
1383 aoid, aclp->z_acl_bytes, 0, tx);
1385 acl_phys.z_acl_extern_obj = aoid;
1386 for (aclnode = list_head(&aclp->z_acl); aclnode;
1387 aclnode = list_next(&aclp->z_acl, aclnode)) {
1388 if (aclnode->z_ace_count == 0)
1390 dmu_write(zfsvfs->z_os, aoid, off,
1391 aclnode->z_size, aclnode->z_acldata, tx);
1392 off += aclnode->z_size;
1395 void *start = acl_phys.z_ace_data;
1397 * Migrating back embedded?
1399 if (acl_phys.z_acl_extern_obj) {
1400 error = dmu_object_free(zfsvfs->z_os,
1401 acl_phys.z_acl_extern_obj, tx);
1404 acl_phys.z_acl_extern_obj = 0;
1407 for (aclnode = list_head(&aclp->z_acl); aclnode;
1408 aclnode = list_next(&aclp->z_acl, aclnode)) {
1409 if (aclnode->z_ace_count == 0)
1411 bcopy(aclnode->z_acldata, start,
1413 start = (caddr_t)start + aclnode->z_size;
1417 * If Old version then swap count/bytes to match old
1418 * layout of znode_acl_phys_t.
1420 if (aclp->z_version == ZFS_ACL_VERSION_INITIAL) {
1421 acl_phys.z_acl_size = aclp->z_acl_count;
1422 acl_phys.z_acl_count = aclp->z_acl_bytes;
1424 acl_phys.z_acl_size = aclp->z_acl_bytes;
1425 acl_phys.z_acl_count = aclp->z_acl_count;
1427 acl_phys.z_acl_version = aclp->z_version;
1429 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ZNODE_ACL(zfsvfs), NULL,
1430 &acl_phys, sizeof (acl_phys));
1434 * Replace ACL wide bits, but first clear them.
1436 zp->z_pflags &= ~ZFS_ACL_WIDE_FLAGS;
1438 zp->z_pflags |= aclp->z_hints;
1440 if (ace_trivial_common(aclp, 0, zfs_ace_walk) == 0)
1441 zp->z_pflags |= ZFS_ACL_TRIVIAL;
1443 zfs_tstamp_update_setup(zp, STATE_CHANGED, NULL, ctime, B_TRUE);
1444 return (sa_bulk_update(zp->z_sa_hdl, bulk, count, tx));
1448 zfs_acl_chmod(zfsvfs_t *zfsvfs, uint64_t mode, zfs_acl_t *aclp)
1452 int new_count, new_bytes;
1455 uint16_t iflags, type;
1456 uint32_t access_mask;
1457 zfs_acl_node_t *newnode;
1458 size_t abstract_size = aclp->z_ops.ace_abstract_size();
1460 uint32_t owner, group, everyone;
1461 uint32_t deny1, deny2, allow0;
1463 new_count = new_bytes = 0;
1465 acl_trivial_access_masks((mode_t)mode, &allow0, &deny1, &deny2,
1466 &owner, &group, &everyone);
1468 newnode = zfs_acl_node_alloc((abstract_size * 6) + aclp->z_acl_bytes);
1470 zacep = newnode->z_acldata;
1472 zfs_set_ace(aclp, zacep, allow0, ALLOW, -1, ACE_OWNER);
1473 zacep = (void *)((uintptr_t)zacep + abstract_size);
1475 new_bytes += abstract_size;
1477 zfs_set_ace(aclp, zacep, deny1, DENY, -1, ACE_OWNER);
1478 zacep = (void *)((uintptr_t)zacep + abstract_size);
1480 new_bytes += abstract_size;
1483 zfs_set_ace(aclp, zacep, deny2, DENY, -1, OWNING_GROUP);
1484 zacep = (void *)((uintptr_t)zacep + abstract_size);
1486 new_bytes += abstract_size;
1489 while ((acep = zfs_acl_next_ace(aclp, acep, &who, &access_mask,
1491 uint16_t inherit_flags;
1493 entry_type = (iflags & ACE_TYPE_FLAGS);
1494 inherit_flags = (iflags & ALL_INHERIT);
1496 if ((entry_type == ACE_OWNER || entry_type == ACE_EVERYONE ||
1497 (entry_type == OWNING_GROUP)) &&
1498 ((inherit_flags & ACE_INHERIT_ONLY_ACE) == 0)) {
1502 if ((type != ALLOW && type != DENY) ||
1503 (inherit_flags & ACE_INHERIT_ONLY_ACE)) {
1505 aclp->z_hints |= ZFS_INHERIT_ACE;
1507 case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE:
1508 case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE:
1509 case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE:
1510 case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE:
1511 aclp->z_hints |= ZFS_ACL_OBJ_ACE;
1517 * Limit permissions to be no greater than
1520 if (zfsvfs->z_acl_inherit == ZFS_ACL_RESTRICTED) {
1521 if (!(mode & S_IRGRP))
1522 access_mask &= ~ACE_READ_DATA;
1523 if (!(mode & S_IWGRP))
1525 ~(ACE_WRITE_DATA|ACE_APPEND_DATA);
1526 if (!(mode & S_IXGRP))
1527 access_mask &= ~ACE_EXECUTE;
1529 ~(ACE_WRITE_OWNER|ACE_WRITE_ACL|
1530 ACE_WRITE_ATTRIBUTES|ACE_WRITE_NAMED_ATTRS);
1533 zfs_set_ace(aclp, zacep, access_mask, type, who, iflags);
1534 ace_size = aclp->z_ops.ace_size(acep);
1535 zacep = (void *)((uintptr_t)zacep + ace_size);
1537 new_bytes += ace_size;
1539 zfs_set_ace(aclp, zacep, owner, 0, -1, ACE_OWNER);
1540 zacep = (void *)((uintptr_t)zacep + abstract_size);
1541 zfs_set_ace(aclp, zacep, group, 0, -1, OWNING_GROUP);
1542 zacep = (void *)((uintptr_t)zacep + abstract_size);
1543 zfs_set_ace(aclp, zacep, everyone, 0, -1, ACE_EVERYONE);
1546 new_bytes += abstract_size * 3;
1547 zfs_acl_release_nodes(aclp);
1548 aclp->z_acl_count = new_count;
1549 aclp->z_acl_bytes = new_bytes;
1550 newnode->z_ace_count = new_count;
1551 newnode->z_size = new_bytes;
1552 list_insert_tail(&aclp->z_acl, newnode);
1556 zfs_acl_chmod_setattr(znode_t *zp, zfs_acl_t **aclp, uint64_t mode)
1558 mutex_enter(&zp->z_acl_lock);
1559 mutex_enter(&zp->z_lock);
1560 *aclp = zfs_acl_alloc(zfs_acl_version_zp(zp));
1561 (*aclp)->z_hints = zp->z_pflags & V4_ACL_WIDE_FLAGS;
1562 zfs_acl_chmod(zp->z_zfsvfs, mode, *aclp);
1563 mutex_exit(&zp->z_lock);
1564 mutex_exit(&zp->z_acl_lock);
1569 * strip off write_owner and write_acl
1572 zfs_restricted_update(zfsvfs_t *zfsvfs, zfs_acl_t *aclp, void *acep)
1574 uint32_t mask = aclp->z_ops.ace_mask_get(acep);
1576 if ((zfsvfs->z_acl_inherit == ZFS_ACL_RESTRICTED) &&
1577 (aclp->z_ops.ace_type_get(acep) == ALLOW)) {
1578 mask &= ~RESTRICTED_CLEAR;
1579 aclp->z_ops.ace_mask_set(acep, mask);
1584 * Should ACE be inherited?
1587 zfs_ace_can_use(vtype_t vtype, uint16_t acep_flags)
1589 int iflags = (acep_flags & 0xf);
1591 if ((vtype == VDIR) && (iflags & ACE_DIRECTORY_INHERIT_ACE))
1593 else if (iflags & ACE_FILE_INHERIT_ACE)
1594 return (!((vtype == VDIR) &&
1595 (iflags & ACE_NO_PROPAGATE_INHERIT_ACE)));
1600 * inherit inheritable ACEs from parent
1603 zfs_acl_inherit(zfsvfs_t *zfsvfs, vtype_t vtype, zfs_acl_t *paclp,
1604 uint64_t mode, boolean_t *need_chmod)
1608 zfs_acl_node_t *aclnode;
1609 zfs_acl_t *aclp = NULL;
1611 uint32_t access_mask;
1612 uint16_t iflags, newflags, type;
1614 void *data1, *data2;
1615 size_t data1sz, data2sz;
1616 boolean_t vdir = vtype == VDIR;
1617 boolean_t vreg = vtype == VREG;
1618 boolean_t passthrough, passthrough_x, noallow;
1621 zfsvfs->z_acl_inherit == ZFS_ACL_PASSTHROUGH_X;
1622 passthrough = passthrough_x ||
1623 zfsvfs->z_acl_inherit == ZFS_ACL_PASSTHROUGH;
1625 zfsvfs->z_acl_inherit == ZFS_ACL_NOALLOW;
1627 *need_chmod = B_TRUE;
1629 aclp = zfs_acl_alloc(paclp->z_version);
1630 if (zfsvfs->z_acl_inherit == ZFS_ACL_DISCARD || vtype == VLNK)
1632 while ((pacep = zfs_acl_next_ace(paclp, pacep, &who,
1633 &access_mask, &iflags, &type))) {
1636 * don't inherit bogus ACEs
1638 if (!zfs_acl_valid_ace_type(type, iflags))
1641 if (noallow && type == ALLOW)
1644 ace_size = aclp->z_ops.ace_size(pacep);
1646 if (!zfs_ace_can_use(vtype, iflags))
1650 * If owner@, group@, or everyone@ inheritable
1651 * then zfs_acl_chmod() isn't needed.
1654 ((iflags & (ACE_OWNER|ACE_EVERYONE)) ||
1655 ((iflags & OWNING_GROUP) ==
1656 OWNING_GROUP)) && (vreg || (vdir && (iflags &
1657 ACE_DIRECTORY_INHERIT_ACE)))) {
1658 *need_chmod = B_FALSE;
1661 if (!vdir && passthrough_x &&
1662 ((mode & (S_IXUSR | S_IXGRP | S_IXOTH)) == 0)) {
1663 access_mask &= ~ACE_EXECUTE;
1666 aclnode = zfs_acl_node_alloc(ace_size);
1667 list_insert_tail(&aclp->z_acl, aclnode);
1668 acep = aclnode->z_acldata;
1670 zfs_set_ace(aclp, acep, access_mask, type,
1671 who, iflags|ACE_INHERITED_ACE);
1674 * Copy special opaque data if any
1676 if ((data1sz = paclp->z_ops.ace_data(pacep, &data1)) != 0) {
1677 VERIFY((data2sz = aclp->z_ops.ace_data(acep,
1678 &data2)) == data1sz);
1679 bcopy(data1, data2, data2sz);
1682 aclp->z_acl_count++;
1683 aclnode->z_ace_count++;
1684 aclp->z_acl_bytes += aclnode->z_size;
1685 newflags = aclp->z_ops.ace_flags_get(acep);
1688 aclp->z_hints |= ZFS_INHERIT_ACE;
1690 if ((iflags & ACE_NO_PROPAGATE_INHERIT_ACE) || !vdir) {
1691 newflags &= ~ALL_INHERIT;
1692 aclp->z_ops.ace_flags_set(acep,
1693 newflags|ACE_INHERITED_ACE);
1694 zfs_restricted_update(zfsvfs, aclp, acep);
1701 * If only FILE_INHERIT is set then turn on
1704 if ((iflags & (ACE_FILE_INHERIT_ACE |
1705 ACE_DIRECTORY_INHERIT_ACE)) == ACE_FILE_INHERIT_ACE) {
1706 newflags |= ACE_INHERIT_ONLY_ACE;
1707 aclp->z_ops.ace_flags_set(acep,
1708 newflags|ACE_INHERITED_ACE);
1710 newflags &= ~ACE_INHERIT_ONLY_ACE;
1711 aclp->z_ops.ace_flags_set(acep,
1712 newflags|ACE_INHERITED_ACE);
1719 * Create file system object initial permissions
1720 * including inheritable ACEs.
1723 zfs_acl_ids_create(znode_t *dzp, int flag, vattr_t *vap, cred_t *cr,
1724 vsecattr_t *vsecp, zfs_acl_ids_t *acl_ids)
1727 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
1730 boolean_t need_chmod = B_TRUE;
1731 boolean_t inherited = B_FALSE;
1733 bzero(acl_ids, sizeof (zfs_acl_ids_t));
1734 acl_ids->z_mode = MAKEIMODE(vap->va_type, vap->va_mode);
1737 if ((error = zfs_vsec_2_aclp(zfsvfs, vap->va_type, vsecp, cr,
1738 &acl_ids->z_fuidp, &acl_ids->z_aclp)) != 0)
1741 * Determine uid and gid.
1743 if ((flag & IS_ROOT_NODE) || zfsvfs->z_replay ||
1744 ((flag & IS_XATTR) && (vap->va_type == VDIR))) {
1745 acl_ids->z_fuid = zfs_fuid_create(zfsvfs,
1746 (uint64_t)vap->va_uid, cr,
1747 ZFS_OWNER, &acl_ids->z_fuidp);
1748 acl_ids->z_fgid = zfs_fuid_create(zfsvfs,
1749 (uint64_t)vap->va_gid, cr,
1750 ZFS_GROUP, &acl_ids->z_fuidp);
1753 acl_ids->z_fuid = zfs_fuid_create_cred(zfsvfs, ZFS_OWNER,
1754 cr, &acl_ids->z_fuidp);
1755 acl_ids->z_fgid = 0;
1756 if (vap->va_mask & AT_GID) {
1757 acl_ids->z_fgid = zfs_fuid_create(zfsvfs,
1758 (uint64_t)vap->va_gid,
1759 cr, ZFS_GROUP, &acl_ids->z_fuidp);
1761 if (acl_ids->z_fgid != dzp->z_gid &&
1762 !groupmember(vap->va_gid, cr) &&
1763 secpolicy_vnode_create_gid(cr) != 0)
1764 acl_ids->z_fgid = 0;
1766 if (acl_ids->z_fgid == 0) {
1767 if (dzp->z_mode & S_ISGID) {
1771 acl_ids->z_fgid = dzp->z_gid;
1772 gid = zfs_fuid_map_id(zfsvfs, acl_ids->z_fgid,
1775 if (zfsvfs->z_use_fuids &&
1776 IS_EPHEMERAL(acl_ids->z_fgid)) {
1777 domain = zfs_fuid_idx_domain(
1778 &zfsvfs->z_fuid_idx,
1779 FUID_INDEX(acl_ids->z_fgid));
1780 rid = FUID_RID(acl_ids->z_fgid);
1781 zfs_fuid_node_add(&acl_ids->z_fuidp,
1783 FUID_INDEX(acl_ids->z_fgid),
1784 acl_ids->z_fgid, ZFS_GROUP);
1787 acl_ids->z_fgid = zfs_fuid_create_cred(zfsvfs,
1788 ZFS_GROUP, cr, &acl_ids->z_fuidp);
1795 * If we're creating a directory, and the parent directory has the
1796 * set-GID bit set, set in on the new directory.
1797 * Otherwise, if the user is neither privileged nor a member of the
1798 * file's new group, clear the file's set-GID bit.
1801 if (!(flag & IS_ROOT_NODE) && (dzp->z_mode & S_ISGID) &&
1802 (vap->va_type == VDIR)) {
1803 acl_ids->z_mode |= S_ISGID;
1805 if ((acl_ids->z_mode & S_ISGID) &&
1806 secpolicy_vnode_setids_setgids(cr, gid) != 0)
1807 acl_ids->z_mode &= ~S_ISGID;
1810 if (acl_ids->z_aclp == NULL) {
1811 mutex_enter(&dzp->z_acl_lock);
1812 mutex_enter(&dzp->z_lock);
1813 if (!(flag & IS_ROOT_NODE) && (ZTOV(dzp)->v_type == VDIR &&
1814 (dzp->z_pflags & ZFS_INHERIT_ACE)) &&
1815 !(dzp->z_pflags & ZFS_XATTR)) {
1816 VERIFY(0 == zfs_acl_node_read(dzp, B_TRUE,
1818 acl_ids->z_aclp = zfs_acl_inherit(zfsvfs,
1819 vap->va_type, paclp, acl_ids->z_mode, &need_chmod);
1823 zfs_acl_alloc(zfs_acl_version_zp(dzp));
1824 acl_ids->z_aclp->z_hints |= ZFS_ACL_TRIVIAL;
1826 mutex_exit(&dzp->z_lock);
1827 mutex_exit(&dzp->z_acl_lock);
1829 acl_ids->z_aclp->z_hints |= (vap->va_type == VDIR) ?
1830 ZFS_ACL_AUTO_INHERIT : 0;
1831 zfs_acl_chmod(zfsvfs, acl_ids->z_mode, acl_ids->z_aclp);
1835 if (inherited || vsecp) {
1836 acl_ids->z_mode = zfs_mode_compute(acl_ids->z_mode,
1837 acl_ids->z_aclp, &acl_ids->z_aclp->z_hints,
1838 acl_ids->z_fuid, acl_ids->z_fgid);
1839 if (ace_trivial_common(acl_ids->z_aclp, 0, zfs_ace_walk) == 0)
1840 acl_ids->z_aclp->z_hints |= ZFS_ACL_TRIVIAL;
1847 * Free ACL and fuid_infop, but not the acl_ids structure
1850 zfs_acl_ids_free(zfs_acl_ids_t *acl_ids)
1852 if (acl_ids->z_aclp)
1853 zfs_acl_free(acl_ids->z_aclp);
1854 if (acl_ids->z_fuidp)
1855 zfs_fuid_info_free(acl_ids->z_fuidp);
1856 acl_ids->z_aclp = NULL;
1857 acl_ids->z_fuidp = NULL;
1861 zfs_acl_ids_overquota(zfsvfs_t *zfsvfs, zfs_acl_ids_t *acl_ids)
1863 return (zfs_fuid_overquota(zfsvfs, B_FALSE, acl_ids->z_fuid) ||
1864 zfs_fuid_overquota(zfsvfs, B_TRUE, acl_ids->z_fgid));
1868 * Retrieve a files ACL
1871 zfs_getacl(znode_t *zp, vsecattr_t *vsecp, boolean_t skipaclchk, cred_t *cr)
1879 mask = vsecp->vsa_mask & (VSA_ACE | VSA_ACECNT |
1880 VSA_ACE_ACLFLAGS | VSA_ACE_ALLTYPES);
1885 if ((error = zfs_zaccess(zp, ACE_READ_ACL, 0, skipaclchk, cr)))
1888 mutex_enter(&zp->z_acl_lock);
1890 error = zfs_acl_node_read(zp, B_FALSE, &aclp, B_FALSE);
1892 mutex_exit(&zp->z_acl_lock);
1897 * Scan ACL to determine number of ACEs
1899 if ((zp->z_pflags & ZFS_ACL_OBJ_ACE) && !(mask & VSA_ACE_ALLTYPES)) {
1902 uint32_t access_mask;
1903 uint16_t type, iflags;
1905 while ((zacep = zfs_acl_next_ace(aclp, zacep,
1906 &who, &access_mask, &iflags, &type))) {
1908 case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE:
1909 case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE:
1910 case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE:
1911 case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE:
1918 vsecp->vsa_aclcnt = count;
1920 count = (int)aclp->z_acl_count;
1922 if (mask & VSA_ACECNT) {
1923 vsecp->vsa_aclcnt = count;
1926 if (mask & VSA_ACE) {
1929 aclsz = count * sizeof (ace_t) +
1930 sizeof (ace_object_t) * largeace;
1932 vsecp->vsa_aclentp = kmem_alloc(aclsz, KM_SLEEP);
1933 vsecp->vsa_aclentsz = aclsz;
1935 if (aclp->z_version == ZFS_ACL_VERSION_FUID)
1936 zfs_copy_fuid_2_ace(zp->z_zfsvfs, aclp, cr,
1937 vsecp->vsa_aclentp, !(mask & VSA_ACE_ALLTYPES));
1939 zfs_acl_node_t *aclnode;
1940 void *start = vsecp->vsa_aclentp;
1942 for (aclnode = list_head(&aclp->z_acl); aclnode;
1943 aclnode = list_next(&aclp->z_acl, aclnode)) {
1944 bcopy(aclnode->z_acldata, start,
1946 start = (caddr_t)start + aclnode->z_size;
1948 ASSERT((caddr_t)start - (caddr_t)vsecp->vsa_aclentp ==
1952 if (mask & VSA_ACE_ACLFLAGS) {
1953 vsecp->vsa_aclflags = 0;
1954 if (zp->z_pflags & ZFS_ACL_DEFAULTED)
1955 vsecp->vsa_aclflags |= ACL_DEFAULTED;
1956 if (zp->z_pflags & ZFS_ACL_PROTECTED)
1957 vsecp->vsa_aclflags |= ACL_PROTECTED;
1958 if (zp->z_pflags & ZFS_ACL_AUTO_INHERIT)
1959 vsecp->vsa_aclflags |= ACL_AUTO_INHERIT;
1962 mutex_exit(&zp->z_acl_lock);
1968 zfs_vsec_2_aclp(zfsvfs_t *zfsvfs, vtype_t obj_type,
1969 vsecattr_t *vsecp, cred_t *cr, zfs_fuid_info_t **fuidp, zfs_acl_t **zaclp)
1972 zfs_acl_node_t *aclnode;
1973 int aclcnt = vsecp->vsa_aclcnt;
1976 if (vsecp->vsa_aclcnt > MAX_ACL_ENTRIES || vsecp->vsa_aclcnt <= 0)
1979 aclp = zfs_acl_alloc(zfs_acl_version(zfsvfs->z_version));
1982 aclnode = zfs_acl_node_alloc(aclcnt * sizeof (zfs_object_ace_t));
1983 if (aclp->z_version == ZFS_ACL_VERSION_INITIAL) {
1984 if ((error = zfs_copy_ace_2_oldace(obj_type, aclp,
1985 (ace_t *)vsecp->vsa_aclentp, aclnode->z_acldata,
1986 aclcnt, &aclnode->z_size)) != 0) {
1988 zfs_acl_node_free(aclnode);
1992 if ((error = zfs_copy_ace_2_fuid(zfsvfs, obj_type, aclp,
1993 vsecp->vsa_aclentp, aclnode->z_acldata, aclcnt,
1994 &aclnode->z_size, fuidp, cr)) != 0) {
1996 zfs_acl_node_free(aclnode);
2000 aclp->z_acl_bytes = aclnode->z_size;
2001 aclnode->z_ace_count = aclcnt;
2002 aclp->z_acl_count = aclcnt;
2003 list_insert_head(&aclp->z_acl, aclnode);
2006 * If flags are being set then add them to z_hints
2008 if (vsecp->vsa_mask & VSA_ACE_ACLFLAGS) {
2009 if (vsecp->vsa_aclflags & ACL_PROTECTED)
2010 aclp->z_hints |= ZFS_ACL_PROTECTED;
2011 if (vsecp->vsa_aclflags & ACL_DEFAULTED)
2012 aclp->z_hints |= ZFS_ACL_DEFAULTED;
2013 if (vsecp->vsa_aclflags & ACL_AUTO_INHERIT)
2014 aclp->z_hints |= ZFS_ACL_AUTO_INHERIT;
2026 zfs_setacl(znode_t *zp, vsecattr_t *vsecp, boolean_t skipaclchk, cred_t *cr)
2028 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
2029 zilog_t *zilog = zfsvfs->z_log;
2030 ulong_t mask = vsecp->vsa_mask & (VSA_ACE | VSA_ACECNT);
2034 zfs_fuid_info_t *fuidp = NULL;
2035 boolean_t fuid_dirtied;
2041 if (zp->z_pflags & ZFS_IMMUTABLE)
2044 if ((error = zfs_zaccess(zp, ACE_WRITE_ACL, 0, skipaclchk, cr)))
2047 error = zfs_vsec_2_aclp(zfsvfs, ZTOV(zp)->v_type, vsecp, cr, &fuidp,
2053 * If ACL wide flags aren't being set then preserve any
2056 if (!(vsecp->vsa_mask & VSA_ACE_ACLFLAGS)) {
2058 (zp->z_pflags & V4_ACL_WIDE_FLAGS);
2061 mutex_enter(&zp->z_acl_lock);
2062 mutex_enter(&zp->z_lock);
2064 tx = dmu_tx_create(zfsvfs->z_os);
2066 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE);
2068 fuid_dirtied = zfsvfs->z_fuid_dirty;
2070 zfs_fuid_txhold(zfsvfs, tx);
2073 * If old version and ACL won't fit in bonus and we aren't
2074 * upgrading then take out necessary DMU holds
2077 if ((acl_obj = zfs_external_acl(zp)) != 0) {
2078 if (zfsvfs->z_version >= ZPL_VERSION_FUID &&
2079 zfs_znode_acl_version(zp) <= ZFS_ACL_VERSION_INITIAL) {
2080 dmu_tx_hold_free(tx, acl_obj, 0,
2082 dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0,
2085 dmu_tx_hold_write(tx, acl_obj, 0, aclp->z_acl_bytes);
2087 } else if (!zp->z_is_sa && aclp->z_acl_bytes > ZFS_ACE_SPACE) {
2088 dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, aclp->z_acl_bytes);
2091 zfs_sa_upgrade_txholds(tx, zp);
2092 error = dmu_tx_assign(tx, TXG_NOWAIT);
2094 mutex_exit(&zp->z_acl_lock);
2095 mutex_exit(&zp->z_lock);
2097 if (error == ERESTART) {
2107 error = zfs_aclset_common(zp, aclp, cr, tx);
2109 ASSERT(zp->z_acl_cached == NULL);
2110 zp->z_acl_cached = aclp;
2113 zfs_fuid_sync(zfsvfs, tx);
2115 zfs_log_acl(zilog, tx, zp, vsecp, fuidp);
2118 zfs_fuid_info_free(fuidp);
2121 mutex_exit(&zp->z_lock);
2122 mutex_exit(&zp->z_acl_lock);
2128 * Check accesses of interest (AoI) against attributes of the dataset
2129 * such as read-only. Returns zero if no AoI conflict with dataset
2130 * attributes, otherwise an appropriate errno is returned.
2133 zfs_zaccess_dataset_check(znode_t *zp, uint32_t v4_mode)
2135 if ((v4_mode & WRITE_MASK) &&
2136 (zp->z_zfsvfs->z_vfs->vfs_flag & VFS_RDONLY) &&
2137 (!IS_DEVVP(ZTOV(zp)) ||
2138 (IS_DEVVP(ZTOV(zp)) && (v4_mode & WRITE_MASK_ATTRS)))) {
2143 * Only check for READONLY on non-directories.
2145 if ((v4_mode & WRITE_MASK_DATA) &&
2146 (((ZTOV(zp)->v_type != VDIR) &&
2147 (zp->z_pflags & (ZFS_READONLY | ZFS_IMMUTABLE))) ||
2148 (ZTOV(zp)->v_type == VDIR &&
2149 (zp->z_pflags & ZFS_IMMUTABLE)))) {
2153 if ((v4_mode & (ACE_DELETE | ACE_DELETE_CHILD)) &&
2154 (zp->z_pflags & ZFS_NOUNLINK)) {
2158 if (((v4_mode & (ACE_READ_DATA|ACE_EXECUTE)) &&
2159 (zp->z_pflags & ZFS_AV_QUARANTINED))) {
2167 * The primary usage of this function is to loop through all of the
2168 * ACEs in the znode, determining what accesses of interest (AoI) to
2169 * the caller are allowed or denied. The AoI are expressed as bits in
2170 * the working_mode parameter. As each ACE is processed, bits covered
2171 * by that ACE are removed from the working_mode. This removal
2172 * facilitates two things. The first is that when the working mode is
2173 * empty (= 0), we know we've looked at all the AoI. The second is
2174 * that the ACE interpretation rules don't allow a later ACE to undo
2175 * something granted or denied by an earlier ACE. Removing the
2176 * discovered access or denial enforces this rule. At the end of
2177 * processing the ACEs, all AoI that were found to be denied are
2178 * placed into the working_mode, giving the caller a mask of denied
2179 * accesses. Returns:
2180 * 0 if all AoI granted
2181 * EACCESS if the denied mask is non-zero
2182 * other error if abnormal failure (e.g., IO error)
2184 * A secondary usage of the function is to determine if any of the
2185 * AoI are granted. If an ACE grants any access in
2186 * the working_mode, we immediately short circuit out of the function.
2187 * This mode is chosen by setting anyaccess to B_TRUE. The
2188 * working_mode is not a denied access mask upon exit if the function
2189 * is used in this manner.
2192 zfs_zaccess_aces_check(znode_t *zp, uint32_t *working_mode,
2193 boolean_t anyaccess, cred_t *cr)
2195 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
2198 uid_t uid = crgetuid(cr);
2200 uint16_t type, iflags;
2201 uint16_t entry_type;
2202 uint32_t access_mask;
2203 uint32_t deny_mask = 0;
2204 zfs_ace_hdr_t *acep = NULL;
2209 zfs_fuid_map_ids(zp, cr, &fowner, &gowner);
2211 mutex_enter(&zp->z_acl_lock);
2213 error = zfs_acl_node_read(zp, B_FALSE, &aclp, B_FALSE);
2215 mutex_exit(&zp->z_acl_lock);
2219 ASSERT(zp->z_acl_cached);
2221 while ((acep = zfs_acl_next_ace(aclp, acep, &who, &access_mask,
2223 uint32_t mask_matched;
2225 if (!zfs_acl_valid_ace_type(type, iflags))
2228 if (ZTOV(zp)->v_type == VDIR && (iflags & ACE_INHERIT_ONLY_ACE))
2231 /* Skip ACE if it does not affect any AoI */
2232 mask_matched = (access_mask & *working_mode);
2236 entry_type = (iflags & ACE_TYPE_FLAGS);
2240 switch (entry_type) {
2248 case ACE_IDENTIFIER_GROUP:
2249 checkit = zfs_groupmember(zfsvfs, who, cr);
2257 if (entry_type == 0) {
2260 newid = zfs_fuid_map_id(zfsvfs, who, cr,
2262 if (newid != IDMAP_WK_CREATOR_OWNER_UID &&
2267 mutex_exit(&zp->z_acl_lock);
2274 DTRACE_PROBE3(zfs__ace__denies,
2276 zfs_ace_hdr_t *, acep,
2277 uint32_t, mask_matched);
2278 deny_mask |= mask_matched;
2280 DTRACE_PROBE3(zfs__ace__allows,
2282 zfs_ace_hdr_t *, acep,
2283 uint32_t, mask_matched);
2285 mutex_exit(&zp->z_acl_lock);
2289 *working_mode &= ~mask_matched;
2293 if (*working_mode == 0)
2297 mutex_exit(&zp->z_acl_lock);
2299 /* Put the found 'denies' back on the working mode */
2301 *working_mode |= deny_mask;
2303 } else if (*working_mode) {
2311 * Return true if any access whatsoever granted, we don't actually
2312 * care what access is granted.
2315 zfs_has_access(znode_t *zp, cred_t *cr)
2317 uint32_t have = ACE_ALL_PERMS;
2319 if (zfs_zaccess_aces_check(zp, &have, B_TRUE, cr) != 0) {
2322 owner = zfs_fuid_map_id(zp->z_zfsvfs, zp->z_uid, cr, ZFS_OWNER);
2323 return (secpolicy_vnode_any_access(cr, ZTOV(zp), owner) == 0);
2329 zfs_zaccess_common(znode_t *zp, uint32_t v4_mode, uint32_t *working_mode,
2330 boolean_t *check_privs, boolean_t skipaclchk, cred_t *cr)
2332 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
2335 *working_mode = v4_mode;
2336 *check_privs = B_TRUE;
2339 * Short circuit empty requests
2341 if (v4_mode == 0 || zfsvfs->z_replay) {
2346 if ((err = zfs_zaccess_dataset_check(zp, v4_mode)) != 0) {
2347 *check_privs = B_FALSE;
2352 * The caller requested that the ACL check be skipped. This
2353 * would only happen if the caller checked VOP_ACCESS() with a
2354 * 32 bit ACE mask and already had the appropriate permissions.
2361 return (zfs_zaccess_aces_check(zp, working_mode, B_FALSE, cr));
2365 zfs_zaccess_append(znode_t *zp, uint32_t *working_mode, boolean_t *check_privs,
2368 if (*working_mode != ACE_WRITE_DATA)
2371 return (zfs_zaccess_common(zp, ACE_APPEND_DATA, working_mode,
2372 check_privs, B_FALSE, cr));
2376 zfs_fastaccesschk_execute(znode_t *zdp, cred_t *cr)
2378 boolean_t owner = B_FALSE;
2379 boolean_t groupmbr = B_FALSE;
2381 uid_t uid = crgetuid(cr);
2384 if (zdp->z_pflags & ZFS_AV_QUARANTINED)
2387 is_attr = ((zdp->z_pflags & ZFS_XATTR) &&
2388 (ZTOV(zdp)->v_type == VDIR));
2393 mutex_enter(&zdp->z_acl_lock);
2395 if (zdp->z_pflags & ZFS_NO_EXECS_DENIED) {
2396 mutex_exit(&zdp->z_acl_lock);
2400 if (FUID_INDEX(zdp->z_uid) != 0 || FUID_INDEX(zdp->z_gid) != 0) {
2401 mutex_exit(&zdp->z_acl_lock);
2405 if (uid == zdp->z_uid) {
2407 if (zdp->z_mode & S_IXUSR) {
2408 mutex_exit(&zdp->z_acl_lock);
2411 mutex_exit(&zdp->z_acl_lock);
2415 if (groupmember(zdp->z_gid, cr)) {
2417 if (zdp->z_mode & S_IXGRP) {
2418 mutex_exit(&zdp->z_acl_lock);
2421 mutex_exit(&zdp->z_acl_lock);
2425 if (!owner && !groupmbr) {
2426 if (zdp->z_mode & S_IXOTH) {
2427 mutex_exit(&zdp->z_acl_lock);
2432 mutex_exit(&zdp->z_acl_lock);
2435 DTRACE_PROBE(zfs__fastpath__execute__access__miss);
2436 ZFS_ENTER(zdp->z_zfsvfs);
2437 error = zfs_zaccess(zdp, ACE_EXECUTE, 0, B_FALSE, cr);
2438 ZFS_EXIT(zdp->z_zfsvfs);
2443 * Determine whether Access should be granted/denied.
2444 * The least priv subsytem is always consulted as a basic privilege
2445 * can define any form of access.
2448 zfs_zaccess(znode_t *zp, int mode, int flags, boolean_t skipaclchk, cred_t *cr)
2450 uint32_t working_mode;
2453 boolean_t check_privs;
2455 znode_t *check_zp = zp;
2459 is_attr = ((zp->z_pflags & ZFS_XATTR) && (ZTOV(zp)->v_type == VDIR));
2462 * If attribute then validate against base file
2467 if ((error = sa_lookup(zp->z_sa_hdl,
2468 SA_ZPL_PARENT(zp->z_zfsvfs), &parent,
2469 sizeof (parent))) != 0)
2472 if ((error = zfs_zget(zp->z_zfsvfs,
2473 parent, &xzp)) != 0) {
2480 * fixup mode to map to xattr perms
2483 if (mode & (ACE_WRITE_DATA|ACE_APPEND_DATA)) {
2484 mode &= ~(ACE_WRITE_DATA|ACE_APPEND_DATA);
2485 mode |= ACE_WRITE_NAMED_ATTRS;
2488 if (mode & (ACE_READ_DATA|ACE_EXECUTE)) {
2489 mode &= ~(ACE_READ_DATA|ACE_EXECUTE);
2490 mode |= ACE_READ_NAMED_ATTRS;
2494 owner = zfs_fuid_map_id(zp->z_zfsvfs, zp->z_uid, cr, ZFS_OWNER);
2496 * Map the bits required to the standard vnode flags VREAD|VWRITE|VEXEC
2497 * in needed_bits. Map the bits mapped by working_mode (currently
2498 * missing) in missing_bits.
2499 * Call secpolicy_vnode_access2() with (needed_bits & ~checkmode),
2504 working_mode = mode;
2505 if ((working_mode & (ACE_READ_ACL|ACE_READ_ATTRIBUTES)) &&
2506 owner == crgetuid(cr))
2507 working_mode &= ~(ACE_READ_ACL|ACE_READ_ATTRIBUTES);
2509 if (working_mode & (ACE_READ_DATA|ACE_READ_NAMED_ATTRS|
2510 ACE_READ_ACL|ACE_READ_ATTRIBUTES|ACE_SYNCHRONIZE))
2511 needed_bits |= VREAD;
2512 if (working_mode & (ACE_WRITE_DATA|ACE_WRITE_NAMED_ATTRS|
2513 ACE_APPEND_DATA|ACE_WRITE_ATTRIBUTES|ACE_SYNCHRONIZE))
2514 needed_bits |= VWRITE;
2515 if (working_mode & ACE_EXECUTE)
2516 needed_bits |= VEXEC;
2518 if ((error = zfs_zaccess_common(check_zp, mode, &working_mode,
2519 &check_privs, skipaclchk, cr)) == 0) {
2522 return (secpolicy_vnode_access2(cr, ZTOV(zp), owner,
2523 needed_bits, needed_bits));
2526 if (error && !check_privs) {
2532 if (error && (flags & V_APPEND)) {
2533 error = zfs_zaccess_append(zp, &working_mode, &check_privs, cr);
2536 if (error && check_privs) {
2537 mode_t checkmode = 0;
2540 * First check for implicit owner permission on
2541 * read_acl/read_attributes
2545 ASSERT(working_mode != 0);
2547 if ((working_mode & (ACE_READ_ACL|ACE_READ_ATTRIBUTES) &&
2548 owner == crgetuid(cr)))
2549 working_mode &= ~(ACE_READ_ACL|ACE_READ_ATTRIBUTES);
2551 if (working_mode & (ACE_READ_DATA|ACE_READ_NAMED_ATTRS|
2552 ACE_READ_ACL|ACE_READ_ATTRIBUTES|ACE_SYNCHRONIZE))
2554 if (working_mode & (ACE_WRITE_DATA|ACE_WRITE_NAMED_ATTRS|
2555 ACE_APPEND_DATA|ACE_WRITE_ATTRIBUTES|ACE_SYNCHRONIZE))
2556 checkmode |= VWRITE;
2557 if (working_mode & ACE_EXECUTE)
2560 error = secpolicy_vnode_access2(cr, ZTOV(check_zp), owner,
2561 needed_bits & ~checkmode, needed_bits);
2563 if (error == 0 && (working_mode & ACE_WRITE_OWNER))
2564 error = secpolicy_vnode_chown(cr, owner);
2565 if (error == 0 && (working_mode & ACE_WRITE_ACL))
2566 error = secpolicy_vnode_setdac(cr, owner);
2568 if (error == 0 && (working_mode &
2569 (ACE_DELETE|ACE_DELETE_CHILD)))
2570 error = secpolicy_vnode_remove(cr);
2572 if (error == 0 && (working_mode & ACE_SYNCHRONIZE)) {
2573 error = secpolicy_vnode_chown(cr, owner);
2577 * See if any bits other than those already checked
2578 * for are still present. If so then return EACCES
2580 if (working_mode & ~(ZFS_CHECKED_MASKS)) {
2584 } else if (error == 0) {
2585 error = secpolicy_vnode_access2(cr, ZTOV(zp), owner,
2586 needed_bits, needed_bits);
2597 * Translate traditional unix VREAD/VWRITE/VEXEC mode into
2598 * native ACL format and call zfs_zaccess()
2601 zfs_zaccess_rwx(znode_t *zp, mode_t mode, int flags, cred_t *cr)
2603 return (zfs_zaccess(zp, zfs_unix_to_v4(mode >> 6), flags, B_FALSE, cr));
2607 * Access function for secpolicy_vnode_setattr
2610 zfs_zaccess_unix(znode_t *zp, mode_t mode, cred_t *cr)
2612 int v4_mode = zfs_unix_to_v4(mode >> 6);
2614 return (zfs_zaccess(zp, v4_mode, 0, B_FALSE, cr));
2618 zfs_delete_final_check(znode_t *zp, znode_t *dzp,
2619 mode_t available_perms, cred_t *cr)
2624 downer = zfs_fuid_map_id(dzp->z_zfsvfs, dzp->z_uid, cr, ZFS_OWNER);
2626 error = secpolicy_vnode_access2(cr, ZTOV(dzp),
2627 downer, available_perms, VWRITE|VEXEC);
2630 error = zfs_sticky_remove_access(dzp, zp, cr);
2636 * Determine whether Access should be granted/deny, without
2637 * consulting least priv subsystem.
2640 * The following chart is the recommended NFSv4 enforcement for
2641 * ability to delete an object.
2643 * -------------------------------------------------------
2644 * | Parent Dir | Target Object Permissions |
2646 * -------------------------------------------------------
2647 * | | ACL Allows | ACL Denies| Delete |
2648 * | | Delete | Delete | unspecified|
2649 * -------------------------------------------------------
2650 * | ACL Allows | Permit | Permit | Permit |
2651 * | DELETE_CHILD | |
2652 * -------------------------------------------------------
2653 * | ACL Denies | Permit | Deny | Deny |
2654 * | DELETE_CHILD | | | |
2655 * -------------------------------------------------------
2656 * | ACL specifies | | | |
2657 * | only allow | Permit | Permit | Permit |
2658 * | write and | | | |
2660 * -------------------------------------------------------
2661 * | ACL denies | | | |
2662 * | write and | Permit | Deny | Deny |
2664 * -------------------------------------------------------
2667 * No search privilege, can't even look up file?
2671 zfs_zaccess_delete(znode_t *dzp, znode_t *zp, cred_t *cr)
2673 uint32_t dzp_working_mode = 0;
2674 uint32_t zp_working_mode = 0;
2675 int dzp_error, zp_error;
2676 mode_t available_perms;
2677 boolean_t dzpcheck_privs = B_TRUE;
2678 boolean_t zpcheck_privs = B_TRUE;
2681 * We want specific DELETE permissions to
2682 * take precedence over WRITE/EXECUTE. We don't
2683 * want an ACL such as this to mess us up.
2684 * user:joe:write_data:deny,user:joe:delete:allow
2686 * However, deny permissions may ultimately be overridden
2687 * by secpolicy_vnode_access().
2689 * We will ask for all of the necessary permissions and then
2690 * look at the working modes from the directory and target object
2691 * to determine what was found.
2694 if (zp->z_pflags & (ZFS_IMMUTABLE | ZFS_NOUNLINK))
2699 * If the directory permissions allow the delete, we are done.
2701 if ((dzp_error = zfs_zaccess_common(dzp, ACE_DELETE_CHILD,
2702 &dzp_working_mode, &dzpcheck_privs, B_FALSE, cr)) == 0)
2706 * If target object has delete permission then we are done
2708 if ((zp_error = zfs_zaccess_common(zp, ACE_DELETE, &zp_working_mode,
2709 &zpcheck_privs, B_FALSE, cr)) == 0)
2712 ASSERT(dzp_error && zp_error);
2714 if (!dzpcheck_privs)
2722 * If directory returns EACCES then delete_child was denied
2723 * due to deny delete_child. In this case send the request through
2724 * secpolicy_vnode_remove(). We don't use zfs_delete_final_check()
2725 * since that *could* allow the delete based on write/execute permission
2726 * and we want delete permissions to override write/execute.
2729 if (dzp_error == EACCES)
2730 return (secpolicy_vnode_remove(cr));
2734 * only need to see if we have write/execute on directory.
2737 dzp_error = zfs_zaccess_common(dzp, ACE_EXECUTE|ACE_WRITE_DATA,
2738 &dzp_working_mode, &dzpcheck_privs, B_FALSE, cr);
2740 if (dzp_error != 0 && !dzpcheck_privs)
2747 available_perms = (dzp_working_mode & ACE_WRITE_DATA) ? 0 : VWRITE;
2748 available_perms |= (dzp_working_mode & ACE_EXECUTE) ? 0 : VEXEC;
2750 return (zfs_delete_final_check(zp, dzp, available_perms, cr));
2755 zfs_zaccess_rename(znode_t *sdzp, znode_t *szp, znode_t *tdzp,
2756 znode_t *tzp, cred_t *cr)
2761 if (szp->z_pflags & ZFS_AV_QUARANTINED)
2764 add_perm = (ZTOV(szp)->v_type == VDIR) ?
2765 ACE_ADD_SUBDIRECTORY : ACE_ADD_FILE;
2768 * Rename permissions are combination of delete permission +
2769 * add file/subdir permission.
2773 * first make sure we do the delete portion.
2775 * If that succeeds then check for add_file/add_subdir permissions
2778 if ((error = zfs_zaccess_delete(sdzp, szp, cr)))
2782 * If we have a tzp, see if we can delete it?
2785 if ((error = zfs_zaccess_delete(tdzp, tzp, cr)))
2790 * Now check for add permissions
2792 error = zfs_zaccess(tdzp, add_perm, 0, B_FALSE, cr);
2797 #endif /* HAVE_ZPL */