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.
26 #include <sys/types.h>
27 #include <sys/param.h>
29 #include <sys/systm.h>
30 #include <sys/sysmacros.h>
31 #include <sys/resource.h>
33 #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>
44 #include <sys/policy.h>
45 #include <sys/zfs_znode.h>
46 #include <sys/zfs_fuid.h>
47 #include <sys/zfs_acl.h>
48 #include <sys/zfs_dir.h>
49 #include <sys/zfs_vfsops.h>
51 #include <sys/dnode.h>
54 #include "fs/fs_subr.h"
56 #define ALLOW ACE_ACCESS_ALLOWED_ACE_TYPE
57 #define DENY ACE_ACCESS_DENIED_ACE_TYPE
58 #define MAX_ACE_TYPE ACE_SYSTEM_ALARM_CALLBACK_OBJECT_ACE_TYPE
59 #define MIN_ACE_TYPE ALLOW
61 #define OWNING_GROUP (ACE_GROUP|ACE_IDENTIFIER_GROUP)
62 #define EVERYONE_ALLOW_MASK (ACE_READ_ACL|ACE_READ_ATTRIBUTES | \
63 ACE_READ_NAMED_ATTRS|ACE_SYNCHRONIZE)
64 #define EVERYONE_DENY_MASK (ACE_WRITE_ACL|ACE_WRITE_OWNER | \
65 ACE_WRITE_ATTRIBUTES|ACE_WRITE_NAMED_ATTRS)
66 #define OWNER_ALLOW_MASK (ACE_WRITE_ACL | ACE_WRITE_OWNER | \
67 ACE_WRITE_ATTRIBUTES|ACE_WRITE_NAMED_ATTRS)
69 #define ZFS_CHECKED_MASKS (ACE_READ_ACL|ACE_READ_ATTRIBUTES|ACE_READ_DATA| \
70 ACE_READ_NAMED_ATTRS|ACE_WRITE_DATA|ACE_WRITE_ATTRIBUTES| \
71 ACE_WRITE_NAMED_ATTRS|ACE_APPEND_DATA|ACE_EXECUTE|ACE_WRITE_OWNER| \
72 ACE_WRITE_ACL|ACE_DELETE|ACE_DELETE_CHILD|ACE_SYNCHRONIZE)
74 #define WRITE_MASK_DATA (ACE_WRITE_DATA|ACE_APPEND_DATA|ACE_WRITE_NAMED_ATTRS)
75 #define WRITE_MASK_ATTRS (ACE_WRITE_ACL|ACE_WRITE_OWNER|ACE_WRITE_ATTRIBUTES| \
76 ACE_DELETE|ACE_DELETE_CHILD)
77 #define WRITE_MASK (WRITE_MASK_DATA|WRITE_MASK_ATTRS)
79 #define OGE_CLEAR (ACE_READ_DATA|ACE_LIST_DIRECTORY|ACE_WRITE_DATA| \
80 ACE_ADD_FILE|ACE_APPEND_DATA|ACE_ADD_SUBDIRECTORY|ACE_EXECUTE)
82 #define OKAY_MASK_BITS (ACE_READ_DATA|ACE_LIST_DIRECTORY|ACE_WRITE_DATA| \
83 ACE_ADD_FILE|ACE_APPEND_DATA|ACE_ADD_SUBDIRECTORY|ACE_EXECUTE)
85 #define ALL_INHERIT (ACE_FILE_INHERIT_ACE|ACE_DIRECTORY_INHERIT_ACE | \
86 ACE_NO_PROPAGATE_INHERIT_ACE|ACE_INHERIT_ONLY_ACE|ACE_INHERITED_ACE)
88 #define RESTRICTED_CLEAR (ACE_WRITE_ACL|ACE_WRITE_OWNER)
90 #define V4_ACL_WIDE_FLAGS (ZFS_ACL_AUTO_INHERIT|ZFS_ACL_DEFAULTED|\
93 #define ZFS_ACL_WIDE_FLAGS (V4_ACL_WIDE_FLAGS|ZFS_ACL_TRIVIAL|ZFS_INHERIT_ACE|\
96 #define ALL_MODE_EXECS (S_IXUSR | S_IXGRP | S_IXOTH)
99 zfs_ace_v0_get_type(void *acep)
101 return (((zfs_oldace_t *)acep)->z_type);
105 zfs_ace_v0_get_flags(void *acep)
107 return (((zfs_oldace_t *)acep)->z_flags);
111 zfs_ace_v0_get_mask(void *acep)
113 return (((zfs_oldace_t *)acep)->z_access_mask);
117 zfs_ace_v0_get_who(void *acep)
119 return (((zfs_oldace_t *)acep)->z_fuid);
123 zfs_ace_v0_set_type(void *acep, uint16_t type)
125 ((zfs_oldace_t *)acep)->z_type = type;
129 zfs_ace_v0_set_flags(void *acep, uint16_t flags)
131 ((zfs_oldace_t *)acep)->z_flags = flags;
135 zfs_ace_v0_set_mask(void *acep, uint32_t mask)
137 ((zfs_oldace_t *)acep)->z_access_mask = mask;
141 zfs_ace_v0_set_who(void *acep, uint64_t who)
143 ((zfs_oldace_t *)acep)->z_fuid = who;
148 zfs_ace_v0_size(void *acep)
150 return (sizeof (zfs_oldace_t));
154 zfs_ace_v0_abstract_size(void)
156 return (sizeof (zfs_oldace_t));
160 zfs_ace_v0_mask_off(void)
162 return (offsetof(zfs_oldace_t, z_access_mask));
167 zfs_ace_v0_data(void *acep, void **datap)
173 static acl_ops_t zfs_acl_v0_ops = {
176 zfs_ace_v0_get_flags,
177 zfs_ace_v0_set_flags,
183 zfs_ace_v0_abstract_size,
189 zfs_ace_fuid_get_type(void *acep)
191 return (((zfs_ace_hdr_t *)acep)->z_type);
195 zfs_ace_fuid_get_flags(void *acep)
197 return (((zfs_ace_hdr_t *)acep)->z_flags);
201 zfs_ace_fuid_get_mask(void *acep)
203 return (((zfs_ace_hdr_t *)acep)->z_access_mask);
207 zfs_ace_fuid_get_who(void *args)
210 zfs_ace_t *acep = args;
212 entry_type = acep->z_hdr.z_flags & ACE_TYPE_FLAGS;
214 if (entry_type == ACE_OWNER || entry_type == OWNING_GROUP ||
215 entry_type == ACE_EVERYONE)
217 return (((zfs_ace_t *)acep)->z_fuid);
221 zfs_ace_fuid_set_type(void *acep, uint16_t type)
223 ((zfs_ace_hdr_t *)acep)->z_type = type;
227 zfs_ace_fuid_set_flags(void *acep, uint16_t flags)
229 ((zfs_ace_hdr_t *)acep)->z_flags = flags;
233 zfs_ace_fuid_set_mask(void *acep, uint32_t mask)
235 ((zfs_ace_hdr_t *)acep)->z_access_mask = mask;
239 zfs_ace_fuid_set_who(void *arg, uint64_t who)
241 zfs_ace_t *acep = arg;
243 uint16_t entry_type = acep->z_hdr.z_flags & ACE_TYPE_FLAGS;
245 if (entry_type == ACE_OWNER || entry_type == OWNING_GROUP ||
246 entry_type == ACE_EVERYONE)
252 zfs_ace_fuid_size(void *acep)
254 zfs_ace_hdr_t *zacep = acep;
257 switch (zacep->z_type) {
258 case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE:
259 case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE:
260 case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE:
261 case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE:
262 return (sizeof (zfs_object_ace_t));
266 (((zfs_ace_hdr_t *)acep)->z_flags & ACE_TYPE_FLAGS);
267 if (entry_type == ACE_OWNER ||
268 entry_type == OWNING_GROUP ||
269 entry_type == ACE_EVERYONE)
270 return (sizeof (zfs_ace_hdr_t));
273 return (sizeof (zfs_ace_t));
278 zfs_ace_fuid_abstract_size(void)
280 return (sizeof (zfs_ace_hdr_t));
284 zfs_ace_fuid_mask_off(void)
286 return (offsetof(zfs_ace_hdr_t, z_access_mask));
290 zfs_ace_fuid_data(void *acep, void **datap)
292 zfs_ace_t *zacep = acep;
293 zfs_object_ace_t *zobjp;
295 switch (zacep->z_hdr.z_type) {
296 case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE:
297 case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE:
298 case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE:
299 case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE:
301 *datap = (caddr_t)zobjp + sizeof (zfs_ace_t);
302 return (sizeof (zfs_object_ace_t) - sizeof (zfs_ace_t));
309 static acl_ops_t zfs_acl_fuid_ops = {
310 zfs_ace_fuid_get_mask,
311 zfs_ace_fuid_set_mask,
312 zfs_ace_fuid_get_flags,
313 zfs_ace_fuid_set_flags,
314 zfs_ace_fuid_get_type,
315 zfs_ace_fuid_set_type,
316 zfs_ace_fuid_get_who,
317 zfs_ace_fuid_set_who,
319 zfs_ace_fuid_abstract_size,
320 zfs_ace_fuid_mask_off,
325 * The following three functions are provided for compatibility with
326 * older ZPL version in order to determine if the file use to have
327 * an external ACL and what version of ACL previously existed on the
328 * file. Would really be nice to not need this, sigh.
331 zfs_external_acl(znode_t *zp)
333 zfs_acl_phys_t acl_phys;
340 * Need to deal with a potential
341 * race where zfs_sa_upgrade could cause
342 * z_isa_sa to change.
344 * If the lookup fails then the state of z_is_sa should have
348 if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_ZNODE_ACL(ZTOZSB(zp)),
349 &acl_phys, sizeof (acl_phys))) == 0)
350 return (acl_phys.z_acl_extern_obj);
353 * after upgrade the SA_ZPL_ZNODE_ACL should have been
356 VERIFY(zp->z_is_sa && error == ENOENT);
362 * Determine size of ACL in bytes
364 * This is more complicated than it should be since we have to deal
365 * with old external ACLs.
368 zfs_acl_znode_info(znode_t *zp, int *aclsize, int *aclcount,
369 zfs_acl_phys_t *aclphys)
371 zfs_sb_t *zsb = ZTOZSB(zp);
376 ASSERT(MUTEX_HELD(&zp->z_acl_lock));
378 if ((error = sa_size(zp->z_sa_hdl, SA_ZPL_DACL_ACES(zsb),
382 if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_DACL_COUNT(zsb),
383 &acl_count, sizeof (acl_count))) != 0)
385 *aclcount = acl_count;
387 if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_ZNODE_ACL(zsb),
388 aclphys, sizeof (*aclphys))) != 0)
391 if (aclphys->z_acl_version == ZFS_ACL_VERSION_INITIAL) {
392 *aclsize = ZFS_ACL_SIZE(aclphys->z_acl_size);
393 *aclcount = aclphys->z_acl_size;
395 *aclsize = aclphys->z_acl_size;
396 *aclcount = aclphys->z_acl_count;
403 zfs_znode_acl_version(znode_t *zp)
405 zfs_acl_phys_t acl_phys;
408 return (ZFS_ACL_VERSION_FUID);
413 * Need to deal with a potential
414 * race where zfs_sa_upgrade could cause
415 * z_isa_sa to change.
417 * If the lookup fails then the state of z_is_sa should have
420 if ((error = sa_lookup(zp->z_sa_hdl,
421 SA_ZPL_ZNODE_ACL(ZTOZSB(zp)),
422 &acl_phys, sizeof (acl_phys))) == 0)
423 return (acl_phys.z_acl_version);
426 * After upgrade SA_ZPL_ZNODE_ACL should have
429 VERIFY(zp->z_is_sa && error == ENOENT);
430 return (ZFS_ACL_VERSION_FUID);
436 zfs_acl_version(int version)
438 if (version < ZPL_VERSION_FUID)
439 return (ZFS_ACL_VERSION_INITIAL);
441 return (ZFS_ACL_VERSION_FUID);
445 zfs_acl_version_zp(znode_t *zp)
447 return (zfs_acl_version(ZTOZSB(zp)->z_version));
451 zfs_acl_alloc(int vers)
455 aclp = kmem_zalloc(sizeof (zfs_acl_t), KM_PUSHPAGE);
456 list_create(&aclp->z_acl, sizeof (zfs_acl_node_t),
457 offsetof(zfs_acl_node_t, z_next));
458 aclp->z_version = vers;
459 if (vers == ZFS_ACL_VERSION_FUID)
460 aclp->z_ops = &zfs_acl_fuid_ops;
462 aclp->z_ops = &zfs_acl_v0_ops;
467 zfs_acl_node_alloc(size_t bytes)
469 zfs_acl_node_t *aclnode;
471 aclnode = kmem_zalloc(sizeof (zfs_acl_node_t), KM_PUSHPAGE);
473 aclnode->z_acldata = kmem_alloc(bytes, KM_PUSHPAGE);
474 aclnode->z_allocdata = aclnode->z_acldata;
475 aclnode->z_allocsize = bytes;
476 aclnode->z_size = bytes;
483 zfs_acl_node_free(zfs_acl_node_t *aclnode)
485 if (aclnode->z_allocsize)
486 kmem_free(aclnode->z_allocdata, aclnode->z_allocsize);
487 kmem_free(aclnode, sizeof (zfs_acl_node_t));
491 zfs_acl_release_nodes(zfs_acl_t *aclp)
493 zfs_acl_node_t *aclnode;
495 while ((aclnode = list_head(&aclp->z_acl))) {
496 list_remove(&aclp->z_acl, aclnode);
497 zfs_acl_node_free(aclnode);
499 aclp->z_acl_count = 0;
500 aclp->z_acl_bytes = 0;
504 zfs_acl_free(zfs_acl_t *aclp)
506 zfs_acl_release_nodes(aclp);
507 list_destroy(&aclp->z_acl);
508 kmem_free(aclp, sizeof (zfs_acl_t));
512 zfs_acl_valid_ace_type(uint_t type, uint_t flags)
519 case ACE_SYSTEM_AUDIT_ACE_TYPE:
520 case ACE_SYSTEM_ALARM_ACE_TYPE:
521 entry_type = flags & ACE_TYPE_FLAGS;
522 return (entry_type == ACE_OWNER ||
523 entry_type == OWNING_GROUP ||
524 entry_type == ACE_EVERYONE || entry_type == 0 ||
525 entry_type == ACE_IDENTIFIER_GROUP);
527 if (type >= MIN_ACE_TYPE && type <= MAX_ACE_TYPE)
534 zfs_ace_valid(umode_t obj_mode, zfs_acl_t *aclp, uint16_t type, uint16_t iflags)
537 * first check type of entry
540 if (!zfs_acl_valid_ace_type(type, iflags))
544 case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE:
545 case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE:
546 case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE:
547 case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE:
548 if (aclp->z_version < ZFS_ACL_VERSION_FUID)
550 aclp->z_hints |= ZFS_ACL_OBJ_ACE;
554 * next check inheritance level flags
557 if (S_ISDIR(obj_mode) &&
558 (iflags & (ACE_FILE_INHERIT_ACE|ACE_DIRECTORY_INHERIT_ACE)))
559 aclp->z_hints |= ZFS_INHERIT_ACE;
561 if (iflags & (ACE_INHERIT_ONLY_ACE|ACE_NO_PROPAGATE_INHERIT_ACE)) {
562 if ((iflags & (ACE_FILE_INHERIT_ACE|
563 ACE_DIRECTORY_INHERIT_ACE)) == 0) {
572 zfs_acl_next_ace(zfs_acl_t *aclp, void *start, uint64_t *who,
573 uint32_t *access_mask, uint16_t *iflags, uint16_t *type)
575 zfs_acl_node_t *aclnode;
580 aclnode = list_head(&aclp->z_acl);
584 aclp->z_next_ace = aclnode->z_acldata;
585 aclp->z_curr_node = aclnode;
586 aclnode->z_ace_idx = 0;
589 aclnode = aclp->z_curr_node;
594 if (aclnode->z_ace_idx >= aclnode->z_ace_count) {
595 aclnode = list_next(&aclp->z_acl, aclnode);
599 aclp->z_curr_node = aclnode;
600 aclnode->z_ace_idx = 0;
601 aclp->z_next_ace = aclnode->z_acldata;
605 if (aclnode->z_ace_idx < aclnode->z_ace_count) {
606 void *acep = aclp->z_next_ace;
610 * Make sure we don't overstep our bounds
612 ace_size = aclp->z_ops->ace_size(acep);
614 if (((caddr_t)acep + ace_size) >
615 ((caddr_t)aclnode->z_acldata + aclnode->z_size)) {
619 *iflags = aclp->z_ops->ace_flags_get(acep);
620 *type = aclp->z_ops->ace_type_get(acep);
621 *access_mask = aclp->z_ops->ace_mask_get(acep);
622 *who = aclp->z_ops->ace_who_get(acep);
623 aclp->z_next_ace = (caddr_t)aclp->z_next_ace + ace_size;
624 aclnode->z_ace_idx++;
626 return ((void *)acep);
633 zfs_ace_walk(void *datap, uint64_t cookie, int aclcnt,
634 uint16_t *flags, uint16_t *type, uint32_t *mask)
636 zfs_acl_t *aclp = datap;
637 zfs_ace_hdr_t *acep = (zfs_ace_hdr_t *)(uintptr_t)cookie;
640 acep = zfs_acl_next_ace(aclp, acep, &who, mask,
642 return ((uint64_t)(uintptr_t)acep);
646 * Copy ACE to internal ZFS format.
647 * While processing the ACL each ACE will be validated for correctness.
648 * ACE FUIDs will be created later.
651 zfs_copy_ace_2_fuid(zfs_sb_t *zsb, umode_t obj_mode, zfs_acl_t *aclp,
652 void *datap, zfs_ace_t *z_acl, uint64_t aclcnt, size_t *size,
653 zfs_fuid_info_t **fuidp, cred_t *cr)
657 zfs_ace_t *aceptr = z_acl;
659 zfs_object_ace_t *zobjacep;
660 ace_object_t *aceobjp;
662 for (i = 0; i != aclcnt; i++) {
663 aceptr->z_hdr.z_access_mask = acep->a_access_mask;
664 aceptr->z_hdr.z_flags = acep->a_flags;
665 aceptr->z_hdr.z_type = acep->a_type;
666 entry_type = aceptr->z_hdr.z_flags & ACE_TYPE_FLAGS;
667 if (entry_type != ACE_OWNER && entry_type != OWNING_GROUP &&
668 entry_type != ACE_EVERYONE) {
669 aceptr->z_fuid = zfs_fuid_create(zsb, acep->a_who,
670 cr, (entry_type == 0) ?
671 ZFS_ACE_USER : ZFS_ACE_GROUP, fuidp);
675 * Make sure ACE is valid
677 if (zfs_ace_valid(obj_mode, aclp, aceptr->z_hdr.z_type,
678 aceptr->z_hdr.z_flags) != B_TRUE)
681 switch (acep->a_type) {
682 case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE:
683 case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE:
684 case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE:
685 case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE:
686 zobjacep = (zfs_object_ace_t *)aceptr;
687 aceobjp = (ace_object_t *)acep;
689 bcopy(aceobjp->a_obj_type, zobjacep->z_object_type,
690 sizeof (aceobjp->a_obj_type));
691 bcopy(aceobjp->a_inherit_obj_type,
692 zobjacep->z_inherit_type,
693 sizeof (aceobjp->a_inherit_obj_type));
694 acep = (ace_t *)((caddr_t)acep + sizeof (ace_object_t));
697 acep = (ace_t *)((caddr_t)acep + sizeof (ace_t));
700 aceptr = (zfs_ace_t *)((caddr_t)aceptr +
701 aclp->z_ops->ace_size(aceptr));
704 *size = (caddr_t)aceptr - (caddr_t)z_acl;
710 * Copy ZFS ACEs to fixed size ace_t layout
713 zfs_copy_fuid_2_ace(zfs_sb_t *zsb, zfs_acl_t *aclp, cred_t *cr,
714 void *datap, int filter)
717 uint32_t access_mask;
718 uint16_t iflags, type;
719 zfs_ace_hdr_t *zacep = NULL;
721 ace_object_t *objacep;
722 zfs_object_ace_t *zobjacep;
726 while ((zacep = zfs_acl_next_ace(aclp, zacep,
727 &who, &access_mask, &iflags, &type))) {
730 case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE:
731 case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE:
732 case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE:
733 case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE:
737 zobjacep = (zfs_object_ace_t *)zacep;
738 objacep = (ace_object_t *)acep;
739 bcopy(zobjacep->z_object_type,
741 sizeof (zobjacep->z_object_type));
742 bcopy(zobjacep->z_inherit_type,
743 objacep->a_inherit_obj_type,
744 sizeof (zobjacep->z_inherit_type));
745 ace_size = sizeof (ace_object_t);
748 ace_size = sizeof (ace_t);
752 entry_type = (iflags & ACE_TYPE_FLAGS);
753 if ((entry_type != ACE_OWNER &&
754 entry_type != OWNING_GROUP &&
755 entry_type != ACE_EVERYONE)) {
756 acep->a_who = zfs_fuid_map_id(zsb, who,
757 cr, (entry_type & ACE_IDENTIFIER_GROUP) ?
758 ZFS_ACE_GROUP : ZFS_ACE_USER);
760 acep->a_who = (uid_t)(int64_t)who;
762 acep->a_access_mask = access_mask;
763 acep->a_flags = iflags;
765 acep = (ace_t *)((caddr_t)acep + ace_size);
770 zfs_copy_ace_2_oldace(umode_t obj_mode, zfs_acl_t *aclp, ace_t *acep,
771 zfs_oldace_t *z_acl, int aclcnt, size_t *size)
774 zfs_oldace_t *aceptr = z_acl;
776 for (i = 0; i != aclcnt; i++, aceptr++) {
777 aceptr->z_access_mask = acep[i].a_access_mask;
778 aceptr->z_type = acep[i].a_type;
779 aceptr->z_flags = acep[i].a_flags;
780 aceptr->z_fuid = acep[i].a_who;
782 * Make sure ACE is valid
784 if (zfs_ace_valid(obj_mode, aclp, aceptr->z_type,
785 aceptr->z_flags) != B_TRUE)
788 *size = (caddr_t)aceptr - (caddr_t)z_acl;
793 * convert old ACL format to new
796 zfs_acl_xform(znode_t *zp, zfs_acl_t *aclp, cred_t *cr)
798 zfs_oldace_t *oldaclp;
800 uint16_t type, iflags;
801 uint32_t access_mask;
804 zfs_acl_node_t *newaclnode;
806 ASSERT(aclp->z_version == ZFS_ACL_VERSION_INITIAL);
808 * First create the ACE in a contiguous piece of memory
809 * for zfs_copy_ace_2_fuid().
811 * We only convert an ACL once, so this won't happen
814 oldaclp = kmem_alloc(sizeof (zfs_oldace_t) * aclp->z_acl_count,
817 while ((cookie = zfs_acl_next_ace(aclp, cookie, &who,
818 &access_mask, &iflags, &type))) {
819 oldaclp[i].z_flags = iflags;
820 oldaclp[i].z_type = type;
821 oldaclp[i].z_fuid = who;
822 oldaclp[i++].z_access_mask = access_mask;
825 newaclnode = zfs_acl_node_alloc(aclp->z_acl_count *
826 sizeof (zfs_object_ace_t));
827 aclp->z_ops = &zfs_acl_fuid_ops;
828 VERIFY(zfs_copy_ace_2_fuid(ZTOZSB(zp), ZTOI(zp)->i_mode,
829 aclp, oldaclp, newaclnode->z_acldata, aclp->z_acl_count,
830 &newaclnode->z_size, NULL, cr) == 0);
831 newaclnode->z_ace_count = aclp->z_acl_count;
832 aclp->z_version = ZFS_ACL_VERSION;
833 kmem_free(oldaclp, aclp->z_acl_count * sizeof (zfs_oldace_t));
836 * Release all previous ACL nodes
839 zfs_acl_release_nodes(aclp);
841 list_insert_head(&aclp->z_acl, newaclnode);
843 aclp->z_acl_bytes = newaclnode->z_size;
844 aclp->z_acl_count = newaclnode->z_ace_count;
849 * Convert unix access mask to v4 access mask
852 zfs_unix_to_v4(uint32_t access_mask)
854 uint32_t new_mask = 0;
856 if (access_mask & S_IXOTH)
857 new_mask |= ACE_EXECUTE;
858 if (access_mask & S_IWOTH)
859 new_mask |= ACE_WRITE_DATA;
860 if (access_mask & S_IROTH)
861 new_mask |= ACE_READ_DATA;
866 zfs_set_ace(zfs_acl_t *aclp, void *acep, uint32_t access_mask,
867 uint16_t access_type, uint64_t fuid, uint16_t entry_type)
869 uint16_t type = entry_type & ACE_TYPE_FLAGS;
871 aclp->z_ops->ace_mask_set(acep, access_mask);
872 aclp->z_ops->ace_type_set(acep, access_type);
873 aclp->z_ops->ace_flags_set(acep, entry_type);
874 if ((type != ACE_OWNER && type != OWNING_GROUP &&
875 type != ACE_EVERYONE))
876 aclp->z_ops->ace_who_set(acep, fuid);
880 * Determine mode of file based on ACL.
881 * Also, create FUIDs for any User/Group ACEs
884 zfs_mode_compute(uint64_t fmode, zfs_acl_t *aclp,
885 uint64_t *pflags, uint64_t fuid, uint64_t fgid)
890 zfs_ace_hdr_t *acep = NULL;
892 uint16_t iflags, type;
893 uint32_t access_mask;
894 boolean_t an_exec_denied = B_FALSE;
896 mode = (fmode & (S_IFMT | S_ISUID | S_ISGID | S_ISVTX));
898 while ((acep = zfs_acl_next_ace(aclp, acep, &who,
899 &access_mask, &iflags, &type))) {
901 if (!zfs_acl_valid_ace_type(type, iflags))
904 entry_type = (iflags & ACE_TYPE_FLAGS);
907 * Skip over owner@, group@ or everyone@ inherit only ACEs
909 if ((iflags & ACE_INHERIT_ONLY_ACE) &&
910 (entry_type == ACE_OWNER || entry_type == ACE_EVERYONE ||
911 entry_type == OWNING_GROUP))
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;
1065 boolean_t drop_lock = B_FALSE;
1067 ASSERT(MUTEX_HELD(&zp->z_acl_lock));
1069 if (zp->z_acl_cached && !will_modify) {
1070 *aclpp = zp->z_acl_cached;
1075 * close race where znode could be upgrade while trying to
1076 * read the znode attributes.
1078 * But this could only happen if the file isn't already an SA
1081 if (!zp->z_is_sa && !have_lock) {
1082 mutex_enter(&zp->z_lock);
1085 version = zfs_znode_acl_version(zp);
1087 if ((error = zfs_acl_znode_info(zp, &aclsize,
1088 &acl_count, &znode_acl)) != 0) {
1092 aclp = zfs_acl_alloc(version);
1094 aclp->z_acl_count = acl_count;
1095 aclp->z_acl_bytes = aclsize;
1097 aclnode = zfs_acl_node_alloc(aclsize);
1098 aclnode->z_ace_count = aclp->z_acl_count;
1099 aclnode->z_size = aclsize;
1102 if (znode_acl.z_acl_extern_obj) {
1103 error = dmu_read(ZTOZSB(zp)->z_os,
1104 znode_acl.z_acl_extern_obj, 0, aclnode->z_size,
1105 aclnode->z_acldata, DMU_READ_PREFETCH);
1107 bcopy(znode_acl.z_ace_data, aclnode->z_acldata,
1111 error = sa_lookup(zp->z_sa_hdl, SA_ZPL_DACL_ACES(ZTOZSB(zp)),
1112 aclnode->z_acldata, aclnode->z_size);
1117 zfs_acl_node_free(aclnode);
1118 /* convert checksum errors into IO errors */
1119 if (error == ECKSUM)
1124 list_insert_head(&aclp->z_acl, aclnode);
1128 zp->z_acl_cached = aclp;
1131 mutex_exit(&zp->z_lock);
1137 zfs_acl_data_locator(void **dataptr, uint32_t *length, uint32_t buflen,
1138 boolean_t start, void *userdata)
1140 zfs_acl_locator_cb_t *cb = (zfs_acl_locator_cb_t *)userdata;
1143 cb->cb_acl_node = list_head(&cb->cb_aclp->z_acl);
1145 cb->cb_acl_node = list_next(&cb->cb_aclp->z_acl,
1148 *dataptr = cb->cb_acl_node->z_acldata;
1149 *length = cb->cb_acl_node->z_size;
1153 zfs_acl_chown_setattr(znode_t *zp)
1158 ASSERT(MUTEX_HELD(&zp->z_lock));
1159 ASSERT(MUTEX_HELD(&zp->z_acl_lock));
1161 if ((error = zfs_acl_node_read(zp, B_TRUE, &aclp, B_FALSE)) == 0)
1162 zp->z_mode = zfs_mode_compute(zp->z_mode, aclp,
1163 &zp->z_pflags, zp->z_uid, zp->z_gid);
1168 acl_trivial_access_masks(mode_t mode, uint32_t *allow0, uint32_t *deny1,
1169 uint32_t *deny2, uint32_t *owner, uint32_t *group, uint32_t *everyone)
1171 *deny1 = *deny2 = *allow0 = *group = 0;
1173 if (!(mode & S_IRUSR) && (mode & (S_IRGRP|S_IROTH)))
1174 *deny1 |= ACE_READ_DATA;
1175 if (!(mode & S_IWUSR) && (mode & (S_IWGRP|S_IWOTH)))
1176 *deny1 |= ACE_WRITE_DATA;
1177 if (!(mode & S_IXUSR) && (mode & (S_IXGRP|S_IXOTH)))
1178 *deny1 |= ACE_EXECUTE;
1180 if (!(mode & S_IRGRP) && (mode & S_IROTH))
1181 *deny2 = ACE_READ_DATA;
1182 if (!(mode & S_IWGRP) && (mode & S_IWOTH))
1183 *deny2 |= ACE_WRITE_DATA;
1184 if (!(mode & S_IXGRP) && (mode & S_IXOTH))
1185 *deny2 |= ACE_EXECUTE;
1187 if ((mode & S_IRUSR) && (!(mode & S_IRGRP) && (mode & S_IROTH)))
1188 *allow0 |= ACE_READ_DATA;
1189 if ((mode & S_IWUSR) && (!(mode & S_IWGRP) && (mode & S_IWOTH)))
1190 *allow0 |= ACE_WRITE_DATA;
1191 if ((mode & S_IXUSR) && (!(mode & S_IXGRP) && (mode & S_IXOTH)))
1192 *allow0 |= ACE_EXECUTE;
1194 *owner = ACE_WRITE_ATTRIBUTES|ACE_WRITE_OWNER|ACE_WRITE_ACL|
1195 ACE_WRITE_NAMED_ATTRS|ACE_READ_ACL|ACE_READ_ATTRIBUTES|
1196 ACE_READ_NAMED_ATTRS|ACE_SYNCHRONIZE;
1198 *owner |= ACE_READ_DATA;
1200 *owner |= ACE_WRITE_DATA|ACE_APPEND_DATA;
1202 *owner |= ACE_EXECUTE;
1204 *group = ACE_READ_ACL|ACE_READ_ATTRIBUTES| ACE_READ_NAMED_ATTRS|
1207 *group |= ACE_READ_DATA;
1209 *group |= ACE_WRITE_DATA|ACE_APPEND_DATA;
1211 *group |= ACE_EXECUTE;
1213 *everyone = ACE_READ_ACL|ACE_READ_ATTRIBUTES| ACE_READ_NAMED_ATTRS|
1216 *everyone |= ACE_READ_DATA;
1218 *everyone |= ACE_WRITE_DATA|ACE_APPEND_DATA;
1220 *everyone |= ACE_EXECUTE;
1225 * determine whether an ace_t acl is trivial
1227 * Trivialness implies that the acl is composed of only
1228 * owner, group, everyone entries. ACL can't
1229 * have read_acl denied, and write_owner/write_acl/write_attributes
1230 * can only be owner@ entry.
1233 ace_trivial_common(void *acep, int aclcnt,
1234 uint64_t (*walk)(void *, uint64_t, int aclcnt,
1235 uint16_t *, uint16_t *, uint32_t *))
1240 uint64_t cookie = 0;
1242 while ((cookie = walk(acep, cookie, aclcnt, &flags, &type, &mask))) {
1243 switch (flags & ACE_TYPE_FLAGS) {
1245 case ACE_GROUP|ACE_IDENTIFIER_GROUP:
1252 if (flags & (ACE_FILE_INHERIT_ACE|
1253 ACE_DIRECTORY_INHERIT_ACE|ACE_NO_PROPAGATE_INHERIT_ACE|
1254 ACE_INHERIT_ONLY_ACE))
1258 * Special check for some special bits
1260 * Don't allow anybody to deny reading basic
1261 * attributes or a files ACL.
1263 if ((mask & (ACE_READ_ACL|ACE_READ_ATTRIBUTES)) &&
1264 (type == ACE_ACCESS_DENIED_ACE_TYPE))
1268 * Delete permissions are never set by default
1270 if (mask & (ACE_DELETE|ACE_DELETE_CHILD))
1273 * only allow owner@ to have
1274 * write_acl/write_owner/write_attributes/write_xattr/
1276 if (type == ACE_ACCESS_ALLOWED_ACE_TYPE &&
1277 (!(flags & ACE_OWNER) && (mask &
1278 (ACE_WRITE_OWNER|ACE_WRITE_ACL| ACE_WRITE_ATTRIBUTES|
1279 ACE_WRITE_NAMED_ATTRS))))
1288 * common code for setting ACLs.
1290 * This function is called from zfs_mode_update, zfs_perm_init, and zfs_setacl.
1291 * zfs_setacl passes a non-NULL inherit pointer (ihp) to indicate that it's
1292 * already checked the acl and knows whether to inherit.
1295 zfs_aclset_common(znode_t *zp, zfs_acl_t *aclp, cred_t *cr, dmu_tx_t *tx)
1298 zfs_sb_t *zsb = ZTOZSB(zp);
1299 dmu_object_type_t otype;
1300 zfs_acl_locator_cb_t locate = { 0 };
1302 sa_bulk_attr_t bulk[5];
1308 mode = zfs_mode_compute(mode, aclp, &zp->z_pflags,
1309 zp->z_uid, zp->z_gid);
1312 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zsb), NULL,
1313 &mode, sizeof (mode));
1314 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zsb), NULL,
1315 &zp->z_pflags, sizeof (zp->z_pflags));
1316 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zsb), NULL,
1317 &ctime, sizeof (ctime));
1319 if (zp->z_acl_cached) {
1320 zfs_acl_free(zp->z_acl_cached);
1321 zp->z_acl_cached = NULL;
1327 if (!zsb->z_use_fuids) {
1328 otype = DMU_OT_OLDACL;
1330 if ((aclp->z_version == ZFS_ACL_VERSION_INITIAL) &&
1331 (zsb->z_version >= ZPL_VERSION_FUID))
1332 zfs_acl_xform(zp, aclp, cr);
1333 ASSERT(aclp->z_version >= ZFS_ACL_VERSION_FUID);
1338 * Arrgh, we have to handle old on disk format
1339 * as well as newer (preferred) SA format.
1342 if (zp->z_is_sa) { /* the easy case, just update the ACL attribute */
1343 locate.cb_aclp = aclp;
1344 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_DACL_ACES(zsb),
1345 zfs_acl_data_locator, &locate, aclp->z_acl_bytes);
1346 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_DACL_COUNT(zsb),
1347 NULL, &aclp->z_acl_count, sizeof (uint64_t));
1348 } else { /* Painful legacy way */
1349 zfs_acl_node_t *aclnode;
1351 zfs_acl_phys_t acl_phys;
1354 if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_ZNODE_ACL(zsb),
1355 &acl_phys, sizeof (acl_phys))) != 0)
1358 aoid = acl_phys.z_acl_extern_obj;
1360 if (aclp->z_acl_bytes > ZFS_ACE_SPACE) {
1362 * If ACL was previously external and we are now
1363 * converting to new ACL format then release old
1364 * ACL object and create a new one.
1367 aclp->z_version != acl_phys.z_acl_version) {
1368 error = dmu_object_free(zsb->z_os, aoid, tx);
1374 aoid = dmu_object_alloc(zsb->z_os,
1375 otype, aclp->z_acl_bytes,
1376 otype == DMU_OT_ACL ?
1377 DMU_OT_SYSACL : DMU_OT_NONE,
1378 otype == DMU_OT_ACL ?
1379 DN_MAX_BONUSLEN : 0, tx);
1381 (void) dmu_object_set_blocksize(zsb->z_os,
1382 aoid, aclp->z_acl_bytes, 0, tx);
1384 acl_phys.z_acl_extern_obj = aoid;
1385 for (aclnode = list_head(&aclp->z_acl); aclnode;
1386 aclnode = list_next(&aclp->z_acl, aclnode)) {
1387 if (aclnode->z_ace_count == 0)
1389 dmu_write(zsb->z_os, aoid, off,
1390 aclnode->z_size, aclnode->z_acldata, tx);
1391 off += aclnode->z_size;
1394 void *start = acl_phys.z_ace_data;
1396 * Migrating back embedded?
1398 if (acl_phys.z_acl_extern_obj) {
1399 error = dmu_object_free(zsb->z_os,
1400 acl_phys.z_acl_extern_obj, tx);
1403 acl_phys.z_acl_extern_obj = 0;
1406 for (aclnode = list_head(&aclp->z_acl); aclnode;
1407 aclnode = list_next(&aclp->z_acl, aclnode)) {
1408 if (aclnode->z_ace_count == 0)
1410 bcopy(aclnode->z_acldata, start,
1412 start = (caddr_t)start + aclnode->z_size;
1416 * If Old version then swap count/bytes to match old
1417 * layout of znode_acl_phys_t.
1419 if (aclp->z_version == ZFS_ACL_VERSION_INITIAL) {
1420 acl_phys.z_acl_size = aclp->z_acl_count;
1421 acl_phys.z_acl_count = aclp->z_acl_bytes;
1423 acl_phys.z_acl_size = aclp->z_acl_bytes;
1424 acl_phys.z_acl_count = aclp->z_acl_count;
1426 acl_phys.z_acl_version = aclp->z_version;
1428 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ZNODE_ACL(zsb), NULL,
1429 &acl_phys, sizeof (acl_phys));
1433 * Replace ACL wide bits, but first clear them.
1435 zp->z_pflags &= ~ZFS_ACL_WIDE_FLAGS;
1437 zp->z_pflags |= aclp->z_hints;
1439 if (ace_trivial_common(aclp, 0, zfs_ace_walk) == 0)
1440 zp->z_pflags |= ZFS_ACL_TRIVIAL;
1442 zfs_tstamp_update_setup(zp, STATE_CHANGED, NULL, ctime, B_TRUE);
1443 return (sa_bulk_update(zp->z_sa_hdl, bulk, count, tx));
1447 zfs_acl_chmod(zfs_sb_t *zsb, uint64_t mode, zfs_acl_t *aclp)
1451 int new_count, new_bytes;
1454 uint16_t iflags, type;
1455 uint32_t access_mask;
1456 zfs_acl_node_t *newnode;
1457 size_t abstract_size = aclp->z_ops->ace_abstract_size();
1459 uint32_t owner, group, everyone;
1460 uint32_t deny1, deny2, allow0;
1462 new_count = new_bytes = 0;
1464 acl_trivial_access_masks((mode_t)mode, &allow0, &deny1, &deny2,
1465 &owner, &group, &everyone);
1467 newnode = zfs_acl_node_alloc((abstract_size * 6) + aclp->z_acl_bytes);
1469 zacep = newnode->z_acldata;
1471 zfs_set_ace(aclp, zacep, allow0, ALLOW, -1, ACE_OWNER);
1472 zacep = (void *)((uintptr_t)zacep + abstract_size);
1474 new_bytes += abstract_size;
1476 zfs_set_ace(aclp, zacep, deny1, DENY, -1, ACE_OWNER);
1477 zacep = (void *)((uintptr_t)zacep + abstract_size);
1479 new_bytes += abstract_size;
1482 zfs_set_ace(aclp, zacep, deny2, DENY, -1, OWNING_GROUP);
1483 zacep = (void *)((uintptr_t)zacep + abstract_size);
1485 new_bytes += abstract_size;
1488 while ((acep = zfs_acl_next_ace(aclp, acep, &who, &access_mask,
1490 uint16_t inherit_flags;
1492 entry_type = (iflags & ACE_TYPE_FLAGS);
1493 inherit_flags = (iflags & ALL_INHERIT);
1495 if ((entry_type == ACE_OWNER || entry_type == ACE_EVERYONE ||
1496 (entry_type == OWNING_GROUP)) &&
1497 ((inherit_flags & ACE_INHERIT_ONLY_ACE) == 0)) {
1501 if ((type != ALLOW && type != DENY) ||
1502 (inherit_flags & ACE_INHERIT_ONLY_ACE)) {
1504 aclp->z_hints |= ZFS_INHERIT_ACE;
1506 case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE:
1507 case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE:
1508 case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE:
1509 case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE:
1510 aclp->z_hints |= ZFS_ACL_OBJ_ACE;
1516 * Limit permissions to be no greater than
1519 if (zsb->z_acl_inherit == ZFS_ACL_RESTRICTED) {
1520 if (!(mode & S_IRGRP))
1521 access_mask &= ~ACE_READ_DATA;
1522 if (!(mode & S_IWGRP))
1524 ~(ACE_WRITE_DATA|ACE_APPEND_DATA);
1525 if (!(mode & S_IXGRP))
1526 access_mask &= ~ACE_EXECUTE;
1528 ~(ACE_WRITE_OWNER|ACE_WRITE_ACL|
1529 ACE_WRITE_ATTRIBUTES|ACE_WRITE_NAMED_ATTRS);
1532 zfs_set_ace(aclp, zacep, access_mask, type, who, iflags);
1533 ace_size = aclp->z_ops->ace_size(acep);
1534 zacep = (void *)((uintptr_t)zacep + ace_size);
1536 new_bytes += ace_size;
1538 zfs_set_ace(aclp, zacep, owner, 0, -1, ACE_OWNER);
1539 zacep = (void *)((uintptr_t)zacep + abstract_size);
1540 zfs_set_ace(aclp, zacep, group, 0, -1, OWNING_GROUP);
1541 zacep = (void *)((uintptr_t)zacep + abstract_size);
1542 zfs_set_ace(aclp, zacep, everyone, 0, -1, ACE_EVERYONE);
1545 new_bytes += abstract_size * 3;
1546 zfs_acl_release_nodes(aclp);
1547 aclp->z_acl_count = new_count;
1548 aclp->z_acl_bytes = new_bytes;
1549 newnode->z_ace_count = new_count;
1550 newnode->z_size = new_bytes;
1551 list_insert_tail(&aclp->z_acl, newnode);
1555 zfs_acl_chmod_setattr(znode_t *zp, zfs_acl_t **aclp, uint64_t mode)
1557 mutex_enter(&zp->z_acl_lock);
1558 mutex_enter(&zp->z_lock);
1559 *aclp = zfs_acl_alloc(zfs_acl_version_zp(zp));
1560 (*aclp)->z_hints = zp->z_pflags & V4_ACL_WIDE_FLAGS;
1561 zfs_acl_chmod(ZTOZSB(zp), mode, *aclp);
1562 mutex_exit(&zp->z_lock);
1563 mutex_exit(&zp->z_acl_lock);
1568 * strip off write_owner and write_acl
1571 zfs_restricted_update(zfs_sb_t *zsb, zfs_acl_t *aclp, void *acep)
1573 uint32_t mask = aclp->z_ops->ace_mask_get(acep);
1575 if ((zsb->z_acl_inherit == ZFS_ACL_RESTRICTED) &&
1576 (aclp->z_ops->ace_type_get(acep) == ALLOW)) {
1577 mask &= ~RESTRICTED_CLEAR;
1578 aclp->z_ops->ace_mask_set(acep, mask);
1583 * Should ACE be inherited?
1586 zfs_ace_can_use(umode_t obj_mode, uint16_t acep_flags)
1588 int iflags = (acep_flags & 0xf);
1590 if (S_ISDIR(obj_mode) && (iflags & ACE_DIRECTORY_INHERIT_ACE))
1592 else if (iflags & ACE_FILE_INHERIT_ACE)
1593 return (!(S_ISDIR(obj_mode) &&
1594 (iflags & ACE_NO_PROPAGATE_INHERIT_ACE)));
1599 * inherit inheritable ACEs from parent
1602 zfs_acl_inherit(zfs_sb_t *zsb, umode_t obj_mode, zfs_acl_t *paclp,
1603 uint64_t mode, boolean_t *need_chmod)
1607 zfs_acl_node_t *aclnode;
1608 zfs_acl_t *aclp = NULL;
1610 uint32_t access_mask;
1611 uint16_t iflags, newflags, type;
1613 void *data1, *data2;
1614 size_t data1sz, data2sz;
1615 boolean_t vdir = S_ISDIR(obj_mode);
1616 boolean_t vreg = S_ISREG(obj_mode);
1617 boolean_t passthrough, passthrough_x, noallow;
1620 zsb->z_acl_inherit == ZFS_ACL_PASSTHROUGH_X;
1621 passthrough = passthrough_x ||
1622 zsb->z_acl_inherit == ZFS_ACL_PASSTHROUGH;
1624 zsb->z_acl_inherit == ZFS_ACL_NOALLOW;
1626 *need_chmod = B_TRUE;
1628 aclp = zfs_acl_alloc(paclp->z_version);
1629 if (zsb->z_acl_inherit == ZFS_ACL_DISCARD || S_ISLNK(obj_mode))
1631 while ((pacep = zfs_acl_next_ace(paclp, pacep, &who,
1632 &access_mask, &iflags, &type))) {
1635 * don't inherit bogus ACEs
1637 if (!zfs_acl_valid_ace_type(type, iflags))
1640 if (noallow && type == ALLOW)
1643 ace_size = aclp->z_ops->ace_size(pacep);
1645 if (!zfs_ace_can_use(obj_mode, iflags))
1649 * If owner@, group@, or everyone@ inheritable
1650 * then zfs_acl_chmod() isn't needed.
1653 ((iflags & (ACE_OWNER|ACE_EVERYONE)) ||
1654 ((iflags & OWNING_GROUP) ==
1655 OWNING_GROUP)) && (vreg || (vdir && (iflags &
1656 ACE_DIRECTORY_INHERIT_ACE)))) {
1657 *need_chmod = B_FALSE;
1660 if (!vdir && passthrough_x &&
1661 ((mode & (S_IXUSR | S_IXGRP | S_IXOTH)) == 0)) {
1662 access_mask &= ~ACE_EXECUTE;
1665 aclnode = zfs_acl_node_alloc(ace_size);
1666 list_insert_tail(&aclp->z_acl, aclnode);
1667 acep = aclnode->z_acldata;
1669 zfs_set_ace(aclp, acep, access_mask, type,
1670 who, iflags|ACE_INHERITED_ACE);
1673 * Copy special opaque data if any
1675 if ((data1sz = paclp->z_ops->ace_data(pacep, &data1)) != 0) {
1676 VERIFY((data2sz = aclp->z_ops->ace_data(acep,
1677 &data2)) == data1sz);
1678 bcopy(data1, data2, data2sz);
1681 aclp->z_acl_count++;
1682 aclnode->z_ace_count++;
1683 aclp->z_acl_bytes += aclnode->z_size;
1684 newflags = aclp->z_ops->ace_flags_get(acep);
1687 aclp->z_hints |= ZFS_INHERIT_ACE;
1689 if ((iflags & ACE_NO_PROPAGATE_INHERIT_ACE) || !vdir) {
1690 newflags &= ~ALL_INHERIT;
1691 aclp->z_ops->ace_flags_set(acep,
1692 newflags|ACE_INHERITED_ACE);
1693 zfs_restricted_update(zsb, aclp, acep);
1700 * If only FILE_INHERIT is set then turn on
1703 if ((iflags & (ACE_FILE_INHERIT_ACE |
1704 ACE_DIRECTORY_INHERIT_ACE)) == ACE_FILE_INHERIT_ACE) {
1705 newflags |= ACE_INHERIT_ONLY_ACE;
1706 aclp->z_ops->ace_flags_set(acep,
1707 newflags|ACE_INHERITED_ACE);
1709 newflags &= ~ACE_INHERIT_ONLY_ACE;
1710 aclp->z_ops->ace_flags_set(acep,
1711 newflags|ACE_INHERITED_ACE);
1718 * Create file system object initial permissions
1719 * including inheritable ACEs.
1722 zfs_acl_ids_create(znode_t *dzp, int flag, vattr_t *vap, cred_t *cr,
1723 vsecattr_t *vsecp, zfs_acl_ids_t *acl_ids)
1726 zfs_sb_t *zsb = ZTOZSB(dzp);
1730 #endif /* HAVE_KSID */
1731 boolean_t need_chmod = B_TRUE;
1732 boolean_t inherited = B_FALSE;
1734 bzero(acl_ids, sizeof (zfs_acl_ids_t));
1735 acl_ids->z_mode = vap->va_mode;
1738 if ((error = zfs_vsec_2_aclp(zsb, vap->va_mode, vsecp,
1739 cr, &acl_ids->z_fuidp, &acl_ids->z_aclp)) != 0)
1742 acl_ids->z_fuid = vap->va_uid;
1743 acl_ids->z_fgid = vap->va_gid;
1746 * Determine uid and gid.
1748 if ((flag & IS_ROOT_NODE) || zsb->z_replay ||
1749 ((flag & IS_XATTR) && (S_ISDIR(vap->va_mode)))) {
1750 acl_ids->z_fuid = zfs_fuid_create(zsb, (uint64_t)vap->va_uid,
1751 cr, ZFS_OWNER, &acl_ids->z_fuidp);
1752 acl_ids->z_fgid = zfs_fuid_create(zsb, (uint64_t)vap->va_gid,
1753 cr, ZFS_GROUP, &acl_ids->z_fuidp);
1756 acl_ids->z_fuid = zfs_fuid_create_cred(zsb, ZFS_OWNER,
1757 cr, &acl_ids->z_fuidp);
1758 acl_ids->z_fgid = 0;
1759 if (vap->va_mask & AT_GID) {
1760 acl_ids->z_fgid = zfs_fuid_create(zsb,
1761 (uint64_t)vap->va_gid,
1762 cr, ZFS_GROUP, &acl_ids->z_fuidp);
1764 if (acl_ids->z_fgid != dzp->z_gid &&
1765 !groupmember(vap->va_gid, cr) &&
1766 secpolicy_vnode_create_gid(cr) != 0)
1767 acl_ids->z_fgid = 0;
1769 if (acl_ids->z_fgid == 0) {
1770 if (dzp->z_mode & S_ISGID) {
1774 acl_ids->z_fgid = dzp->z_gid;
1775 gid = zfs_fuid_map_id(zsb, acl_ids->z_fgid,
1778 if (zsb->z_use_fuids &&
1779 IS_EPHEMERAL(acl_ids->z_fgid)) {
1780 domain = zfs_fuid_idx_domain(
1782 FUID_INDEX(acl_ids->z_fgid));
1783 rid = FUID_RID(acl_ids->z_fgid);
1784 zfs_fuid_node_add(&acl_ids->z_fuidp,
1786 FUID_INDEX(acl_ids->z_fgid),
1787 acl_ids->z_fgid, ZFS_GROUP);
1790 acl_ids->z_fgid = zfs_fuid_create_cred(zsb,
1791 ZFS_GROUP, cr, &acl_ids->z_fuidp);
1796 #endif /* HAVE_KSID */
1799 * If we're creating a directory, and the parent directory has the
1800 * set-GID bit set, set in on the new directory.
1801 * Otherwise, if the user is neither privileged nor a member of the
1802 * file's new group, clear the file's set-GID bit.
1805 if (!(flag & IS_ROOT_NODE) && (dzp->z_mode & S_ISGID) &&
1806 (S_ISDIR(vap->va_mode))) {
1807 acl_ids->z_mode |= S_ISGID;
1809 if ((acl_ids->z_mode & S_ISGID) &&
1810 secpolicy_vnode_setids_setgids(cr, gid) != 0)
1811 acl_ids->z_mode &= ~S_ISGID;
1814 if (acl_ids->z_aclp == NULL) {
1815 mutex_enter(&dzp->z_acl_lock);
1816 mutex_enter(&dzp->z_lock);
1817 if (!(flag & IS_ROOT_NODE) && (S_ISDIR(ZTOI(dzp)->i_mode) &&
1818 (dzp->z_pflags & ZFS_INHERIT_ACE)) &&
1819 !(dzp->z_pflags & ZFS_XATTR)) {
1820 VERIFY(0 == zfs_acl_node_read(dzp, B_TRUE,
1822 acl_ids->z_aclp = zfs_acl_inherit(zsb,
1823 vap->va_mode, paclp, acl_ids->z_mode, &need_chmod);
1827 zfs_acl_alloc(zfs_acl_version_zp(dzp));
1828 acl_ids->z_aclp->z_hints |= ZFS_ACL_TRIVIAL;
1830 mutex_exit(&dzp->z_lock);
1831 mutex_exit(&dzp->z_acl_lock);
1833 acl_ids->z_aclp->z_hints |= S_ISDIR(vap->va_mode) ?
1834 ZFS_ACL_AUTO_INHERIT : 0;
1835 zfs_acl_chmod(zsb, acl_ids->z_mode, acl_ids->z_aclp);
1839 if (inherited || vsecp) {
1840 acl_ids->z_mode = zfs_mode_compute(acl_ids->z_mode,
1841 acl_ids->z_aclp, &acl_ids->z_aclp->z_hints,
1842 acl_ids->z_fuid, acl_ids->z_fgid);
1843 if (ace_trivial_common(acl_ids->z_aclp, 0, zfs_ace_walk) == 0)
1844 acl_ids->z_aclp->z_hints |= ZFS_ACL_TRIVIAL;
1851 * Free ACL and fuid_infop, but not the acl_ids structure
1854 zfs_acl_ids_free(zfs_acl_ids_t *acl_ids)
1856 if (acl_ids->z_aclp)
1857 zfs_acl_free(acl_ids->z_aclp);
1858 if (acl_ids->z_fuidp)
1859 zfs_fuid_info_free(acl_ids->z_fuidp);
1860 acl_ids->z_aclp = NULL;
1861 acl_ids->z_fuidp = NULL;
1865 zfs_acl_ids_overquota(zfs_sb_t *zsb, zfs_acl_ids_t *acl_ids)
1867 return (zfs_fuid_overquota(zsb, B_FALSE, acl_ids->z_fuid) ||
1868 zfs_fuid_overquota(zsb, B_TRUE, acl_ids->z_fgid));
1872 * Retrieve a files ACL
1875 zfs_getacl(znode_t *zp, vsecattr_t *vsecp, boolean_t skipaclchk, cred_t *cr)
1883 mask = vsecp->vsa_mask & (VSA_ACE | VSA_ACECNT |
1884 VSA_ACE_ACLFLAGS | VSA_ACE_ALLTYPES);
1889 if ((error = zfs_zaccess(zp, ACE_READ_ACL, 0, skipaclchk, cr)))
1892 mutex_enter(&zp->z_acl_lock);
1894 error = zfs_acl_node_read(zp, B_FALSE, &aclp, B_FALSE);
1896 mutex_exit(&zp->z_acl_lock);
1901 * Scan ACL to determine number of ACEs
1903 if ((zp->z_pflags & ZFS_ACL_OBJ_ACE) && !(mask & VSA_ACE_ALLTYPES)) {
1906 uint32_t access_mask;
1907 uint16_t type, iflags;
1909 while ((zacep = zfs_acl_next_ace(aclp, zacep,
1910 &who, &access_mask, &iflags, &type))) {
1912 case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE:
1913 case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE:
1914 case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE:
1915 case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE:
1922 vsecp->vsa_aclcnt = count;
1924 count = (int)aclp->z_acl_count;
1926 if (mask & VSA_ACECNT) {
1927 vsecp->vsa_aclcnt = count;
1930 if (mask & VSA_ACE) {
1933 aclsz = count * sizeof (ace_t) +
1934 sizeof (ace_object_t) * largeace;
1936 vsecp->vsa_aclentp = kmem_alloc(aclsz, KM_SLEEP);
1937 vsecp->vsa_aclentsz = aclsz;
1939 if (aclp->z_version == ZFS_ACL_VERSION_FUID)
1940 zfs_copy_fuid_2_ace(ZTOZSB(zp), aclp, cr,
1941 vsecp->vsa_aclentp, !(mask & VSA_ACE_ALLTYPES));
1943 zfs_acl_node_t *aclnode;
1944 void *start = vsecp->vsa_aclentp;
1946 for (aclnode = list_head(&aclp->z_acl); aclnode;
1947 aclnode = list_next(&aclp->z_acl, aclnode)) {
1948 bcopy(aclnode->z_acldata, start,
1950 start = (caddr_t)start + aclnode->z_size;
1952 ASSERT((caddr_t)start - (caddr_t)vsecp->vsa_aclentp ==
1956 if (mask & VSA_ACE_ACLFLAGS) {
1957 vsecp->vsa_aclflags = 0;
1958 if (zp->z_pflags & ZFS_ACL_DEFAULTED)
1959 vsecp->vsa_aclflags |= ACL_DEFAULTED;
1960 if (zp->z_pflags & ZFS_ACL_PROTECTED)
1961 vsecp->vsa_aclflags |= ACL_PROTECTED;
1962 if (zp->z_pflags & ZFS_ACL_AUTO_INHERIT)
1963 vsecp->vsa_aclflags |= ACL_AUTO_INHERIT;
1966 mutex_exit(&zp->z_acl_lock);
1972 zfs_vsec_2_aclp(zfs_sb_t *zsb, umode_t obj_mode,
1973 vsecattr_t *vsecp, cred_t *cr, zfs_fuid_info_t **fuidp, zfs_acl_t **zaclp)
1976 zfs_acl_node_t *aclnode;
1977 int aclcnt = vsecp->vsa_aclcnt;
1980 if (vsecp->vsa_aclcnt > MAX_ACL_ENTRIES || vsecp->vsa_aclcnt <= 0)
1983 aclp = zfs_acl_alloc(zfs_acl_version(zsb->z_version));
1986 aclnode = zfs_acl_node_alloc(aclcnt * sizeof (zfs_object_ace_t));
1987 if (aclp->z_version == ZFS_ACL_VERSION_INITIAL) {
1988 if ((error = zfs_copy_ace_2_oldace(obj_mode, aclp,
1989 (ace_t *)vsecp->vsa_aclentp, aclnode->z_acldata,
1990 aclcnt, &aclnode->z_size)) != 0) {
1992 zfs_acl_node_free(aclnode);
1996 if ((error = zfs_copy_ace_2_fuid(zsb, obj_mode, aclp,
1997 vsecp->vsa_aclentp, aclnode->z_acldata, aclcnt,
1998 &aclnode->z_size, fuidp, cr)) != 0) {
2000 zfs_acl_node_free(aclnode);
2004 aclp->z_acl_bytes = aclnode->z_size;
2005 aclnode->z_ace_count = aclcnt;
2006 aclp->z_acl_count = aclcnt;
2007 list_insert_head(&aclp->z_acl, aclnode);
2010 * If flags are being set then add them to z_hints
2012 if (vsecp->vsa_mask & VSA_ACE_ACLFLAGS) {
2013 if (vsecp->vsa_aclflags & ACL_PROTECTED)
2014 aclp->z_hints |= ZFS_ACL_PROTECTED;
2015 if (vsecp->vsa_aclflags & ACL_DEFAULTED)
2016 aclp->z_hints |= ZFS_ACL_DEFAULTED;
2017 if (vsecp->vsa_aclflags & ACL_AUTO_INHERIT)
2018 aclp->z_hints |= ZFS_ACL_AUTO_INHERIT;
2030 zfs_setacl(znode_t *zp, vsecattr_t *vsecp, boolean_t skipaclchk, cred_t *cr)
2032 zfs_sb_t *zsb = ZTOZSB(zp);
2033 zilog_t *zilog = zsb->z_log;
2034 ulong_t mask = vsecp->vsa_mask & (VSA_ACE | VSA_ACECNT);
2038 zfs_fuid_info_t *fuidp = NULL;
2039 boolean_t fuid_dirtied;
2045 if (zp->z_pflags & ZFS_IMMUTABLE)
2048 if ((error = zfs_zaccess(zp, ACE_WRITE_ACL, 0, skipaclchk, cr)))
2051 error = zfs_vsec_2_aclp(zsb, ZTOI(zp)->i_mode, vsecp, cr, &fuidp,
2057 * If ACL wide flags aren't being set then preserve any
2060 if (!(vsecp->vsa_mask & VSA_ACE_ACLFLAGS)) {
2062 (zp->z_pflags & V4_ACL_WIDE_FLAGS);
2065 mutex_enter(&zp->z_acl_lock);
2066 mutex_enter(&zp->z_lock);
2068 tx = dmu_tx_create(zsb->z_os);
2070 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE);
2072 fuid_dirtied = zsb->z_fuid_dirty;
2074 zfs_fuid_txhold(zsb, tx);
2077 * If old version and ACL won't fit in bonus and we aren't
2078 * upgrading then take out necessary DMU holds
2081 if ((acl_obj = zfs_external_acl(zp)) != 0) {
2082 if (zsb->z_version >= ZPL_VERSION_FUID &&
2083 zfs_znode_acl_version(zp) <= ZFS_ACL_VERSION_INITIAL) {
2084 dmu_tx_hold_free(tx, acl_obj, 0,
2086 dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0,
2089 dmu_tx_hold_write(tx, acl_obj, 0, aclp->z_acl_bytes);
2091 } else if (!zp->z_is_sa && aclp->z_acl_bytes > ZFS_ACE_SPACE) {
2092 dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, aclp->z_acl_bytes);
2095 zfs_sa_upgrade_txholds(tx, zp);
2096 error = dmu_tx_assign(tx, TXG_NOWAIT);
2098 mutex_exit(&zp->z_acl_lock);
2099 mutex_exit(&zp->z_lock);
2101 if (error == ERESTART) {
2111 error = zfs_aclset_common(zp, aclp, cr, tx);
2113 ASSERT(zp->z_acl_cached == NULL);
2114 zp->z_acl_cached = aclp;
2117 zfs_fuid_sync(zsb, tx);
2119 zfs_log_acl(zilog, tx, zp, vsecp, fuidp);
2122 zfs_fuid_info_free(fuidp);
2125 mutex_exit(&zp->z_lock);
2126 mutex_exit(&zp->z_acl_lock);
2132 * Check accesses of interest (AoI) against attributes of the dataset
2133 * such as read-only. Returns zero if no AoI conflict with dataset
2134 * attributes, otherwise an appropriate errno is returned.
2137 zfs_zaccess_dataset_check(znode_t *zp, uint32_t v4_mode)
2139 if ((v4_mode & WRITE_MASK) && (zfs_is_readonly(ZTOZSB(zp))) &&
2140 (!S_ISDEV(ZTOI(zp)->i_mode) ||
2141 (S_ISDEV(ZTOI(zp)->i_mode) && (v4_mode & WRITE_MASK_ATTRS)))) {
2146 * Only check for READONLY on non-directories.
2148 if ((v4_mode & WRITE_MASK_DATA) &&
2149 ((!S_ISDIR(ZTOI(zp)->i_mode) &&
2150 (zp->z_pflags & (ZFS_READONLY | ZFS_IMMUTABLE))) ||
2151 (S_ISDIR(ZTOI(zp)->i_mode) &&
2152 (zp->z_pflags & ZFS_IMMUTABLE)))) {
2156 if ((v4_mode & (ACE_DELETE | ACE_DELETE_CHILD)) &&
2157 (zp->z_pflags & ZFS_NOUNLINK)) {
2161 if (((v4_mode & (ACE_READ_DATA|ACE_EXECUTE)) &&
2162 (zp->z_pflags & ZFS_AV_QUARANTINED))) {
2170 * The primary usage of this function is to loop through all of the
2171 * ACEs in the znode, determining what accesses of interest (AoI) to
2172 * the caller are allowed or denied. The AoI are expressed as bits in
2173 * the working_mode parameter. As each ACE is processed, bits covered
2174 * by that ACE are removed from the working_mode. This removal
2175 * facilitates two things. The first is that when the working mode is
2176 * empty (= 0), we know we've looked at all the AoI. The second is
2177 * that the ACE interpretation rules don't allow a later ACE to undo
2178 * something granted or denied by an earlier ACE. Removing the
2179 * discovered access or denial enforces this rule. At the end of
2180 * processing the ACEs, all AoI that were found to be denied are
2181 * placed into the working_mode, giving the caller a mask of denied
2182 * accesses. Returns:
2183 * 0 if all AoI granted
2184 * EACCESS if the denied mask is non-zero
2185 * other error if abnormal failure (e.g., IO error)
2187 * A secondary usage of the function is to determine if any of the
2188 * AoI are granted. If an ACE grants any access in
2189 * the working_mode, we immediately short circuit out of the function.
2190 * This mode is chosen by setting anyaccess to B_TRUE. The
2191 * working_mode is not a denied access mask upon exit if the function
2192 * is used in this manner.
2195 zfs_zaccess_aces_check(znode_t *zp, uint32_t *working_mode,
2196 boolean_t anyaccess, cred_t *cr)
2198 zfs_sb_t *zsb = ZTOZSB(zp);
2201 uid_t uid = crgetuid(cr);
2203 uint16_t type, iflags;
2204 uint16_t entry_type;
2205 uint32_t access_mask;
2206 uint32_t deny_mask = 0;
2207 zfs_ace_hdr_t *acep = NULL;
2212 zfs_fuid_map_ids(zp, cr, &fowner, &gowner);
2214 mutex_enter(&zp->z_acl_lock);
2216 error = zfs_acl_node_read(zp, B_FALSE, &aclp, B_FALSE);
2218 mutex_exit(&zp->z_acl_lock);
2222 ASSERT(zp->z_acl_cached);
2224 while ((acep = zfs_acl_next_ace(aclp, acep, &who, &access_mask,
2226 uint32_t mask_matched;
2228 if (!zfs_acl_valid_ace_type(type, iflags))
2231 if (S_ISDIR(ZTOI(zp)->i_mode) &&
2232 (iflags & ACE_INHERIT_ONLY_ACE))
2235 /* Skip ACE if it does not affect any AoI */
2236 mask_matched = (access_mask & *working_mode);
2240 entry_type = (iflags & ACE_TYPE_FLAGS);
2244 switch (entry_type) {
2252 case ACE_IDENTIFIER_GROUP:
2253 checkit = zfs_groupmember(zsb, who, cr);
2261 if (entry_type == 0) {
2264 newid = zfs_fuid_map_id(zsb, who, cr,
2266 if (newid != IDMAP_WK_CREATOR_OWNER_UID &&
2271 mutex_exit(&zp->z_acl_lock);
2278 DTRACE_PROBE3(zfs__ace__denies,
2280 zfs_ace_hdr_t *, acep,
2281 uint32_t, mask_matched);
2282 deny_mask |= mask_matched;
2284 DTRACE_PROBE3(zfs__ace__allows,
2286 zfs_ace_hdr_t *, acep,
2287 uint32_t, mask_matched);
2289 mutex_exit(&zp->z_acl_lock);
2293 *working_mode &= ~mask_matched;
2297 if (*working_mode == 0)
2301 mutex_exit(&zp->z_acl_lock);
2303 /* Put the found 'denies' back on the working mode */
2305 *working_mode |= deny_mask;
2307 } else if (*working_mode) {
2315 * Return true if any access whatsoever granted, we don't actually
2316 * care what access is granted.
2319 zfs_has_access(znode_t *zp, cred_t *cr)
2321 uint32_t have = ACE_ALL_PERMS;
2323 if (zfs_zaccess_aces_check(zp, &have, B_TRUE, cr) != 0) {
2326 owner = zfs_fuid_map_id(ZTOZSB(zp), zp->z_uid, cr, ZFS_OWNER);
2327 return (secpolicy_vnode_any_access(cr, ZTOI(zp), owner) == 0);
2333 zfs_zaccess_common(znode_t *zp, uint32_t v4_mode, uint32_t *working_mode,
2334 boolean_t *check_privs, boolean_t skipaclchk, cred_t *cr)
2336 zfs_sb_t *zsb = ZTOZSB(zp);
2339 *working_mode = v4_mode;
2340 *check_privs = B_TRUE;
2343 * Short circuit empty requests
2345 if (v4_mode == 0 || zsb->z_replay) {
2350 if ((err = zfs_zaccess_dataset_check(zp, v4_mode)) != 0) {
2351 *check_privs = B_FALSE;
2356 * The caller requested that the ACL check be skipped. This
2357 * would only happen if the caller checked VOP_ACCESS() with a
2358 * 32 bit ACE mask and already had the appropriate permissions.
2365 return (zfs_zaccess_aces_check(zp, working_mode, B_FALSE, cr));
2369 zfs_zaccess_append(znode_t *zp, uint32_t *working_mode, boolean_t *check_privs,
2372 if (*working_mode != ACE_WRITE_DATA)
2375 return (zfs_zaccess_common(zp, ACE_APPEND_DATA, working_mode,
2376 check_privs, B_FALSE, cr));
2380 zfs_fastaccesschk_execute(znode_t *zdp, cred_t *cr)
2382 boolean_t owner = B_FALSE;
2383 boolean_t groupmbr = B_FALSE;
2385 uid_t uid = crgetuid(cr);
2388 if (zdp->z_pflags & ZFS_AV_QUARANTINED)
2391 is_attr = ((zdp->z_pflags & ZFS_XATTR) &&
2392 (S_ISDIR(ZTOI(zdp)->i_mode)));
2397 mutex_enter(&zdp->z_acl_lock);
2399 if (zdp->z_pflags & ZFS_NO_EXECS_DENIED) {
2400 mutex_exit(&zdp->z_acl_lock);
2404 if (FUID_INDEX(zdp->z_uid) != 0 || FUID_INDEX(zdp->z_gid) != 0) {
2405 mutex_exit(&zdp->z_acl_lock);
2409 if (uid == zdp->z_uid) {
2411 if (zdp->z_mode & S_IXUSR) {
2412 mutex_exit(&zdp->z_acl_lock);
2415 mutex_exit(&zdp->z_acl_lock);
2419 if (groupmember(zdp->z_gid, cr)) {
2421 if (zdp->z_mode & S_IXGRP) {
2422 mutex_exit(&zdp->z_acl_lock);
2425 mutex_exit(&zdp->z_acl_lock);
2429 if (!owner && !groupmbr) {
2430 if (zdp->z_mode & S_IXOTH) {
2431 mutex_exit(&zdp->z_acl_lock);
2436 mutex_exit(&zdp->z_acl_lock);
2439 DTRACE_PROBE(zfs__fastpath__execute__access__miss);
2440 ZFS_ENTER(ZTOZSB(zdp));
2441 error = zfs_zaccess(zdp, ACE_EXECUTE, 0, B_FALSE, cr);
2442 ZFS_EXIT(ZTOZSB(zdp));
2447 * Determine whether Access should be granted/denied.
2448 * The least priv subsytem is always consulted as a basic privilege
2449 * can define any form of access.
2452 zfs_zaccess(znode_t *zp, int mode, int flags, boolean_t skipaclchk, cred_t *cr)
2454 uint32_t working_mode;
2456 boolean_t check_privs;
2457 znode_t *check_zp = zp;
2462 * If attribute then validate against base file
2464 if ((zp->z_pflags & ZFS_XATTR) && S_ISDIR(ZTOI(zp)->i_mode)) {
2467 rw_enter(&zp->z_xattr_lock, RW_READER);
2468 if (zp->z_xattr_parent) {
2469 check_zp = zp->z_xattr_parent;
2470 rw_exit(&zp->z_xattr_lock);
2473 * Verify a lookup yields the same znode.
2475 ASSERT3S(sa_lookup(zp->z_sa_hdl, SA_ZPL_PARENT(
2476 ZTOZSB(zp)), &parent, sizeof (parent)), ==, 0);
2477 ASSERT3U(check_zp->z_id, ==, parent);
2479 rw_exit(&zp->z_xattr_lock);
2481 error = sa_lookup(zp->z_sa_hdl, SA_ZPL_PARENT(
2482 ZTOZSB(zp)), &parent, sizeof (parent));
2487 * Cache the lookup on the parent file znode as
2488 * zp->z_xattr_parent and hold a reference. This
2489 * effectively pins the parent in memory until all
2490 * child xattr znodes have been destroyed and
2491 * release their references in zfs_inode_destroy().
2493 error = zfs_zget(ZTOZSB(zp), parent, &check_zp);
2497 rw_enter(&zp->z_xattr_lock, RW_WRITER);
2498 if (zp->z_xattr_parent == NULL)
2499 zp->z_xattr_parent = check_zp;
2500 rw_exit(&zp->z_xattr_lock);
2504 * fixup mode to map to xattr perms
2507 if (mode & (ACE_WRITE_DATA|ACE_APPEND_DATA)) {
2508 mode &= ~(ACE_WRITE_DATA|ACE_APPEND_DATA);
2509 mode |= ACE_WRITE_NAMED_ATTRS;
2512 if (mode & (ACE_READ_DATA|ACE_EXECUTE)) {
2513 mode &= ~(ACE_READ_DATA|ACE_EXECUTE);
2514 mode |= ACE_READ_NAMED_ATTRS;
2518 owner = zfs_fuid_map_id(ZTOZSB(zp), zp->z_uid, cr, ZFS_OWNER);
2520 * Map the bits required to the standard inode flags
2521 * S_IRUSR|S_IWUSR|S_IXUSR in the needed_bits. Map the bits
2522 * mapped by working_mode (currently missing) in missing_bits.
2523 * Call secpolicy_vnode_access2() with (needed_bits & ~checkmode),
2528 working_mode = mode;
2529 if ((working_mode & (ACE_READ_ACL|ACE_READ_ATTRIBUTES)) &&
2530 owner == crgetuid(cr))
2531 working_mode &= ~(ACE_READ_ACL|ACE_READ_ATTRIBUTES);
2533 if (working_mode & (ACE_READ_DATA|ACE_READ_NAMED_ATTRS|
2534 ACE_READ_ACL|ACE_READ_ATTRIBUTES|ACE_SYNCHRONIZE))
2535 needed_bits |= S_IRUSR;
2536 if (working_mode & (ACE_WRITE_DATA|ACE_WRITE_NAMED_ATTRS|
2537 ACE_APPEND_DATA|ACE_WRITE_ATTRIBUTES|ACE_SYNCHRONIZE))
2538 needed_bits |= S_IWUSR;
2539 if (working_mode & ACE_EXECUTE)
2540 needed_bits |= S_IXUSR;
2542 if ((error = zfs_zaccess_common(check_zp, mode, &working_mode,
2543 &check_privs, skipaclchk, cr)) == 0) {
2544 return (secpolicy_vnode_access2(cr, ZTOI(zp), owner,
2545 needed_bits, needed_bits));
2548 if (error && !check_privs) {
2552 if (error && (flags & V_APPEND)) {
2553 error = zfs_zaccess_append(zp, &working_mode, &check_privs, cr);
2556 if (error && check_privs) {
2557 mode_t checkmode = 0;
2560 * First check for implicit owner permission on
2561 * read_acl/read_attributes
2565 ASSERT(working_mode != 0);
2567 if ((working_mode & (ACE_READ_ACL|ACE_READ_ATTRIBUTES) &&
2568 owner == crgetuid(cr)))
2569 working_mode &= ~(ACE_READ_ACL|ACE_READ_ATTRIBUTES);
2571 if (working_mode & (ACE_READ_DATA|ACE_READ_NAMED_ATTRS|
2572 ACE_READ_ACL|ACE_READ_ATTRIBUTES|ACE_SYNCHRONIZE))
2573 checkmode |= S_IRUSR;
2574 if (working_mode & (ACE_WRITE_DATA|ACE_WRITE_NAMED_ATTRS|
2575 ACE_APPEND_DATA|ACE_WRITE_ATTRIBUTES|ACE_SYNCHRONIZE))
2576 checkmode |= S_IWUSR;
2577 if (working_mode & ACE_EXECUTE)
2578 checkmode |= S_IXUSR;
2580 error = secpolicy_vnode_access2(cr, ZTOI(check_zp), owner,
2581 needed_bits & ~checkmode, needed_bits);
2583 if (error == 0 && (working_mode & ACE_WRITE_OWNER))
2584 error = secpolicy_vnode_chown(cr, owner);
2585 if (error == 0 && (working_mode & ACE_WRITE_ACL))
2586 error = secpolicy_vnode_setdac(cr, owner);
2588 if (error == 0 && (working_mode &
2589 (ACE_DELETE|ACE_DELETE_CHILD)))
2590 error = secpolicy_vnode_remove(cr);
2592 if (error == 0 && (working_mode & ACE_SYNCHRONIZE)) {
2593 error = secpolicy_vnode_chown(cr, owner);
2597 * See if any bits other than those already checked
2598 * for are still present. If so then return EACCES
2600 if (working_mode & ~(ZFS_CHECKED_MASKS)) {
2604 } else if (error == 0) {
2605 error = secpolicy_vnode_access2(cr, ZTOI(zp), owner,
2606 needed_bits, needed_bits);
2613 * Translate traditional unix S_IRUSR/S_IWUSR/S_IXUSR mode into
2614 * native ACL format and call zfs_zaccess()
2617 zfs_zaccess_rwx(znode_t *zp, mode_t mode, int flags, cred_t *cr)
2619 return (zfs_zaccess(zp, zfs_unix_to_v4(mode >> 6), flags, B_FALSE, cr));
2623 * Access function for secpolicy_vnode_setattr
2626 zfs_zaccess_unix(znode_t *zp, mode_t mode, cred_t *cr)
2628 int v4_mode = zfs_unix_to_v4(mode >> 6);
2630 return (zfs_zaccess(zp, v4_mode, 0, B_FALSE, cr));
2634 zfs_delete_final_check(znode_t *zp, znode_t *dzp,
2635 mode_t available_perms, cred_t *cr)
2640 downer = zfs_fuid_map_id(ZTOZSB(dzp), dzp->z_uid, cr, ZFS_OWNER);
2642 error = secpolicy_vnode_access2(cr, ZTOI(dzp),
2643 downer, available_perms, S_IWUSR|S_IXUSR);
2646 error = zfs_sticky_remove_access(dzp, zp, cr);
2652 * Determine whether Access should be granted/deny, without
2653 * consulting least priv subsystem.
2656 * The following chart is the recommended NFSv4 enforcement for
2657 * ability to delete an object.
2659 * -------------------------------------------------------
2660 * | Parent Dir | Target Object Permissions |
2662 * -------------------------------------------------------
2663 * | | ACL Allows | ACL Denies| Delete |
2664 * | | Delete | Delete | unspecified|
2665 * -------------------------------------------------------
2666 * | ACL Allows | Permit | Permit | Permit |
2667 * | DELETE_CHILD | |
2668 * -------------------------------------------------------
2669 * | ACL Denies | Permit | Deny | Deny |
2670 * | DELETE_CHILD | | | |
2671 * -------------------------------------------------------
2672 * | ACL specifies | | | |
2673 * | only allow | Permit | Permit | Permit |
2674 * | write and | | | |
2676 * -------------------------------------------------------
2677 * | ACL denies | | | |
2678 * | write and | Permit | Deny | Deny |
2680 * -------------------------------------------------------
2683 * No search privilege, can't even look up file?
2687 zfs_zaccess_delete(znode_t *dzp, znode_t *zp, cred_t *cr)
2689 uint32_t dzp_working_mode = 0;
2690 uint32_t zp_working_mode = 0;
2691 int dzp_error, zp_error;
2692 mode_t available_perms;
2693 boolean_t dzpcheck_privs = B_TRUE;
2694 boolean_t zpcheck_privs = B_TRUE;
2697 * We want specific DELETE permissions to
2698 * take precedence over WRITE/EXECUTE. We don't
2699 * want an ACL such as this to mess us up.
2700 * user:joe:write_data:deny,user:joe:delete:allow
2702 * However, deny permissions may ultimately be overridden
2703 * by secpolicy_vnode_access().
2705 * We will ask for all of the necessary permissions and then
2706 * look at the working modes from the directory and target object
2707 * to determine what was found.
2710 if (zp->z_pflags & (ZFS_IMMUTABLE | ZFS_NOUNLINK))
2715 * If the directory permissions allow the delete, we are done.
2717 if ((dzp_error = zfs_zaccess_common(dzp, ACE_DELETE_CHILD,
2718 &dzp_working_mode, &dzpcheck_privs, B_FALSE, cr)) == 0)
2722 * If target object has delete permission then we are done
2724 if ((zp_error = zfs_zaccess_common(zp, ACE_DELETE, &zp_working_mode,
2725 &zpcheck_privs, B_FALSE, cr)) == 0)
2728 ASSERT(dzp_error && zp_error);
2730 if (!dzpcheck_privs)
2738 * If directory returns EACCES then delete_child was denied
2739 * due to deny delete_child. In this case send the request through
2740 * secpolicy_vnode_remove(). We don't use zfs_delete_final_check()
2741 * since that *could* allow the delete based on write/execute permission
2742 * and we want delete permissions to override write/execute.
2745 if (dzp_error == EACCES)
2746 return (secpolicy_vnode_remove(cr));
2750 * only need to see if we have write/execute on directory.
2753 dzp_error = zfs_zaccess_common(dzp, ACE_EXECUTE|ACE_WRITE_DATA,
2754 &dzp_working_mode, &dzpcheck_privs, B_FALSE, cr);
2756 if (dzp_error != 0 && !dzpcheck_privs)
2763 available_perms = (dzp_working_mode & ACE_WRITE_DATA) ? 0 : S_IWUSR;
2764 available_perms |= (dzp_working_mode & ACE_EXECUTE) ? 0 : S_IXUSR;
2766 return (zfs_delete_final_check(zp, dzp, available_perms, cr));
2771 zfs_zaccess_rename(znode_t *sdzp, znode_t *szp, znode_t *tdzp,
2772 znode_t *tzp, cred_t *cr)
2777 if (szp->z_pflags & ZFS_AV_QUARANTINED)
2780 add_perm = S_ISDIR(ZTOI(szp)->i_mode) ?
2781 ACE_ADD_SUBDIRECTORY : ACE_ADD_FILE;
2784 * Rename permissions are combination of delete permission +
2785 * add file/subdir permission.
2789 * first make sure we do the delete portion.
2791 * If that succeeds then check for add_file/add_subdir permissions
2794 if ((error = zfs_zaccess_delete(sdzp, szp, cr)))
2798 * If we have a tzp, see if we can delete it?
2801 if ((error = zfs_zaccess_delete(tdzp, tzp, cr)))
2806 * Now check for add permissions
2808 error = zfs_zaccess(tdzp, add_perm, 0, B_FALSE, cr);