4 * The contents of this file are subject to the terms of the
5 * Common Development and Distribution License (the "License").
6 * You may not use this file except in compliance with the License.
8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9 * or http://www.opensolaris.org/os/licensing.
10 * See the License for the specific language governing permissions
11 * and limitations under the License.
13 * When distributing Covered Code, include this CDDL HEADER in each
14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15 * If applicable, add the following below this CDDL HEADER, with the
16 * fields enclosed by brackets "[]" replaced with your own identifying
17 * information: Portions Copyright [yyyy] [name of copyright owner]
22 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
23 * Copyright (c) 2013 by Delphix. All rights reserved.
27 #include <sys/types.h>
28 #include <sys/param.h>
30 #include <sys/systm.h>
31 #include <sys/sysmacros.h>
32 #include <sys/resource.h>
34 #include <sys/vnode.h>
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(ZTOZSB(zp
)),
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 zfs_sb_t
*zsb
= ZTOZSB(zp
);
377 ASSERT(MUTEX_HELD(&zp
->z_acl_lock
));
379 if ((error
= sa_size(zp
->z_sa_hdl
, SA_ZPL_DACL_ACES(zsb
),
383 if ((error
= sa_lookup(zp
->z_sa_hdl
, SA_ZPL_DACL_COUNT(zsb
),
384 &acl_count
, sizeof (acl_count
))) != 0)
386 *aclcount
= acl_count
;
388 if ((error
= sa_lookup(zp
->z_sa_hdl
, SA_ZPL_ZNODE_ACL(zsb
),
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(ZTOZSB(zp
)),
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(ZTOZSB(zp
)->z_version
));
452 zfs_acl_alloc(int vers
)
456 aclp
= kmem_zalloc(sizeof (zfs_acl_t
), KM_PUSHPAGE
);
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_PUSHPAGE
);
474 aclnode
->z_acldata
= kmem_alloc(bytes
, KM_PUSHPAGE
);
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(umode_t obj_mode
, 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 (S_ISDIR(obj_mode
) &&
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(zfs_sb_t
*zsb
, umode_t obj_mode
, 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(zsb
, 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_mode
, aclp
, aceptr
->z_hdr
.z_type
,
679 aceptr
->z_hdr
.z_flags
) != B_TRUE
)
680 return (SET_ERROR(EINVAL
));
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(zfs_sb_t
*zsb
, 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(zsb
, 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(umode_t obj_mode
, 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_mode
, aclp
, aceptr
->z_type
,
786 aceptr
->z_flags
) != B_TRUE
)
787 return (SET_ERROR(EINVAL
));
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(ZTOZSB(zp
), ZTOI(zp
)->i_mode
,
830 aclp
, 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(ZTOZSB(zp
)->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(ZTOZSB(zp
)),
1113 aclnode
->z_acldata
, aclnode
->z_size
);
1118 zfs_acl_node_free(aclnode
);
1119 /* convert checksum errors into IO errors */
1120 if (error
== ECKSUM
)
1121 error
= SET_ERROR(EIO
);
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 if (ZTOZSB(zp
)->z_acl_type
== ZFS_ACLTYPE_POSIXACL
)
1162 ASSERT(MUTEX_HELD(&zp
->z_lock
));
1163 ASSERT(MUTEX_HELD(&zp
->z_acl_lock
));
1165 if ((error
= zfs_acl_node_read(zp
, B_TRUE
, &aclp
, B_FALSE
)) == 0)
1166 zp
->z_mode
= zfs_mode_compute(zp
->z_mode
, aclp
,
1167 &zp
->z_pflags
, zp
->z_uid
, zp
->z_gid
);
1173 acl_trivial_access_masks(mode_t mode
, uint32_t *allow0
, uint32_t *deny1
,
1174 uint32_t *deny2
, uint32_t *owner
, uint32_t *group
, uint32_t *everyone
)
1176 *deny1
= *deny2
= *allow0
= *group
= 0;
1178 if (!(mode
& S_IRUSR
) && (mode
& (S_IRGRP
|S_IROTH
)))
1179 *deny1
|= ACE_READ_DATA
;
1180 if (!(mode
& S_IWUSR
) && (mode
& (S_IWGRP
|S_IWOTH
)))
1181 *deny1
|= ACE_WRITE_DATA
;
1182 if (!(mode
& S_IXUSR
) && (mode
& (S_IXGRP
|S_IXOTH
)))
1183 *deny1
|= ACE_EXECUTE
;
1185 if (!(mode
& S_IRGRP
) && (mode
& S_IROTH
))
1186 *deny2
= ACE_READ_DATA
;
1187 if (!(mode
& S_IWGRP
) && (mode
& S_IWOTH
))
1188 *deny2
|= ACE_WRITE_DATA
;
1189 if (!(mode
& S_IXGRP
) && (mode
& S_IXOTH
))
1190 *deny2
|= ACE_EXECUTE
;
1192 if ((mode
& S_IRUSR
) && (!(mode
& S_IRGRP
) && (mode
& S_IROTH
)))
1193 *allow0
|= ACE_READ_DATA
;
1194 if ((mode
& S_IWUSR
) && (!(mode
& S_IWGRP
) && (mode
& S_IWOTH
)))
1195 *allow0
|= ACE_WRITE_DATA
;
1196 if ((mode
& S_IXUSR
) && (!(mode
& S_IXGRP
) && (mode
& S_IXOTH
)))
1197 *allow0
|= ACE_EXECUTE
;
1199 *owner
= ACE_WRITE_ATTRIBUTES
|ACE_WRITE_OWNER
|ACE_WRITE_ACL
|
1200 ACE_WRITE_NAMED_ATTRS
|ACE_READ_ACL
|ACE_READ_ATTRIBUTES
|
1201 ACE_READ_NAMED_ATTRS
|ACE_SYNCHRONIZE
;
1203 *owner
|= ACE_READ_DATA
;
1205 *owner
|= ACE_WRITE_DATA
|ACE_APPEND_DATA
;
1207 *owner
|= ACE_EXECUTE
;
1209 *group
= ACE_READ_ACL
|ACE_READ_ATTRIBUTES
| ACE_READ_NAMED_ATTRS
|
1212 *group
|= ACE_READ_DATA
;
1214 *group
|= ACE_WRITE_DATA
|ACE_APPEND_DATA
;
1216 *group
|= ACE_EXECUTE
;
1218 *everyone
= ACE_READ_ACL
|ACE_READ_ATTRIBUTES
| ACE_READ_NAMED_ATTRS
|
1221 *everyone
|= ACE_READ_DATA
;
1223 *everyone
|= ACE_WRITE_DATA
|ACE_APPEND_DATA
;
1225 *everyone
|= ACE_EXECUTE
;
1230 * determine whether an ace_t acl is trivial
1232 * Trivialness implies that the acl is composed of only
1233 * owner, group, everyone entries. ACL can't
1234 * have read_acl denied, and write_owner/write_acl/write_attributes
1235 * can only be owner@ entry.
1238 ace_trivial_common(void *acep
, int aclcnt
,
1239 uint64_t (*walk
)(void *, uint64_t, int aclcnt
,
1240 uint16_t *, uint16_t *, uint32_t *))
1245 uint64_t cookie
= 0;
1247 while ((cookie
= walk(acep
, cookie
, aclcnt
, &flags
, &type
, &mask
))) {
1248 switch (flags
& ACE_TYPE_FLAGS
) {
1250 case ACE_GROUP
|ACE_IDENTIFIER_GROUP
:
1257 if (flags
& (ACE_FILE_INHERIT_ACE
|
1258 ACE_DIRECTORY_INHERIT_ACE
|ACE_NO_PROPAGATE_INHERIT_ACE
|
1259 ACE_INHERIT_ONLY_ACE
))
1263 * Special check for some special bits
1265 * Don't allow anybody to deny reading basic
1266 * attributes or a files ACL.
1268 if ((mask
& (ACE_READ_ACL
|ACE_READ_ATTRIBUTES
)) &&
1269 (type
== ACE_ACCESS_DENIED_ACE_TYPE
))
1273 * Delete permissions are never set by default
1275 if (mask
& (ACE_DELETE
|ACE_DELETE_CHILD
))
1278 * only allow owner@ to have
1279 * write_acl/write_owner/write_attributes/write_xattr/
1281 if (type
== ACE_ACCESS_ALLOWED_ACE_TYPE
&&
1282 (!(flags
& ACE_OWNER
) && (mask
&
1283 (ACE_WRITE_OWNER
|ACE_WRITE_ACL
| ACE_WRITE_ATTRIBUTES
|
1284 ACE_WRITE_NAMED_ATTRS
))))
1293 * common code for setting ACLs.
1295 * This function is called from zfs_mode_update, zfs_perm_init, and zfs_setacl.
1296 * zfs_setacl passes a non-NULL inherit pointer (ihp) to indicate that it's
1297 * already checked the acl and knows whether to inherit.
1300 zfs_aclset_common(znode_t
*zp
, zfs_acl_t
*aclp
, cred_t
*cr
, dmu_tx_t
*tx
)
1303 zfs_sb_t
*zsb
= ZTOZSB(zp
);
1304 dmu_object_type_t otype
;
1305 zfs_acl_locator_cb_t locate
= { 0 };
1307 sa_bulk_attr_t bulk
[5];
1313 mode
= zfs_mode_compute(mode
, aclp
, &zp
->z_pflags
,
1314 zp
->z_uid
, zp
->z_gid
);
1317 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_MODE(zsb
), NULL
,
1318 &mode
, sizeof (mode
));
1319 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_FLAGS(zsb
), NULL
,
1320 &zp
->z_pflags
, sizeof (zp
->z_pflags
));
1321 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_CTIME(zsb
), NULL
,
1322 &ctime
, sizeof (ctime
));
1324 if (zp
->z_acl_cached
) {
1325 zfs_acl_free(zp
->z_acl_cached
);
1326 zp
->z_acl_cached
= NULL
;
1332 if (!zsb
->z_use_fuids
) {
1333 otype
= DMU_OT_OLDACL
;
1335 if ((aclp
->z_version
== ZFS_ACL_VERSION_INITIAL
) &&
1336 (zsb
->z_version
>= ZPL_VERSION_FUID
))
1337 zfs_acl_xform(zp
, aclp
, cr
);
1338 ASSERT(aclp
->z_version
>= ZFS_ACL_VERSION_FUID
);
1343 * Arrgh, we have to handle old on disk format
1344 * as well as newer (preferred) SA format.
1347 if (zp
->z_is_sa
) { /* the easy case, just update the ACL attribute */
1348 locate
.cb_aclp
= aclp
;
1349 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_DACL_ACES(zsb
),
1350 zfs_acl_data_locator
, &locate
, aclp
->z_acl_bytes
);
1351 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_DACL_COUNT(zsb
),
1352 NULL
, &aclp
->z_acl_count
, sizeof (uint64_t));
1353 } else { /* Painful legacy way */
1354 zfs_acl_node_t
*aclnode
;
1356 zfs_acl_phys_t acl_phys
;
1359 if ((error
= sa_lookup(zp
->z_sa_hdl
, SA_ZPL_ZNODE_ACL(zsb
),
1360 &acl_phys
, sizeof (acl_phys
))) != 0)
1363 aoid
= acl_phys
.z_acl_extern_obj
;
1365 if (aclp
->z_acl_bytes
> ZFS_ACE_SPACE
) {
1367 * If ACL was previously external and we are now
1368 * converting to new ACL format then release old
1369 * ACL object and create a new one.
1372 aclp
->z_version
!= acl_phys
.z_acl_version
) {
1373 error
= dmu_object_free(zsb
->z_os
, aoid
, tx
);
1379 aoid
= dmu_object_alloc(zsb
->z_os
,
1380 otype
, aclp
->z_acl_bytes
,
1381 otype
== DMU_OT_ACL
?
1382 DMU_OT_SYSACL
: DMU_OT_NONE
,
1383 otype
== DMU_OT_ACL
?
1384 DN_MAX_BONUSLEN
: 0, tx
);
1386 (void) dmu_object_set_blocksize(zsb
->z_os
,
1387 aoid
, aclp
->z_acl_bytes
, 0, tx
);
1389 acl_phys
.z_acl_extern_obj
= aoid
;
1390 for (aclnode
= list_head(&aclp
->z_acl
); aclnode
;
1391 aclnode
= list_next(&aclp
->z_acl
, aclnode
)) {
1392 if (aclnode
->z_ace_count
== 0)
1394 dmu_write(zsb
->z_os
, aoid
, off
,
1395 aclnode
->z_size
, aclnode
->z_acldata
, tx
);
1396 off
+= aclnode
->z_size
;
1399 void *start
= acl_phys
.z_ace_data
;
1401 * Migrating back embedded?
1403 if (acl_phys
.z_acl_extern_obj
) {
1404 error
= dmu_object_free(zsb
->z_os
,
1405 acl_phys
.z_acl_extern_obj
, tx
);
1408 acl_phys
.z_acl_extern_obj
= 0;
1411 for (aclnode
= list_head(&aclp
->z_acl
); aclnode
;
1412 aclnode
= list_next(&aclp
->z_acl
, aclnode
)) {
1413 if (aclnode
->z_ace_count
== 0)
1415 bcopy(aclnode
->z_acldata
, start
,
1417 start
= (caddr_t
)start
+ aclnode
->z_size
;
1421 * If Old version then swap count/bytes to match old
1422 * layout of znode_acl_phys_t.
1424 if (aclp
->z_version
== ZFS_ACL_VERSION_INITIAL
) {
1425 acl_phys
.z_acl_size
= aclp
->z_acl_count
;
1426 acl_phys
.z_acl_count
= aclp
->z_acl_bytes
;
1428 acl_phys
.z_acl_size
= aclp
->z_acl_bytes
;
1429 acl_phys
.z_acl_count
= aclp
->z_acl_count
;
1431 acl_phys
.z_acl_version
= aclp
->z_version
;
1433 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_ZNODE_ACL(zsb
), NULL
,
1434 &acl_phys
, sizeof (acl_phys
));
1438 * Replace ACL wide bits, but first clear them.
1440 zp
->z_pflags
&= ~ZFS_ACL_WIDE_FLAGS
;
1442 zp
->z_pflags
|= aclp
->z_hints
;
1444 if (ace_trivial_common(aclp
, 0, zfs_ace_walk
) == 0)
1445 zp
->z_pflags
|= ZFS_ACL_TRIVIAL
;
1447 zfs_tstamp_update_setup(zp
, STATE_CHANGED
, NULL
, ctime
, B_TRUE
);
1448 return (sa_bulk_update(zp
->z_sa_hdl
, bulk
, count
, tx
));
1452 zfs_acl_chmod(zfs_sb_t
*zsb
, uint64_t mode
, zfs_acl_t
*aclp
)
1456 int new_count
, new_bytes
;
1459 uint16_t iflags
, type
;
1460 uint32_t access_mask
;
1461 zfs_acl_node_t
*newnode
;
1462 size_t abstract_size
= aclp
->z_ops
->ace_abstract_size();
1464 uint32_t owner
, group
, everyone
;
1465 uint32_t deny1
, deny2
, allow0
;
1467 new_count
= new_bytes
= 0;
1469 acl_trivial_access_masks((mode_t
)mode
, &allow0
, &deny1
, &deny2
,
1470 &owner
, &group
, &everyone
);
1472 newnode
= zfs_acl_node_alloc((abstract_size
* 6) + aclp
->z_acl_bytes
);
1474 zacep
= newnode
->z_acldata
;
1476 zfs_set_ace(aclp
, zacep
, allow0
, ALLOW
, -1, ACE_OWNER
);
1477 zacep
= (void *)((uintptr_t)zacep
+ abstract_size
);
1479 new_bytes
+= abstract_size
;
1482 zfs_set_ace(aclp
, zacep
, deny1
, DENY
, -1, ACE_OWNER
);
1483 zacep
= (void *)((uintptr_t)zacep
+ abstract_size
);
1485 new_bytes
+= abstract_size
;
1488 zfs_set_ace(aclp
, zacep
, deny2
, DENY
, -1, OWNING_GROUP
);
1489 zacep
= (void *)((uintptr_t)zacep
+ abstract_size
);
1491 new_bytes
+= abstract_size
;
1494 while ((acep
= zfs_acl_next_ace(aclp
, acep
, &who
, &access_mask
,
1496 uint16_t inherit_flags
;
1498 entry_type
= (iflags
& ACE_TYPE_FLAGS
);
1499 inherit_flags
= (iflags
& ALL_INHERIT
);
1501 if ((entry_type
== ACE_OWNER
|| entry_type
== ACE_EVERYONE
||
1502 (entry_type
== OWNING_GROUP
)) &&
1503 ((inherit_flags
& ACE_INHERIT_ONLY_ACE
) == 0)) {
1507 if ((type
!= ALLOW
&& type
!= DENY
) ||
1508 (inherit_flags
& ACE_INHERIT_ONLY_ACE
)) {
1510 aclp
->z_hints
|= ZFS_INHERIT_ACE
;
1512 case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE
:
1513 case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE
:
1514 case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE
:
1515 case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE
:
1516 aclp
->z_hints
|= ZFS_ACL_OBJ_ACE
;
1522 * Limit permissions to be no greater than
1525 if (zsb
->z_acl_inherit
== ZFS_ACL_RESTRICTED
) {
1526 if (!(mode
& S_IRGRP
))
1527 access_mask
&= ~ACE_READ_DATA
;
1528 if (!(mode
& S_IWGRP
))
1530 ~(ACE_WRITE_DATA
|ACE_APPEND_DATA
);
1531 if (!(mode
& S_IXGRP
))
1532 access_mask
&= ~ACE_EXECUTE
;
1534 ~(ACE_WRITE_OWNER
|ACE_WRITE_ACL
|
1535 ACE_WRITE_ATTRIBUTES
|ACE_WRITE_NAMED_ATTRS
);
1538 zfs_set_ace(aclp
, zacep
, access_mask
, type
, who
, iflags
);
1539 ace_size
= aclp
->z_ops
->ace_size(acep
);
1540 zacep
= (void *)((uintptr_t)zacep
+ ace_size
);
1542 new_bytes
+= ace_size
;
1544 zfs_set_ace(aclp
, zacep
, owner
, 0, -1, ACE_OWNER
);
1545 zacep
= (void *)((uintptr_t)zacep
+ abstract_size
);
1546 zfs_set_ace(aclp
, zacep
, group
, 0, -1, OWNING_GROUP
);
1547 zacep
= (void *)((uintptr_t)zacep
+ abstract_size
);
1548 zfs_set_ace(aclp
, zacep
, everyone
, 0, -1, ACE_EVERYONE
);
1551 new_bytes
+= abstract_size
* 3;
1552 zfs_acl_release_nodes(aclp
);
1553 aclp
->z_acl_count
= new_count
;
1554 aclp
->z_acl_bytes
= new_bytes
;
1555 newnode
->z_ace_count
= new_count
;
1556 newnode
->z_size
= new_bytes
;
1557 list_insert_tail(&aclp
->z_acl
, newnode
);
1561 zfs_acl_chmod_setattr(znode_t
*zp
, zfs_acl_t
**aclp
, uint64_t mode
)
1563 mutex_enter(&zp
->z_acl_lock
);
1564 mutex_enter(&zp
->z_lock
);
1565 *aclp
= zfs_acl_alloc(zfs_acl_version_zp(zp
));
1566 (*aclp
)->z_hints
= zp
->z_pflags
& V4_ACL_WIDE_FLAGS
;
1567 zfs_acl_chmod(ZTOZSB(zp
), mode
, *aclp
);
1568 mutex_exit(&zp
->z_lock
);
1569 mutex_exit(&zp
->z_acl_lock
);
1574 * strip off write_owner and write_acl
1577 zfs_restricted_update(zfs_sb_t
*zsb
, zfs_acl_t
*aclp
, void *acep
)
1579 uint32_t mask
= aclp
->z_ops
->ace_mask_get(acep
);
1581 if ((zsb
->z_acl_inherit
== ZFS_ACL_RESTRICTED
) &&
1582 (aclp
->z_ops
->ace_type_get(acep
) == ALLOW
)) {
1583 mask
&= ~RESTRICTED_CLEAR
;
1584 aclp
->z_ops
->ace_mask_set(acep
, mask
);
1589 * Should ACE be inherited?
1592 zfs_ace_can_use(umode_t obj_mode
, uint16_t acep_flags
)
1594 int iflags
= (acep_flags
& 0xf);
1596 if (S_ISDIR(obj_mode
) && (iflags
& ACE_DIRECTORY_INHERIT_ACE
))
1598 else if (iflags
& ACE_FILE_INHERIT_ACE
)
1599 return (!(S_ISDIR(obj_mode
) &&
1600 (iflags
& ACE_NO_PROPAGATE_INHERIT_ACE
)));
1605 * inherit inheritable ACEs from parent
1608 zfs_acl_inherit(zfs_sb_t
*zsb
, umode_t obj_mode
, zfs_acl_t
*paclp
,
1609 uint64_t mode
, boolean_t
*need_chmod
)
1613 zfs_acl_node_t
*aclnode
;
1614 zfs_acl_t
*aclp
= NULL
;
1616 uint32_t access_mask
;
1617 uint16_t iflags
, newflags
, type
;
1619 void *data1
, *data2
;
1620 size_t data1sz
, data2sz
;
1621 boolean_t vdir
= S_ISDIR(obj_mode
);
1622 boolean_t vreg
= S_ISREG(obj_mode
);
1623 boolean_t passthrough
, passthrough_x
, noallow
;
1626 zsb
->z_acl_inherit
== ZFS_ACL_PASSTHROUGH_X
;
1627 passthrough
= passthrough_x
||
1628 zsb
->z_acl_inherit
== ZFS_ACL_PASSTHROUGH
;
1630 zsb
->z_acl_inherit
== ZFS_ACL_NOALLOW
;
1632 *need_chmod
= B_TRUE
;
1634 aclp
= zfs_acl_alloc(paclp
->z_version
);
1635 if (zsb
->z_acl_inherit
== ZFS_ACL_DISCARD
|| S_ISLNK(obj_mode
))
1637 while ((pacep
= zfs_acl_next_ace(paclp
, pacep
, &who
,
1638 &access_mask
, &iflags
, &type
))) {
1641 * don't inherit bogus ACEs
1643 if (!zfs_acl_valid_ace_type(type
, iflags
))
1646 if (noallow
&& type
== ALLOW
)
1649 ace_size
= aclp
->z_ops
->ace_size(pacep
);
1651 if (!zfs_ace_can_use(obj_mode
, iflags
))
1655 * If owner@, group@, or everyone@ inheritable
1656 * then zfs_acl_chmod() isn't needed.
1659 ((iflags
& (ACE_OWNER
|ACE_EVERYONE
)) ||
1660 ((iflags
& OWNING_GROUP
) ==
1661 OWNING_GROUP
)) && (vreg
|| (vdir
&& (iflags
&
1662 ACE_DIRECTORY_INHERIT_ACE
)))) {
1663 *need_chmod
= B_FALSE
;
1666 if (!vdir
&& passthrough_x
&&
1667 ((mode
& (S_IXUSR
| S_IXGRP
| S_IXOTH
)) == 0)) {
1668 access_mask
&= ~ACE_EXECUTE
;
1671 aclnode
= zfs_acl_node_alloc(ace_size
);
1672 list_insert_tail(&aclp
->z_acl
, aclnode
);
1673 acep
= aclnode
->z_acldata
;
1675 zfs_set_ace(aclp
, acep
, access_mask
, type
,
1676 who
, iflags
|ACE_INHERITED_ACE
);
1679 * Copy special opaque data if any
1681 if ((data1sz
= paclp
->z_ops
->ace_data(pacep
, &data1
)) != 0) {
1682 VERIFY((data2sz
= aclp
->z_ops
->ace_data(acep
,
1683 &data2
)) == data1sz
);
1684 bcopy(data1
, data2
, data2sz
);
1687 aclp
->z_acl_count
++;
1688 aclnode
->z_ace_count
++;
1689 aclp
->z_acl_bytes
+= aclnode
->z_size
;
1690 newflags
= aclp
->z_ops
->ace_flags_get(acep
);
1693 aclp
->z_hints
|= ZFS_INHERIT_ACE
;
1695 if ((iflags
& ACE_NO_PROPAGATE_INHERIT_ACE
) || !vdir
) {
1696 newflags
&= ~ALL_INHERIT
;
1697 aclp
->z_ops
->ace_flags_set(acep
,
1698 newflags
|ACE_INHERITED_ACE
);
1699 zfs_restricted_update(zsb
, aclp
, acep
);
1706 * If only FILE_INHERIT is set then turn on
1709 if ((iflags
& (ACE_FILE_INHERIT_ACE
|
1710 ACE_DIRECTORY_INHERIT_ACE
)) == ACE_FILE_INHERIT_ACE
) {
1711 newflags
|= ACE_INHERIT_ONLY_ACE
;
1712 aclp
->z_ops
->ace_flags_set(acep
,
1713 newflags
|ACE_INHERITED_ACE
);
1715 newflags
&= ~ACE_INHERIT_ONLY_ACE
;
1716 aclp
->z_ops
->ace_flags_set(acep
,
1717 newflags
|ACE_INHERITED_ACE
);
1724 * Create file system object initial permissions
1725 * including inheritable ACEs.
1728 zfs_acl_ids_create(znode_t
*dzp
, int flag
, vattr_t
*vap
, cred_t
*cr
,
1729 vsecattr_t
*vsecp
, zfs_acl_ids_t
*acl_ids
)
1732 zfs_sb_t
*zsb
= ZTOZSB(dzp
);
1736 #endif /* HAVE_KSID */
1737 boolean_t need_chmod
= B_TRUE
;
1738 boolean_t inherited
= B_FALSE
;
1740 bzero(acl_ids
, sizeof (zfs_acl_ids_t
));
1741 acl_ids
->z_mode
= vap
->va_mode
;
1744 if ((error
= zfs_vsec_2_aclp(zsb
, vap
->va_mode
, vsecp
,
1745 cr
, &acl_ids
->z_fuidp
, &acl_ids
->z_aclp
)) != 0)
1748 acl_ids
->z_fuid
= vap
->va_uid
;
1749 acl_ids
->z_fgid
= vap
->va_gid
;
1752 * Determine uid and gid.
1754 if ((flag
& IS_ROOT_NODE
) || zsb
->z_replay
||
1755 ((flag
& IS_XATTR
) && (S_ISDIR(vap
->va_mode
)))) {
1756 acl_ids
->z_fuid
= zfs_fuid_create(zsb
, (uint64_t)vap
->va_uid
,
1757 cr
, ZFS_OWNER
, &acl_ids
->z_fuidp
);
1758 acl_ids
->z_fgid
= zfs_fuid_create(zsb
, (uint64_t)vap
->va_gid
,
1759 cr
, ZFS_GROUP
, &acl_ids
->z_fuidp
);
1762 acl_ids
->z_fuid
= zfs_fuid_create_cred(zsb
, ZFS_OWNER
,
1763 cr
, &acl_ids
->z_fuidp
);
1764 acl_ids
->z_fgid
= 0;
1765 if (vap
->va_mask
& AT_GID
) {
1766 acl_ids
->z_fgid
= zfs_fuid_create(zsb
,
1767 (uint64_t)vap
->va_gid
,
1768 cr
, ZFS_GROUP
, &acl_ids
->z_fuidp
);
1770 if (acl_ids
->z_fgid
!= dzp
->z_gid
&&
1771 !groupmember(vap
->va_gid
, cr
) &&
1772 secpolicy_vnode_create_gid(cr
) != 0)
1773 acl_ids
->z_fgid
= 0;
1775 if (acl_ids
->z_fgid
== 0) {
1776 if (dzp
->z_mode
& S_ISGID
) {
1780 acl_ids
->z_fgid
= dzp
->z_gid
;
1781 gid
= zfs_fuid_map_id(zsb
, acl_ids
->z_fgid
,
1784 if (zsb
->z_use_fuids
&&
1785 IS_EPHEMERAL(acl_ids
->z_fgid
)) {
1786 domain
= zfs_fuid_idx_domain(
1788 FUID_INDEX(acl_ids
->z_fgid
));
1789 rid
= FUID_RID(acl_ids
->z_fgid
);
1790 zfs_fuid_node_add(&acl_ids
->z_fuidp
,
1792 FUID_INDEX(acl_ids
->z_fgid
),
1793 acl_ids
->z_fgid
, ZFS_GROUP
);
1796 acl_ids
->z_fgid
= zfs_fuid_create_cred(zsb
,
1797 ZFS_GROUP
, cr
, &acl_ids
->z_fuidp
);
1802 #endif /* HAVE_KSID */
1805 * If we're creating a directory, and the parent directory has the
1806 * set-GID bit set, set in on the new directory.
1807 * Otherwise, if the user is neither privileged nor a member of the
1808 * file's new group, clear the file's set-GID bit.
1811 if (!(flag
& IS_ROOT_NODE
) && (dzp
->z_mode
& S_ISGID
) &&
1812 (S_ISDIR(vap
->va_mode
))) {
1813 acl_ids
->z_mode
|= S_ISGID
;
1815 if ((acl_ids
->z_mode
& S_ISGID
) &&
1816 secpolicy_vnode_setids_setgids(cr
, gid
) != 0)
1817 acl_ids
->z_mode
&= ~S_ISGID
;
1820 if (acl_ids
->z_aclp
== NULL
) {
1821 mutex_enter(&dzp
->z_acl_lock
);
1822 mutex_enter(&dzp
->z_lock
);
1823 if (!(flag
& IS_ROOT_NODE
) && (S_ISDIR(ZTOI(dzp
)->i_mode
) &&
1824 (dzp
->z_pflags
& ZFS_INHERIT_ACE
)) &&
1825 !(dzp
->z_pflags
& ZFS_XATTR
)) {
1826 VERIFY(0 == zfs_acl_node_read(dzp
, B_TRUE
,
1828 acl_ids
->z_aclp
= zfs_acl_inherit(zsb
,
1829 vap
->va_mode
, paclp
, acl_ids
->z_mode
, &need_chmod
);
1833 zfs_acl_alloc(zfs_acl_version_zp(dzp
));
1834 acl_ids
->z_aclp
->z_hints
|= ZFS_ACL_TRIVIAL
;
1836 mutex_exit(&dzp
->z_lock
);
1837 mutex_exit(&dzp
->z_acl_lock
);
1839 acl_ids
->z_aclp
->z_hints
|= S_ISDIR(vap
->va_mode
) ?
1840 ZFS_ACL_AUTO_INHERIT
: 0;
1841 zfs_acl_chmod(zsb
, acl_ids
->z_mode
, acl_ids
->z_aclp
);
1845 if (inherited
|| vsecp
) {
1846 acl_ids
->z_mode
= zfs_mode_compute(acl_ids
->z_mode
,
1847 acl_ids
->z_aclp
, &acl_ids
->z_aclp
->z_hints
,
1848 acl_ids
->z_fuid
, acl_ids
->z_fgid
);
1849 if (ace_trivial_common(acl_ids
->z_aclp
, 0, zfs_ace_walk
) == 0)
1850 acl_ids
->z_aclp
->z_hints
|= ZFS_ACL_TRIVIAL
;
1857 * Free ACL and fuid_infop, but not the acl_ids structure
1860 zfs_acl_ids_free(zfs_acl_ids_t
*acl_ids
)
1862 if (acl_ids
->z_aclp
)
1863 zfs_acl_free(acl_ids
->z_aclp
);
1864 if (acl_ids
->z_fuidp
)
1865 zfs_fuid_info_free(acl_ids
->z_fuidp
);
1866 acl_ids
->z_aclp
= NULL
;
1867 acl_ids
->z_fuidp
= NULL
;
1871 zfs_acl_ids_overquota(zfs_sb_t
*zsb
, zfs_acl_ids_t
*acl_ids
)
1873 return (zfs_fuid_overquota(zsb
, B_FALSE
, acl_ids
->z_fuid
) ||
1874 zfs_fuid_overquota(zsb
, B_TRUE
, acl_ids
->z_fgid
));
1878 * Retrieve a file's ACL
1881 zfs_getacl(znode_t
*zp
, vsecattr_t
*vsecp
, boolean_t skipaclchk
, cred_t
*cr
)
1889 mask
= vsecp
->vsa_mask
& (VSA_ACE
| VSA_ACECNT
|
1890 VSA_ACE_ACLFLAGS
| VSA_ACE_ALLTYPES
);
1893 return (SET_ERROR(ENOSYS
));
1895 if ((error
= zfs_zaccess(zp
, ACE_READ_ACL
, 0, skipaclchk
, cr
)))
1898 mutex_enter(&zp
->z_acl_lock
);
1900 error
= zfs_acl_node_read(zp
, B_FALSE
, &aclp
, B_FALSE
);
1902 mutex_exit(&zp
->z_acl_lock
);
1907 * Scan ACL to determine number of ACEs
1909 if ((zp
->z_pflags
& ZFS_ACL_OBJ_ACE
) && !(mask
& VSA_ACE_ALLTYPES
)) {
1912 uint32_t access_mask
;
1913 uint16_t type
, iflags
;
1915 while ((zacep
= zfs_acl_next_ace(aclp
, zacep
,
1916 &who
, &access_mask
, &iflags
, &type
))) {
1918 case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE
:
1919 case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE
:
1920 case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE
:
1921 case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE
:
1928 vsecp
->vsa_aclcnt
= count
;
1930 count
= (int)aclp
->z_acl_count
;
1932 if (mask
& VSA_ACECNT
) {
1933 vsecp
->vsa_aclcnt
= count
;
1936 if (mask
& VSA_ACE
) {
1939 aclsz
= count
* sizeof (ace_t
) +
1940 sizeof (ace_object_t
) * largeace
;
1942 vsecp
->vsa_aclentp
= kmem_alloc(aclsz
, KM_SLEEP
);
1943 vsecp
->vsa_aclentsz
= aclsz
;
1945 if (aclp
->z_version
== ZFS_ACL_VERSION_FUID
)
1946 zfs_copy_fuid_2_ace(ZTOZSB(zp
), aclp
, cr
,
1947 vsecp
->vsa_aclentp
, !(mask
& VSA_ACE_ALLTYPES
));
1949 zfs_acl_node_t
*aclnode
;
1950 void *start
= vsecp
->vsa_aclentp
;
1952 for (aclnode
= list_head(&aclp
->z_acl
); aclnode
;
1953 aclnode
= list_next(&aclp
->z_acl
, aclnode
)) {
1954 bcopy(aclnode
->z_acldata
, start
,
1956 start
= (caddr_t
)start
+ aclnode
->z_size
;
1958 ASSERT((caddr_t
)start
- (caddr_t
)vsecp
->vsa_aclentp
==
1962 if (mask
& VSA_ACE_ACLFLAGS
) {
1963 vsecp
->vsa_aclflags
= 0;
1964 if (zp
->z_pflags
& ZFS_ACL_DEFAULTED
)
1965 vsecp
->vsa_aclflags
|= ACL_DEFAULTED
;
1966 if (zp
->z_pflags
& ZFS_ACL_PROTECTED
)
1967 vsecp
->vsa_aclflags
|= ACL_PROTECTED
;
1968 if (zp
->z_pflags
& ZFS_ACL_AUTO_INHERIT
)
1969 vsecp
->vsa_aclflags
|= ACL_AUTO_INHERIT
;
1972 mutex_exit(&zp
->z_acl_lock
);
1978 zfs_vsec_2_aclp(zfs_sb_t
*zsb
, umode_t obj_mode
,
1979 vsecattr_t
*vsecp
, cred_t
*cr
, zfs_fuid_info_t
**fuidp
, zfs_acl_t
**zaclp
)
1982 zfs_acl_node_t
*aclnode
;
1983 int aclcnt
= vsecp
->vsa_aclcnt
;
1986 if (vsecp
->vsa_aclcnt
> MAX_ACL_ENTRIES
|| vsecp
->vsa_aclcnt
<= 0)
1987 return (SET_ERROR(EINVAL
));
1989 aclp
= zfs_acl_alloc(zfs_acl_version(zsb
->z_version
));
1992 aclnode
= zfs_acl_node_alloc(aclcnt
* sizeof (zfs_object_ace_t
));
1993 if (aclp
->z_version
== ZFS_ACL_VERSION_INITIAL
) {
1994 if ((error
= zfs_copy_ace_2_oldace(obj_mode
, aclp
,
1995 (ace_t
*)vsecp
->vsa_aclentp
, aclnode
->z_acldata
,
1996 aclcnt
, &aclnode
->z_size
)) != 0) {
1998 zfs_acl_node_free(aclnode
);
2002 if ((error
= zfs_copy_ace_2_fuid(zsb
, obj_mode
, aclp
,
2003 vsecp
->vsa_aclentp
, aclnode
->z_acldata
, aclcnt
,
2004 &aclnode
->z_size
, fuidp
, cr
)) != 0) {
2006 zfs_acl_node_free(aclnode
);
2010 aclp
->z_acl_bytes
= aclnode
->z_size
;
2011 aclnode
->z_ace_count
= aclcnt
;
2012 aclp
->z_acl_count
= aclcnt
;
2013 list_insert_head(&aclp
->z_acl
, aclnode
);
2016 * If flags are being set then add them to z_hints
2018 if (vsecp
->vsa_mask
& VSA_ACE_ACLFLAGS
) {
2019 if (vsecp
->vsa_aclflags
& ACL_PROTECTED
)
2020 aclp
->z_hints
|= ZFS_ACL_PROTECTED
;
2021 if (vsecp
->vsa_aclflags
& ACL_DEFAULTED
)
2022 aclp
->z_hints
|= ZFS_ACL_DEFAULTED
;
2023 if (vsecp
->vsa_aclflags
& ACL_AUTO_INHERIT
)
2024 aclp
->z_hints
|= ZFS_ACL_AUTO_INHERIT
;
2036 zfs_setacl(znode_t
*zp
, vsecattr_t
*vsecp
, boolean_t skipaclchk
, cred_t
*cr
)
2038 zfs_sb_t
*zsb
= ZTOZSB(zp
);
2039 zilog_t
*zilog
= zsb
->z_log
;
2040 ulong_t mask
= vsecp
->vsa_mask
& (VSA_ACE
| VSA_ACECNT
);
2044 zfs_fuid_info_t
*fuidp
= NULL
;
2045 boolean_t fuid_dirtied
;
2049 return (SET_ERROR(ENOSYS
));
2051 if (zp
->z_pflags
& ZFS_IMMUTABLE
)
2052 return (SET_ERROR(EPERM
));
2054 if ((error
= zfs_zaccess(zp
, ACE_WRITE_ACL
, 0, skipaclchk
, cr
)))
2057 error
= zfs_vsec_2_aclp(zsb
, ZTOI(zp
)->i_mode
, vsecp
, cr
, &fuidp
,
2063 * If ACL wide flags aren't being set then preserve any
2066 if (!(vsecp
->vsa_mask
& VSA_ACE_ACLFLAGS
)) {
2068 (zp
->z_pflags
& V4_ACL_WIDE_FLAGS
);
2071 mutex_enter(&zp
->z_acl_lock
);
2072 mutex_enter(&zp
->z_lock
);
2074 tx
= dmu_tx_create(zsb
->z_os
);
2076 dmu_tx_hold_sa(tx
, zp
->z_sa_hdl
, B_TRUE
);
2078 fuid_dirtied
= zsb
->z_fuid_dirty
;
2080 zfs_fuid_txhold(zsb
, tx
);
2083 * If old version and ACL won't fit in bonus and we aren't
2084 * upgrading then take out necessary DMU holds
2087 if ((acl_obj
= zfs_external_acl(zp
)) != 0) {
2088 if (zsb
->z_version
>= ZPL_VERSION_FUID
&&
2089 zfs_znode_acl_version(zp
) <= ZFS_ACL_VERSION_INITIAL
) {
2090 dmu_tx_hold_free(tx
, acl_obj
, 0,
2092 dmu_tx_hold_write(tx
, DMU_NEW_OBJECT
, 0,
2095 dmu_tx_hold_write(tx
, acl_obj
, 0, aclp
->z_acl_bytes
);
2097 } else if (!zp
->z_is_sa
&& aclp
->z_acl_bytes
> ZFS_ACE_SPACE
) {
2098 dmu_tx_hold_write(tx
, DMU_NEW_OBJECT
, 0, aclp
->z_acl_bytes
);
2101 zfs_sa_upgrade_txholds(tx
, zp
);
2102 error
= dmu_tx_assign(tx
, TXG_NOWAIT
);
2104 mutex_exit(&zp
->z_acl_lock
);
2105 mutex_exit(&zp
->z_lock
);
2107 if (error
== ERESTART
) {
2117 error
= zfs_aclset_common(zp
, aclp
, cr
, tx
);
2119 ASSERT(zp
->z_acl_cached
== NULL
);
2120 zp
->z_acl_cached
= aclp
;
2123 zfs_fuid_sync(zsb
, tx
);
2125 zfs_log_acl(zilog
, tx
, zp
, vsecp
, fuidp
);
2128 zfs_fuid_info_free(fuidp
);
2131 mutex_exit(&zp
->z_lock
);
2132 mutex_exit(&zp
->z_acl_lock
);
2138 * Check accesses of interest (AoI) against attributes of the dataset
2139 * such as read-only. Returns zero if no AoI conflict with dataset
2140 * attributes, otherwise an appropriate errno is returned.
2143 zfs_zaccess_dataset_check(znode_t
*zp
, uint32_t v4_mode
)
2145 if ((v4_mode
& WRITE_MASK
) && (zfs_is_readonly(ZTOZSB(zp
))) &&
2146 (!S_ISDEV(ZTOI(zp
)->i_mode
) ||
2147 (S_ISDEV(ZTOI(zp
)->i_mode
) && (v4_mode
& WRITE_MASK_ATTRS
)))) {
2148 return (SET_ERROR(EROFS
));
2152 * Only check for READONLY on non-directories.
2154 if ((v4_mode
& WRITE_MASK_DATA
) &&
2155 ((!S_ISDIR(ZTOI(zp
)->i_mode
) &&
2156 (zp
->z_pflags
& (ZFS_READONLY
| ZFS_IMMUTABLE
))) ||
2157 (S_ISDIR(ZTOI(zp
)->i_mode
) &&
2158 (zp
->z_pflags
& ZFS_IMMUTABLE
)))) {
2159 return (SET_ERROR(EPERM
));
2162 if ((v4_mode
& (ACE_DELETE
| ACE_DELETE_CHILD
)) &&
2163 (zp
->z_pflags
& ZFS_NOUNLINK
)) {
2164 return (SET_ERROR(EPERM
));
2167 if (((v4_mode
& (ACE_READ_DATA
|ACE_EXECUTE
)) &&
2168 (zp
->z_pflags
& ZFS_AV_QUARANTINED
))) {
2169 return (SET_ERROR(EACCES
));
2176 * The primary usage of this function is to loop through all of the
2177 * ACEs in the znode, determining what accesses of interest (AoI) to
2178 * the caller are allowed or denied. The AoI are expressed as bits in
2179 * the working_mode parameter. As each ACE is processed, bits covered
2180 * by that ACE are removed from the working_mode. This removal
2181 * facilitates two things. The first is that when the working mode is
2182 * empty (= 0), we know we've looked at all the AoI. The second is
2183 * that the ACE interpretation rules don't allow a later ACE to undo
2184 * something granted or denied by an earlier ACE. Removing the
2185 * discovered access or denial enforces this rule. At the end of
2186 * processing the ACEs, all AoI that were found to be denied are
2187 * placed into the working_mode, giving the caller a mask of denied
2188 * accesses. Returns:
2189 * 0 if all AoI granted
2190 * EACCESS if the denied mask is non-zero
2191 * other error if abnormal failure (e.g., IO error)
2193 * A secondary usage of the function is to determine if any of the
2194 * AoI are granted. If an ACE grants any access in
2195 * the working_mode, we immediately short circuit out of the function.
2196 * This mode is chosen by setting anyaccess to B_TRUE. The
2197 * working_mode is not a denied access mask upon exit if the function
2198 * is used in this manner.
2201 zfs_zaccess_aces_check(znode_t
*zp
, uint32_t *working_mode
,
2202 boolean_t anyaccess
, cred_t
*cr
)
2204 zfs_sb_t
*zsb
= ZTOZSB(zp
);
2207 uid_t uid
= crgetuid(cr
);
2209 uint16_t type
, iflags
;
2210 uint16_t entry_type
;
2211 uint32_t access_mask
;
2212 uint32_t deny_mask
= 0;
2213 zfs_ace_hdr_t
*acep
= NULL
;
2218 zfs_fuid_map_ids(zp
, cr
, &fowner
, &gowner
);
2220 mutex_enter(&zp
->z_acl_lock
);
2222 error
= zfs_acl_node_read(zp
, B_FALSE
, &aclp
, B_FALSE
);
2224 mutex_exit(&zp
->z_acl_lock
);
2228 ASSERT(zp
->z_acl_cached
);
2230 while ((acep
= zfs_acl_next_ace(aclp
, acep
, &who
, &access_mask
,
2232 uint32_t mask_matched
;
2234 if (!zfs_acl_valid_ace_type(type
, iflags
))
2237 if (S_ISDIR(ZTOI(zp
)->i_mode
) &&
2238 (iflags
& ACE_INHERIT_ONLY_ACE
))
2241 /* Skip ACE if it does not affect any AoI */
2242 mask_matched
= (access_mask
& *working_mode
);
2246 entry_type
= (iflags
& ACE_TYPE_FLAGS
);
2250 switch (entry_type
) {
2258 case ACE_IDENTIFIER_GROUP
:
2259 checkit
= zfs_groupmember(zsb
, who
, cr
);
2267 if (entry_type
== 0) {
2270 newid
= zfs_fuid_map_id(zsb
, who
, cr
,
2272 if (newid
!= IDMAP_WK_CREATOR_OWNER_UID
&&
2277 mutex_exit(&zp
->z_acl_lock
);
2278 return (SET_ERROR(EIO
));
2284 DTRACE_PROBE3(zfs__ace__denies
,
2286 zfs_ace_hdr_t
*, acep
,
2287 uint32_t, mask_matched
);
2288 deny_mask
|= mask_matched
;
2290 DTRACE_PROBE3(zfs__ace__allows
,
2292 zfs_ace_hdr_t
*, acep
,
2293 uint32_t, mask_matched
);
2295 mutex_exit(&zp
->z_acl_lock
);
2299 *working_mode
&= ~mask_matched
;
2303 if (*working_mode
== 0)
2307 mutex_exit(&zp
->z_acl_lock
);
2309 /* Put the found 'denies' back on the working mode */
2311 *working_mode
|= deny_mask
;
2312 return (SET_ERROR(EACCES
));
2313 } else if (*working_mode
) {
2321 * Return true if any access whatsoever granted, we don't actually
2322 * care what access is granted.
2325 zfs_has_access(znode_t
*zp
, cred_t
*cr
)
2327 uint32_t have
= ACE_ALL_PERMS
;
2329 if (zfs_zaccess_aces_check(zp
, &have
, B_TRUE
, cr
) != 0) {
2332 owner
= zfs_fuid_map_id(ZTOZSB(zp
), zp
->z_uid
, cr
, ZFS_OWNER
);
2333 return (secpolicy_vnode_any_access(cr
, ZTOI(zp
), owner
) == 0);
2339 zfs_zaccess_common(znode_t
*zp
, uint32_t v4_mode
, uint32_t *working_mode
,
2340 boolean_t
*check_privs
, boolean_t skipaclchk
, cred_t
*cr
)
2342 zfs_sb_t
*zsb
= ZTOZSB(zp
);
2345 *working_mode
= v4_mode
;
2346 *check_privs
= B_TRUE
;
2349 * Short circuit empty requests
2351 if (v4_mode
== 0 || zsb
->z_replay
) {
2356 if ((err
= zfs_zaccess_dataset_check(zp
, v4_mode
)) != 0) {
2357 *check_privs
= B_FALSE
;
2362 * The caller requested that the ACL check be skipped. This
2363 * would only happen if the caller checked VOP_ACCESS() with a
2364 * 32 bit ACE mask and already had the appropriate permissions.
2371 return (zfs_zaccess_aces_check(zp
, working_mode
, B_FALSE
, cr
));
2375 zfs_zaccess_append(znode_t
*zp
, uint32_t *working_mode
, boolean_t
*check_privs
,
2378 if (*working_mode
!= ACE_WRITE_DATA
)
2379 return (SET_ERROR(EACCES
));
2381 return (zfs_zaccess_common(zp
, ACE_APPEND_DATA
, working_mode
,
2382 check_privs
, B_FALSE
, cr
));
2386 zfs_fastaccesschk_execute(znode_t
*zdp
, cred_t
*cr
)
2388 boolean_t owner
= B_FALSE
;
2389 boolean_t groupmbr
= B_FALSE
;
2391 uid_t uid
= crgetuid(cr
);
2394 if (zdp
->z_pflags
& ZFS_AV_QUARANTINED
)
2395 return (SET_ERROR(EACCES
));
2397 is_attr
= ((zdp
->z_pflags
& ZFS_XATTR
) &&
2398 (S_ISDIR(ZTOI(zdp
)->i_mode
)));
2403 mutex_enter(&zdp
->z_acl_lock
);
2405 if (zdp
->z_pflags
& ZFS_NO_EXECS_DENIED
) {
2406 mutex_exit(&zdp
->z_acl_lock
);
2410 if (FUID_INDEX(zdp
->z_uid
) != 0 || FUID_INDEX(zdp
->z_gid
) != 0) {
2411 mutex_exit(&zdp
->z_acl_lock
);
2415 if (uid
== zdp
->z_uid
) {
2417 if (zdp
->z_mode
& S_IXUSR
) {
2418 mutex_exit(&zdp
->z_acl_lock
);
2421 mutex_exit(&zdp
->z_acl_lock
);
2425 if (groupmember(zdp
->z_gid
, cr
)) {
2427 if (zdp
->z_mode
& S_IXGRP
) {
2428 mutex_exit(&zdp
->z_acl_lock
);
2431 mutex_exit(&zdp
->z_acl_lock
);
2435 if (!owner
&& !groupmbr
) {
2436 if (zdp
->z_mode
& S_IXOTH
) {
2437 mutex_exit(&zdp
->z_acl_lock
);
2442 mutex_exit(&zdp
->z_acl_lock
);
2445 DTRACE_PROBE(zfs__fastpath__execute__access__miss
);
2446 ZFS_ENTER(ZTOZSB(zdp
));
2447 error
= zfs_zaccess(zdp
, ACE_EXECUTE
, 0, B_FALSE
, cr
);
2448 ZFS_EXIT(ZTOZSB(zdp
));
2453 * Determine whether Access should be granted/denied.
2455 * The least priv subsytem is always consulted as a basic privilege
2456 * can define any form of access.
2459 zfs_zaccess(znode_t
*zp
, int mode
, int flags
, boolean_t skipaclchk
, cred_t
*cr
)
2461 uint32_t working_mode
;
2463 boolean_t check_privs
;
2464 znode_t
*check_zp
= zp
;
2469 * If attribute then validate against base file
2471 if ((zp
->z_pflags
& ZFS_XATTR
) && S_ISDIR(ZTOI(zp
)->i_mode
)) {
2474 rw_enter(&zp
->z_xattr_lock
, RW_READER
);
2475 if (zp
->z_xattr_parent
) {
2476 check_zp
= zp
->z_xattr_parent
;
2477 rw_exit(&zp
->z_xattr_lock
);
2480 * Verify a lookup yields the same znode.
2482 ASSERT3S(sa_lookup(zp
->z_sa_hdl
, SA_ZPL_PARENT(
2483 ZTOZSB(zp
)), &parent
, sizeof (parent
)), ==, 0);
2484 ASSERT3U(check_zp
->z_id
, ==, parent
);
2486 rw_exit(&zp
->z_xattr_lock
);
2488 error
= sa_lookup(zp
->z_sa_hdl
, SA_ZPL_PARENT(
2489 ZTOZSB(zp
)), &parent
, sizeof (parent
));
2494 * Cache the lookup on the parent file znode as
2495 * zp->z_xattr_parent and hold a reference. This
2496 * effectively pins the parent in memory until all
2497 * child xattr znodes have been destroyed and
2498 * release their references in zfs_inode_destroy().
2500 error
= zfs_zget(ZTOZSB(zp
), parent
, &check_zp
);
2504 rw_enter(&zp
->z_xattr_lock
, RW_WRITER
);
2505 if (zp
->z_xattr_parent
== NULL
)
2506 zp
->z_xattr_parent
= check_zp
;
2507 rw_exit(&zp
->z_xattr_lock
);
2511 * fixup mode to map to xattr perms
2514 if (mode
& (ACE_WRITE_DATA
|ACE_APPEND_DATA
)) {
2515 mode
&= ~(ACE_WRITE_DATA
|ACE_APPEND_DATA
);
2516 mode
|= ACE_WRITE_NAMED_ATTRS
;
2519 if (mode
& (ACE_READ_DATA
|ACE_EXECUTE
)) {
2520 mode
&= ~(ACE_READ_DATA
|ACE_EXECUTE
);
2521 mode
|= ACE_READ_NAMED_ATTRS
;
2525 owner
= zfs_fuid_map_id(ZTOZSB(zp
), zp
->z_uid
, cr
, ZFS_OWNER
);
2527 * Map the bits required to the standard inode flags
2528 * S_IRUSR|S_IWUSR|S_IXUSR in the needed_bits. Map the bits
2529 * mapped by working_mode (currently missing) in missing_bits.
2530 * Call secpolicy_vnode_access2() with (needed_bits & ~checkmode),
2535 working_mode
= mode
;
2536 if ((working_mode
& (ACE_READ_ACL
|ACE_READ_ATTRIBUTES
)) &&
2537 owner
== crgetuid(cr
))
2538 working_mode
&= ~(ACE_READ_ACL
|ACE_READ_ATTRIBUTES
);
2540 if (working_mode
& (ACE_READ_DATA
|ACE_READ_NAMED_ATTRS
|
2541 ACE_READ_ACL
|ACE_READ_ATTRIBUTES
|ACE_SYNCHRONIZE
))
2542 needed_bits
|= S_IRUSR
;
2543 if (working_mode
& (ACE_WRITE_DATA
|ACE_WRITE_NAMED_ATTRS
|
2544 ACE_APPEND_DATA
|ACE_WRITE_ATTRIBUTES
|ACE_SYNCHRONIZE
))
2545 needed_bits
|= S_IWUSR
;
2546 if (working_mode
& ACE_EXECUTE
)
2547 needed_bits
|= S_IXUSR
;
2549 if ((error
= zfs_zaccess_common(check_zp
, mode
, &working_mode
,
2550 &check_privs
, skipaclchk
, cr
)) == 0) {
2551 return (secpolicy_vnode_access2(cr
, ZTOI(zp
), owner
,
2552 needed_bits
, needed_bits
));
2555 if (error
&& !check_privs
) {
2559 if (error
&& (flags
& V_APPEND
)) {
2560 error
= zfs_zaccess_append(zp
, &working_mode
, &check_privs
, cr
);
2563 if (error
&& check_privs
) {
2564 mode_t checkmode
= 0;
2567 * First check for implicit owner permission on
2568 * read_acl/read_attributes
2572 ASSERT(working_mode
!= 0);
2574 if ((working_mode
& (ACE_READ_ACL
|ACE_READ_ATTRIBUTES
) &&
2575 owner
== crgetuid(cr
)))
2576 working_mode
&= ~(ACE_READ_ACL
|ACE_READ_ATTRIBUTES
);
2578 if (working_mode
& (ACE_READ_DATA
|ACE_READ_NAMED_ATTRS
|
2579 ACE_READ_ACL
|ACE_READ_ATTRIBUTES
|ACE_SYNCHRONIZE
))
2580 checkmode
|= S_IRUSR
;
2581 if (working_mode
& (ACE_WRITE_DATA
|ACE_WRITE_NAMED_ATTRS
|
2582 ACE_APPEND_DATA
|ACE_WRITE_ATTRIBUTES
|ACE_SYNCHRONIZE
))
2583 checkmode
|= S_IWUSR
;
2584 if (working_mode
& ACE_EXECUTE
)
2585 checkmode
|= S_IXUSR
;
2587 error
= secpolicy_vnode_access2(cr
, ZTOI(check_zp
), owner
,
2588 needed_bits
& ~checkmode
, needed_bits
);
2590 if (error
== 0 && (working_mode
& ACE_WRITE_OWNER
))
2591 error
= secpolicy_vnode_chown(cr
, owner
);
2592 if (error
== 0 && (working_mode
& ACE_WRITE_ACL
))
2593 error
= secpolicy_vnode_setdac(cr
, owner
);
2595 if (error
== 0 && (working_mode
&
2596 (ACE_DELETE
|ACE_DELETE_CHILD
)))
2597 error
= secpolicy_vnode_remove(cr
);
2599 if (error
== 0 && (working_mode
& ACE_SYNCHRONIZE
)) {
2600 error
= secpolicy_vnode_chown(cr
, owner
);
2604 * See if any bits other than those already checked
2605 * for are still present. If so then return EACCES
2607 if (working_mode
& ~(ZFS_CHECKED_MASKS
)) {
2608 error
= SET_ERROR(EACCES
);
2611 } else if (error
== 0) {
2612 error
= secpolicy_vnode_access2(cr
, ZTOI(zp
), owner
,
2613 needed_bits
, needed_bits
);
2620 * Translate traditional unix S_IRUSR/S_IWUSR/S_IXUSR mode into
2621 * native ACL format and call zfs_zaccess()
2624 zfs_zaccess_rwx(znode_t
*zp
, mode_t mode
, int flags
, cred_t
*cr
)
2626 return (zfs_zaccess(zp
, zfs_unix_to_v4(mode
>> 6), flags
, B_FALSE
, cr
));
2630 * Access function for secpolicy_vnode_setattr
2633 zfs_zaccess_unix(znode_t
*zp
, mode_t mode
, cred_t
*cr
)
2635 int v4_mode
= zfs_unix_to_v4(mode
>> 6);
2637 return (zfs_zaccess(zp
, v4_mode
, 0, B_FALSE
, cr
));
2641 zfs_delete_final_check(znode_t
*zp
, znode_t
*dzp
,
2642 mode_t available_perms
, cred_t
*cr
)
2647 downer
= zfs_fuid_map_id(ZTOZSB(dzp
), dzp
->z_uid
, cr
, ZFS_OWNER
);
2649 error
= secpolicy_vnode_access2(cr
, ZTOI(dzp
),
2650 downer
, available_perms
, S_IWUSR
|S_IXUSR
);
2653 error
= zfs_sticky_remove_access(dzp
, zp
, cr
);
2659 * Determine whether Access should be granted/deny, without
2660 * consulting least priv subsystem.
2662 * The following chart is the recommended NFSv4 enforcement for
2663 * ability to delete an object.
2665 * -------------------------------------------------------
2666 * | Parent Dir | Target Object Permissions |
2668 * -------------------------------------------------------
2669 * | | ACL Allows | ACL Denies| Delete |
2670 * | | Delete | Delete | unspecified|
2671 * -------------------------------------------------------
2672 * | ACL Allows | Permit | Permit | Permit |
2673 * | DELETE_CHILD | |
2674 * -------------------------------------------------------
2675 * | ACL Denies | Permit | Deny | Deny |
2676 * | DELETE_CHILD | | | |
2677 * -------------------------------------------------------
2678 * | ACL specifies | | | |
2679 * | only allow | Permit | Permit | Permit |
2680 * | write and | | | |
2682 * -------------------------------------------------------
2683 * | ACL denies | | | |
2684 * | write and | Permit | Deny | Deny |
2686 * -------------------------------------------------------
2689 * No search privilege, can't even look up file?
2693 zfs_zaccess_delete(znode_t
*dzp
, znode_t
*zp
, cred_t
*cr
)
2695 uint32_t dzp_working_mode
= 0;
2696 uint32_t zp_working_mode
= 0;
2697 int dzp_error
, zp_error
;
2698 mode_t available_perms
;
2699 boolean_t dzpcheck_privs
= B_TRUE
;
2700 boolean_t zpcheck_privs
= B_TRUE
;
2703 * We want specific DELETE permissions to
2704 * take precedence over WRITE/EXECUTE. We don't
2705 * want an ACL such as this to mess us up.
2706 * user:joe:write_data:deny,user:joe:delete:allow
2708 * However, deny permissions may ultimately be overridden
2709 * by secpolicy_vnode_access().
2711 * We will ask for all of the necessary permissions and then
2712 * look at the working modes from the directory and target object
2713 * to determine what was found.
2716 if (zp
->z_pflags
& (ZFS_IMMUTABLE
| ZFS_NOUNLINK
))
2717 return (SET_ERROR(EPERM
));
2721 * If the directory permissions allow the delete, we are done.
2723 if ((dzp_error
= zfs_zaccess_common(dzp
, ACE_DELETE_CHILD
,
2724 &dzp_working_mode
, &dzpcheck_privs
, B_FALSE
, cr
)) == 0)
2728 * If target object has delete permission then we are done
2730 if ((zp_error
= zfs_zaccess_common(zp
, ACE_DELETE
, &zp_working_mode
,
2731 &zpcheck_privs
, B_FALSE
, cr
)) == 0)
2734 ASSERT(dzp_error
&& zp_error
);
2736 if (!dzpcheck_privs
)
2744 * If directory returns EACCES then delete_child was denied
2745 * due to deny delete_child. In this case send the request through
2746 * secpolicy_vnode_remove(). We don't use zfs_delete_final_check()
2747 * since that *could* allow the delete based on write/execute permission
2748 * and we want delete permissions to override write/execute.
2751 if (dzp_error
== EACCES
)
2752 return (secpolicy_vnode_remove(cr
));
2756 * only need to see if we have write/execute on directory.
2759 dzp_error
= zfs_zaccess_common(dzp
, ACE_EXECUTE
|ACE_WRITE_DATA
,
2760 &dzp_working_mode
, &dzpcheck_privs
, B_FALSE
, cr
);
2762 if (dzp_error
!= 0 && !dzpcheck_privs
)
2769 available_perms
= (dzp_working_mode
& ACE_WRITE_DATA
) ? 0 : S_IWUSR
;
2770 available_perms
|= (dzp_working_mode
& ACE_EXECUTE
) ? 0 : S_IXUSR
;
2772 return (zfs_delete_final_check(zp
, dzp
, available_perms
, cr
));
2777 zfs_zaccess_rename(znode_t
*sdzp
, znode_t
*szp
, znode_t
*tdzp
,
2778 znode_t
*tzp
, cred_t
*cr
)
2783 if (szp
->z_pflags
& ZFS_AV_QUARANTINED
)
2784 return (SET_ERROR(EACCES
));
2786 add_perm
= S_ISDIR(ZTOI(szp
)->i_mode
) ?
2787 ACE_ADD_SUBDIRECTORY
: ACE_ADD_FILE
;
2790 * Rename permissions are combination of delete permission +
2791 * add file/subdir permission.
2795 * first make sure we do the delete portion.
2797 * If that succeeds then check for add_file/add_subdir permissions
2800 if ((error
= zfs_zaccess_delete(sdzp
, szp
, cr
)))
2804 * If we have a tzp, see if we can delete it?
2807 if ((error
= zfs_zaccess_delete(tdzp
, tzp
, cr
)))
2812 * Now check for add permissions
2814 error
= zfs_zaccess(tdzp
, add_perm
, 0, B_FALSE
, cr
);