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 <sys/trace_acl.h>
57 #include "fs/fs_subr.h"
59 #define ALLOW ACE_ACCESS_ALLOWED_ACE_TYPE
60 #define DENY ACE_ACCESS_DENIED_ACE_TYPE
61 #define MAX_ACE_TYPE ACE_SYSTEM_ALARM_CALLBACK_OBJECT_ACE_TYPE
62 #define MIN_ACE_TYPE ALLOW
64 #define OWNING_GROUP (ACE_GROUP|ACE_IDENTIFIER_GROUP)
65 #define EVERYONE_ALLOW_MASK (ACE_READ_ACL|ACE_READ_ATTRIBUTES | \
66 ACE_READ_NAMED_ATTRS|ACE_SYNCHRONIZE)
67 #define EVERYONE_DENY_MASK (ACE_WRITE_ACL|ACE_WRITE_OWNER | \
68 ACE_WRITE_ATTRIBUTES|ACE_WRITE_NAMED_ATTRS)
69 #define OWNER_ALLOW_MASK (ACE_WRITE_ACL | ACE_WRITE_OWNER | \
70 ACE_WRITE_ATTRIBUTES|ACE_WRITE_NAMED_ATTRS)
72 #define ZFS_CHECKED_MASKS (ACE_READ_ACL|ACE_READ_ATTRIBUTES|ACE_READ_DATA| \
73 ACE_READ_NAMED_ATTRS|ACE_WRITE_DATA|ACE_WRITE_ATTRIBUTES| \
74 ACE_WRITE_NAMED_ATTRS|ACE_APPEND_DATA|ACE_EXECUTE|ACE_WRITE_OWNER| \
75 ACE_WRITE_ACL|ACE_DELETE|ACE_DELETE_CHILD|ACE_SYNCHRONIZE)
77 #define WRITE_MASK_DATA (ACE_WRITE_DATA|ACE_APPEND_DATA|ACE_WRITE_NAMED_ATTRS)
78 #define WRITE_MASK_ATTRS (ACE_WRITE_ACL|ACE_WRITE_OWNER|ACE_WRITE_ATTRIBUTES| \
79 ACE_DELETE|ACE_DELETE_CHILD)
80 #define WRITE_MASK (WRITE_MASK_DATA|WRITE_MASK_ATTRS)
82 #define OGE_CLEAR (ACE_READ_DATA|ACE_LIST_DIRECTORY|ACE_WRITE_DATA| \
83 ACE_ADD_FILE|ACE_APPEND_DATA|ACE_ADD_SUBDIRECTORY|ACE_EXECUTE)
85 #define OKAY_MASK_BITS (ACE_READ_DATA|ACE_LIST_DIRECTORY|ACE_WRITE_DATA| \
86 ACE_ADD_FILE|ACE_APPEND_DATA|ACE_ADD_SUBDIRECTORY|ACE_EXECUTE)
88 #define ALL_INHERIT (ACE_FILE_INHERIT_ACE|ACE_DIRECTORY_INHERIT_ACE | \
89 ACE_NO_PROPAGATE_INHERIT_ACE|ACE_INHERIT_ONLY_ACE|ACE_INHERITED_ACE)
91 #define RESTRICTED_CLEAR (ACE_WRITE_ACL|ACE_WRITE_OWNER)
93 #define V4_ACL_WIDE_FLAGS (ZFS_ACL_AUTO_INHERIT|ZFS_ACL_DEFAULTED|\
96 #define ZFS_ACL_WIDE_FLAGS (V4_ACL_WIDE_FLAGS|ZFS_ACL_TRIVIAL|ZFS_INHERIT_ACE|\
99 #define ALL_MODE_EXECS (S_IXUSR | S_IXGRP | S_IXOTH)
102 zfs_ace_v0_get_type(void *acep
)
104 return (((zfs_oldace_t
*)acep
)->z_type
);
108 zfs_ace_v0_get_flags(void *acep
)
110 return (((zfs_oldace_t
*)acep
)->z_flags
);
114 zfs_ace_v0_get_mask(void *acep
)
116 return (((zfs_oldace_t
*)acep
)->z_access_mask
);
120 zfs_ace_v0_get_who(void *acep
)
122 return (((zfs_oldace_t
*)acep
)->z_fuid
);
126 zfs_ace_v0_set_type(void *acep
, uint16_t type
)
128 ((zfs_oldace_t
*)acep
)->z_type
= type
;
132 zfs_ace_v0_set_flags(void *acep
, uint16_t flags
)
134 ((zfs_oldace_t
*)acep
)->z_flags
= flags
;
138 zfs_ace_v0_set_mask(void *acep
, uint32_t mask
)
140 ((zfs_oldace_t
*)acep
)->z_access_mask
= mask
;
144 zfs_ace_v0_set_who(void *acep
, uint64_t who
)
146 ((zfs_oldace_t
*)acep
)->z_fuid
= who
;
151 zfs_ace_v0_size(void *acep
)
153 return (sizeof (zfs_oldace_t
));
157 zfs_ace_v0_abstract_size(void)
159 return (sizeof (zfs_oldace_t
));
163 zfs_ace_v0_mask_off(void)
165 return (offsetof(zfs_oldace_t
, z_access_mask
));
170 zfs_ace_v0_data(void *acep
, void **datap
)
176 static acl_ops_t zfs_acl_v0_ops
= {
179 zfs_ace_v0_get_flags
,
180 zfs_ace_v0_set_flags
,
186 zfs_ace_v0_abstract_size
,
192 zfs_ace_fuid_get_type(void *acep
)
194 return (((zfs_ace_hdr_t
*)acep
)->z_type
);
198 zfs_ace_fuid_get_flags(void *acep
)
200 return (((zfs_ace_hdr_t
*)acep
)->z_flags
);
204 zfs_ace_fuid_get_mask(void *acep
)
206 return (((zfs_ace_hdr_t
*)acep
)->z_access_mask
);
210 zfs_ace_fuid_get_who(void *args
)
213 zfs_ace_t
*acep
= args
;
215 entry_type
= acep
->z_hdr
.z_flags
& ACE_TYPE_FLAGS
;
217 if (entry_type
== ACE_OWNER
|| entry_type
== OWNING_GROUP
||
218 entry_type
== ACE_EVERYONE
)
220 return (((zfs_ace_t
*)acep
)->z_fuid
);
224 zfs_ace_fuid_set_type(void *acep
, uint16_t type
)
226 ((zfs_ace_hdr_t
*)acep
)->z_type
= type
;
230 zfs_ace_fuid_set_flags(void *acep
, uint16_t flags
)
232 ((zfs_ace_hdr_t
*)acep
)->z_flags
= flags
;
236 zfs_ace_fuid_set_mask(void *acep
, uint32_t mask
)
238 ((zfs_ace_hdr_t
*)acep
)->z_access_mask
= mask
;
242 zfs_ace_fuid_set_who(void *arg
, uint64_t who
)
244 zfs_ace_t
*acep
= arg
;
246 uint16_t entry_type
= acep
->z_hdr
.z_flags
& ACE_TYPE_FLAGS
;
248 if (entry_type
== ACE_OWNER
|| entry_type
== OWNING_GROUP
||
249 entry_type
== ACE_EVERYONE
)
255 zfs_ace_fuid_size(void *acep
)
257 zfs_ace_hdr_t
*zacep
= acep
;
260 switch (zacep
->z_type
) {
261 case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE
:
262 case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE
:
263 case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE
:
264 case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE
:
265 return (sizeof (zfs_object_ace_t
));
269 (((zfs_ace_hdr_t
*)acep
)->z_flags
& ACE_TYPE_FLAGS
);
270 if (entry_type
== ACE_OWNER
||
271 entry_type
== OWNING_GROUP
||
272 entry_type
== ACE_EVERYONE
)
273 return (sizeof (zfs_ace_hdr_t
));
276 return (sizeof (zfs_ace_t
));
281 zfs_ace_fuid_abstract_size(void)
283 return (sizeof (zfs_ace_hdr_t
));
287 zfs_ace_fuid_mask_off(void)
289 return (offsetof(zfs_ace_hdr_t
, z_access_mask
));
293 zfs_ace_fuid_data(void *acep
, void **datap
)
295 zfs_ace_t
*zacep
= acep
;
296 zfs_object_ace_t
*zobjp
;
298 switch (zacep
->z_hdr
.z_type
) {
299 case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE
:
300 case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE
:
301 case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE
:
302 case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE
:
304 *datap
= (caddr_t
)zobjp
+ sizeof (zfs_ace_t
);
305 return (sizeof (zfs_object_ace_t
) - sizeof (zfs_ace_t
));
312 static acl_ops_t zfs_acl_fuid_ops
= {
313 zfs_ace_fuid_get_mask
,
314 zfs_ace_fuid_set_mask
,
315 zfs_ace_fuid_get_flags
,
316 zfs_ace_fuid_set_flags
,
317 zfs_ace_fuid_get_type
,
318 zfs_ace_fuid_set_type
,
319 zfs_ace_fuid_get_who
,
320 zfs_ace_fuid_set_who
,
322 zfs_ace_fuid_abstract_size
,
323 zfs_ace_fuid_mask_off
,
328 * The following three functions are provided for compatibility with
329 * older ZPL version in order to determine if the file use to have
330 * an external ACL and what version of ACL previously existed on the
331 * file. Would really be nice to not need this, sigh.
334 zfs_external_acl(znode_t
*zp
)
336 zfs_acl_phys_t acl_phys
;
343 * Need to deal with a potential
344 * race where zfs_sa_upgrade could cause
345 * z_isa_sa to change.
347 * If the lookup fails then the state of z_is_sa should have
351 if ((error
= sa_lookup(zp
->z_sa_hdl
, SA_ZPL_ZNODE_ACL(ZTOZSB(zp
)),
352 &acl_phys
, sizeof (acl_phys
))) == 0)
353 return (acl_phys
.z_acl_extern_obj
);
356 * after upgrade the SA_ZPL_ZNODE_ACL should have been
359 VERIFY(zp
->z_is_sa
&& error
== ENOENT
);
365 * Determine size of ACL in bytes
367 * This is more complicated than it should be since we have to deal
368 * with old external ACLs.
371 zfs_acl_znode_info(znode_t
*zp
, int *aclsize
, int *aclcount
,
372 zfs_acl_phys_t
*aclphys
)
374 zfs_sb_t
*zsb
= ZTOZSB(zp
);
379 ASSERT(MUTEX_HELD(&zp
->z_acl_lock
));
381 if ((error
= sa_size(zp
->z_sa_hdl
, SA_ZPL_DACL_ACES(zsb
),
385 if ((error
= sa_lookup(zp
->z_sa_hdl
, SA_ZPL_DACL_COUNT(zsb
),
386 &acl_count
, sizeof (acl_count
))) != 0)
388 *aclcount
= acl_count
;
390 if ((error
= sa_lookup(zp
->z_sa_hdl
, SA_ZPL_ZNODE_ACL(zsb
),
391 aclphys
, sizeof (*aclphys
))) != 0)
394 if (aclphys
->z_acl_version
== ZFS_ACL_VERSION_INITIAL
) {
395 *aclsize
= ZFS_ACL_SIZE(aclphys
->z_acl_size
);
396 *aclcount
= aclphys
->z_acl_size
;
398 *aclsize
= aclphys
->z_acl_size
;
399 *aclcount
= aclphys
->z_acl_count
;
406 zfs_znode_acl_version(znode_t
*zp
)
408 zfs_acl_phys_t acl_phys
;
411 return (ZFS_ACL_VERSION_FUID
);
416 * Need to deal with a potential
417 * race where zfs_sa_upgrade could cause
418 * z_isa_sa to change.
420 * If the lookup fails then the state of z_is_sa should have
423 if ((error
= sa_lookup(zp
->z_sa_hdl
,
424 SA_ZPL_ZNODE_ACL(ZTOZSB(zp
)),
425 &acl_phys
, sizeof (acl_phys
))) == 0)
426 return (acl_phys
.z_acl_version
);
429 * After upgrade SA_ZPL_ZNODE_ACL should have
432 VERIFY(zp
->z_is_sa
&& error
== ENOENT
);
433 return (ZFS_ACL_VERSION_FUID
);
439 zfs_acl_version(int version
)
441 if (version
< ZPL_VERSION_FUID
)
442 return (ZFS_ACL_VERSION_INITIAL
);
444 return (ZFS_ACL_VERSION_FUID
);
448 zfs_acl_version_zp(znode_t
*zp
)
450 return (zfs_acl_version(ZTOZSB(zp
)->z_version
));
454 zfs_acl_alloc(int vers
)
458 aclp
= kmem_zalloc(sizeof (zfs_acl_t
), KM_SLEEP
);
459 list_create(&aclp
->z_acl
, sizeof (zfs_acl_node_t
),
460 offsetof(zfs_acl_node_t
, z_next
));
461 aclp
->z_version
= vers
;
462 if (vers
== ZFS_ACL_VERSION_FUID
)
463 aclp
->z_ops
= &zfs_acl_fuid_ops
;
465 aclp
->z_ops
= &zfs_acl_v0_ops
;
470 zfs_acl_node_alloc(size_t bytes
)
472 zfs_acl_node_t
*aclnode
;
474 aclnode
= kmem_zalloc(sizeof (zfs_acl_node_t
), KM_SLEEP
);
476 aclnode
->z_acldata
= kmem_alloc(bytes
, KM_SLEEP
);
477 aclnode
->z_allocdata
= aclnode
->z_acldata
;
478 aclnode
->z_allocsize
= bytes
;
479 aclnode
->z_size
= bytes
;
486 zfs_acl_node_free(zfs_acl_node_t
*aclnode
)
488 if (aclnode
->z_allocsize
)
489 kmem_free(aclnode
->z_allocdata
, aclnode
->z_allocsize
);
490 kmem_free(aclnode
, sizeof (zfs_acl_node_t
));
494 zfs_acl_release_nodes(zfs_acl_t
*aclp
)
496 zfs_acl_node_t
*aclnode
;
498 while ((aclnode
= list_head(&aclp
->z_acl
))) {
499 list_remove(&aclp
->z_acl
, aclnode
);
500 zfs_acl_node_free(aclnode
);
502 aclp
->z_acl_count
= 0;
503 aclp
->z_acl_bytes
= 0;
507 zfs_acl_free(zfs_acl_t
*aclp
)
509 zfs_acl_release_nodes(aclp
);
510 list_destroy(&aclp
->z_acl
);
511 kmem_free(aclp
, sizeof (zfs_acl_t
));
515 zfs_acl_valid_ace_type(uint_t type
, uint_t flags
)
522 case ACE_SYSTEM_AUDIT_ACE_TYPE
:
523 case ACE_SYSTEM_ALARM_ACE_TYPE
:
524 entry_type
= flags
& ACE_TYPE_FLAGS
;
525 return (entry_type
== ACE_OWNER
||
526 entry_type
== OWNING_GROUP
||
527 entry_type
== ACE_EVERYONE
|| entry_type
== 0 ||
528 entry_type
== ACE_IDENTIFIER_GROUP
);
530 if (type
>= MIN_ACE_TYPE
&& type
<= MAX_ACE_TYPE
)
537 zfs_ace_valid(umode_t obj_mode
, zfs_acl_t
*aclp
, uint16_t type
, uint16_t iflags
)
540 * first check type of entry
543 if (!zfs_acl_valid_ace_type(type
, iflags
))
547 case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE
:
548 case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE
:
549 case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE
:
550 case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE
:
551 if (aclp
->z_version
< ZFS_ACL_VERSION_FUID
)
553 aclp
->z_hints
|= ZFS_ACL_OBJ_ACE
;
557 * next check inheritance level flags
560 if (S_ISDIR(obj_mode
) &&
561 (iflags
& (ACE_FILE_INHERIT_ACE
|ACE_DIRECTORY_INHERIT_ACE
)))
562 aclp
->z_hints
|= ZFS_INHERIT_ACE
;
564 if (iflags
& (ACE_INHERIT_ONLY_ACE
|ACE_NO_PROPAGATE_INHERIT_ACE
)) {
565 if ((iflags
& (ACE_FILE_INHERIT_ACE
|
566 ACE_DIRECTORY_INHERIT_ACE
)) == 0) {
575 zfs_acl_next_ace(zfs_acl_t
*aclp
, void *start
, uint64_t *who
,
576 uint32_t *access_mask
, uint16_t *iflags
, uint16_t *type
)
578 zfs_acl_node_t
*aclnode
;
583 aclnode
= list_head(&aclp
->z_acl
);
587 aclp
->z_next_ace
= aclnode
->z_acldata
;
588 aclp
->z_curr_node
= aclnode
;
589 aclnode
->z_ace_idx
= 0;
592 aclnode
= aclp
->z_curr_node
;
597 if (aclnode
->z_ace_idx
>= aclnode
->z_ace_count
) {
598 aclnode
= list_next(&aclp
->z_acl
, aclnode
);
602 aclp
->z_curr_node
= aclnode
;
603 aclnode
->z_ace_idx
= 0;
604 aclp
->z_next_ace
= aclnode
->z_acldata
;
608 if (aclnode
->z_ace_idx
< aclnode
->z_ace_count
) {
609 void *acep
= aclp
->z_next_ace
;
613 * Make sure we don't overstep our bounds
615 ace_size
= aclp
->z_ops
->ace_size(acep
);
617 if (((caddr_t
)acep
+ ace_size
) >
618 ((caddr_t
)aclnode
->z_acldata
+ aclnode
->z_size
)) {
622 *iflags
= aclp
->z_ops
->ace_flags_get(acep
);
623 *type
= aclp
->z_ops
->ace_type_get(acep
);
624 *access_mask
= aclp
->z_ops
->ace_mask_get(acep
);
625 *who
= aclp
->z_ops
->ace_who_get(acep
);
626 aclp
->z_next_ace
= (caddr_t
)aclp
->z_next_ace
+ ace_size
;
627 aclnode
->z_ace_idx
++;
629 return ((void *)acep
);
636 zfs_ace_walk(void *datap
, uint64_t cookie
, int aclcnt
,
637 uint16_t *flags
, uint16_t *type
, uint32_t *mask
)
639 zfs_acl_t
*aclp
= datap
;
640 zfs_ace_hdr_t
*acep
= (zfs_ace_hdr_t
*)(uintptr_t)cookie
;
643 acep
= zfs_acl_next_ace(aclp
, acep
, &who
, mask
,
645 return ((uint64_t)(uintptr_t)acep
);
649 * Copy ACE to internal ZFS format.
650 * While processing the ACL each ACE will be validated for correctness.
651 * ACE FUIDs will be created later.
654 zfs_copy_ace_2_fuid(zfs_sb_t
*zsb
, umode_t obj_mode
, zfs_acl_t
*aclp
,
655 void *datap
, zfs_ace_t
*z_acl
, uint64_t aclcnt
, size_t *size
,
656 zfs_fuid_info_t
**fuidp
, cred_t
*cr
)
660 zfs_ace_t
*aceptr
= z_acl
;
662 zfs_object_ace_t
*zobjacep
;
663 ace_object_t
*aceobjp
;
665 for (i
= 0; i
!= aclcnt
; i
++) {
666 aceptr
->z_hdr
.z_access_mask
= acep
->a_access_mask
;
667 aceptr
->z_hdr
.z_flags
= acep
->a_flags
;
668 aceptr
->z_hdr
.z_type
= acep
->a_type
;
669 entry_type
= aceptr
->z_hdr
.z_flags
& ACE_TYPE_FLAGS
;
670 if (entry_type
!= ACE_OWNER
&& entry_type
!= OWNING_GROUP
&&
671 entry_type
!= ACE_EVERYONE
) {
672 aceptr
->z_fuid
= zfs_fuid_create(zsb
, acep
->a_who
,
673 cr
, (entry_type
== 0) ?
674 ZFS_ACE_USER
: ZFS_ACE_GROUP
, fuidp
);
678 * Make sure ACE is valid
680 if (zfs_ace_valid(obj_mode
, aclp
, aceptr
->z_hdr
.z_type
,
681 aceptr
->z_hdr
.z_flags
) != B_TRUE
)
682 return (SET_ERROR(EINVAL
));
684 switch (acep
->a_type
) {
685 case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE
:
686 case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE
:
687 case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE
:
688 case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE
:
689 zobjacep
= (zfs_object_ace_t
*)aceptr
;
690 aceobjp
= (ace_object_t
*)acep
;
692 bcopy(aceobjp
->a_obj_type
, zobjacep
->z_object_type
,
693 sizeof (aceobjp
->a_obj_type
));
694 bcopy(aceobjp
->a_inherit_obj_type
,
695 zobjacep
->z_inherit_type
,
696 sizeof (aceobjp
->a_inherit_obj_type
));
697 acep
= (ace_t
*)((caddr_t
)acep
+ sizeof (ace_object_t
));
700 acep
= (ace_t
*)((caddr_t
)acep
+ sizeof (ace_t
));
703 aceptr
= (zfs_ace_t
*)((caddr_t
)aceptr
+
704 aclp
->z_ops
->ace_size(aceptr
));
707 *size
= (caddr_t
)aceptr
- (caddr_t
)z_acl
;
713 * Copy ZFS ACEs to fixed size ace_t layout
716 zfs_copy_fuid_2_ace(zfs_sb_t
*zsb
, zfs_acl_t
*aclp
, cred_t
*cr
,
717 void *datap
, int filter
)
720 uint32_t access_mask
;
721 uint16_t iflags
, type
;
722 zfs_ace_hdr_t
*zacep
= NULL
;
724 ace_object_t
*objacep
;
725 zfs_object_ace_t
*zobjacep
;
729 while ((zacep
= zfs_acl_next_ace(aclp
, zacep
,
730 &who
, &access_mask
, &iflags
, &type
))) {
733 case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE
:
734 case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE
:
735 case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE
:
736 case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE
:
740 zobjacep
= (zfs_object_ace_t
*)zacep
;
741 objacep
= (ace_object_t
*)acep
;
742 bcopy(zobjacep
->z_object_type
,
744 sizeof (zobjacep
->z_object_type
));
745 bcopy(zobjacep
->z_inherit_type
,
746 objacep
->a_inherit_obj_type
,
747 sizeof (zobjacep
->z_inherit_type
));
748 ace_size
= sizeof (ace_object_t
);
751 ace_size
= sizeof (ace_t
);
755 entry_type
= (iflags
& ACE_TYPE_FLAGS
);
756 if ((entry_type
!= ACE_OWNER
&&
757 entry_type
!= OWNING_GROUP
&&
758 entry_type
!= ACE_EVERYONE
)) {
759 acep
->a_who
= zfs_fuid_map_id(zsb
, who
,
760 cr
, (entry_type
& ACE_IDENTIFIER_GROUP
) ?
761 ZFS_ACE_GROUP
: ZFS_ACE_USER
);
763 acep
->a_who
= (uid_t
)(int64_t)who
;
765 acep
->a_access_mask
= access_mask
;
766 acep
->a_flags
= iflags
;
768 acep
= (ace_t
*)((caddr_t
)acep
+ ace_size
);
773 zfs_copy_ace_2_oldace(umode_t obj_mode
, zfs_acl_t
*aclp
, ace_t
*acep
,
774 zfs_oldace_t
*z_acl
, int aclcnt
, size_t *size
)
777 zfs_oldace_t
*aceptr
= z_acl
;
779 for (i
= 0; i
!= aclcnt
; i
++, aceptr
++) {
780 aceptr
->z_access_mask
= acep
[i
].a_access_mask
;
781 aceptr
->z_type
= acep
[i
].a_type
;
782 aceptr
->z_flags
= acep
[i
].a_flags
;
783 aceptr
->z_fuid
= acep
[i
].a_who
;
785 * Make sure ACE is valid
787 if (zfs_ace_valid(obj_mode
, aclp
, aceptr
->z_type
,
788 aceptr
->z_flags
) != B_TRUE
)
789 return (SET_ERROR(EINVAL
));
791 *size
= (caddr_t
)aceptr
- (caddr_t
)z_acl
;
796 * convert old ACL format to new
799 zfs_acl_xform(znode_t
*zp
, zfs_acl_t
*aclp
, cred_t
*cr
)
801 zfs_oldace_t
*oldaclp
;
803 uint16_t type
, iflags
;
804 uint32_t access_mask
;
807 zfs_acl_node_t
*newaclnode
;
809 ASSERT(aclp
->z_version
== ZFS_ACL_VERSION_INITIAL
);
811 * First create the ACE in a contiguous piece of memory
812 * for zfs_copy_ace_2_fuid().
814 * We only convert an ACL once, so this won't happen
817 oldaclp
= kmem_alloc(sizeof (zfs_oldace_t
) * aclp
->z_acl_count
,
820 while ((cookie
= zfs_acl_next_ace(aclp
, cookie
, &who
,
821 &access_mask
, &iflags
, &type
))) {
822 oldaclp
[i
].z_flags
= iflags
;
823 oldaclp
[i
].z_type
= type
;
824 oldaclp
[i
].z_fuid
= who
;
825 oldaclp
[i
++].z_access_mask
= access_mask
;
828 newaclnode
= zfs_acl_node_alloc(aclp
->z_acl_count
*
829 sizeof (zfs_object_ace_t
));
830 aclp
->z_ops
= &zfs_acl_fuid_ops
;
831 VERIFY(zfs_copy_ace_2_fuid(ZTOZSB(zp
), ZTOI(zp
)->i_mode
,
832 aclp
, oldaclp
, newaclnode
->z_acldata
, aclp
->z_acl_count
,
833 &newaclnode
->z_size
, NULL
, cr
) == 0);
834 newaclnode
->z_ace_count
= aclp
->z_acl_count
;
835 aclp
->z_version
= ZFS_ACL_VERSION
;
836 kmem_free(oldaclp
, aclp
->z_acl_count
* sizeof (zfs_oldace_t
));
839 * Release all previous ACL nodes
842 zfs_acl_release_nodes(aclp
);
844 list_insert_head(&aclp
->z_acl
, newaclnode
);
846 aclp
->z_acl_bytes
= newaclnode
->z_size
;
847 aclp
->z_acl_count
= newaclnode
->z_ace_count
;
852 * Convert unix access mask to v4 access mask
855 zfs_unix_to_v4(uint32_t access_mask
)
857 uint32_t new_mask
= 0;
859 if (access_mask
& S_IXOTH
)
860 new_mask
|= ACE_EXECUTE
;
861 if (access_mask
& S_IWOTH
)
862 new_mask
|= ACE_WRITE_DATA
;
863 if (access_mask
& S_IROTH
)
864 new_mask
|= ACE_READ_DATA
;
869 zfs_set_ace(zfs_acl_t
*aclp
, void *acep
, uint32_t access_mask
,
870 uint16_t access_type
, uint64_t fuid
, uint16_t entry_type
)
872 uint16_t type
= entry_type
& ACE_TYPE_FLAGS
;
874 aclp
->z_ops
->ace_mask_set(acep
, access_mask
);
875 aclp
->z_ops
->ace_type_set(acep
, access_type
);
876 aclp
->z_ops
->ace_flags_set(acep
, entry_type
);
877 if ((type
!= ACE_OWNER
&& type
!= OWNING_GROUP
&&
878 type
!= ACE_EVERYONE
))
879 aclp
->z_ops
->ace_who_set(acep
, fuid
);
883 * Determine mode of file based on ACL.
884 * Also, create FUIDs for any User/Group ACEs
887 zfs_mode_compute(uint64_t fmode
, zfs_acl_t
*aclp
,
888 uint64_t *pflags
, uint64_t fuid
, uint64_t fgid
)
893 zfs_ace_hdr_t
*acep
= NULL
;
895 uint16_t iflags
, type
;
896 uint32_t access_mask
;
897 boolean_t an_exec_denied
= B_FALSE
;
899 mode
= (fmode
& (S_IFMT
| S_ISUID
| S_ISGID
| S_ISVTX
));
901 while ((acep
= zfs_acl_next_ace(aclp
, acep
, &who
,
902 &access_mask
, &iflags
, &type
))) {
904 if (!zfs_acl_valid_ace_type(type
, iflags
))
907 entry_type
= (iflags
& ACE_TYPE_FLAGS
);
910 * Skip over owner@, group@ or everyone@ inherit only ACEs
912 if ((iflags
& ACE_INHERIT_ONLY_ACE
) &&
913 (entry_type
== ACE_OWNER
|| entry_type
== ACE_EVERYONE
||
914 entry_type
== OWNING_GROUP
))
917 if (entry_type
== ACE_OWNER
|| (entry_type
== 0 &&
919 if ((access_mask
& ACE_READ_DATA
) &&
920 (!(seen
& S_IRUSR
))) {
926 if ((access_mask
& ACE_WRITE_DATA
) &&
927 (!(seen
& S_IWUSR
))) {
933 if ((access_mask
& ACE_EXECUTE
) &&
934 (!(seen
& S_IXUSR
))) {
940 } else if (entry_type
== OWNING_GROUP
||
941 (entry_type
== ACE_IDENTIFIER_GROUP
&& who
== fgid
)) {
942 if ((access_mask
& ACE_READ_DATA
) &&
943 (!(seen
& S_IRGRP
))) {
949 if ((access_mask
& ACE_WRITE_DATA
) &&
950 (!(seen
& S_IWGRP
))) {
956 if ((access_mask
& ACE_EXECUTE
) &&
957 (!(seen
& S_IXGRP
))) {
963 } else if (entry_type
== ACE_EVERYONE
) {
964 if ((access_mask
& ACE_READ_DATA
)) {
965 if (!(seen
& S_IRUSR
)) {
971 if (!(seen
& S_IRGRP
)) {
977 if (!(seen
& S_IROTH
)) {
984 if ((access_mask
& ACE_WRITE_DATA
)) {
985 if (!(seen
& S_IWUSR
)) {
991 if (!(seen
& S_IWGRP
)) {
997 if (!(seen
& S_IWOTH
)) {
1004 if ((access_mask
& ACE_EXECUTE
)) {
1005 if (!(seen
& S_IXUSR
)) {
1007 if (type
== ALLOW
) {
1011 if (!(seen
& S_IXGRP
)) {
1013 if (type
== ALLOW
) {
1017 if (!(seen
& S_IXOTH
)) {
1019 if (type
== ALLOW
) {
1026 * Only care if this IDENTIFIER_GROUP or
1027 * USER ACE denies execute access to someone,
1028 * mode is not affected
1030 if ((access_mask
& ACE_EXECUTE
) && type
== DENY
)
1031 an_exec_denied
= B_TRUE
;
1036 * Failure to allow is effectively a deny, so execute permission
1037 * is denied if it was never mentioned or if we explicitly
1038 * weren't allowed it.
1040 if (!an_exec_denied
&&
1041 ((seen
& ALL_MODE_EXECS
) != ALL_MODE_EXECS
||
1042 (mode
& ALL_MODE_EXECS
) != ALL_MODE_EXECS
))
1043 an_exec_denied
= B_TRUE
;
1046 *pflags
&= ~ZFS_NO_EXECS_DENIED
;
1048 *pflags
|= ZFS_NO_EXECS_DENIED
;
1054 * Read an external acl object. If the intent is to modify, always
1055 * create a new acl and leave any cached acl in place.
1058 zfs_acl_node_read(znode_t
*zp
, boolean_t have_lock
, zfs_acl_t
**aclpp
,
1059 boolean_t will_modify
)
1064 zfs_acl_node_t
*aclnode
;
1065 zfs_acl_phys_t znode_acl
;
1068 boolean_t drop_lock
= B_FALSE
;
1070 ASSERT(MUTEX_HELD(&zp
->z_acl_lock
));
1072 if (zp
->z_acl_cached
&& !will_modify
) {
1073 *aclpp
= zp
->z_acl_cached
;
1078 * close race where znode could be upgrade while trying to
1079 * read the znode attributes.
1081 * But this could only happen if the file isn't already an SA
1084 if (!zp
->z_is_sa
&& !have_lock
) {
1085 mutex_enter(&zp
->z_lock
);
1088 version
= zfs_znode_acl_version(zp
);
1090 if ((error
= zfs_acl_znode_info(zp
, &aclsize
,
1091 &acl_count
, &znode_acl
)) != 0) {
1095 aclp
= zfs_acl_alloc(version
);
1097 aclp
->z_acl_count
= acl_count
;
1098 aclp
->z_acl_bytes
= aclsize
;
1100 aclnode
= zfs_acl_node_alloc(aclsize
);
1101 aclnode
->z_ace_count
= aclp
->z_acl_count
;
1102 aclnode
->z_size
= aclsize
;
1105 if (znode_acl
.z_acl_extern_obj
) {
1106 error
= dmu_read(ZTOZSB(zp
)->z_os
,
1107 znode_acl
.z_acl_extern_obj
, 0, aclnode
->z_size
,
1108 aclnode
->z_acldata
, DMU_READ_PREFETCH
);
1110 bcopy(znode_acl
.z_ace_data
, aclnode
->z_acldata
,
1114 error
= sa_lookup(zp
->z_sa_hdl
, SA_ZPL_DACL_ACES(ZTOZSB(zp
)),
1115 aclnode
->z_acldata
, aclnode
->z_size
);
1120 zfs_acl_node_free(aclnode
);
1121 /* convert checksum errors into IO errors */
1122 if (error
== ECKSUM
)
1123 error
= SET_ERROR(EIO
);
1127 list_insert_head(&aclp
->z_acl
, aclnode
);
1131 zp
->z_acl_cached
= aclp
;
1134 mutex_exit(&zp
->z_lock
);
1140 zfs_acl_data_locator(void **dataptr
, uint32_t *length
, uint32_t buflen
,
1141 boolean_t start
, void *userdata
)
1143 zfs_acl_locator_cb_t
*cb
= (zfs_acl_locator_cb_t
*)userdata
;
1146 cb
->cb_acl_node
= list_head(&cb
->cb_aclp
->z_acl
);
1148 cb
->cb_acl_node
= list_next(&cb
->cb_aclp
->z_acl
,
1151 *dataptr
= cb
->cb_acl_node
->z_acldata
;
1152 *length
= cb
->cb_acl_node
->z_size
;
1156 zfs_acl_chown_setattr(znode_t
*zp
)
1161 if (ZTOZSB(zp
)->z_acl_type
== ZFS_ACLTYPE_POSIXACL
)
1164 ASSERT(MUTEX_HELD(&zp
->z_lock
));
1165 ASSERT(MUTEX_HELD(&zp
->z_acl_lock
));
1167 error
= zfs_acl_node_read(zp
, B_TRUE
, &aclp
, B_FALSE
);
1168 if (error
== 0 && aclp
->z_acl_count
> 0)
1169 zp
->z_mode
= ZTOI(zp
)->i_mode
=
1170 zfs_mode_compute(zp
->z_mode
, aclp
,
1171 &zp
->z_pflags
, KUID_TO_SUID(ZTOI(zp
)->i_uid
),
1172 KGID_TO_SGID(ZTOI(zp
)->i_gid
));
1175 * Some ZFS implementations (ZEVO) create neither a ZNODE_ACL
1176 * nor a DACL_ACES SA in which case ENOENT is returned from
1177 * zfs_acl_node_read() when the SA can't be located.
1178 * Allow chown/chgrp to succeed in these cases rather than
1179 * returning an error that makes no sense in the context of
1182 if (error
== ENOENT
)
1189 acl_trivial_access_masks(mode_t mode
, uint32_t *allow0
, uint32_t *deny1
,
1190 uint32_t *deny2
, uint32_t *owner
, uint32_t *group
, uint32_t *everyone
)
1192 *deny1
= *deny2
= *allow0
= *group
= 0;
1194 if (!(mode
& S_IRUSR
) && (mode
& (S_IRGRP
|S_IROTH
)))
1195 *deny1
|= ACE_READ_DATA
;
1196 if (!(mode
& S_IWUSR
) && (mode
& (S_IWGRP
|S_IWOTH
)))
1197 *deny1
|= ACE_WRITE_DATA
;
1198 if (!(mode
& S_IXUSR
) && (mode
& (S_IXGRP
|S_IXOTH
)))
1199 *deny1
|= ACE_EXECUTE
;
1201 if (!(mode
& S_IRGRP
) && (mode
& S_IROTH
))
1202 *deny2
= ACE_READ_DATA
;
1203 if (!(mode
& S_IWGRP
) && (mode
& S_IWOTH
))
1204 *deny2
|= ACE_WRITE_DATA
;
1205 if (!(mode
& S_IXGRP
) && (mode
& S_IXOTH
))
1206 *deny2
|= ACE_EXECUTE
;
1208 if ((mode
& S_IRUSR
) && (!(mode
& S_IRGRP
) && (mode
& S_IROTH
)))
1209 *allow0
|= ACE_READ_DATA
;
1210 if ((mode
& S_IWUSR
) && (!(mode
& S_IWGRP
) && (mode
& S_IWOTH
)))
1211 *allow0
|= ACE_WRITE_DATA
;
1212 if ((mode
& S_IXUSR
) && (!(mode
& S_IXGRP
) && (mode
& S_IXOTH
)))
1213 *allow0
|= ACE_EXECUTE
;
1215 *owner
= ACE_WRITE_ATTRIBUTES
|ACE_WRITE_OWNER
|ACE_WRITE_ACL
|
1216 ACE_WRITE_NAMED_ATTRS
|ACE_READ_ACL
|ACE_READ_ATTRIBUTES
|
1217 ACE_READ_NAMED_ATTRS
|ACE_SYNCHRONIZE
;
1219 *owner
|= ACE_READ_DATA
;
1221 *owner
|= ACE_WRITE_DATA
|ACE_APPEND_DATA
;
1223 *owner
|= ACE_EXECUTE
;
1225 *group
= ACE_READ_ACL
|ACE_READ_ATTRIBUTES
| ACE_READ_NAMED_ATTRS
|
1228 *group
|= ACE_READ_DATA
;
1230 *group
|= ACE_WRITE_DATA
|ACE_APPEND_DATA
;
1232 *group
|= ACE_EXECUTE
;
1234 *everyone
= ACE_READ_ACL
|ACE_READ_ATTRIBUTES
| ACE_READ_NAMED_ATTRS
|
1237 *everyone
|= ACE_READ_DATA
;
1239 *everyone
|= ACE_WRITE_DATA
|ACE_APPEND_DATA
;
1241 *everyone
|= ACE_EXECUTE
;
1246 * determine whether an ace_t acl is trivial
1248 * Trivialness implies that the acl is composed of only
1249 * owner, group, everyone entries. ACL can't
1250 * have read_acl denied, and write_owner/write_acl/write_attributes
1251 * can only be owner@ entry.
1254 ace_trivial_common(void *acep
, int aclcnt
,
1255 uint64_t (*walk
)(void *, uint64_t, int aclcnt
,
1256 uint16_t *, uint16_t *, uint32_t *))
1261 uint64_t cookie
= 0;
1263 while ((cookie
= walk(acep
, cookie
, aclcnt
, &flags
, &type
, &mask
))) {
1264 switch (flags
& ACE_TYPE_FLAGS
) {
1266 case ACE_GROUP
|ACE_IDENTIFIER_GROUP
:
1273 if (flags
& (ACE_FILE_INHERIT_ACE
|
1274 ACE_DIRECTORY_INHERIT_ACE
|ACE_NO_PROPAGATE_INHERIT_ACE
|
1275 ACE_INHERIT_ONLY_ACE
))
1279 * Special check for some special bits
1281 * Don't allow anybody to deny reading basic
1282 * attributes or a files ACL.
1284 if ((mask
& (ACE_READ_ACL
|ACE_READ_ATTRIBUTES
)) &&
1285 (type
== ACE_ACCESS_DENIED_ACE_TYPE
))
1289 * Delete permissions are never set by default
1291 if (mask
& (ACE_DELETE
|ACE_DELETE_CHILD
))
1294 * only allow owner@ to have
1295 * write_acl/write_owner/write_attributes/write_xattr/
1297 if (type
== ACE_ACCESS_ALLOWED_ACE_TYPE
&&
1298 (!(flags
& ACE_OWNER
) && (mask
&
1299 (ACE_WRITE_OWNER
|ACE_WRITE_ACL
| ACE_WRITE_ATTRIBUTES
|
1300 ACE_WRITE_NAMED_ATTRS
))))
1309 * common code for setting ACLs.
1311 * This function is called from zfs_mode_update, zfs_perm_init, and zfs_setacl.
1312 * zfs_setacl passes a non-NULL inherit pointer (ihp) to indicate that it's
1313 * already checked the acl and knows whether to inherit.
1316 zfs_aclset_common(znode_t
*zp
, zfs_acl_t
*aclp
, cred_t
*cr
, dmu_tx_t
*tx
)
1319 zfs_sb_t
*zsb
= ZTOZSB(zp
);
1320 dmu_object_type_t otype
;
1321 zfs_acl_locator_cb_t locate
= { 0 };
1323 sa_bulk_attr_t bulk
[5];
1329 mode
= zfs_mode_compute(mode
, aclp
, &zp
->z_pflags
,
1330 KUID_TO_SUID(ZTOI(zp
)->i_uid
), KGID_TO_SGID(ZTOI(zp
)->i_gid
));
1332 zp
->z_mode
= ZTOI(zp
)->i_mode
= mode
;
1333 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_MODE(zsb
), NULL
,
1334 &mode
, sizeof (mode
));
1335 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_FLAGS(zsb
), NULL
,
1336 &zp
->z_pflags
, sizeof (zp
->z_pflags
));
1337 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_CTIME(zsb
), NULL
,
1338 &ctime
, sizeof (ctime
));
1340 if (zp
->z_acl_cached
) {
1341 zfs_acl_free(zp
->z_acl_cached
);
1342 zp
->z_acl_cached
= NULL
;
1348 if (!zsb
->z_use_fuids
) {
1349 otype
= DMU_OT_OLDACL
;
1351 if ((aclp
->z_version
== ZFS_ACL_VERSION_INITIAL
) &&
1352 (zsb
->z_version
>= ZPL_VERSION_FUID
))
1353 zfs_acl_xform(zp
, aclp
, cr
);
1354 ASSERT(aclp
->z_version
>= ZFS_ACL_VERSION_FUID
);
1359 * Arrgh, we have to handle old on disk format
1360 * as well as newer (preferred) SA format.
1363 if (zp
->z_is_sa
) { /* the easy case, just update the ACL attribute */
1364 locate
.cb_aclp
= aclp
;
1365 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_DACL_ACES(zsb
),
1366 zfs_acl_data_locator
, &locate
, aclp
->z_acl_bytes
);
1367 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_DACL_COUNT(zsb
),
1368 NULL
, &aclp
->z_acl_count
, sizeof (uint64_t));
1369 } else { /* Painful legacy way */
1370 zfs_acl_node_t
*aclnode
;
1372 zfs_acl_phys_t acl_phys
;
1375 if ((error
= sa_lookup(zp
->z_sa_hdl
, SA_ZPL_ZNODE_ACL(zsb
),
1376 &acl_phys
, sizeof (acl_phys
))) != 0)
1379 aoid
= acl_phys
.z_acl_extern_obj
;
1381 if (aclp
->z_acl_bytes
> ZFS_ACE_SPACE
) {
1383 * If ACL was previously external and we are now
1384 * converting to new ACL format then release old
1385 * ACL object and create a new one.
1388 aclp
->z_version
!= acl_phys
.z_acl_version
) {
1389 error
= dmu_object_free(zsb
->z_os
, aoid
, tx
);
1395 aoid
= dmu_object_alloc(zsb
->z_os
,
1396 otype
, aclp
->z_acl_bytes
,
1397 otype
== DMU_OT_ACL
?
1398 DMU_OT_SYSACL
: DMU_OT_NONE
,
1399 otype
== DMU_OT_ACL
?
1400 DN_OLD_MAX_BONUSLEN
: 0, tx
);
1402 (void) dmu_object_set_blocksize(zsb
->z_os
,
1403 aoid
, aclp
->z_acl_bytes
, 0, tx
);
1405 acl_phys
.z_acl_extern_obj
= aoid
;
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 dmu_write(zsb
->z_os
, aoid
, off
,
1411 aclnode
->z_size
, aclnode
->z_acldata
, tx
);
1412 off
+= aclnode
->z_size
;
1415 void *start
= acl_phys
.z_ace_data
;
1417 * Migrating back embedded?
1419 if (acl_phys
.z_acl_extern_obj
) {
1420 error
= dmu_object_free(zsb
->z_os
,
1421 acl_phys
.z_acl_extern_obj
, tx
);
1424 acl_phys
.z_acl_extern_obj
= 0;
1427 for (aclnode
= list_head(&aclp
->z_acl
); aclnode
;
1428 aclnode
= list_next(&aclp
->z_acl
, aclnode
)) {
1429 if (aclnode
->z_ace_count
== 0)
1431 bcopy(aclnode
->z_acldata
, start
,
1433 start
= (caddr_t
)start
+ aclnode
->z_size
;
1437 * If Old version then swap count/bytes to match old
1438 * layout of znode_acl_phys_t.
1440 if (aclp
->z_version
== ZFS_ACL_VERSION_INITIAL
) {
1441 acl_phys
.z_acl_size
= aclp
->z_acl_count
;
1442 acl_phys
.z_acl_count
= aclp
->z_acl_bytes
;
1444 acl_phys
.z_acl_size
= aclp
->z_acl_bytes
;
1445 acl_phys
.z_acl_count
= aclp
->z_acl_count
;
1447 acl_phys
.z_acl_version
= aclp
->z_version
;
1449 SA_ADD_BULK_ATTR(bulk
, count
, SA_ZPL_ZNODE_ACL(zsb
), NULL
,
1450 &acl_phys
, sizeof (acl_phys
));
1454 * Replace ACL wide bits, but first clear them.
1456 zp
->z_pflags
&= ~ZFS_ACL_WIDE_FLAGS
;
1458 zp
->z_pflags
|= aclp
->z_hints
;
1460 if (ace_trivial_common(aclp
, 0, zfs_ace_walk
) == 0)
1461 zp
->z_pflags
|= ZFS_ACL_TRIVIAL
;
1463 zfs_tstamp_update_setup(zp
, STATE_CHANGED
, NULL
, ctime
);
1464 return (sa_bulk_update(zp
->z_sa_hdl
, bulk
, count
, tx
));
1468 zfs_acl_chmod(zfs_sb_t
*zsb
, uint64_t mode
, zfs_acl_t
*aclp
)
1472 int new_count
, new_bytes
;
1475 uint16_t iflags
, type
;
1476 uint32_t access_mask
;
1477 zfs_acl_node_t
*newnode
;
1478 size_t abstract_size
= aclp
->z_ops
->ace_abstract_size();
1480 uint32_t owner
, group
, everyone
;
1481 uint32_t deny1
, deny2
, allow0
;
1483 new_count
= new_bytes
= 0;
1485 acl_trivial_access_masks((mode_t
)mode
, &allow0
, &deny1
, &deny2
,
1486 &owner
, &group
, &everyone
);
1488 newnode
= zfs_acl_node_alloc((abstract_size
* 6) + aclp
->z_acl_bytes
);
1490 zacep
= newnode
->z_acldata
;
1492 zfs_set_ace(aclp
, zacep
, allow0
, ALLOW
, -1, ACE_OWNER
);
1493 zacep
= (void *)((uintptr_t)zacep
+ abstract_size
);
1495 new_bytes
+= abstract_size
;
1498 zfs_set_ace(aclp
, zacep
, deny1
, DENY
, -1, ACE_OWNER
);
1499 zacep
= (void *)((uintptr_t)zacep
+ abstract_size
);
1501 new_bytes
+= abstract_size
;
1504 zfs_set_ace(aclp
, zacep
, deny2
, DENY
, -1, OWNING_GROUP
);
1505 zacep
= (void *)((uintptr_t)zacep
+ abstract_size
);
1507 new_bytes
+= abstract_size
;
1510 while ((acep
= zfs_acl_next_ace(aclp
, acep
, &who
, &access_mask
,
1512 uint16_t inherit_flags
;
1514 entry_type
= (iflags
& ACE_TYPE_FLAGS
);
1515 inherit_flags
= (iflags
& ALL_INHERIT
);
1517 if ((entry_type
== ACE_OWNER
|| entry_type
== ACE_EVERYONE
||
1518 (entry_type
== OWNING_GROUP
)) &&
1519 ((inherit_flags
& ACE_INHERIT_ONLY_ACE
) == 0)) {
1523 if ((type
!= ALLOW
&& type
!= DENY
) ||
1524 (inherit_flags
& ACE_INHERIT_ONLY_ACE
)) {
1526 aclp
->z_hints
|= ZFS_INHERIT_ACE
;
1528 case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE
:
1529 case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE
:
1530 case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE
:
1531 case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE
:
1532 aclp
->z_hints
|= ZFS_ACL_OBJ_ACE
;
1538 * Limit permissions to be no greater than
1541 if (zsb
->z_acl_inherit
== ZFS_ACL_RESTRICTED
) {
1542 if (!(mode
& S_IRGRP
))
1543 access_mask
&= ~ACE_READ_DATA
;
1544 if (!(mode
& S_IWGRP
))
1546 ~(ACE_WRITE_DATA
|ACE_APPEND_DATA
);
1547 if (!(mode
& S_IXGRP
))
1548 access_mask
&= ~ACE_EXECUTE
;
1550 ~(ACE_WRITE_OWNER
|ACE_WRITE_ACL
|
1551 ACE_WRITE_ATTRIBUTES
|ACE_WRITE_NAMED_ATTRS
);
1554 zfs_set_ace(aclp
, zacep
, access_mask
, type
, who
, iflags
);
1555 ace_size
= aclp
->z_ops
->ace_size(acep
);
1556 zacep
= (void *)((uintptr_t)zacep
+ ace_size
);
1558 new_bytes
+= ace_size
;
1560 zfs_set_ace(aclp
, zacep
, owner
, 0, -1, ACE_OWNER
);
1561 zacep
= (void *)((uintptr_t)zacep
+ abstract_size
);
1562 zfs_set_ace(aclp
, zacep
, group
, 0, -1, OWNING_GROUP
);
1563 zacep
= (void *)((uintptr_t)zacep
+ abstract_size
);
1564 zfs_set_ace(aclp
, zacep
, everyone
, 0, -1, ACE_EVERYONE
);
1567 new_bytes
+= abstract_size
* 3;
1568 zfs_acl_release_nodes(aclp
);
1569 aclp
->z_acl_count
= new_count
;
1570 aclp
->z_acl_bytes
= new_bytes
;
1571 newnode
->z_ace_count
= new_count
;
1572 newnode
->z_size
= new_bytes
;
1573 list_insert_tail(&aclp
->z_acl
, newnode
);
1577 zfs_acl_chmod_setattr(znode_t
*zp
, zfs_acl_t
**aclp
, uint64_t mode
)
1579 mutex_enter(&zp
->z_acl_lock
);
1580 mutex_enter(&zp
->z_lock
);
1581 *aclp
= zfs_acl_alloc(zfs_acl_version_zp(zp
));
1582 (*aclp
)->z_hints
= zp
->z_pflags
& V4_ACL_WIDE_FLAGS
;
1583 zfs_acl_chmod(ZTOZSB(zp
), mode
, *aclp
);
1584 mutex_exit(&zp
->z_lock
);
1585 mutex_exit(&zp
->z_acl_lock
);
1590 * strip off write_owner and write_acl
1593 zfs_restricted_update(zfs_sb_t
*zsb
, zfs_acl_t
*aclp
, void *acep
)
1595 uint32_t mask
= aclp
->z_ops
->ace_mask_get(acep
);
1597 if ((zsb
->z_acl_inherit
== ZFS_ACL_RESTRICTED
) &&
1598 (aclp
->z_ops
->ace_type_get(acep
) == ALLOW
)) {
1599 mask
&= ~RESTRICTED_CLEAR
;
1600 aclp
->z_ops
->ace_mask_set(acep
, mask
);
1605 * Should ACE be inherited?
1608 zfs_ace_can_use(umode_t obj_mode
, uint16_t acep_flags
)
1610 int iflags
= (acep_flags
& 0xf);
1612 if (S_ISDIR(obj_mode
) && (iflags
& ACE_DIRECTORY_INHERIT_ACE
))
1614 else if (iflags
& ACE_FILE_INHERIT_ACE
)
1615 return (!(S_ISDIR(obj_mode
) &&
1616 (iflags
& ACE_NO_PROPAGATE_INHERIT_ACE
)));
1621 * inherit inheritable ACEs from parent
1624 zfs_acl_inherit(zfs_sb_t
*zsb
, umode_t obj_mode
, zfs_acl_t
*paclp
,
1625 uint64_t mode
, boolean_t
*need_chmod
)
1629 zfs_acl_node_t
*aclnode
;
1630 zfs_acl_t
*aclp
= NULL
;
1632 uint32_t access_mask
;
1633 uint16_t iflags
, newflags
, type
;
1635 void *data1
, *data2
;
1636 size_t data1sz
, data2sz
;
1637 boolean_t vdir
= S_ISDIR(obj_mode
);
1638 boolean_t vreg
= S_ISREG(obj_mode
);
1639 boolean_t passthrough
, passthrough_x
, noallow
;
1642 zsb
->z_acl_inherit
== ZFS_ACL_PASSTHROUGH_X
;
1643 passthrough
= passthrough_x
||
1644 zsb
->z_acl_inherit
== ZFS_ACL_PASSTHROUGH
;
1646 zsb
->z_acl_inherit
== ZFS_ACL_NOALLOW
;
1648 *need_chmod
= B_TRUE
;
1650 aclp
= zfs_acl_alloc(paclp
->z_version
);
1651 if (zsb
->z_acl_inherit
== ZFS_ACL_DISCARD
|| S_ISLNK(obj_mode
))
1653 while ((pacep
= zfs_acl_next_ace(paclp
, pacep
, &who
,
1654 &access_mask
, &iflags
, &type
))) {
1657 * don't inherit bogus ACEs
1659 if (!zfs_acl_valid_ace_type(type
, iflags
))
1662 if (noallow
&& type
== ALLOW
)
1665 ace_size
= aclp
->z_ops
->ace_size(pacep
);
1667 if (!zfs_ace_can_use(obj_mode
, iflags
))
1671 * If owner@, group@, or everyone@ inheritable
1672 * then zfs_acl_chmod() isn't needed.
1675 ((iflags
& (ACE_OWNER
|ACE_EVERYONE
)) ||
1676 ((iflags
& OWNING_GROUP
) ==
1677 OWNING_GROUP
)) && (vreg
|| (vdir
&& (iflags
&
1678 ACE_DIRECTORY_INHERIT_ACE
)))) {
1679 *need_chmod
= B_FALSE
;
1682 if (!vdir
&& passthrough_x
&&
1683 ((mode
& (S_IXUSR
| S_IXGRP
| S_IXOTH
)) == 0)) {
1684 access_mask
&= ~ACE_EXECUTE
;
1687 aclnode
= zfs_acl_node_alloc(ace_size
);
1688 list_insert_tail(&aclp
->z_acl
, aclnode
);
1689 acep
= aclnode
->z_acldata
;
1691 zfs_set_ace(aclp
, acep
, access_mask
, type
,
1692 who
, iflags
|ACE_INHERITED_ACE
);
1695 * Copy special opaque data if any
1697 if ((data1sz
= paclp
->z_ops
->ace_data(pacep
, &data1
)) != 0) {
1698 VERIFY((data2sz
= aclp
->z_ops
->ace_data(acep
,
1699 &data2
)) == data1sz
);
1700 bcopy(data1
, data2
, data2sz
);
1703 aclp
->z_acl_count
++;
1704 aclnode
->z_ace_count
++;
1705 aclp
->z_acl_bytes
+= aclnode
->z_size
;
1706 newflags
= aclp
->z_ops
->ace_flags_get(acep
);
1709 aclp
->z_hints
|= ZFS_INHERIT_ACE
;
1711 if ((iflags
& ACE_NO_PROPAGATE_INHERIT_ACE
) || !vdir
) {
1712 newflags
&= ~ALL_INHERIT
;
1713 aclp
->z_ops
->ace_flags_set(acep
,
1714 newflags
|ACE_INHERITED_ACE
);
1715 zfs_restricted_update(zsb
, aclp
, acep
);
1722 * If only FILE_INHERIT is set then turn on
1725 if ((iflags
& (ACE_FILE_INHERIT_ACE
|
1726 ACE_DIRECTORY_INHERIT_ACE
)) == ACE_FILE_INHERIT_ACE
) {
1727 newflags
|= ACE_INHERIT_ONLY_ACE
;
1728 aclp
->z_ops
->ace_flags_set(acep
,
1729 newflags
|ACE_INHERITED_ACE
);
1731 newflags
&= ~ACE_INHERIT_ONLY_ACE
;
1732 aclp
->z_ops
->ace_flags_set(acep
,
1733 newflags
|ACE_INHERITED_ACE
);
1740 * Create file system object initial permissions
1741 * including inheritable ACEs.
1744 zfs_acl_ids_create(znode_t
*dzp
, int flag
, vattr_t
*vap
, cred_t
*cr
,
1745 vsecattr_t
*vsecp
, zfs_acl_ids_t
*acl_ids
)
1748 zfs_sb_t
*zsb
= ZTOZSB(dzp
);
1750 gid_t gid
= vap
->va_gid
;
1751 boolean_t need_chmod
= B_TRUE
;
1752 boolean_t inherited
= B_FALSE
;
1754 bzero(acl_ids
, sizeof (zfs_acl_ids_t
));
1755 acl_ids
->z_mode
= vap
->va_mode
;
1758 if ((error
= zfs_vsec_2_aclp(zsb
, vap
->va_mode
, vsecp
,
1759 cr
, &acl_ids
->z_fuidp
, &acl_ids
->z_aclp
)) != 0)
1762 acl_ids
->z_fuid
= vap
->va_uid
;
1763 acl_ids
->z_fgid
= vap
->va_gid
;
1766 * Determine uid and gid.
1768 if ((flag
& IS_ROOT_NODE
) || zsb
->z_replay
||
1769 ((flag
& IS_XATTR
) && (S_ISDIR(vap
->va_mode
)))) {
1770 acl_ids
->z_fuid
= zfs_fuid_create(zsb
, (uint64_t)vap
->va_uid
,
1771 cr
, ZFS_OWNER
, &acl_ids
->z_fuidp
);
1772 acl_ids
->z_fgid
= zfs_fuid_create(zsb
, (uint64_t)vap
->va_gid
,
1773 cr
, ZFS_GROUP
, &acl_ids
->z_fuidp
);
1776 acl_ids
->z_fuid
= zfs_fuid_create_cred(zsb
, ZFS_OWNER
,
1777 cr
, &acl_ids
->z_fuidp
);
1778 acl_ids
->z_fgid
= 0;
1779 if (vap
->va_mask
& AT_GID
) {
1780 acl_ids
->z_fgid
= zfs_fuid_create(zsb
,
1781 (uint64_t)vap
->va_gid
,
1782 cr
, ZFS_GROUP
, &acl_ids
->z_fuidp
);
1784 if (acl_ids
->z_fgid
!= KGID_TO_SGID(ZTOI(dzp
)->i_gid
) &&
1785 !groupmember(vap
->va_gid
, cr
) &&
1786 secpolicy_vnode_create_gid(cr
) != 0)
1787 acl_ids
->z_fgid
= 0;
1789 if (acl_ids
->z_fgid
== 0) {
1790 if (dzp
->z_mode
& S_ISGID
) {
1794 acl_ids
->z_fgid
= KGID_TO_SGID(
1796 gid
= zfs_fuid_map_id(zsb
, acl_ids
->z_fgid
,
1799 if (zsb
->z_use_fuids
&&
1800 IS_EPHEMERAL(acl_ids
->z_fgid
)) {
1801 domain
= zfs_fuid_idx_domain(
1803 FUID_INDEX(acl_ids
->z_fgid
));
1804 rid
= FUID_RID(acl_ids
->z_fgid
);
1805 zfs_fuid_node_add(&acl_ids
->z_fuidp
,
1807 FUID_INDEX(acl_ids
->z_fgid
),
1808 acl_ids
->z_fgid
, ZFS_GROUP
);
1811 acl_ids
->z_fgid
= zfs_fuid_create_cred(zsb
,
1812 ZFS_GROUP
, cr
, &acl_ids
->z_fuidp
);
1817 #endif /* HAVE_KSID */
1820 * If we're creating a directory, and the parent directory has the
1821 * set-GID bit set, set in on the new directory.
1822 * Otherwise, if the user is neither privileged nor a member of the
1823 * file's new group, clear the file's set-GID bit.
1826 if (!(flag
& IS_ROOT_NODE
) && (dzp
->z_mode
& S_ISGID
) &&
1827 (S_ISDIR(vap
->va_mode
))) {
1828 acl_ids
->z_mode
|= S_ISGID
;
1830 if ((acl_ids
->z_mode
& S_ISGID
) &&
1831 secpolicy_vnode_setids_setgids(cr
, gid
) != 0)
1832 acl_ids
->z_mode
&= ~S_ISGID
;
1835 if (acl_ids
->z_aclp
== NULL
) {
1836 mutex_enter(&dzp
->z_acl_lock
);
1837 mutex_enter(&dzp
->z_lock
);
1838 if (!(flag
& IS_ROOT_NODE
) && (S_ISDIR(ZTOI(dzp
)->i_mode
) &&
1839 (dzp
->z_pflags
& ZFS_INHERIT_ACE
)) &&
1840 !(dzp
->z_pflags
& ZFS_XATTR
)) {
1841 VERIFY(0 == zfs_acl_node_read(dzp
, B_TRUE
,
1843 acl_ids
->z_aclp
= zfs_acl_inherit(zsb
,
1844 vap
->va_mode
, paclp
, acl_ids
->z_mode
, &need_chmod
);
1848 zfs_acl_alloc(zfs_acl_version_zp(dzp
));
1849 acl_ids
->z_aclp
->z_hints
|= ZFS_ACL_TRIVIAL
;
1851 mutex_exit(&dzp
->z_lock
);
1852 mutex_exit(&dzp
->z_acl_lock
);
1854 acl_ids
->z_aclp
->z_hints
|= S_ISDIR(vap
->va_mode
) ?
1855 ZFS_ACL_AUTO_INHERIT
: 0;
1856 zfs_acl_chmod(zsb
, acl_ids
->z_mode
, acl_ids
->z_aclp
);
1860 if (inherited
|| vsecp
) {
1861 acl_ids
->z_mode
= zfs_mode_compute(acl_ids
->z_mode
,
1862 acl_ids
->z_aclp
, &acl_ids
->z_aclp
->z_hints
,
1863 acl_ids
->z_fuid
, acl_ids
->z_fgid
);
1864 if (ace_trivial_common(acl_ids
->z_aclp
, 0, zfs_ace_walk
) == 0)
1865 acl_ids
->z_aclp
->z_hints
|= ZFS_ACL_TRIVIAL
;
1872 * Free ACL and fuid_infop, but not the acl_ids structure
1875 zfs_acl_ids_free(zfs_acl_ids_t
*acl_ids
)
1877 if (acl_ids
->z_aclp
)
1878 zfs_acl_free(acl_ids
->z_aclp
);
1879 if (acl_ids
->z_fuidp
)
1880 zfs_fuid_info_free(acl_ids
->z_fuidp
);
1881 acl_ids
->z_aclp
= NULL
;
1882 acl_ids
->z_fuidp
= NULL
;
1886 zfs_acl_ids_overquota(zfs_sb_t
*zsb
, zfs_acl_ids_t
*acl_ids
)
1888 return (zfs_fuid_overquota(zsb
, B_FALSE
, acl_ids
->z_fuid
) ||
1889 zfs_fuid_overquota(zsb
, B_TRUE
, acl_ids
->z_fgid
) ||
1890 zfs_fuid_overobjquota(zsb
, B_FALSE
, acl_ids
->z_fuid
) ||
1891 zfs_fuid_overobjquota(zsb
, B_TRUE
, acl_ids
->z_fgid
));
1895 * Retrieve a file's ACL
1898 zfs_getacl(znode_t
*zp
, vsecattr_t
*vsecp
, boolean_t skipaclchk
, cred_t
*cr
)
1906 mask
= vsecp
->vsa_mask
& (VSA_ACE
| VSA_ACECNT
|
1907 VSA_ACE_ACLFLAGS
| VSA_ACE_ALLTYPES
);
1910 return (SET_ERROR(ENOSYS
));
1912 if ((error
= zfs_zaccess(zp
, ACE_READ_ACL
, 0, skipaclchk
, cr
)))
1915 mutex_enter(&zp
->z_acl_lock
);
1917 error
= zfs_acl_node_read(zp
, B_FALSE
, &aclp
, B_FALSE
);
1919 mutex_exit(&zp
->z_acl_lock
);
1924 * Scan ACL to determine number of ACEs
1926 if ((zp
->z_pflags
& ZFS_ACL_OBJ_ACE
) && !(mask
& VSA_ACE_ALLTYPES
)) {
1929 uint32_t access_mask
;
1930 uint16_t type
, iflags
;
1932 while ((zacep
= zfs_acl_next_ace(aclp
, zacep
,
1933 &who
, &access_mask
, &iflags
, &type
))) {
1935 case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE
:
1936 case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE
:
1937 case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE
:
1938 case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE
:
1945 vsecp
->vsa_aclcnt
= count
;
1947 count
= (int)aclp
->z_acl_count
;
1949 if (mask
& VSA_ACECNT
) {
1950 vsecp
->vsa_aclcnt
= count
;
1953 if (mask
& VSA_ACE
) {
1956 aclsz
= count
* sizeof (ace_t
) +
1957 sizeof (ace_object_t
) * largeace
;
1959 vsecp
->vsa_aclentp
= kmem_alloc(aclsz
, KM_SLEEP
);
1960 vsecp
->vsa_aclentsz
= aclsz
;
1962 if (aclp
->z_version
== ZFS_ACL_VERSION_FUID
)
1963 zfs_copy_fuid_2_ace(ZTOZSB(zp
), aclp
, cr
,
1964 vsecp
->vsa_aclentp
, !(mask
& VSA_ACE_ALLTYPES
));
1966 zfs_acl_node_t
*aclnode
;
1967 void *start
= vsecp
->vsa_aclentp
;
1969 for (aclnode
= list_head(&aclp
->z_acl
); aclnode
;
1970 aclnode
= list_next(&aclp
->z_acl
, aclnode
)) {
1971 bcopy(aclnode
->z_acldata
, start
,
1973 start
= (caddr_t
)start
+ aclnode
->z_size
;
1975 ASSERT((caddr_t
)start
- (caddr_t
)vsecp
->vsa_aclentp
==
1979 if (mask
& VSA_ACE_ACLFLAGS
) {
1980 vsecp
->vsa_aclflags
= 0;
1981 if (zp
->z_pflags
& ZFS_ACL_DEFAULTED
)
1982 vsecp
->vsa_aclflags
|= ACL_DEFAULTED
;
1983 if (zp
->z_pflags
& ZFS_ACL_PROTECTED
)
1984 vsecp
->vsa_aclflags
|= ACL_PROTECTED
;
1985 if (zp
->z_pflags
& ZFS_ACL_AUTO_INHERIT
)
1986 vsecp
->vsa_aclflags
|= ACL_AUTO_INHERIT
;
1989 mutex_exit(&zp
->z_acl_lock
);
1995 zfs_vsec_2_aclp(zfs_sb_t
*zsb
, umode_t obj_mode
,
1996 vsecattr_t
*vsecp
, cred_t
*cr
, zfs_fuid_info_t
**fuidp
, zfs_acl_t
**zaclp
)
1999 zfs_acl_node_t
*aclnode
;
2000 int aclcnt
= vsecp
->vsa_aclcnt
;
2003 if (vsecp
->vsa_aclcnt
> MAX_ACL_ENTRIES
|| vsecp
->vsa_aclcnt
<= 0)
2004 return (SET_ERROR(EINVAL
));
2006 aclp
= zfs_acl_alloc(zfs_acl_version(zsb
->z_version
));
2009 aclnode
= zfs_acl_node_alloc(aclcnt
* sizeof (zfs_object_ace_t
));
2010 if (aclp
->z_version
== ZFS_ACL_VERSION_INITIAL
) {
2011 if ((error
= zfs_copy_ace_2_oldace(obj_mode
, aclp
,
2012 (ace_t
*)vsecp
->vsa_aclentp
, aclnode
->z_acldata
,
2013 aclcnt
, &aclnode
->z_size
)) != 0) {
2015 zfs_acl_node_free(aclnode
);
2019 if ((error
= zfs_copy_ace_2_fuid(zsb
, obj_mode
, aclp
,
2020 vsecp
->vsa_aclentp
, aclnode
->z_acldata
, aclcnt
,
2021 &aclnode
->z_size
, fuidp
, cr
)) != 0) {
2023 zfs_acl_node_free(aclnode
);
2027 aclp
->z_acl_bytes
= aclnode
->z_size
;
2028 aclnode
->z_ace_count
= aclcnt
;
2029 aclp
->z_acl_count
= aclcnt
;
2030 list_insert_head(&aclp
->z_acl
, aclnode
);
2033 * If flags are being set then add them to z_hints
2035 if (vsecp
->vsa_mask
& VSA_ACE_ACLFLAGS
) {
2036 if (vsecp
->vsa_aclflags
& ACL_PROTECTED
)
2037 aclp
->z_hints
|= ZFS_ACL_PROTECTED
;
2038 if (vsecp
->vsa_aclflags
& ACL_DEFAULTED
)
2039 aclp
->z_hints
|= ZFS_ACL_DEFAULTED
;
2040 if (vsecp
->vsa_aclflags
& ACL_AUTO_INHERIT
)
2041 aclp
->z_hints
|= ZFS_ACL_AUTO_INHERIT
;
2053 zfs_setacl(znode_t
*zp
, vsecattr_t
*vsecp
, boolean_t skipaclchk
, cred_t
*cr
)
2055 zfs_sb_t
*zsb
= ZTOZSB(zp
);
2056 zilog_t
*zilog
= zsb
->z_log
;
2057 ulong_t mask
= vsecp
->vsa_mask
& (VSA_ACE
| VSA_ACECNT
);
2061 zfs_fuid_info_t
*fuidp
= NULL
;
2062 boolean_t fuid_dirtied
;
2066 return (SET_ERROR(ENOSYS
));
2068 if (zp
->z_pflags
& ZFS_IMMUTABLE
)
2069 return (SET_ERROR(EPERM
));
2071 if ((error
= zfs_zaccess(zp
, ACE_WRITE_ACL
, 0, skipaclchk
, cr
)))
2074 error
= zfs_vsec_2_aclp(zsb
, ZTOI(zp
)->i_mode
, vsecp
, cr
, &fuidp
,
2080 * If ACL wide flags aren't being set then preserve any
2083 if (!(vsecp
->vsa_mask
& VSA_ACE_ACLFLAGS
)) {
2085 (zp
->z_pflags
& V4_ACL_WIDE_FLAGS
);
2088 mutex_enter(&zp
->z_acl_lock
);
2089 mutex_enter(&zp
->z_lock
);
2091 tx
= dmu_tx_create(zsb
->z_os
);
2093 dmu_tx_hold_sa(tx
, zp
->z_sa_hdl
, B_TRUE
);
2095 fuid_dirtied
= zsb
->z_fuid_dirty
;
2097 zfs_fuid_txhold(zsb
, tx
);
2100 * If old version and ACL won't fit in bonus and we aren't
2101 * upgrading then take out necessary DMU holds
2104 if ((acl_obj
= zfs_external_acl(zp
)) != 0) {
2105 if (zsb
->z_version
>= ZPL_VERSION_FUID
&&
2106 zfs_znode_acl_version(zp
) <= ZFS_ACL_VERSION_INITIAL
) {
2107 dmu_tx_hold_free(tx
, acl_obj
, 0,
2109 dmu_tx_hold_write(tx
, DMU_NEW_OBJECT
, 0,
2112 dmu_tx_hold_write(tx
, acl_obj
, 0, aclp
->z_acl_bytes
);
2114 } else if (!zp
->z_is_sa
&& aclp
->z_acl_bytes
> ZFS_ACE_SPACE
) {
2115 dmu_tx_hold_write(tx
, DMU_NEW_OBJECT
, 0, aclp
->z_acl_bytes
);
2118 zfs_sa_upgrade_txholds(tx
, zp
);
2119 error
= dmu_tx_assign(tx
, TXG_NOWAIT
);
2121 mutex_exit(&zp
->z_acl_lock
);
2122 mutex_exit(&zp
->z_lock
);
2124 if (error
== ERESTART
) {
2134 error
= zfs_aclset_common(zp
, aclp
, cr
, tx
);
2136 ASSERT(zp
->z_acl_cached
== NULL
);
2137 zp
->z_acl_cached
= aclp
;
2140 zfs_fuid_sync(zsb
, tx
);
2142 zfs_log_acl(zilog
, tx
, zp
, vsecp
, fuidp
);
2145 zfs_fuid_info_free(fuidp
);
2148 mutex_exit(&zp
->z_lock
);
2149 mutex_exit(&zp
->z_acl_lock
);
2155 * Check accesses of interest (AoI) against attributes of the dataset
2156 * such as read-only. Returns zero if no AoI conflict with dataset
2157 * attributes, otherwise an appropriate errno is returned.
2160 zfs_zaccess_dataset_check(znode_t
*zp
, uint32_t v4_mode
)
2162 if ((v4_mode
& WRITE_MASK
) && (zfs_is_readonly(ZTOZSB(zp
))) &&
2163 (!S_ISDEV(ZTOI(zp
)->i_mode
) ||
2164 (S_ISDEV(ZTOI(zp
)->i_mode
) && (v4_mode
& WRITE_MASK_ATTRS
)))) {
2165 return (SET_ERROR(EROFS
));
2169 * Only check for READONLY on non-directories.
2171 if ((v4_mode
& WRITE_MASK_DATA
) &&
2172 ((!S_ISDIR(ZTOI(zp
)->i_mode
) &&
2173 (zp
->z_pflags
& (ZFS_READONLY
| ZFS_IMMUTABLE
))) ||
2174 (S_ISDIR(ZTOI(zp
)->i_mode
) &&
2175 (zp
->z_pflags
& ZFS_IMMUTABLE
)))) {
2176 return (SET_ERROR(EPERM
));
2179 if ((v4_mode
& (ACE_DELETE
| ACE_DELETE_CHILD
)) &&
2180 (zp
->z_pflags
& ZFS_NOUNLINK
)) {
2181 return (SET_ERROR(EPERM
));
2184 if (((v4_mode
& (ACE_READ_DATA
|ACE_EXECUTE
)) &&
2185 (zp
->z_pflags
& ZFS_AV_QUARANTINED
))) {
2186 return (SET_ERROR(EACCES
));
2193 * The primary usage of this function is to loop through all of the
2194 * ACEs in the znode, determining what accesses of interest (AoI) to
2195 * the caller are allowed or denied. The AoI are expressed as bits in
2196 * the working_mode parameter. As each ACE is processed, bits covered
2197 * by that ACE are removed from the working_mode. This removal
2198 * facilitates two things. The first is that when the working mode is
2199 * empty (= 0), we know we've looked at all the AoI. The second is
2200 * that the ACE interpretation rules don't allow a later ACE to undo
2201 * something granted or denied by an earlier ACE. Removing the
2202 * discovered access or denial enforces this rule. At the end of
2203 * processing the ACEs, all AoI that were found to be denied are
2204 * placed into the working_mode, giving the caller a mask of denied
2205 * accesses. Returns:
2206 * 0 if all AoI granted
2207 * EACCESS if the denied mask is non-zero
2208 * other error if abnormal failure (e.g., IO error)
2210 * A secondary usage of the function is to determine if any of the
2211 * AoI are granted. If an ACE grants any access in
2212 * the working_mode, we immediately short circuit out of the function.
2213 * This mode is chosen by setting anyaccess to B_TRUE. The
2214 * working_mode is not a denied access mask upon exit if the function
2215 * is used in this manner.
2218 zfs_zaccess_aces_check(znode_t
*zp
, uint32_t *working_mode
,
2219 boolean_t anyaccess
, cred_t
*cr
)
2221 zfs_sb_t
*zsb
= ZTOZSB(zp
);
2224 uid_t uid
= crgetuid(cr
);
2226 uint16_t type
, iflags
;
2227 uint16_t entry_type
;
2228 uint32_t access_mask
;
2229 uint32_t deny_mask
= 0;
2230 zfs_ace_hdr_t
*acep
= NULL
;
2235 zfs_fuid_map_ids(zp
, cr
, &fowner
, &gowner
);
2237 mutex_enter(&zp
->z_acl_lock
);
2239 error
= zfs_acl_node_read(zp
, B_FALSE
, &aclp
, B_FALSE
);
2241 mutex_exit(&zp
->z_acl_lock
);
2245 ASSERT(zp
->z_acl_cached
);
2247 while ((acep
= zfs_acl_next_ace(aclp
, acep
, &who
, &access_mask
,
2249 uint32_t mask_matched
;
2251 if (!zfs_acl_valid_ace_type(type
, iflags
))
2254 if (S_ISDIR(ZTOI(zp
)->i_mode
) &&
2255 (iflags
& ACE_INHERIT_ONLY_ACE
))
2258 /* Skip ACE if it does not affect any AoI */
2259 mask_matched
= (access_mask
& *working_mode
);
2263 entry_type
= (iflags
& ACE_TYPE_FLAGS
);
2267 switch (entry_type
) {
2275 case ACE_IDENTIFIER_GROUP
:
2276 checkit
= zfs_groupmember(zsb
, who
, cr
);
2284 if (entry_type
== 0) {
2287 newid
= zfs_fuid_map_id(zsb
, who
, cr
,
2289 if (newid
!= IDMAP_WK_CREATOR_OWNER_UID
&&
2294 mutex_exit(&zp
->z_acl_lock
);
2295 return (SET_ERROR(EIO
));
2301 DTRACE_PROBE3(zfs__ace__denies
,
2303 zfs_ace_hdr_t
*, acep
,
2304 uint32_t, mask_matched
);
2305 deny_mask
|= mask_matched
;
2307 DTRACE_PROBE3(zfs__ace__allows
,
2309 zfs_ace_hdr_t
*, acep
,
2310 uint32_t, mask_matched
);
2312 mutex_exit(&zp
->z_acl_lock
);
2316 *working_mode
&= ~mask_matched
;
2320 if (*working_mode
== 0)
2324 mutex_exit(&zp
->z_acl_lock
);
2326 /* Put the found 'denies' back on the working mode */
2328 *working_mode
|= deny_mask
;
2329 return (SET_ERROR(EACCES
));
2330 } else if (*working_mode
) {
2338 * Return true if any access whatsoever granted, we don't actually
2339 * care what access is granted.
2342 zfs_has_access(znode_t
*zp
, cred_t
*cr
)
2344 uint32_t have
= ACE_ALL_PERMS
;
2346 if (zfs_zaccess_aces_check(zp
, &have
, B_TRUE
, cr
) != 0) {
2349 owner
= zfs_fuid_map_id(ZTOZSB(zp
),
2350 KUID_TO_SUID(ZTOI(zp
)->i_uid
), cr
, ZFS_OWNER
);
2351 return (secpolicy_vnode_any_access(cr
, ZTOI(zp
), owner
) == 0);
2357 zfs_zaccess_common(znode_t
*zp
, uint32_t v4_mode
, uint32_t *working_mode
,
2358 boolean_t
*check_privs
, boolean_t skipaclchk
, cred_t
*cr
)
2360 zfs_sb_t
*zsb
= ZTOZSB(zp
);
2363 *working_mode
= v4_mode
;
2364 *check_privs
= B_TRUE
;
2367 * Short circuit empty requests
2369 if (v4_mode
== 0 || zsb
->z_replay
) {
2374 if ((err
= zfs_zaccess_dataset_check(zp
, v4_mode
)) != 0) {
2375 *check_privs
= B_FALSE
;
2380 * The caller requested that the ACL check be skipped. This
2381 * would only happen if the caller checked VOP_ACCESS() with a
2382 * 32 bit ACE mask and already had the appropriate permissions.
2389 return (zfs_zaccess_aces_check(zp
, working_mode
, B_FALSE
, cr
));
2393 zfs_zaccess_append(znode_t
*zp
, uint32_t *working_mode
, boolean_t
*check_privs
,
2396 if (*working_mode
!= ACE_WRITE_DATA
)
2397 return (SET_ERROR(EACCES
));
2399 return (zfs_zaccess_common(zp
, ACE_APPEND_DATA
, working_mode
,
2400 check_privs
, B_FALSE
, cr
));
2404 zfs_fastaccesschk_execute(znode_t
*zdp
, cred_t
*cr
)
2406 boolean_t owner
= B_FALSE
;
2407 boolean_t groupmbr
= B_FALSE
;
2409 uid_t uid
= crgetuid(cr
);
2412 if (zdp
->z_pflags
& ZFS_AV_QUARANTINED
)
2413 return (SET_ERROR(EACCES
));
2415 is_attr
= ((zdp
->z_pflags
& ZFS_XATTR
) &&
2416 (S_ISDIR(ZTOI(zdp
)->i_mode
)));
2421 mutex_enter(&zdp
->z_acl_lock
);
2423 if (zdp
->z_pflags
& ZFS_NO_EXECS_DENIED
) {
2424 mutex_exit(&zdp
->z_acl_lock
);
2428 if (KUID_TO_SUID(ZTOI(zdp
)->i_uid
) != 0 ||
2429 KGID_TO_SGID(ZTOI(zdp
)->i_gid
) != 0) {
2430 mutex_exit(&zdp
->z_acl_lock
);
2434 if (uid
== KUID_TO_SUID(ZTOI(zdp
)->i_uid
)) {
2436 if (zdp
->z_mode
& S_IXUSR
) {
2437 mutex_exit(&zdp
->z_acl_lock
);
2440 mutex_exit(&zdp
->z_acl_lock
);
2444 if (groupmember(KGID_TO_SGID(ZTOI(zdp
)->i_gid
), cr
)) {
2446 if (zdp
->z_mode
& S_IXGRP
) {
2447 mutex_exit(&zdp
->z_acl_lock
);
2450 mutex_exit(&zdp
->z_acl_lock
);
2454 if (!owner
&& !groupmbr
) {
2455 if (zdp
->z_mode
& S_IXOTH
) {
2456 mutex_exit(&zdp
->z_acl_lock
);
2461 mutex_exit(&zdp
->z_acl_lock
);
2464 DTRACE_PROBE(zfs__fastpath__execute__access__miss
);
2465 ZFS_ENTER(ZTOZSB(zdp
));
2466 error
= zfs_zaccess(zdp
, ACE_EXECUTE
, 0, B_FALSE
, cr
);
2467 ZFS_EXIT(ZTOZSB(zdp
));
2472 * Determine whether Access should be granted/denied.
2474 * The least priv subsytem is always consulted as a basic privilege
2475 * can define any form of access.
2478 zfs_zaccess(znode_t
*zp
, int mode
, int flags
, boolean_t skipaclchk
, cred_t
*cr
)
2480 uint32_t working_mode
;
2483 boolean_t check_privs
;
2485 znode_t
*check_zp
= zp
;
2489 is_attr
= ((zp
->z_pflags
& ZFS_XATTR
) && S_ISDIR(ZTOI(zp
)->i_mode
));
2492 * If attribute then validate against base file
2495 if ((error
= zfs_zget(ZTOZSB(zp
),
2496 zp
->z_xattr_parent
, &xzp
)) != 0) {
2503 * fixup mode to map to xattr perms
2506 if (mode
& (ACE_WRITE_DATA
|ACE_APPEND_DATA
)) {
2507 mode
&= ~(ACE_WRITE_DATA
|ACE_APPEND_DATA
);
2508 mode
|= ACE_WRITE_NAMED_ATTRS
;
2511 if (mode
& (ACE_READ_DATA
|ACE_EXECUTE
)) {
2512 mode
&= ~(ACE_READ_DATA
|ACE_EXECUTE
);
2513 mode
|= ACE_READ_NAMED_ATTRS
;
2517 owner
= zfs_fuid_map_id(ZTOZSB(zp
), KUID_TO_SUID(ZTOI(zp
)->i_uid
),
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) {
2546 return (secpolicy_vnode_access2(cr
, ZTOI(zp
), owner
,
2547 needed_bits
, needed_bits
));
2550 if (error
&& !check_privs
) {
2556 if (error
&& (flags
& V_APPEND
)) {
2557 error
= zfs_zaccess_append(zp
, &working_mode
, &check_privs
, cr
);
2560 if (error
&& check_privs
) {
2561 mode_t checkmode
= 0;
2564 * First check for implicit owner permission on
2565 * read_acl/read_attributes
2569 ASSERT(working_mode
!= 0);
2571 if ((working_mode
& (ACE_READ_ACL
|ACE_READ_ATTRIBUTES
) &&
2572 owner
== crgetuid(cr
)))
2573 working_mode
&= ~(ACE_READ_ACL
|ACE_READ_ATTRIBUTES
);
2575 if (working_mode
& (ACE_READ_DATA
|ACE_READ_NAMED_ATTRS
|
2576 ACE_READ_ACL
|ACE_READ_ATTRIBUTES
|ACE_SYNCHRONIZE
))
2577 checkmode
|= S_IRUSR
;
2578 if (working_mode
& (ACE_WRITE_DATA
|ACE_WRITE_NAMED_ATTRS
|
2579 ACE_APPEND_DATA
|ACE_WRITE_ATTRIBUTES
|ACE_SYNCHRONIZE
))
2580 checkmode
|= S_IWUSR
;
2581 if (working_mode
& ACE_EXECUTE
)
2582 checkmode
|= S_IXUSR
;
2584 error
= secpolicy_vnode_access2(cr
, ZTOI(check_zp
), owner
,
2585 needed_bits
& ~checkmode
, needed_bits
);
2587 if (error
== 0 && (working_mode
& ACE_WRITE_OWNER
))
2588 error
= secpolicy_vnode_chown(cr
, owner
);
2589 if (error
== 0 && (working_mode
& ACE_WRITE_ACL
))
2590 error
= secpolicy_vnode_setdac(cr
, owner
);
2592 if (error
== 0 && (working_mode
&
2593 (ACE_DELETE
|ACE_DELETE_CHILD
)))
2594 error
= secpolicy_vnode_remove(cr
);
2596 if (error
== 0 && (working_mode
& ACE_SYNCHRONIZE
)) {
2597 error
= secpolicy_vnode_chown(cr
, owner
);
2601 * See if any bits other than those already checked
2602 * for are still present. If so then return EACCES
2604 if (working_mode
& ~(ZFS_CHECKED_MASKS
)) {
2605 error
= SET_ERROR(EACCES
);
2608 } else if (error
== 0) {
2609 error
= secpolicy_vnode_access2(cr
, ZTOI(zp
), owner
,
2610 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
), KUID_TO_SUID(ZTOI(dzp
)->i_uid
),
2650 error
= secpolicy_vnode_access2(cr
, ZTOI(dzp
),
2651 downer
, available_perms
, S_IWUSR
|S_IXUSR
);
2654 error
= zfs_sticky_remove_access(dzp
, zp
, cr
);
2660 * Determine whether Access should be granted/deny, without
2661 * consulting least priv subsystem.
2663 * The following chart is the recommended NFSv4 enforcement for
2664 * ability to delete an object.
2666 * -------------------------------------------------------
2667 * | Parent Dir | Target Object Permissions |
2669 * -------------------------------------------------------
2670 * | | ACL Allows | ACL Denies| Delete |
2671 * | | Delete | Delete | unspecified|
2672 * -------------------------------------------------------
2673 * | ACL Allows | Permit | Permit | Permit |
2674 * | DELETE_CHILD | |
2675 * -------------------------------------------------------
2676 * | ACL Denies | Permit | Deny | Deny |
2677 * | DELETE_CHILD | | | |
2678 * -------------------------------------------------------
2679 * | ACL specifies | | | |
2680 * | only allow | Permit | Permit | Permit |
2681 * | write and | | | |
2683 * -------------------------------------------------------
2684 * | ACL denies | | | |
2685 * | write and | Permit | Deny | Deny |
2687 * -------------------------------------------------------
2690 * No search privilege, can't even look up file?
2694 zfs_zaccess_delete(znode_t
*dzp
, znode_t
*zp
, cred_t
*cr
)
2696 uint32_t dzp_working_mode
= 0;
2697 uint32_t zp_working_mode
= 0;
2698 int dzp_error
, zp_error
;
2699 mode_t available_perms
;
2700 boolean_t dzpcheck_privs
= B_TRUE
;
2701 boolean_t zpcheck_privs
= B_TRUE
;
2704 * We want specific DELETE permissions to
2705 * take precedence over WRITE/EXECUTE. We don't
2706 * want an ACL such as this to mess us up.
2707 * user:joe:write_data:deny,user:joe:delete:allow
2709 * However, deny permissions may ultimately be overridden
2710 * by secpolicy_vnode_access().
2712 * We will ask for all of the necessary permissions and then
2713 * look at the working modes from the directory and target object
2714 * to determine what was found.
2717 if (zp
->z_pflags
& (ZFS_IMMUTABLE
| ZFS_NOUNLINK
))
2718 return (SET_ERROR(EPERM
));
2722 * If the directory permissions allow the delete, we are done.
2724 if ((dzp_error
= zfs_zaccess_common(dzp
, ACE_DELETE_CHILD
,
2725 &dzp_working_mode
, &dzpcheck_privs
, B_FALSE
, cr
)) == 0)
2729 * If target object has delete permission then we are done
2731 if ((zp_error
= zfs_zaccess_common(zp
, ACE_DELETE
, &zp_working_mode
,
2732 &zpcheck_privs
, B_FALSE
, cr
)) == 0)
2735 ASSERT(dzp_error
&& zp_error
);
2737 if (!dzpcheck_privs
)
2745 * If directory returns EACCES then delete_child was denied
2746 * due to deny delete_child. In this case send the request through
2747 * secpolicy_vnode_remove(). We don't use zfs_delete_final_check()
2748 * since that *could* allow the delete based on write/execute permission
2749 * and we want delete permissions to override write/execute.
2752 if (dzp_error
== EACCES
)
2753 return (secpolicy_vnode_remove(cr
));
2757 * only need to see if we have write/execute on directory.
2760 dzp_error
= zfs_zaccess_common(dzp
, ACE_EXECUTE
|ACE_WRITE_DATA
,
2761 &dzp_working_mode
, &dzpcheck_privs
, B_FALSE
, cr
);
2763 if (dzp_error
!= 0 && !dzpcheck_privs
)
2770 available_perms
= (dzp_working_mode
& ACE_WRITE_DATA
) ? 0 : S_IWUSR
;
2771 available_perms
|= (dzp_working_mode
& ACE_EXECUTE
) ? 0 : S_IXUSR
;
2773 return (zfs_delete_final_check(zp
, dzp
, available_perms
, cr
));
2778 zfs_zaccess_rename(znode_t
*sdzp
, znode_t
*szp
, znode_t
*tdzp
,
2779 znode_t
*tzp
, cred_t
*cr
)
2784 if (szp
->z_pflags
& ZFS_AV_QUARANTINED
)
2785 return (SET_ERROR(EACCES
));
2787 add_perm
= S_ISDIR(ZTOI(szp
)->i_mode
) ?
2788 ACE_ADD_SUBDIRECTORY
: ACE_ADD_FILE
;
2791 * Rename permissions are combination of delete permission +
2792 * add file/subdir permission.
2796 * first make sure we do the delete portion.
2798 * If that succeeds then check for add_file/add_subdir permissions
2801 if ((error
= zfs_zaccess_delete(sdzp
, szp
, cr
)))
2805 * If we have a tzp, see if we can delete it?
2808 if ((error
= zfs_zaccess_delete(tdzp
, tzp
, cr
)))
2813 * Now check for add permissions
2815 error
= zfs_zaccess(tdzp
, add_perm
, 0, B_FALSE
, cr
);