]> git.proxmox.com Git - mirror_zfs-debian.git/blob - module/zfs/zfs_acl.c
New upstream version 0.7.9
[mirror_zfs-debian.git] / module / zfs / zfs_acl.c
1 /*
2 * CDDL HEADER START
3 *
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.
7 *
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.
12 *
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]
18 *
19 * CDDL HEADER END
20 */
21 /*
22 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
23 * Copyright (c) 2013 by Delphix. All rights reserved.
24 */
25
26
27 #include <sys/types.h>
28 #include <sys/param.h>
29 #include <sys/time.h>
30 #include <sys/systm.h>
31 #include <sys/sysmacros.h>
32 #include <sys/resource.h>
33 #include <sys/vfs.h>
34 #include <sys/vnode.h>
35 #include <sys/sid.h>
36 #include <sys/file.h>
37 #include <sys/stat.h>
38 #include <sys/kmem.h>
39 #include <sys/cmn_err.h>
40 #include <sys/errno.h>
41 #include <sys/unistd.h>
42 #include <sys/sdt.h>
43 #include <sys/fs/zfs.h>
44 #include <sys/mode.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>
51 #include <sys/dmu.h>
52 #include <sys/dnode.h>
53 #include <sys/zap.h>
54 #include <sys/sa.h>
55 #include <sys/trace_acl.h>
56 #include <sys/zpl.h>
57 #include "fs/fs_subr.h"
58
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
63
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)
71
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)
76
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)
81
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)
84
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)
87
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)
90
91 #define RESTRICTED_CLEAR (ACE_WRITE_ACL|ACE_WRITE_OWNER)
92
93 #define V4_ACL_WIDE_FLAGS (ZFS_ACL_AUTO_INHERIT|ZFS_ACL_DEFAULTED|\
94 ZFS_ACL_PROTECTED)
95
96 #define ZFS_ACL_WIDE_FLAGS (V4_ACL_WIDE_FLAGS|ZFS_ACL_TRIVIAL|ZFS_INHERIT_ACE|\
97 ZFS_ACL_OBJ_ACE)
98
99 #define ALL_MODE_EXECS (S_IXUSR | S_IXGRP | S_IXOTH)
100
101 static uint16_t
102 zfs_ace_v0_get_type(void *acep)
103 {
104 return (((zfs_oldace_t *)acep)->z_type);
105 }
106
107 static uint16_t
108 zfs_ace_v0_get_flags(void *acep)
109 {
110 return (((zfs_oldace_t *)acep)->z_flags);
111 }
112
113 static uint32_t
114 zfs_ace_v0_get_mask(void *acep)
115 {
116 return (((zfs_oldace_t *)acep)->z_access_mask);
117 }
118
119 static uint64_t
120 zfs_ace_v0_get_who(void *acep)
121 {
122 return (((zfs_oldace_t *)acep)->z_fuid);
123 }
124
125 static void
126 zfs_ace_v0_set_type(void *acep, uint16_t type)
127 {
128 ((zfs_oldace_t *)acep)->z_type = type;
129 }
130
131 static void
132 zfs_ace_v0_set_flags(void *acep, uint16_t flags)
133 {
134 ((zfs_oldace_t *)acep)->z_flags = flags;
135 }
136
137 static void
138 zfs_ace_v0_set_mask(void *acep, uint32_t mask)
139 {
140 ((zfs_oldace_t *)acep)->z_access_mask = mask;
141 }
142
143 static void
144 zfs_ace_v0_set_who(void *acep, uint64_t who)
145 {
146 ((zfs_oldace_t *)acep)->z_fuid = who;
147 }
148
149 /*ARGSUSED*/
150 static size_t
151 zfs_ace_v0_size(void *acep)
152 {
153 return (sizeof (zfs_oldace_t));
154 }
155
156 static size_t
157 zfs_ace_v0_abstract_size(void)
158 {
159 return (sizeof (zfs_oldace_t));
160 }
161
162 static int
163 zfs_ace_v0_mask_off(void)
164 {
165 return (offsetof(zfs_oldace_t, z_access_mask));
166 }
167
168 /*ARGSUSED*/
169 static int
170 zfs_ace_v0_data(void *acep, void **datap)
171 {
172 *datap = NULL;
173 return (0);
174 }
175
176 static acl_ops_t zfs_acl_v0_ops = {
177 .ace_mask_get = zfs_ace_v0_get_mask,
178 .ace_mask_set = zfs_ace_v0_set_mask,
179 .ace_flags_get = zfs_ace_v0_get_flags,
180 .ace_flags_set = zfs_ace_v0_set_flags,
181 .ace_type_get = zfs_ace_v0_get_type,
182 .ace_type_set = zfs_ace_v0_set_type,
183 .ace_who_get = zfs_ace_v0_get_who,
184 .ace_who_set = zfs_ace_v0_set_who,
185 .ace_size = zfs_ace_v0_size,
186 .ace_abstract_size = zfs_ace_v0_abstract_size,
187 .ace_mask_off = zfs_ace_v0_mask_off,
188 .ace_data = zfs_ace_v0_data
189 };
190
191 static uint16_t
192 zfs_ace_fuid_get_type(void *acep)
193 {
194 return (((zfs_ace_hdr_t *)acep)->z_type);
195 }
196
197 static uint16_t
198 zfs_ace_fuid_get_flags(void *acep)
199 {
200 return (((zfs_ace_hdr_t *)acep)->z_flags);
201 }
202
203 static uint32_t
204 zfs_ace_fuid_get_mask(void *acep)
205 {
206 return (((zfs_ace_hdr_t *)acep)->z_access_mask);
207 }
208
209 static uint64_t
210 zfs_ace_fuid_get_who(void *args)
211 {
212 uint16_t entry_type;
213 zfs_ace_t *acep = args;
214
215 entry_type = acep->z_hdr.z_flags & ACE_TYPE_FLAGS;
216
217 if (entry_type == ACE_OWNER || entry_type == OWNING_GROUP ||
218 entry_type == ACE_EVERYONE)
219 return (-1);
220 return (((zfs_ace_t *)acep)->z_fuid);
221 }
222
223 static void
224 zfs_ace_fuid_set_type(void *acep, uint16_t type)
225 {
226 ((zfs_ace_hdr_t *)acep)->z_type = type;
227 }
228
229 static void
230 zfs_ace_fuid_set_flags(void *acep, uint16_t flags)
231 {
232 ((zfs_ace_hdr_t *)acep)->z_flags = flags;
233 }
234
235 static void
236 zfs_ace_fuid_set_mask(void *acep, uint32_t mask)
237 {
238 ((zfs_ace_hdr_t *)acep)->z_access_mask = mask;
239 }
240
241 static void
242 zfs_ace_fuid_set_who(void *arg, uint64_t who)
243 {
244 zfs_ace_t *acep = arg;
245
246 uint16_t entry_type = acep->z_hdr.z_flags & ACE_TYPE_FLAGS;
247
248 if (entry_type == ACE_OWNER || entry_type == OWNING_GROUP ||
249 entry_type == ACE_EVERYONE)
250 return;
251 acep->z_fuid = who;
252 }
253
254 static size_t
255 zfs_ace_fuid_size(void *acep)
256 {
257 zfs_ace_hdr_t *zacep = acep;
258 uint16_t entry_type;
259
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));
266 case ALLOW:
267 case DENY:
268 entry_type =
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));
274 /*FALLTHROUGH*/
275 default:
276 return (sizeof (zfs_ace_t));
277 }
278 }
279
280 static size_t
281 zfs_ace_fuid_abstract_size(void)
282 {
283 return (sizeof (zfs_ace_hdr_t));
284 }
285
286 static int
287 zfs_ace_fuid_mask_off(void)
288 {
289 return (offsetof(zfs_ace_hdr_t, z_access_mask));
290 }
291
292 static int
293 zfs_ace_fuid_data(void *acep, void **datap)
294 {
295 zfs_ace_t *zacep = acep;
296 zfs_object_ace_t *zobjp;
297
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:
303 zobjp = acep;
304 *datap = (caddr_t)zobjp + sizeof (zfs_ace_t);
305 return (sizeof (zfs_object_ace_t) - sizeof (zfs_ace_t));
306 default:
307 *datap = NULL;
308 return (0);
309 }
310 }
311
312 static acl_ops_t zfs_acl_fuid_ops = {
313 .ace_mask_get = zfs_ace_fuid_get_mask,
314 .ace_mask_set = zfs_ace_fuid_set_mask,
315 .ace_flags_get = zfs_ace_fuid_get_flags,
316 .ace_flags_set = zfs_ace_fuid_set_flags,
317 .ace_type_get = zfs_ace_fuid_get_type,
318 .ace_type_set = zfs_ace_fuid_set_type,
319 .ace_who_get = zfs_ace_fuid_get_who,
320 .ace_who_set = zfs_ace_fuid_set_who,
321 .ace_size = zfs_ace_fuid_size,
322 .ace_abstract_size = zfs_ace_fuid_abstract_size,
323 .ace_mask_off = zfs_ace_fuid_mask_off,
324 .ace_data = zfs_ace_fuid_data
325 };
326
327 /*
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.
332 */
333 uint64_t
334 zfs_external_acl(znode_t *zp)
335 {
336 zfs_acl_phys_t acl_phys;
337 int error;
338
339 if (zp->z_is_sa)
340 return (0);
341
342 /*
343 * Need to deal with a potential
344 * race where zfs_sa_upgrade could cause
345 * z_isa_sa to change.
346 *
347 * If the lookup fails then the state of z_is_sa should have
348 * changed.
349 */
350
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);
354 else {
355 /*
356 * after upgrade the SA_ZPL_ZNODE_ACL should have been
357 * removed
358 */
359 VERIFY(zp->z_is_sa && error == ENOENT);
360 return (0);
361 }
362 }
363
364 /*
365 * Determine size of ACL in bytes
366 *
367 * This is more complicated than it should be since we have to deal
368 * with old external ACLs.
369 */
370 static int
371 zfs_acl_znode_info(znode_t *zp, int *aclsize, int *aclcount,
372 zfs_acl_phys_t *aclphys)
373 {
374 zfsvfs_t *zfsvfs = ZTOZSB(zp);
375 uint64_t acl_count;
376 int size;
377 int error;
378
379 ASSERT(MUTEX_HELD(&zp->z_acl_lock));
380 if (zp->z_is_sa) {
381 if ((error = sa_size(zp->z_sa_hdl, SA_ZPL_DACL_ACES(zfsvfs),
382 &size)) != 0)
383 return (error);
384 *aclsize = size;
385 if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_DACL_COUNT(zfsvfs),
386 &acl_count, sizeof (acl_count))) != 0)
387 return (error);
388 *aclcount = acl_count;
389 } else {
390 if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_ZNODE_ACL(zfsvfs),
391 aclphys, sizeof (*aclphys))) != 0)
392 return (error);
393
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;
397 } else {
398 *aclsize = aclphys->z_acl_size;
399 *aclcount = aclphys->z_acl_count;
400 }
401 }
402 return (0);
403 }
404
405 int
406 zfs_znode_acl_version(znode_t *zp)
407 {
408 zfs_acl_phys_t acl_phys;
409
410 if (zp->z_is_sa)
411 return (ZFS_ACL_VERSION_FUID);
412 else {
413 int error;
414
415 /*
416 * Need to deal with a potential
417 * race where zfs_sa_upgrade could cause
418 * z_isa_sa to change.
419 *
420 * If the lookup fails then the state of z_is_sa should have
421 * changed.
422 */
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);
427 else {
428 /*
429 * After upgrade SA_ZPL_ZNODE_ACL should have
430 * been removed.
431 */
432 VERIFY(zp->z_is_sa && error == ENOENT);
433 return (ZFS_ACL_VERSION_FUID);
434 }
435 }
436 }
437
438 static int
439 zfs_acl_version(int version)
440 {
441 if (version < ZPL_VERSION_FUID)
442 return (ZFS_ACL_VERSION_INITIAL);
443 else
444 return (ZFS_ACL_VERSION_FUID);
445 }
446
447 static int
448 zfs_acl_version_zp(znode_t *zp)
449 {
450 return (zfs_acl_version(ZTOZSB(zp)->z_version));
451 }
452
453 zfs_acl_t *
454 zfs_acl_alloc(int vers)
455 {
456 zfs_acl_t *aclp;
457
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;
464 else
465 aclp->z_ops = &zfs_acl_v0_ops;
466 return (aclp);
467 }
468
469 zfs_acl_node_t *
470 zfs_acl_node_alloc(size_t bytes)
471 {
472 zfs_acl_node_t *aclnode;
473
474 aclnode = kmem_zalloc(sizeof (zfs_acl_node_t), KM_SLEEP);
475 if (bytes) {
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;
480 }
481
482 return (aclnode);
483 }
484
485 static void
486 zfs_acl_node_free(zfs_acl_node_t *aclnode)
487 {
488 if (aclnode->z_allocsize)
489 kmem_free(aclnode->z_allocdata, aclnode->z_allocsize);
490 kmem_free(aclnode, sizeof (zfs_acl_node_t));
491 }
492
493 static void
494 zfs_acl_release_nodes(zfs_acl_t *aclp)
495 {
496 zfs_acl_node_t *aclnode;
497
498 while ((aclnode = list_head(&aclp->z_acl))) {
499 list_remove(&aclp->z_acl, aclnode);
500 zfs_acl_node_free(aclnode);
501 }
502 aclp->z_acl_count = 0;
503 aclp->z_acl_bytes = 0;
504 }
505
506 void
507 zfs_acl_free(zfs_acl_t *aclp)
508 {
509 zfs_acl_release_nodes(aclp);
510 list_destroy(&aclp->z_acl);
511 kmem_free(aclp, sizeof (zfs_acl_t));
512 }
513
514 static boolean_t
515 zfs_acl_valid_ace_type(uint_t type, uint_t flags)
516 {
517 uint16_t entry_type;
518
519 switch (type) {
520 case ALLOW:
521 case DENY:
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);
529 default:
530 if (type >= MIN_ACE_TYPE && type <= MAX_ACE_TYPE)
531 return (B_TRUE);
532 }
533 return (B_FALSE);
534 }
535
536 static boolean_t
537 zfs_ace_valid(umode_t obj_mode, zfs_acl_t *aclp, uint16_t type, uint16_t iflags)
538 {
539 /*
540 * first check type of entry
541 */
542
543 if (!zfs_acl_valid_ace_type(type, iflags))
544 return (B_FALSE);
545
546 switch (type) {
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)
552 return (B_FALSE);
553 aclp->z_hints |= ZFS_ACL_OBJ_ACE;
554 }
555
556 /*
557 * next check inheritance level flags
558 */
559
560 if (S_ISDIR(obj_mode) &&
561 (iflags & (ACE_FILE_INHERIT_ACE|ACE_DIRECTORY_INHERIT_ACE)))
562 aclp->z_hints |= ZFS_INHERIT_ACE;
563
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) {
567 return (B_FALSE);
568 }
569 }
570
571 return (B_TRUE);
572 }
573
574 static void *
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)
577 {
578 zfs_acl_node_t *aclnode;
579
580 ASSERT(aclp);
581
582 if (start == NULL) {
583 aclnode = list_head(&aclp->z_acl);
584 if (aclnode == NULL)
585 return (NULL);
586
587 aclp->z_next_ace = aclnode->z_acldata;
588 aclp->z_curr_node = aclnode;
589 aclnode->z_ace_idx = 0;
590 }
591
592 aclnode = aclp->z_curr_node;
593
594 if (aclnode == NULL)
595 return (NULL);
596
597 if (aclnode->z_ace_idx >= aclnode->z_ace_count) {
598 aclnode = list_next(&aclp->z_acl, aclnode);
599 if (aclnode == NULL)
600 return (NULL);
601 else {
602 aclp->z_curr_node = aclnode;
603 aclnode->z_ace_idx = 0;
604 aclp->z_next_ace = aclnode->z_acldata;
605 }
606 }
607
608 if (aclnode->z_ace_idx < aclnode->z_ace_count) {
609 void *acep = aclp->z_next_ace;
610 size_t ace_size;
611
612 /*
613 * Make sure we don't overstep our bounds
614 */
615 ace_size = aclp->z_ops->ace_size(acep);
616
617 if (((caddr_t)acep + ace_size) >
618 ((caddr_t)aclnode->z_acldata + aclnode->z_size)) {
619 return (NULL);
620 }
621
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++;
628
629 return ((void *)acep);
630 }
631 return (NULL);
632 }
633
634 /*ARGSUSED*/
635 static uint64_t
636 zfs_ace_walk(void *datap, uint64_t cookie, int aclcnt,
637 uint16_t *flags, uint16_t *type, uint32_t *mask)
638 {
639 zfs_acl_t *aclp = datap;
640 zfs_ace_hdr_t *acep = (zfs_ace_hdr_t *)(uintptr_t)cookie;
641 uint64_t who;
642
643 acep = zfs_acl_next_ace(aclp, acep, &who, mask,
644 flags, type);
645 return ((uint64_t)(uintptr_t)acep);
646 }
647
648 /*
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.
652 */
653 int
654 zfs_copy_ace_2_fuid(zfsvfs_t *zfsvfs, 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)
657 {
658 int i;
659 uint16_t entry_type;
660 zfs_ace_t *aceptr = z_acl;
661 ace_t *acep = datap;
662 zfs_object_ace_t *zobjacep;
663 ace_object_t *aceobjp;
664
665 for (i = 0; i != aclcnt; i++) {
666 aceptr->z_hdr.z_access_mask = acep->a_access_mask;
667 aceptr->z_hdr.z_flags = acep->a_flags;
668 aceptr->z_hdr.z_type = acep->a_type;
669 entry_type = aceptr->z_hdr.z_flags & ACE_TYPE_FLAGS;
670 if (entry_type != ACE_OWNER && entry_type != OWNING_GROUP &&
671 entry_type != ACE_EVERYONE) {
672 aceptr->z_fuid = zfs_fuid_create(zfsvfs, acep->a_who,
673 cr, (entry_type == 0) ?
674 ZFS_ACE_USER : ZFS_ACE_GROUP, fuidp);
675 }
676
677 /*
678 * Make sure ACE is valid
679 */
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));
683
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;
691
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));
698 break;
699 default:
700 acep = (ace_t *)((caddr_t)acep + sizeof (ace_t));
701 }
702
703 aceptr = (zfs_ace_t *)((caddr_t)aceptr +
704 aclp->z_ops->ace_size(aceptr));
705 }
706
707 *size = (caddr_t)aceptr - (caddr_t)z_acl;
708
709 return (0);
710 }
711
712 /*
713 * Copy ZFS ACEs to fixed size ace_t layout
714 */
715 static void
716 zfs_copy_fuid_2_ace(zfsvfs_t *zfsvfs, zfs_acl_t *aclp, cred_t *cr,
717 void *datap, int filter)
718 {
719 uint64_t who;
720 uint32_t access_mask;
721 uint16_t iflags, type;
722 zfs_ace_hdr_t *zacep = NULL;
723 ace_t *acep = datap;
724 ace_object_t *objacep;
725 zfs_object_ace_t *zobjacep;
726 size_t ace_size;
727 uint16_t entry_type;
728
729 while ((zacep = zfs_acl_next_ace(aclp, zacep,
730 &who, &access_mask, &iflags, &type))) {
731
732 switch (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:
737 if (filter) {
738 continue;
739 }
740 zobjacep = (zfs_object_ace_t *)zacep;
741 objacep = (ace_object_t *)acep;
742 bcopy(zobjacep->z_object_type,
743 objacep->a_obj_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);
749 break;
750 default:
751 ace_size = sizeof (ace_t);
752 break;
753 }
754
755 entry_type = (iflags & ACE_TYPE_FLAGS);
756 if ((entry_type != ACE_OWNER &&
757 entry_type != OWNING_GROUP &&
758 entry_type != ACE_EVERYONE)) {
759 acep->a_who = zfs_fuid_map_id(zfsvfs, who,
760 cr, (entry_type & ACE_IDENTIFIER_GROUP) ?
761 ZFS_ACE_GROUP : ZFS_ACE_USER);
762 } else {
763 acep->a_who = (uid_t)(int64_t)who;
764 }
765 acep->a_access_mask = access_mask;
766 acep->a_flags = iflags;
767 acep->a_type = type;
768 acep = (ace_t *)((caddr_t)acep + ace_size);
769 }
770 }
771
772 static int
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)
775 {
776 int i;
777 zfs_oldace_t *aceptr = z_acl;
778
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;
784 /*
785 * Make sure ACE is valid
786 */
787 if (zfs_ace_valid(obj_mode, aclp, aceptr->z_type,
788 aceptr->z_flags) != B_TRUE)
789 return (SET_ERROR(EINVAL));
790 }
791 *size = (caddr_t)aceptr - (caddr_t)z_acl;
792 return (0);
793 }
794
795 /*
796 * convert old ACL format to new
797 */
798 void
799 zfs_acl_xform(znode_t *zp, zfs_acl_t *aclp, cred_t *cr)
800 {
801 zfs_oldace_t *oldaclp;
802 int i;
803 uint16_t type, iflags;
804 uint32_t access_mask;
805 uint64_t who;
806 void *cookie = NULL;
807 zfs_acl_node_t *newaclnode;
808
809 ASSERT(aclp->z_version == ZFS_ACL_VERSION_INITIAL);
810 /*
811 * First create the ACE in a contiguous piece of memory
812 * for zfs_copy_ace_2_fuid().
813 *
814 * We only convert an ACL once, so this won't happen
815 * everytime.
816 */
817 oldaclp = kmem_alloc(sizeof (zfs_oldace_t) * aclp->z_acl_count,
818 KM_SLEEP);
819 i = 0;
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;
826 }
827
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));
837
838 /*
839 * Release all previous ACL nodes
840 */
841
842 zfs_acl_release_nodes(aclp);
843
844 list_insert_head(&aclp->z_acl, newaclnode);
845
846 aclp->z_acl_bytes = newaclnode->z_size;
847 aclp->z_acl_count = newaclnode->z_ace_count;
848
849 }
850
851 /*
852 * Convert unix access mask to v4 access mask
853 */
854 static uint32_t
855 zfs_unix_to_v4(uint32_t access_mask)
856 {
857 uint32_t new_mask = 0;
858
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;
865 return (new_mask);
866 }
867
868 static void
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)
871 {
872 uint16_t type = entry_type & ACE_TYPE_FLAGS;
873
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);
880 }
881
882 /*
883 * Determine mode of file based on ACL.
884 * Also, create FUIDs for any User/Group ACEs
885 */
886 uint64_t
887 zfs_mode_compute(uint64_t fmode, zfs_acl_t *aclp,
888 uint64_t *pflags, uint64_t fuid, uint64_t fgid)
889 {
890 int entry_type;
891 mode_t mode;
892 mode_t seen = 0;
893 zfs_ace_hdr_t *acep = NULL;
894 uint64_t who;
895 uint16_t iflags, type;
896 uint32_t access_mask;
897 boolean_t an_exec_denied = B_FALSE;
898
899 mode = (fmode & (S_IFMT | S_ISUID | S_ISGID | S_ISVTX));
900
901 while ((acep = zfs_acl_next_ace(aclp, acep, &who,
902 &access_mask, &iflags, &type))) {
903
904 if (!zfs_acl_valid_ace_type(type, iflags))
905 continue;
906
907 entry_type = (iflags & ACE_TYPE_FLAGS);
908
909 /*
910 * Skip over owner@, group@ or everyone@ inherit only ACEs
911 */
912 if ((iflags & ACE_INHERIT_ONLY_ACE) &&
913 (entry_type == ACE_OWNER || entry_type == ACE_EVERYONE ||
914 entry_type == OWNING_GROUP))
915 continue;
916
917 if (entry_type == ACE_OWNER || (entry_type == 0 &&
918 who == fuid)) {
919 if ((access_mask & ACE_READ_DATA) &&
920 (!(seen & S_IRUSR))) {
921 seen |= S_IRUSR;
922 if (type == ALLOW) {
923 mode |= S_IRUSR;
924 }
925 }
926 if ((access_mask & ACE_WRITE_DATA) &&
927 (!(seen & S_IWUSR))) {
928 seen |= S_IWUSR;
929 if (type == ALLOW) {
930 mode |= S_IWUSR;
931 }
932 }
933 if ((access_mask & ACE_EXECUTE) &&
934 (!(seen & S_IXUSR))) {
935 seen |= S_IXUSR;
936 if (type == ALLOW) {
937 mode |= S_IXUSR;
938 }
939 }
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))) {
944 seen |= S_IRGRP;
945 if (type == ALLOW) {
946 mode |= S_IRGRP;
947 }
948 }
949 if ((access_mask & ACE_WRITE_DATA) &&
950 (!(seen & S_IWGRP))) {
951 seen |= S_IWGRP;
952 if (type == ALLOW) {
953 mode |= S_IWGRP;
954 }
955 }
956 if ((access_mask & ACE_EXECUTE) &&
957 (!(seen & S_IXGRP))) {
958 seen |= S_IXGRP;
959 if (type == ALLOW) {
960 mode |= S_IXGRP;
961 }
962 }
963 } else if (entry_type == ACE_EVERYONE) {
964 if ((access_mask & ACE_READ_DATA)) {
965 if (!(seen & S_IRUSR)) {
966 seen |= S_IRUSR;
967 if (type == ALLOW) {
968 mode |= S_IRUSR;
969 }
970 }
971 if (!(seen & S_IRGRP)) {
972 seen |= S_IRGRP;
973 if (type == ALLOW) {
974 mode |= S_IRGRP;
975 }
976 }
977 if (!(seen & S_IROTH)) {
978 seen |= S_IROTH;
979 if (type == ALLOW) {
980 mode |= S_IROTH;
981 }
982 }
983 }
984 if ((access_mask & ACE_WRITE_DATA)) {
985 if (!(seen & S_IWUSR)) {
986 seen |= S_IWUSR;
987 if (type == ALLOW) {
988 mode |= S_IWUSR;
989 }
990 }
991 if (!(seen & S_IWGRP)) {
992 seen |= S_IWGRP;
993 if (type == ALLOW) {
994 mode |= S_IWGRP;
995 }
996 }
997 if (!(seen & S_IWOTH)) {
998 seen |= S_IWOTH;
999 if (type == ALLOW) {
1000 mode |= S_IWOTH;
1001 }
1002 }
1003 }
1004 if ((access_mask & ACE_EXECUTE)) {
1005 if (!(seen & S_IXUSR)) {
1006 seen |= S_IXUSR;
1007 if (type == ALLOW) {
1008 mode |= S_IXUSR;
1009 }
1010 }
1011 if (!(seen & S_IXGRP)) {
1012 seen |= S_IXGRP;
1013 if (type == ALLOW) {
1014 mode |= S_IXGRP;
1015 }
1016 }
1017 if (!(seen & S_IXOTH)) {
1018 seen |= S_IXOTH;
1019 if (type == ALLOW) {
1020 mode |= S_IXOTH;
1021 }
1022 }
1023 }
1024 } else {
1025 /*
1026 * Only care if this IDENTIFIER_GROUP or
1027 * USER ACE denies execute access to someone,
1028 * mode is not affected
1029 */
1030 if ((access_mask & ACE_EXECUTE) && type == DENY)
1031 an_exec_denied = B_TRUE;
1032 }
1033 }
1034
1035 /*
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.
1039 */
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;
1044
1045 if (an_exec_denied)
1046 *pflags &= ~ZFS_NO_EXECS_DENIED;
1047 else
1048 *pflags |= ZFS_NO_EXECS_DENIED;
1049
1050 return (mode);
1051 }
1052
1053 /*
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.
1056 */
1057 static int
1058 zfs_acl_node_read(znode_t *zp, boolean_t have_lock, zfs_acl_t **aclpp,
1059 boolean_t will_modify)
1060 {
1061 zfs_acl_t *aclp;
1062 int aclsize = 0;
1063 int acl_count = 0;
1064 zfs_acl_node_t *aclnode;
1065 zfs_acl_phys_t znode_acl;
1066 int version;
1067 int error;
1068 boolean_t drop_lock = B_FALSE;
1069
1070 ASSERT(MUTEX_HELD(&zp->z_acl_lock));
1071
1072 if (zp->z_acl_cached && !will_modify) {
1073 *aclpp = zp->z_acl_cached;
1074 return (0);
1075 }
1076
1077 /*
1078 * close race where znode could be upgrade while trying to
1079 * read the znode attributes.
1080 *
1081 * But this could only happen if the file isn't already an SA
1082 * znode
1083 */
1084 if (!zp->z_is_sa && !have_lock) {
1085 mutex_enter(&zp->z_lock);
1086 drop_lock = B_TRUE;
1087 }
1088 version = zfs_znode_acl_version(zp);
1089
1090 if ((error = zfs_acl_znode_info(zp, &aclsize,
1091 &acl_count, &znode_acl)) != 0) {
1092 goto done;
1093 }
1094
1095 aclp = zfs_acl_alloc(version);
1096
1097 aclp->z_acl_count = acl_count;
1098 aclp->z_acl_bytes = aclsize;
1099
1100 aclnode = zfs_acl_node_alloc(aclsize);
1101 aclnode->z_ace_count = aclp->z_acl_count;
1102 aclnode->z_size = aclsize;
1103
1104 if (!zp->z_is_sa) {
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);
1109 } else {
1110 bcopy(znode_acl.z_ace_data, aclnode->z_acldata,
1111 aclnode->z_size);
1112 }
1113 } else {
1114 error = sa_lookup(zp->z_sa_hdl, SA_ZPL_DACL_ACES(ZTOZSB(zp)),
1115 aclnode->z_acldata, aclnode->z_size);
1116 }
1117
1118 if (error != 0) {
1119 zfs_acl_free(aclp);
1120 zfs_acl_node_free(aclnode);
1121 /* convert checksum errors into IO errors */
1122 if (error == ECKSUM)
1123 error = SET_ERROR(EIO);
1124 goto done;
1125 }
1126
1127 list_insert_head(&aclp->z_acl, aclnode);
1128
1129 *aclpp = aclp;
1130 if (!will_modify)
1131 zp->z_acl_cached = aclp;
1132 done:
1133 if (drop_lock)
1134 mutex_exit(&zp->z_lock);
1135 return (error);
1136 }
1137
1138 /*ARGSUSED*/
1139 void
1140 zfs_acl_data_locator(void **dataptr, uint32_t *length, uint32_t buflen,
1141 boolean_t start, void *userdata)
1142 {
1143 zfs_acl_locator_cb_t *cb = (zfs_acl_locator_cb_t *)userdata;
1144
1145 if (start) {
1146 cb->cb_acl_node = list_head(&cb->cb_aclp->z_acl);
1147 } else {
1148 cb->cb_acl_node = list_next(&cb->cb_aclp->z_acl,
1149 cb->cb_acl_node);
1150 }
1151 *dataptr = cb->cb_acl_node->z_acldata;
1152 *length = cb->cb_acl_node->z_size;
1153 }
1154
1155 int
1156 zfs_acl_chown_setattr(znode_t *zp)
1157 {
1158 int error;
1159 zfs_acl_t *aclp;
1160
1161 if (ZTOZSB(zp)->z_acl_type == ZFS_ACLTYPE_POSIXACL)
1162 return (0);
1163
1164 ASSERT(MUTEX_HELD(&zp->z_lock));
1165 ASSERT(MUTEX_HELD(&zp->z_acl_lock));
1166
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));
1173
1174 /*
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
1180 * the caller.
1181 */
1182 if (error == ENOENT)
1183 return (0);
1184
1185 return (error);
1186 }
1187
1188 static void
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)
1191 {
1192 *deny1 = *deny2 = *allow0 = *group = 0;
1193
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;
1200
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;
1207
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;
1214
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;
1218 if (mode & S_IRUSR)
1219 *owner |= ACE_READ_DATA;
1220 if (mode & S_IWUSR)
1221 *owner |= ACE_WRITE_DATA|ACE_APPEND_DATA;
1222 if (mode & S_IXUSR)
1223 *owner |= ACE_EXECUTE;
1224
1225 *group = ACE_READ_ACL|ACE_READ_ATTRIBUTES| ACE_READ_NAMED_ATTRS|
1226 ACE_SYNCHRONIZE;
1227 if (mode & S_IRGRP)
1228 *group |= ACE_READ_DATA;
1229 if (mode & S_IWGRP)
1230 *group |= ACE_WRITE_DATA|ACE_APPEND_DATA;
1231 if (mode & S_IXGRP)
1232 *group |= ACE_EXECUTE;
1233
1234 *everyone = ACE_READ_ACL|ACE_READ_ATTRIBUTES| ACE_READ_NAMED_ATTRS|
1235 ACE_SYNCHRONIZE;
1236 if (mode & S_IROTH)
1237 *everyone |= ACE_READ_DATA;
1238 if (mode & S_IWOTH)
1239 *everyone |= ACE_WRITE_DATA|ACE_APPEND_DATA;
1240 if (mode & S_IXOTH)
1241 *everyone |= ACE_EXECUTE;
1242 }
1243
1244 /*
1245 * ace_trivial:
1246 * determine whether an ace_t acl is trivial
1247 *
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.
1252 */
1253 static int
1254 ace_trivial_common(void *acep, int aclcnt,
1255 uint64_t (*walk)(void *, uint64_t, int aclcnt,
1256 uint16_t *, uint16_t *, uint32_t *))
1257 {
1258 uint16_t flags;
1259 uint32_t mask;
1260 uint16_t type;
1261 uint64_t cookie = 0;
1262
1263 while ((cookie = walk(acep, cookie, aclcnt, &flags, &type, &mask))) {
1264 switch (flags & ACE_TYPE_FLAGS) {
1265 case ACE_OWNER:
1266 case ACE_GROUP|ACE_IDENTIFIER_GROUP:
1267 case ACE_EVERYONE:
1268 break;
1269 default:
1270 return (1);
1271 }
1272
1273 if (flags & (ACE_FILE_INHERIT_ACE|
1274 ACE_DIRECTORY_INHERIT_ACE|ACE_NO_PROPAGATE_INHERIT_ACE|
1275 ACE_INHERIT_ONLY_ACE))
1276 return (1);
1277
1278 /*
1279 * Special check for some special bits
1280 *
1281 * Don't allow anybody to deny reading basic
1282 * attributes or a files ACL.
1283 */
1284 if ((mask & (ACE_READ_ACL|ACE_READ_ATTRIBUTES)) &&
1285 (type == ACE_ACCESS_DENIED_ACE_TYPE))
1286 return (1);
1287
1288 /*
1289 * Delete permissions are never set by default
1290 */
1291 if (mask & (ACE_DELETE|ACE_DELETE_CHILD))
1292 return (1);
1293 /*
1294 * only allow owner@ to have
1295 * write_acl/write_owner/write_attributes/write_xattr/
1296 */
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))))
1301 return (1);
1302
1303 }
1304
1305 return (0);
1306 }
1307
1308 /*
1309 * common code for setting ACLs.
1310 *
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.
1314 */
1315 int
1316 zfs_aclset_common(znode_t *zp, zfs_acl_t *aclp, cred_t *cr, dmu_tx_t *tx)
1317 {
1318 int error;
1319 zfsvfs_t *zfsvfs = ZTOZSB(zp);
1320 dmu_object_type_t otype;
1321 zfs_acl_locator_cb_t locate = { 0 };
1322 uint64_t mode;
1323 sa_bulk_attr_t bulk[5];
1324 uint64_t ctime[2];
1325 int count = 0;
1326 zfs_acl_phys_t acl_phys;
1327
1328 mode = zp->z_mode;
1329
1330 mode = zfs_mode_compute(mode, aclp, &zp->z_pflags,
1331 KUID_TO_SUID(ZTOI(zp)->i_uid), KGID_TO_SGID(ZTOI(zp)->i_gid));
1332
1333 zp->z_mode = ZTOI(zp)->i_mode = mode;
1334 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs), NULL,
1335 &mode, sizeof (mode));
1336 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL,
1337 &zp->z_pflags, sizeof (zp->z_pflags));
1338 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL,
1339 &ctime, sizeof (ctime));
1340
1341 if (zp->z_acl_cached) {
1342 zfs_acl_free(zp->z_acl_cached);
1343 zp->z_acl_cached = NULL;
1344 }
1345
1346 /*
1347 * Upgrade needed?
1348 */
1349 if (!zfsvfs->z_use_fuids) {
1350 otype = DMU_OT_OLDACL;
1351 } else {
1352 if ((aclp->z_version == ZFS_ACL_VERSION_INITIAL) &&
1353 (zfsvfs->z_version >= ZPL_VERSION_FUID))
1354 zfs_acl_xform(zp, aclp, cr);
1355 ASSERT(aclp->z_version >= ZFS_ACL_VERSION_FUID);
1356 otype = DMU_OT_ACL;
1357 }
1358
1359 /*
1360 * Arrgh, we have to handle old on disk format
1361 * as well as newer (preferred) SA format.
1362 */
1363
1364 if (zp->z_is_sa) { /* the easy case, just update the ACL attribute */
1365 locate.cb_aclp = aclp;
1366 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_DACL_ACES(zfsvfs),
1367 zfs_acl_data_locator, &locate, aclp->z_acl_bytes);
1368 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_DACL_COUNT(zfsvfs),
1369 NULL, &aclp->z_acl_count, sizeof (uint64_t));
1370 } else { /* Painful legacy way */
1371 zfs_acl_node_t *aclnode;
1372 uint64_t off = 0;
1373 uint64_t aoid;
1374
1375 if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_ZNODE_ACL(zfsvfs),
1376 &acl_phys, sizeof (acl_phys))) != 0)
1377 return (error);
1378
1379 aoid = acl_phys.z_acl_extern_obj;
1380
1381 if (aclp->z_acl_bytes > ZFS_ACE_SPACE) {
1382 /*
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.
1386 */
1387 if (aoid &&
1388 aclp->z_version != acl_phys.z_acl_version) {
1389 error = dmu_object_free(zfsvfs->z_os, aoid, tx);
1390 if (error)
1391 return (error);
1392 aoid = 0;
1393 }
1394 if (aoid == 0) {
1395 aoid = dmu_object_alloc(zfsvfs->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);
1401 } else {
1402 (void) dmu_object_set_blocksize(zfsvfs->z_os,
1403 aoid, aclp->z_acl_bytes, 0, tx);
1404 }
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)
1409 continue;
1410 dmu_write(zfsvfs->z_os, aoid, off,
1411 aclnode->z_size, aclnode->z_acldata, tx);
1412 off += aclnode->z_size;
1413 }
1414 } else {
1415 void *start = acl_phys.z_ace_data;
1416 /*
1417 * Migrating back embedded?
1418 */
1419 if (acl_phys.z_acl_extern_obj) {
1420 error = dmu_object_free(zfsvfs->z_os,
1421 acl_phys.z_acl_extern_obj, tx);
1422 if (error)
1423 return (error);
1424 acl_phys.z_acl_extern_obj = 0;
1425 }
1426
1427 for (aclnode = list_head(&aclp->z_acl); aclnode;
1428 aclnode = list_next(&aclp->z_acl, aclnode)) {
1429 if (aclnode->z_ace_count == 0)
1430 continue;
1431 bcopy(aclnode->z_acldata, start,
1432 aclnode->z_size);
1433 start = (caddr_t)start + aclnode->z_size;
1434 }
1435 }
1436 /*
1437 * If Old version then swap count/bytes to match old
1438 * layout of znode_acl_phys_t.
1439 */
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;
1443 } else {
1444 acl_phys.z_acl_size = aclp->z_acl_bytes;
1445 acl_phys.z_acl_count = aclp->z_acl_count;
1446 }
1447 acl_phys.z_acl_version = aclp->z_version;
1448
1449 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ZNODE_ACL(zfsvfs), NULL,
1450 &acl_phys, sizeof (acl_phys));
1451 }
1452
1453 /*
1454 * Replace ACL wide bits, but first clear them.
1455 */
1456 zp->z_pflags &= ~ZFS_ACL_WIDE_FLAGS;
1457
1458 zp->z_pflags |= aclp->z_hints;
1459
1460 if (ace_trivial_common(aclp, 0, zfs_ace_walk) == 0)
1461 zp->z_pflags |= ZFS_ACL_TRIVIAL;
1462
1463 zfs_tstamp_update_setup(zp, STATE_CHANGED, NULL, ctime);
1464 return (sa_bulk_update(zp->z_sa_hdl, bulk, count, tx));
1465 }
1466
1467 static void
1468 zfs_acl_chmod(zfsvfs_t *zfsvfs, uint64_t mode, zfs_acl_t *aclp)
1469 {
1470 void *acep = NULL;
1471 uint64_t who;
1472 int new_count, new_bytes;
1473 int ace_size;
1474 int entry_type;
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();
1479 void *zacep;
1480 uint32_t owner, group, everyone;
1481 uint32_t deny1, deny2, allow0;
1482
1483 new_count = new_bytes = 0;
1484
1485 acl_trivial_access_masks((mode_t)mode, &allow0, &deny1, &deny2,
1486 &owner, &group, &everyone);
1487
1488 newnode = zfs_acl_node_alloc((abstract_size * 6) + aclp->z_acl_bytes);
1489
1490 zacep = newnode->z_acldata;
1491 if (allow0) {
1492 zfs_set_ace(aclp, zacep, allow0, ALLOW, -1, ACE_OWNER);
1493 zacep = (void *)((uintptr_t)zacep + abstract_size);
1494 new_count++;
1495 new_bytes += abstract_size;
1496 }
1497 if (deny1) {
1498 zfs_set_ace(aclp, zacep, deny1, DENY, -1, ACE_OWNER);
1499 zacep = (void *)((uintptr_t)zacep + abstract_size);
1500 new_count++;
1501 new_bytes += abstract_size;
1502 }
1503 if (deny2) {
1504 zfs_set_ace(aclp, zacep, deny2, DENY, -1, OWNING_GROUP);
1505 zacep = (void *)((uintptr_t)zacep + abstract_size);
1506 new_count++;
1507 new_bytes += abstract_size;
1508 }
1509
1510 while ((acep = zfs_acl_next_ace(aclp, acep, &who, &access_mask,
1511 &iflags, &type))) {
1512 uint16_t inherit_flags;
1513
1514 entry_type = (iflags & ACE_TYPE_FLAGS);
1515 inherit_flags = (iflags & ALL_INHERIT);
1516
1517 if ((entry_type == ACE_OWNER || entry_type == ACE_EVERYONE ||
1518 (entry_type == OWNING_GROUP)) &&
1519 ((inherit_flags & ACE_INHERIT_ONLY_ACE) == 0)) {
1520 continue;
1521 }
1522
1523 if ((type != ALLOW && type != DENY) ||
1524 (inherit_flags & ACE_INHERIT_ONLY_ACE)) {
1525 if (inherit_flags)
1526 aclp->z_hints |= ZFS_INHERIT_ACE;
1527 switch (type) {
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;
1533 break;
1534 }
1535 } else {
1536
1537 /*
1538 * Limit permissions to be no greater than
1539 * group permissions
1540 */
1541 if (zfsvfs->z_acl_inherit == ZFS_ACL_RESTRICTED) {
1542 if (!(mode & S_IRGRP))
1543 access_mask &= ~ACE_READ_DATA;
1544 if (!(mode & S_IWGRP))
1545 access_mask &=
1546 ~(ACE_WRITE_DATA|ACE_APPEND_DATA);
1547 if (!(mode & S_IXGRP))
1548 access_mask &= ~ACE_EXECUTE;
1549 access_mask &=
1550 ~(ACE_WRITE_OWNER|ACE_WRITE_ACL|
1551 ACE_WRITE_ATTRIBUTES|ACE_WRITE_NAMED_ATTRS);
1552 }
1553 }
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);
1557 new_count++;
1558 new_bytes += ace_size;
1559 }
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);
1565
1566 new_count += 3;
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);
1574 }
1575
1576 void
1577 zfs_acl_chmod_setattr(znode_t *zp, zfs_acl_t **aclp, uint64_t mode)
1578 {
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);
1586 ASSERT(*aclp);
1587 }
1588
1589 /*
1590 * strip off write_owner and write_acl
1591 */
1592 static void
1593 zfs_restricted_update(zfsvfs_t *zfsvfs, zfs_acl_t *aclp, void *acep)
1594 {
1595 uint32_t mask = aclp->z_ops->ace_mask_get(acep);
1596
1597 if ((zfsvfs->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);
1601 }
1602 }
1603
1604 /*
1605 * Should ACE be inherited?
1606 */
1607 static int
1608 zfs_ace_can_use(umode_t obj_mode, uint16_t acep_flags)
1609 {
1610 int iflags = (acep_flags & 0xf);
1611
1612 if (S_ISDIR(obj_mode) && (iflags & ACE_DIRECTORY_INHERIT_ACE))
1613 return (1);
1614 else if (iflags & ACE_FILE_INHERIT_ACE)
1615 return (!(S_ISDIR(obj_mode) &&
1616 (iflags & ACE_NO_PROPAGATE_INHERIT_ACE)));
1617 return (0);
1618 }
1619
1620 /*
1621 * inherit inheritable ACEs from parent
1622 */
1623 static zfs_acl_t *
1624 zfs_acl_inherit(zfsvfs_t *zfsvfs, umode_t obj_mode, zfs_acl_t *paclp,
1625 uint64_t mode, boolean_t *need_chmod)
1626 {
1627 void *pacep;
1628 void *acep;
1629 zfs_acl_node_t *aclnode;
1630 zfs_acl_t *aclp = NULL;
1631 uint64_t who;
1632 uint32_t access_mask;
1633 uint16_t iflags, newflags, type;
1634 size_t ace_size;
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;
1640
1641 passthrough_x =
1642 zfsvfs->z_acl_inherit == ZFS_ACL_PASSTHROUGH_X;
1643 passthrough = passthrough_x ||
1644 zfsvfs->z_acl_inherit == ZFS_ACL_PASSTHROUGH;
1645 noallow =
1646 zfsvfs->z_acl_inherit == ZFS_ACL_NOALLOW;
1647
1648 *need_chmod = B_TRUE;
1649 pacep = NULL;
1650 aclp = zfs_acl_alloc(paclp->z_version);
1651 if (zfsvfs->z_acl_inherit == ZFS_ACL_DISCARD || S_ISLNK(obj_mode))
1652 return (aclp);
1653 while ((pacep = zfs_acl_next_ace(paclp, pacep, &who,
1654 &access_mask, &iflags, &type))) {
1655
1656 /*
1657 * don't inherit bogus ACEs
1658 */
1659 if (!zfs_acl_valid_ace_type(type, iflags))
1660 continue;
1661
1662 if (noallow && type == ALLOW)
1663 continue;
1664
1665 ace_size = aclp->z_ops->ace_size(pacep);
1666
1667 if (!zfs_ace_can_use(obj_mode, iflags))
1668 continue;
1669
1670 /*
1671 * If owner@, group@, or everyone@ inheritable
1672 * then zfs_acl_chmod() isn't needed.
1673 */
1674 if (passthrough &&
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;
1680 }
1681
1682 if (!vdir && passthrough_x &&
1683 ((mode & (S_IXUSR | S_IXGRP | S_IXOTH)) == 0)) {
1684 access_mask &= ~ACE_EXECUTE;
1685 }
1686
1687 aclnode = zfs_acl_node_alloc(ace_size);
1688 list_insert_tail(&aclp->z_acl, aclnode);
1689 acep = aclnode->z_acldata;
1690
1691 zfs_set_ace(aclp, acep, access_mask, type,
1692 who, iflags|ACE_INHERITED_ACE);
1693
1694 /*
1695 * Copy special opaque data if any
1696 */
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);
1701 }
1702
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);
1707
1708 if (vdir)
1709 aclp->z_hints |= ZFS_INHERIT_ACE;
1710
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(zfsvfs, aclp, acep);
1716 continue;
1717 }
1718
1719 ASSERT(vdir);
1720
1721 /*
1722 * If only FILE_INHERIT is set then turn on
1723 * inherit_only
1724 */
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);
1730 } else {
1731 newflags &= ~ACE_INHERIT_ONLY_ACE;
1732 aclp->z_ops->ace_flags_set(acep,
1733 newflags|ACE_INHERITED_ACE);
1734 }
1735 }
1736 return (aclp);
1737 }
1738
1739 /*
1740 * Create file system object initial permissions
1741 * including inheritable ACEs.
1742 */
1743 int
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)
1746 {
1747 int error;
1748 zfsvfs_t *zfsvfs = ZTOZSB(dzp);
1749 zfs_acl_t *paclp;
1750 gid_t gid = vap->va_gid;
1751 boolean_t need_chmod = B_TRUE;
1752 boolean_t inherited = B_FALSE;
1753
1754 bzero(acl_ids, sizeof (zfs_acl_ids_t));
1755 acl_ids->z_mode = vap->va_mode;
1756
1757 if (vsecp)
1758 if ((error = zfs_vsec_2_aclp(zfsvfs, vap->va_mode, vsecp,
1759 cr, &acl_ids->z_fuidp, &acl_ids->z_aclp)) != 0)
1760 return (error);
1761
1762 acl_ids->z_fuid = vap->va_uid;
1763 acl_ids->z_fgid = vap->va_gid;
1764 #ifdef HAVE_KSID
1765 /*
1766 * Determine uid and gid.
1767 */
1768 if ((flag & IS_ROOT_NODE) || zfsvfs->z_replay ||
1769 ((flag & IS_XATTR) && (S_ISDIR(vap->va_mode)))) {
1770 acl_ids->z_fuid = zfs_fuid_create(zfsvfs, (uint64_t)vap->va_uid,
1771 cr, ZFS_OWNER, &acl_ids->z_fuidp);
1772 acl_ids->z_fgid = zfs_fuid_create(zfsvfs, (uint64_t)vap->va_gid,
1773 cr, ZFS_GROUP, &acl_ids->z_fuidp);
1774 gid = vap->va_gid;
1775 } else {
1776 acl_ids->z_fuid = zfs_fuid_create_cred(zfsvfs, 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(zfsvfs,
1781 (uint64_t)vap->va_gid,
1782 cr, ZFS_GROUP, &acl_ids->z_fuidp);
1783 gid = vap->va_gid;
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;
1788 }
1789 if (acl_ids->z_fgid == 0) {
1790 if (dzp->z_mode & S_ISGID) {
1791 char *domain;
1792 uint32_t rid;
1793
1794 acl_ids->z_fgid = KGID_TO_SGID(
1795 ZTOI(dzp)->i_gid);
1796 gid = zfs_fuid_map_id(zfsvfs, acl_ids->z_fgid,
1797 cr, ZFS_GROUP);
1798
1799 if (zfsvfs->z_use_fuids &&
1800 IS_EPHEMERAL(acl_ids->z_fgid)) {
1801 domain = zfs_fuid_idx_domain(
1802 &zfsvfs->z_fuid_idx,
1803 FUID_INDEX(acl_ids->z_fgid));
1804 rid = FUID_RID(acl_ids->z_fgid);
1805 zfs_fuid_node_add(&acl_ids->z_fuidp,
1806 domain, rid,
1807 FUID_INDEX(acl_ids->z_fgid),
1808 acl_ids->z_fgid, ZFS_GROUP);
1809 }
1810 } else {
1811 acl_ids->z_fgid = zfs_fuid_create_cred(zfsvfs,
1812 ZFS_GROUP, cr, &acl_ids->z_fuidp);
1813 gid = crgetgid(cr);
1814 }
1815 }
1816 }
1817 #endif /* HAVE_KSID */
1818
1819 /*
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.
1824 */
1825
1826 if (!(flag & IS_ROOT_NODE) && (dzp->z_mode & S_ISGID) &&
1827 (S_ISDIR(vap->va_mode))) {
1828 acl_ids->z_mode |= S_ISGID;
1829 } else {
1830 if ((acl_ids->z_mode & S_ISGID) &&
1831 secpolicy_vnode_setids_setgids(cr, gid) != 0)
1832 acl_ids->z_mode &= ~S_ISGID;
1833 }
1834
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,
1842 &paclp, B_FALSE));
1843 acl_ids->z_aclp = zfs_acl_inherit(zfsvfs,
1844 vap->va_mode, paclp, acl_ids->z_mode, &need_chmod);
1845 inherited = B_TRUE;
1846 } else {
1847 acl_ids->z_aclp =
1848 zfs_acl_alloc(zfs_acl_version_zp(dzp));
1849 acl_ids->z_aclp->z_hints |= ZFS_ACL_TRIVIAL;
1850 }
1851 mutex_exit(&dzp->z_lock);
1852 mutex_exit(&dzp->z_acl_lock);
1853 if (need_chmod) {
1854 acl_ids->z_aclp->z_hints |= S_ISDIR(vap->va_mode) ?
1855 ZFS_ACL_AUTO_INHERIT : 0;
1856 zfs_acl_chmod(zfsvfs, acl_ids->z_mode, acl_ids->z_aclp);
1857 }
1858 }
1859
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;
1866 }
1867
1868 return (0);
1869 }
1870
1871 /*
1872 * Free ACL and fuid_infop, but not the acl_ids structure
1873 */
1874 void
1875 zfs_acl_ids_free(zfs_acl_ids_t *acl_ids)
1876 {
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;
1883 }
1884
1885 boolean_t
1886 zfs_acl_ids_overquota(zfsvfs_t *zfsvfs, zfs_acl_ids_t *acl_ids)
1887 {
1888 return (zfs_fuid_overquota(zfsvfs, B_FALSE, acl_ids->z_fuid) ||
1889 zfs_fuid_overquota(zfsvfs, B_TRUE, acl_ids->z_fgid) ||
1890 zfs_fuid_overobjquota(zfsvfs, B_FALSE, acl_ids->z_fuid) ||
1891 zfs_fuid_overobjquota(zfsvfs, B_TRUE, acl_ids->z_fgid));
1892 }
1893
1894 /*
1895 * Retrieve a file's ACL
1896 */
1897 int
1898 zfs_getacl(znode_t *zp, vsecattr_t *vsecp, boolean_t skipaclchk, cred_t *cr)
1899 {
1900 zfs_acl_t *aclp;
1901 ulong_t mask;
1902 int error;
1903 int count = 0;
1904 int largeace = 0;
1905
1906 mask = vsecp->vsa_mask & (VSA_ACE | VSA_ACECNT |
1907 VSA_ACE_ACLFLAGS | VSA_ACE_ALLTYPES);
1908
1909 if (mask == 0)
1910 return (SET_ERROR(ENOSYS));
1911
1912 if ((error = zfs_zaccess(zp, ACE_READ_ACL, 0, skipaclchk, cr)))
1913 return (error);
1914
1915 mutex_enter(&zp->z_acl_lock);
1916
1917 error = zfs_acl_node_read(zp, B_FALSE, &aclp, B_FALSE);
1918 if (error != 0) {
1919 mutex_exit(&zp->z_acl_lock);
1920 return (error);
1921 }
1922
1923 /*
1924 * Scan ACL to determine number of ACEs
1925 */
1926 if ((zp->z_pflags & ZFS_ACL_OBJ_ACE) && !(mask & VSA_ACE_ALLTYPES)) {
1927 void *zacep = NULL;
1928 uint64_t who;
1929 uint32_t access_mask;
1930 uint16_t type, iflags;
1931
1932 while ((zacep = zfs_acl_next_ace(aclp, zacep,
1933 &who, &access_mask, &iflags, &type))) {
1934 switch (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:
1939 largeace++;
1940 continue;
1941 default:
1942 count++;
1943 }
1944 }
1945 vsecp->vsa_aclcnt = count;
1946 } else
1947 count = (int)aclp->z_acl_count;
1948
1949 if (mask & VSA_ACECNT) {
1950 vsecp->vsa_aclcnt = count;
1951 }
1952
1953 if (mask & VSA_ACE) {
1954 size_t aclsz;
1955
1956 aclsz = count * sizeof (ace_t) +
1957 sizeof (ace_object_t) * largeace;
1958
1959 vsecp->vsa_aclentp = kmem_alloc(aclsz, KM_SLEEP);
1960 vsecp->vsa_aclentsz = aclsz;
1961
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));
1965 else {
1966 zfs_acl_node_t *aclnode;
1967 void *start = vsecp->vsa_aclentp;
1968
1969 for (aclnode = list_head(&aclp->z_acl); aclnode;
1970 aclnode = list_next(&aclp->z_acl, aclnode)) {
1971 bcopy(aclnode->z_acldata, start,
1972 aclnode->z_size);
1973 start = (caddr_t)start + aclnode->z_size;
1974 }
1975 ASSERT((caddr_t)start - (caddr_t)vsecp->vsa_aclentp ==
1976 aclp->z_acl_bytes);
1977 }
1978 }
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;
1987 }
1988
1989 mutex_exit(&zp->z_acl_lock);
1990
1991 return (0);
1992 }
1993
1994 int
1995 zfs_vsec_2_aclp(zfsvfs_t *zfsvfs, umode_t obj_mode,
1996 vsecattr_t *vsecp, cred_t *cr, zfs_fuid_info_t **fuidp, zfs_acl_t **zaclp)
1997 {
1998 zfs_acl_t *aclp;
1999 zfs_acl_node_t *aclnode;
2000 int aclcnt = vsecp->vsa_aclcnt;
2001 int error;
2002
2003 if (vsecp->vsa_aclcnt > MAX_ACL_ENTRIES || vsecp->vsa_aclcnt <= 0)
2004 return (SET_ERROR(EINVAL));
2005
2006 aclp = zfs_acl_alloc(zfs_acl_version(zfsvfs->z_version));
2007
2008 aclp->z_hints = 0;
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) {
2014 zfs_acl_free(aclp);
2015 zfs_acl_node_free(aclnode);
2016 return (error);
2017 }
2018 } else {
2019 if ((error = zfs_copy_ace_2_fuid(zfsvfs, obj_mode, aclp,
2020 vsecp->vsa_aclentp, aclnode->z_acldata, aclcnt,
2021 &aclnode->z_size, fuidp, cr)) != 0) {
2022 zfs_acl_free(aclp);
2023 zfs_acl_node_free(aclnode);
2024 return (error);
2025 }
2026 }
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);
2031
2032 /*
2033 * If flags are being set then add them to z_hints
2034 */
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;
2042 }
2043
2044 *zaclp = aclp;
2045
2046 return (0);
2047 }
2048
2049 /*
2050 * Set a file's ACL
2051 */
2052 int
2053 zfs_setacl(znode_t *zp, vsecattr_t *vsecp, boolean_t skipaclchk, cred_t *cr)
2054 {
2055 zfsvfs_t *zfsvfs = ZTOZSB(zp);
2056 zilog_t *zilog = zfsvfs->z_log;
2057 ulong_t mask = vsecp->vsa_mask & (VSA_ACE | VSA_ACECNT);
2058 dmu_tx_t *tx;
2059 int error;
2060 zfs_acl_t *aclp;
2061 zfs_fuid_info_t *fuidp = NULL;
2062 boolean_t fuid_dirtied;
2063 uint64_t acl_obj;
2064
2065 if (mask == 0)
2066 return (SET_ERROR(ENOSYS));
2067
2068 if (zp->z_pflags & ZFS_IMMUTABLE)
2069 return (SET_ERROR(EPERM));
2070
2071 if ((error = zfs_zaccess(zp, ACE_WRITE_ACL, 0, skipaclchk, cr)))
2072 return (error);
2073
2074 error = zfs_vsec_2_aclp(zfsvfs, ZTOI(zp)->i_mode, vsecp, cr, &fuidp,
2075 &aclp);
2076 if (error)
2077 return (error);
2078
2079 /*
2080 * If ACL wide flags aren't being set then preserve any
2081 * existing flags.
2082 */
2083 if (!(vsecp->vsa_mask & VSA_ACE_ACLFLAGS)) {
2084 aclp->z_hints |=
2085 (zp->z_pflags & V4_ACL_WIDE_FLAGS);
2086 }
2087 top:
2088 mutex_enter(&zp->z_acl_lock);
2089 mutex_enter(&zp->z_lock);
2090
2091 tx = dmu_tx_create(zfsvfs->z_os);
2092
2093 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE);
2094
2095 fuid_dirtied = zfsvfs->z_fuid_dirty;
2096 if (fuid_dirtied)
2097 zfs_fuid_txhold(zfsvfs, tx);
2098
2099 /*
2100 * If old version and ACL won't fit in bonus and we aren't
2101 * upgrading then take out necessary DMU holds
2102 */
2103
2104 if ((acl_obj = zfs_external_acl(zp)) != 0) {
2105 if (zfsvfs->z_version >= ZPL_VERSION_FUID &&
2106 zfs_znode_acl_version(zp) <= ZFS_ACL_VERSION_INITIAL) {
2107 dmu_tx_hold_free(tx, acl_obj, 0,
2108 DMU_OBJECT_END);
2109 dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0,
2110 aclp->z_acl_bytes);
2111 } else {
2112 dmu_tx_hold_write(tx, acl_obj, 0, aclp->z_acl_bytes);
2113 }
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);
2116 }
2117
2118 zfs_sa_upgrade_txholds(tx, zp);
2119 error = dmu_tx_assign(tx, TXG_NOWAIT);
2120 if (error) {
2121 mutex_exit(&zp->z_acl_lock);
2122 mutex_exit(&zp->z_lock);
2123
2124 if (error == ERESTART) {
2125 dmu_tx_wait(tx);
2126 dmu_tx_abort(tx);
2127 goto top;
2128 }
2129 dmu_tx_abort(tx);
2130 zfs_acl_free(aclp);
2131 return (error);
2132 }
2133
2134 error = zfs_aclset_common(zp, aclp, cr, tx);
2135 ASSERT(error == 0);
2136 ASSERT(zp->z_acl_cached == NULL);
2137 zp->z_acl_cached = aclp;
2138
2139 if (fuid_dirtied)
2140 zfs_fuid_sync(zfsvfs, tx);
2141
2142 zfs_log_acl(zilog, tx, zp, vsecp, fuidp);
2143
2144 if (fuidp)
2145 zfs_fuid_info_free(fuidp);
2146 dmu_tx_commit(tx);
2147
2148 mutex_exit(&zp->z_lock);
2149 mutex_exit(&zp->z_acl_lock);
2150
2151 return (error);
2152 }
2153
2154 /*
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.
2158 */
2159 static int
2160 zfs_zaccess_dataset_check(znode_t *zp, uint32_t v4_mode)
2161 {
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));
2166 }
2167
2168 /*
2169 * Only check for READONLY on non-directories.
2170 */
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));
2177 }
2178
2179 if ((v4_mode & (ACE_DELETE | ACE_DELETE_CHILD)) &&
2180 (zp->z_pflags & ZFS_NOUNLINK)) {
2181 return (SET_ERROR(EPERM));
2182 }
2183
2184 if (((v4_mode & (ACE_READ_DATA|ACE_EXECUTE)) &&
2185 (zp->z_pflags & ZFS_AV_QUARANTINED))) {
2186 return (SET_ERROR(EACCES));
2187 }
2188
2189 return (0);
2190 }
2191
2192 /*
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)
2209 *
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.
2216 */
2217 static int
2218 zfs_zaccess_aces_check(znode_t *zp, uint32_t *working_mode,
2219 boolean_t anyaccess, cred_t *cr)
2220 {
2221 zfsvfs_t *zfsvfs = ZTOZSB(zp);
2222 zfs_acl_t *aclp;
2223 int error;
2224 uid_t uid = crgetuid(cr);
2225 uint64_t who;
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;
2231 boolean_t checkit;
2232 uid_t gowner;
2233 uid_t fowner;
2234
2235 zfs_fuid_map_ids(zp, cr, &fowner, &gowner);
2236
2237 mutex_enter(&zp->z_acl_lock);
2238
2239 error = zfs_acl_node_read(zp, B_FALSE, &aclp, B_FALSE);
2240 if (error != 0) {
2241 mutex_exit(&zp->z_acl_lock);
2242 return (error);
2243 }
2244
2245 ASSERT(zp->z_acl_cached);
2246
2247 while ((acep = zfs_acl_next_ace(aclp, acep, &who, &access_mask,
2248 &iflags, &type))) {
2249 uint32_t mask_matched;
2250
2251 if (!zfs_acl_valid_ace_type(type, iflags))
2252 continue;
2253
2254 if (S_ISDIR(ZTOI(zp)->i_mode) &&
2255 (iflags & ACE_INHERIT_ONLY_ACE))
2256 continue;
2257
2258 /* Skip ACE if it does not affect any AoI */
2259 mask_matched = (access_mask & *working_mode);
2260 if (!mask_matched)
2261 continue;
2262
2263 entry_type = (iflags & ACE_TYPE_FLAGS);
2264
2265 checkit = B_FALSE;
2266
2267 switch (entry_type) {
2268 case ACE_OWNER:
2269 if (uid == fowner)
2270 checkit = B_TRUE;
2271 break;
2272 case OWNING_GROUP:
2273 who = gowner;
2274 /*FALLTHROUGH*/
2275 case ACE_IDENTIFIER_GROUP:
2276 checkit = zfs_groupmember(zfsvfs, who, cr);
2277 break;
2278 case ACE_EVERYONE:
2279 checkit = B_TRUE;
2280 break;
2281
2282 /* USER Entry */
2283 default:
2284 if (entry_type == 0) {
2285 uid_t newid;
2286
2287 newid = zfs_fuid_map_id(zfsvfs, who, cr,
2288 ZFS_ACE_USER);
2289 if (newid != IDMAP_WK_CREATOR_OWNER_UID &&
2290 uid == newid)
2291 checkit = B_TRUE;
2292 break;
2293 } else {
2294 mutex_exit(&zp->z_acl_lock);
2295 return (SET_ERROR(EIO));
2296 }
2297 }
2298
2299 if (checkit) {
2300 if (type == DENY) {
2301 DTRACE_PROBE3(zfs__ace__denies,
2302 znode_t *, zp,
2303 zfs_ace_hdr_t *, acep,
2304 uint32_t, mask_matched);
2305 deny_mask |= mask_matched;
2306 } else {
2307 DTRACE_PROBE3(zfs__ace__allows,
2308 znode_t *, zp,
2309 zfs_ace_hdr_t *, acep,
2310 uint32_t, mask_matched);
2311 if (anyaccess) {
2312 mutex_exit(&zp->z_acl_lock);
2313 return (0);
2314 }
2315 }
2316 *working_mode &= ~mask_matched;
2317 }
2318
2319 /* Are we done? */
2320 if (*working_mode == 0)
2321 break;
2322 }
2323
2324 mutex_exit(&zp->z_acl_lock);
2325
2326 /* Put the found 'denies' back on the working mode */
2327 if (deny_mask) {
2328 *working_mode |= deny_mask;
2329 return (SET_ERROR(EACCES));
2330 } else if (*working_mode) {
2331 return (-1);
2332 }
2333
2334 return (0);
2335 }
2336
2337 /*
2338 * Return true if any access whatsoever granted, we don't actually
2339 * care what access is granted.
2340 */
2341 boolean_t
2342 zfs_has_access(znode_t *zp, cred_t *cr)
2343 {
2344 uint32_t have = ACE_ALL_PERMS;
2345
2346 if (zfs_zaccess_aces_check(zp, &have, B_TRUE, cr) != 0) {
2347 uid_t owner;
2348
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);
2352 }
2353 return (B_TRUE);
2354 }
2355
2356 static int
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)
2359 {
2360 zfsvfs_t *zfsvfs = ZTOZSB(zp);
2361 int err;
2362
2363 *working_mode = v4_mode;
2364 *check_privs = B_TRUE;
2365
2366 /*
2367 * Short circuit empty requests
2368 */
2369 if (v4_mode == 0 || zfsvfs->z_replay) {
2370 *working_mode = 0;
2371 return (0);
2372 }
2373
2374 if ((err = zfs_zaccess_dataset_check(zp, v4_mode)) != 0) {
2375 *check_privs = B_FALSE;
2376 return (err);
2377 }
2378
2379 /*
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.
2383 */
2384 if (skipaclchk) {
2385 *working_mode = 0;
2386 return (0);
2387 }
2388
2389 return (zfs_zaccess_aces_check(zp, working_mode, B_FALSE, cr));
2390 }
2391
2392 static int
2393 zfs_zaccess_append(znode_t *zp, uint32_t *working_mode, boolean_t *check_privs,
2394 cred_t *cr)
2395 {
2396 if (*working_mode != ACE_WRITE_DATA)
2397 return (SET_ERROR(EACCES));
2398
2399 return (zfs_zaccess_common(zp, ACE_APPEND_DATA, working_mode,
2400 check_privs, B_FALSE, cr));
2401 }
2402
2403 int
2404 zfs_fastaccesschk_execute(znode_t *zdp, cred_t *cr)
2405 {
2406 boolean_t owner = B_FALSE;
2407 boolean_t groupmbr = B_FALSE;
2408 boolean_t is_attr;
2409 uid_t uid = crgetuid(cr);
2410 int error;
2411
2412 if (zdp->z_pflags & ZFS_AV_QUARANTINED)
2413 return (SET_ERROR(EACCES));
2414
2415 is_attr = ((zdp->z_pflags & ZFS_XATTR) &&
2416 (S_ISDIR(ZTOI(zdp)->i_mode)));
2417 if (is_attr)
2418 goto slow;
2419
2420
2421 mutex_enter(&zdp->z_acl_lock);
2422
2423 if (zdp->z_pflags & ZFS_NO_EXECS_DENIED) {
2424 mutex_exit(&zdp->z_acl_lock);
2425 return (0);
2426 }
2427
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);
2431 goto slow;
2432 }
2433
2434 if (uid == KUID_TO_SUID(ZTOI(zdp)->i_uid)) {
2435 owner = B_TRUE;
2436 if (zdp->z_mode & S_IXUSR) {
2437 mutex_exit(&zdp->z_acl_lock);
2438 return (0);
2439 } else {
2440 mutex_exit(&zdp->z_acl_lock);
2441 goto slow;
2442 }
2443 }
2444 if (groupmember(KGID_TO_SGID(ZTOI(zdp)->i_gid), cr)) {
2445 groupmbr = B_TRUE;
2446 if (zdp->z_mode & S_IXGRP) {
2447 mutex_exit(&zdp->z_acl_lock);
2448 return (0);
2449 } else {
2450 mutex_exit(&zdp->z_acl_lock);
2451 goto slow;
2452 }
2453 }
2454 if (!owner && !groupmbr) {
2455 if (zdp->z_mode & S_IXOTH) {
2456 mutex_exit(&zdp->z_acl_lock);
2457 return (0);
2458 }
2459 }
2460
2461 mutex_exit(&zdp->z_acl_lock);
2462
2463 slow:
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));
2468 return (error);
2469 }
2470
2471 /*
2472 * Determine whether Access should be granted/denied.
2473 *
2474 * The least priv subsytem is always consulted as a basic privilege
2475 * can define any form of access.
2476 */
2477 int
2478 zfs_zaccess(znode_t *zp, int mode, int flags, boolean_t skipaclchk, cred_t *cr)
2479 {
2480 uint32_t working_mode;
2481 int error;
2482 int is_attr;
2483 boolean_t check_privs;
2484 znode_t *xzp;
2485 znode_t *check_zp = zp;
2486 mode_t needed_bits;
2487 uid_t owner;
2488
2489 is_attr = ((zp->z_pflags & ZFS_XATTR) && S_ISDIR(ZTOI(zp)->i_mode));
2490
2491 /*
2492 * If attribute then validate against base file
2493 */
2494 if (is_attr) {
2495 if ((error = zfs_zget(ZTOZSB(zp),
2496 zp->z_xattr_parent, &xzp)) != 0) {
2497 return (error);
2498 }
2499
2500 check_zp = xzp;
2501
2502 /*
2503 * fixup mode to map to xattr perms
2504 */
2505
2506 if (mode & (ACE_WRITE_DATA|ACE_APPEND_DATA)) {
2507 mode &= ~(ACE_WRITE_DATA|ACE_APPEND_DATA);
2508 mode |= ACE_WRITE_NAMED_ATTRS;
2509 }
2510
2511 if (mode & (ACE_READ_DATA|ACE_EXECUTE)) {
2512 mode &= ~(ACE_READ_DATA|ACE_EXECUTE);
2513 mode |= ACE_READ_NAMED_ATTRS;
2514 }
2515 }
2516
2517 owner = zfs_fuid_map_id(ZTOZSB(zp), KUID_TO_SUID(ZTOI(zp)->i_uid),
2518 cr, ZFS_OWNER);
2519 /*
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),
2524 * needed_bits.
2525 */
2526 needed_bits = 0;
2527
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);
2532
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;
2541
2542 if ((error = zfs_zaccess_common(check_zp, mode, &working_mode,
2543 &check_privs, skipaclchk, cr)) == 0) {
2544 if (is_attr)
2545 iput(ZTOI(xzp));
2546 return (secpolicy_vnode_access2(cr, ZTOI(zp), owner,
2547 needed_bits, needed_bits));
2548 }
2549
2550 if (error && !check_privs) {
2551 if (is_attr)
2552 iput(ZTOI(xzp));
2553 return (error);
2554 }
2555
2556 if (error && (flags & V_APPEND)) {
2557 error = zfs_zaccess_append(zp, &working_mode, &check_privs, cr);
2558 }
2559
2560 if (error && check_privs) {
2561 mode_t checkmode = 0;
2562
2563 /*
2564 * First check for implicit owner permission on
2565 * read_acl/read_attributes
2566 */
2567
2568 error = 0;
2569 ASSERT(working_mode != 0);
2570
2571 if ((working_mode & (ACE_READ_ACL|ACE_READ_ATTRIBUTES) &&
2572 owner == crgetuid(cr)))
2573 working_mode &= ~(ACE_READ_ACL|ACE_READ_ATTRIBUTES);
2574
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;
2583
2584 error = secpolicy_vnode_access2(cr, ZTOI(check_zp), owner,
2585 needed_bits & ~checkmode, needed_bits);
2586
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);
2591
2592 if (error == 0 && (working_mode &
2593 (ACE_DELETE|ACE_DELETE_CHILD)))
2594 error = secpolicy_vnode_remove(cr);
2595
2596 if (error == 0 && (working_mode & ACE_SYNCHRONIZE)) {
2597 error = secpolicy_vnode_chown(cr, owner);
2598 }
2599 if (error == 0) {
2600 /*
2601 * See if any bits other than those already checked
2602 * for are still present. If so then return EACCES
2603 */
2604 if (working_mode & ~(ZFS_CHECKED_MASKS)) {
2605 error = SET_ERROR(EACCES);
2606 }
2607 }
2608 } else if (error == 0) {
2609 error = secpolicy_vnode_access2(cr, ZTOI(zp), owner,
2610 needed_bits, needed_bits);
2611 }
2612
2613 if (is_attr)
2614 iput(ZTOI(xzp));
2615
2616 return (error);
2617 }
2618
2619 /*
2620 * Translate traditional unix S_IRUSR/S_IWUSR/S_IXUSR mode into
2621 * native ACL format and call zfs_zaccess()
2622 */
2623 int
2624 zfs_zaccess_rwx(znode_t *zp, mode_t mode, int flags, cred_t *cr)
2625 {
2626 return (zfs_zaccess(zp, zfs_unix_to_v4(mode >> 6), flags, B_FALSE, cr));
2627 }
2628
2629 /*
2630 * Access function for secpolicy_vnode_setattr
2631 */
2632 int
2633 zfs_zaccess_unix(znode_t *zp, mode_t mode, cred_t *cr)
2634 {
2635 int v4_mode = zfs_unix_to_v4(mode >> 6);
2636
2637 return (zfs_zaccess(zp, v4_mode, 0, B_FALSE, cr));
2638 }
2639
2640 static int
2641 zfs_delete_final_check(znode_t *zp, znode_t *dzp,
2642 mode_t available_perms, cred_t *cr)
2643 {
2644 int error;
2645 uid_t downer;
2646
2647 downer = zfs_fuid_map_id(ZTOZSB(dzp), KUID_TO_SUID(ZTOI(dzp)->i_uid),
2648 cr, ZFS_OWNER);
2649
2650 error = secpolicy_vnode_access2(cr, ZTOI(dzp),
2651 downer, available_perms, S_IWUSR|S_IXUSR);
2652
2653 if (error == 0)
2654 error = zfs_sticky_remove_access(dzp, zp, cr);
2655
2656 return (error);
2657 }
2658
2659 /*
2660 * Determine whether Access should be granted/deny, without
2661 * consulting least priv subsystem.
2662 *
2663 * The following chart is the recommended NFSv4 enforcement for
2664 * ability to delete an object.
2665 *
2666 * -------------------------------------------------------
2667 * | Parent Dir | Target Object Permissions |
2668 * | 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 | | | |
2682 * | execute | | | |
2683 * -------------------------------------------------------
2684 * | ACL denies | | | |
2685 * | write and | Permit | Deny | Deny |
2686 * | execute | | | |
2687 * -------------------------------------------------------
2688 * ^
2689 * |
2690 * No search privilege, can't even look up file?
2691 *
2692 */
2693 int
2694 zfs_zaccess_delete(znode_t *dzp, znode_t *zp, cred_t *cr)
2695 {
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;
2702
2703 /*
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
2708 *
2709 * However, deny permissions may ultimately be overridden
2710 * by secpolicy_vnode_access().
2711 *
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.
2715 */
2716
2717 if (zp->z_pflags & (ZFS_IMMUTABLE | ZFS_NOUNLINK))
2718 return (SET_ERROR(EPERM));
2719
2720 /*
2721 * First row
2722 * If the directory permissions allow the delete, we are done.
2723 */
2724 if ((dzp_error = zfs_zaccess_common(dzp, ACE_DELETE_CHILD,
2725 &dzp_working_mode, &dzpcheck_privs, B_FALSE, cr)) == 0)
2726 return (0);
2727
2728 /*
2729 * If target object has delete permission then we are done
2730 */
2731 if ((zp_error = zfs_zaccess_common(zp, ACE_DELETE, &zp_working_mode,
2732 &zpcheck_privs, B_FALSE, cr)) == 0)
2733 return (0);
2734
2735 ASSERT(dzp_error && zp_error);
2736
2737 if (!dzpcheck_privs)
2738 return (dzp_error);
2739 if (!zpcheck_privs)
2740 return (zp_error);
2741
2742 /*
2743 * Second row
2744 *
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.
2750 */
2751
2752 if (dzp_error == EACCES)
2753 return (secpolicy_vnode_remove(cr));
2754
2755 /*
2756 * Third Row
2757 * only need to see if we have write/execute on directory.
2758 */
2759
2760 dzp_error = zfs_zaccess_common(dzp, ACE_EXECUTE|ACE_WRITE_DATA,
2761 &dzp_working_mode, &dzpcheck_privs, B_FALSE, cr);
2762
2763 if (dzp_error != 0 && !dzpcheck_privs)
2764 return (dzp_error);
2765
2766 /*
2767 * Fourth row
2768 */
2769
2770 available_perms = (dzp_working_mode & ACE_WRITE_DATA) ? 0 : S_IWUSR;
2771 available_perms |= (dzp_working_mode & ACE_EXECUTE) ? 0 : S_IXUSR;
2772
2773 return (zfs_delete_final_check(zp, dzp, available_perms, cr));
2774
2775 }
2776
2777 int
2778 zfs_zaccess_rename(znode_t *sdzp, znode_t *szp, znode_t *tdzp,
2779 znode_t *tzp, cred_t *cr)
2780 {
2781 int add_perm;
2782 int error;
2783
2784 if (szp->z_pflags & ZFS_AV_QUARANTINED)
2785 return (SET_ERROR(EACCES));
2786
2787 add_perm = S_ISDIR(ZTOI(szp)->i_mode) ?
2788 ACE_ADD_SUBDIRECTORY : ACE_ADD_FILE;
2789
2790 /*
2791 * Rename permissions are combination of delete permission +
2792 * add file/subdir permission.
2793 */
2794
2795 /*
2796 * first make sure we do the delete portion.
2797 *
2798 * If that succeeds then check for add_file/add_subdir permissions
2799 */
2800
2801 if ((error = zfs_zaccess_delete(sdzp, szp, cr)))
2802 return (error);
2803
2804 /*
2805 * If we have a tzp, see if we can delete it?
2806 */
2807 if (tzp) {
2808 if ((error = zfs_zaccess_delete(tdzp, tzp, cr)))
2809 return (error);
2810 }
2811
2812 /*
2813 * Now check for add permissions
2814 */
2815 error = zfs_zaccess(tdzp, add_perm, 0, B_FALSE, cr);
2816
2817 return (error);
2818 }