]> git.proxmox.com Git - mirror_zfs.git/blob - module/zfs/zfs_acl.c
projects / mirror_zfs.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
Illumos #3598
[mirror_zfs.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 "fs/fs_subr.h"
56
57 #define ALLOW ACE_ACCESS_ALLOWED_ACE_TYPE
58 #define DENY ACE_ACCESS_DENIED_ACE_TYPE
59 #define MAX_ACE_TYPE ACE_SYSTEM_ALARM_CALLBACK_OBJECT_ACE_TYPE
60 #define MIN_ACE_TYPE ALLOW
61
62 #define OWNING_GROUP (ACE_GROUP|ACE_IDENTIFIER_GROUP)
63 #define EVERYONE_ALLOW_MASK (ACE_READ_ACL|ACE_READ_ATTRIBUTES | \
64 ACE_READ_NAMED_ATTRS|ACE_SYNCHRONIZE)
65 #define EVERYONE_DENY_MASK (ACE_WRITE_ACL|ACE_WRITE_OWNER | \
66 ACE_WRITE_ATTRIBUTES|ACE_WRITE_NAMED_ATTRS)
67 #define OWNER_ALLOW_MASK (ACE_WRITE_ACL | ACE_WRITE_OWNER | \
68 ACE_WRITE_ATTRIBUTES|ACE_WRITE_NAMED_ATTRS)
69
70 #define ZFS_CHECKED_MASKS (ACE_READ_ACL|ACE_READ_ATTRIBUTES|ACE_READ_DATA| \
71 ACE_READ_NAMED_ATTRS|ACE_WRITE_DATA|ACE_WRITE_ATTRIBUTES| \
72 ACE_WRITE_NAMED_ATTRS|ACE_APPEND_DATA|ACE_EXECUTE|ACE_WRITE_OWNER| \
73 ACE_WRITE_ACL|ACE_DELETE|ACE_DELETE_CHILD|ACE_SYNCHRONIZE)
74
75 #define WRITE_MASK_DATA (ACE_WRITE_DATA|ACE_APPEND_DATA|ACE_WRITE_NAMED_ATTRS)
76 #define WRITE_MASK_ATTRS (ACE_WRITE_ACL|ACE_WRITE_OWNER|ACE_WRITE_ATTRIBUTES| \
77 ACE_DELETE|ACE_DELETE_CHILD)
78 #define WRITE_MASK (WRITE_MASK_DATA|WRITE_MASK_ATTRS)
79
80 #define OGE_CLEAR (ACE_READ_DATA|ACE_LIST_DIRECTORY|ACE_WRITE_DATA| \
81 ACE_ADD_FILE|ACE_APPEND_DATA|ACE_ADD_SUBDIRECTORY|ACE_EXECUTE)
82
83 #define OKAY_MASK_BITS (ACE_READ_DATA|ACE_LIST_DIRECTORY|ACE_WRITE_DATA| \
84 ACE_ADD_FILE|ACE_APPEND_DATA|ACE_ADD_SUBDIRECTORY|ACE_EXECUTE)
85
86 #define ALL_INHERIT (ACE_FILE_INHERIT_ACE|ACE_DIRECTORY_INHERIT_ACE | \
87 ACE_NO_PROPAGATE_INHERIT_ACE|ACE_INHERIT_ONLY_ACE|ACE_INHERITED_ACE)
88
89 #define RESTRICTED_CLEAR (ACE_WRITE_ACL|ACE_WRITE_OWNER)
90
91 #define V4_ACL_WIDE_FLAGS (ZFS_ACL_AUTO_INHERIT|ZFS_ACL_DEFAULTED|\
92 ZFS_ACL_PROTECTED)
93
94 #define ZFS_ACL_WIDE_FLAGS (V4_ACL_WIDE_FLAGS|ZFS_ACL_TRIVIAL|ZFS_INHERIT_ACE|\
95 ZFS_ACL_OBJ_ACE)
96
97 #define ALL_MODE_EXECS (S_IXUSR | S_IXGRP | S_IXOTH)
98
99 static uint16_t
100 zfs_ace_v0_get_type(void *acep)
101 {
102 return (((zfs_oldace_t *)acep)->z_type);
103 }
104
105 static uint16_t
106 zfs_ace_v0_get_flags(void *acep)
107 {
108 return (((zfs_oldace_t *)acep)->z_flags);
109 }
110
111 static uint32_t
112 zfs_ace_v0_get_mask(void *acep)
113 {
114 return (((zfs_oldace_t *)acep)->z_access_mask);
115 }
116
117 static uint64_t
118 zfs_ace_v0_get_who(void *acep)
119 {
120 return (((zfs_oldace_t *)acep)->z_fuid);
121 }
122
123 static void
124 zfs_ace_v0_set_type(void *acep, uint16_t type)
125 {
126 ((zfs_oldace_t *)acep)->z_type = type;
127 }
128
129 static void
130 zfs_ace_v0_set_flags(void *acep, uint16_t flags)
131 {
132 ((zfs_oldace_t *)acep)->z_flags = flags;
133 }
134
135 static void
136 zfs_ace_v0_set_mask(void *acep, uint32_t mask)
137 {
138 ((zfs_oldace_t *)acep)->z_access_mask = mask;
139 }
140
141 static void
142 zfs_ace_v0_set_who(void *acep, uint64_t who)
143 {
144 ((zfs_oldace_t *)acep)->z_fuid = who;
145 }
146
147 /*ARGSUSED*/
148 static size_t
149 zfs_ace_v0_size(void *acep)
150 {
151 return (sizeof (zfs_oldace_t));
152 }
153
154 static size_t
155 zfs_ace_v0_abstract_size(void)
156 {
157 return (sizeof (zfs_oldace_t));
158 }
159
160 static int
161 zfs_ace_v0_mask_off(void)
162 {
163 return (offsetof(zfs_oldace_t, z_access_mask));
164 }
165
166 /*ARGSUSED*/
167 static int
168 zfs_ace_v0_data(void *acep, void **datap)
169 {
170 *datap = NULL;
171 return (0);
172 }
173
174 static acl_ops_t zfs_acl_v0_ops = {
175 zfs_ace_v0_get_mask,
176 zfs_ace_v0_set_mask,
177 zfs_ace_v0_get_flags,
178 zfs_ace_v0_set_flags,
179 zfs_ace_v0_get_type,
180 zfs_ace_v0_set_type,
181 zfs_ace_v0_get_who,
182 zfs_ace_v0_set_who,
183 zfs_ace_v0_size,
184 zfs_ace_v0_abstract_size,
185 zfs_ace_v0_mask_off,
186 zfs_ace_v0_data
187 };
188
189 static uint16_t
190 zfs_ace_fuid_get_type(void *acep)
191 {
192 return (((zfs_ace_hdr_t *)acep)->z_type);
193 }
194
195 static uint16_t
196 zfs_ace_fuid_get_flags(void *acep)
197 {
198 return (((zfs_ace_hdr_t *)acep)->z_flags);
199 }
200
201 static uint32_t
202 zfs_ace_fuid_get_mask(void *acep)
203 {
204 return (((zfs_ace_hdr_t *)acep)->z_access_mask);
205 }
206
207 static uint64_t
208 zfs_ace_fuid_get_who(void *args)
209 {
210 uint16_t entry_type;
211 zfs_ace_t *acep = args;
212
213 entry_type = acep->z_hdr.z_flags & ACE_TYPE_FLAGS;
214
215 if (entry_type == ACE_OWNER || entry_type == OWNING_GROUP ||
216 entry_type == ACE_EVERYONE)
217 return (-1);
218 return (((zfs_ace_t *)acep)->z_fuid);
219 }
220
221 static void
222 zfs_ace_fuid_set_type(void *acep, uint16_t type)
223 {
224 ((zfs_ace_hdr_t *)acep)->z_type = type;
225 }
226
227 static void
228 zfs_ace_fuid_set_flags(void *acep, uint16_t flags)
229 {
230 ((zfs_ace_hdr_t *)acep)->z_flags = flags;
231 }
232
233 static void
234 zfs_ace_fuid_set_mask(void *acep, uint32_t mask)
235 {
236 ((zfs_ace_hdr_t *)acep)->z_access_mask = mask;
237 }
238
239 static void
240 zfs_ace_fuid_set_who(void *arg, uint64_t who)
241 {
242 zfs_ace_t *acep = arg;
243
244 uint16_t entry_type = acep->z_hdr.z_flags & ACE_TYPE_FLAGS;
245
246 if (entry_type == ACE_OWNER || entry_type == OWNING_GROUP ||
247 entry_type == ACE_EVERYONE)
248 return;
249 acep->z_fuid = who;
250 }
251
252 static size_t
253 zfs_ace_fuid_size(void *acep)
254 {
255 zfs_ace_hdr_t *zacep = acep;
256 uint16_t entry_type;
257
258 switch (zacep->z_type) {
259 case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE:
260 case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE:
261 case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE:
262 case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE:
263 return (sizeof (zfs_object_ace_t));
264 case ALLOW:
265 case DENY:
266 entry_type =
267 (((zfs_ace_hdr_t *)acep)->z_flags & ACE_TYPE_FLAGS);
268 if (entry_type == ACE_OWNER ||
269 entry_type == OWNING_GROUP ||
270 entry_type == ACE_EVERYONE)
271 return (sizeof (zfs_ace_hdr_t));
272 /*FALLTHROUGH*/
273 default:
274 return (sizeof (zfs_ace_t));
275 }
276 }
277
278 static size_t
279 zfs_ace_fuid_abstract_size(void)
280 {
281 return (sizeof (zfs_ace_hdr_t));
282 }
283
284 static int
285 zfs_ace_fuid_mask_off(void)
286 {
287 return (offsetof(zfs_ace_hdr_t, z_access_mask));
288 }
289
290 static int
291 zfs_ace_fuid_data(void *acep, void **datap)
292 {
293 zfs_ace_t *zacep = acep;
294 zfs_object_ace_t *zobjp;
295
296 switch (zacep->z_hdr.z_type) {
297 case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE:
298 case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE:
299 case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE:
300 case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE:
301 zobjp = acep;
302 *datap = (caddr_t)zobjp + sizeof (zfs_ace_t);
303 return (sizeof (zfs_object_ace_t) - sizeof (zfs_ace_t));
304 default:
305 *datap = NULL;
306 return (0);
307 }
308 }
309
310 static acl_ops_t zfs_acl_fuid_ops = {
311 zfs_ace_fuid_get_mask,
312 zfs_ace_fuid_set_mask,
313 zfs_ace_fuid_get_flags,
314 zfs_ace_fuid_set_flags,
315 zfs_ace_fuid_get_type,
316 zfs_ace_fuid_set_type,
317 zfs_ace_fuid_get_who,
318 zfs_ace_fuid_set_who,
319 zfs_ace_fuid_size,
320 zfs_ace_fuid_abstract_size,
321 zfs_ace_fuid_mask_off,
322 zfs_ace_fuid_data
323 };
324
325 /*
326 * The following three functions are provided for compatibility with
327 * older ZPL version in order to determine if the file use to have
328 * an external ACL and what version of ACL previously existed on the
329 * file. Would really be nice to not need this, sigh.
330 */
331 uint64_t
332 zfs_external_acl(znode_t *zp)
333 {
334 zfs_acl_phys_t acl_phys;
335 int error;
336
337 if (zp->z_is_sa)
338 return (0);
339
340 /*
341 * Need to deal with a potential
342 * race where zfs_sa_upgrade could cause
343 * z_isa_sa to change.
344 *
345 * If the lookup fails then the state of z_is_sa should have
346 * changed.
347 */
348
349 if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_ZNODE_ACL(ZTOZSB(zp)),
350 &acl_phys, sizeof (acl_phys))) == 0)
351 return (acl_phys.z_acl_extern_obj);
352 else {
353 /*
354 * after upgrade the SA_ZPL_ZNODE_ACL should have been
355 * removed
356 */
357 VERIFY(zp->z_is_sa && error == ENOENT);
358 return (0);
359 }
360 }
361
362 /*
363 * Determine size of ACL in bytes
364 *
365 * This is more complicated than it should be since we have to deal
366 * with old external ACLs.
367 */
368 static int
369 zfs_acl_znode_info(znode_t *zp, int *aclsize, int *aclcount,
370 zfs_acl_phys_t *aclphys)
371 {
372 zfs_sb_t *zsb = ZTOZSB(zp);
373 uint64_t acl_count;
374 int size;
375 int error;
376
377 ASSERT(MUTEX_HELD(&zp->z_acl_lock));
378 if (zp->z_is_sa) {
379 if ((error = sa_size(zp->z_sa_hdl, SA_ZPL_DACL_ACES(zsb),
380 &size)) != 0)
381 return (error);
382 *aclsize = size;
383 if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_DACL_COUNT(zsb),
384 &acl_count, sizeof (acl_count))) != 0)
385 return (error);
386 *aclcount = acl_count;
387 } else {
388 if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_ZNODE_ACL(zsb),
389 aclphys, sizeof (*aclphys))) != 0)
390 return (error);
391
392 if (aclphys->z_acl_version == ZFS_ACL_VERSION_INITIAL) {
393 *aclsize = ZFS_ACL_SIZE(aclphys->z_acl_size);
394 *aclcount = aclphys->z_acl_size;
395 } else {
396 *aclsize = aclphys->z_acl_size;
397 *aclcount = aclphys->z_acl_count;
398 }
399 }
400 return (0);
401 }
402
403 int
404 zfs_znode_acl_version(znode_t *zp)
405 {
406 zfs_acl_phys_t acl_phys;
407
408 if (zp->z_is_sa)
409 return (ZFS_ACL_VERSION_FUID);
410 else {
411 int error;
412
413 /*
414 * Need to deal with a potential
415 * race where zfs_sa_upgrade could cause
416 * z_isa_sa to change.
417 *
418 * If the lookup fails then the state of z_is_sa should have
419 * changed.
420 */
421 if ((error = sa_lookup(zp->z_sa_hdl,
422 SA_ZPL_ZNODE_ACL(ZTOZSB(zp)),
423 &acl_phys, sizeof (acl_phys))) == 0)
424 return (acl_phys.z_acl_version);
425 else {
426 /*
427 * After upgrade SA_ZPL_ZNODE_ACL should have
428 * been removed.
429 */
430 VERIFY(zp->z_is_sa && error == ENOENT);
431 return (ZFS_ACL_VERSION_FUID);
432 }
433 }
434 }
435
436 static int
437 zfs_acl_version(int version)
438 {
439 if (version < ZPL_VERSION_FUID)
440 return (ZFS_ACL_VERSION_INITIAL);
441 else
442 return (ZFS_ACL_VERSION_FUID);
443 }
444
445 static int
446 zfs_acl_version_zp(znode_t *zp)
447 {
448 return (zfs_acl_version(ZTOZSB(zp)->z_version));
449 }
450
451 zfs_acl_t *
452 zfs_acl_alloc(int vers)
453 {
454 zfs_acl_t *aclp;
455
456 aclp = kmem_zalloc(sizeof (zfs_acl_t), KM_PUSHPAGE);
457 list_create(&aclp->z_acl, sizeof (zfs_acl_node_t),
458 offsetof(zfs_acl_node_t, z_next));
459 aclp->z_version = vers;
460 if (vers == ZFS_ACL_VERSION_FUID)
461 aclp->z_ops = &zfs_acl_fuid_ops;
462 else
463 aclp->z_ops = &zfs_acl_v0_ops;
464 return (aclp);
465 }
466
467 zfs_acl_node_t *
468 zfs_acl_node_alloc(size_t bytes)
469 {
470 zfs_acl_node_t *aclnode;
471
472 aclnode = kmem_zalloc(sizeof (zfs_acl_node_t), KM_PUSHPAGE);
473 if (bytes) {
474 aclnode->z_acldata = kmem_alloc(bytes, KM_PUSHPAGE);
475 aclnode->z_allocdata = aclnode->z_acldata;
476 aclnode->z_allocsize = bytes;
477 aclnode->z_size = bytes;
478 }
479
480 return (aclnode);
481 }
482
483 static void
484 zfs_acl_node_free(zfs_acl_node_t *aclnode)
485 {
486 if (aclnode->z_allocsize)
487 kmem_free(aclnode->z_allocdata, aclnode->z_allocsize);
488 kmem_free(aclnode, sizeof (zfs_acl_node_t));
489 }
490
491 static void
492 zfs_acl_release_nodes(zfs_acl_t *aclp)
493 {
494 zfs_acl_node_t *aclnode;
495
496 while ((aclnode = list_head(&aclp->z_acl))) {
497 list_remove(&aclp->z_acl, aclnode);
498 zfs_acl_node_free(aclnode);
499 }
500 aclp->z_acl_count = 0;
501 aclp->z_acl_bytes = 0;
502 }
503
504 void
505 zfs_acl_free(zfs_acl_t *aclp)
506 {
507 zfs_acl_release_nodes(aclp);
508 list_destroy(&aclp->z_acl);
509 kmem_free(aclp, sizeof (zfs_acl_t));
510 }
511
512 static boolean_t
513 zfs_acl_valid_ace_type(uint_t type, uint_t flags)
514 {
515 uint16_t entry_type;
516
517 switch (type) {
518 case ALLOW:
519 case DENY:
520 case ACE_SYSTEM_AUDIT_ACE_TYPE:
521 case ACE_SYSTEM_ALARM_ACE_TYPE:
522 entry_type = flags & ACE_TYPE_FLAGS;
523 return (entry_type == ACE_OWNER ||
524 entry_type == OWNING_GROUP ||
525 entry_type == ACE_EVERYONE || entry_type == 0 ||
526 entry_type == ACE_IDENTIFIER_GROUP);
527 default:
528 if (type >= MIN_ACE_TYPE && type <= MAX_ACE_TYPE)
529 return (B_TRUE);
530 }
531 return (B_FALSE);
532 }
533
534 static boolean_t
535 zfs_ace_valid(umode_t obj_mode, zfs_acl_t *aclp, uint16_t type, uint16_t iflags)
536 {
537 /*
538 * first check type of entry
539 */
540
541 if (!zfs_acl_valid_ace_type(type, iflags))
542 return (B_FALSE);
543
544 switch (type) {
545 case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE:
546 case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE:
547 case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE:
548 case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE:
549 if (aclp->z_version < ZFS_ACL_VERSION_FUID)
550 return (B_FALSE);
551 aclp->z_hints |= ZFS_ACL_OBJ_ACE;
552 }
553
554 /*
555 * next check inheritance level flags
556 */
557
558 if (S_ISDIR(obj_mode) &&
559 (iflags & (ACE_FILE_INHERIT_ACE|ACE_DIRECTORY_INHERIT_ACE)))
560 aclp->z_hints |= ZFS_INHERIT_ACE;
561
562 if (iflags & (ACE_INHERIT_ONLY_ACE|ACE_NO_PROPAGATE_INHERIT_ACE)) {
563 if ((iflags & (ACE_FILE_INHERIT_ACE|
564 ACE_DIRECTORY_INHERIT_ACE)) == 0) {
565 return (B_FALSE);
566 }
567 }
568
569 return (B_TRUE);
570 }
571
572 static void *
573 zfs_acl_next_ace(zfs_acl_t *aclp, void *start, uint64_t *who,
574 uint32_t *access_mask, uint16_t *iflags, uint16_t *type)
575 {
576 zfs_acl_node_t *aclnode;
577
578 ASSERT(aclp);
579
580 if (start == NULL) {
581 aclnode = list_head(&aclp->z_acl);
582 if (aclnode == NULL)
583 return (NULL);
584
585 aclp->z_next_ace = aclnode->z_acldata;
586 aclp->z_curr_node = aclnode;
587 aclnode->z_ace_idx = 0;
588 }
589
590 aclnode = aclp->z_curr_node;
591
592 if (aclnode == NULL)
593 return (NULL);
594
595 if (aclnode->z_ace_idx >= aclnode->z_ace_count) {
596 aclnode = list_next(&aclp->z_acl, aclnode);
597 if (aclnode == NULL)
598 return (NULL);
599 else {
600 aclp->z_curr_node = aclnode;
601 aclnode->z_ace_idx = 0;
602 aclp->z_next_ace = aclnode->z_acldata;
603 }
604 }
605
606 if (aclnode->z_ace_idx < aclnode->z_ace_count) {
607 void *acep = aclp->z_next_ace;
608 size_t ace_size;
609
610 /*
611 * Make sure we don't overstep our bounds
612 */
613 ace_size = aclp->z_ops->ace_size(acep);
614
615 if (((caddr_t)acep + ace_size) >
616 ((caddr_t)aclnode->z_acldata + aclnode->z_size)) {
617 return (NULL);
618 }
619
620 *iflags = aclp->z_ops->ace_flags_get(acep);
621 *type = aclp->z_ops->ace_type_get(acep);
622 *access_mask = aclp->z_ops->ace_mask_get(acep);
623 *who = aclp->z_ops->ace_who_get(acep);
624 aclp->z_next_ace = (caddr_t)aclp->z_next_ace + ace_size;
625 aclnode->z_ace_idx++;
626
627 return ((void *)acep);
628 }
629 return (NULL);
630 }
631
632 /*ARGSUSED*/
633 static uint64_t
634 zfs_ace_walk(void *datap, uint64_t cookie, int aclcnt,
635 uint16_t *flags, uint16_t *type, uint32_t *mask)
636 {
637 zfs_acl_t *aclp = datap;
638 zfs_ace_hdr_t *acep = (zfs_ace_hdr_t *)(uintptr_t)cookie;
639 uint64_t who;
640
641 acep = zfs_acl_next_ace(aclp, acep, &who, mask,
642 flags, type);
643 return ((uint64_t)(uintptr_t)acep);
644 }
645
646 /*
647 * Copy ACE to internal ZFS format.
648 * While processing the ACL each ACE will be validated for correctness.
649 * ACE FUIDs will be created later.
650 */
651 int
652 zfs_copy_ace_2_fuid(zfs_sb_t *zsb, umode_t obj_mode, zfs_acl_t *aclp,
653 void *datap, zfs_ace_t *z_acl, uint64_t aclcnt, size_t *size,
654 zfs_fuid_info_t **fuidp, cred_t *cr)
655 {
656 int i;
657 uint16_t entry_type;
658 zfs_ace_t *aceptr = z_acl;
659 ace_t *acep = datap;
660 zfs_object_ace_t *zobjacep;
661 ace_object_t *aceobjp;
662
663 for (i = 0; i != aclcnt; i++) {
664 aceptr->z_hdr.z_access_mask = acep->a_access_mask;
665 aceptr->z_hdr.z_flags = acep->a_flags;
666 aceptr->z_hdr.z_type = acep->a_type;
667 entry_type = aceptr->z_hdr.z_flags & ACE_TYPE_FLAGS;
668 if (entry_type != ACE_OWNER && entry_type != OWNING_GROUP &&
669 entry_type != ACE_EVERYONE) {
670 aceptr->z_fuid = zfs_fuid_create(zsb, acep->a_who,
671 cr, (entry_type == 0) ?
672 ZFS_ACE_USER : ZFS_ACE_GROUP, fuidp);
673 }
674
675 /*
676 * Make sure ACE is valid
677 */
678 if (zfs_ace_valid(obj_mode, aclp, aceptr->z_hdr.z_type,
679 aceptr->z_hdr.z_flags) != B_TRUE)
680 return (SET_ERROR(EINVAL));
681
682 switch (acep->a_type) {
683 case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE:
684 case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE:
685 case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE:
686 case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE:
687 zobjacep = (zfs_object_ace_t *)aceptr;
688 aceobjp = (ace_object_t *)acep;
689
690 bcopy(aceobjp->a_obj_type, zobjacep->z_object_type,
691 sizeof (aceobjp->a_obj_type));
692 bcopy(aceobjp->a_inherit_obj_type,
693 zobjacep->z_inherit_type,
694 sizeof (aceobjp->a_inherit_obj_type));
695 acep = (ace_t *)((caddr_t)acep + sizeof (ace_object_t));
696 break;
697 default:
698 acep = (ace_t *)((caddr_t)acep + sizeof (ace_t));
699 }
700
701 aceptr = (zfs_ace_t *)((caddr_t)aceptr +
702 aclp->z_ops->ace_size(aceptr));
703 }
704
705 *size = (caddr_t)aceptr - (caddr_t)z_acl;
706
707 return (0);
708 }
709
710 /*
711 * Copy ZFS ACEs to fixed size ace_t layout
712 */
713 static void
714 zfs_copy_fuid_2_ace(zfs_sb_t *zsb, zfs_acl_t *aclp, cred_t *cr,
715 void *datap, int filter)
716 {
717 uint64_t who;
718 uint32_t access_mask;
719 uint16_t iflags, type;
720 zfs_ace_hdr_t *zacep = NULL;
721 ace_t *acep = datap;
722 ace_object_t *objacep;
723 zfs_object_ace_t *zobjacep;
724 size_t ace_size;
725 uint16_t entry_type;
726
727 while ((zacep = zfs_acl_next_ace(aclp, zacep,
728 &who, &access_mask, &iflags, &type))) {
729
730 switch (type) {
731 case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE:
732 case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE:
733 case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE:
734 case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE:
735 if (filter) {
736 continue;
737 }
738 zobjacep = (zfs_object_ace_t *)zacep;
739 objacep = (ace_object_t *)acep;
740 bcopy(zobjacep->z_object_type,
741 objacep->a_obj_type,
742 sizeof (zobjacep->z_object_type));
743 bcopy(zobjacep->z_inherit_type,
744 objacep->a_inherit_obj_type,
745 sizeof (zobjacep->z_inherit_type));
746 ace_size = sizeof (ace_object_t);
747 break;
748 default:
749 ace_size = sizeof (ace_t);
750 break;
751 }
752
753 entry_type = (iflags & ACE_TYPE_FLAGS);
754 if ((entry_type != ACE_OWNER &&
755 entry_type != OWNING_GROUP &&
756 entry_type != ACE_EVERYONE)) {
757 acep->a_who = zfs_fuid_map_id(zsb, who,
758 cr, (entry_type & ACE_IDENTIFIER_GROUP) ?
759 ZFS_ACE_GROUP : ZFS_ACE_USER);
760 } else {
761 acep->a_who = (uid_t)(int64_t)who;
762 }
763 acep->a_access_mask = access_mask;
764 acep->a_flags = iflags;
765 acep->a_type = type;
766 acep = (ace_t *)((caddr_t)acep + ace_size);
767 }
768 }
769
770 static int
771 zfs_copy_ace_2_oldace(umode_t obj_mode, zfs_acl_t *aclp, ace_t *acep,
772 zfs_oldace_t *z_acl, int aclcnt, size_t *size)
773 {
774 int i;
775 zfs_oldace_t *aceptr = z_acl;
776
777 for (i = 0; i != aclcnt; i++, aceptr++) {
778 aceptr->z_access_mask = acep[i].a_access_mask;
779 aceptr->z_type = acep[i].a_type;
780 aceptr->z_flags = acep[i].a_flags;
781 aceptr->z_fuid = acep[i].a_who;
782 /*
783 * Make sure ACE is valid
784 */
785 if (zfs_ace_valid(obj_mode, aclp, aceptr->z_type,
786 aceptr->z_flags) != B_TRUE)
787 return (SET_ERROR(EINVAL));
788 }
789 *size = (caddr_t)aceptr - (caddr_t)z_acl;
790 return (0);
791 }
792
793 /*
794 * convert old ACL format to new
795 */
796 void
797 zfs_acl_xform(znode_t *zp, zfs_acl_t *aclp, cred_t *cr)
798 {
799 zfs_oldace_t *oldaclp;
800 int i;
801 uint16_t type, iflags;
802 uint32_t access_mask;
803 uint64_t who;
804 void *cookie = NULL;
805 zfs_acl_node_t *newaclnode;
806
807 ASSERT(aclp->z_version == ZFS_ACL_VERSION_INITIAL);
808 /*
809 * First create the ACE in a contiguous piece of memory
810 * for zfs_copy_ace_2_fuid().
811 *
812 * We only convert an ACL once, so this won't happen
813 * everytime.
814 */
815 oldaclp = kmem_alloc(sizeof (zfs_oldace_t) * aclp->z_acl_count,
816 KM_SLEEP);
817 i = 0;
818 while ((cookie = zfs_acl_next_ace(aclp, cookie, &who,
819 &access_mask, &iflags, &type))) {
820 oldaclp[i].z_flags = iflags;
821 oldaclp[i].z_type = type;
822 oldaclp[i].z_fuid = who;
823 oldaclp[i++].z_access_mask = access_mask;
824 }
825
826 newaclnode = zfs_acl_node_alloc(aclp->z_acl_count *
827 sizeof (zfs_object_ace_t));
828 aclp->z_ops = &zfs_acl_fuid_ops;
829 VERIFY(zfs_copy_ace_2_fuid(ZTOZSB(zp), ZTOI(zp)->i_mode,
830 aclp, oldaclp, newaclnode->z_acldata, aclp->z_acl_count,
831 &newaclnode->z_size, NULL, cr) == 0);
832 newaclnode->z_ace_count = aclp->z_acl_count;
833 aclp->z_version = ZFS_ACL_VERSION;
834 kmem_free(oldaclp, aclp->z_acl_count * sizeof (zfs_oldace_t));
835
836 /*
837 * Release all previous ACL nodes
838 */
839
840 zfs_acl_release_nodes(aclp);
841
842 list_insert_head(&aclp->z_acl, newaclnode);
843
844 aclp->z_acl_bytes = newaclnode->z_size;
845 aclp->z_acl_count = newaclnode->z_ace_count;
846
847 }
848
849 /*
850 * Convert unix access mask to v4 access mask
851 */
852 static uint32_t
853 zfs_unix_to_v4(uint32_t access_mask)
854 {
855 uint32_t new_mask = 0;
856
857 if (access_mask & S_IXOTH)
858 new_mask |= ACE_EXECUTE;
859 if (access_mask & S_IWOTH)
860 new_mask |= ACE_WRITE_DATA;
861 if (access_mask & S_IROTH)
862 new_mask |= ACE_READ_DATA;
863 return (new_mask);
864 }
865
866 static void
867 zfs_set_ace(zfs_acl_t *aclp, void *acep, uint32_t access_mask,
868 uint16_t access_type, uint64_t fuid, uint16_t entry_type)
869 {
870 uint16_t type = entry_type & ACE_TYPE_FLAGS;
871
872 aclp->z_ops->ace_mask_set(acep, access_mask);
873 aclp->z_ops->ace_type_set(acep, access_type);
874 aclp->z_ops->ace_flags_set(acep, entry_type);
875 if ((type != ACE_OWNER && type != OWNING_GROUP &&
876 type != ACE_EVERYONE))
877 aclp->z_ops->ace_who_set(acep, fuid);
878 }
879
880 /*
881 * Determine mode of file based on ACL.
882 * Also, create FUIDs for any User/Group ACEs
883 */
884 uint64_t
885 zfs_mode_compute(uint64_t fmode, zfs_acl_t *aclp,
886 uint64_t *pflags, uint64_t fuid, uint64_t fgid)
887 {
888 int entry_type;
889 mode_t mode;
890 mode_t seen = 0;
891 zfs_ace_hdr_t *acep = NULL;
892 uint64_t who;
893 uint16_t iflags, type;
894 uint32_t access_mask;
895 boolean_t an_exec_denied = B_FALSE;
896
897 mode = (fmode & (S_IFMT | S_ISUID | S_ISGID | S_ISVTX));
898
899 while ((acep = zfs_acl_next_ace(aclp, acep, &who,
900 &access_mask, &iflags, &type))) {
901
902 if (!zfs_acl_valid_ace_type(type, iflags))
903 continue;
904
905 entry_type = (iflags & ACE_TYPE_FLAGS);
906
907 /*
908 * Skip over owner@, group@ or everyone@ inherit only ACEs
909 */
910 if ((iflags & ACE_INHERIT_ONLY_ACE) &&
911 (entry_type == ACE_OWNER || entry_type == ACE_EVERYONE ||
912 entry_type == OWNING_GROUP))
913 continue;
914
915 if (entry_type == ACE_OWNER || (entry_type == 0 &&
916 who == fuid)) {
917 if ((access_mask & ACE_READ_DATA) &&
918 (!(seen & S_IRUSR))) {
919 seen |= S_IRUSR;
920 if (type == ALLOW) {
921 mode |= S_IRUSR;
922 }
923 }
924 if ((access_mask & ACE_WRITE_DATA) &&
925 (!(seen & S_IWUSR))) {
926 seen |= S_IWUSR;
927 if (type == ALLOW) {
928 mode |= S_IWUSR;
929 }
930 }
931 if ((access_mask & ACE_EXECUTE) &&
932 (!(seen & S_IXUSR))) {
933 seen |= S_IXUSR;
934 if (type == ALLOW) {
935 mode |= S_IXUSR;
936 }
937 }
938 } else if (entry_type == OWNING_GROUP ||
939 (entry_type == ACE_IDENTIFIER_GROUP && who == fgid)) {
940 if ((access_mask & ACE_READ_DATA) &&
941 (!(seen & S_IRGRP))) {
942 seen |= S_IRGRP;
943 if (type == ALLOW) {
944 mode |= S_IRGRP;
945 }
946 }
947 if ((access_mask & ACE_WRITE_DATA) &&
948 (!(seen & S_IWGRP))) {
949 seen |= S_IWGRP;
950 if (type == ALLOW) {
951 mode |= S_IWGRP;
952 }
953 }
954 if ((access_mask & ACE_EXECUTE) &&
955 (!(seen & S_IXGRP))) {
956 seen |= S_IXGRP;
957 if (type == ALLOW) {
958 mode |= S_IXGRP;
959 }
960 }
961 } else if (entry_type == ACE_EVERYONE) {
962 if ((access_mask & ACE_READ_DATA)) {
963 if (!(seen & S_IRUSR)) {
964 seen |= S_IRUSR;
965 if (type == ALLOW) {
966 mode |= S_IRUSR;
967 }
968 }
969 if (!(seen & S_IRGRP)) {
970 seen |= S_IRGRP;
971 if (type == ALLOW) {
972 mode |= S_IRGRP;
973 }
974 }
975 if (!(seen & S_IROTH)) {
976 seen |= S_IROTH;
977 if (type == ALLOW) {
978 mode |= S_IROTH;
979 }
980 }
981 }
982 if ((access_mask & ACE_WRITE_DATA)) {
983 if (!(seen & S_IWUSR)) {
984 seen |= S_IWUSR;
985 if (type == ALLOW) {
986 mode |= S_IWUSR;
987 }
988 }
989 if (!(seen & S_IWGRP)) {
990 seen |= S_IWGRP;
991 if (type == ALLOW) {
992 mode |= S_IWGRP;
993 }
994 }
995 if (!(seen & S_IWOTH)) {
996 seen |= S_IWOTH;
997 if (type == ALLOW) {
998 mode |= S_IWOTH;
999 }
1000 }
1001 }
1002 if ((access_mask & ACE_EXECUTE)) {
1003 if (!(seen & S_IXUSR)) {
1004 seen |= S_IXUSR;
1005 if (type == ALLOW) {
1006 mode |= S_IXUSR;
1007 }
1008 }
1009 if (!(seen & S_IXGRP)) {
1010 seen |= S_IXGRP;
1011 if (type == ALLOW) {
1012 mode |= S_IXGRP;
1013 }
1014 }
1015 if (!(seen & S_IXOTH)) {
1016 seen |= S_IXOTH;
1017 if (type == ALLOW) {
1018 mode |= S_IXOTH;
1019 }
1020 }
1021 }
1022 } else {
1023 /*
1024 * Only care if this IDENTIFIER_GROUP or
1025 * USER ACE denies execute access to someone,
1026 * mode is not affected
1027 */
1028 if ((access_mask & ACE_EXECUTE) && type == DENY)
1029 an_exec_denied = B_TRUE;
1030 }
1031 }
1032
1033 /*
1034 * Failure to allow is effectively a deny, so execute permission
1035 * is denied if it was never mentioned or if we explicitly
1036 * weren't allowed it.
1037 */
1038 if (!an_exec_denied &&
1039 ((seen & ALL_MODE_EXECS) != ALL_MODE_EXECS ||
1040 (mode & ALL_MODE_EXECS) != ALL_MODE_EXECS))
1041 an_exec_denied = B_TRUE;
1042
1043 if (an_exec_denied)
1044 *pflags &= ~ZFS_NO_EXECS_DENIED;
1045 else
1046 *pflags |= ZFS_NO_EXECS_DENIED;
1047
1048 return (mode);
1049 }
1050
1051 /*
1052 * Read an external acl object. If the intent is to modify, always
1053 * create a new acl and leave any cached acl in place.
1054 */
1055 static int
1056 zfs_acl_node_read(znode_t *zp, boolean_t have_lock, zfs_acl_t **aclpp,
1057 boolean_t will_modify)
1058 {
1059 zfs_acl_t *aclp;
1060 int aclsize = 0;
1061 int acl_count = 0;
1062 zfs_acl_node_t *aclnode;
1063 zfs_acl_phys_t znode_acl;
1064 int version;
1065 int error;
1066 boolean_t drop_lock = B_FALSE;
1067
1068 ASSERT(MUTEX_HELD(&zp->z_acl_lock));
1069
1070 if (zp->z_acl_cached && !will_modify) {
1071 *aclpp = zp->z_acl_cached;
1072 return (0);
1073 }
1074
1075 /*
1076 * close race where znode could be upgrade while trying to
1077 * read the znode attributes.
1078 *
1079 * But this could only happen if the file isn't already an SA
1080 * znode
1081 */
1082 if (!zp->z_is_sa && !have_lock) {
1083 mutex_enter(&zp->z_lock);
1084 drop_lock = B_TRUE;
1085 }
1086 version = zfs_znode_acl_version(zp);
1087
1088 if ((error = zfs_acl_znode_info(zp, &aclsize,
1089 &acl_count, &znode_acl)) != 0) {
1090 goto done;
1091 }
1092
1093 aclp = zfs_acl_alloc(version);
1094
1095 aclp->z_acl_count = acl_count;
1096 aclp->z_acl_bytes = aclsize;
1097
1098 aclnode = zfs_acl_node_alloc(aclsize);
1099 aclnode->z_ace_count = aclp->z_acl_count;
1100 aclnode->z_size = aclsize;
1101
1102 if (!zp->z_is_sa) {
1103 if (znode_acl.z_acl_extern_obj) {
1104 error = dmu_read(ZTOZSB(zp)->z_os,
1105 znode_acl.z_acl_extern_obj, 0, aclnode->z_size,
1106 aclnode->z_acldata, DMU_READ_PREFETCH);
1107 } else {
1108 bcopy(znode_acl.z_ace_data, aclnode->z_acldata,
1109 aclnode->z_size);
1110 }
1111 } else {
1112 error = sa_lookup(zp->z_sa_hdl, SA_ZPL_DACL_ACES(ZTOZSB(zp)),
1113 aclnode->z_acldata, aclnode->z_size);
1114 }
1115
1116 if (error != 0) {
1117 zfs_acl_free(aclp);
1118 zfs_acl_node_free(aclnode);
1119 /* convert checksum errors into IO errors */
1120 if (error == ECKSUM)
1121 error = SET_ERROR(EIO);
1122 goto done;
1123 }
1124
1125 list_insert_head(&aclp->z_acl, aclnode);
1126
1127 *aclpp = aclp;
1128 if (!will_modify)
1129 zp->z_acl_cached = aclp;
1130 done:
1131 if (drop_lock)
1132 mutex_exit(&zp->z_lock);
1133 return (error);
1134 }
1135
1136 /*ARGSUSED*/
1137 void
1138 zfs_acl_data_locator(void **dataptr, uint32_t *length, uint32_t buflen,
1139 boolean_t start, void *userdata)
1140 {
1141 zfs_acl_locator_cb_t *cb = (zfs_acl_locator_cb_t *)userdata;
1142
1143 if (start) {
1144 cb->cb_acl_node = list_head(&cb->cb_aclp->z_acl);
1145 } else {
1146 cb->cb_acl_node = list_next(&cb->cb_aclp->z_acl,
1147 cb->cb_acl_node);
1148 }
1149 *dataptr = cb->cb_acl_node->z_acldata;
1150 *length = cb->cb_acl_node->z_size;
1151 }
1152
1153 int
1154 zfs_acl_chown_setattr(znode_t *zp)
1155 {
1156 int error;
1157 zfs_acl_t *aclp;
1158
1159 if (ZTOZSB(zp)->z_acl_type == ZFS_ACLTYPE_POSIXACL)
1160 return 0;
1161
1162 ASSERT(MUTEX_HELD(&zp->z_lock));
1163 ASSERT(MUTEX_HELD(&zp->z_acl_lock));
1164
1165 if ((error = zfs_acl_node_read(zp, B_TRUE, &aclp, B_FALSE)) == 0)
1166 zp->z_mode = zfs_mode_compute(zp->z_mode, aclp,
1167 &zp->z_pflags, zp->z_uid, zp->z_gid);
1168 return (error);
1169 }
1170
1171 static void
1172 acl_trivial_access_masks(mode_t mode, uint32_t *allow0, uint32_t *deny1,
1173 uint32_t *deny2, uint32_t *owner, uint32_t *group, uint32_t *everyone)
1174 {
1175 *deny1 = *deny2 = *allow0 = *group = 0;
1176
1177 if (!(mode & S_IRUSR) && (mode & (S_IRGRP|S_IROTH)))
1178 *deny1 |= ACE_READ_DATA;
1179 if (!(mode & S_IWUSR) && (mode & (S_IWGRP|S_IWOTH)))
1180 *deny1 |= ACE_WRITE_DATA;
1181 if (!(mode & S_IXUSR) && (mode & (S_IXGRP|S_IXOTH)))
1182 *deny1 |= ACE_EXECUTE;
1183
1184 if (!(mode & S_IRGRP) && (mode & S_IROTH))
1185 *deny2 = ACE_READ_DATA;
1186 if (!(mode & S_IWGRP) && (mode & S_IWOTH))
1187 *deny2 |= ACE_WRITE_DATA;
1188 if (!(mode & S_IXGRP) && (mode & S_IXOTH))
1189 *deny2 |= ACE_EXECUTE;
1190
1191 if ((mode & S_IRUSR) && (!(mode & S_IRGRP) && (mode & S_IROTH)))
1192 *allow0 |= ACE_READ_DATA;
1193 if ((mode & S_IWUSR) && (!(mode & S_IWGRP) && (mode & S_IWOTH)))
1194 *allow0 |= ACE_WRITE_DATA;
1195 if ((mode & S_IXUSR) && (!(mode & S_IXGRP) && (mode & S_IXOTH)))
1196 *allow0 |= ACE_EXECUTE;
1197
1198 *owner = ACE_WRITE_ATTRIBUTES|ACE_WRITE_OWNER|ACE_WRITE_ACL|
1199 ACE_WRITE_NAMED_ATTRS|ACE_READ_ACL|ACE_READ_ATTRIBUTES|
1200 ACE_READ_NAMED_ATTRS|ACE_SYNCHRONIZE;
1201 if (mode & S_IRUSR)
1202 *owner |= ACE_READ_DATA;
1203 if (mode & S_IWUSR)
1204 *owner |= ACE_WRITE_DATA|ACE_APPEND_DATA;
1205 if (mode & S_IXUSR)
1206 *owner |= ACE_EXECUTE;
1207
1208 *group = ACE_READ_ACL|ACE_READ_ATTRIBUTES| ACE_READ_NAMED_ATTRS|
1209 ACE_SYNCHRONIZE;
1210 if (mode & S_IRGRP)
1211 *group |= ACE_READ_DATA;
1212 if (mode & S_IWGRP)
1213 *group |= ACE_WRITE_DATA|ACE_APPEND_DATA;
1214 if (mode & S_IXGRP)
1215 *group |= ACE_EXECUTE;
1216
1217 *everyone = ACE_READ_ACL|ACE_READ_ATTRIBUTES| ACE_READ_NAMED_ATTRS|
1218 ACE_SYNCHRONIZE;
1219 if (mode & S_IROTH)
1220 *everyone |= ACE_READ_DATA;
1221 if (mode & S_IWOTH)
1222 *everyone |= ACE_WRITE_DATA|ACE_APPEND_DATA;
1223 if (mode & S_IXOTH)
1224 *everyone |= ACE_EXECUTE;
1225 }
1226
1227 /*
1228 * ace_trivial:
1229 * determine whether an ace_t acl is trivial
1230 *
1231 * Trivialness implies that the acl is composed of only
1232 * owner, group, everyone entries. ACL can't
1233 * have read_acl denied, and write_owner/write_acl/write_attributes
1234 * can only be owner@ entry.
1235 */
1236 static int
1237 ace_trivial_common(void *acep, int aclcnt,
1238 uint64_t (*walk)(void *, uint64_t, int aclcnt,
1239 uint16_t *, uint16_t *, uint32_t *))
1240 {
1241 uint16_t flags;
1242 uint32_t mask;
1243 uint16_t type;
1244 uint64_t cookie = 0;
1245
1246 while ((cookie = walk(acep, cookie, aclcnt, &flags, &type, &mask))) {
1247 switch (flags & ACE_TYPE_FLAGS) {
1248 case ACE_OWNER:
1249 case ACE_GROUP|ACE_IDENTIFIER_GROUP:
1250 case ACE_EVERYONE:
1251 break;
1252 default:
1253 return (1);
1254 }
1255
1256 if (flags & (ACE_FILE_INHERIT_ACE|
1257 ACE_DIRECTORY_INHERIT_ACE|ACE_NO_PROPAGATE_INHERIT_ACE|
1258 ACE_INHERIT_ONLY_ACE))
1259 return (1);
1260
1261 /*
1262 * Special check for some special bits
1263 *
1264 * Don't allow anybody to deny reading basic
1265 * attributes or a files ACL.
1266 */
1267 if ((mask & (ACE_READ_ACL|ACE_READ_ATTRIBUTES)) &&
1268 (type == ACE_ACCESS_DENIED_ACE_TYPE))
1269 return (1);
1270
1271 /*
1272 * Delete permissions are never set by default
1273 */
1274 if (mask & (ACE_DELETE|ACE_DELETE_CHILD))
1275 return (1);
1276 /*
1277 * only allow owner@ to have
1278 * write_acl/write_owner/write_attributes/write_xattr/
1279 */
1280 if (type == ACE_ACCESS_ALLOWED_ACE_TYPE &&
1281 (!(flags & ACE_OWNER) && (mask &
1282 (ACE_WRITE_OWNER|ACE_WRITE_ACL| ACE_WRITE_ATTRIBUTES|
1283 ACE_WRITE_NAMED_ATTRS))))
1284 return (1);
1285
1286 }
1287
1288 return (0);
1289 }
1290
1291 /*
1292 * common code for setting ACLs.
1293 *
1294 * This function is called from zfs_mode_update, zfs_perm_init, and zfs_setacl.
1295 * zfs_setacl passes a non-NULL inherit pointer (ihp) to indicate that it's
1296 * already checked the acl and knows whether to inherit.
1297 */
1298 int
1299 zfs_aclset_common(znode_t *zp, zfs_acl_t *aclp, cred_t *cr, dmu_tx_t *tx)
1300 {
1301 int error;
1302 zfs_sb_t *zsb = ZTOZSB(zp);
1303 dmu_object_type_t otype;
1304 zfs_acl_locator_cb_t locate = { 0 };
1305 uint64_t mode;
1306 sa_bulk_attr_t bulk[5];
1307 uint64_t ctime[2];
1308 int count = 0;
1309
1310 mode = zp->z_mode;
1311
1312 mode = zfs_mode_compute(mode, aclp, &zp->z_pflags,
1313 zp->z_uid, zp->z_gid);
1314
1315 zp->z_mode = mode;
1316 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zsb), NULL,
1317 &mode, sizeof (mode));
1318 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zsb), NULL,
1319 &zp->z_pflags, sizeof (zp->z_pflags));
1320 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zsb), NULL,
1321 &ctime, sizeof (ctime));
1322
1323 if (zp->z_acl_cached) {
1324 zfs_acl_free(zp->z_acl_cached);
1325 zp->z_acl_cached = NULL;
1326 }
1327
1328 /*
1329 * Upgrade needed?
1330 */
1331 if (!zsb->z_use_fuids) {
1332 otype = DMU_OT_OLDACL;
1333 } else {
1334 if ((aclp->z_version == ZFS_ACL_VERSION_INITIAL) &&
1335 (zsb->z_version >= ZPL_VERSION_FUID))
1336 zfs_acl_xform(zp, aclp, cr);
1337 ASSERT(aclp->z_version >= ZFS_ACL_VERSION_FUID);
1338 otype = DMU_OT_ACL;
1339 }
1340
1341 /*
1342 * Arrgh, we have to handle old on disk format
1343 * as well as newer (preferred) SA format.
1344 */
1345
1346 if (zp->z_is_sa) { /* the easy case, just update the ACL attribute */
1347 locate.cb_aclp = aclp;
1348 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_DACL_ACES(zsb),
1349 zfs_acl_data_locator, &locate, aclp->z_acl_bytes);
1350 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_DACL_COUNT(zsb),
1351 NULL, &aclp->z_acl_count, sizeof (uint64_t));
1352 } else { /* Painful legacy way */
1353 zfs_acl_node_t *aclnode;
1354 uint64_t off = 0;
1355 zfs_acl_phys_t acl_phys;
1356 uint64_t aoid;
1357
1358 if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_ZNODE_ACL(zsb),
1359 &acl_phys, sizeof (acl_phys))) != 0)
1360 return (error);
1361
1362 aoid = acl_phys.z_acl_extern_obj;
1363
1364 if (aclp->z_acl_bytes > ZFS_ACE_SPACE) {
1365 /*
1366 * If ACL was previously external and we are now
1367 * converting to new ACL format then release old
1368 * ACL object and create a new one.
1369 */
1370 if (aoid &&
1371 aclp->z_version != acl_phys.z_acl_version) {
1372 error = dmu_object_free(zsb->z_os, aoid, tx);
1373 if (error)
1374 return (error);
1375 aoid = 0;
1376 }
1377 if (aoid == 0) {
1378 aoid = dmu_object_alloc(zsb->z_os,
1379 otype, aclp->z_acl_bytes,
1380 otype == DMU_OT_ACL ?
1381 DMU_OT_SYSACL : DMU_OT_NONE,
1382 otype == DMU_OT_ACL ?
1383 DN_MAX_BONUSLEN : 0, tx);
1384 } else {
1385 (void) dmu_object_set_blocksize(zsb->z_os,
1386 aoid, aclp->z_acl_bytes, 0, tx);
1387 }
1388 acl_phys.z_acl_extern_obj = aoid;
1389 for (aclnode = list_head(&aclp->z_acl); aclnode;
1390 aclnode = list_next(&aclp->z_acl, aclnode)) {
1391 if (aclnode->z_ace_count == 0)
1392 continue;
1393 dmu_write(zsb->z_os, aoid, off,
1394 aclnode->z_size, aclnode->z_acldata, tx);
1395 off += aclnode->z_size;
1396 }
1397 } else {
1398 void *start = acl_phys.z_ace_data;
1399 /*
1400 * Migrating back embedded?
1401 */
1402 if (acl_phys.z_acl_extern_obj) {
1403 error = dmu_object_free(zsb->z_os,
1404 acl_phys.z_acl_extern_obj, tx);
1405 if (error)
1406 return (error);
1407 acl_phys.z_acl_extern_obj = 0;
1408 }
1409
1410 for (aclnode = list_head(&aclp->z_acl); aclnode;
1411 aclnode = list_next(&aclp->z_acl, aclnode)) {
1412 if (aclnode->z_ace_count == 0)
1413 continue;
1414 bcopy(aclnode->z_acldata, start,
1415 aclnode->z_size);
1416 start = (caddr_t)start + aclnode->z_size;
1417 }
1418 }
1419 /*
1420 * If Old version then swap count/bytes to match old
1421 * layout of znode_acl_phys_t.
1422 */
1423 if (aclp->z_version == ZFS_ACL_VERSION_INITIAL) {
1424 acl_phys.z_acl_size = aclp->z_acl_count;
1425 acl_phys.z_acl_count = aclp->z_acl_bytes;
1426 } else {
1427 acl_phys.z_acl_size = aclp->z_acl_bytes;
1428 acl_phys.z_acl_count = aclp->z_acl_count;
1429 }
1430 acl_phys.z_acl_version = aclp->z_version;
1431
1432 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ZNODE_ACL(zsb), NULL,
1433 &acl_phys, sizeof (acl_phys));
1434 }
1435
1436 /*
1437 * Replace ACL wide bits, but first clear them.
1438 */
1439 zp->z_pflags &= ~ZFS_ACL_WIDE_FLAGS;
1440
1441 zp->z_pflags |= aclp->z_hints;
1442
1443 if (ace_trivial_common(aclp, 0, zfs_ace_walk) == 0)
1444 zp->z_pflags |= ZFS_ACL_TRIVIAL;
1445
1446 zfs_tstamp_update_setup(zp, STATE_CHANGED, NULL, ctime, B_TRUE);
1447 return (sa_bulk_update(zp->z_sa_hdl, bulk, count, tx));
1448 }
1449
1450 static void
1451 zfs_acl_chmod(zfs_sb_t *zsb, uint64_t mode, zfs_acl_t *aclp)
1452 {
1453 void *acep = NULL;
1454 uint64_t who;
1455 int new_count, new_bytes;
1456 int ace_size;
1457 int entry_type;
1458 uint16_t iflags, type;
1459 uint32_t access_mask;
1460 zfs_acl_node_t *newnode;
1461 size_t abstract_size = aclp->z_ops->ace_abstract_size();
1462 void *zacep;
1463 uint32_t owner, group, everyone;
1464 uint32_t deny1, deny2, allow0;
1465
1466 new_count = new_bytes = 0;
1467
1468 acl_trivial_access_masks((mode_t)mode, &allow0, &deny1, &deny2,
1469 &owner, &group, &everyone);
1470
1471 newnode = zfs_acl_node_alloc((abstract_size * 6) + aclp->z_acl_bytes);
1472
1473 zacep = newnode->z_acldata;
1474 if (allow0) {
1475 zfs_set_ace(aclp, zacep, allow0, ALLOW, -1, ACE_OWNER);
1476 zacep = (void *)((uintptr_t)zacep + abstract_size);
1477 new_count++;
1478 new_bytes += abstract_size;
1479 } if (deny1) {
1480 zfs_set_ace(aclp, zacep, deny1, DENY, -1, ACE_OWNER);
1481 zacep = (void *)((uintptr_t)zacep + abstract_size);
1482 new_count++;
1483 new_bytes += abstract_size;
1484 }
1485 if (deny2) {
1486 zfs_set_ace(aclp, zacep, deny2, DENY, -1, OWNING_GROUP);
1487 zacep = (void *)((uintptr_t)zacep + abstract_size);
1488 new_count++;
1489 new_bytes += abstract_size;
1490 }
1491
1492 while ((acep = zfs_acl_next_ace(aclp, acep, &who, &access_mask,
1493 &iflags, &type))) {
1494 uint16_t inherit_flags;
1495
1496 entry_type = (iflags & ACE_TYPE_FLAGS);
1497 inherit_flags = (iflags & ALL_INHERIT);
1498
1499 if ((entry_type == ACE_OWNER || entry_type == ACE_EVERYONE ||
1500 (entry_type == OWNING_GROUP)) &&
1501 ((inherit_flags & ACE_INHERIT_ONLY_ACE) == 0)) {
1502 continue;
1503 }
1504
1505 if ((type != ALLOW && type != DENY) ||
1506 (inherit_flags & ACE_INHERIT_ONLY_ACE)) {
1507 if (inherit_flags)
1508 aclp->z_hints |= ZFS_INHERIT_ACE;
1509 switch (type) {
1510 case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE:
1511 case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE:
1512 case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE:
1513 case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE:
1514 aclp->z_hints |= ZFS_ACL_OBJ_ACE;
1515 break;
1516 }
1517 } else {
1518
1519 /*
1520 * Limit permissions to be no greater than
1521 * group permissions
1522 */
1523 if (zsb->z_acl_inherit == ZFS_ACL_RESTRICTED) {
1524 if (!(mode & S_IRGRP))
1525 access_mask &= ~ACE_READ_DATA;
1526 if (!(mode & S_IWGRP))
1527 access_mask &=
1528 ~(ACE_WRITE_DATA|ACE_APPEND_DATA);
1529 if (!(mode & S_IXGRP))
1530 access_mask &= ~ACE_EXECUTE;
1531 access_mask &=
1532 ~(ACE_WRITE_OWNER|ACE_WRITE_ACL|
1533 ACE_WRITE_ATTRIBUTES|ACE_WRITE_NAMED_ATTRS);
1534 }
1535 }
1536 zfs_set_ace(aclp, zacep, access_mask, type, who, iflags);
1537 ace_size = aclp->z_ops->ace_size(acep);
1538 zacep = (void *)((uintptr_t)zacep + ace_size);
1539 new_count++;
1540 new_bytes += ace_size;
1541 }
1542 zfs_set_ace(aclp, zacep, owner, 0, -1, ACE_OWNER);
1543 zacep = (void *)((uintptr_t)zacep + abstract_size);
1544 zfs_set_ace(aclp, zacep, group, 0, -1, OWNING_GROUP);
1545 zacep = (void *)((uintptr_t)zacep + abstract_size);
1546 zfs_set_ace(aclp, zacep, everyone, 0, -1, ACE_EVERYONE);
1547
1548 new_count += 3;
1549 new_bytes += abstract_size * 3;
1550 zfs_acl_release_nodes(aclp);
1551 aclp->z_acl_count = new_count;
1552 aclp->z_acl_bytes = new_bytes;
1553 newnode->z_ace_count = new_count;
1554 newnode->z_size = new_bytes;
1555 list_insert_tail(&aclp->z_acl, newnode);
1556 }
1557
1558 void
1559 zfs_acl_chmod_setattr(znode_t *zp, zfs_acl_t **aclp, uint64_t mode)
1560 {
1561 mutex_enter(&zp->z_acl_lock);
1562 mutex_enter(&zp->z_lock);
1563 *aclp = zfs_acl_alloc(zfs_acl_version_zp(zp));
1564 (*aclp)->z_hints = zp->z_pflags & V4_ACL_WIDE_FLAGS;
1565 zfs_acl_chmod(ZTOZSB(zp), mode, *aclp);
1566 mutex_exit(&zp->z_lock);
1567 mutex_exit(&zp->z_acl_lock);
1568 ASSERT(*aclp);
1569 }
1570
1571 /*
1572 * strip off write_owner and write_acl
1573 */
1574 static void
1575 zfs_restricted_update(zfs_sb_t *zsb, zfs_acl_t *aclp, void *acep)
1576 {
1577 uint32_t mask = aclp->z_ops->ace_mask_get(acep);
1578
1579 if ((zsb->z_acl_inherit == ZFS_ACL_RESTRICTED) &&
1580 (aclp->z_ops->ace_type_get(acep) == ALLOW)) {
1581 mask &= ~RESTRICTED_CLEAR;
1582 aclp->z_ops->ace_mask_set(acep, mask);
1583 }
1584 }
1585
1586 /*
1587 * Should ACE be inherited?
1588 */
1589 static int
1590 zfs_ace_can_use(umode_t obj_mode, uint16_t acep_flags)
1591 {
1592 int iflags = (acep_flags & 0xf);
1593
1594 if (S_ISDIR(obj_mode) && (iflags & ACE_DIRECTORY_INHERIT_ACE))
1595 return (1);
1596 else if (iflags & ACE_FILE_INHERIT_ACE)
1597 return (!(S_ISDIR(obj_mode) &&
1598 (iflags & ACE_NO_PROPAGATE_INHERIT_ACE)));
1599 return (0);
1600 }
1601
1602 /*
1603 * inherit inheritable ACEs from parent
1604 */
1605 static zfs_acl_t *
1606 zfs_acl_inherit(zfs_sb_t *zsb, umode_t obj_mode, zfs_acl_t *paclp,
1607 uint64_t mode, boolean_t *need_chmod)
1608 {
1609 void *pacep;
1610 void *acep;
1611 zfs_acl_node_t *aclnode;
1612 zfs_acl_t *aclp = NULL;
1613 uint64_t who;
1614 uint32_t access_mask;
1615 uint16_t iflags, newflags, type;
1616 size_t ace_size;
1617 void *data1, *data2;
1618 size_t data1sz, data2sz;
1619 boolean_t vdir = S_ISDIR(obj_mode);
1620 boolean_t vreg = S_ISREG(obj_mode);
1621 boolean_t passthrough, passthrough_x, noallow;
1622
1623 passthrough_x =
1624 zsb->z_acl_inherit == ZFS_ACL_PASSTHROUGH_X;
1625 passthrough = passthrough_x ||
1626 zsb->z_acl_inherit == ZFS_ACL_PASSTHROUGH;
1627 noallow =
1628 zsb->z_acl_inherit == ZFS_ACL_NOALLOW;
1629
1630 *need_chmod = B_TRUE;
1631 pacep = NULL;
1632 aclp = zfs_acl_alloc(paclp->z_version);
1633 if (zsb->z_acl_inherit == ZFS_ACL_DISCARD || S_ISLNK(obj_mode))
1634 return (aclp);
1635 while ((pacep = zfs_acl_next_ace(paclp, pacep, &who,
1636 &access_mask, &iflags, &type))) {
1637
1638 /*
1639 * don't inherit bogus ACEs
1640 */
1641 if (!zfs_acl_valid_ace_type(type, iflags))
1642 continue;
1643
1644 if (noallow && type == ALLOW)
1645 continue;
1646
1647 ace_size = aclp->z_ops->ace_size(pacep);
1648
1649 if (!zfs_ace_can_use(obj_mode, iflags))
1650 continue;
1651
1652 /*
1653 * If owner@, group@, or everyone@ inheritable
1654 * then zfs_acl_chmod() isn't needed.
1655 */
1656 if (passthrough &&
1657 ((iflags & (ACE_OWNER|ACE_EVERYONE)) ||
1658 ((iflags & OWNING_GROUP) ==
1659 OWNING_GROUP)) && (vreg || (vdir && (iflags &
1660 ACE_DIRECTORY_INHERIT_ACE)))) {
1661 *need_chmod = B_FALSE;
1662 }
1663
1664 if (!vdir && passthrough_x &&
1665 ((mode & (S_IXUSR | S_IXGRP | S_IXOTH)) == 0)) {
1666 access_mask &= ~ACE_EXECUTE;
1667 }
1668
1669 aclnode = zfs_acl_node_alloc(ace_size);
1670 list_insert_tail(&aclp->z_acl, aclnode);
1671 acep = aclnode->z_acldata;
1672
1673 zfs_set_ace(aclp, acep, access_mask, type,
1674 who, iflags|ACE_INHERITED_ACE);
1675
1676 /*
1677 * Copy special opaque data if any
1678 */
1679 if ((data1sz = paclp->z_ops->ace_data(pacep, &data1)) != 0) {
1680 VERIFY((data2sz = aclp->z_ops->ace_data(acep,
1681 &data2)) == data1sz);
1682 bcopy(data1, data2, data2sz);
1683 }
1684
1685 aclp->z_acl_count++;
1686 aclnode->z_ace_count++;
1687 aclp->z_acl_bytes += aclnode->z_size;
1688 newflags = aclp->z_ops->ace_flags_get(acep);
1689
1690 if (vdir)
1691 aclp->z_hints |= ZFS_INHERIT_ACE;
1692
1693 if ((iflags & ACE_NO_PROPAGATE_INHERIT_ACE) || !vdir) {
1694 newflags &= ~ALL_INHERIT;
1695 aclp->z_ops->ace_flags_set(acep,
1696 newflags|ACE_INHERITED_ACE);
1697 zfs_restricted_update(zsb, aclp, acep);
1698 continue;
1699 }
1700
1701 ASSERT(vdir);
1702
1703 /*
1704 * If only FILE_INHERIT is set then turn on
1705 * inherit_only
1706 */
1707 if ((iflags & (ACE_FILE_INHERIT_ACE |
1708 ACE_DIRECTORY_INHERIT_ACE)) == ACE_FILE_INHERIT_ACE) {
1709 newflags |= ACE_INHERIT_ONLY_ACE;
1710 aclp->z_ops->ace_flags_set(acep,
1711 newflags|ACE_INHERITED_ACE);
1712 } else {
1713 newflags &= ~ACE_INHERIT_ONLY_ACE;
1714 aclp->z_ops->ace_flags_set(acep,
1715 newflags|ACE_INHERITED_ACE);
1716 }
1717 }
1718 return (aclp);
1719 }
1720
1721 /*
1722 * Create file system object initial permissions
1723 * including inheritable ACEs.
1724 */
1725 int
1726 zfs_acl_ids_create(znode_t *dzp, int flag, vattr_t *vap, cred_t *cr,
1727 vsecattr_t *vsecp, zfs_acl_ids_t *acl_ids)
1728 {
1729 int error;
1730 zfs_sb_t *zsb = ZTOZSB(dzp);
1731 zfs_acl_t *paclp;
1732 #ifdef HAVE_KSID
1733 gid_t gid;
1734 #endif /* HAVE_KSID */
1735 boolean_t need_chmod = B_TRUE;
1736 boolean_t inherited = B_FALSE;
1737
1738 bzero(acl_ids, sizeof (zfs_acl_ids_t));
1739 acl_ids->z_mode = vap->va_mode;
1740
1741 if (vsecp)
1742 if ((error = zfs_vsec_2_aclp(zsb, vap->va_mode, vsecp,
1743 cr, &acl_ids->z_fuidp, &acl_ids->z_aclp)) != 0)
1744 return (error);
1745
1746 acl_ids->z_fuid = vap->va_uid;
1747 acl_ids->z_fgid = vap->va_gid;
1748 #ifdef HAVE_KSID
1749 /*
1750 * Determine uid and gid.
1751 */
1752 if ((flag & IS_ROOT_NODE) || zsb->z_replay ||
1753 ((flag & IS_XATTR) && (S_ISDIR(vap->va_mode)))) {
1754 acl_ids->z_fuid = zfs_fuid_create(zsb, (uint64_t)vap->va_uid,
1755 cr, ZFS_OWNER, &acl_ids->z_fuidp);
1756 acl_ids->z_fgid = zfs_fuid_create(zsb, (uint64_t)vap->va_gid,
1757 cr, ZFS_GROUP, &acl_ids->z_fuidp);
1758 gid = vap->va_gid;
1759 } else {
1760 acl_ids->z_fuid = zfs_fuid_create_cred(zsb, ZFS_OWNER,
1761 cr, &acl_ids->z_fuidp);
1762 acl_ids->z_fgid = 0;
1763 if (vap->va_mask & AT_GID) {
1764 acl_ids->z_fgid = zfs_fuid_create(zsb,
1765 (uint64_t)vap->va_gid,
1766 cr, ZFS_GROUP, &acl_ids->z_fuidp);
1767 gid = vap->va_gid;
1768 if (acl_ids->z_fgid != dzp->z_gid &&
1769 !groupmember(vap->va_gid, cr) &&
1770 secpolicy_vnode_create_gid(cr) != 0)
1771 acl_ids->z_fgid = 0;
1772 }
1773 if (acl_ids->z_fgid == 0) {
1774 if (dzp->z_mode & S_ISGID) {
1775 char *domain;
1776 uint32_t rid;
1777
1778 acl_ids->z_fgid = dzp->z_gid;
1779 gid = zfs_fuid_map_id(zsb, acl_ids->z_fgid,
1780 cr, ZFS_GROUP);
1781
1782 if (zsb->z_use_fuids &&
1783 IS_EPHEMERAL(acl_ids->z_fgid)) {
1784 domain = zfs_fuid_idx_domain(
1785 &zsb->z_fuid_idx,
1786 FUID_INDEX(acl_ids->z_fgid));
1787 rid = FUID_RID(acl_ids->z_fgid);
1788 zfs_fuid_node_add(&acl_ids->z_fuidp,
1789 domain, rid,
1790 FUID_INDEX(acl_ids->z_fgid),
1791 acl_ids->z_fgid, ZFS_GROUP);
1792 }
1793 } else {
1794 acl_ids->z_fgid = zfs_fuid_create_cred(zsb,
1795 ZFS_GROUP, cr, &acl_ids->z_fuidp);
1796 gid = crgetgid(cr);
1797 }
1798 }
1799 }
1800 #endif /* HAVE_KSID */
1801
1802 /*
1803 * If we're creating a directory, and the parent directory has the
1804 * set-GID bit set, set in on the new directory.
1805 * Otherwise, if the user is neither privileged nor a member of the
1806 * file's new group, clear the file's set-GID bit.
1807 */
1808
1809 if (!(flag & IS_ROOT_NODE) && (dzp->z_mode & S_ISGID) &&
1810 (S_ISDIR(vap->va_mode))) {
1811 acl_ids->z_mode |= S_ISGID;
1812 } else {
1813 if ((acl_ids->z_mode & S_ISGID) &&
1814 secpolicy_vnode_setids_setgids(cr, gid) != 0)
1815 acl_ids->z_mode &= ~S_ISGID;
1816 }
1817
1818 if (acl_ids->z_aclp == NULL) {
1819 mutex_enter(&dzp->z_acl_lock);
1820 mutex_enter(&dzp->z_lock);
1821 if (!(flag & IS_ROOT_NODE) && (S_ISDIR(ZTOI(dzp)->i_mode) &&
1822 (dzp->z_pflags & ZFS_INHERIT_ACE)) &&
1823 !(dzp->z_pflags & ZFS_XATTR)) {
1824 VERIFY(0 == zfs_acl_node_read(dzp, B_TRUE,
1825 &paclp, B_FALSE));
1826 acl_ids->z_aclp = zfs_acl_inherit(zsb,
1827 vap->va_mode, paclp, acl_ids->z_mode, &need_chmod);
1828 inherited = B_TRUE;
1829 } else {
1830 acl_ids->z_aclp =
1831 zfs_acl_alloc(zfs_acl_version_zp(dzp));
1832 acl_ids->z_aclp->z_hints |= ZFS_ACL_TRIVIAL;
1833 }
1834 mutex_exit(&dzp->z_lock);
1835 mutex_exit(&dzp->z_acl_lock);
1836 if (need_chmod) {
1837 acl_ids->z_aclp->z_hints |= S_ISDIR(vap->va_mode) ?
1838 ZFS_ACL_AUTO_INHERIT : 0;
1839 zfs_acl_chmod(zsb, acl_ids->z_mode, acl_ids->z_aclp);
1840 }
1841 }
1842
1843 if (inherited || vsecp) {
1844 acl_ids->z_mode = zfs_mode_compute(acl_ids->z_mode,
1845 acl_ids->z_aclp, &acl_ids->z_aclp->z_hints,
1846 acl_ids->z_fuid, acl_ids->z_fgid);
1847 if (ace_trivial_common(acl_ids->z_aclp, 0, zfs_ace_walk) == 0)
1848 acl_ids->z_aclp->z_hints |= ZFS_ACL_TRIVIAL;
1849 }
1850
1851 return (0);
1852 }
1853
1854 /*
1855 * Free ACL and fuid_infop, but not the acl_ids structure
1856 */
1857 void
1858 zfs_acl_ids_free(zfs_acl_ids_t *acl_ids)
1859 {
1860 if (acl_ids->z_aclp)
1861 zfs_acl_free(acl_ids->z_aclp);
1862 if (acl_ids->z_fuidp)
1863 zfs_fuid_info_free(acl_ids->z_fuidp);
1864 acl_ids->z_aclp = NULL;
1865 acl_ids->z_fuidp = NULL;
1866 }
1867
1868 boolean_t
1869 zfs_acl_ids_overquota(zfs_sb_t *zsb, zfs_acl_ids_t *acl_ids)
1870 {
1871 return (zfs_fuid_overquota(zsb, B_FALSE, acl_ids->z_fuid) ||
1872 zfs_fuid_overquota(zsb, B_TRUE, acl_ids->z_fgid));
1873 }
1874
1875 /*
1876 * Retrieve a files ACL
1877 */
1878 int
1879 zfs_getacl(znode_t *zp, vsecattr_t *vsecp, boolean_t skipaclchk, cred_t *cr)
1880 {
1881 zfs_acl_t *aclp;
1882 ulong_t mask;
1883 int error;
1884 int count = 0;
1885 int largeace = 0;
1886
1887 mask = vsecp->vsa_mask & (VSA_ACE | VSA_ACECNT |
1888 VSA_ACE_ACLFLAGS | VSA_ACE_ALLTYPES);
1889
1890 if (mask == 0)
1891 return (SET_ERROR(ENOSYS));
1892
1893 if ((error = zfs_zaccess(zp, ACE_READ_ACL, 0, skipaclchk, cr)))
1894 return (error);
1895
1896 mutex_enter(&zp->z_acl_lock);
1897
1898 error = zfs_acl_node_read(zp, B_FALSE, &aclp, B_FALSE);
1899 if (error != 0) {
1900 mutex_exit(&zp->z_acl_lock);
1901 return (error);
1902 }
1903
1904 /*
1905 * Scan ACL to determine number of ACEs
1906 */
1907 if ((zp->z_pflags & ZFS_ACL_OBJ_ACE) && !(mask & VSA_ACE_ALLTYPES)) {
1908 void *zacep = NULL;
1909 uint64_t who;
1910 uint32_t access_mask;
1911 uint16_t type, iflags;
1912
1913 while ((zacep = zfs_acl_next_ace(aclp, zacep,
1914 &who, &access_mask, &iflags, &type))) {
1915 switch (type) {
1916 case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE:
1917 case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE:
1918 case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE:
1919 case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE:
1920 largeace++;
1921 continue;
1922 default:
1923 count++;
1924 }
1925 }
1926 vsecp->vsa_aclcnt = count;
1927 } else
1928 count = (int)aclp->z_acl_count;
1929
1930 if (mask & VSA_ACECNT) {
1931 vsecp->vsa_aclcnt = count;
1932 }
1933
1934 if (mask & VSA_ACE) {
1935 size_t aclsz;
1936
1937 aclsz = count * sizeof (ace_t) +
1938 sizeof (ace_object_t) * largeace;
1939
1940 vsecp->vsa_aclentp = kmem_alloc(aclsz, KM_SLEEP);
1941 vsecp->vsa_aclentsz = aclsz;
1942
1943 if (aclp->z_version == ZFS_ACL_VERSION_FUID)
1944 zfs_copy_fuid_2_ace(ZTOZSB(zp), aclp, cr,
1945 vsecp->vsa_aclentp, !(mask & VSA_ACE_ALLTYPES));
1946 else {
1947 zfs_acl_node_t *aclnode;
1948 void *start = vsecp->vsa_aclentp;
1949
1950 for (aclnode = list_head(&aclp->z_acl); aclnode;
1951 aclnode = list_next(&aclp->z_acl, aclnode)) {
1952 bcopy(aclnode->z_acldata, start,
1953 aclnode->z_size);
1954 start = (caddr_t)start + aclnode->z_size;
1955 }
1956 ASSERT((caddr_t)start - (caddr_t)vsecp->vsa_aclentp ==
1957 aclp->z_acl_bytes);
1958 }
1959 }
1960 if (mask & VSA_ACE_ACLFLAGS) {
1961 vsecp->vsa_aclflags = 0;
1962 if (zp->z_pflags & ZFS_ACL_DEFAULTED)
1963 vsecp->vsa_aclflags |= ACL_DEFAULTED;
1964 if (zp->z_pflags & ZFS_ACL_PROTECTED)
1965 vsecp->vsa_aclflags |= ACL_PROTECTED;
1966 if (zp->z_pflags & ZFS_ACL_AUTO_INHERIT)
1967 vsecp->vsa_aclflags |= ACL_AUTO_INHERIT;
1968 }
1969
1970 mutex_exit(&zp->z_acl_lock);
1971
1972 return (0);
1973 }
1974
1975 int
1976 zfs_vsec_2_aclp(zfs_sb_t *zsb, umode_t obj_mode,
1977 vsecattr_t *vsecp, cred_t *cr, zfs_fuid_info_t **fuidp, zfs_acl_t **zaclp)
1978 {
1979 zfs_acl_t *aclp;
1980 zfs_acl_node_t *aclnode;
1981 int aclcnt = vsecp->vsa_aclcnt;
1982 int error;
1983
1984 if (vsecp->vsa_aclcnt > MAX_ACL_ENTRIES || vsecp->vsa_aclcnt <= 0)
1985 return (SET_ERROR(EINVAL));
1986
1987 aclp = zfs_acl_alloc(zfs_acl_version(zsb->z_version));
1988
1989 aclp->z_hints = 0;
1990 aclnode = zfs_acl_node_alloc(aclcnt * sizeof (zfs_object_ace_t));
1991 if (aclp->z_version == ZFS_ACL_VERSION_INITIAL) {
1992 if ((error = zfs_copy_ace_2_oldace(obj_mode, aclp,
1993 (ace_t *)vsecp->vsa_aclentp, aclnode->z_acldata,
1994 aclcnt, &aclnode->z_size)) != 0) {
1995 zfs_acl_free(aclp);
1996 zfs_acl_node_free(aclnode);
1997 return (error);
1998 }
1999 } else {
2000 if ((error = zfs_copy_ace_2_fuid(zsb, obj_mode, aclp,
2001 vsecp->vsa_aclentp, aclnode->z_acldata, aclcnt,
2002 &aclnode->z_size, fuidp, cr)) != 0) {
2003 zfs_acl_free(aclp);
2004 zfs_acl_node_free(aclnode);
2005 return (error);
2006 }
2007 }
2008 aclp->z_acl_bytes = aclnode->z_size;
2009 aclnode->z_ace_count = aclcnt;
2010 aclp->z_acl_count = aclcnt;
2011 list_insert_head(&aclp->z_acl, aclnode);
2012
2013 /*
2014 * If flags are being set then add them to z_hints
2015 */
2016 if (vsecp->vsa_mask & VSA_ACE_ACLFLAGS) {
2017 if (vsecp->vsa_aclflags & ACL_PROTECTED)
2018 aclp->z_hints |= ZFS_ACL_PROTECTED;
2019 if (vsecp->vsa_aclflags & ACL_DEFAULTED)
2020 aclp->z_hints |= ZFS_ACL_DEFAULTED;
2021 if (vsecp->vsa_aclflags & ACL_AUTO_INHERIT)
2022 aclp->z_hints |= ZFS_ACL_AUTO_INHERIT;
2023 }
2024
2025 *zaclp = aclp;
2026
2027 return (0);
2028 }
2029
2030 /*
2031 * Set a files ACL
2032 */
2033 int
2034 zfs_setacl(znode_t *zp, vsecattr_t *vsecp, boolean_t skipaclchk, cred_t *cr)
2035 {
2036 zfs_sb_t *zsb = ZTOZSB(zp);
2037 zilog_t *zilog = zsb->z_log;
2038 ulong_t mask = vsecp->vsa_mask & (VSA_ACE | VSA_ACECNT);
2039 dmu_tx_t *tx;
2040 int error;
2041 zfs_acl_t *aclp;
2042 zfs_fuid_info_t *fuidp = NULL;
2043 boolean_t fuid_dirtied;
2044 uint64_t acl_obj;
2045
2046 if (mask == 0)
2047 return (SET_ERROR(ENOSYS));
2048
2049 if (zp->z_pflags & ZFS_IMMUTABLE)
2050 return (SET_ERROR(EPERM));
2051
2052 if ((error = zfs_zaccess(zp, ACE_WRITE_ACL, 0, skipaclchk, cr)))
2053 return (error);
2054
2055 error = zfs_vsec_2_aclp(zsb, ZTOI(zp)->i_mode, vsecp, cr, &fuidp,
2056 &aclp);
2057 if (error)
2058 return (error);
2059
2060 /*
2061 * If ACL wide flags aren't being set then preserve any
2062 * existing flags.
2063 */
2064 if (!(vsecp->vsa_mask & VSA_ACE_ACLFLAGS)) {
2065 aclp->z_hints |=
2066 (zp->z_pflags & V4_ACL_WIDE_FLAGS);
2067 }
2068 top:
2069 mutex_enter(&zp->z_acl_lock);
2070 mutex_enter(&zp->z_lock);
2071
2072 tx = dmu_tx_create(zsb->z_os);
2073
2074 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE);
2075
2076 fuid_dirtied = zsb->z_fuid_dirty;
2077 if (fuid_dirtied)
2078 zfs_fuid_txhold(zsb, tx);
2079
2080 /*
2081 * If old version and ACL won't fit in bonus and we aren't
2082 * upgrading then take out necessary DMU holds
2083 */
2084
2085 if ((acl_obj = zfs_external_acl(zp)) != 0) {
2086 if (zsb->z_version >= ZPL_VERSION_FUID &&
2087 zfs_znode_acl_version(zp) <= ZFS_ACL_VERSION_INITIAL) {
2088 dmu_tx_hold_free(tx, acl_obj, 0,
2089 DMU_OBJECT_END);
2090 dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0,
2091 aclp->z_acl_bytes);
2092 } else {
2093 dmu_tx_hold_write(tx, acl_obj, 0, aclp->z_acl_bytes);
2094 }
2095 } else if (!zp->z_is_sa && aclp->z_acl_bytes > ZFS_ACE_SPACE) {
2096 dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, aclp->z_acl_bytes);
2097 }
2098
2099 zfs_sa_upgrade_txholds(tx, zp);
2100 error = dmu_tx_assign(tx, TXG_NOWAIT);
2101 if (error) {
2102 mutex_exit(&zp->z_acl_lock);
2103 mutex_exit(&zp->z_lock);
2104
2105 if (error == ERESTART) {
2106 dmu_tx_wait(tx);
2107 dmu_tx_abort(tx);
2108 goto top;
2109 }
2110 dmu_tx_abort(tx);
2111 zfs_acl_free(aclp);
2112 return (error);
2113 }
2114
2115 error = zfs_aclset_common(zp, aclp, cr, tx);
2116 ASSERT(error == 0);
2117 ASSERT(zp->z_acl_cached == NULL);
2118 zp->z_acl_cached = aclp;
2119
2120 if (fuid_dirtied)
2121 zfs_fuid_sync(zsb, tx);
2122
2123 zfs_log_acl(zilog, tx, zp, vsecp, fuidp);
2124
2125 if (fuidp)
2126 zfs_fuid_info_free(fuidp);
2127 dmu_tx_commit(tx);
2128
2129 mutex_exit(&zp->z_lock);
2130 mutex_exit(&zp->z_acl_lock);
2131
2132 return (error);
2133 }
2134
2135 /*
2136 * Check accesses of interest (AoI) against attributes of the dataset
2137 * such as read-only. Returns zero if no AoI conflict with dataset
2138 * attributes, otherwise an appropriate errno is returned.
2139 */
2140 static int
2141 zfs_zaccess_dataset_check(znode_t *zp, uint32_t v4_mode)
2142 {
2143 if ((v4_mode & WRITE_MASK) && (zfs_is_readonly(ZTOZSB(zp))) &&
2144 (!S_ISDEV(ZTOI(zp)->i_mode) ||
2145 (S_ISDEV(ZTOI(zp)->i_mode) && (v4_mode & WRITE_MASK_ATTRS)))) {
2146 return (SET_ERROR(EROFS));
2147 }
2148
2149 /*
2150 * Only check for READONLY on non-directories.
2151 */
2152 if ((v4_mode & WRITE_MASK_DATA) &&
2153 ((!S_ISDIR(ZTOI(zp)->i_mode) &&
2154 (zp->z_pflags & (ZFS_READONLY | ZFS_IMMUTABLE))) ||
2155 (S_ISDIR(ZTOI(zp)->i_mode) &&
2156 (zp->z_pflags & ZFS_IMMUTABLE)))) {
2157 return (SET_ERROR(EPERM));
2158 }
2159
2160 if ((v4_mode & (ACE_DELETE | ACE_DELETE_CHILD)) &&
2161 (zp->z_pflags & ZFS_NOUNLINK)) {
2162 return (SET_ERROR(EPERM));
2163 }
2164
2165 if (((v4_mode & (ACE_READ_DATA|ACE_EXECUTE)) &&
2166 (zp->z_pflags & ZFS_AV_QUARANTINED))) {
2167 return (SET_ERROR(EACCES));
2168 }
2169
2170 return (0);
2171 }
2172
2173 /*
2174 * The primary usage of this function is to loop through all of the
2175 * ACEs in the znode, determining what accesses of interest (AoI) to
2176 * the caller are allowed or denied. The AoI are expressed as bits in
2177 * the working_mode parameter. As each ACE is processed, bits covered
2178 * by that ACE are removed from the working_mode. This removal
2179 * facilitates two things. The first is that when the working mode is
2180 * empty (= 0), we know we've looked at all the AoI. The second is
2181 * that the ACE interpretation rules don't allow a later ACE to undo
2182 * something granted or denied by an earlier ACE. Removing the
2183 * discovered access or denial enforces this rule. At the end of
2184 * processing the ACEs, all AoI that were found to be denied are
2185 * placed into the working_mode, giving the caller a mask of denied
2186 * accesses. Returns:
2187 * 0 if all AoI granted
2188 * EACCESS if the denied mask is non-zero
2189 * other error if abnormal failure (e.g., IO error)
2190 *
2191 * A secondary usage of the function is to determine if any of the
2192 * AoI are granted. If an ACE grants any access in
2193 * the working_mode, we immediately short circuit out of the function.
2194 * This mode is chosen by setting anyaccess to B_TRUE. The
2195 * working_mode is not a denied access mask upon exit if the function
2196 * is used in this manner.
2197 */
2198 static int
2199 zfs_zaccess_aces_check(znode_t *zp, uint32_t *working_mode,
2200 boolean_t anyaccess, cred_t *cr)
2201 {
2202 zfs_sb_t *zsb = ZTOZSB(zp);
2203 zfs_acl_t *aclp;
2204 int error;
2205 uid_t uid = crgetuid(cr);
2206 uint64_t who;
2207 uint16_t type, iflags;
2208 uint16_t entry_type;
2209 uint32_t access_mask;
2210 uint32_t deny_mask = 0;
2211 zfs_ace_hdr_t *acep = NULL;
2212 boolean_t checkit;
2213 uid_t gowner;
2214 uid_t fowner;
2215
2216 zfs_fuid_map_ids(zp, cr, &fowner, &gowner);
2217
2218 mutex_enter(&zp->z_acl_lock);
2219
2220 error = zfs_acl_node_read(zp, B_FALSE, &aclp, B_FALSE);
2221 if (error != 0) {
2222 mutex_exit(&zp->z_acl_lock);
2223 return (error);
2224 }
2225
2226 ASSERT(zp->z_acl_cached);
2227
2228 while ((acep = zfs_acl_next_ace(aclp, acep, &who, &access_mask,
2229 &iflags, &type))) {
2230 uint32_t mask_matched;
2231
2232 if (!zfs_acl_valid_ace_type(type, iflags))
2233 continue;
2234
2235 if (S_ISDIR(ZTOI(zp)->i_mode) &&
2236 (iflags & ACE_INHERIT_ONLY_ACE))
2237 continue;
2238
2239 /* Skip ACE if it does not affect any AoI */
2240 mask_matched = (access_mask & *working_mode);
2241 if (!mask_matched)
2242 continue;
2243
2244 entry_type = (iflags & ACE_TYPE_FLAGS);
2245
2246 checkit = B_FALSE;
2247
2248 switch (entry_type) {
2249 case ACE_OWNER:
2250 if (uid == fowner)
2251 checkit = B_TRUE;
2252 break;
2253 case OWNING_GROUP:
2254 who = gowner;
2255 /*FALLTHROUGH*/
2256 case ACE_IDENTIFIER_GROUP:
2257 checkit = zfs_groupmember(zsb, who, cr);
2258 break;
2259 case ACE_EVERYONE:
2260 checkit = B_TRUE;
2261 break;
2262
2263 /* USER Entry */
2264 default:
2265 if (entry_type == 0) {
2266 uid_t newid;
2267
2268 newid = zfs_fuid_map_id(zsb, who, cr,
2269 ZFS_ACE_USER);
2270 if (newid != IDMAP_WK_CREATOR_OWNER_UID &&
2271 uid == newid)
2272 checkit = B_TRUE;
2273 break;
2274 } else {
2275 mutex_exit(&zp->z_acl_lock);
2276 return (SET_ERROR(EIO));
2277 }
2278 }
2279
2280 if (checkit) {
2281 if (type == DENY) {
2282 DTRACE_PROBE3(zfs__ace__denies,
2283 znode_t *, zp,
2284 zfs_ace_hdr_t *, acep,
2285 uint32_t, mask_matched);
2286 deny_mask |= mask_matched;
2287 } else {
2288 DTRACE_PROBE3(zfs__ace__allows,
2289 znode_t *, zp,
2290 zfs_ace_hdr_t *, acep,
2291 uint32_t, mask_matched);
2292 if (anyaccess) {
2293 mutex_exit(&zp->z_acl_lock);
2294 return (0);
2295 }
2296 }
2297 *working_mode &= ~mask_matched;
2298 }
2299
2300 /* Are we done? */
2301 if (*working_mode == 0)
2302 break;
2303 }
2304
2305 mutex_exit(&zp->z_acl_lock);
2306
2307 /* Put the found 'denies' back on the working mode */
2308 if (deny_mask) {
2309 *working_mode |= deny_mask;
2310 return (SET_ERROR(EACCES));
2311 } else if (*working_mode) {
2312 return (-1);
2313 }
2314
2315 return (0);
2316 }
2317
2318 /*
2319 * Return true if any access whatsoever granted, we don't actually
2320 * care what access is granted.
2321 */
2322 boolean_t
2323 zfs_has_access(znode_t *zp, cred_t *cr)
2324 {
2325 uint32_t have = ACE_ALL_PERMS;
2326
2327 if (zfs_zaccess_aces_check(zp, &have, B_TRUE, cr) != 0) {
2328 uid_t owner;
2329
2330 owner = zfs_fuid_map_id(ZTOZSB(zp), zp->z_uid, cr, ZFS_OWNER);
2331 return (secpolicy_vnode_any_access(cr, ZTOI(zp), owner) == 0);
2332 }
2333 return (B_TRUE);
2334 }
2335
2336 static int
2337 zfs_zaccess_common(znode_t *zp, uint32_t v4_mode, uint32_t *working_mode,
2338 boolean_t *check_privs, boolean_t skipaclchk, cred_t *cr)
2339 {
2340 zfs_sb_t *zsb = ZTOZSB(zp);
2341 int err;
2342
2343 *working_mode = v4_mode;
2344 *check_privs = B_TRUE;
2345
2346 /*
2347 * Short circuit empty requests
2348 */
2349 if (v4_mode == 0 || zsb->z_replay) {
2350 *working_mode = 0;
2351 return (0);
2352 }
2353
2354 if ((err = zfs_zaccess_dataset_check(zp, v4_mode)) != 0) {
2355 *check_privs = B_FALSE;
2356 return (err);
2357 }
2358
2359 /*
2360 * The caller requested that the ACL check be skipped. This
2361 * would only happen if the caller checked VOP_ACCESS() with a
2362 * 32 bit ACE mask and already had the appropriate permissions.
2363 */
2364 if (skipaclchk) {
2365 *working_mode = 0;
2366 return (0);
2367 }
2368
2369 return (zfs_zaccess_aces_check(zp, working_mode, B_FALSE, cr));
2370 }
2371
2372 static int
2373 zfs_zaccess_append(znode_t *zp, uint32_t *working_mode, boolean_t *check_privs,
2374 cred_t *cr)
2375 {
2376 if (*working_mode != ACE_WRITE_DATA)
2377 return (SET_ERROR(EACCES));
2378
2379 return (zfs_zaccess_common(zp, ACE_APPEND_DATA, working_mode,
2380 check_privs, B_FALSE, cr));
2381 }
2382
2383 int
2384 zfs_fastaccesschk_execute(znode_t *zdp, cred_t *cr)
2385 {
2386 boolean_t owner = B_FALSE;
2387 boolean_t groupmbr = B_FALSE;
2388 boolean_t is_attr;
2389 uid_t uid = crgetuid(cr);
2390 int error;
2391
2392 if (zdp->z_pflags & ZFS_AV_QUARANTINED)
2393 return (SET_ERROR(EACCES));
2394
2395 is_attr = ((zdp->z_pflags & ZFS_XATTR) &&
2396 (S_ISDIR(ZTOI(zdp)->i_mode)));
2397 if (is_attr)
2398 goto slow;
2399
2400
2401 mutex_enter(&zdp->z_acl_lock);
2402
2403 if (zdp->z_pflags & ZFS_NO_EXECS_DENIED) {
2404 mutex_exit(&zdp->z_acl_lock);
2405 return (0);
2406 }
2407
2408 if (FUID_INDEX(zdp->z_uid) != 0 || FUID_INDEX(zdp->z_gid) != 0) {
2409 mutex_exit(&zdp->z_acl_lock);
2410 goto slow;
2411 }
2412
2413 if (uid == zdp->z_uid) {
2414 owner = B_TRUE;
2415 if (zdp->z_mode & S_IXUSR) {
2416 mutex_exit(&zdp->z_acl_lock);
2417 return (0);
2418 } else {
2419 mutex_exit(&zdp->z_acl_lock);
2420 goto slow;
2421 }
2422 }
2423 if (groupmember(zdp->z_gid, cr)) {
2424 groupmbr = B_TRUE;
2425 if (zdp->z_mode & S_IXGRP) {
2426 mutex_exit(&zdp->z_acl_lock);
2427 return (0);
2428 } else {
2429 mutex_exit(&zdp->z_acl_lock);
2430 goto slow;
2431 }
2432 }
2433 if (!owner && !groupmbr) {
2434 if (zdp->z_mode & S_IXOTH) {
2435 mutex_exit(&zdp->z_acl_lock);
2436 return (0);
2437 }
2438 }
2439
2440 mutex_exit(&zdp->z_acl_lock);
2441
2442 slow:
2443 DTRACE_PROBE(zfs__fastpath__execute__access__miss);
2444 ZFS_ENTER(ZTOZSB(zdp));
2445 error = zfs_zaccess(zdp, ACE_EXECUTE, 0, B_FALSE, cr);
2446 ZFS_EXIT(ZTOZSB(zdp));
2447 return (error);
2448 }
2449
2450 /*
2451 * Determine whether Access should be granted/denied.
2452 * The least priv subsytem is always consulted as a basic privilege
2453 * can define any form of access.
2454 */
2455 int
2456 zfs_zaccess(znode_t *zp, int mode, int flags, boolean_t skipaclchk, cred_t *cr)
2457 {
2458 uint32_t working_mode;
2459 int error;
2460 boolean_t check_privs;
2461 znode_t *check_zp = zp;
2462 mode_t needed_bits;
2463 uid_t owner;
2464
2465 /*
2466 * If attribute then validate against base file
2467 */
2468 if ((zp->z_pflags & ZFS_XATTR) && S_ISDIR(ZTOI(zp)->i_mode)) {
2469 uint64_t parent;
2470
2471 rw_enter(&zp->z_xattr_lock, RW_READER);
2472 if (zp->z_xattr_parent) {
2473 check_zp = zp->z_xattr_parent;
2474 rw_exit(&zp->z_xattr_lock);
2475
2476 /*
2477 * Verify a lookup yields the same znode.
2478 */
2479 ASSERT3S(sa_lookup(zp->z_sa_hdl, SA_ZPL_PARENT(
2480 ZTOZSB(zp)), &parent, sizeof (parent)), ==, 0);
2481 ASSERT3U(check_zp->z_id, ==, parent);
2482 } else {
2483 rw_exit(&zp->z_xattr_lock);
2484
2485 error = sa_lookup(zp->z_sa_hdl, SA_ZPL_PARENT(
2486 ZTOZSB(zp)), &parent, sizeof (parent));
2487 if (error)
2488 return (error);
2489
2490 /*
2491 * Cache the lookup on the parent file znode as
2492 * zp->z_xattr_parent and hold a reference. This
2493 * effectively pins the parent in memory until all
2494 * child xattr znodes have been destroyed and
2495 * release their references in zfs_inode_destroy().
2496 */
2497 error = zfs_zget(ZTOZSB(zp), parent, &check_zp);
2498 if (error)
2499 return (error);
2500
2501 rw_enter(&zp->z_xattr_lock, RW_WRITER);
2502 if (zp->z_xattr_parent == NULL)
2503 zp->z_xattr_parent = check_zp;
2504 rw_exit(&zp->z_xattr_lock);
2505 }
2506
2507 /*
2508 * fixup mode to map to xattr perms
2509 */
2510
2511 if (mode & (ACE_WRITE_DATA|ACE_APPEND_DATA)) {
2512 mode &= ~(ACE_WRITE_DATA|ACE_APPEND_DATA);
2513 mode |= ACE_WRITE_NAMED_ATTRS;
2514 }
2515
2516 if (mode & (ACE_READ_DATA|ACE_EXECUTE)) {
2517 mode &= ~(ACE_READ_DATA|ACE_EXECUTE);
2518 mode |= ACE_READ_NAMED_ATTRS;
2519 }
2520 }
2521
2522 owner = zfs_fuid_map_id(ZTOZSB(zp), zp->z_uid, cr, ZFS_OWNER);
2523 /*
2524 * Map the bits required to the standard inode flags
2525 * S_IRUSR|S_IWUSR|S_IXUSR in the needed_bits. Map the bits
2526 * mapped by working_mode (currently missing) in missing_bits.
2527 * Call secpolicy_vnode_access2() with (needed_bits & ~checkmode),
2528 * needed_bits.
2529 */
2530 needed_bits = 0;
2531
2532 working_mode = mode;
2533 if ((working_mode & (ACE_READ_ACL|ACE_READ_ATTRIBUTES)) &&
2534 owner == crgetuid(cr))
2535 working_mode &= ~(ACE_READ_ACL|ACE_READ_ATTRIBUTES);
2536
2537 if (working_mode & (ACE_READ_DATA|ACE_READ_NAMED_ATTRS|
2538 ACE_READ_ACL|ACE_READ_ATTRIBUTES|ACE_SYNCHRONIZE))
2539 needed_bits |= S_IRUSR;
2540 if (working_mode & (ACE_WRITE_DATA|ACE_WRITE_NAMED_ATTRS|
2541 ACE_APPEND_DATA|ACE_WRITE_ATTRIBUTES|ACE_SYNCHRONIZE))
2542 needed_bits |= S_IWUSR;
2543 if (working_mode & ACE_EXECUTE)
2544 needed_bits |= S_IXUSR;
2545
2546 if ((error = zfs_zaccess_common(check_zp, mode, &working_mode,
2547 &check_privs, skipaclchk, cr)) == 0) {
2548 return (secpolicy_vnode_access2(cr, ZTOI(zp), owner,
2549 needed_bits, needed_bits));
2550 }
2551
2552 if (error && !check_privs) {
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 return (error);
2614 }
2615
2616 /*
2617 * Translate traditional unix S_IRUSR/S_IWUSR/S_IXUSR mode into
2618 * native ACL format and call zfs_zaccess()
2619 */
2620 int
2621 zfs_zaccess_rwx(znode_t *zp, mode_t mode, int flags, cred_t *cr)
2622 {
2623 return (zfs_zaccess(zp, zfs_unix_to_v4(mode >> 6), flags, B_FALSE, cr));
2624 }
2625
2626 /*
2627 * Access function for secpolicy_vnode_setattr
2628 */
2629 int
2630 zfs_zaccess_unix(znode_t *zp, mode_t mode, cred_t *cr)
2631 {
2632 int v4_mode = zfs_unix_to_v4(mode >> 6);
2633
2634 return (zfs_zaccess(zp, v4_mode, 0, B_FALSE, cr));
2635 }
2636
2637 static int
2638 zfs_delete_final_check(znode_t *zp, znode_t *dzp,
2639 mode_t available_perms, cred_t *cr)
2640 {
2641 int error;
2642 uid_t downer;
2643
2644 downer = zfs_fuid_map_id(ZTOZSB(dzp), dzp->z_uid, cr, ZFS_OWNER);
2645
2646 error = secpolicy_vnode_access2(cr, ZTOI(dzp),
2647 downer, available_perms, S_IWUSR|S_IXUSR);
2648
2649 if (error == 0)
2650 error = zfs_sticky_remove_access(dzp, zp, cr);
2651
2652 return (error);
2653 }
2654
2655 /*
2656 * Determine whether Access should be granted/deny, without
2657 * consulting least priv subsystem.
2658 *
2659 *
2660 * The following chart is the recommended NFSv4 enforcement for
2661 * ability to delete an object.
2662 *
2663 * -------------------------------------------------------
2664 * | Parent Dir | Target Object Permissions |
2665 * | permissions | |
2666 * -------------------------------------------------------
2667 * | | ACL Allows | ACL Denies| Delete |
2668 * | | Delete | Delete | unspecified|
2669 * -------------------------------------------------------
2670 * | ACL Allows | Permit | Permit | Permit |
2671 * | DELETE_CHILD | |
2672 * -------------------------------------------------------
2673 * | ACL Denies | Permit | Deny | Deny |
2674 * | DELETE_CHILD | | | |
2675 * -------------------------------------------------------
2676 * | ACL specifies | | | |
2677 * | only allow | Permit | Permit | Permit |
2678 * | write and | | | |
2679 * | execute | | | |
2680 * -------------------------------------------------------
2681 * | ACL denies | | | |
2682 * | write and | Permit | Deny | Deny |
2683 * | execute | | | |
2684 * -------------------------------------------------------
2685 * ^
2686 * |
2687 * No search privilege, can't even look up file?
2688 *
2689 */
2690 int
2691 zfs_zaccess_delete(znode_t *dzp, znode_t *zp, cred_t *cr)
2692 {
2693 uint32_t dzp_working_mode = 0;
2694 uint32_t zp_working_mode = 0;
2695 int dzp_error, zp_error;
2696 mode_t available_perms;
2697 boolean_t dzpcheck_privs = B_TRUE;
2698 boolean_t zpcheck_privs = B_TRUE;
2699
2700 /*
2701 * We want specific DELETE permissions to
2702 * take precedence over WRITE/EXECUTE. We don't
2703 * want an ACL such as this to mess us up.
2704 * user:joe:write_data:deny,user:joe:delete:allow
2705 *
2706 * However, deny permissions may ultimately be overridden
2707 * by secpolicy_vnode_access().
2708 *
2709 * We will ask for all of the necessary permissions and then
2710 * look at the working modes from the directory and target object
2711 * to determine what was found.
2712 */
2713
2714 if (zp->z_pflags & (ZFS_IMMUTABLE | ZFS_NOUNLINK))
2715 return (SET_ERROR(EPERM));
2716
2717 /*
2718 * First row
2719 * If the directory permissions allow the delete, we are done.
2720 */
2721 if ((dzp_error = zfs_zaccess_common(dzp, ACE_DELETE_CHILD,
2722 &dzp_working_mode, &dzpcheck_privs, B_FALSE, cr)) == 0)
2723 return (0);
2724
2725 /*
2726 * If target object has delete permission then we are done
2727 */
2728 if ((zp_error = zfs_zaccess_common(zp, ACE_DELETE, &zp_working_mode,
2729 &zpcheck_privs, B_FALSE, cr)) == 0)
2730 return (0);
2731
2732 ASSERT(dzp_error && zp_error);
2733
2734 if (!dzpcheck_privs)
2735 return (dzp_error);
2736 if (!zpcheck_privs)
2737 return (zp_error);
2738
2739 /*
2740 * Second row
2741 *
2742 * If directory returns EACCES then delete_child was denied
2743 * due to deny delete_child. In this case send the request through
2744 * secpolicy_vnode_remove(). We don't use zfs_delete_final_check()
2745 * since that *could* allow the delete based on write/execute permission
2746 * and we want delete permissions to override write/execute.
2747 */
2748
2749 if (dzp_error == EACCES)
2750 return (secpolicy_vnode_remove(cr));
2751
2752 /*
2753 * Third Row
2754 * only need to see if we have write/execute on directory.
2755 */
2756
2757 dzp_error = zfs_zaccess_common(dzp, ACE_EXECUTE|ACE_WRITE_DATA,
2758 &dzp_working_mode, &dzpcheck_privs, B_FALSE, cr);
2759
2760 if (dzp_error != 0 && !dzpcheck_privs)
2761 return (dzp_error);
2762
2763 /*
2764 * Fourth row
2765 */
2766
2767 available_perms = (dzp_working_mode & ACE_WRITE_DATA) ? 0 : S_IWUSR;
2768 available_perms |= (dzp_working_mode & ACE_EXECUTE) ? 0 : S_IXUSR;
2769
2770 return (zfs_delete_final_check(zp, dzp, available_perms, cr));
2771
2772 }
2773
2774 int
2775 zfs_zaccess_rename(znode_t *sdzp, znode_t *szp, znode_t *tdzp,
2776 znode_t *tzp, cred_t *cr)
2777 {
2778 int add_perm;
2779 int error;
2780
2781 if (szp->z_pflags & ZFS_AV_QUARANTINED)
2782 return (SET_ERROR(EACCES));
2783
2784 add_perm = S_ISDIR(ZTOI(szp)->i_mode) ?
2785 ACE_ADD_SUBDIRECTORY : ACE_ADD_FILE;
2786
2787 /*
2788 * Rename permissions are combination of delete permission +
2789 * add file/subdir permission.
2790 */
2791
2792 /*
2793 * first make sure we do the delete portion.
2794 *
2795 * If that succeeds then check for add_file/add_subdir permissions
2796 */
2797
2798 if ((error = zfs_zaccess_delete(sdzp, szp, cr)))
2799 return (error);
2800
2801 /*
2802 * If we have a tzp, see if we can delete it?
2803 */
2804 if (tzp) {
2805 if ((error = zfs_zaccess_delete(tdzp, tzp, cr)))
2806 return (error);
2807 }
2808
2809 /*
2810 * Now check for add permissions
2811 */
2812 error = zfs_zaccess(tdzp, add_perm, 0, B_FALSE, cr);
2813
2814 return (error);
2815 }
ZFS On Linux mirror for PVE