]> 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
Fix NFS credential
[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 <sys/trace_acl.h>
56 #include "fs/fs_subr.h"
57
58 #define ALLOW ACE_ACCESS_ALLOWED_ACE_TYPE
59 #define DENY ACE_ACCESS_DENIED_ACE_TYPE
60 #define MAX_ACE_TYPE ACE_SYSTEM_ALARM_CALLBACK_OBJECT_ACE_TYPE
61 #define MIN_ACE_TYPE ALLOW
62
63 #define OWNING_GROUP (ACE_GROUP|ACE_IDENTIFIER_GROUP)
64 #define EVERYONE_ALLOW_MASK (ACE_READ_ACL|ACE_READ_ATTRIBUTES | \
65 ACE_READ_NAMED_ATTRS|ACE_SYNCHRONIZE)
66 #define EVERYONE_DENY_MASK (ACE_WRITE_ACL|ACE_WRITE_OWNER | \
67 ACE_WRITE_ATTRIBUTES|ACE_WRITE_NAMED_ATTRS)
68 #define OWNER_ALLOW_MASK (ACE_WRITE_ACL | ACE_WRITE_OWNER | \
69 ACE_WRITE_ATTRIBUTES|ACE_WRITE_NAMED_ATTRS)
70
71 #define ZFS_CHECKED_MASKS (ACE_READ_ACL|ACE_READ_ATTRIBUTES|ACE_READ_DATA| \
72 ACE_READ_NAMED_ATTRS|ACE_WRITE_DATA|ACE_WRITE_ATTRIBUTES| \
73 ACE_WRITE_NAMED_ATTRS|ACE_APPEND_DATA|ACE_EXECUTE|ACE_WRITE_OWNER| \
74 ACE_WRITE_ACL|ACE_DELETE|ACE_DELETE_CHILD|ACE_SYNCHRONIZE)
75
76 #define WRITE_MASK_DATA (ACE_WRITE_DATA|ACE_APPEND_DATA|ACE_WRITE_NAMED_ATTRS)
77 #define WRITE_MASK_ATTRS (ACE_WRITE_ACL|ACE_WRITE_OWNER|ACE_WRITE_ATTRIBUTES| \
78 ACE_DELETE|ACE_DELETE_CHILD)
79 #define WRITE_MASK (WRITE_MASK_DATA|WRITE_MASK_ATTRS)
80
81 #define OGE_CLEAR (ACE_READ_DATA|ACE_LIST_DIRECTORY|ACE_WRITE_DATA| \
82 ACE_ADD_FILE|ACE_APPEND_DATA|ACE_ADD_SUBDIRECTORY|ACE_EXECUTE)
83
84 #define OKAY_MASK_BITS (ACE_READ_DATA|ACE_LIST_DIRECTORY|ACE_WRITE_DATA| \
85 ACE_ADD_FILE|ACE_APPEND_DATA|ACE_ADD_SUBDIRECTORY|ACE_EXECUTE)
86
87 #define ALL_INHERIT (ACE_FILE_INHERIT_ACE|ACE_DIRECTORY_INHERIT_ACE | \
88 ACE_NO_PROPAGATE_INHERIT_ACE|ACE_INHERIT_ONLY_ACE|ACE_INHERITED_ACE)
89
90 #define RESTRICTED_CLEAR (ACE_WRITE_ACL|ACE_WRITE_OWNER)
91
92 #define V4_ACL_WIDE_FLAGS (ZFS_ACL_AUTO_INHERIT|ZFS_ACL_DEFAULTED|\
93 ZFS_ACL_PROTECTED)
94
95 #define ZFS_ACL_WIDE_FLAGS (V4_ACL_WIDE_FLAGS|ZFS_ACL_TRIVIAL|ZFS_INHERIT_ACE|\
96 ZFS_ACL_OBJ_ACE)
97
98 #define ALL_MODE_EXECS (S_IXUSR | S_IXGRP | S_IXOTH)
99
100 static uint16_t
101 zfs_ace_v0_get_type(void *acep)
102 {
103 return (((zfs_oldace_t *)acep)->z_type);
104 }
105
106 static uint16_t
107 zfs_ace_v0_get_flags(void *acep)
108 {
109 return (((zfs_oldace_t *)acep)->z_flags);
110 }
111
112 static uint32_t
113 zfs_ace_v0_get_mask(void *acep)
114 {
115 return (((zfs_oldace_t *)acep)->z_access_mask);
116 }
117
118 static uint64_t
119 zfs_ace_v0_get_who(void *acep)
120 {
121 return (((zfs_oldace_t *)acep)->z_fuid);
122 }
123
124 static void
125 zfs_ace_v0_set_type(void *acep, uint16_t type)
126 {
127 ((zfs_oldace_t *)acep)->z_type = type;
128 }
129
130 static void
131 zfs_ace_v0_set_flags(void *acep, uint16_t flags)
132 {
133 ((zfs_oldace_t *)acep)->z_flags = flags;
134 }
135
136 static void
137 zfs_ace_v0_set_mask(void *acep, uint32_t mask)
138 {
139 ((zfs_oldace_t *)acep)->z_access_mask = mask;
140 }
141
142 static void
143 zfs_ace_v0_set_who(void *acep, uint64_t who)
144 {
145 ((zfs_oldace_t *)acep)->z_fuid = who;
146 }
147
148 /*ARGSUSED*/
149 static size_t
150 zfs_ace_v0_size(void *acep)
151 {
152 return (sizeof (zfs_oldace_t));
153 }
154
155 static size_t
156 zfs_ace_v0_abstract_size(void)
157 {
158 return (sizeof (zfs_oldace_t));
159 }
160
161 static int
162 zfs_ace_v0_mask_off(void)
163 {
164 return (offsetof(zfs_oldace_t, z_access_mask));
165 }
166
167 /*ARGSUSED*/
168 static int
169 zfs_ace_v0_data(void *acep, void **datap)
170 {
171 *datap = NULL;
172 return (0);
173 }
174
175 static acl_ops_t zfs_acl_v0_ops = {
176 zfs_ace_v0_get_mask,
177 zfs_ace_v0_set_mask,
178 zfs_ace_v0_get_flags,
179 zfs_ace_v0_set_flags,
180 zfs_ace_v0_get_type,
181 zfs_ace_v0_set_type,
182 zfs_ace_v0_get_who,
183 zfs_ace_v0_set_who,
184 zfs_ace_v0_size,
185 zfs_ace_v0_abstract_size,
186 zfs_ace_v0_mask_off,
187 zfs_ace_v0_data
188 };
189
190 static uint16_t
191 zfs_ace_fuid_get_type(void *acep)
192 {
193 return (((zfs_ace_hdr_t *)acep)->z_type);
194 }
195
196 static uint16_t
197 zfs_ace_fuid_get_flags(void *acep)
198 {
199 return (((zfs_ace_hdr_t *)acep)->z_flags);
200 }
201
202 static uint32_t
203 zfs_ace_fuid_get_mask(void *acep)
204 {
205 return (((zfs_ace_hdr_t *)acep)->z_access_mask);
206 }
207
208 static uint64_t
209 zfs_ace_fuid_get_who(void *args)
210 {
211 uint16_t entry_type;
212 zfs_ace_t *acep = args;
213
214 entry_type = acep->z_hdr.z_flags & ACE_TYPE_FLAGS;
215
216 if (entry_type == ACE_OWNER || entry_type == OWNING_GROUP ||
217 entry_type == ACE_EVERYONE)
218 return (-1);
219 return (((zfs_ace_t *)acep)->z_fuid);
220 }
221
222 static void
223 zfs_ace_fuid_set_type(void *acep, uint16_t type)
224 {
225 ((zfs_ace_hdr_t *)acep)->z_type = type;
226 }
227
228 static void
229 zfs_ace_fuid_set_flags(void *acep, uint16_t flags)
230 {
231 ((zfs_ace_hdr_t *)acep)->z_flags = flags;
232 }
233
234 static void
235 zfs_ace_fuid_set_mask(void *acep, uint32_t mask)
236 {
237 ((zfs_ace_hdr_t *)acep)->z_access_mask = mask;
238 }
239
240 static void
241 zfs_ace_fuid_set_who(void *arg, uint64_t who)
242 {
243 zfs_ace_t *acep = arg;
244
245 uint16_t entry_type = acep->z_hdr.z_flags & ACE_TYPE_FLAGS;
246
247 if (entry_type == ACE_OWNER || entry_type == OWNING_GROUP ||
248 entry_type == ACE_EVERYONE)
249 return;
250 acep->z_fuid = who;
251 }
252
253 static size_t
254 zfs_ace_fuid_size(void *acep)
255 {
256 zfs_ace_hdr_t *zacep = acep;
257 uint16_t entry_type;
258
259 switch (zacep->z_type) {
260 case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE:
261 case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE:
262 case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE:
263 case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE:
264 return (sizeof (zfs_object_ace_t));
265 case ALLOW:
266 case DENY:
267 entry_type =
268 (((zfs_ace_hdr_t *)acep)->z_flags & ACE_TYPE_FLAGS);
269 if (entry_type == ACE_OWNER ||
270 entry_type == OWNING_GROUP ||
271 entry_type == ACE_EVERYONE)
272 return (sizeof (zfs_ace_hdr_t));
273 /*FALLTHROUGH*/
274 default:
275 return (sizeof (zfs_ace_t));
276 }
277 }
278
279 static size_t
280 zfs_ace_fuid_abstract_size(void)
281 {
282 return (sizeof (zfs_ace_hdr_t));
283 }
284
285 static int
286 zfs_ace_fuid_mask_off(void)
287 {
288 return (offsetof(zfs_ace_hdr_t, z_access_mask));
289 }
290
291 static int
292 zfs_ace_fuid_data(void *acep, void **datap)
293 {
294 zfs_ace_t *zacep = acep;
295 zfs_object_ace_t *zobjp;
296
297 switch (zacep->z_hdr.z_type) {
298 case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE:
299 case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE:
300 case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE:
301 case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE:
302 zobjp = acep;
303 *datap = (caddr_t)zobjp + sizeof (zfs_ace_t);
304 return (sizeof (zfs_object_ace_t) - sizeof (zfs_ace_t));
305 default:
306 *datap = NULL;
307 return (0);
308 }
309 }
310
311 static acl_ops_t zfs_acl_fuid_ops = {
312 zfs_ace_fuid_get_mask,
313 zfs_ace_fuid_set_mask,
314 zfs_ace_fuid_get_flags,
315 zfs_ace_fuid_set_flags,
316 zfs_ace_fuid_get_type,
317 zfs_ace_fuid_set_type,
318 zfs_ace_fuid_get_who,
319 zfs_ace_fuid_set_who,
320 zfs_ace_fuid_size,
321 zfs_ace_fuid_abstract_size,
322 zfs_ace_fuid_mask_off,
323 zfs_ace_fuid_data
324 };
325
326 /*
327 * The following three functions are provided for compatibility with
328 * older ZPL version in order to determine if the file use to have
329 * an external ACL and what version of ACL previously existed on the
330 * file. Would really be nice to not need this, sigh.
331 */
332 uint64_t
333 zfs_external_acl(znode_t *zp)
334 {
335 zfs_acl_phys_t acl_phys;
336 int error;
337
338 if (zp->z_is_sa)
339 return (0);
340
341 /*
342 * Need to deal with a potential
343 * race where zfs_sa_upgrade could cause
344 * z_isa_sa to change.
345 *
346 * If the lookup fails then the state of z_is_sa should have
347 * changed.
348 */
349
350 if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_ZNODE_ACL(ZTOZSB(zp)),
351 &acl_phys, sizeof (acl_phys))) == 0)
352 return (acl_phys.z_acl_extern_obj);
353 else {
354 /*
355 * after upgrade the SA_ZPL_ZNODE_ACL should have been
356 * removed
357 */
358 VERIFY(zp->z_is_sa && error == ENOENT);
359 return (0);
360 }
361 }
362
363 /*
364 * Determine size of ACL in bytes
365 *
366 * This is more complicated than it should be since we have to deal
367 * with old external ACLs.
368 */
369 static int
370 zfs_acl_znode_info(znode_t *zp, int *aclsize, int *aclcount,
371 zfs_acl_phys_t *aclphys)
372 {
373 zfs_sb_t *zsb = ZTOZSB(zp);
374 uint64_t acl_count;
375 int size;
376 int error;
377
378 ASSERT(MUTEX_HELD(&zp->z_acl_lock));
379 if (zp->z_is_sa) {
380 if ((error = sa_size(zp->z_sa_hdl, SA_ZPL_DACL_ACES(zsb),
381 &size)) != 0)
382 return (error);
383 *aclsize = size;
384 if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_DACL_COUNT(zsb),
385 &acl_count, sizeof (acl_count))) != 0)
386 return (error);
387 *aclcount = acl_count;
388 } else {
389 if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_ZNODE_ACL(zsb),
390 aclphys, sizeof (*aclphys))) != 0)
391 return (error);
392
393 if (aclphys->z_acl_version == ZFS_ACL_VERSION_INITIAL) {
394 *aclsize = ZFS_ACL_SIZE(aclphys->z_acl_size);
395 *aclcount = aclphys->z_acl_size;
396 } else {
397 *aclsize = aclphys->z_acl_size;
398 *aclcount = aclphys->z_acl_count;
399 }
400 }
401 return (0);
402 }
403
404 int
405 zfs_znode_acl_version(znode_t *zp)
406 {
407 zfs_acl_phys_t acl_phys;
408
409 if (zp->z_is_sa)
410 return (ZFS_ACL_VERSION_FUID);
411 else {
412 int error;
413
414 /*
415 * Need to deal with a potential
416 * race where zfs_sa_upgrade could cause
417 * z_isa_sa to change.
418 *
419 * If the lookup fails then the state of z_is_sa should have
420 * changed.
421 */
422 if ((error = sa_lookup(zp->z_sa_hdl,
423 SA_ZPL_ZNODE_ACL(ZTOZSB(zp)),
424 &acl_phys, sizeof (acl_phys))) == 0)
425 return (acl_phys.z_acl_version);
426 else {
427 /*
428 * After upgrade SA_ZPL_ZNODE_ACL should have
429 * been removed.
430 */
431 VERIFY(zp->z_is_sa && error == ENOENT);
432 return (ZFS_ACL_VERSION_FUID);
433 }
434 }
435 }
436
437 static int
438 zfs_acl_version(int version)
439 {
440 if (version < ZPL_VERSION_FUID)
441 return (ZFS_ACL_VERSION_INITIAL);
442 else
443 return (ZFS_ACL_VERSION_FUID);
444 }
445
446 static int
447 zfs_acl_version_zp(znode_t *zp)
448 {
449 return (zfs_acl_version(ZTOZSB(zp)->z_version));
450 }
451
452 zfs_acl_t *
453 zfs_acl_alloc(int vers)
454 {
455 zfs_acl_t *aclp;
456
457 aclp = kmem_zalloc(sizeof (zfs_acl_t), KM_SLEEP);
458 list_create(&aclp->z_acl, sizeof (zfs_acl_node_t),
459 offsetof(zfs_acl_node_t, z_next));
460 aclp->z_version = vers;
461 if (vers == ZFS_ACL_VERSION_FUID)
462 aclp->z_ops = &zfs_acl_fuid_ops;
463 else
464 aclp->z_ops = &zfs_acl_v0_ops;
465 return (aclp);
466 }
467
468 zfs_acl_node_t *
469 zfs_acl_node_alloc(size_t bytes)
470 {
471 zfs_acl_node_t *aclnode;
472
473 aclnode = kmem_zalloc(sizeof (zfs_acl_node_t), KM_SLEEP);
474 if (bytes) {
475 aclnode->z_acldata = kmem_alloc(bytes, KM_SLEEP);
476 aclnode->z_allocdata = aclnode->z_acldata;
477 aclnode->z_allocsize = bytes;
478 aclnode->z_size = bytes;
479 }
480
481 return (aclnode);
482 }
483
484 static void
485 zfs_acl_node_free(zfs_acl_node_t *aclnode)
486 {
487 if (aclnode->z_allocsize)
488 kmem_free(aclnode->z_allocdata, aclnode->z_allocsize);
489 kmem_free(aclnode, sizeof (zfs_acl_node_t));
490 }
491
492 static void
493 zfs_acl_release_nodes(zfs_acl_t *aclp)
494 {
495 zfs_acl_node_t *aclnode;
496
497 while ((aclnode = list_head(&aclp->z_acl))) {
498 list_remove(&aclp->z_acl, aclnode);
499 zfs_acl_node_free(aclnode);
500 }
501 aclp->z_acl_count = 0;
502 aclp->z_acl_bytes = 0;
503 }
504
505 void
506 zfs_acl_free(zfs_acl_t *aclp)
507 {
508 zfs_acl_release_nodes(aclp);
509 list_destroy(&aclp->z_acl);
510 kmem_free(aclp, sizeof (zfs_acl_t));
511 }
512
513 static boolean_t
514 zfs_acl_valid_ace_type(uint_t type, uint_t flags)
515 {
516 uint16_t entry_type;
517
518 switch (type) {
519 case ALLOW:
520 case DENY:
521 case ACE_SYSTEM_AUDIT_ACE_TYPE:
522 case ACE_SYSTEM_ALARM_ACE_TYPE:
523 entry_type = flags & ACE_TYPE_FLAGS;
524 return (entry_type == ACE_OWNER ||
525 entry_type == OWNING_GROUP ||
526 entry_type == ACE_EVERYONE || entry_type == 0 ||
527 entry_type == ACE_IDENTIFIER_GROUP);
528 default:
529 if (type >= MIN_ACE_TYPE && type <= MAX_ACE_TYPE)
530 return (B_TRUE);
531 }
532 return (B_FALSE);
533 }
534
535 static boolean_t
536 zfs_ace_valid(umode_t obj_mode, zfs_acl_t *aclp, uint16_t type, uint16_t iflags)
537 {
538 /*
539 * first check type of entry
540 */
541
542 if (!zfs_acl_valid_ace_type(type, iflags))
543 return (B_FALSE);
544
545 switch (type) {
546 case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE:
547 case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE:
548 case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE:
549 case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE:
550 if (aclp->z_version < ZFS_ACL_VERSION_FUID)
551 return (B_FALSE);
552 aclp->z_hints |= ZFS_ACL_OBJ_ACE;
553 }
554
555 /*
556 * next check inheritance level flags
557 */
558
559 if (S_ISDIR(obj_mode) &&
560 (iflags & (ACE_FILE_INHERIT_ACE|ACE_DIRECTORY_INHERIT_ACE)))
561 aclp->z_hints |= ZFS_INHERIT_ACE;
562
563 if (iflags & (ACE_INHERIT_ONLY_ACE|ACE_NO_PROPAGATE_INHERIT_ACE)) {
564 if ((iflags & (ACE_FILE_INHERIT_ACE|
565 ACE_DIRECTORY_INHERIT_ACE)) == 0) {
566 return (B_FALSE);
567 }
568 }
569
570 return (B_TRUE);
571 }
572
573 static void *
574 zfs_acl_next_ace(zfs_acl_t *aclp, void *start, uint64_t *who,
575 uint32_t *access_mask, uint16_t *iflags, uint16_t *type)
576 {
577 zfs_acl_node_t *aclnode;
578
579 ASSERT(aclp);
580
581 if (start == NULL) {
582 aclnode = list_head(&aclp->z_acl);
583 if (aclnode == NULL)
584 return (NULL);
585
586 aclp->z_next_ace = aclnode->z_acldata;
587 aclp->z_curr_node = aclnode;
588 aclnode->z_ace_idx = 0;
589 }
590
591 aclnode = aclp->z_curr_node;
592
593 if (aclnode == NULL)
594 return (NULL);
595
596 if (aclnode->z_ace_idx >= aclnode->z_ace_count) {
597 aclnode = list_next(&aclp->z_acl, aclnode);
598 if (aclnode == NULL)
599 return (NULL);
600 else {
601 aclp->z_curr_node = aclnode;
602 aclnode->z_ace_idx = 0;
603 aclp->z_next_ace = aclnode->z_acldata;
604 }
605 }
606
607 if (aclnode->z_ace_idx < aclnode->z_ace_count) {
608 void *acep = aclp->z_next_ace;
609 size_t ace_size;
610
611 /*
612 * Make sure we don't overstep our bounds
613 */
614 ace_size = aclp->z_ops->ace_size(acep);
615
616 if (((caddr_t)acep + ace_size) >
617 ((caddr_t)aclnode->z_acldata + aclnode->z_size)) {
618 return (NULL);
619 }
620
621 *iflags = aclp->z_ops->ace_flags_get(acep);
622 *type = aclp->z_ops->ace_type_get(acep);
623 *access_mask = aclp->z_ops->ace_mask_get(acep);
624 *who = aclp->z_ops->ace_who_get(acep);
625 aclp->z_next_ace = (caddr_t)aclp->z_next_ace + ace_size;
626 aclnode->z_ace_idx++;
627
628 return ((void *)acep);
629 }
630 return (NULL);
631 }
632
633 /*ARGSUSED*/
634 static uint64_t
635 zfs_ace_walk(void *datap, uint64_t cookie, int aclcnt,
636 uint16_t *flags, uint16_t *type, uint32_t *mask)
637 {
638 zfs_acl_t *aclp = datap;
639 zfs_ace_hdr_t *acep = (zfs_ace_hdr_t *)(uintptr_t)cookie;
640 uint64_t who;
641
642 acep = zfs_acl_next_ace(aclp, acep, &who, mask,
643 flags, type);
644 return ((uint64_t)(uintptr_t)acep);
645 }
646
647 /*
648 * Copy ACE to internal ZFS format.
649 * While processing the ACL each ACE will be validated for correctness.
650 * ACE FUIDs will be created later.
651 */
652 int
653 zfs_copy_ace_2_fuid(zfs_sb_t *zsb, umode_t obj_mode, zfs_acl_t *aclp,
654 void *datap, zfs_ace_t *z_acl, uint64_t aclcnt, size_t *size,
655 zfs_fuid_info_t **fuidp, cred_t *cr)
656 {
657 int i;
658 uint16_t entry_type;
659 zfs_ace_t *aceptr = z_acl;
660 ace_t *acep = datap;
661 zfs_object_ace_t *zobjacep;
662 ace_object_t *aceobjp;
663
664 for (i = 0; i != aclcnt; i++) {
665 aceptr->z_hdr.z_access_mask = acep->a_access_mask;
666 aceptr->z_hdr.z_flags = acep->a_flags;
667 aceptr->z_hdr.z_type = acep->a_type;
668 entry_type = aceptr->z_hdr.z_flags & ACE_TYPE_FLAGS;
669 if (entry_type != ACE_OWNER && entry_type != OWNING_GROUP &&
670 entry_type != ACE_EVERYONE) {
671 aceptr->z_fuid = zfs_fuid_create(zsb, acep->a_who,
672 cr, (entry_type == 0) ?
673 ZFS_ACE_USER : ZFS_ACE_GROUP, fuidp);
674 }
675
676 /*
677 * Make sure ACE is valid
678 */
679 if (zfs_ace_valid(obj_mode, aclp, aceptr->z_hdr.z_type,
680 aceptr->z_hdr.z_flags) != B_TRUE)
681 return (SET_ERROR(EINVAL));
682
683 switch (acep->a_type) {
684 case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE:
685 case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE:
686 case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE:
687 case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE:
688 zobjacep = (zfs_object_ace_t *)aceptr;
689 aceobjp = (ace_object_t *)acep;
690
691 bcopy(aceobjp->a_obj_type, zobjacep->z_object_type,
692 sizeof (aceobjp->a_obj_type));
693 bcopy(aceobjp->a_inherit_obj_type,
694 zobjacep->z_inherit_type,
695 sizeof (aceobjp->a_inherit_obj_type));
696 acep = (ace_t *)((caddr_t)acep + sizeof (ace_object_t));
697 break;
698 default:
699 acep = (ace_t *)((caddr_t)acep + sizeof (ace_t));
700 }
701
702 aceptr = (zfs_ace_t *)((caddr_t)aceptr +
703 aclp->z_ops->ace_size(aceptr));
704 }
705
706 *size = (caddr_t)aceptr - (caddr_t)z_acl;
707
708 return (0);
709 }
710
711 /*
712 * Copy ZFS ACEs to fixed size ace_t layout
713 */
714 static void
715 zfs_copy_fuid_2_ace(zfs_sb_t *zsb, zfs_acl_t *aclp, cred_t *cr,
716 void *datap, int filter)
717 {
718 uint64_t who;
719 uint32_t access_mask;
720 uint16_t iflags, type;
721 zfs_ace_hdr_t *zacep = NULL;
722 ace_t *acep = datap;
723 ace_object_t *objacep;
724 zfs_object_ace_t *zobjacep;
725 size_t ace_size;
726 uint16_t entry_type;
727
728 while ((zacep = zfs_acl_next_ace(aclp, zacep,
729 &who, &access_mask, &iflags, &type))) {
730
731 switch (type) {
732 case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE:
733 case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE:
734 case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE:
735 case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE:
736 if (filter) {
737 continue;
738 }
739 zobjacep = (zfs_object_ace_t *)zacep;
740 objacep = (ace_object_t *)acep;
741 bcopy(zobjacep->z_object_type,
742 objacep->a_obj_type,
743 sizeof (zobjacep->z_object_type));
744 bcopy(zobjacep->z_inherit_type,
745 objacep->a_inherit_obj_type,
746 sizeof (zobjacep->z_inherit_type));
747 ace_size = sizeof (ace_object_t);
748 break;
749 default:
750 ace_size = sizeof (ace_t);
751 break;
752 }
753
754 entry_type = (iflags & ACE_TYPE_FLAGS);
755 if ((entry_type != ACE_OWNER &&
756 entry_type != OWNING_GROUP &&
757 entry_type != ACE_EVERYONE)) {
758 acep->a_who = zfs_fuid_map_id(zsb, who,
759 cr, (entry_type & ACE_IDENTIFIER_GROUP) ?
760 ZFS_ACE_GROUP : ZFS_ACE_USER);
761 } else {
762 acep->a_who = (uid_t)(int64_t)who;
763 }
764 acep->a_access_mask = access_mask;
765 acep->a_flags = iflags;
766 acep->a_type = type;
767 acep = (ace_t *)((caddr_t)acep + ace_size);
768 }
769 }
770
771 static int
772 zfs_copy_ace_2_oldace(umode_t obj_mode, zfs_acl_t *aclp, ace_t *acep,
773 zfs_oldace_t *z_acl, int aclcnt, size_t *size)
774 {
775 int i;
776 zfs_oldace_t *aceptr = z_acl;
777
778 for (i = 0; i != aclcnt; i++, aceptr++) {
779 aceptr->z_access_mask = acep[i].a_access_mask;
780 aceptr->z_type = acep[i].a_type;
781 aceptr->z_flags = acep[i].a_flags;
782 aceptr->z_fuid = acep[i].a_who;
783 /*
784 * Make sure ACE is valid
785 */
786 if (zfs_ace_valid(obj_mode, aclp, aceptr->z_type,
787 aceptr->z_flags) != B_TRUE)
788 return (SET_ERROR(EINVAL));
789 }
790 *size = (caddr_t)aceptr - (caddr_t)z_acl;
791 return (0);
792 }
793
794 /*
795 * convert old ACL format to new
796 */
797 void
798 zfs_acl_xform(znode_t *zp, zfs_acl_t *aclp, cred_t *cr)
799 {
800 zfs_oldace_t *oldaclp;
801 int i;
802 uint16_t type, iflags;
803 uint32_t access_mask;
804 uint64_t who;
805 void *cookie = NULL;
806 zfs_acl_node_t *newaclnode;
807
808 ASSERT(aclp->z_version == ZFS_ACL_VERSION_INITIAL);
809 /*
810 * First create the ACE in a contiguous piece of memory
811 * for zfs_copy_ace_2_fuid().
812 *
813 * We only convert an ACL once, so this won't happen
814 * everytime.
815 */
816 oldaclp = kmem_alloc(sizeof (zfs_oldace_t) * aclp->z_acl_count,
817 KM_SLEEP);
818 i = 0;
819 while ((cookie = zfs_acl_next_ace(aclp, cookie, &who,
820 &access_mask, &iflags, &type))) {
821 oldaclp[i].z_flags = iflags;
822 oldaclp[i].z_type = type;
823 oldaclp[i].z_fuid = who;
824 oldaclp[i++].z_access_mask = access_mask;
825 }
826
827 newaclnode = zfs_acl_node_alloc(aclp->z_acl_count *
828 sizeof (zfs_object_ace_t));
829 aclp->z_ops = &zfs_acl_fuid_ops;
830 VERIFY(zfs_copy_ace_2_fuid(ZTOZSB(zp), ZTOI(zp)->i_mode,
831 aclp, oldaclp, newaclnode->z_acldata, aclp->z_acl_count,
832 &newaclnode->z_size, NULL, cr) == 0);
833 newaclnode->z_ace_count = aclp->z_acl_count;
834 aclp->z_version = ZFS_ACL_VERSION;
835 kmem_free(oldaclp, aclp->z_acl_count * sizeof (zfs_oldace_t));
836
837 /*
838 * Release all previous ACL nodes
839 */
840
841 zfs_acl_release_nodes(aclp);
842
843 list_insert_head(&aclp->z_acl, newaclnode);
844
845 aclp->z_acl_bytes = newaclnode->z_size;
846 aclp->z_acl_count = newaclnode->z_ace_count;
847
848 }
849
850 /*
851 * Convert unix access mask to v4 access mask
852 */
853 static uint32_t
854 zfs_unix_to_v4(uint32_t access_mask)
855 {
856 uint32_t new_mask = 0;
857
858 if (access_mask & S_IXOTH)
859 new_mask |= ACE_EXECUTE;
860 if (access_mask & S_IWOTH)
861 new_mask |= ACE_WRITE_DATA;
862 if (access_mask & S_IROTH)
863 new_mask |= ACE_READ_DATA;
864 return (new_mask);
865 }
866
867 static void
868 zfs_set_ace(zfs_acl_t *aclp, void *acep, uint32_t access_mask,
869 uint16_t access_type, uint64_t fuid, uint16_t entry_type)
870 {
871 uint16_t type = entry_type & ACE_TYPE_FLAGS;
872
873 aclp->z_ops->ace_mask_set(acep, access_mask);
874 aclp->z_ops->ace_type_set(acep, access_type);
875 aclp->z_ops->ace_flags_set(acep, entry_type);
876 if ((type != ACE_OWNER && type != OWNING_GROUP &&
877 type != ACE_EVERYONE))
878 aclp->z_ops->ace_who_set(acep, fuid);
879 }
880
881 /*
882 * Determine mode of file based on ACL.
883 * Also, create FUIDs for any User/Group ACEs
884 */
885 uint64_t
886 zfs_mode_compute(uint64_t fmode, zfs_acl_t *aclp,
887 uint64_t *pflags, uint64_t fuid, uint64_t fgid)
888 {
889 int entry_type;
890 mode_t mode;
891 mode_t seen = 0;
892 zfs_ace_hdr_t *acep = NULL;
893 uint64_t who;
894 uint16_t iflags, type;
895 uint32_t access_mask;
896 boolean_t an_exec_denied = B_FALSE;
897
898 mode = (fmode & (S_IFMT | S_ISUID | S_ISGID | S_ISVTX));
899
900 while ((acep = zfs_acl_next_ace(aclp, acep, &who,
901 &access_mask, &iflags, &type))) {
902
903 if (!zfs_acl_valid_ace_type(type, iflags))
904 continue;
905
906 entry_type = (iflags & ACE_TYPE_FLAGS);
907
908 /*
909 * Skip over owner@, group@ or everyone@ inherit only ACEs
910 */
911 if ((iflags & ACE_INHERIT_ONLY_ACE) &&
912 (entry_type == ACE_OWNER || entry_type == ACE_EVERYONE ||
913 entry_type == OWNING_GROUP))
914 continue;
915
916 if (entry_type == ACE_OWNER || (entry_type == 0 &&
917 who == fuid)) {
918 if ((access_mask & ACE_READ_DATA) &&
919 (!(seen & S_IRUSR))) {
920 seen |= S_IRUSR;
921 if (type == ALLOW) {
922 mode |= S_IRUSR;
923 }
924 }
925 if ((access_mask & ACE_WRITE_DATA) &&
926 (!(seen & S_IWUSR))) {
927 seen |= S_IWUSR;
928 if (type == ALLOW) {
929 mode |= S_IWUSR;
930 }
931 }
932 if ((access_mask & ACE_EXECUTE) &&
933 (!(seen & S_IXUSR))) {
934 seen |= S_IXUSR;
935 if (type == ALLOW) {
936 mode |= S_IXUSR;
937 }
938 }
939 } else if (entry_type == OWNING_GROUP ||
940 (entry_type == ACE_IDENTIFIER_GROUP && who == fgid)) {
941 if ((access_mask & ACE_READ_DATA) &&
942 (!(seen & S_IRGRP))) {
943 seen |= S_IRGRP;
944 if (type == ALLOW) {
945 mode |= S_IRGRP;
946 }
947 }
948 if ((access_mask & ACE_WRITE_DATA) &&
949 (!(seen & S_IWGRP))) {
950 seen |= S_IWGRP;
951 if (type == ALLOW) {
952 mode |= S_IWGRP;
953 }
954 }
955 if ((access_mask & ACE_EXECUTE) &&
956 (!(seen & S_IXGRP))) {
957 seen |= S_IXGRP;
958 if (type == ALLOW) {
959 mode |= S_IXGRP;
960 }
961 }
962 } else if (entry_type == ACE_EVERYONE) {
963 if ((access_mask & ACE_READ_DATA)) {
964 if (!(seen & S_IRUSR)) {
965 seen |= S_IRUSR;
966 if (type == ALLOW) {
967 mode |= S_IRUSR;
968 }
969 }
970 if (!(seen & S_IRGRP)) {
971 seen |= S_IRGRP;
972 if (type == ALLOW) {
973 mode |= S_IRGRP;
974 }
975 }
976 if (!(seen & S_IROTH)) {
977 seen |= S_IROTH;
978 if (type == ALLOW) {
979 mode |= S_IROTH;
980 }
981 }
982 }
983 if ((access_mask & ACE_WRITE_DATA)) {
984 if (!(seen & S_IWUSR)) {
985 seen |= S_IWUSR;
986 if (type == ALLOW) {
987 mode |= S_IWUSR;
988 }
989 }
990 if (!(seen & S_IWGRP)) {
991 seen |= S_IWGRP;
992 if (type == ALLOW) {
993 mode |= S_IWGRP;
994 }
995 }
996 if (!(seen & S_IWOTH)) {
997 seen |= S_IWOTH;
998 if (type == ALLOW) {
999 mode |= S_IWOTH;
1000 }
1001 }
1002 }
1003 if ((access_mask & ACE_EXECUTE)) {
1004 if (!(seen & S_IXUSR)) {
1005 seen |= S_IXUSR;
1006 if (type == ALLOW) {
1007 mode |= S_IXUSR;
1008 }
1009 }
1010 if (!(seen & S_IXGRP)) {
1011 seen |= S_IXGRP;
1012 if (type == ALLOW) {
1013 mode |= S_IXGRP;
1014 }
1015 }
1016 if (!(seen & S_IXOTH)) {
1017 seen |= S_IXOTH;
1018 if (type == ALLOW) {
1019 mode |= S_IXOTH;
1020 }
1021 }
1022 }
1023 } else {
1024 /*
1025 * Only care if this IDENTIFIER_GROUP or
1026 * USER ACE denies execute access to someone,
1027 * mode is not affected
1028 */
1029 if ((access_mask & ACE_EXECUTE) && type == DENY)
1030 an_exec_denied = B_TRUE;
1031 }
1032 }
1033
1034 /*
1035 * Failure to allow is effectively a deny, so execute permission
1036 * is denied if it was never mentioned or if we explicitly
1037 * weren't allowed it.
1038 */
1039 if (!an_exec_denied &&
1040 ((seen & ALL_MODE_EXECS) != ALL_MODE_EXECS ||
1041 (mode & ALL_MODE_EXECS) != ALL_MODE_EXECS))
1042 an_exec_denied = B_TRUE;
1043
1044 if (an_exec_denied)
1045 *pflags &= ~ZFS_NO_EXECS_DENIED;
1046 else
1047 *pflags |= ZFS_NO_EXECS_DENIED;
1048
1049 return (mode);
1050 }
1051
1052 /*
1053 * Read an external acl object. If the intent is to modify, always
1054 * create a new acl and leave any cached acl in place.
1055 */
1056 static int
1057 zfs_acl_node_read(znode_t *zp, boolean_t have_lock, zfs_acl_t **aclpp,
1058 boolean_t will_modify)
1059 {
1060 zfs_acl_t *aclp;
1061 int aclsize = 0;
1062 int acl_count = 0;
1063 zfs_acl_node_t *aclnode;
1064 zfs_acl_phys_t znode_acl;
1065 int version;
1066 int error;
1067 boolean_t drop_lock = B_FALSE;
1068
1069 ASSERT(MUTEX_HELD(&zp->z_acl_lock));
1070
1071 if (zp->z_acl_cached && !will_modify) {
1072 *aclpp = zp->z_acl_cached;
1073 return (0);
1074 }
1075
1076 /*
1077 * close race where znode could be upgrade while trying to
1078 * read the znode attributes.
1079 *
1080 * But this could only happen if the file isn't already an SA
1081 * znode
1082 */
1083 if (!zp->z_is_sa && !have_lock) {
1084 mutex_enter(&zp->z_lock);
1085 drop_lock = B_TRUE;
1086 }
1087 version = zfs_znode_acl_version(zp);
1088
1089 if ((error = zfs_acl_znode_info(zp, &aclsize,
1090 &acl_count, &znode_acl)) != 0) {
1091 goto done;
1092 }
1093
1094 aclp = zfs_acl_alloc(version);
1095
1096 aclp->z_acl_count = acl_count;
1097 aclp->z_acl_bytes = aclsize;
1098
1099 aclnode = zfs_acl_node_alloc(aclsize);
1100 aclnode->z_ace_count = aclp->z_acl_count;
1101 aclnode->z_size = aclsize;
1102
1103 if (!zp->z_is_sa) {
1104 if (znode_acl.z_acl_extern_obj) {
1105 error = dmu_read(ZTOZSB(zp)->z_os,
1106 znode_acl.z_acl_extern_obj, 0, aclnode->z_size,
1107 aclnode->z_acldata, DMU_READ_PREFETCH);
1108 } else {
1109 bcopy(znode_acl.z_ace_data, aclnode->z_acldata,
1110 aclnode->z_size);
1111 }
1112 } else {
1113 error = sa_lookup(zp->z_sa_hdl, SA_ZPL_DACL_ACES(ZTOZSB(zp)),
1114 aclnode->z_acldata, aclnode->z_size);
1115 }
1116
1117 if (error != 0) {
1118 zfs_acl_free(aclp);
1119 zfs_acl_node_free(aclnode);
1120 /* convert checksum errors into IO errors */
1121 if (error == ECKSUM)
1122 error = SET_ERROR(EIO);
1123 goto done;
1124 }
1125
1126 list_insert_head(&aclp->z_acl, aclnode);
1127
1128 *aclpp = aclp;
1129 if (!will_modify)
1130 zp->z_acl_cached = aclp;
1131 done:
1132 if (drop_lock)
1133 mutex_exit(&zp->z_lock);
1134 return (error);
1135 }
1136
1137 /*ARGSUSED*/
1138 void
1139 zfs_acl_data_locator(void **dataptr, uint32_t *length, uint32_t buflen,
1140 boolean_t start, void *userdata)
1141 {
1142 zfs_acl_locator_cb_t *cb = (zfs_acl_locator_cb_t *)userdata;
1143
1144 if (start) {
1145 cb->cb_acl_node = list_head(&cb->cb_aclp->z_acl);
1146 } else {
1147 cb->cb_acl_node = list_next(&cb->cb_aclp->z_acl,
1148 cb->cb_acl_node);
1149 }
1150 *dataptr = cb->cb_acl_node->z_acldata;
1151 *length = cb->cb_acl_node->z_size;
1152 }
1153
1154 int
1155 zfs_acl_chown_setattr(znode_t *zp)
1156 {
1157 int error;
1158 zfs_acl_t *aclp;
1159
1160 if (ZTOZSB(zp)->z_acl_type == ZFS_ACLTYPE_POSIXACL)
1161 return (0);
1162
1163 ASSERT(MUTEX_HELD(&zp->z_lock));
1164 ASSERT(MUTEX_HELD(&zp->z_acl_lock));
1165
1166 error = zfs_acl_node_read(zp, B_TRUE, &aclp, B_FALSE);
1167 if (error == 0 && aclp->z_acl_count > 0)
1168 zp->z_mode = zfs_mode_compute(zp->z_mode, aclp,
1169 &zp->z_pflags, zp->z_uid, zp->z_gid);
1170
1171 /*
1172 * Some ZFS implementations (ZEVO) create neither a ZNODE_ACL
1173 * nor a DACL_ACES SA in which case ENOENT is returned from
1174 * zfs_acl_node_read() when the SA can't be located.
1175 * Allow chown/chgrp to succeed in these cases rather than
1176 * returning an error that makes no sense in the context of
1177 * the caller.
1178 */
1179 if (error == ENOENT)
1180 return (0);
1181
1182 return (error);
1183 }
1184
1185 static void
1186 acl_trivial_access_masks(mode_t mode, uint32_t *allow0, uint32_t *deny1,
1187 uint32_t *deny2, uint32_t *owner, uint32_t *group, uint32_t *everyone)
1188 {
1189 *deny1 = *deny2 = *allow0 = *group = 0;
1190
1191 if (!(mode & S_IRUSR) && (mode & (S_IRGRP|S_IROTH)))
1192 *deny1 |= ACE_READ_DATA;
1193 if (!(mode & S_IWUSR) && (mode & (S_IWGRP|S_IWOTH)))
1194 *deny1 |= ACE_WRITE_DATA;
1195 if (!(mode & S_IXUSR) && (mode & (S_IXGRP|S_IXOTH)))
1196 *deny1 |= ACE_EXECUTE;
1197
1198 if (!(mode & S_IRGRP) && (mode & S_IROTH))
1199 *deny2 = ACE_READ_DATA;
1200 if (!(mode & S_IWGRP) && (mode & S_IWOTH))
1201 *deny2 |= ACE_WRITE_DATA;
1202 if (!(mode & S_IXGRP) && (mode & S_IXOTH))
1203 *deny2 |= ACE_EXECUTE;
1204
1205 if ((mode & S_IRUSR) && (!(mode & S_IRGRP) && (mode & S_IROTH)))
1206 *allow0 |= ACE_READ_DATA;
1207 if ((mode & S_IWUSR) && (!(mode & S_IWGRP) && (mode & S_IWOTH)))
1208 *allow0 |= ACE_WRITE_DATA;
1209 if ((mode & S_IXUSR) && (!(mode & S_IXGRP) && (mode & S_IXOTH)))
1210 *allow0 |= ACE_EXECUTE;
1211
1212 *owner = ACE_WRITE_ATTRIBUTES|ACE_WRITE_OWNER|ACE_WRITE_ACL|
1213 ACE_WRITE_NAMED_ATTRS|ACE_READ_ACL|ACE_READ_ATTRIBUTES|
1214 ACE_READ_NAMED_ATTRS|ACE_SYNCHRONIZE;
1215 if (mode & S_IRUSR)
1216 *owner |= ACE_READ_DATA;
1217 if (mode & S_IWUSR)
1218 *owner |= ACE_WRITE_DATA|ACE_APPEND_DATA;
1219 if (mode & S_IXUSR)
1220 *owner |= ACE_EXECUTE;
1221
1222 *group = ACE_READ_ACL|ACE_READ_ATTRIBUTES| ACE_READ_NAMED_ATTRS|
1223 ACE_SYNCHRONIZE;
1224 if (mode & S_IRGRP)
1225 *group |= ACE_READ_DATA;
1226 if (mode & S_IWGRP)
1227 *group |= ACE_WRITE_DATA|ACE_APPEND_DATA;
1228 if (mode & S_IXGRP)
1229 *group |= ACE_EXECUTE;
1230
1231 *everyone = ACE_READ_ACL|ACE_READ_ATTRIBUTES| ACE_READ_NAMED_ATTRS|
1232 ACE_SYNCHRONIZE;
1233 if (mode & S_IROTH)
1234 *everyone |= ACE_READ_DATA;
1235 if (mode & S_IWOTH)
1236 *everyone |= ACE_WRITE_DATA|ACE_APPEND_DATA;
1237 if (mode & S_IXOTH)
1238 *everyone |= ACE_EXECUTE;
1239 }
1240
1241 /*
1242 * ace_trivial:
1243 * determine whether an ace_t acl is trivial
1244 *
1245 * Trivialness implies that the acl is composed of only
1246 * owner, group, everyone entries. ACL can't
1247 * have read_acl denied, and write_owner/write_acl/write_attributes
1248 * can only be owner@ entry.
1249 */
1250 static int
1251 ace_trivial_common(void *acep, int aclcnt,
1252 uint64_t (*walk)(void *, uint64_t, int aclcnt,
1253 uint16_t *, uint16_t *, uint32_t *))
1254 {
1255 uint16_t flags;
1256 uint32_t mask;
1257 uint16_t type;
1258 uint64_t cookie = 0;
1259
1260 while ((cookie = walk(acep, cookie, aclcnt, &flags, &type, &mask))) {
1261 switch (flags & ACE_TYPE_FLAGS) {
1262 case ACE_OWNER:
1263 case ACE_GROUP|ACE_IDENTIFIER_GROUP:
1264 case ACE_EVERYONE:
1265 break;
1266 default:
1267 return (1);
1268 }
1269
1270 if (flags & (ACE_FILE_INHERIT_ACE|
1271 ACE_DIRECTORY_INHERIT_ACE|ACE_NO_PROPAGATE_INHERIT_ACE|
1272 ACE_INHERIT_ONLY_ACE))
1273 return (1);
1274
1275 /*
1276 * Special check for some special bits
1277 *
1278 * Don't allow anybody to deny reading basic
1279 * attributes or a files ACL.
1280 */
1281 if ((mask & (ACE_READ_ACL|ACE_READ_ATTRIBUTES)) &&
1282 (type == ACE_ACCESS_DENIED_ACE_TYPE))
1283 return (1);
1284
1285 /*
1286 * Delete permissions are never set by default
1287 */
1288 if (mask & (ACE_DELETE|ACE_DELETE_CHILD))
1289 return (1);
1290 /*
1291 * only allow owner@ to have
1292 * write_acl/write_owner/write_attributes/write_xattr/
1293 */
1294 if (type == ACE_ACCESS_ALLOWED_ACE_TYPE &&
1295 (!(flags & ACE_OWNER) && (mask &
1296 (ACE_WRITE_OWNER|ACE_WRITE_ACL| ACE_WRITE_ATTRIBUTES|
1297 ACE_WRITE_NAMED_ATTRS))))
1298 return (1);
1299
1300 }
1301
1302 return (0);
1303 }
1304
1305 /*
1306 * common code for setting ACLs.
1307 *
1308 * This function is called from zfs_mode_update, zfs_perm_init, and zfs_setacl.
1309 * zfs_setacl passes a non-NULL inherit pointer (ihp) to indicate that it's
1310 * already checked the acl and knows whether to inherit.
1311 */
1312 int
1313 zfs_aclset_common(znode_t *zp, zfs_acl_t *aclp, cred_t *cr, dmu_tx_t *tx)
1314 {
1315 int error;
1316 zfs_sb_t *zsb = ZTOZSB(zp);
1317 dmu_object_type_t otype;
1318 zfs_acl_locator_cb_t locate = { 0 };
1319 uint64_t mode;
1320 sa_bulk_attr_t bulk[5];
1321 uint64_t ctime[2];
1322 int count = 0;
1323
1324 mode = zp->z_mode;
1325
1326 mode = zfs_mode_compute(mode, aclp, &zp->z_pflags,
1327 zp->z_uid, zp->z_gid);
1328
1329 zp->z_mode = mode;
1330 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zsb), NULL,
1331 &mode, sizeof (mode));
1332 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zsb), NULL,
1333 &zp->z_pflags, sizeof (zp->z_pflags));
1334 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zsb), NULL,
1335 &ctime, sizeof (ctime));
1336
1337 if (zp->z_acl_cached) {
1338 zfs_acl_free(zp->z_acl_cached);
1339 zp->z_acl_cached = NULL;
1340 }
1341
1342 /*
1343 * Upgrade needed?
1344 */
1345 if (!zsb->z_use_fuids) {
1346 otype = DMU_OT_OLDACL;
1347 } else {
1348 if ((aclp->z_version == ZFS_ACL_VERSION_INITIAL) &&
1349 (zsb->z_version >= ZPL_VERSION_FUID))
1350 zfs_acl_xform(zp, aclp, cr);
1351 ASSERT(aclp->z_version >= ZFS_ACL_VERSION_FUID);
1352 otype = DMU_OT_ACL;
1353 }
1354
1355 /*
1356 * Arrgh, we have to handle old on disk format
1357 * as well as newer (preferred) SA format.
1358 */
1359
1360 if (zp->z_is_sa) { /* the easy case, just update the ACL attribute */
1361 locate.cb_aclp = aclp;
1362 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_DACL_ACES(zsb),
1363 zfs_acl_data_locator, &locate, aclp->z_acl_bytes);
1364 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_DACL_COUNT(zsb),
1365 NULL, &aclp->z_acl_count, sizeof (uint64_t));
1366 } else { /* Painful legacy way */
1367 zfs_acl_node_t *aclnode;
1368 uint64_t off = 0;
1369 zfs_acl_phys_t acl_phys;
1370 uint64_t aoid;
1371
1372 if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_ZNODE_ACL(zsb),
1373 &acl_phys, sizeof (acl_phys))) != 0)
1374 return (error);
1375
1376 aoid = acl_phys.z_acl_extern_obj;
1377
1378 if (aclp->z_acl_bytes > ZFS_ACE_SPACE) {
1379 /*
1380 * If ACL was previously external and we are now
1381 * converting to new ACL format then release old
1382 * ACL object and create a new one.
1383 */
1384 if (aoid &&
1385 aclp->z_version != acl_phys.z_acl_version) {
1386 error = dmu_object_free(zsb->z_os, aoid, tx);
1387 if (error)
1388 return (error);
1389 aoid = 0;
1390 }
1391 if (aoid == 0) {
1392 aoid = dmu_object_alloc(zsb->z_os,
1393 otype, aclp->z_acl_bytes,
1394 otype == DMU_OT_ACL ?
1395 DMU_OT_SYSACL : DMU_OT_NONE,
1396 otype == DMU_OT_ACL ?
1397 DN_MAX_BONUSLEN : 0, tx);
1398 } else {
1399 (void) dmu_object_set_blocksize(zsb->z_os,
1400 aoid, aclp->z_acl_bytes, 0, tx);
1401 }
1402 acl_phys.z_acl_extern_obj = aoid;
1403 for (aclnode = list_head(&aclp->z_acl); aclnode;
1404 aclnode = list_next(&aclp->z_acl, aclnode)) {
1405 if (aclnode->z_ace_count == 0)
1406 continue;
1407 dmu_write(zsb->z_os, aoid, off,
1408 aclnode->z_size, aclnode->z_acldata, tx);
1409 off += aclnode->z_size;
1410 }
1411 } else {
1412 void *start = acl_phys.z_ace_data;
1413 /*
1414 * Migrating back embedded?
1415 */
1416 if (acl_phys.z_acl_extern_obj) {
1417 error = dmu_object_free(zsb->z_os,
1418 acl_phys.z_acl_extern_obj, tx);
1419 if (error)
1420 return (error);
1421 acl_phys.z_acl_extern_obj = 0;
1422 }
1423
1424 for (aclnode = list_head(&aclp->z_acl); aclnode;
1425 aclnode = list_next(&aclp->z_acl, aclnode)) {
1426 if (aclnode->z_ace_count == 0)
1427 continue;
1428 bcopy(aclnode->z_acldata, start,
1429 aclnode->z_size);
1430 start = (caddr_t)start + aclnode->z_size;
1431 }
1432 }
1433 /*
1434 * If Old version then swap count/bytes to match old
1435 * layout of znode_acl_phys_t.
1436 */
1437 if (aclp->z_version == ZFS_ACL_VERSION_INITIAL) {
1438 acl_phys.z_acl_size = aclp->z_acl_count;
1439 acl_phys.z_acl_count = aclp->z_acl_bytes;
1440 } else {
1441 acl_phys.z_acl_size = aclp->z_acl_bytes;
1442 acl_phys.z_acl_count = aclp->z_acl_count;
1443 }
1444 acl_phys.z_acl_version = aclp->z_version;
1445
1446 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ZNODE_ACL(zsb), NULL,
1447 &acl_phys, sizeof (acl_phys));
1448 }
1449
1450 /*
1451 * Replace ACL wide bits, but first clear them.
1452 */
1453 zp->z_pflags &= ~ZFS_ACL_WIDE_FLAGS;
1454
1455 zp->z_pflags |= aclp->z_hints;
1456
1457 if (ace_trivial_common(aclp, 0, zfs_ace_walk) == 0)
1458 zp->z_pflags |= ZFS_ACL_TRIVIAL;
1459
1460 zfs_tstamp_update_setup(zp, STATE_CHANGED, NULL, ctime);
1461 return (sa_bulk_update(zp->z_sa_hdl, bulk, count, tx));
1462 }
1463
1464 static void
1465 zfs_acl_chmod(zfs_sb_t *zsb, uint64_t mode, zfs_acl_t *aclp)
1466 {
1467 void *acep = NULL;
1468 uint64_t who;
1469 int new_count, new_bytes;
1470 int ace_size;
1471 int entry_type;
1472 uint16_t iflags, type;
1473 uint32_t access_mask;
1474 zfs_acl_node_t *newnode;
1475 size_t abstract_size = aclp->z_ops->ace_abstract_size();
1476 void *zacep;
1477 uint32_t owner, group, everyone;
1478 uint32_t deny1, deny2, allow0;
1479
1480 new_count = new_bytes = 0;
1481
1482 acl_trivial_access_masks((mode_t)mode, &allow0, &deny1, &deny2,
1483 &owner, &group, &everyone);
1484
1485 newnode = zfs_acl_node_alloc((abstract_size * 6) + aclp->z_acl_bytes);
1486
1487 zacep = newnode->z_acldata;
1488 if (allow0) {
1489 zfs_set_ace(aclp, zacep, allow0, ALLOW, -1, ACE_OWNER);
1490 zacep = (void *)((uintptr_t)zacep + abstract_size);
1491 new_count++;
1492 new_bytes += abstract_size;
1493 }
1494 if (deny1) {
1495 zfs_set_ace(aclp, zacep, deny1, DENY, -1, ACE_OWNER);
1496 zacep = (void *)((uintptr_t)zacep + abstract_size);
1497 new_count++;
1498 new_bytes += abstract_size;
1499 }
1500 if (deny2) {
1501 zfs_set_ace(aclp, zacep, deny2, DENY, -1, OWNING_GROUP);
1502 zacep = (void *)((uintptr_t)zacep + abstract_size);
1503 new_count++;
1504 new_bytes += abstract_size;
1505 }
1506
1507 while ((acep = zfs_acl_next_ace(aclp, acep, &who, &access_mask,
1508 &iflags, &type))) {
1509 uint16_t inherit_flags;
1510
1511 entry_type = (iflags & ACE_TYPE_FLAGS);
1512 inherit_flags = (iflags & ALL_INHERIT);
1513
1514 if ((entry_type == ACE_OWNER || entry_type == ACE_EVERYONE ||
1515 (entry_type == OWNING_GROUP)) &&
1516 ((inherit_flags & ACE_INHERIT_ONLY_ACE) == 0)) {
1517 continue;
1518 }
1519
1520 if ((type != ALLOW && type != DENY) ||
1521 (inherit_flags & ACE_INHERIT_ONLY_ACE)) {
1522 if (inherit_flags)
1523 aclp->z_hints |= ZFS_INHERIT_ACE;
1524 switch (type) {
1525 case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE:
1526 case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE:
1527 case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE:
1528 case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE:
1529 aclp->z_hints |= ZFS_ACL_OBJ_ACE;
1530 break;
1531 }
1532 } else {
1533
1534 /*
1535 * Limit permissions to be no greater than
1536 * group permissions
1537 */
1538 if (zsb->z_acl_inherit == ZFS_ACL_RESTRICTED) {
1539 if (!(mode & S_IRGRP))
1540 access_mask &= ~ACE_READ_DATA;
1541 if (!(mode & S_IWGRP))
1542 access_mask &=
1543 ~(ACE_WRITE_DATA|ACE_APPEND_DATA);
1544 if (!(mode & S_IXGRP))
1545 access_mask &= ~ACE_EXECUTE;
1546 access_mask &=
1547 ~(ACE_WRITE_OWNER|ACE_WRITE_ACL|
1548 ACE_WRITE_ATTRIBUTES|ACE_WRITE_NAMED_ATTRS);
1549 }
1550 }
1551 zfs_set_ace(aclp, zacep, access_mask, type, who, iflags);
1552 ace_size = aclp->z_ops->ace_size(acep);
1553 zacep = (void *)((uintptr_t)zacep + ace_size);
1554 new_count++;
1555 new_bytes += ace_size;
1556 }
1557 zfs_set_ace(aclp, zacep, owner, 0, -1, ACE_OWNER);
1558 zacep = (void *)((uintptr_t)zacep + abstract_size);
1559 zfs_set_ace(aclp, zacep, group, 0, -1, OWNING_GROUP);
1560 zacep = (void *)((uintptr_t)zacep + abstract_size);
1561 zfs_set_ace(aclp, zacep, everyone, 0, -1, ACE_EVERYONE);
1562
1563 new_count += 3;
1564 new_bytes += abstract_size * 3;
1565 zfs_acl_release_nodes(aclp);
1566 aclp->z_acl_count = new_count;
1567 aclp->z_acl_bytes = new_bytes;
1568 newnode->z_ace_count = new_count;
1569 newnode->z_size = new_bytes;
1570 list_insert_tail(&aclp->z_acl, newnode);
1571 }
1572
1573 void
1574 zfs_acl_chmod_setattr(znode_t *zp, zfs_acl_t **aclp, uint64_t mode)
1575 {
1576 mutex_enter(&zp->z_acl_lock);
1577 mutex_enter(&zp->z_lock);
1578 *aclp = zfs_acl_alloc(zfs_acl_version_zp(zp));
1579 (*aclp)->z_hints = zp->z_pflags & V4_ACL_WIDE_FLAGS;
1580 zfs_acl_chmod(ZTOZSB(zp), mode, *aclp);
1581 mutex_exit(&zp->z_lock);
1582 mutex_exit(&zp->z_acl_lock);
1583 ASSERT(*aclp);
1584 }
1585
1586 /*
1587 * strip off write_owner and write_acl
1588 */
1589 static void
1590 zfs_restricted_update(zfs_sb_t *zsb, zfs_acl_t *aclp, void *acep)
1591 {
1592 uint32_t mask = aclp->z_ops->ace_mask_get(acep);
1593
1594 if ((zsb->z_acl_inherit == ZFS_ACL_RESTRICTED) &&
1595 (aclp->z_ops->ace_type_get(acep) == ALLOW)) {
1596 mask &= ~RESTRICTED_CLEAR;
1597 aclp->z_ops->ace_mask_set(acep, mask);
1598 }
1599 }
1600
1601 /*
1602 * Should ACE be inherited?
1603 */
1604 static int
1605 zfs_ace_can_use(umode_t obj_mode, uint16_t acep_flags)
1606 {
1607 int iflags = (acep_flags & 0xf);
1608
1609 if (S_ISDIR(obj_mode) && (iflags & ACE_DIRECTORY_INHERIT_ACE))
1610 return (1);
1611 else if (iflags & ACE_FILE_INHERIT_ACE)
1612 return (!(S_ISDIR(obj_mode) &&
1613 (iflags & ACE_NO_PROPAGATE_INHERIT_ACE)));
1614 return (0);
1615 }
1616
1617 /*
1618 * inherit inheritable ACEs from parent
1619 */
1620 static zfs_acl_t *
1621 zfs_acl_inherit(zfs_sb_t *zsb, umode_t obj_mode, zfs_acl_t *paclp,
1622 uint64_t mode, boolean_t *need_chmod)
1623 {
1624 void *pacep;
1625 void *acep;
1626 zfs_acl_node_t *aclnode;
1627 zfs_acl_t *aclp = NULL;
1628 uint64_t who;
1629 uint32_t access_mask;
1630 uint16_t iflags, newflags, type;
1631 size_t ace_size;
1632 void *data1, *data2;
1633 size_t data1sz, data2sz;
1634 boolean_t vdir = S_ISDIR(obj_mode);
1635 boolean_t vreg = S_ISREG(obj_mode);
1636 boolean_t passthrough, passthrough_x, noallow;
1637
1638 passthrough_x =
1639 zsb->z_acl_inherit == ZFS_ACL_PASSTHROUGH_X;
1640 passthrough = passthrough_x ||
1641 zsb->z_acl_inherit == ZFS_ACL_PASSTHROUGH;
1642 noallow =
1643 zsb->z_acl_inherit == ZFS_ACL_NOALLOW;
1644
1645 *need_chmod = B_TRUE;
1646 pacep = NULL;
1647 aclp = zfs_acl_alloc(paclp->z_version);
1648 if (zsb->z_acl_inherit == ZFS_ACL_DISCARD || S_ISLNK(obj_mode))
1649 return (aclp);
1650 while ((pacep = zfs_acl_next_ace(paclp, pacep, &who,
1651 &access_mask, &iflags, &type))) {
1652
1653 /*
1654 * don't inherit bogus ACEs
1655 */
1656 if (!zfs_acl_valid_ace_type(type, iflags))
1657 continue;
1658
1659 if (noallow && type == ALLOW)
1660 continue;
1661
1662 ace_size = aclp->z_ops->ace_size(pacep);
1663
1664 if (!zfs_ace_can_use(obj_mode, iflags))
1665 continue;
1666
1667 /*
1668 * If owner@, group@, or everyone@ inheritable
1669 * then zfs_acl_chmod() isn't needed.
1670 */
1671 if (passthrough &&
1672 ((iflags & (ACE_OWNER|ACE_EVERYONE)) ||
1673 ((iflags & OWNING_GROUP) ==
1674 OWNING_GROUP)) && (vreg || (vdir && (iflags &
1675 ACE_DIRECTORY_INHERIT_ACE)))) {
1676 *need_chmod = B_FALSE;
1677 }
1678
1679 if (!vdir && passthrough_x &&
1680 ((mode & (S_IXUSR | S_IXGRP | S_IXOTH)) == 0)) {
1681 access_mask &= ~ACE_EXECUTE;
1682 }
1683
1684 aclnode = zfs_acl_node_alloc(ace_size);
1685 list_insert_tail(&aclp->z_acl, aclnode);
1686 acep = aclnode->z_acldata;
1687
1688 zfs_set_ace(aclp, acep, access_mask, type,
1689 who, iflags|ACE_INHERITED_ACE);
1690
1691 /*
1692 * Copy special opaque data if any
1693 */
1694 if ((data1sz = paclp->z_ops->ace_data(pacep, &data1)) != 0) {
1695 VERIFY((data2sz = aclp->z_ops->ace_data(acep,
1696 &data2)) == data1sz);
1697 bcopy(data1, data2, data2sz);
1698 }
1699
1700 aclp->z_acl_count++;
1701 aclnode->z_ace_count++;
1702 aclp->z_acl_bytes += aclnode->z_size;
1703 newflags = aclp->z_ops->ace_flags_get(acep);
1704
1705 if (vdir)
1706 aclp->z_hints |= ZFS_INHERIT_ACE;
1707
1708 if ((iflags & ACE_NO_PROPAGATE_INHERIT_ACE) || !vdir) {
1709 newflags &= ~ALL_INHERIT;
1710 aclp->z_ops->ace_flags_set(acep,
1711 newflags|ACE_INHERITED_ACE);
1712 zfs_restricted_update(zsb, aclp, acep);
1713 continue;
1714 }
1715
1716 ASSERT(vdir);
1717
1718 /*
1719 * If only FILE_INHERIT is set then turn on
1720 * inherit_only
1721 */
1722 if ((iflags & (ACE_FILE_INHERIT_ACE |
1723 ACE_DIRECTORY_INHERIT_ACE)) == ACE_FILE_INHERIT_ACE) {
1724 newflags |= ACE_INHERIT_ONLY_ACE;
1725 aclp->z_ops->ace_flags_set(acep,
1726 newflags|ACE_INHERITED_ACE);
1727 } else {
1728 newflags &= ~ACE_INHERIT_ONLY_ACE;
1729 aclp->z_ops->ace_flags_set(acep,
1730 newflags|ACE_INHERITED_ACE);
1731 }
1732 }
1733 return (aclp);
1734 }
1735
1736 /*
1737 * Create file system object initial permissions
1738 * including inheritable ACEs.
1739 */
1740 int
1741 zfs_acl_ids_create(znode_t *dzp, int flag, vattr_t *vap, cred_t *cr,
1742 vsecattr_t *vsecp, zfs_acl_ids_t *acl_ids)
1743 {
1744 int error;
1745 zfs_sb_t *zsb = ZTOZSB(dzp);
1746 zfs_acl_t *paclp;
1747 gid_t gid = vap->va_gid;
1748 boolean_t need_chmod = B_TRUE;
1749 boolean_t inherited = B_FALSE;
1750
1751 bzero(acl_ids, sizeof (zfs_acl_ids_t));
1752 acl_ids->z_mode = vap->va_mode;
1753
1754 if (vsecp)
1755 if ((error = zfs_vsec_2_aclp(zsb, vap->va_mode, vsecp,
1756 cr, &acl_ids->z_fuidp, &acl_ids->z_aclp)) != 0)
1757 return (error);
1758
1759 acl_ids->z_fuid = vap->va_uid;
1760 acl_ids->z_fgid = vap->va_gid;
1761 #ifdef HAVE_KSID
1762 /*
1763 * Determine uid and gid.
1764 */
1765 if ((flag & IS_ROOT_NODE) || zsb->z_replay ||
1766 ((flag & IS_XATTR) && (S_ISDIR(vap->va_mode)))) {
1767 acl_ids->z_fuid = zfs_fuid_create(zsb, (uint64_t)vap->va_uid,
1768 cr, ZFS_OWNER, &acl_ids->z_fuidp);
1769 acl_ids->z_fgid = zfs_fuid_create(zsb, (uint64_t)vap->va_gid,
1770 cr, ZFS_GROUP, &acl_ids->z_fuidp);
1771 gid = vap->va_gid;
1772 } else {
1773 acl_ids->z_fuid = zfs_fuid_create_cred(zsb, ZFS_OWNER,
1774 cr, &acl_ids->z_fuidp);
1775 acl_ids->z_fgid = 0;
1776 if (vap->va_mask & AT_GID) {
1777 acl_ids->z_fgid = zfs_fuid_create(zsb,
1778 (uint64_t)vap->va_gid,
1779 cr, ZFS_GROUP, &acl_ids->z_fuidp);
1780 gid = vap->va_gid;
1781 if (acl_ids->z_fgid != dzp->z_gid &&
1782 !groupmember(vap->va_gid, cr) &&
1783 secpolicy_vnode_create_gid(cr) != 0)
1784 acl_ids->z_fgid = 0;
1785 }
1786 if (acl_ids->z_fgid == 0) {
1787 if (dzp->z_mode & S_ISGID) {
1788 char *domain;
1789 uint32_t rid;
1790
1791 acl_ids->z_fgid = dzp->z_gid;
1792 gid = zfs_fuid_map_id(zsb, acl_ids->z_fgid,
1793 cr, ZFS_GROUP);
1794
1795 if (zsb->z_use_fuids &&
1796 IS_EPHEMERAL(acl_ids->z_fgid)) {
1797 domain = zfs_fuid_idx_domain(
1798 &zsb->z_fuid_idx,
1799 FUID_INDEX(acl_ids->z_fgid));
1800 rid = FUID_RID(acl_ids->z_fgid);
1801 zfs_fuid_node_add(&acl_ids->z_fuidp,
1802 domain, rid,
1803 FUID_INDEX(acl_ids->z_fgid),
1804 acl_ids->z_fgid, ZFS_GROUP);
1805 }
1806 } else {
1807 acl_ids->z_fgid = zfs_fuid_create_cred(zsb,
1808 ZFS_GROUP, cr, &acl_ids->z_fuidp);
1809 gid = crgetgid(cr);
1810 }
1811 }
1812 }
1813 #endif /* HAVE_KSID */
1814
1815 /*
1816 * If we're creating a directory, and the parent directory has the
1817 * set-GID bit set, set in on the new directory.
1818 * Otherwise, if the user is neither privileged nor a member of the
1819 * file's new group, clear the file's set-GID bit.
1820 */
1821
1822 if (!(flag & IS_ROOT_NODE) && (dzp->z_mode & S_ISGID) &&
1823 (S_ISDIR(vap->va_mode))) {
1824 acl_ids->z_mode |= S_ISGID;
1825 } else {
1826 if ((acl_ids->z_mode & S_ISGID) &&
1827 secpolicy_vnode_setids_setgids(cr, gid) != 0)
1828 acl_ids->z_mode &= ~S_ISGID;
1829 }
1830
1831 if (acl_ids->z_aclp == NULL) {
1832 mutex_enter(&dzp->z_acl_lock);
1833 mutex_enter(&dzp->z_lock);
1834 if (!(flag & IS_ROOT_NODE) && (S_ISDIR(ZTOI(dzp)->i_mode) &&
1835 (dzp->z_pflags & ZFS_INHERIT_ACE)) &&
1836 !(dzp->z_pflags & ZFS_XATTR)) {
1837 VERIFY(0 == zfs_acl_node_read(dzp, B_TRUE,
1838 &paclp, B_FALSE));
1839 acl_ids->z_aclp = zfs_acl_inherit(zsb,
1840 vap->va_mode, paclp, acl_ids->z_mode, &need_chmod);
1841 inherited = B_TRUE;
1842 } else {
1843 acl_ids->z_aclp =
1844 zfs_acl_alloc(zfs_acl_version_zp(dzp));
1845 acl_ids->z_aclp->z_hints |= ZFS_ACL_TRIVIAL;
1846 }
1847 mutex_exit(&dzp->z_lock);
1848 mutex_exit(&dzp->z_acl_lock);
1849 if (need_chmod) {
1850 acl_ids->z_aclp->z_hints |= S_ISDIR(vap->va_mode) ?
1851 ZFS_ACL_AUTO_INHERIT : 0;
1852 zfs_acl_chmod(zsb, acl_ids->z_mode, acl_ids->z_aclp);
1853 }
1854 }
1855
1856 if (inherited || vsecp) {
1857 acl_ids->z_mode = zfs_mode_compute(acl_ids->z_mode,
1858 acl_ids->z_aclp, &acl_ids->z_aclp->z_hints,
1859 acl_ids->z_fuid, acl_ids->z_fgid);
1860 if (ace_trivial_common(acl_ids->z_aclp, 0, zfs_ace_walk) == 0)
1861 acl_ids->z_aclp->z_hints |= ZFS_ACL_TRIVIAL;
1862 }
1863
1864 return (0);
1865 }
1866
1867 /*
1868 * Free ACL and fuid_infop, but not the acl_ids structure
1869 */
1870 void
1871 zfs_acl_ids_free(zfs_acl_ids_t *acl_ids)
1872 {
1873 if (acl_ids->z_aclp)
1874 zfs_acl_free(acl_ids->z_aclp);
1875 if (acl_ids->z_fuidp)
1876 zfs_fuid_info_free(acl_ids->z_fuidp);
1877 acl_ids->z_aclp = NULL;
1878 acl_ids->z_fuidp = NULL;
1879 }
1880
1881 boolean_t
1882 zfs_acl_ids_overquota(zfs_sb_t *zsb, zfs_acl_ids_t *acl_ids)
1883 {
1884 return (zfs_fuid_overquota(zsb, B_FALSE, acl_ids->z_fuid) ||
1885 zfs_fuid_overquota(zsb, B_TRUE, acl_ids->z_fgid));
1886 }
1887
1888 /*
1889 * Retrieve a file's ACL
1890 */
1891 int
1892 zfs_getacl(znode_t *zp, vsecattr_t *vsecp, boolean_t skipaclchk, cred_t *cr)
1893 {
1894 zfs_acl_t *aclp;
1895 ulong_t mask;
1896 int error;
1897 int count = 0;
1898 int largeace = 0;
1899
1900 mask = vsecp->vsa_mask & (VSA_ACE | VSA_ACECNT |
1901 VSA_ACE_ACLFLAGS | VSA_ACE_ALLTYPES);
1902
1903 if (mask == 0)
1904 return (SET_ERROR(ENOSYS));
1905
1906 if ((error = zfs_zaccess(zp, ACE_READ_ACL, 0, skipaclchk, cr)))
1907 return (error);
1908
1909 mutex_enter(&zp->z_acl_lock);
1910
1911 error = zfs_acl_node_read(zp, B_FALSE, &aclp, B_FALSE);
1912 if (error != 0) {
1913 mutex_exit(&zp->z_acl_lock);
1914 return (error);
1915 }
1916
1917 /*
1918 * Scan ACL to determine number of ACEs
1919 */
1920 if ((zp->z_pflags & ZFS_ACL_OBJ_ACE) && !(mask & VSA_ACE_ALLTYPES)) {
1921 void *zacep = NULL;
1922 uint64_t who;
1923 uint32_t access_mask;
1924 uint16_t type, iflags;
1925
1926 while ((zacep = zfs_acl_next_ace(aclp, zacep,
1927 &who, &access_mask, &iflags, &type))) {
1928 switch (type) {
1929 case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE:
1930 case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE:
1931 case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE:
1932 case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE:
1933 largeace++;
1934 continue;
1935 default:
1936 count++;
1937 }
1938 }
1939 vsecp->vsa_aclcnt = count;
1940 } else
1941 count = (int)aclp->z_acl_count;
1942
1943 if (mask & VSA_ACECNT) {
1944 vsecp->vsa_aclcnt = count;
1945 }
1946
1947 if (mask & VSA_ACE) {
1948 size_t aclsz;
1949
1950 aclsz = count * sizeof (ace_t) +
1951 sizeof (ace_object_t) * largeace;
1952
1953 vsecp->vsa_aclentp = kmem_alloc(aclsz, KM_SLEEP);
1954 vsecp->vsa_aclentsz = aclsz;
1955
1956 if (aclp->z_version == ZFS_ACL_VERSION_FUID)
1957 zfs_copy_fuid_2_ace(ZTOZSB(zp), aclp, cr,
1958 vsecp->vsa_aclentp, !(mask & VSA_ACE_ALLTYPES));
1959 else {
1960 zfs_acl_node_t *aclnode;
1961 void *start = vsecp->vsa_aclentp;
1962
1963 for (aclnode = list_head(&aclp->z_acl); aclnode;
1964 aclnode = list_next(&aclp->z_acl, aclnode)) {
1965 bcopy(aclnode->z_acldata, start,
1966 aclnode->z_size);
1967 start = (caddr_t)start + aclnode->z_size;
1968 }
1969 ASSERT((caddr_t)start - (caddr_t)vsecp->vsa_aclentp ==
1970 aclp->z_acl_bytes);
1971 }
1972 }
1973 if (mask & VSA_ACE_ACLFLAGS) {
1974 vsecp->vsa_aclflags = 0;
1975 if (zp->z_pflags & ZFS_ACL_DEFAULTED)
1976 vsecp->vsa_aclflags |= ACL_DEFAULTED;
1977 if (zp->z_pflags & ZFS_ACL_PROTECTED)
1978 vsecp->vsa_aclflags |= ACL_PROTECTED;
1979 if (zp->z_pflags & ZFS_ACL_AUTO_INHERIT)
1980 vsecp->vsa_aclflags |= ACL_AUTO_INHERIT;
1981 }
1982
1983 mutex_exit(&zp->z_acl_lock);
1984
1985 return (0);
1986 }
1987
1988 int
1989 zfs_vsec_2_aclp(zfs_sb_t *zsb, umode_t obj_mode,
1990 vsecattr_t *vsecp, cred_t *cr, zfs_fuid_info_t **fuidp, zfs_acl_t **zaclp)
1991 {
1992 zfs_acl_t *aclp;
1993 zfs_acl_node_t *aclnode;
1994 int aclcnt = vsecp->vsa_aclcnt;
1995 int error;
1996
1997 if (vsecp->vsa_aclcnt > MAX_ACL_ENTRIES || vsecp->vsa_aclcnt <= 0)
1998 return (SET_ERROR(EINVAL));
1999
2000 aclp = zfs_acl_alloc(zfs_acl_version(zsb->z_version));
2001
2002 aclp->z_hints = 0;
2003 aclnode = zfs_acl_node_alloc(aclcnt * sizeof (zfs_object_ace_t));
2004 if (aclp->z_version == ZFS_ACL_VERSION_INITIAL) {
2005 if ((error = zfs_copy_ace_2_oldace(obj_mode, aclp,
2006 (ace_t *)vsecp->vsa_aclentp, aclnode->z_acldata,
2007 aclcnt, &aclnode->z_size)) != 0) {
2008 zfs_acl_free(aclp);
2009 zfs_acl_node_free(aclnode);
2010 return (error);
2011 }
2012 } else {
2013 if ((error = zfs_copy_ace_2_fuid(zsb, obj_mode, aclp,
2014 vsecp->vsa_aclentp, aclnode->z_acldata, aclcnt,
2015 &aclnode->z_size, fuidp, cr)) != 0) {
2016 zfs_acl_free(aclp);
2017 zfs_acl_node_free(aclnode);
2018 return (error);
2019 }
2020 }
2021 aclp->z_acl_bytes = aclnode->z_size;
2022 aclnode->z_ace_count = aclcnt;
2023 aclp->z_acl_count = aclcnt;
2024 list_insert_head(&aclp->z_acl, aclnode);
2025
2026 /*
2027 * If flags are being set then add them to z_hints
2028 */
2029 if (vsecp->vsa_mask & VSA_ACE_ACLFLAGS) {
2030 if (vsecp->vsa_aclflags & ACL_PROTECTED)
2031 aclp->z_hints |= ZFS_ACL_PROTECTED;
2032 if (vsecp->vsa_aclflags & ACL_DEFAULTED)
2033 aclp->z_hints |= ZFS_ACL_DEFAULTED;
2034 if (vsecp->vsa_aclflags & ACL_AUTO_INHERIT)
2035 aclp->z_hints |= ZFS_ACL_AUTO_INHERIT;
2036 }
2037
2038 *zaclp = aclp;
2039
2040 return (0);
2041 }
2042
2043 /*
2044 * Set a file's ACL
2045 */
2046 int
2047 zfs_setacl(znode_t *zp, vsecattr_t *vsecp, boolean_t skipaclchk, cred_t *cr)
2048 {
2049 zfs_sb_t *zsb = ZTOZSB(zp);
2050 zilog_t *zilog = zsb->z_log;
2051 ulong_t mask = vsecp->vsa_mask & (VSA_ACE | VSA_ACECNT);
2052 dmu_tx_t *tx;
2053 int error;
2054 zfs_acl_t *aclp;
2055 zfs_fuid_info_t *fuidp = NULL;
2056 boolean_t fuid_dirtied;
2057 uint64_t acl_obj;
2058
2059 if (mask == 0)
2060 return (SET_ERROR(ENOSYS));
2061
2062 if (zp->z_pflags & ZFS_IMMUTABLE)
2063 return (SET_ERROR(EPERM));
2064
2065 if ((error = zfs_zaccess(zp, ACE_WRITE_ACL, 0, skipaclchk, cr)))
2066 return (error);
2067
2068 error = zfs_vsec_2_aclp(zsb, ZTOI(zp)->i_mode, vsecp, cr, &fuidp,
2069 &aclp);
2070 if (error)
2071 return (error);
2072
2073 /*
2074 * If ACL wide flags aren't being set then preserve any
2075 * existing flags.
2076 */
2077 if (!(vsecp->vsa_mask & VSA_ACE_ACLFLAGS)) {
2078 aclp->z_hints |=
2079 (zp->z_pflags & V4_ACL_WIDE_FLAGS);
2080 }
2081 top:
2082 mutex_enter(&zp->z_acl_lock);
2083 mutex_enter(&zp->z_lock);
2084
2085 tx = dmu_tx_create(zsb->z_os);
2086
2087 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE);
2088
2089 fuid_dirtied = zsb->z_fuid_dirty;
2090 if (fuid_dirtied)
2091 zfs_fuid_txhold(zsb, tx);
2092
2093 /*
2094 * If old version and ACL won't fit in bonus and we aren't
2095 * upgrading then take out necessary DMU holds
2096 */
2097
2098 if ((acl_obj = zfs_external_acl(zp)) != 0) {
2099 if (zsb->z_version >= ZPL_VERSION_FUID &&
2100 zfs_znode_acl_version(zp) <= ZFS_ACL_VERSION_INITIAL) {
2101 dmu_tx_hold_free(tx, acl_obj, 0,
2102 DMU_OBJECT_END);
2103 dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0,
2104 aclp->z_acl_bytes);
2105 } else {
2106 dmu_tx_hold_write(tx, acl_obj, 0, aclp->z_acl_bytes);
2107 }
2108 } else if (!zp->z_is_sa && aclp->z_acl_bytes > ZFS_ACE_SPACE) {
2109 dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, aclp->z_acl_bytes);
2110 }
2111
2112 zfs_sa_upgrade_txholds(tx, zp);
2113 error = dmu_tx_assign(tx, TXG_NOWAIT);
2114 if (error) {
2115 mutex_exit(&zp->z_acl_lock);
2116 mutex_exit(&zp->z_lock);
2117
2118 if (error == ERESTART) {
2119 dmu_tx_wait(tx);
2120 dmu_tx_abort(tx);
2121 goto top;
2122 }
2123 dmu_tx_abort(tx);
2124 zfs_acl_free(aclp);
2125 return (error);
2126 }
2127
2128 error = zfs_aclset_common(zp, aclp, cr, tx);
2129 ASSERT(error == 0);
2130 ASSERT(zp->z_acl_cached == NULL);
2131 zp->z_acl_cached = aclp;
2132
2133 if (fuid_dirtied)
2134 zfs_fuid_sync(zsb, tx);
2135
2136 zfs_log_acl(zilog, tx, zp, vsecp, fuidp);
2137
2138 if (fuidp)
2139 zfs_fuid_info_free(fuidp);
2140 dmu_tx_commit(tx);
2141
2142 mutex_exit(&zp->z_lock);
2143 mutex_exit(&zp->z_acl_lock);
2144
2145 return (error);
2146 }
2147
2148 /*
2149 * Check accesses of interest (AoI) against attributes of the dataset
2150 * such as read-only. Returns zero if no AoI conflict with dataset
2151 * attributes, otherwise an appropriate errno is returned.
2152 */
2153 static int
2154 zfs_zaccess_dataset_check(znode_t *zp, uint32_t v4_mode)
2155 {
2156 if ((v4_mode & WRITE_MASK) && (zfs_is_readonly(ZTOZSB(zp))) &&
2157 (!S_ISDEV(ZTOI(zp)->i_mode) ||
2158 (S_ISDEV(ZTOI(zp)->i_mode) && (v4_mode & WRITE_MASK_ATTRS)))) {
2159 return (SET_ERROR(EROFS));
2160 }
2161
2162 /*
2163 * Only check for READONLY on non-directories.
2164 */
2165 if ((v4_mode & WRITE_MASK_DATA) &&
2166 ((!S_ISDIR(ZTOI(zp)->i_mode) &&
2167 (zp->z_pflags & (ZFS_READONLY | ZFS_IMMUTABLE))) ||
2168 (S_ISDIR(ZTOI(zp)->i_mode) &&
2169 (zp->z_pflags & ZFS_IMMUTABLE)))) {
2170 return (SET_ERROR(EPERM));
2171 }
2172
2173 if ((v4_mode & (ACE_DELETE | ACE_DELETE_CHILD)) &&
2174 (zp->z_pflags & ZFS_NOUNLINK)) {
2175 return (SET_ERROR(EPERM));
2176 }
2177
2178 if (((v4_mode & (ACE_READ_DATA|ACE_EXECUTE)) &&
2179 (zp->z_pflags & ZFS_AV_QUARANTINED))) {
2180 return (SET_ERROR(EACCES));
2181 }
2182
2183 return (0);
2184 }
2185
2186 /*
2187 * The primary usage of this function is to loop through all of the
2188 * ACEs in the znode, determining what accesses of interest (AoI) to
2189 * the caller are allowed or denied. The AoI are expressed as bits in
2190 * the working_mode parameter. As each ACE is processed, bits covered
2191 * by that ACE are removed from the working_mode. This removal
2192 * facilitates two things. The first is that when the working mode is
2193 * empty (= 0), we know we've looked at all the AoI. The second is
2194 * that the ACE interpretation rules don't allow a later ACE to undo
2195 * something granted or denied by an earlier ACE. Removing the
2196 * discovered access or denial enforces this rule. At the end of
2197 * processing the ACEs, all AoI that were found to be denied are
2198 * placed into the working_mode, giving the caller a mask of denied
2199 * accesses. Returns:
2200 * 0 if all AoI granted
2201 * EACCESS if the denied mask is non-zero
2202 * other error if abnormal failure (e.g., IO error)
2203 *
2204 * A secondary usage of the function is to determine if any of the
2205 * AoI are granted. If an ACE grants any access in
2206 * the working_mode, we immediately short circuit out of the function.
2207 * This mode is chosen by setting anyaccess to B_TRUE. The
2208 * working_mode is not a denied access mask upon exit if the function
2209 * is used in this manner.
2210 */
2211 static int
2212 zfs_zaccess_aces_check(znode_t *zp, uint32_t *working_mode,
2213 boolean_t anyaccess, cred_t *cr)
2214 {
2215 zfs_sb_t *zsb = ZTOZSB(zp);
2216 zfs_acl_t *aclp;
2217 int error;
2218 uid_t uid = crgetuid(cr);
2219 uint64_t who;
2220 uint16_t type, iflags;
2221 uint16_t entry_type;
2222 uint32_t access_mask;
2223 uint32_t deny_mask = 0;
2224 zfs_ace_hdr_t *acep = NULL;
2225 boolean_t checkit;
2226 uid_t gowner;
2227 uid_t fowner;
2228
2229 zfs_fuid_map_ids(zp, cr, &fowner, &gowner);
2230
2231 mutex_enter(&zp->z_acl_lock);
2232
2233 error = zfs_acl_node_read(zp, B_FALSE, &aclp, B_FALSE);
2234 if (error != 0) {
2235 mutex_exit(&zp->z_acl_lock);
2236 return (error);
2237 }
2238
2239 ASSERT(zp->z_acl_cached);
2240
2241 while ((acep = zfs_acl_next_ace(aclp, acep, &who, &access_mask,
2242 &iflags, &type))) {
2243 uint32_t mask_matched;
2244
2245 if (!zfs_acl_valid_ace_type(type, iflags))
2246 continue;
2247
2248 if (S_ISDIR(ZTOI(zp)->i_mode) &&
2249 (iflags & ACE_INHERIT_ONLY_ACE))
2250 continue;
2251
2252 /* Skip ACE if it does not affect any AoI */
2253 mask_matched = (access_mask & *working_mode);
2254 if (!mask_matched)
2255 continue;
2256
2257 entry_type = (iflags & ACE_TYPE_FLAGS);
2258
2259 checkit = B_FALSE;
2260
2261 switch (entry_type) {
2262 case ACE_OWNER:
2263 if (uid == fowner)
2264 checkit = B_TRUE;
2265 break;
2266 case OWNING_GROUP:
2267 who = gowner;
2268 /*FALLTHROUGH*/
2269 case ACE_IDENTIFIER_GROUP:
2270 checkit = zfs_groupmember(zsb, who, cr);
2271 break;
2272 case ACE_EVERYONE:
2273 checkit = B_TRUE;
2274 break;
2275
2276 /* USER Entry */
2277 default:
2278 if (entry_type == 0) {
2279 uid_t newid;
2280
2281 newid = zfs_fuid_map_id(zsb, who, cr,
2282 ZFS_ACE_USER);
2283 if (newid != IDMAP_WK_CREATOR_OWNER_UID &&
2284 uid == newid)
2285 checkit = B_TRUE;
2286 break;
2287 } else {
2288 mutex_exit(&zp->z_acl_lock);
2289 return (SET_ERROR(EIO));
2290 }
2291 }
2292
2293 if (checkit) {
2294 if (type == DENY) {
2295 DTRACE_PROBE3(zfs__ace__denies,
2296 znode_t *, zp,
2297 zfs_ace_hdr_t *, acep,
2298 uint32_t, mask_matched);
2299 deny_mask |= mask_matched;
2300 } else {
2301 DTRACE_PROBE3(zfs__ace__allows,
2302 znode_t *, zp,
2303 zfs_ace_hdr_t *, acep,
2304 uint32_t, mask_matched);
2305 if (anyaccess) {
2306 mutex_exit(&zp->z_acl_lock);
2307 return (0);
2308 }
2309 }
2310 *working_mode &= ~mask_matched;
2311 }
2312
2313 /* Are we done? */
2314 if (*working_mode == 0)
2315 break;
2316 }
2317
2318 mutex_exit(&zp->z_acl_lock);
2319
2320 /* Put the found 'denies' back on the working mode */
2321 if (deny_mask) {
2322 *working_mode |= deny_mask;
2323 return (SET_ERROR(EACCES));
2324 } else if (*working_mode) {
2325 return (-1);
2326 }
2327
2328 return (0);
2329 }
2330
2331 /*
2332 * Return true if any access whatsoever granted, we don't actually
2333 * care what access is granted.
2334 */
2335 boolean_t
2336 zfs_has_access(znode_t *zp, cred_t *cr)
2337 {
2338 uint32_t have = ACE_ALL_PERMS;
2339
2340 if (zfs_zaccess_aces_check(zp, &have, B_TRUE, cr) != 0) {
2341 uid_t owner;
2342
2343 owner = zfs_fuid_map_id(ZTOZSB(zp), zp->z_uid, cr, ZFS_OWNER);
2344 return (secpolicy_vnode_any_access(cr, ZTOI(zp), owner) == 0);
2345 }
2346 return (B_TRUE);
2347 }
2348
2349 static int
2350 zfs_zaccess_common(znode_t *zp, uint32_t v4_mode, uint32_t *working_mode,
2351 boolean_t *check_privs, boolean_t skipaclchk, cred_t *cr)
2352 {
2353 zfs_sb_t *zsb = ZTOZSB(zp);
2354 int err;
2355
2356 *working_mode = v4_mode;
2357 *check_privs = B_TRUE;
2358
2359 /*
2360 * Short circuit empty requests
2361 */
2362 if (v4_mode == 0 || zsb->z_replay) {
2363 *working_mode = 0;
2364 return (0);
2365 }
2366
2367 if ((err = zfs_zaccess_dataset_check(zp, v4_mode)) != 0) {
2368 *check_privs = B_FALSE;
2369 return (err);
2370 }
2371
2372 /*
2373 * The caller requested that the ACL check be skipped. This
2374 * would only happen if the caller checked VOP_ACCESS() with a
2375 * 32 bit ACE mask and already had the appropriate permissions.
2376 */
2377 if (skipaclchk) {
2378 *working_mode = 0;
2379 return (0);
2380 }
2381
2382 return (zfs_zaccess_aces_check(zp, working_mode, B_FALSE, cr));
2383 }
2384
2385 static int
2386 zfs_zaccess_append(znode_t *zp, uint32_t *working_mode, boolean_t *check_privs,
2387 cred_t *cr)
2388 {
2389 if (*working_mode != ACE_WRITE_DATA)
2390 return (SET_ERROR(EACCES));
2391
2392 return (zfs_zaccess_common(zp, ACE_APPEND_DATA, working_mode,
2393 check_privs, B_FALSE, cr));
2394 }
2395
2396 int
2397 zfs_fastaccesschk_execute(znode_t *zdp, cred_t *cr)
2398 {
2399 boolean_t owner = B_FALSE;
2400 boolean_t groupmbr = B_FALSE;
2401 boolean_t is_attr;
2402 uid_t uid = crgetuid(cr);
2403 int error;
2404
2405 if (zdp->z_pflags & ZFS_AV_QUARANTINED)
2406 return (SET_ERROR(EACCES));
2407
2408 is_attr = ((zdp->z_pflags & ZFS_XATTR) &&
2409 (S_ISDIR(ZTOI(zdp)->i_mode)));
2410 if (is_attr)
2411 goto slow;
2412
2413
2414 mutex_enter(&zdp->z_acl_lock);
2415
2416 if (zdp->z_pflags & ZFS_NO_EXECS_DENIED) {
2417 mutex_exit(&zdp->z_acl_lock);
2418 return (0);
2419 }
2420
2421 if (FUID_INDEX(zdp->z_uid) != 0 || FUID_INDEX(zdp->z_gid) != 0) {
2422 mutex_exit(&zdp->z_acl_lock);
2423 goto slow;
2424 }
2425
2426 if (uid == zdp->z_uid) {
2427 owner = B_TRUE;
2428 if (zdp->z_mode & S_IXUSR) {
2429 mutex_exit(&zdp->z_acl_lock);
2430 return (0);
2431 } else {
2432 mutex_exit(&zdp->z_acl_lock);
2433 goto slow;
2434 }
2435 }
2436 if (groupmember(zdp->z_gid, cr)) {
2437 groupmbr = B_TRUE;
2438 if (zdp->z_mode & S_IXGRP) {
2439 mutex_exit(&zdp->z_acl_lock);
2440 return (0);
2441 } else {
2442 mutex_exit(&zdp->z_acl_lock);
2443 goto slow;
2444 }
2445 }
2446 if (!owner && !groupmbr) {
2447 if (zdp->z_mode & S_IXOTH) {
2448 mutex_exit(&zdp->z_acl_lock);
2449 return (0);
2450 }
2451 }
2452
2453 mutex_exit(&zdp->z_acl_lock);
2454
2455 slow:
2456 DTRACE_PROBE(zfs__fastpath__execute__access__miss);
2457 ZFS_ENTER(ZTOZSB(zdp));
2458 error = zfs_zaccess(zdp, ACE_EXECUTE, 0, B_FALSE, cr);
2459 ZFS_EXIT(ZTOZSB(zdp));
2460 return (error);
2461 }
2462
2463 /*
2464 * Determine whether Access should be granted/denied.
2465 *
2466 * The least priv subsytem is always consulted as a basic privilege
2467 * can define any form of access.
2468 */
2469 int
2470 zfs_zaccess(znode_t *zp, int mode, int flags, boolean_t skipaclchk, cred_t *cr)
2471 {
2472 uint32_t working_mode;
2473 int error;
2474 boolean_t check_privs;
2475 znode_t *check_zp = zp;
2476 mode_t needed_bits;
2477 uid_t owner;
2478
2479 /*
2480 * If attribute then validate against base file
2481 */
2482 if ((zp->z_pflags & ZFS_XATTR) && S_ISDIR(ZTOI(zp)->i_mode)) {
2483 uint64_t parent;
2484
2485 rw_enter(&zp->z_xattr_lock, RW_READER);
2486 if (zp->z_xattr_parent) {
2487 check_zp = zp->z_xattr_parent;
2488 rw_exit(&zp->z_xattr_lock);
2489
2490 /*
2491 * Verify a lookup yields the same znode.
2492 */
2493 ASSERT3S(sa_lookup(zp->z_sa_hdl, SA_ZPL_PARENT(
2494 ZTOZSB(zp)), &parent, sizeof (parent)), ==, 0);
2495 ASSERT3U(check_zp->z_id, ==, parent);
2496 } else {
2497 rw_exit(&zp->z_xattr_lock);
2498
2499 error = sa_lookup(zp->z_sa_hdl, SA_ZPL_PARENT(
2500 ZTOZSB(zp)), &parent, sizeof (parent));
2501 if (error)
2502 return (error);
2503
2504 /*
2505 * Cache the lookup on the parent file znode as
2506 * zp->z_xattr_parent and hold a reference. This
2507 * effectively pins the parent in memory until all
2508 * child xattr znodes have been destroyed and
2509 * release their references in zfs_inode_destroy().
2510 */
2511 error = zfs_zget(ZTOZSB(zp), parent, &check_zp);
2512 if (error)
2513 return (error);
2514
2515 rw_enter(&zp->z_xattr_lock, RW_WRITER);
2516 if (zp->z_xattr_parent == NULL)
2517 zp->z_xattr_parent = check_zp;
2518 rw_exit(&zp->z_xattr_lock);
2519 }
2520
2521 /*
2522 * fixup mode to map to xattr perms
2523 */
2524
2525 if (mode & (ACE_WRITE_DATA|ACE_APPEND_DATA)) {
2526 mode &= ~(ACE_WRITE_DATA|ACE_APPEND_DATA);
2527 mode |= ACE_WRITE_NAMED_ATTRS;
2528 }
2529
2530 if (mode & (ACE_READ_DATA|ACE_EXECUTE)) {
2531 mode &= ~(ACE_READ_DATA|ACE_EXECUTE);
2532 mode |= ACE_READ_NAMED_ATTRS;
2533 }
2534 }
2535
2536 owner = zfs_fuid_map_id(ZTOZSB(zp), zp->z_uid, cr, ZFS_OWNER);
2537 /*
2538 * Map the bits required to the standard inode flags
2539 * S_IRUSR|S_IWUSR|S_IXUSR in the needed_bits. Map the bits
2540 * mapped by working_mode (currently missing) in missing_bits.
2541 * Call secpolicy_vnode_access2() with (needed_bits & ~checkmode),
2542 * needed_bits.
2543 */
2544 needed_bits = 0;
2545
2546 working_mode = mode;
2547 if ((working_mode & (ACE_READ_ACL|ACE_READ_ATTRIBUTES)) &&
2548 owner == crgetuid(cr))
2549 working_mode &= ~(ACE_READ_ACL|ACE_READ_ATTRIBUTES);
2550
2551 if (working_mode & (ACE_READ_DATA|ACE_READ_NAMED_ATTRS|
2552 ACE_READ_ACL|ACE_READ_ATTRIBUTES|ACE_SYNCHRONIZE))
2553 needed_bits |= S_IRUSR;
2554 if (working_mode & (ACE_WRITE_DATA|ACE_WRITE_NAMED_ATTRS|
2555 ACE_APPEND_DATA|ACE_WRITE_ATTRIBUTES|ACE_SYNCHRONIZE))
2556 needed_bits |= S_IWUSR;
2557 if (working_mode & ACE_EXECUTE)
2558 needed_bits |= S_IXUSR;
2559
2560 if ((error = zfs_zaccess_common(check_zp, mode, &working_mode,
2561 &check_privs, skipaclchk, cr)) == 0) {
2562 return (secpolicy_vnode_access2(cr, ZTOI(zp), owner,
2563 needed_bits, needed_bits));
2564 }
2565
2566 if (error && !check_privs) {
2567 return (error);
2568 }
2569
2570 if (error && (flags & V_APPEND)) {
2571 error = zfs_zaccess_append(zp, &working_mode, &check_privs, cr);
2572 }
2573
2574 if (error && check_privs) {
2575 mode_t checkmode = 0;
2576
2577 /*
2578 * First check for implicit owner permission on
2579 * read_acl/read_attributes
2580 */
2581
2582 error = 0;
2583 ASSERT(working_mode != 0);
2584
2585 if ((working_mode & (ACE_READ_ACL|ACE_READ_ATTRIBUTES) &&
2586 owner == crgetuid(cr)))
2587 working_mode &= ~(ACE_READ_ACL|ACE_READ_ATTRIBUTES);
2588
2589 if (working_mode & (ACE_READ_DATA|ACE_READ_NAMED_ATTRS|
2590 ACE_READ_ACL|ACE_READ_ATTRIBUTES|ACE_SYNCHRONIZE))
2591 checkmode |= S_IRUSR;
2592 if (working_mode & (ACE_WRITE_DATA|ACE_WRITE_NAMED_ATTRS|
2593 ACE_APPEND_DATA|ACE_WRITE_ATTRIBUTES|ACE_SYNCHRONIZE))
2594 checkmode |= S_IWUSR;
2595 if (working_mode & ACE_EXECUTE)
2596 checkmode |= S_IXUSR;
2597
2598 error = secpolicy_vnode_access2(cr, ZTOI(check_zp), owner,
2599 needed_bits & ~checkmode, needed_bits);
2600
2601 if (error == 0 && (working_mode & ACE_WRITE_OWNER))
2602 error = secpolicy_vnode_chown(cr, owner);
2603 if (error == 0 && (working_mode & ACE_WRITE_ACL))
2604 error = secpolicy_vnode_setdac(cr, owner);
2605
2606 if (error == 0 && (working_mode &
2607 (ACE_DELETE|ACE_DELETE_CHILD)))
2608 error = secpolicy_vnode_remove(cr);
2609
2610 if (error == 0 && (working_mode & ACE_SYNCHRONIZE)) {
2611 error = secpolicy_vnode_chown(cr, owner);
2612 }
2613 if (error == 0) {
2614 /*
2615 * See if any bits other than those already checked
2616 * for are still present. If so then return EACCES
2617 */
2618 if (working_mode & ~(ZFS_CHECKED_MASKS)) {
2619 error = SET_ERROR(EACCES);
2620 }
2621 }
2622 } else if (error == 0) {
2623 error = secpolicy_vnode_access2(cr, ZTOI(zp), owner,
2624 needed_bits, needed_bits);
2625 }
2626
2627 return (error);
2628 }
2629
2630 /*
2631 * Translate traditional unix S_IRUSR/S_IWUSR/S_IXUSR mode into
2632 * native ACL format and call zfs_zaccess()
2633 */
2634 int
2635 zfs_zaccess_rwx(znode_t *zp, mode_t mode, int flags, cred_t *cr)
2636 {
2637 return (zfs_zaccess(zp, zfs_unix_to_v4(mode >> 6), flags, B_FALSE, cr));
2638 }
2639
2640 /*
2641 * Access function for secpolicy_vnode_setattr
2642 */
2643 int
2644 zfs_zaccess_unix(znode_t *zp, mode_t mode, cred_t *cr)
2645 {
2646 int v4_mode = zfs_unix_to_v4(mode >> 6);
2647
2648 return (zfs_zaccess(zp, v4_mode, 0, B_FALSE, cr));
2649 }
2650
2651 static int
2652 zfs_delete_final_check(znode_t *zp, znode_t *dzp,
2653 mode_t available_perms, cred_t *cr)
2654 {
2655 int error;
2656 uid_t downer;
2657
2658 downer = zfs_fuid_map_id(ZTOZSB(dzp), dzp->z_uid, cr, ZFS_OWNER);
2659
2660 error = secpolicy_vnode_access2(cr, ZTOI(dzp),
2661 downer, available_perms, S_IWUSR|S_IXUSR);
2662
2663 if (error == 0)
2664 error = zfs_sticky_remove_access(dzp, zp, cr);
2665
2666 return (error);
2667 }
2668
2669 /*
2670 * Determine whether Access should be granted/deny, without
2671 * consulting least priv subsystem.
2672 *
2673 * The following chart is the recommended NFSv4 enforcement for
2674 * ability to delete an object.
2675 *
2676 * -------------------------------------------------------
2677 * | Parent Dir | Target Object Permissions |
2678 * | permissions | |
2679 * -------------------------------------------------------
2680 * | | ACL Allows | ACL Denies| Delete |
2681 * | | Delete | Delete | unspecified|
2682 * -------------------------------------------------------
2683 * | ACL Allows | Permit | Permit | Permit |
2684 * | DELETE_CHILD | |
2685 * -------------------------------------------------------
2686 * | ACL Denies | Permit | Deny | Deny |
2687 * | DELETE_CHILD | | | |
2688 * -------------------------------------------------------
2689 * | ACL specifies | | | |
2690 * | only allow | Permit | Permit | Permit |
2691 * | write and | | | |
2692 * | execute | | | |
2693 * -------------------------------------------------------
2694 * | ACL denies | | | |
2695 * | write and | Permit | Deny | Deny |
2696 * | execute | | | |
2697 * -------------------------------------------------------
2698 * ^
2699 * |
2700 * No search privilege, can't even look up file?
2701 *
2702 */
2703 int
2704 zfs_zaccess_delete(znode_t *dzp, znode_t *zp, cred_t *cr)
2705 {
2706 uint32_t dzp_working_mode = 0;
2707 uint32_t zp_working_mode = 0;
2708 int dzp_error, zp_error;
2709 mode_t available_perms;
2710 boolean_t dzpcheck_privs = B_TRUE;
2711 boolean_t zpcheck_privs = B_TRUE;
2712
2713 /*
2714 * We want specific DELETE permissions to
2715 * take precedence over WRITE/EXECUTE. We don't
2716 * want an ACL such as this to mess us up.
2717 * user:joe:write_data:deny,user:joe:delete:allow
2718 *
2719 * However, deny permissions may ultimately be overridden
2720 * by secpolicy_vnode_access().
2721 *
2722 * We will ask for all of the necessary permissions and then
2723 * look at the working modes from the directory and target object
2724 * to determine what was found.
2725 */
2726
2727 if (zp->z_pflags & (ZFS_IMMUTABLE | ZFS_NOUNLINK))
2728 return (SET_ERROR(EPERM));
2729
2730 /*
2731 * First row
2732 * If the directory permissions allow the delete, we are done.
2733 */
2734 if ((dzp_error = zfs_zaccess_common(dzp, ACE_DELETE_CHILD,
2735 &dzp_working_mode, &dzpcheck_privs, B_FALSE, cr)) == 0)
2736 return (0);
2737
2738 /*
2739 * If target object has delete permission then we are done
2740 */
2741 if ((zp_error = zfs_zaccess_common(zp, ACE_DELETE, &zp_working_mode,
2742 &zpcheck_privs, B_FALSE, cr)) == 0)
2743 return (0);
2744
2745 ASSERT(dzp_error && zp_error);
2746
2747 if (!dzpcheck_privs)
2748 return (dzp_error);
2749 if (!zpcheck_privs)
2750 return (zp_error);
2751
2752 /*
2753 * Second row
2754 *
2755 * If directory returns EACCES then delete_child was denied
2756 * due to deny delete_child. In this case send the request through
2757 * secpolicy_vnode_remove(). We don't use zfs_delete_final_check()
2758 * since that *could* allow the delete based on write/execute permission
2759 * and we want delete permissions to override write/execute.
2760 */
2761
2762 if (dzp_error == EACCES)
2763 return (secpolicy_vnode_remove(cr));
2764
2765 /*
2766 * Third Row
2767 * only need to see if we have write/execute on directory.
2768 */
2769
2770 dzp_error = zfs_zaccess_common(dzp, ACE_EXECUTE|ACE_WRITE_DATA,
2771 &dzp_working_mode, &dzpcheck_privs, B_FALSE, cr);
2772
2773 if (dzp_error != 0 && !dzpcheck_privs)
2774 return (dzp_error);
2775
2776 /*
2777 * Fourth row
2778 */
2779
2780 available_perms = (dzp_working_mode & ACE_WRITE_DATA) ? 0 : S_IWUSR;
2781 available_perms |= (dzp_working_mode & ACE_EXECUTE) ? 0 : S_IXUSR;
2782
2783 return (zfs_delete_final_check(zp, dzp, available_perms, cr));
2784
2785 }
2786
2787 int
2788 zfs_zaccess_rename(znode_t *sdzp, znode_t *szp, znode_t *tdzp,
2789 znode_t *tzp, cred_t *cr)
2790 {
2791 int add_perm;
2792 int error;
2793
2794 if (szp->z_pflags & ZFS_AV_QUARANTINED)
2795 return (SET_ERROR(EACCES));
2796
2797 add_perm = S_ISDIR(ZTOI(szp)->i_mode) ?
2798 ACE_ADD_SUBDIRECTORY : ACE_ADD_FILE;
2799
2800 /*
2801 * Rename permissions are combination of delete permission +
2802 * add file/subdir permission.
2803 */
2804
2805 /*
2806 * first make sure we do the delete portion.
2807 *
2808 * If that succeeds then check for add_file/add_subdir permissions
2809 */
2810
2811 if ((error = zfs_zaccess_delete(sdzp, szp, cr)))
2812 return (error);
2813
2814 /*
2815 * If we have a tzp, see if we can delete it?
2816 */
2817 if (tzp) {
2818 if ((error = zfs_zaccess_delete(tdzp, tzp, cr)))
2819 return (error);
2820 }
2821
2822 /*
2823 * Now check for add permissions
2824 */
2825 error = zfs_zaccess(tdzp, add_perm, 0, B_FALSE, cr);
2826
2827 return (error);
2828 }
ZFS On Linux mirror for PVE