4 * The contents of this file are subject to the terms of the
5 * Common Development and Distribution License (the "License").
6 * You may not use this file except in compliance with the License.
8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9 * or https://opensource.org/licenses/CDDL-1.0.
10 * See the License for the specific language governing permissions
11 * and limitations under the License.
13 * When distributing Covered Code, include this CDDL HEADER in each
14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15 * If applicable, add the following below this CDDL HEADER, with the
16 * fields enclosed by brackets "[]" replaced with your own identifying
17 * information: Portions Copyright [yyyy] [name of copyright owner]
22 * Copyright (c) 2007, 2010, Oracle and/or its affiliates. All rights reserved.
25 #include <sys/zfs_context.h>
29 #include <sys/nvpair.h>
32 #include <sys/zfs_vfsops.h>
33 #include <sys/zfs_znode.h>
35 #include <sys/zfs_fuid.h>
38 * FUID Domain table(s).
40 * The FUID table is stored as a packed nvlist of an array
41 * of nvlists which contain an index, domain string and offset
43 * During file system initialization the nvlist(s) are read and
44 * two AVL trees are created. One tree is keyed by the index number
45 * and the other by the domain string. Nodes are never removed from
46 * trees, but new entries may be added. If a new entry is added then
47 * the zfsvfs->z_fuid_dirty flag is set to true and the caller will then
48 * be responsible for calling zfs_fuid_sync() to sync the changes to disk.
52 #define FUID_IDX "fuid_idx"
53 #define FUID_DOMAIN "fuid_domain"
54 #define FUID_OFFSET "fuid_offset"
55 #define FUID_NVP_ARRAY "fuid_nvlist"
57 typedef struct fuid_domain {
64 static const char *const nulldomain = "";
67 * Compare two indexes.
70 idx_compare(const void *arg1, const void *arg2)
72 const fuid_domain_t *node1 = (const fuid_domain_t *)arg1;
73 const fuid_domain_t *node2 = (const fuid_domain_t *)arg2;
75 return (TREE_CMP(node1->f_idx, node2->f_idx));
79 * Compare two domain strings.
82 domain_compare(const void *arg1, const void *arg2)
84 const fuid_domain_t *node1 = (const fuid_domain_t *)arg1;
85 const fuid_domain_t *node2 = (const fuid_domain_t *)arg2;
88 val = strcmp(node1->f_ksid->kd_name, node2->f_ksid->kd_name);
90 return (TREE_ISIGN(val));
94 zfs_fuid_avl_tree_create(avl_tree_t *idx_tree, avl_tree_t *domain_tree)
96 avl_create(idx_tree, idx_compare,
97 sizeof (fuid_domain_t), offsetof(fuid_domain_t, f_idxnode));
98 avl_create(domain_tree, domain_compare,
99 sizeof (fuid_domain_t), offsetof(fuid_domain_t, f_domnode));
103 * load initial fuid domain and idx trees. This function is used by
104 * both the kernel and zdb.
107 zfs_fuid_table_load(objset_t *os, uint64_t fuid_obj, avl_tree_t *idx_tree,
108 avl_tree_t *domain_tree)
113 ASSERT(fuid_obj != 0);
114 VERIFY(0 == dmu_bonus_hold(os, fuid_obj,
116 fuid_size = *(uint64_t *)db->db_data;
117 dmu_buf_rele(db, FTAG);
121 nvlist_t *nvp = NULL;
126 packed = kmem_alloc(fuid_size, KM_SLEEP);
127 VERIFY(dmu_read(os, fuid_obj, 0,
128 fuid_size, packed, DMU_READ_PREFETCH) == 0);
129 VERIFY(nvlist_unpack(packed, fuid_size,
131 VERIFY(nvlist_lookup_nvlist_array(nvp, FUID_NVP_ARRAY,
132 &fuidnvp, &count) == 0);
134 for (i = 0; i != count; i++) {
135 fuid_domain_t *domnode;
139 VERIFY(nvlist_lookup_string(fuidnvp[i], FUID_DOMAIN,
141 VERIFY(nvlist_lookup_uint64(fuidnvp[i], FUID_IDX,
144 domnode = kmem_alloc(sizeof (fuid_domain_t), KM_SLEEP);
146 domnode->f_idx = idx;
147 domnode->f_ksid = ksid_lookupdomain(domain);
148 avl_add(idx_tree, domnode);
149 avl_add(domain_tree, domnode);
152 kmem_free(packed, fuid_size);
158 zfs_fuid_table_destroy(avl_tree_t *idx_tree, avl_tree_t *domain_tree)
160 fuid_domain_t *domnode;
164 while ((domnode = avl_destroy_nodes(domain_tree, &cookie)))
165 ksiddomain_rele(domnode->f_ksid);
167 avl_destroy(domain_tree);
169 while ((domnode = avl_destroy_nodes(idx_tree, &cookie)))
170 kmem_free(domnode, sizeof (fuid_domain_t));
171 avl_destroy(idx_tree);
175 zfs_fuid_idx_domain(avl_tree_t *idx_tree, uint32_t idx)
177 fuid_domain_t searchnode, *findnode;
180 searchnode.f_idx = idx;
182 findnode = avl_find(idx_tree, &searchnode, &loc);
184 return (findnode ? findnode->f_ksid->kd_name : nulldomain);
189 * Load the fuid table(s) into memory.
192 zfs_fuid_init(zfsvfs_t *zfsvfs)
194 rw_enter(&zfsvfs->z_fuid_lock, RW_WRITER);
196 if (zfsvfs->z_fuid_loaded) {
197 rw_exit(&zfsvfs->z_fuid_lock);
201 zfs_fuid_avl_tree_create(&zfsvfs->z_fuid_idx, &zfsvfs->z_fuid_domain);
203 (void) zap_lookup(zfsvfs->z_os, MASTER_NODE_OBJ,
204 ZFS_FUID_TABLES, 8, 1, &zfsvfs->z_fuid_obj);
205 if (zfsvfs->z_fuid_obj != 0) {
206 zfsvfs->z_fuid_size = zfs_fuid_table_load(zfsvfs->z_os,
207 zfsvfs->z_fuid_obj, &zfsvfs->z_fuid_idx,
208 &zfsvfs->z_fuid_domain);
211 zfsvfs->z_fuid_loaded = B_TRUE;
212 rw_exit(&zfsvfs->z_fuid_lock);
216 * sync out AVL trees to persistent storage.
219 zfs_fuid_sync(zfsvfs_t *zfsvfs, dmu_tx_t *tx)
226 fuid_domain_t *domnode;
230 if (!zfsvfs->z_fuid_dirty) {
234 rw_enter(&zfsvfs->z_fuid_lock, RW_WRITER);
237 * First see if table needs to be created?
239 if (zfsvfs->z_fuid_obj == 0) {
240 zfsvfs->z_fuid_obj = dmu_object_alloc(zfsvfs->z_os,
241 DMU_OT_FUID, 1 << 14, DMU_OT_FUID_SIZE,
242 sizeof (uint64_t), tx);
243 VERIFY(zap_add(zfsvfs->z_os, MASTER_NODE_OBJ,
244 ZFS_FUID_TABLES, sizeof (uint64_t), 1,
245 &zfsvfs->z_fuid_obj, tx) == 0);
248 VERIFY(nvlist_alloc(&nvp, NV_UNIQUE_NAME, KM_SLEEP) == 0);
250 numnodes = avl_numnodes(&zfsvfs->z_fuid_idx);
251 fuids = kmem_alloc(numnodes * sizeof (void *), KM_SLEEP);
252 for (i = 0, domnode = avl_first(&zfsvfs->z_fuid_domain); domnode; i++,
253 domnode = AVL_NEXT(&zfsvfs->z_fuid_domain, domnode)) {
254 VERIFY(nvlist_alloc(&fuids[i], NV_UNIQUE_NAME, KM_SLEEP) == 0);
255 VERIFY(nvlist_add_uint64(fuids[i], FUID_IDX,
256 domnode->f_idx) == 0);
257 VERIFY(nvlist_add_uint64(fuids[i], FUID_OFFSET, 0) == 0);
258 VERIFY(nvlist_add_string(fuids[i], FUID_DOMAIN,
259 domnode->f_ksid->kd_name) == 0);
261 fnvlist_add_nvlist_array(nvp, FUID_NVP_ARRAY,
262 (const nvlist_t * const *)fuids, numnodes);
263 for (i = 0; i != numnodes; i++)
264 nvlist_free(fuids[i]);
265 kmem_free(fuids, numnodes * sizeof (void *));
266 VERIFY(nvlist_size(nvp, &nvsize, NV_ENCODE_XDR) == 0);
267 packed = kmem_alloc(nvsize, KM_SLEEP);
268 VERIFY(nvlist_pack(nvp, &packed, &nvsize,
269 NV_ENCODE_XDR, KM_SLEEP) == 0);
271 zfsvfs->z_fuid_size = nvsize;
272 dmu_write(zfsvfs->z_os, zfsvfs->z_fuid_obj, 0,
273 zfsvfs->z_fuid_size, packed, tx);
274 kmem_free(packed, zfsvfs->z_fuid_size);
275 VERIFY(0 == dmu_bonus_hold(zfsvfs->z_os, zfsvfs->z_fuid_obj,
277 dmu_buf_will_dirty(db, tx);
278 *(uint64_t *)db->db_data = zfsvfs->z_fuid_size;
279 dmu_buf_rele(db, FTAG);
281 zfsvfs->z_fuid_dirty = B_FALSE;
282 rw_exit(&zfsvfs->z_fuid_lock);
286 * Query domain table for a given domain.
288 * If domain isn't found and addok is set, it is added to AVL trees and
289 * the zfsvfs->z_fuid_dirty flag will be set to TRUE. It will then be
290 * necessary for the caller or another thread to detect the dirty table
291 * and sync out the changes.
294 zfs_fuid_find_by_domain(zfsvfs_t *zfsvfs, const char *domain,
295 const char **retdomain, boolean_t addok)
297 fuid_domain_t searchnode, *findnode;
299 krw_t rw = RW_READER;
302 * If the dummy "nobody" domain then return an index of 0
303 * to cause the created FUID to be a standard POSIX id
304 * for the user nobody.
306 if (domain[0] == '\0') {
308 *retdomain = nulldomain;
312 searchnode.f_ksid = ksid_lookupdomain(domain);
314 *retdomain = searchnode.f_ksid->kd_name;
315 if (!zfsvfs->z_fuid_loaded)
316 zfs_fuid_init(zfsvfs);
319 rw_enter(&zfsvfs->z_fuid_lock, rw);
320 findnode = avl_find(&zfsvfs->z_fuid_domain, &searchnode, &loc);
323 rw_exit(&zfsvfs->z_fuid_lock);
324 ksiddomain_rele(searchnode.f_ksid);
325 return (findnode->f_idx);
327 fuid_domain_t *domnode;
330 if (rw == RW_READER && !rw_tryupgrade(&zfsvfs->z_fuid_lock)) {
331 rw_exit(&zfsvfs->z_fuid_lock);
336 domnode = kmem_alloc(sizeof (fuid_domain_t), KM_SLEEP);
337 domnode->f_ksid = searchnode.f_ksid;
339 retidx = domnode->f_idx = avl_numnodes(&zfsvfs->z_fuid_idx) + 1;
341 avl_add(&zfsvfs->z_fuid_domain, domnode);
342 avl_add(&zfsvfs->z_fuid_idx, domnode);
343 zfsvfs->z_fuid_dirty = B_TRUE;
344 rw_exit(&zfsvfs->z_fuid_lock);
347 rw_exit(&zfsvfs->z_fuid_lock);
353 * Query domain table by index, returning domain string
355 * Returns a pointer from an avl node of the domain string.
359 zfs_fuid_find_by_idx(zfsvfs_t *zfsvfs, uint32_t idx)
363 if (idx == 0 || !zfsvfs->z_use_fuids)
366 if (!zfsvfs->z_fuid_loaded)
367 zfs_fuid_init(zfsvfs);
369 rw_enter(&zfsvfs->z_fuid_lock, RW_READER);
371 if (zfsvfs->z_fuid_obj || zfsvfs->z_fuid_dirty)
372 domain = zfs_fuid_idx_domain(&zfsvfs->z_fuid_idx, idx);
375 rw_exit(&zfsvfs->z_fuid_lock);
382 zfs_fuid_map_ids(znode_t *zp, cred_t *cr, uid_t *uidp, uid_t *gidp)
384 *uidp = zfs_fuid_map_id(ZTOZSB(zp), KUID_TO_SUID(ZTOUID(zp)),
386 *gidp = zfs_fuid_map_id(ZTOZSB(zp), KGID_TO_SGID(ZTOGID(zp)),
392 zfs_fuid_map_id(zfsvfs_t *zfsvfs, uint64_t fuid,
393 cred_t *cr, zfs_fuid_type_t type)
395 uint32_t index = FUID_INDEX(fuid);
402 #elif defined(__linux__)
404 zfs_fuid_map_id(zfsvfs_t *zfsvfs, uint64_t fuid,
405 cred_t *cr, zfs_fuid_type_t type)
408 * The Linux port only supports POSIX IDs, use the passed id.
415 zfs_fuid_map_id(zfsvfs_t *zfsvfs, uint64_t fuid,
416 cred_t *cr, zfs_fuid_type_t type)
418 uint32_t index = FUID_INDEX(fuid);
425 domain = zfs_fuid_find_by_idx(zfsvfs, index);
426 ASSERT(domain != NULL);
428 if (type == ZFS_OWNER || type == ZFS_ACE_USER) {
429 (void) kidmap_getuidbysid(crgetzone(cr), domain,
430 FUID_RID(fuid), &id);
432 (void) kidmap_getgidbysid(crgetzone(cr), domain,
433 FUID_RID(fuid), &id);
440 * Add a FUID node to the list of fuid's being created for this
443 * If ACL has multiple domains, then keep only one copy of each unique
447 zfs_fuid_node_add(zfs_fuid_info_t **fuidpp, const char *domain, uint32_t rid,
448 uint64_t idx, uint64_t id, zfs_fuid_type_t type)
451 zfs_fuid_domain_t *fuid_domain;
452 zfs_fuid_info_t *fuidp;
454 boolean_t found = B_FALSE;
457 *fuidpp = zfs_fuid_info_alloc();
461 * First find fuid domain index in linked list
463 * If one isn't found then create an entry.
466 for (fuididx = 1, fuid_domain = list_head(&fuidp->z_domains);
467 fuid_domain; fuid_domain = list_next(&fuidp->z_domains,
468 fuid_domain), fuididx++) {
469 if (idx == fuid_domain->z_domidx) {
476 fuid_domain = kmem_alloc(sizeof (zfs_fuid_domain_t), KM_SLEEP);
477 fuid_domain->z_domain = domain;
478 fuid_domain->z_domidx = idx;
479 list_insert_tail(&fuidp->z_domains, fuid_domain);
480 fuidp->z_domain_str_sz += strlen(domain) + 1;
481 fuidp->z_domain_cnt++;
484 if (type == ZFS_ACE_USER || type == ZFS_ACE_GROUP) {
487 * Now allocate fuid entry and add it on the end of the list
490 fuid = kmem_alloc(sizeof (zfs_fuid_t), KM_SLEEP);
492 fuid->z_domidx = idx;
493 fuid->z_logfuid = FUID_ENCODE(fuididx, rid);
495 list_insert_tail(&fuidp->z_fuids, fuid);
498 if (type == ZFS_OWNER)
499 fuidp->z_fuid_owner = FUID_ENCODE(fuididx, rid);
501 fuidp->z_fuid_group = FUID_ENCODE(fuididx, rid);
507 * Create a file system FUID, based on information in the users cred
509 * If cred contains KSID_OWNER then it should be used to determine
510 * the uid otherwise cred's uid will be used. By default cred's gid
511 * is used unless it's an ephemeral ID in which case KSID_GROUP will
512 * be used if it exists.
515 zfs_fuid_create_cred(zfsvfs_t *zfsvfs, zfs_fuid_type_t type,
516 cred_t *cr, zfs_fuid_info_t **fuidp)
521 const char *kdomain, *domain;
524 VERIFY(type == ZFS_OWNER || type == ZFS_GROUP);
526 ksid = crgetsid(cr, (type == ZFS_OWNER) ? KSID_OWNER : KSID_GROUP);
528 if (!zfsvfs->z_use_fuids || (ksid == NULL)) {
529 id = (type == ZFS_OWNER) ? crgetuid(cr) : crgetgid(cr);
531 if (IS_EPHEMERAL(id))
532 return ((type == ZFS_OWNER) ? UID_NOBODY : GID_NOBODY);
534 return ((uint64_t)id);
538 * ksid is present and FUID is supported
540 id = (type == ZFS_OWNER) ? ksid_getid(ksid) : crgetgid(cr);
542 if (!IS_EPHEMERAL(id))
543 return ((uint64_t)id);
545 if (type == ZFS_GROUP)
546 id = ksid_getid(ksid);
548 rid = ksid_getrid(ksid);
549 domain = ksid_getdomain(ksid);
551 idx = zfs_fuid_find_by_domain(zfsvfs, domain, &kdomain, B_TRUE);
553 zfs_fuid_node_add(fuidp, kdomain, rid, idx, id, type);
555 return (FUID_ENCODE(idx, rid));
557 #endif /* HAVE_KSID */
560 * Create a file system FUID for an ACL ace
561 * or a chown/chgrp of the file.
562 * This is similar to zfs_fuid_create_cred, except that
563 * we can't find the domain + rid information in the
564 * cred. Instead we have to query Winchester for the
567 * During replay operations the domain+rid information is
568 * found in the zfs_fuid_info_t that the replay code has
569 * attached to the zfsvfs of the file system.
572 zfs_fuid_create(zfsvfs_t *zfsvfs, uint64_t id, cred_t *cr,
573 zfs_fuid_type_t type, zfs_fuid_info_t **fuidpp)
576 const char *domain, *kdomain;
577 uint32_t fuid_idx = FUID_INDEX(id);
580 uint64_t idx = UID_NOBODY;
581 zfs_fuid_t *zfuid = NULL;
582 zfs_fuid_info_t *fuidp = NULL;
585 * If POSIX ID, or entry is already a FUID then
588 * We may also be handed an already FUID'ized id via
592 if (!zfsvfs->z_use_fuids || !IS_EPHEMERAL(id) || fuid_idx != 0)
595 if (zfsvfs->z_replay) {
596 fuidp = zfsvfs->z_fuid_replay;
599 * If we are passed an ephemeral id, but no
600 * fuid_info was logged then return NOBODY.
601 * This is most likely a result of idmap service
602 * not being available.
607 VERIFY3U(type, >=, ZFS_OWNER);
608 VERIFY3U(type, <=, ZFS_ACE_GROUP);
613 zfuid = list_head(&fuidp->z_fuids);
614 rid = FUID_RID(zfuid->z_logfuid);
615 idx = FUID_INDEX(zfuid->z_logfuid);
618 rid = FUID_RID(fuidp->z_fuid_owner);
619 idx = FUID_INDEX(fuidp->z_fuid_owner);
622 rid = FUID_RID(fuidp->z_fuid_group);
623 idx = FUID_INDEX(fuidp->z_fuid_group);
626 domain = fuidp->z_domain_table[idx - 1];
628 if (type == ZFS_OWNER || type == ZFS_ACE_USER)
629 status = kidmap_getsidbyuid(crgetzone(cr), id,
632 status = kidmap_getsidbygid(crgetzone(cr), id,
637 * When returning nobody we will need to
638 * make a dummy fuid table entry for logging
646 idx = zfs_fuid_find_by_domain(zfsvfs, domain, &kdomain, B_TRUE);
648 if (!zfsvfs->z_replay)
649 zfs_fuid_node_add(fuidpp, kdomain,
651 else if (zfuid != NULL) {
652 list_remove(&fuidp->z_fuids, zfuid);
653 kmem_free(zfuid, sizeof (zfs_fuid_t));
655 return (FUID_ENCODE(idx, rid));
658 * The Linux port only supports POSIX IDs, use the passed id.
665 zfs_fuid_destroy(zfsvfs_t *zfsvfs)
667 rw_enter(&zfsvfs->z_fuid_lock, RW_WRITER);
668 if (!zfsvfs->z_fuid_loaded) {
669 rw_exit(&zfsvfs->z_fuid_lock);
672 zfs_fuid_table_destroy(&zfsvfs->z_fuid_idx, &zfsvfs->z_fuid_domain);
673 rw_exit(&zfsvfs->z_fuid_lock);
677 * Allocate zfs_fuid_info for tracking FUIDs created during
678 * zfs_mknode, VOP_SETATTR() or VOP_SETSECATTR()
681 zfs_fuid_info_alloc(void)
683 zfs_fuid_info_t *fuidp;
685 fuidp = kmem_zalloc(sizeof (zfs_fuid_info_t), KM_SLEEP);
686 list_create(&fuidp->z_domains, sizeof (zfs_fuid_domain_t),
687 offsetof(zfs_fuid_domain_t, z_next));
688 list_create(&fuidp->z_fuids, sizeof (zfs_fuid_t),
689 offsetof(zfs_fuid_t, z_next));
694 * Release all memory associated with zfs_fuid_info_t
697 zfs_fuid_info_free(zfs_fuid_info_t *fuidp)
700 zfs_fuid_domain_t *zdomain;
702 while ((zfuid = list_remove_head(&fuidp->z_fuids)) != NULL)
703 kmem_free(zfuid, sizeof (zfs_fuid_t));
705 if (fuidp->z_domain_table != NULL)
706 kmem_free(fuidp->z_domain_table,
707 (sizeof (char *)) * fuidp->z_domain_cnt);
709 while ((zdomain = list_remove_head(&fuidp->z_domains)) != NULL)
710 kmem_free(zdomain, sizeof (zfs_fuid_domain_t));
712 kmem_free(fuidp, sizeof (zfs_fuid_info_t));
716 * Check to see if id is a groupmember. If cred
717 * has ksid info then sidlist is checked first
718 * and if still not found then POSIX groups are checked
720 * Will use a straight FUID compare when possible.
723 zfs_groupmember(zfsvfs_t *zfsvfs, uint64_t id, cred_t *cr)
728 ksid_t *ksid = crgetsid(cr, KSID_GROUP);
729 ksidlist_t *ksidlist = crgetsidlist(cr);
731 if (ksid && ksidlist) {
734 uint32_t idx = FUID_INDEX(id);
735 uint32_t rid = FUID_RID(id);
737 ksid_groups = ksidlist->ksl_sids;
739 for (i = 0; i != ksidlist->ksl_nsid; i++) {
741 if (id != IDMAP_WK_CREATOR_GROUP_GID &&
742 id == ksid_groups[i].ks_id) {
748 domain = zfs_fuid_find_by_idx(zfsvfs, idx);
749 ASSERT(domain != NULL);
752 IDMAP_WK_CREATOR_SID_AUTHORITY) == 0)
756 ksid_groups[i].ks_domain->kd_name) == 0) &&
757 rid == ksid_groups[i].ks_rid)
765 * Not found in ksidlist, check posix groups
767 gid = zfs_fuid_map_id(zfsvfs, id, cr, ZFS_GROUP);
768 return (groupmember(gid, cr));
772 zfs_fuid_txhold(zfsvfs_t *zfsvfs, dmu_tx_t *tx)
774 if (zfsvfs->z_fuid_obj == 0) {
775 dmu_tx_hold_bonus(tx, DMU_NEW_OBJECT);
776 dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0,
777 FUID_SIZE_ESTIMATE(zfsvfs));
778 dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, FALSE, NULL);
780 dmu_tx_hold_bonus(tx, zfsvfs->z_fuid_obj);
781 dmu_tx_hold_write(tx, zfsvfs->z_fuid_obj, 0,
782 FUID_SIZE_ESTIMATE(zfsvfs));
787 * buf must be big enough (eg, 32 bytes)
790 zfs_id_to_fuidstr(zfsvfs_t *zfsvfs, const char *domain, uid_t rid,
791 char *buf, size_t len, boolean_t addok)
796 if (domain && domain[0]) {
797 domainid = zfs_fuid_find_by_domain(zfsvfs, domain, NULL, addok);
799 return (SET_ERROR(ENOENT));
801 fuid = FUID_ENCODE(domainid, rid);
802 (void) snprintf(buf, len, "%llx", (longlong_t)fuid);