4 * The contents of this file are subject to the terms of the
5 * Common Development and Distribution License (the "License").
6 * You may not use this file except in compliance with the License.
8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9 * or http://www.opensolaris.org/os/licensing.
10 * See the License for the specific language governing permissions
11 * and limitations under the License.
13 * When distributing Covered Code, include this CDDL HEADER in each
14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15 * If applicable, add the following below this CDDL HEADER, with the
16 * fields enclosed by brackets "[]" replaced with your own identifying
17 * information: Portions Copyright [yyyy] [name of copyright owner]
22 * Copyright (c) 2007, 2010, Oracle and/or its affiliates. All rights reserved.
25 #include <sys/zfs_context.h>
29 #include <sys/refcount.h>
30 #include <sys/nvpair.h>
33 #include <sys/zfs_vfsops.h>
34 #include <sys/zfs_znode.h>
36 #include <sys/zfs_fuid.h>
39 * FUID Domain table(s).
41 * The FUID table is stored as a packed nvlist of an array
42 * of nvlists which contain an index, domain string and offset
44 * During file system initialization the nvlist(s) are read and
45 * two AVL trees are created. One tree is keyed by the index number
46 * and the other by the domain string. Nodes are never removed from
47 * trees, but new entries may be added. If a new entry is added then
48 * the zfsvfs->z_fuid_dirty flag is set to true and the caller will then
49 * be responsible for calling zfs_fuid_sync() to sync the changes to disk.
53 #define FUID_IDX "fuid_idx"
54 #define FUID_DOMAIN "fuid_domain"
55 #define FUID_OFFSET "fuid_offset"
56 #define FUID_NVP_ARRAY "fuid_nvlist"
58 typedef struct fuid_domain {
65 static char *nulldomain = "";
68 * Compare two indexes.
71 idx_compare(const void *arg1, const void *arg2)
73 const fuid_domain_t *node1 = (const fuid_domain_t *)arg1;
74 const fuid_domain_t *node2 = (const fuid_domain_t *)arg2;
76 return (AVL_CMP(node1->f_idx, node2->f_idx));
80 * Compare two domain strings.
83 domain_compare(const void *arg1, const void *arg2)
85 const fuid_domain_t *node1 = (const fuid_domain_t *)arg1;
86 const fuid_domain_t *node2 = (const fuid_domain_t *)arg2;
89 val = strcmp(node1->f_ksid->kd_name, node2->f_ksid->kd_name);
91 return (AVL_ISIGN(val));
95 zfs_fuid_avl_tree_create(avl_tree_t *idx_tree, avl_tree_t *domain_tree)
97 avl_create(idx_tree, idx_compare,
98 sizeof (fuid_domain_t), offsetof(fuid_domain_t, f_idxnode));
99 avl_create(domain_tree, domain_compare,
100 sizeof (fuid_domain_t), offsetof(fuid_domain_t, f_domnode));
104 * load initial fuid domain and idx trees. This function is used by
105 * both the kernel and zdb.
108 zfs_fuid_table_load(objset_t *os, uint64_t fuid_obj, avl_tree_t *idx_tree,
109 avl_tree_t *domain_tree)
114 ASSERT(fuid_obj != 0);
115 VERIFY(0 == dmu_bonus_hold(os, fuid_obj,
117 fuid_size = *(uint64_t *)db->db_data;
118 dmu_buf_rele(db, FTAG);
122 nvlist_t *nvp = NULL;
127 packed = kmem_alloc(fuid_size, KM_SLEEP);
128 VERIFY(dmu_read(os, fuid_obj, 0,
129 fuid_size, packed, DMU_READ_PREFETCH) == 0);
130 VERIFY(nvlist_unpack(packed, fuid_size,
132 VERIFY(nvlist_lookup_nvlist_array(nvp, FUID_NVP_ARRAY,
133 &fuidnvp, &count) == 0);
135 for (i = 0; i != count; i++) {
136 fuid_domain_t *domnode;
140 VERIFY(nvlist_lookup_string(fuidnvp[i], FUID_DOMAIN,
142 VERIFY(nvlist_lookup_uint64(fuidnvp[i], FUID_IDX,
145 domnode = kmem_alloc(sizeof (fuid_domain_t), KM_SLEEP);
147 domnode->f_idx = idx;
148 domnode->f_ksid = ksid_lookupdomain(domain);
149 avl_add(idx_tree, domnode);
150 avl_add(domain_tree, domnode);
153 kmem_free(packed, fuid_size);
159 zfs_fuid_table_destroy(avl_tree_t *idx_tree, avl_tree_t *domain_tree)
161 fuid_domain_t *domnode;
165 while ((domnode = avl_destroy_nodes(domain_tree, &cookie)))
166 ksiddomain_rele(domnode->f_ksid);
168 avl_destroy(domain_tree);
170 while ((domnode = avl_destroy_nodes(idx_tree, &cookie)))
171 kmem_free(domnode, sizeof (fuid_domain_t));
172 avl_destroy(idx_tree);
176 zfs_fuid_idx_domain(avl_tree_t *idx_tree, uint32_t idx)
178 fuid_domain_t searchnode, *findnode;
181 searchnode.f_idx = idx;
183 findnode = avl_find(idx_tree, &searchnode, &loc);
185 return (findnode ? findnode->f_ksid->kd_name : nulldomain);
190 * Load the fuid table(s) into memory.
193 zfs_fuid_init(zfsvfs_t *zfsvfs)
195 rw_enter(&zfsvfs->z_fuid_lock, RW_WRITER);
197 if (zfsvfs->z_fuid_loaded) {
198 rw_exit(&zfsvfs->z_fuid_lock);
202 zfs_fuid_avl_tree_create(&zfsvfs->z_fuid_idx, &zfsvfs->z_fuid_domain);
204 (void) zap_lookup(zfsvfs->z_os, MASTER_NODE_OBJ,
205 ZFS_FUID_TABLES, 8, 1, &zfsvfs->z_fuid_obj);
206 if (zfsvfs->z_fuid_obj != 0) {
207 zfsvfs->z_fuid_size = zfs_fuid_table_load(zfsvfs->z_os,
208 zfsvfs->z_fuid_obj, &zfsvfs->z_fuid_idx,
209 &zfsvfs->z_fuid_domain);
212 zfsvfs->z_fuid_loaded = B_TRUE;
213 rw_exit(&zfsvfs->z_fuid_lock);
217 * sync out AVL trees to persistent storage.
220 zfs_fuid_sync(zfsvfs_t *zfsvfs, dmu_tx_t *tx)
227 fuid_domain_t *domnode;
231 if (!zfsvfs->z_fuid_dirty) {
235 rw_enter(&zfsvfs->z_fuid_lock, RW_WRITER);
238 * First see if table needs to be created?
240 if (zfsvfs->z_fuid_obj == 0) {
241 zfsvfs->z_fuid_obj = dmu_object_alloc(zfsvfs->z_os,
242 DMU_OT_FUID, 1 << 14, DMU_OT_FUID_SIZE,
243 sizeof (uint64_t), tx);
244 VERIFY(zap_add(zfsvfs->z_os, MASTER_NODE_OBJ,
245 ZFS_FUID_TABLES, sizeof (uint64_t), 1,
246 &zfsvfs->z_fuid_obj, tx) == 0);
249 VERIFY(nvlist_alloc(&nvp, NV_UNIQUE_NAME, KM_SLEEP) == 0);
251 numnodes = avl_numnodes(&zfsvfs->z_fuid_idx);
252 fuids = kmem_alloc(numnodes * sizeof (void *), KM_SLEEP);
253 for (i = 0, domnode = avl_first(&zfsvfs->z_fuid_domain); domnode; i++,
254 domnode = AVL_NEXT(&zfsvfs->z_fuid_domain, domnode)) {
255 VERIFY(nvlist_alloc(&fuids[i], NV_UNIQUE_NAME, KM_SLEEP) == 0);
256 VERIFY(nvlist_add_uint64(fuids[i], FUID_IDX,
257 domnode->f_idx) == 0);
258 VERIFY(nvlist_add_uint64(fuids[i], FUID_OFFSET, 0) == 0);
259 VERIFY(nvlist_add_string(fuids[i], FUID_DOMAIN,
260 domnode->f_ksid->kd_name) == 0);
262 VERIFY(nvlist_add_nvlist_array(nvp, FUID_NVP_ARRAY,
263 fuids, numnodes) == 0);
264 for (i = 0; i != numnodes; i++)
265 nvlist_free(fuids[i]);
266 kmem_free(fuids, numnodes * sizeof (void *));
267 VERIFY(nvlist_size(nvp, &nvsize, NV_ENCODE_XDR) == 0);
268 packed = kmem_alloc(nvsize, KM_SLEEP);
269 VERIFY(nvlist_pack(nvp, &packed, &nvsize,
270 NV_ENCODE_XDR, KM_SLEEP) == 0);
272 zfsvfs->z_fuid_size = nvsize;
273 dmu_write(zfsvfs->z_os, zfsvfs->z_fuid_obj, 0,
274 zfsvfs->z_fuid_size, packed, tx);
275 kmem_free(packed, zfsvfs->z_fuid_size);
276 VERIFY(0 == dmu_bonus_hold(zfsvfs->z_os, zfsvfs->z_fuid_obj,
278 dmu_buf_will_dirty(db, tx);
279 *(uint64_t *)db->db_data = zfsvfs->z_fuid_size;
280 dmu_buf_rele(db, FTAG);
282 zfsvfs->z_fuid_dirty = B_FALSE;
283 rw_exit(&zfsvfs->z_fuid_lock);
287 * Query domain table for a given domain.
289 * If domain isn't found and addok is set, it is added to AVL trees and
290 * the zfsvfs->z_fuid_dirty flag will be set to TRUE. It will then be
291 * necessary for the caller or another thread to detect the dirty table
292 * and sync out the changes.
295 zfs_fuid_find_by_domain(zfsvfs_t *zfsvfs, const char *domain,
296 char **retdomain, boolean_t addok)
298 fuid_domain_t searchnode, *findnode;
300 krw_t rw = RW_READER;
303 * If the dummy "nobody" domain then return an index of 0
304 * to cause the created FUID to be a standard POSIX id
305 * for the user nobody.
307 if (domain[0] == '\0') {
309 *retdomain = nulldomain;
313 searchnode.f_ksid = ksid_lookupdomain(domain);
315 *retdomain = searchnode.f_ksid->kd_name;
316 if (!zfsvfs->z_fuid_loaded)
317 zfs_fuid_init(zfsvfs);
320 rw_enter(&zfsvfs->z_fuid_lock, rw);
321 findnode = avl_find(&zfsvfs->z_fuid_domain, &searchnode, &loc);
324 rw_exit(&zfsvfs->z_fuid_lock);
325 ksiddomain_rele(searchnode.f_ksid);
326 return (findnode->f_idx);
328 fuid_domain_t *domnode;
331 if (rw == RW_READER && !rw_tryupgrade(&zfsvfs->z_fuid_lock)) {
332 rw_exit(&zfsvfs->z_fuid_lock);
337 domnode = kmem_alloc(sizeof (fuid_domain_t), KM_SLEEP);
338 domnode->f_ksid = searchnode.f_ksid;
340 retidx = domnode->f_idx = avl_numnodes(&zfsvfs->z_fuid_idx) + 1;
342 avl_add(&zfsvfs->z_fuid_domain, domnode);
343 avl_add(&zfsvfs->z_fuid_idx, domnode);
344 zfsvfs->z_fuid_dirty = B_TRUE;
345 rw_exit(&zfsvfs->z_fuid_lock);
348 rw_exit(&zfsvfs->z_fuid_lock);
354 * Query domain table by index, returning domain string
356 * Returns a pointer from an avl node of the domain string.
360 zfs_fuid_find_by_idx(zfsvfs_t *zfsvfs, uint32_t idx)
364 if (idx == 0 || !zfsvfs->z_use_fuids)
367 if (!zfsvfs->z_fuid_loaded)
368 zfs_fuid_init(zfsvfs);
370 rw_enter(&zfsvfs->z_fuid_lock, RW_READER);
372 if (zfsvfs->z_fuid_obj || zfsvfs->z_fuid_dirty)
373 domain = zfs_fuid_idx_domain(&zfsvfs->z_fuid_idx, idx);
376 rw_exit(&zfsvfs->z_fuid_lock);
383 zfs_fuid_map_ids(znode_t *zp, cred_t *cr, uid_t *uidp, uid_t *gidp)
385 *uidp = zfs_fuid_map_id(ZTOZSB(zp), KUID_TO_SUID(ZTOI(zp)->i_uid),
387 *gidp = zfs_fuid_map_id(ZTOZSB(zp), KGID_TO_SGID(ZTOI(zp)->i_gid),
392 zfs_fuid_map_id(zfsvfs_t *zfsvfs, uint64_t fuid,
393 cred_t *cr, zfs_fuid_type_t type)
396 uint32_t index = FUID_INDEX(fuid);
403 domain = zfs_fuid_find_by_idx(zfsvfs, index);
404 ASSERT(domain != NULL);
406 if (type == ZFS_OWNER || type == ZFS_ACE_USER) {
407 (void) kidmap_getuidbysid(crgetzone(cr), domain,
408 FUID_RID(fuid), &id);
410 (void) kidmap_getgidbysid(crgetzone(cr), domain,
411 FUID_RID(fuid), &id);
416 * The Linux port only supports POSIX IDs, use the passed id.
419 #endif /* HAVE_KSID */
423 * Add a FUID node to the list of fuid's being created for this
426 * If ACL has multiple domains, then keep only one copy of each unique
430 zfs_fuid_node_add(zfs_fuid_info_t **fuidpp, const char *domain, uint32_t rid,
431 uint64_t idx, uint64_t id, zfs_fuid_type_t type)
434 zfs_fuid_domain_t *fuid_domain;
435 zfs_fuid_info_t *fuidp;
437 boolean_t found = B_FALSE;
440 *fuidpp = zfs_fuid_info_alloc();
444 * First find fuid domain index in linked list
446 * If one isn't found then create an entry.
449 for (fuididx = 1, fuid_domain = list_head(&fuidp->z_domains);
450 fuid_domain; fuid_domain = list_next(&fuidp->z_domains,
451 fuid_domain), fuididx++) {
452 if (idx == fuid_domain->z_domidx) {
459 fuid_domain = kmem_alloc(sizeof (zfs_fuid_domain_t), KM_SLEEP);
460 fuid_domain->z_domain = domain;
461 fuid_domain->z_domidx = idx;
462 list_insert_tail(&fuidp->z_domains, fuid_domain);
463 fuidp->z_domain_str_sz += strlen(domain) + 1;
464 fuidp->z_domain_cnt++;
467 if (type == ZFS_ACE_USER || type == ZFS_ACE_GROUP) {
470 * Now allocate fuid entry and add it on the end of the list
473 fuid = kmem_alloc(sizeof (zfs_fuid_t), KM_SLEEP);
475 fuid->z_domidx = idx;
476 fuid->z_logfuid = FUID_ENCODE(fuididx, rid);
478 list_insert_tail(&fuidp->z_fuids, fuid);
481 if (type == ZFS_OWNER)
482 fuidp->z_fuid_owner = FUID_ENCODE(fuididx, rid);
484 fuidp->z_fuid_group = FUID_ENCODE(fuididx, rid);
490 * Create a file system FUID, based on information in the users cred
492 * If cred contains KSID_OWNER then it should be used to determine
493 * the uid otherwise cred's uid will be used. By default cred's gid
494 * is used unless it's an ephemeral ID in which case KSID_GROUP will
495 * be used if it exists.
498 zfs_fuid_create_cred(zfsvfs_t *zfsvfs, zfs_fuid_type_t type,
499 cred_t *cr, zfs_fuid_info_t **fuidp)
508 VERIFY(type == ZFS_OWNER || type == ZFS_GROUP);
510 ksid = crgetsid(cr, (type == ZFS_OWNER) ? KSID_OWNER : KSID_GROUP);
512 if (!zfsvfs->z_use_fuids || (ksid == NULL)) {
513 id = (type == ZFS_OWNER) ? crgetuid(cr) : crgetgid(cr);
515 if (IS_EPHEMERAL(id))
516 return ((type == ZFS_OWNER) ? UID_NOBODY : GID_NOBODY);
518 return ((uint64_t)id);
522 * ksid is present and FUID is supported
524 id = (type == ZFS_OWNER) ? ksid_getid(ksid) : crgetgid(cr);
526 if (!IS_EPHEMERAL(id))
527 return ((uint64_t)id);
529 if (type == ZFS_GROUP)
530 id = ksid_getid(ksid);
532 rid = ksid_getrid(ksid);
533 domain = ksid_getdomain(ksid);
535 idx = zfs_fuid_find_by_domain(zfsvfs, domain, &kdomain, B_TRUE);
537 zfs_fuid_node_add(fuidp, kdomain, rid, idx, id, type);
539 return (FUID_ENCODE(idx, rid));
541 #endif /* HAVE_KSID */
544 * Create a file system FUID for an ACL ace
545 * or a chown/chgrp of the file.
546 * This is similar to zfs_fuid_create_cred, except that
547 * we can't find the domain + rid information in the
548 * cred. Instead we have to query Winchester for the
551 * During replay operations the domain+rid information is
552 * found in the zfs_fuid_info_t that the replay code has
553 * attached to the zfsvfs of the file system.
556 zfs_fuid_create(zfsvfs_t *zfsvfs, uint64_t id, cred_t *cr,
557 zfs_fuid_type_t type, zfs_fuid_info_t **fuidpp)
562 uint32_t fuid_idx = FUID_INDEX(id);
566 zfs_fuid_t *zfuid = NULL;
567 zfs_fuid_info_t *fuidp = NULL;
570 * If POSIX ID, or entry is already a FUID then
573 * We may also be handed an already FUID'ized id via
577 if (!zfsvfs->z_use_fuids || !IS_EPHEMERAL(id) || fuid_idx != 0)
580 if (zfsvfs->z_replay) {
581 fuidp = zfsvfs->z_fuid_replay;
584 * If we are passed an ephemeral id, but no
585 * fuid_info was logged then return NOBODY.
586 * This is most likely a result of idmap service
587 * not being available.
592 VERIFY3U(type, >=, ZFS_OWNER);
593 VERIFY3U(type, <=, ZFS_ACE_GROUP);
598 zfuid = list_head(&fuidp->z_fuids);
599 rid = FUID_RID(zfuid->z_logfuid);
600 idx = FUID_INDEX(zfuid->z_logfuid);
603 rid = FUID_RID(fuidp->z_fuid_owner);
604 idx = FUID_INDEX(fuidp->z_fuid_owner);
607 rid = FUID_RID(fuidp->z_fuid_group);
608 idx = FUID_INDEX(fuidp->z_fuid_group);
611 domain = fuidp->z_domain_table[idx - 1];
613 if (type == ZFS_OWNER || type == ZFS_ACE_USER)
614 status = kidmap_getsidbyuid(crgetzone(cr), id,
617 status = kidmap_getsidbygid(crgetzone(cr), id,
622 * When returning nobody we will need to
623 * make a dummy fuid table entry for logging
631 idx = zfs_fuid_find_by_domain(zfsvfs, domain, &kdomain, B_TRUE);
633 if (!zfsvfs->z_replay)
634 zfs_fuid_node_add(fuidpp, kdomain,
636 else if (zfuid != NULL) {
637 list_remove(&fuidp->z_fuids, zfuid);
638 kmem_free(zfuid, sizeof (zfs_fuid_t));
640 return (FUID_ENCODE(idx, rid));
643 * The Linux port only supports POSIX IDs, use the passed id.
650 zfs_fuid_destroy(zfsvfs_t *zfsvfs)
652 rw_enter(&zfsvfs->z_fuid_lock, RW_WRITER);
653 if (!zfsvfs->z_fuid_loaded) {
654 rw_exit(&zfsvfs->z_fuid_lock);
657 zfs_fuid_table_destroy(&zfsvfs->z_fuid_idx, &zfsvfs->z_fuid_domain);
658 rw_exit(&zfsvfs->z_fuid_lock);
662 * Allocate zfs_fuid_info for tracking FUIDs created during
663 * zfs_mknode, VOP_SETATTR() or VOP_SETSECATTR()
666 zfs_fuid_info_alloc(void)
668 zfs_fuid_info_t *fuidp;
670 fuidp = kmem_zalloc(sizeof (zfs_fuid_info_t), KM_SLEEP);
671 list_create(&fuidp->z_domains, sizeof (zfs_fuid_domain_t),
672 offsetof(zfs_fuid_domain_t, z_next));
673 list_create(&fuidp->z_fuids, sizeof (zfs_fuid_t),
674 offsetof(zfs_fuid_t, z_next));
679 * Release all memory associated with zfs_fuid_info_t
682 zfs_fuid_info_free(zfs_fuid_info_t *fuidp)
685 zfs_fuid_domain_t *zdomain;
687 while ((zfuid = list_head(&fuidp->z_fuids)) != NULL) {
688 list_remove(&fuidp->z_fuids, zfuid);
689 kmem_free(zfuid, sizeof (zfs_fuid_t));
692 if (fuidp->z_domain_table != NULL)
693 kmem_free(fuidp->z_domain_table,
694 (sizeof (char *)) * fuidp->z_domain_cnt);
696 while ((zdomain = list_head(&fuidp->z_domains)) != NULL) {
697 list_remove(&fuidp->z_domains, zdomain);
698 kmem_free(zdomain, sizeof (zfs_fuid_domain_t));
701 kmem_free(fuidp, sizeof (zfs_fuid_info_t));
705 * Check to see if id is a groupmember. If cred
706 * has ksid info then sidlist is checked first
707 * and if still not found then POSIX groups are checked
709 * Will use a straight FUID compare when possible.
712 zfs_groupmember(zfsvfs_t *zfsvfs, uint64_t id, cred_t *cr)
715 ksid_t *ksid = crgetsid(cr, KSID_GROUP);
716 ksidlist_t *ksidlist = crgetsidlist(cr);
719 if (ksid && ksidlist) {
722 uint32_t idx = FUID_INDEX(id);
723 uint32_t rid = FUID_RID(id);
725 ksid_groups = ksidlist->ksl_sids;
727 for (i = 0; i != ksidlist->ksl_nsid; i++) {
729 if (id != IDMAP_WK_CREATOR_GROUP_GID &&
730 id == ksid_groups[i].ks_id) {
736 domain = zfs_fuid_find_by_idx(zfsvfs, idx);
737 ASSERT(domain != NULL);
740 IDMAP_WK_CREATOR_SID_AUTHORITY) == 0)
744 ksid_groups[i].ks_domain->kd_name) == 0) &&
745 rid == ksid_groups[i].ks_rid)
752 * Not found in ksidlist, check posix groups
754 gid = zfs_fuid_map_id(zfsvfs, id, cr, ZFS_GROUP);
755 return (groupmember(gid, cr));
762 zfs_fuid_txhold(zfsvfs_t *zfsvfs, dmu_tx_t *tx)
764 if (zfsvfs->z_fuid_obj == 0) {
765 dmu_tx_hold_bonus(tx, DMU_NEW_OBJECT);
766 dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0,
767 FUID_SIZE_ESTIMATE(zfsvfs));
768 dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, FALSE, NULL);
770 dmu_tx_hold_bonus(tx, zfsvfs->z_fuid_obj);
771 dmu_tx_hold_write(tx, zfsvfs->z_fuid_obj, 0,
772 FUID_SIZE_ESTIMATE(zfsvfs));