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>
32 #include <sys/kidmap.h>
34 #include <sys/zfs_vfsops.h>
35 #include <sys/zfs_znode.h>
37 #include <sys/zfs_fuid.h>
40 * FUID Domain table(s).
42 * The FUID table is stored as a packed nvlist of an array
43 * of nvlists which contain an index, domain string and offset
45 * During file system initialization the nvlist(s) are read and
46 * two AVL trees are created. One tree is keyed by the index number
47 * and the other by the domain string. Nodes are never removed from
48 * trees, but new entries may be added. If a new entry is added then
49 * the zfsvfs->z_fuid_dirty flag is set to true and the caller will then
50 * be responsible for calling zfs_fuid_sync() to sync the changes to disk.
54 #define FUID_IDX "fuid_idx"
55 #define FUID_DOMAIN "fuid_domain"
56 #define FUID_OFFSET "fuid_offset"
57 #define FUID_NVP_ARRAY "fuid_nvlist"
59 typedef struct fuid_domain {
66 static char *nulldomain = "";
69 * Compare two indexes.
72 idx_compare(const void *arg1, const void *arg2)
74 const fuid_domain_t *node1 = (const fuid_domain_t *)arg1;
75 const fuid_domain_t *node2 = (const fuid_domain_t *)arg2;
77 return (AVL_CMP(node1->f_idx, node2->f_idx));
81 * Compare two domain strings.
84 domain_compare(const void *arg1, const void *arg2)
86 const fuid_domain_t *node1 = (const fuid_domain_t *)arg1;
87 const fuid_domain_t *node2 = (const fuid_domain_t *)arg2;
90 val = strcmp(node1->f_ksid->kd_name, node2->f_ksid->kd_name);
92 return (AVL_ISIGN(val));
96 zfs_fuid_avl_tree_create(avl_tree_t *idx_tree, avl_tree_t *domain_tree)
98 avl_create(idx_tree, idx_compare,
99 sizeof (fuid_domain_t), offsetof(fuid_domain_t, f_idxnode));
100 avl_create(domain_tree, domain_compare,
101 sizeof (fuid_domain_t), offsetof(fuid_domain_t, f_domnode));
105 * load initial fuid domain and idx trees. This function is used by
106 * both the kernel and zdb.
109 zfs_fuid_table_load(objset_t *os, uint64_t fuid_obj, avl_tree_t *idx_tree,
110 avl_tree_t *domain_tree)
115 ASSERT(fuid_obj != 0);
116 VERIFY(0 == dmu_bonus_hold(os, fuid_obj,
118 fuid_size = *(uint64_t *)db->db_data;
119 dmu_buf_rele(db, FTAG);
123 nvlist_t *nvp = NULL;
128 packed = kmem_alloc(fuid_size, KM_SLEEP);
129 VERIFY(dmu_read(os, fuid_obj, 0,
130 fuid_size, packed, DMU_READ_PREFETCH) == 0);
131 VERIFY(nvlist_unpack(packed, fuid_size,
133 VERIFY(nvlist_lookup_nvlist_array(nvp, FUID_NVP_ARRAY,
134 &fuidnvp, &count) == 0);
136 for (i = 0; i != count; i++) {
137 fuid_domain_t *domnode;
141 VERIFY(nvlist_lookup_string(fuidnvp[i], FUID_DOMAIN,
143 VERIFY(nvlist_lookup_uint64(fuidnvp[i], FUID_IDX,
146 domnode = kmem_alloc(sizeof (fuid_domain_t), KM_SLEEP);
148 domnode->f_idx = idx;
149 domnode->f_ksid = ksid_lookupdomain(domain);
150 avl_add(idx_tree, domnode);
151 avl_add(domain_tree, domnode);
154 kmem_free(packed, fuid_size);
160 zfs_fuid_table_destroy(avl_tree_t *idx_tree, avl_tree_t *domain_tree)
162 fuid_domain_t *domnode;
166 while (domnode = avl_destroy_nodes(domain_tree, &cookie))
167 ksiddomain_rele(domnode->f_ksid);
169 avl_destroy(domain_tree);
171 while (domnode = avl_destroy_nodes(idx_tree, &cookie))
172 kmem_free(domnode, sizeof (fuid_domain_t));
173 avl_destroy(idx_tree);
177 zfs_fuid_idx_domain(avl_tree_t *idx_tree, uint32_t idx)
179 fuid_domain_t searchnode, *findnode;
182 searchnode.f_idx = idx;
184 findnode = avl_find(idx_tree, &searchnode, &loc);
186 return (findnode ? findnode->f_ksid->kd_name : nulldomain);
191 * Load the fuid table(s) into memory.
194 zfs_fuid_init(zfsvfs_t *zfsvfs)
196 rw_enter(&zfsvfs->z_fuid_lock, RW_WRITER);
198 if (zfsvfs->z_fuid_loaded) {
199 rw_exit(&zfsvfs->z_fuid_lock);
203 zfs_fuid_avl_tree_create(&zfsvfs->z_fuid_idx, &zfsvfs->z_fuid_domain);
205 (void) zap_lookup(zfsvfs->z_os, MASTER_NODE_OBJ,
206 ZFS_FUID_TABLES, 8, 1, &zfsvfs->z_fuid_obj);
207 if (zfsvfs->z_fuid_obj != 0) {
208 zfsvfs->z_fuid_size = zfs_fuid_table_load(zfsvfs->z_os,
209 zfsvfs->z_fuid_obj, &zfsvfs->z_fuid_idx,
210 &zfsvfs->z_fuid_domain);
213 zfsvfs->z_fuid_loaded = B_TRUE;
214 rw_exit(&zfsvfs->z_fuid_lock);
218 * sync out AVL trees to persistent storage.
221 zfs_fuid_sync(zfsvfs_t *zfsvfs, dmu_tx_t *tx)
228 fuid_domain_t *domnode;
232 if (!zfsvfs->z_fuid_dirty) {
236 rw_enter(&zfsvfs->z_fuid_lock, RW_WRITER);
239 * First see if table needs to be created?
241 if (zfsvfs->z_fuid_obj == 0) {
242 zfsvfs->z_fuid_obj = dmu_object_alloc(zfsvfs->z_os,
243 DMU_OT_FUID, 1 << 14, DMU_OT_FUID_SIZE,
244 sizeof (uint64_t), tx);
245 VERIFY(zap_add(zfsvfs->z_os, MASTER_NODE_OBJ,
246 ZFS_FUID_TABLES, sizeof (uint64_t), 1,
247 &zfsvfs->z_fuid_obj, tx) == 0);
250 VERIFY(nvlist_alloc(&nvp, NV_UNIQUE_NAME, KM_SLEEP) == 0);
252 numnodes = avl_numnodes(&zfsvfs->z_fuid_idx);
253 fuids = kmem_alloc(numnodes * sizeof (void *), KM_SLEEP);
254 for (i = 0, domnode = avl_first(&zfsvfs->z_fuid_domain); domnode; i++,
255 domnode = AVL_NEXT(&zfsvfs->z_fuid_domain, domnode)) {
256 VERIFY(nvlist_alloc(&fuids[i], NV_UNIQUE_NAME, KM_SLEEP) == 0);
257 VERIFY(nvlist_add_uint64(fuids[i], FUID_IDX,
258 domnode->f_idx) == 0);
259 VERIFY(nvlist_add_uint64(fuids[i], FUID_OFFSET, 0) == 0);
260 VERIFY(nvlist_add_string(fuids[i], FUID_DOMAIN,
261 domnode->f_ksid->kd_name) == 0);
263 VERIFY(nvlist_add_nvlist_array(nvp, FUID_NVP_ARRAY,
264 fuids, numnodes) == 0);
265 for (i = 0; i != numnodes; i++)
266 nvlist_free(fuids[i]);
267 kmem_free(fuids, numnodes * sizeof (void *));
268 VERIFY(nvlist_size(nvp, &nvsize, NV_ENCODE_XDR) == 0);
269 packed = kmem_alloc(nvsize, KM_SLEEP);
270 VERIFY(nvlist_pack(nvp, &packed, &nvsize,
271 NV_ENCODE_XDR, KM_SLEEP) == 0);
273 zfsvfs->z_fuid_size = nvsize;
274 dmu_write(zfsvfs->z_os, zfsvfs->z_fuid_obj, 0,
275 zfsvfs->z_fuid_size, packed, tx);
276 kmem_free(packed, zfsvfs->z_fuid_size);
277 VERIFY(0 == dmu_bonus_hold(zfsvfs->z_os, zfsvfs->z_fuid_obj,
279 dmu_buf_will_dirty(db, tx);
280 *(uint64_t *)db->db_data = zfsvfs->z_fuid_size;
281 dmu_buf_rele(db, FTAG);
283 zfsvfs->z_fuid_dirty = B_FALSE;
284 rw_exit(&zfsvfs->z_fuid_lock);
288 * Query domain table for a given domain.
290 * If domain isn't found and addok is set, it is added to AVL trees and
291 * the zfsvfs->z_fuid_dirty flag will be set to TRUE. It will then be
292 * necessary for the caller or another thread to detect the dirty table
293 * and sync out the changes.
296 zfs_fuid_find_by_domain(zfsvfs_t *zfsvfs, const char *domain,
297 char **retdomain, boolean_t addok)
299 fuid_domain_t searchnode, *findnode;
301 krw_t rw = RW_READER;
304 * If the dummy "nobody" domain then return an index of 0
305 * to cause the created FUID to be a standard POSIX id
306 * for the user nobody.
308 if (domain[0] == '\0') {
310 *retdomain = nulldomain;
314 searchnode.f_ksid = ksid_lookupdomain(domain);
316 *retdomain = searchnode.f_ksid->kd_name;
317 if (!zfsvfs->z_fuid_loaded)
318 zfs_fuid_init(zfsvfs);
321 rw_enter(&zfsvfs->z_fuid_lock, rw);
322 findnode = avl_find(&zfsvfs->z_fuid_domain, &searchnode, &loc);
325 rw_exit(&zfsvfs->z_fuid_lock);
326 ksiddomain_rele(searchnode.f_ksid);
327 return (findnode->f_idx);
329 fuid_domain_t *domnode;
332 if (rw == RW_READER && !rw_tryupgrade(&zfsvfs->z_fuid_lock)) {
333 rw_exit(&zfsvfs->z_fuid_lock);
338 domnode = kmem_alloc(sizeof (fuid_domain_t), KM_SLEEP);
339 domnode->f_ksid = searchnode.f_ksid;
341 retidx = domnode->f_idx = avl_numnodes(&zfsvfs->z_fuid_idx) + 1;
343 avl_add(&zfsvfs->z_fuid_domain, domnode);
344 avl_add(&zfsvfs->z_fuid_idx, domnode);
345 zfsvfs->z_fuid_dirty = B_TRUE;
346 rw_exit(&zfsvfs->z_fuid_lock);
349 rw_exit(&zfsvfs->z_fuid_lock);
355 * Query domain table by index, returning domain string
357 * Returns a pointer from an avl node of the domain string.
361 zfs_fuid_find_by_idx(zfsvfs_t *zfsvfs, uint32_t idx)
365 if (idx == 0 || !zfsvfs->z_use_fuids)
368 if (!zfsvfs->z_fuid_loaded)
369 zfs_fuid_init(zfsvfs);
371 rw_enter(&zfsvfs->z_fuid_lock, RW_READER);
373 if (zfsvfs->z_fuid_obj || zfsvfs->z_fuid_dirty)
374 domain = zfs_fuid_idx_domain(&zfsvfs->z_fuid_idx, idx);
377 rw_exit(&zfsvfs->z_fuid_lock);
384 zfs_fuid_map_ids(znode_t *zp, cred_t *cr, uid_t *uidp, uid_t *gidp)
386 *uidp = zfs_fuid_map_id(zp->z_zfsvfs, zp->z_uid, cr, ZFS_OWNER);
387 *gidp = zfs_fuid_map_id(zp->z_zfsvfs, zp->z_gid, cr, ZFS_GROUP);
391 zfs_fuid_map_id(zfsvfs_t *zfsvfs, uint64_t fuid,
392 cred_t *cr, zfs_fuid_type_t type)
394 uint32_t index = FUID_INDEX(fuid);
401 domain = zfs_fuid_find_by_idx(zfsvfs, index);
402 ASSERT(domain != NULL);
405 if (type == ZFS_OWNER || type == ZFS_ACE_USER) {
406 (void) kidmap_getuidbysid(crgetzone(cr), domain,
407 FUID_RID(fuid), &id);
409 (void) kidmap_getgidbysid(crgetzone(cr), domain,
410 FUID_RID(fuid), &id);
419 * Add a FUID node to the list of fuid's being created for this
422 * If ACL has multiple domains, then keep only one copy of each unique
426 zfs_fuid_node_add(zfs_fuid_info_t **fuidpp, const char *domain, uint32_t rid,
427 uint64_t idx, uint64_t id, zfs_fuid_type_t type)
430 zfs_fuid_domain_t *fuid_domain;
431 zfs_fuid_info_t *fuidp;
433 boolean_t found = B_FALSE;
436 *fuidpp = zfs_fuid_info_alloc();
440 * First find fuid domain index in linked list
442 * If one isn't found then create an entry.
445 for (fuididx = 1, fuid_domain = list_head(&fuidp->z_domains);
446 fuid_domain; fuid_domain = list_next(&fuidp->z_domains,
447 fuid_domain), fuididx++) {
448 if (idx == fuid_domain->z_domidx) {
455 fuid_domain = kmem_alloc(sizeof (zfs_fuid_domain_t), KM_SLEEP);
456 fuid_domain->z_domain = domain;
457 fuid_domain->z_domidx = idx;
458 list_insert_tail(&fuidp->z_domains, fuid_domain);
459 fuidp->z_domain_str_sz += strlen(domain) + 1;
460 fuidp->z_domain_cnt++;
463 if (type == ZFS_ACE_USER || type == ZFS_ACE_GROUP) {
466 * Now allocate fuid entry and add it on the end of the list
469 fuid = kmem_alloc(sizeof (zfs_fuid_t), KM_SLEEP);
471 fuid->z_domidx = idx;
472 fuid->z_logfuid = FUID_ENCODE(fuididx, rid);
474 list_insert_tail(&fuidp->z_fuids, fuid);
477 if (type == ZFS_OWNER)
478 fuidp->z_fuid_owner = FUID_ENCODE(fuididx, rid);
480 fuidp->z_fuid_group = FUID_ENCODE(fuididx, rid);
485 * Create a file system FUID, based on information in the users cred
487 * If cred contains KSID_OWNER then it should be used to determine
488 * the uid otherwise cred's uid will be used. By default cred's gid
489 * is used unless it's an ephemeral ID in which case KSID_GROUP will
490 * be used if it exists.
493 zfs_fuid_create_cred(zfsvfs_t *zfsvfs, zfs_fuid_type_t type,
494 cred_t *cr, zfs_fuid_info_t **fuidp)
503 VERIFY(type == ZFS_OWNER || type == ZFS_GROUP);
505 ksid = crgetsid(cr, (type == ZFS_OWNER) ? KSID_OWNER : KSID_GROUP);
507 if (!zfsvfs->z_use_fuids || (ksid == NULL)) {
508 id = (type == ZFS_OWNER) ? crgetuid(cr) : crgetgid(cr);
510 if (IS_EPHEMERAL(id))
511 return ((type == ZFS_OWNER) ? UID_NOBODY : GID_NOBODY);
513 return ((uint64_t)id);
517 * ksid is present and FUID is supported
519 id = (type == ZFS_OWNER) ? ksid_getid(ksid) : crgetgid(cr);
521 if (!IS_EPHEMERAL(id))
522 return ((uint64_t)id);
524 if (type == ZFS_GROUP)
525 id = ksid_getid(ksid);
527 rid = ksid_getrid(ksid);
528 domain = ksid_getdomain(ksid);
530 idx = zfs_fuid_find_by_domain(zfsvfs, domain, &kdomain, B_TRUE);
532 zfs_fuid_node_add(fuidp, kdomain, rid, idx, id, type);
534 return (FUID_ENCODE(idx, rid));
538 * Create a file system FUID for an ACL ace
539 * or a chown/chgrp of the file.
540 * This is similar to zfs_fuid_create_cred, except that
541 * we can't find the domain + rid information in the
542 * cred. Instead we have to query Winchester for the
545 * During replay operations the domain+rid information is
546 * found in the zfs_fuid_info_t that the replay code has
547 * attached to the zfsvfs of the file system.
550 zfs_fuid_create(zfsvfs_t *zfsvfs, uint64_t id, cred_t *cr,
551 zfs_fuid_type_t type, zfs_fuid_info_t **fuidpp)
555 uint32_t fuid_idx = FUID_INDEX(id);
559 zfs_fuid_t *zfuid = NULL;
560 zfs_fuid_info_t *fuidp = NULL;
563 * If POSIX ID, or entry is already a FUID then
566 * We may also be handed an already FUID'ized id via
570 if (!zfsvfs->z_use_fuids || !IS_EPHEMERAL(id) || fuid_idx != 0)
573 if (zfsvfs->z_replay) {
574 fuidp = zfsvfs->z_fuid_replay;
577 * If we are passed an ephemeral id, but no
578 * fuid_info was logged then return NOBODY.
579 * This is most likely a result of idmap service
580 * not being available.
585 VERIFY3U(type, >=, ZFS_OWNER);
586 VERIFY3U(type, <=, ZFS_ACE_GROUP);
591 zfuid = list_head(&fuidp->z_fuids);
592 rid = FUID_RID(zfuid->z_logfuid);
593 idx = FUID_INDEX(zfuid->z_logfuid);
596 rid = FUID_RID(fuidp->z_fuid_owner);
597 idx = FUID_INDEX(fuidp->z_fuid_owner);
600 rid = FUID_RID(fuidp->z_fuid_group);
601 idx = FUID_INDEX(fuidp->z_fuid_group);
604 domain = fuidp->z_domain_table[idx - 1];
606 if (type == ZFS_OWNER || type == ZFS_ACE_USER)
607 status = kidmap_getsidbyuid(crgetzone(cr), id,
610 status = kidmap_getsidbygid(crgetzone(cr), id,
615 * When returning nobody we will need to
616 * make a dummy fuid table entry for logging
624 idx = zfs_fuid_find_by_domain(zfsvfs, domain, &kdomain, B_TRUE);
626 if (!zfsvfs->z_replay)
627 zfs_fuid_node_add(fuidpp, kdomain,
629 else if (zfuid != NULL) {
630 list_remove(&fuidp->z_fuids, zfuid);
631 kmem_free(zfuid, sizeof (zfs_fuid_t));
633 return (FUID_ENCODE(idx, rid));
637 zfs_fuid_destroy(zfsvfs_t *zfsvfs)
639 rw_enter(&zfsvfs->z_fuid_lock, RW_WRITER);
640 if (!zfsvfs->z_fuid_loaded) {
641 rw_exit(&zfsvfs->z_fuid_lock);
644 zfs_fuid_table_destroy(&zfsvfs->z_fuid_idx, &zfsvfs->z_fuid_domain);
645 rw_exit(&zfsvfs->z_fuid_lock);
649 * Allocate zfs_fuid_info for tracking FUIDs created during
650 * zfs_mknode, VOP_SETATTR() or VOP_SETSECATTR()
653 zfs_fuid_info_alloc(void)
655 zfs_fuid_info_t *fuidp;
657 fuidp = kmem_zalloc(sizeof (zfs_fuid_info_t), KM_SLEEP);
658 list_create(&fuidp->z_domains, sizeof (zfs_fuid_domain_t),
659 offsetof(zfs_fuid_domain_t, z_next));
660 list_create(&fuidp->z_fuids, sizeof (zfs_fuid_t),
661 offsetof(zfs_fuid_t, z_next));
666 * Release all memory associated with zfs_fuid_info_t
669 zfs_fuid_info_free(zfs_fuid_info_t *fuidp)
672 zfs_fuid_domain_t *zdomain;
674 while ((zfuid = list_head(&fuidp->z_fuids)) != NULL) {
675 list_remove(&fuidp->z_fuids, zfuid);
676 kmem_free(zfuid, sizeof (zfs_fuid_t));
679 if (fuidp->z_domain_table != NULL)
680 kmem_free(fuidp->z_domain_table,
681 (sizeof (char **)) * fuidp->z_domain_cnt);
683 while ((zdomain = list_head(&fuidp->z_domains)) != NULL) {
684 list_remove(&fuidp->z_domains, zdomain);
685 kmem_free(zdomain, sizeof (zfs_fuid_domain_t));
688 kmem_free(fuidp, sizeof (zfs_fuid_info_t));
692 * Check to see if id is a groupmember. If cred
693 * has ksid info then sidlist is checked first
694 * and if still not found then POSIX groups are checked
696 * Will use a straight FUID compare when possible.
699 zfs_groupmember(zfsvfs_t *zfsvfs, uint64_t id, cred_t *cr)
702 ksid_t *ksid = crgetsid(cr, KSID_GROUP);
703 ksidlist_t *ksidlist = crgetsidlist(cr);
708 if (ksid && ksidlist) {
711 uint32_t idx = FUID_INDEX(id);
712 uint32_t rid = FUID_RID(id);
714 ksid_groups = ksidlist->ksl_sids;
716 for (i = 0; i != ksidlist->ksl_nsid; i++) {
718 if (id != IDMAP_WK_CREATOR_GROUP_GID &&
719 id == ksid_groups[i].ks_id) {
725 domain = zfs_fuid_find_by_idx(zfsvfs, idx);
726 ASSERT(domain != NULL);
729 IDMAP_WK_CREATOR_SID_AUTHORITY) == 0)
733 ksid_groups[i].ks_domain->kd_name) == 0) &&
734 rid == ksid_groups[i].ks_rid)
742 * Not found in ksidlist, check posix groups
744 gid = zfs_fuid_map_id(zfsvfs, id, cr, ZFS_GROUP);
745 return (groupmember(gid, cr));
749 zfs_fuid_txhold(zfsvfs_t *zfsvfs, dmu_tx_t *tx)
751 if (zfsvfs->z_fuid_obj == 0) {
752 dmu_tx_hold_bonus(tx, DMU_NEW_OBJECT);
753 dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0,
754 FUID_SIZE_ESTIMATE(zfsvfs));
755 dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, FALSE, NULL);
757 dmu_tx_hold_bonus(tx, zfsvfs->z_fuid_obj);
758 dmu_tx_hold_write(tx, zfsvfs->z_fuid_obj, 0,
759 FUID_SIZE_ESTIMATE(zfsvfs));