zfs_fuid.c revision 11935:538c866aaac6
160107Sobrien/*
217695Swosch * CDDL HEADER START
317695Swosch *
417695Swosch * The contents of this file are subject to the terms of the
517695Swosch * Common Development and Distribution License (the "License").
617695Swosch * You may not use this file except in compliance with the License.
748663Syokota *
817695Swosch * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
917695Swosch * or http://www.opensolaris.org/os/licensing.
1017695Swosch * See the License for the specific language governing permissions
1117695Swosch * and limitations under the License.
1217695Swosch *
1348663Syokota * When distributing Covered Code, include this CDDL HEADER in each
1419554Sjoerg * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
1526656Sache * If applicable, add the following below this CDDL HEADER, with the
1617695Swosch * fields enclosed by brackets "[]" replaced with your own identifying
1748663Syokota * information: Portions Copyright [yyyy] [name of copyright owner]
1848663Syokota *
1948663Syokota * CDDL HEADER END
2017695Swosch */
2117695Swosch/*
2248663Syokota * Copyright 2010 Sun Microsystems, Inc.  All rights reserved.
2348663Syokota * Use is subject to license terms.
2461118Sroberto */
2517695Swosch
2617695Swosch#include <sys/zfs_context.h>
2717695Swosch#include <sys/dmu.h>
2817695Swosch#include <sys/avl.h>
2917695Swosch#include <sys/zap.h>
3017695Swosch#include <sys/refcount.h>
3117695Swosch#include <sys/nvpair.h>
3217695Swosch#ifdef _KERNEL
3317695Swosch#include <sys/kidmap.h>
3417695Swosch#include <sys/sid.h>
3517695Swosch#include <sys/zfs_vfsops.h>
3648663Syokota#include <sys/zfs_znode.h>
3717695Swosch#endif
3817695Swosch#include <sys/zfs_fuid.h>
3917695Swosch
4017695Swosch/*
4117695Swosch * FUID Domain table(s).
4217695Swosch *
4317695Swosch * The FUID table is stored as a packed nvlist of an array
4417695Swosch * of nvlists which contain an index, domain string and offset
4548663Syokota *
4648663Syokota * During file system initialization the nvlist(s) are read and
4748663Syokota * two AVL trees are created.  One tree is keyed by the index number
4817695Swosch * and the other by the domain string.  Nodes are never removed from
4948663Syokota * trees, but new entries may be added.  If a new entry is added then
5048663Syokota * the zfsvfs->z_fuid_dirty flag is set to true and the caller will then
5117695Swosch * be responsible for calling zfs_fuid_sync() to sync the changes to disk.
5217695Swosch *
5317695Swosch */
5417695Swosch
5517695Swosch#define	FUID_IDX	"fuid_idx"
5648663Syokota#define	FUID_DOMAIN	"fuid_domain"
5717695Swosch#define	FUID_OFFSET	"fuid_offset"
5817695Swosch#define	FUID_NVP_ARRAY	"fuid_nvlist"
5917695Swosch
6017695Swoschtypedef struct fuid_domain {
6143334Syokota	avl_node_t	f_domnode;
6217695Swosch	avl_node_t	f_idxnode;
6332822Syokota	ksiddomain_t	*f_ksid;
6417695Swosch	uint64_t	f_idx;
6517695Swosch} fuid_domain_t;
6617695Swosch
6717695Swoschstatic char *nulldomain = "";
6817695Swosch
6917695Swosch/*
7017695Swosch * Compare two indexes.
7117695Swosch */
7217695Swoschstatic int
7317695Swoschidx_compare(const void *arg1, const void *arg2)
7417695Swosch{
7543334Syokota	const fuid_domain_t *node1 = arg1;
7643334Syokota	const fuid_domain_t *node2 = arg2;
7717695Swosch
7817695Swosch	if (node1->f_idx < node2->f_idx)
7917695Swosch		return (-1);
8017695Swosch	else if (node1->f_idx > node2->f_idx)
8117695Swosch		return (1);
8217695Swosch	return (0);
8317695Swosch}
8417695Swosch
8517695Swosch/*
8617695Swosch * Compare two domain strings.
8717695Swosch */
8817695Swoschstatic int
8917695Swoschdomain_compare(const void *arg1, const void *arg2)
9043334Syokota{
9117695Swosch	const fuid_domain_t *node1 = arg1;
9248663Syokota	const fuid_domain_t *node2 = arg2;
9317695Swosch	int val;
9417695Swosch
9518194Ssos	val = strcmp(node1->f_ksid->kd_name, node2->f_ksid->kd_name);
9617695Swosch	if (val == 0)
9717695Swosch		return (0);
9843334Syokota	return (val > 0 ? 1 : -1);
9917695Swosch}
10017695Swosch
10117695Swoschvoid
10217695Swoschzfs_fuid_avl_tree_create(avl_tree_t *idx_tree, avl_tree_t *domain_tree)
10317695Swosch{
10417695Swosch	avl_create(idx_tree, idx_compare,
10517695Swosch	    sizeof (fuid_domain_t), offsetof(fuid_domain_t, f_idxnode));
10617695Swosch	avl_create(domain_tree, domain_compare,
10717695Swosch	    sizeof (fuid_domain_t), offsetof(fuid_domain_t, f_domnode));
10817695Swosch}
10917695Swosch
11043334Syokota/*
11143334Syokota * load initial fuid domain and idx trees.  This function is used by
11243334Syokota * both the kernel and zdb.
11343334Syokota */
11443334Syokotauint64_t
115zfs_fuid_table_load(objset_t *os, uint64_t fuid_obj, avl_tree_t *idx_tree,
116    avl_tree_t *domain_tree)
117{
118	dmu_buf_t *db;
119	uint64_t fuid_size;
120
121	ASSERT(fuid_obj != 0);
122	VERIFY(0 == dmu_bonus_hold(os, fuid_obj,
123	    FTAG, &db));
124	fuid_size = *(uint64_t *)db->db_data;
125	dmu_buf_rele(db, FTAG);
126
127	if (fuid_size)  {
128		nvlist_t **fuidnvp;
129		nvlist_t *nvp = NULL;
130		uint_t count;
131		char *packed;
132		int i;
133
134		packed = kmem_alloc(fuid_size, KM_SLEEP);
135		VERIFY(dmu_read(os, fuid_obj, 0,
136		    fuid_size, packed, DMU_READ_PREFETCH) == 0);
137		VERIFY(nvlist_unpack(packed, fuid_size,
138		    &nvp, 0) == 0);
139		VERIFY(nvlist_lookup_nvlist_array(nvp, FUID_NVP_ARRAY,
140		    &fuidnvp, &count) == 0);
141
142		for (i = 0; i != count; i++) {
143			fuid_domain_t *domnode;
144			char *domain;
145			uint64_t idx;
146
147			VERIFY(nvlist_lookup_string(fuidnvp[i], FUID_DOMAIN,
148			    &domain) == 0);
149			VERIFY(nvlist_lookup_uint64(fuidnvp[i], FUID_IDX,
150			    &idx) == 0);
151
152			domnode = kmem_alloc(sizeof (fuid_domain_t), KM_SLEEP);
153
154			domnode->f_idx = idx;
155			domnode->f_ksid = ksid_lookupdomain(domain);
156			avl_add(idx_tree, domnode);
157			avl_add(domain_tree, domnode);
158		}
159		nvlist_free(nvp);
160		kmem_free(packed, fuid_size);
161	}
162	return (fuid_size);
163}
164
165void
166zfs_fuid_table_destroy(avl_tree_t *idx_tree, avl_tree_t *domain_tree)
167{
168	fuid_domain_t *domnode;
169	void *cookie;
170
171	cookie = NULL;
172	while (domnode = avl_destroy_nodes(domain_tree, &cookie))
173		ksiddomain_rele(domnode->f_ksid);
174
175	avl_destroy(domain_tree);
176	cookie = NULL;
177	while (domnode = avl_destroy_nodes(idx_tree, &cookie))
178		kmem_free(domnode, sizeof (fuid_domain_t));
179	avl_destroy(idx_tree);
180}
181
182char *
183zfs_fuid_idx_domain(avl_tree_t *idx_tree, uint32_t idx)
184{
185	fuid_domain_t searchnode, *findnode;
186	avl_index_t loc;
187
188	searchnode.f_idx = idx;
189
190	findnode = avl_find(idx_tree, &searchnode, &loc);
191
192	return (findnode ? findnode->f_ksid->kd_name : nulldomain);
193}
194
195#ifdef _KERNEL
196/*
197 * Load the fuid table(s) into memory.
198 */
199static void
200zfs_fuid_init(zfsvfs_t *zfsvfs)
201{
202	rw_enter(&zfsvfs->z_fuid_lock, RW_WRITER);
203
204	if (zfsvfs->z_fuid_loaded) {
205		rw_exit(&zfsvfs->z_fuid_lock);
206		return;
207	}
208
209	zfs_fuid_avl_tree_create(&zfsvfs->z_fuid_idx, &zfsvfs->z_fuid_domain);
210
211	(void) zap_lookup(zfsvfs->z_os, MASTER_NODE_OBJ,
212	    ZFS_FUID_TABLES, 8, 1, &zfsvfs->z_fuid_obj);
213	if (zfsvfs->z_fuid_obj != 0) {
214		zfsvfs->z_fuid_size = zfs_fuid_table_load(zfsvfs->z_os,
215		    zfsvfs->z_fuid_obj, &zfsvfs->z_fuid_idx,
216		    &zfsvfs->z_fuid_domain);
217	}
218
219	zfsvfs->z_fuid_loaded = B_TRUE;
220	rw_exit(&zfsvfs->z_fuid_lock);
221}
222
223/*
224 * sync out AVL trees to persistent storage.
225 */
226void
227zfs_fuid_sync(zfsvfs_t *zfsvfs, dmu_tx_t *tx)
228{
229	nvlist_t *nvp;
230	nvlist_t **fuids;
231	size_t nvsize = 0;
232	char *packed;
233	dmu_buf_t *db;
234	fuid_domain_t *domnode;
235	int numnodes;
236	int i;
237
238	if (!zfsvfs->z_fuid_dirty) {
239		return;
240	}
241
242	rw_enter(&zfsvfs->z_fuid_lock, RW_WRITER);
243
244	/*
245	 * First see if table needs to be created?
246	 */
247	if (zfsvfs->z_fuid_obj == 0) {
248		zfsvfs->z_fuid_obj = dmu_object_alloc(zfsvfs->z_os,
249		    DMU_OT_FUID, 1 << 14, DMU_OT_FUID_SIZE,
250		    sizeof (uint64_t), tx);
251		VERIFY(zap_add(zfsvfs->z_os, MASTER_NODE_OBJ,
252		    ZFS_FUID_TABLES, sizeof (uint64_t), 1,
253		    &zfsvfs->z_fuid_obj, tx) == 0);
254	}
255
256	VERIFY(nvlist_alloc(&nvp, NV_UNIQUE_NAME, KM_SLEEP) == 0);
257
258	numnodes = avl_numnodes(&zfsvfs->z_fuid_idx);
259	fuids = kmem_alloc(numnodes * sizeof (void *), KM_SLEEP);
260	for (i = 0, domnode = avl_first(&zfsvfs->z_fuid_domain); domnode; i++,
261	    domnode = AVL_NEXT(&zfsvfs->z_fuid_domain, domnode)) {
262		VERIFY(nvlist_alloc(&fuids[i], NV_UNIQUE_NAME, KM_SLEEP) == 0);
263		VERIFY(nvlist_add_uint64(fuids[i], FUID_IDX,
264		    domnode->f_idx) == 0);
265		VERIFY(nvlist_add_uint64(fuids[i], FUID_OFFSET, 0) == 0);
266		VERIFY(nvlist_add_string(fuids[i], FUID_DOMAIN,
267		    domnode->f_ksid->kd_name) == 0);
268	}
269	VERIFY(nvlist_add_nvlist_array(nvp, FUID_NVP_ARRAY,
270	    fuids, numnodes) == 0);
271	for (i = 0; i != numnodes; i++)
272		nvlist_free(fuids[i]);
273	kmem_free(fuids, numnodes * sizeof (void *));
274	VERIFY(nvlist_size(nvp, &nvsize, NV_ENCODE_XDR) == 0);
275	packed = kmem_alloc(nvsize, KM_SLEEP);
276	VERIFY(nvlist_pack(nvp, &packed, &nvsize,
277	    NV_ENCODE_XDR, KM_SLEEP) == 0);
278	nvlist_free(nvp);
279	zfsvfs->z_fuid_size = nvsize;
280	dmu_write(zfsvfs->z_os, zfsvfs->z_fuid_obj, 0,
281	    zfsvfs->z_fuid_size, packed, tx);
282	kmem_free(packed, zfsvfs->z_fuid_size);
283	VERIFY(0 == dmu_bonus_hold(zfsvfs->z_os, zfsvfs->z_fuid_obj,
284	    FTAG, &db));
285	dmu_buf_will_dirty(db, tx);
286	*(uint64_t *)db->db_data = zfsvfs->z_fuid_size;
287	dmu_buf_rele(db, FTAG);
288
289	zfsvfs->z_fuid_dirty = B_FALSE;
290	rw_exit(&zfsvfs->z_fuid_lock);
291}
292
293/*
294 * Query domain table for a given domain.
295 *
296 * If domain isn't found and addok is set, it is added to AVL trees and
297 * the zfsvfs->z_fuid_dirty flag will be set to TRUE.  It will then be
298 * necessary for the caller or another thread to detect the dirty table
299 * and sync out the changes.
300 */
301int
302zfs_fuid_find_by_domain(zfsvfs_t *zfsvfs, const char *domain,
303    char **retdomain, boolean_t addok)
304{
305	fuid_domain_t searchnode, *findnode;
306	avl_index_t loc;
307	krw_t rw = RW_READER;
308
309	/*
310	 * If the dummy "nobody" domain then return an index of 0
311	 * to cause the created FUID to be a standard POSIX id
312	 * for the user nobody.
313	 */
314	if (domain[0] == '\0') {
315		if (retdomain)
316			*retdomain = nulldomain;
317		return (0);
318	}
319
320	searchnode.f_ksid = ksid_lookupdomain(domain);
321	if (retdomain)
322		*retdomain = searchnode.f_ksid->kd_name;
323	if (!zfsvfs->z_fuid_loaded)
324		zfs_fuid_init(zfsvfs);
325
326retry:
327	rw_enter(&zfsvfs->z_fuid_lock, rw);
328	findnode = avl_find(&zfsvfs->z_fuid_domain, &searchnode, &loc);
329
330	if (findnode) {
331		rw_exit(&zfsvfs->z_fuid_lock);
332		ksiddomain_rele(searchnode.f_ksid);
333		return (findnode->f_idx);
334	} else if (addok) {
335		fuid_domain_t *domnode;
336		uint64_t retidx;
337
338		if (rw == RW_READER && !rw_tryupgrade(&zfsvfs->z_fuid_lock)) {
339			rw_exit(&zfsvfs->z_fuid_lock);
340			rw = RW_WRITER;
341			goto retry;
342		}
343
344		domnode = kmem_alloc(sizeof (fuid_domain_t), KM_SLEEP);
345		domnode->f_ksid = searchnode.f_ksid;
346
347		retidx = domnode->f_idx = avl_numnodes(&zfsvfs->z_fuid_idx) + 1;
348
349		avl_add(&zfsvfs->z_fuid_domain, domnode);
350		avl_add(&zfsvfs->z_fuid_idx, domnode);
351		zfsvfs->z_fuid_dirty = B_TRUE;
352		rw_exit(&zfsvfs->z_fuid_lock);
353		return (retidx);
354	} else {
355		rw_exit(&zfsvfs->z_fuid_lock);
356		return (-1);
357	}
358}
359
360/*
361 * Query domain table by index, returning domain string
362 *
363 * Returns a pointer from an avl node of the domain string.
364 *
365 */
366const char *
367zfs_fuid_find_by_idx(zfsvfs_t *zfsvfs, uint32_t idx)
368{
369	char *domain;
370
371	if (idx == 0 || !zfsvfs->z_use_fuids)
372		return (NULL);
373
374	if (!zfsvfs->z_fuid_loaded)
375		zfs_fuid_init(zfsvfs);
376
377	rw_enter(&zfsvfs->z_fuid_lock, RW_READER);
378
379	if (zfsvfs->z_fuid_obj)
380		domain = zfs_fuid_idx_domain(&zfsvfs->z_fuid_idx, idx);
381	else
382		domain = nulldomain;
383	rw_exit(&zfsvfs->z_fuid_lock);
384
385	ASSERT(domain);
386	return (domain);
387}
388
389void
390zfs_fuid_map_ids(znode_t *zp, cred_t *cr, uid_t *uidp, uid_t *gidp)
391{
392	uint64_t fuid, fgid;
393	sa_bulk_attr_t bulk[2];
394	int count = 0;
395
396	if (IS_EPHEMERAL(zp->z_uid) || IS_EPHEMERAL(zp->z_gid)) {
397		SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zp->z_zfsvfs),
398		    NULL, &fuid, 8);
399		SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zp->z_zfsvfs),
400		    NULL, &fgid, 8);
401		VERIFY(0 == sa_bulk_lookup(zp->z_sa_hdl, bulk, count));
402	}
403	if (IS_EPHEMERAL(zp->z_uid))
404		*uidp = zfs_fuid_map_id(zp->z_zfsvfs, zp->z_uid, cr, ZFS_OWNER);
405	else
406		*uidp = zp->z_uid;
407	if (IS_EPHEMERAL(zp->z_gid))
408		*gidp = zfs_fuid_map_id(zp->z_zfsvfs,
409		    zp->z_gid, cr, ZFS_GROUP);
410	else
411		*gidp = zp->z_gid;
412}
413
414uid_t
415zfs_fuid_map_id(zfsvfs_t *zfsvfs, uint64_t fuid,
416    cred_t *cr, zfs_fuid_type_t type)
417{
418	uint32_t index = FUID_INDEX(fuid);
419	const char *domain;
420	uid_t id;
421
422	if (index == 0)
423		return (fuid);
424
425	domain = zfs_fuid_find_by_idx(zfsvfs, index);
426	ASSERT(domain != NULL);
427
428	if (type == ZFS_OWNER || type == ZFS_ACE_USER) {
429		(void) kidmap_getuidbysid(crgetzone(cr), domain,
430		    FUID_RID(fuid), &id);
431	} else {
432		(void) kidmap_getgidbysid(crgetzone(cr), domain,
433		    FUID_RID(fuid), &id);
434	}
435	return (id);
436}
437
438/*
439 * Add a FUID node to the list of fuid's being created for this
440 * ACL
441 *
442 * If ACL has multiple domains, then keep only one copy of each unique
443 * domain.
444 */
445void
446zfs_fuid_node_add(zfs_fuid_info_t **fuidpp, const char *domain, uint32_t rid,
447    uint64_t idx, uint64_t id, zfs_fuid_type_t type)
448{
449	zfs_fuid_t *fuid;
450	zfs_fuid_domain_t *fuid_domain;
451	zfs_fuid_info_t *fuidp;
452	uint64_t fuididx;
453	boolean_t found = B_FALSE;
454
455	if (*fuidpp == NULL)
456		*fuidpp = zfs_fuid_info_alloc();
457
458	fuidp = *fuidpp;
459	/*
460	 * First find fuid domain index in linked list
461	 *
462	 * If one isn't found then create an entry.
463	 */
464
465	for (fuididx = 1, fuid_domain = list_head(&fuidp->z_domains);
466	    fuid_domain; fuid_domain = list_next(&fuidp->z_domains,
467	    fuid_domain), fuididx++) {
468		if (idx == fuid_domain->z_domidx) {
469			found = B_TRUE;
470			break;
471		}
472	}
473
474	if (!found) {
475		fuid_domain = kmem_alloc(sizeof (zfs_fuid_domain_t), KM_SLEEP);
476		fuid_domain->z_domain = domain;
477		fuid_domain->z_domidx = idx;
478		list_insert_tail(&fuidp->z_domains, fuid_domain);
479		fuidp->z_domain_str_sz += strlen(domain) + 1;
480		fuidp->z_domain_cnt++;
481	}
482
483	if (type == ZFS_ACE_USER || type == ZFS_ACE_GROUP) {
484
485		/*
486		 * Now allocate fuid entry and add it on the end of the list
487		 */
488
489		fuid = kmem_alloc(sizeof (zfs_fuid_t), KM_SLEEP);
490		fuid->z_id = id;
491		fuid->z_domidx = idx;
492		fuid->z_logfuid = FUID_ENCODE(fuididx, rid);
493
494		list_insert_tail(&fuidp->z_fuids, fuid);
495		fuidp->z_fuid_cnt++;
496	} else {
497		if (type == ZFS_OWNER)
498			fuidp->z_fuid_owner = FUID_ENCODE(fuididx, rid);
499		else
500			fuidp->z_fuid_group = FUID_ENCODE(fuididx, rid);
501	}
502}
503
504/*
505 * Create a file system FUID, based on information in the users cred
506 */
507uint64_t
508zfs_fuid_create_cred(zfsvfs_t *zfsvfs, zfs_fuid_type_t type,
509    cred_t *cr, zfs_fuid_info_t **fuidp)
510{
511	uint64_t	idx;
512	ksid_t		*ksid;
513	uint32_t	rid;
514	char 		*kdomain;
515	const char	*domain;
516	uid_t		id;
517
518	VERIFY(type == ZFS_OWNER || type == ZFS_GROUP);
519
520	ksid = crgetsid(cr, (type == ZFS_OWNER) ? KSID_OWNER : KSID_GROUP);
521	if (ksid) {
522		id = ksid_getid(ksid);
523	} else {
524		if (type == ZFS_OWNER)
525			id = crgetuid(cr);
526		else
527			id = crgetgid(cr);
528
529		if (IS_EPHEMERAL(id)) {
530			return ((uint64_t)(type == ZFS_OWNER ?
531			    UID_NOBODY : GID_NOBODY));
532		}
533	}
534
535	if (!zfsvfs->z_use_fuids || (!IS_EPHEMERAL(id)))
536		return ((uint64_t)id);
537
538	rid = ksid_getrid(ksid);
539	domain = ksid_getdomain(ksid);
540
541	idx = zfs_fuid_find_by_domain(zfsvfs, domain, &kdomain, B_TRUE);
542
543	zfs_fuid_node_add(fuidp, kdomain, rid, idx, id, type);
544
545	return (FUID_ENCODE(idx, rid));
546}
547
548/*
549 * Create a file system FUID for an ACL ace
550 * or a chown/chgrp of the file.
551 * This is similar to zfs_fuid_create_cred, except that
552 * we can't find the domain + rid information in the
553 * cred.  Instead we have to query Winchester for the
554 * domain and rid.
555 *
556 * During replay operations the domain+rid information is
557 * found in the zfs_fuid_info_t that the replay code has
558 * attached to the zfsvfs of the file system.
559 */
560uint64_t
561zfs_fuid_create(zfsvfs_t *zfsvfs, uint64_t id, cred_t *cr,
562    zfs_fuid_type_t type, zfs_fuid_info_t **fuidpp)
563{
564	const char *domain;
565	char *kdomain;
566	uint32_t fuid_idx = FUID_INDEX(id);
567	uint32_t rid;
568	idmap_stat status;
569	uint64_t idx;
570	zfs_fuid_t *zfuid = NULL;
571	zfs_fuid_info_t *fuidp;
572
573	/*
574	 * If POSIX ID, or entry is already a FUID then
575	 * just return the id
576	 *
577	 * We may also be handed an already FUID'ized id via
578	 * chmod.
579	 */
580
581	if (!zfsvfs->z_use_fuids || !IS_EPHEMERAL(id) || fuid_idx != 0)
582		return (id);
583
584	if (zfsvfs->z_replay) {
585		fuidp = zfsvfs->z_fuid_replay;
586
587		/*
588		 * If we are passed an ephemeral id, but no
589		 * fuid_info was logged then return NOBODY.
590		 * This is most likely a result of idmap service
591		 * not being available.
592		 */
593		if (fuidp == NULL)
594			return (UID_NOBODY);
595
596		switch (type) {
597		case ZFS_ACE_USER:
598		case ZFS_ACE_GROUP:
599			zfuid = list_head(&fuidp->z_fuids);
600			rid = FUID_RID(zfuid->z_logfuid);
601			idx = FUID_INDEX(zfuid->z_logfuid);
602			break;
603		case ZFS_OWNER:
604			rid = FUID_RID(fuidp->z_fuid_owner);
605			idx = FUID_INDEX(fuidp->z_fuid_owner);
606			break;
607		case ZFS_GROUP:
608			rid = FUID_RID(fuidp->z_fuid_group);
609			idx = FUID_INDEX(fuidp->z_fuid_group);
610			break;
611		};
612		domain = fuidp->z_domain_table[idx -1];
613	} else {
614		if (type == ZFS_OWNER || type == ZFS_ACE_USER)
615			status = kidmap_getsidbyuid(crgetzone(cr), id,
616			    &domain, &rid);
617		else
618			status = kidmap_getsidbygid(crgetzone(cr), id,
619			    &domain, &rid);
620
621		if (status != 0) {
622			/*
623			 * When returning nobody we will need to
624			 * make a dummy fuid table entry for logging
625			 * purposes.
626			 */
627			rid = UID_NOBODY;
628			domain = nulldomain;
629		}
630	}
631
632	idx = zfs_fuid_find_by_domain(zfsvfs, domain, &kdomain, B_TRUE);
633
634	if (!zfsvfs->z_replay)
635		zfs_fuid_node_add(fuidpp, kdomain,
636		    rid, idx, id, type);
637	else if (zfuid != NULL) {
638		list_remove(&fuidp->z_fuids, zfuid);
639		kmem_free(zfuid, sizeof (zfs_fuid_t));
640	}
641	return (FUID_ENCODE(idx, rid));
642}
643
644void
645zfs_fuid_destroy(zfsvfs_t *zfsvfs)
646{
647	rw_enter(&zfsvfs->z_fuid_lock, RW_WRITER);
648	if (!zfsvfs->z_fuid_loaded) {
649		rw_exit(&zfsvfs->z_fuid_lock);
650		return;
651	}
652	zfs_fuid_table_destroy(&zfsvfs->z_fuid_idx, &zfsvfs->z_fuid_domain);
653	rw_exit(&zfsvfs->z_fuid_lock);
654}
655
656/*
657 * Allocate zfs_fuid_info for tracking FUIDs created during
658 * zfs_mknode, VOP_SETATTR() or VOP_SETSECATTR()
659 */
660zfs_fuid_info_t *
661zfs_fuid_info_alloc(void)
662{
663	zfs_fuid_info_t *fuidp;
664
665	fuidp = kmem_zalloc(sizeof (zfs_fuid_info_t), KM_SLEEP);
666	list_create(&fuidp->z_domains, sizeof (zfs_fuid_domain_t),
667	    offsetof(zfs_fuid_domain_t, z_next));
668	list_create(&fuidp->z_fuids, sizeof (zfs_fuid_t),
669	    offsetof(zfs_fuid_t, z_next));
670	return (fuidp);
671}
672
673/*
674 * Release all memory associated with zfs_fuid_info_t
675 */
676void
677zfs_fuid_info_free(zfs_fuid_info_t *fuidp)
678{
679	zfs_fuid_t *zfuid;
680	zfs_fuid_domain_t *zdomain;
681
682	while ((zfuid = list_head(&fuidp->z_fuids)) != NULL) {
683		list_remove(&fuidp->z_fuids, zfuid);
684		kmem_free(zfuid, sizeof (zfs_fuid_t));
685	}
686
687	if (fuidp->z_domain_table != NULL)
688		kmem_free(fuidp->z_domain_table,
689		    (sizeof (char **)) * fuidp->z_domain_cnt);
690
691	while ((zdomain = list_head(&fuidp->z_domains)) != NULL) {
692		list_remove(&fuidp->z_domains, zdomain);
693		kmem_free(zdomain, sizeof (zfs_fuid_domain_t));
694	}
695
696	kmem_free(fuidp, sizeof (zfs_fuid_info_t));
697}
698
699/*
700 * Check to see if id is a groupmember.  If cred
701 * has ksid info then sidlist is checked first
702 * and if still not found then POSIX groups are checked
703 *
704 * Will use a straight FUID compare when possible.
705 */
706boolean_t
707zfs_groupmember(zfsvfs_t *zfsvfs, uint64_t id, cred_t *cr)
708{
709	ksid_t		*ksid = crgetsid(cr, KSID_GROUP);
710	ksidlist_t	*ksidlist = crgetsidlist(cr);
711	uid_t		gid;
712
713	if (ksid && ksidlist) {
714		int 		i;
715		ksid_t		*ksid_groups;
716		uint32_t	idx = FUID_INDEX(id);
717		uint32_t	rid = FUID_RID(id);
718
719		ksid_groups = ksidlist->ksl_sids;
720
721		for (i = 0; i != ksidlist->ksl_nsid; i++) {
722			if (idx == 0) {
723				if (id != IDMAP_WK_CREATOR_GROUP_GID &&
724				    id == ksid_groups[i].ks_id) {
725					return (B_TRUE);
726				}
727			} else {
728				const char *domain;
729
730				domain = zfs_fuid_find_by_idx(zfsvfs, idx);
731				ASSERT(domain != NULL);
732
733				if (strcmp(domain,
734				    IDMAP_WK_CREATOR_SID_AUTHORITY) == 0)
735					return (B_FALSE);
736
737				if ((strcmp(domain,
738				    ksid_groups[i].ks_domain->kd_name) == 0) &&
739				    rid == ksid_groups[i].ks_rid)
740					return (B_TRUE);
741			}
742		}
743	}
744
745	/*
746	 * Not found in ksidlist, check posix groups
747	 */
748	gid = zfs_fuid_map_id(zfsvfs, id, cr, ZFS_GROUP);
749	return (groupmember(gid, cr));
750}
751
752void
753zfs_fuid_txhold(zfsvfs_t *zfsvfs, dmu_tx_t *tx)
754{
755	if (zfsvfs->z_fuid_obj == 0) {
756		dmu_tx_hold_bonus(tx, DMU_NEW_OBJECT);
757		dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0,
758		    FUID_SIZE_ESTIMATE(zfsvfs));
759		dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, FALSE, NULL);
760	} else {
761		dmu_tx_hold_bonus(tx, zfsvfs->z_fuid_obj);
762		dmu_tx_hold_write(tx, zfsvfs->z_fuid_obj, 0,
763		    FUID_SIZE_ESTIMATE(zfsvfs));
764	}
765}
766#endif
767