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) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
25 /* Portions Copyright 2007 Jeremy Teo */
28 #include <sys/types.h>
29 #include <sys/param.h>
31 #include <sys/systm.h>
32 #include <sys/sysmacros.h>
33 #include <sys/resource.h>
34 #include <sys/mntent.h>
35 #include <sys/mkdev.h>
36 #include <sys/u8_textprep.h>
37 #include <sys/dsl_dataset.h>
39 #include <sys/vfs_opreg.h>
40 #include <sys/vnode.h>
43 #include <sys/errno.h>
44 #include <sys/unistd.h>
46 #include <sys/atomic.h>
48 #include "fs/fs_subr.h"
49 #include <sys/zfs_dir.h>
50 #include <sys/zfs_acl.h>
51 #include <sys/zfs_ioctl.h>
52 #include <sys/zfs_rlock.h>
53 #include <sys/zfs_fuid.h>
54 #include <sys/zfs_vnops.h>
55 #include <sys/dnode.h>
56 #include <sys/fs/zfs.h>
57 #include <sys/kidmap.h>
62 #include <sys/refcount.h>
65 #include <sys/zfs_znode.h>
67 #include <sys/zfs_sa.h>
68 #include <sys/zfs_stat.h>
71 #include "zfs_comutil.h"
74 * Define ZNODE_STATS to turn on statistic gathering. By default, it is only
75 * turned on when DEBUG is also defined.
82 #define ZNODE_STAT_ADD(stat) ((stat)++)
84 #define ZNODE_STAT_ADD(stat) /* nothing */
85 #endif /* ZNODE_STATS */
88 * Functions needed for userland (ie: libzpool) are not put under
89 * #ifdef_KERNEL; the rest of the functions have dependencies
90 * (such as VFS logic) that will not compile easily in userland.
94 static kmem_cache_t *znode_cache = NULL;
98 zfs_znode_cache_constructor(void *buf, void *arg, int kmflags)
102 inode_init_once(ZTOI(zp));
103 list_link_init(&zp->z_link_node);
105 mutex_init(&zp->z_lock, NULL, MUTEX_DEFAULT, NULL);
106 rw_init(&zp->z_parent_lock, NULL, RW_DEFAULT, NULL);
107 rw_init(&zp->z_name_lock, NULL, RW_DEFAULT, NULL);
108 mutex_init(&zp->z_acl_lock, NULL, MUTEX_DEFAULT, NULL);
110 mutex_init(&zp->z_range_lock, NULL, MUTEX_DEFAULT, NULL);
111 avl_create(&zp->z_range_avl, zfs_range_compare,
112 sizeof (rl_t), offsetof(rl_t, r_node));
114 zp->z_dirlocks = NULL;
115 zp->z_acl_cached = NULL;
122 zfs_znode_cache_destructor(void *buf, void *arg)
126 ASSERT(!list_link_active(&zp->z_link_node));
127 mutex_destroy(&zp->z_lock);
128 rw_destroy(&zp->z_parent_lock);
129 rw_destroy(&zp->z_name_lock);
130 mutex_destroy(&zp->z_acl_lock);
131 avl_destroy(&zp->z_range_avl);
132 mutex_destroy(&zp->z_range_lock);
134 ASSERT(zp->z_dirlocks == NULL);
135 ASSERT(zp->z_acl_cached == NULL);
144 ASSERT(znode_cache == NULL);
145 znode_cache = kmem_cache_create("zfs_znode_cache",
146 sizeof (znode_t), 0, zfs_znode_cache_constructor,
147 zfs_znode_cache_destructor, NULL, NULL, NULL, KMC_KMEM);
157 kmem_cache_destroy(znode_cache);
162 zfs_create_share_dir(zfs_sb_t *zsb, dmu_tx_t *tx)
165 zfs_acl_ids_t acl_ids;
172 vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE;
173 vattr.va_mode = S_IFDIR | 0555;
174 vattr.va_uid = crgetuid(kcred);
175 vattr.va_gid = crgetgid(kcred);
177 sharezp = kmem_cache_alloc(znode_cache, KM_SLEEP);
178 sharezp->z_moved = 0;
179 sharezp->z_unlinked = 0;
180 sharezp->z_atime_dirty = 0;
181 sharezp->z_zfsvfs = zfsvfs;
182 sharezp->z_is_sa = zfsvfs->z_use_sa;
188 VERIFY(0 == zfs_acl_ids_create(sharezp, IS_ROOT_NODE, &vattr,
189 kcred, NULL, &acl_ids));
190 zfs_mknode(sharezp, &vattr, tx, kcred, IS_ROOT_NODE, &zp, &acl_ids);
191 ASSERT3P(zp, ==, sharezp);
192 ASSERT(!vn_in_dnlc(ZTOV(sharezp))); /* not valid to move */
193 POINTER_INVALIDATE(&sharezp->z_zfsvfs);
194 error = zap_add(zfsvfs->z_os, MASTER_NODE_OBJ,
195 ZFS_SHARES_DIR, 8, 1, &sharezp->z_id, tx);
196 zfsvfs->z_shares_dir = sharezp->z_id;
198 zfs_acl_ids_free(&acl_ids);
199 // ZTOV(sharezp)->v_count = 0;
200 sa_handle_destroy(sharezp->z_sa_hdl);
201 kmem_cache_free(znode_cache, sharezp);
206 #endif /* HAVE_SHARE */
210 zfs_znode_sa_init(zfs_sb_t *zsb, znode_t *zp,
211 dmu_buf_t *db, dmu_object_type_t obj_type, sa_handle_t *sa_hdl)
213 ASSERT(MUTEX_HELD(ZFS_OBJ_MUTEX(zsb, zp->z_id)));
215 mutex_enter(&zp->z_lock);
217 ASSERT(zp->z_sa_hdl == NULL);
218 ASSERT(zp->z_acl_cached == NULL);
219 if (sa_hdl == NULL) {
220 VERIFY(0 == sa_handle_get_from_db(zsb->z_os, db, zp,
221 SA_HDL_SHARED, &zp->z_sa_hdl));
223 zp->z_sa_hdl = sa_hdl;
224 sa_set_userp(sa_hdl, zp);
227 zp->z_is_sa = (obj_type == DMU_OT_SA) ? B_TRUE : B_FALSE;
229 mutex_exit(&zp->z_lock);
233 zfs_znode_dmu_fini(znode_t *zp)
235 ASSERT(MUTEX_HELD(ZFS_OBJ_MUTEX(ZTOZSB(zp), zp->z_id)) ||
237 RW_WRITE_HELD(&ZTOZSB(zp)->z_teardown_inactive_lock));
239 sa_handle_destroy(zp->z_sa_hdl);
244 * Called by new_inode() to allocate a new inode.
247 zfs_inode_alloc(struct super_block *sb, struct inode **ip)
251 zp = kmem_cache_alloc(znode_cache, KM_SLEEP);
258 * Called in multiple places when an inode should be destroyed.
261 zfs_inode_destroy(struct inode *ip)
263 znode_t *zp = ITOZ(ip);
264 zfs_sb_t *zsb = ZTOZSB(zp);
266 mutex_enter(&zsb->z_znodes_lock);
267 list_remove(&zsb->z_all_znodes, zp);
268 mutex_exit(&zsb->z_znodes_lock);
270 if (zp->z_acl_cached) {
271 zfs_acl_free(zp->z_acl_cached);
272 zp->z_acl_cached = NULL;
275 kmem_cache_free(znode_cache, zp);
279 zfs_inode_set_ops(zfs_sb_t *zsb, struct inode *ip)
283 switch (ip->i_mode & S_IFMT) {
285 ip->i_op = &zpl_inode_operations;
286 ip->i_fop = &zpl_file_operations;
287 ip->i_mapping->a_ops = &zpl_address_space_operations;
291 ip->i_op = &zpl_dir_inode_operations;
292 ip->i_fop = &zpl_dir_file_operations;
293 ITOZ(ip)->z_zn_prefetch = B_TRUE;
297 ip->i_op = &zpl_symlink_inode_operations;
304 VERIFY(sa_lookup(ITOZ(ip)->z_sa_hdl, SA_ZPL_RDEV(zsb),
305 &rdev, sizeof (rdev)) == 0);
306 init_special_inode(ip, ip->i_mode, rdev);
307 ip->i_op = &zpl_special_inode_operations;
311 printk("ZFS: Invalid mode: 0x%x\n", ip->i_mode);
317 * Construct a znode+inode and initialize.
319 * This does not do a call to dmu_set_user() that is
320 * up to the caller to do, in case you don't want to
324 zfs_znode_alloc(zfs_sb_t *zsb, dmu_buf_t *db, int blksz,
325 dmu_object_type_t obj_type, uint64_t obj, sa_handle_t *hdl)
330 sa_bulk_attr_t bulk[9];
335 ip = new_inode(zsb->z_sb);
340 ASSERT(zp->z_dirlocks == NULL);
344 zp->z_atime_dirty = 0;
346 zp->z_id = db->db_object;
348 zp->z_seq = 0x7A4653;
351 zfs_znode_sa_init(zsb, zp, db, obj_type, hdl);
353 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zsb), NULL, &zp->z_mode, 8);
354 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GEN(zsb), NULL, &zp->z_gen, 8);
355 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zsb), NULL, &zp->z_size, 8);
356 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zsb), NULL, &zp->z_links, 8);
357 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zsb), NULL,
359 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_PARENT(zsb), NULL,
361 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zsb), NULL,
363 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zsb), NULL, &zp->z_uid, 8);
364 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zsb), NULL, &zp->z_gid, 8);
366 if (sa_bulk_lookup(zp->z_sa_hdl, bulk, count) != 0 || zp->z_gen == 0) {
368 sa_handle_destroy(zp->z_sa_hdl);
374 ip->i_mode = zp->z_mode;
375 ip->i_mtime = ip->i_atime = ip->i_ctime = CURRENT_TIME_SEC;
376 zfs_inode_set_ops(zsb, ip);
378 if (insert_inode_locked(ip))
381 mutex_enter(&zsb->z_znodes_lock);
382 list_insert_tail(&zsb->z_all_znodes, zp);
384 mutex_exit(&zsb->z_znodes_lock);
386 unlock_new_inode(ip);
390 unlock_new_inode(ip);
396 * Update the embedded inode given the znode. We should work toward
397 * eliminating this function as soon as possible by removing values
398 * which are duplicated between the znode and inode. If the generic
399 * inode has the correct field it should be used, and the ZFS code
400 * updated to access the inode. This can be done incrementally.
403 zfs_inode_update(znode_t *zp)
408 uint64_t atime[2], mtime[2], ctime[2];
414 sa_lookup(zp->z_sa_hdl, SA_ZPL_ATIME(zsb), &atime, 16);
415 sa_lookup(zp->z_sa_hdl, SA_ZPL_MTIME(zsb), &mtime, 16);
416 sa_lookup(zp->z_sa_hdl, SA_ZPL_CTIME(zsb), &ctime, 16);
418 spin_lock(&ip->i_lock);
419 ip->i_generation = zp->z_gen;
420 ip->i_uid = zp->z_uid;
421 ip->i_gid = zp->z_gid;
422 ip->i_nlink = zp->z_links;
423 ip->i_mode = zp->z_mode;
424 ip->i_blkbits = SPA_MINBLOCKSHIFT;
425 dmu_object_size_from_db(sa_get_db(zp->z_sa_hdl), &blksize,
426 (u_longlong_t *)&ip->i_blocks);
428 ZFS_TIME_DECODE(&ip->i_atime, atime);
429 ZFS_TIME_DECODE(&ip->i_mtime, mtime);
430 ZFS_TIME_DECODE(&ip->i_ctime, ctime);
432 i_size_write(ip, zp->z_size);
433 spin_unlock(&ip->i_lock);
436 static uint64_t empty_xattr;
437 static uint64_t pad[4];
438 static zfs_acl_phys_t acl_phys;
440 * Create a new DMU object to hold a zfs znode.
442 * IN: dzp - parent directory for new znode
443 * vap - file attributes for new znode
444 * tx - dmu transaction id for zap operations
445 * cr - credentials of caller
447 * IS_ROOT_NODE - new object will be root
448 * IS_XATTR - new object is an attribute
449 * bonuslen - length of bonus buffer
450 * setaclp - File/Dir initial ACL
451 * fuidp - Tracks fuid allocation.
453 * OUT: zpp - allocated znode
457 zfs_mknode(znode_t *dzp, vattr_t *vap, dmu_tx_t *tx, cred_t *cr,
458 uint_t flag, znode_t **zpp, zfs_acl_ids_t *acl_ids)
460 uint64_t crtime[2], atime[2], mtime[2], ctime[2];
461 uint64_t mode, size, links, parent, pflags;
462 uint64_t dzp_pflags = 0;
464 zfs_sb_t *zsb = ZTOZSB(dzp);
471 dmu_object_type_t obj_type;
472 sa_bulk_attr_t *sa_attrs;
474 zfs_acl_locator_cb_t locate = { 0 };
477 obj = vap->va_nodeid;
478 now = vap->va_ctime; /* see zfs_replay_create() */
479 gen = vap->va_nblocks; /* ditto */
483 gen = dmu_tx_get_txg(tx);
486 obj_type = zsb->z_use_sa ? DMU_OT_SA : DMU_OT_ZNODE;
487 bonuslen = (obj_type == DMU_OT_SA) ?
488 DN_MAX_BONUSLEN : ZFS_OLD_ZNODE_PHYS_SIZE;
491 * Create a new DMU object.
494 * There's currently no mechanism for pre-reading the blocks that will
495 * be needed to allocate a new object, so we accept the small chance
496 * that there will be an i/o error and we will fail one of the
499 if (S_ISDIR(vap->va_mode)) {
501 err = zap_create_claim_norm(zsb->z_os, obj,
502 zsb->z_norm, DMU_OT_DIRECTORY_CONTENTS,
503 obj_type, bonuslen, tx);
504 ASSERT3U(err, ==, 0);
506 obj = zap_create_norm(zsb->z_os,
507 zsb->z_norm, DMU_OT_DIRECTORY_CONTENTS,
508 obj_type, bonuslen, tx);
512 err = dmu_object_claim(zsb->z_os, obj,
513 DMU_OT_PLAIN_FILE_CONTENTS, 0,
514 obj_type, bonuslen, tx);
515 ASSERT3U(err, ==, 0);
517 obj = dmu_object_alloc(zsb->z_os,
518 DMU_OT_PLAIN_FILE_CONTENTS, 0,
519 obj_type, bonuslen, tx);
523 ZFS_OBJ_HOLD_ENTER(zsb, obj);
524 VERIFY(0 == sa_buf_hold(zsb->z_os, obj, NULL, &db));
527 * If this is the root, fix up the half-initialized parent pointer
528 * to reference the just-allocated physical data area.
530 if (flag & IS_ROOT_NODE) {
533 dzp_pflags = dzp->z_pflags;
537 * If parent is an xattr, so am I.
539 if (dzp_pflags & ZFS_XATTR) {
543 if (zsb->z_use_fuids)
544 pflags = ZFS_ARCHIVE | ZFS_AV_MODIFIED;
548 if (S_ISDIR(vap->va_mode)) {
549 size = 2; /* contents ("." and "..") */
550 links = (flag & (IS_ROOT_NODE | IS_XATTR)) ? 2 : 1;
555 if (S_ISBLK(vap->va_mode) || S_ISCHR(vap->va_mode) ||
556 S_ISFIFO(vap->va_mode) || S_ISSOCK(vap->va_mode))
560 mode = acl_ids->z_mode;
565 * No execs denied will be deterimed when zfs_mode_compute() is called.
567 pflags |= acl_ids->z_aclp->z_hints &
568 (ZFS_ACL_TRIVIAL|ZFS_INHERIT_ACE|ZFS_ACL_AUTO_INHERIT|
569 ZFS_ACL_DEFAULTED|ZFS_ACL_PROTECTED);
571 ZFS_TIME_ENCODE(&now, crtime);
572 ZFS_TIME_ENCODE(&now, ctime);
574 if (vap->va_mask & ATTR_ATIME) {
575 ZFS_TIME_ENCODE(&vap->va_atime, atime);
577 ZFS_TIME_ENCODE(&now, atime);
580 if (vap->va_mask & ATTR_MTIME) {
581 ZFS_TIME_ENCODE(&vap->va_mtime, mtime);
583 ZFS_TIME_ENCODE(&now, mtime);
586 /* Now add in all of the "SA" attributes */
587 VERIFY(0 == sa_handle_get_from_db(zsb->z_os, db, NULL, SA_HDL_SHARED,
591 * Setup the array of attributes to be replaced/set on the new file
593 * order for DMU_OT_ZNODE is critical since it needs to be constructed
594 * in the old znode_phys_t format. Don't change this ordering
596 sa_attrs = kmem_alloc(sizeof(sa_bulk_attr_t) * ZPL_END, KM_SLEEP);
598 if (obj_type == DMU_OT_ZNODE) {
599 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ATIME(zsb),
601 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MTIME(zsb),
603 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CTIME(zsb),
605 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CRTIME(zsb),
607 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GEN(zsb),
609 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MODE(zsb),
611 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_SIZE(zsb),
613 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PARENT(zsb),
616 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MODE(zsb),
618 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_SIZE(zsb),
620 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GEN(zsb),
622 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_UID(zsb),
623 NULL, &acl_ids->z_fuid, 8);
624 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GID(zsb),
625 NULL, &acl_ids->z_fgid, 8);
626 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PARENT(zsb),
628 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_FLAGS(zsb),
630 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ATIME(zsb),
632 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MTIME(zsb),
634 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CTIME(zsb),
636 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CRTIME(zsb),
640 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_LINKS(zsb), NULL, &links, 8);
642 if (obj_type == DMU_OT_ZNODE) {
643 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_XATTR(zsb), NULL,
646 if (obj_type == DMU_OT_ZNODE ||
647 (S_ISBLK(vap->va_mode) || S_ISCHR(vap->va_mode) ||
648 S_ISFIFO(vap->va_mode) || S_ISSOCK(vap->va_mode))) {
649 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_RDEV(zsb),
652 if (obj_type == DMU_OT_ZNODE) {
653 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_FLAGS(zsb),
655 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_UID(zsb), NULL,
656 &acl_ids->z_fuid, 8);
657 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GID(zsb), NULL,
658 &acl_ids->z_fgid, 8);
659 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PAD(zsb), NULL, pad,
660 sizeof (uint64_t) * 4);
661 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ZNODE_ACL(zsb), NULL,
662 &acl_phys, sizeof (zfs_acl_phys_t));
663 } else if (acl_ids->z_aclp->z_version >= ZFS_ACL_VERSION_FUID) {
664 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_DACL_COUNT(zsb), NULL,
665 &acl_ids->z_aclp->z_acl_count, 8);
666 locate.cb_aclp = acl_ids->z_aclp;
667 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_DACL_ACES(zsb),
668 zfs_acl_data_locator, &locate,
669 acl_ids->z_aclp->z_acl_bytes);
670 mode = zfs_mode_compute(mode, acl_ids->z_aclp, &pflags,
671 acl_ids->z_fuid, acl_ids->z_fgid);
674 VERIFY(sa_replace_all_by_template(sa_hdl, sa_attrs, cnt, tx) == 0);
676 if (!(flag & IS_ROOT_NODE)) {
677 *zpp = zfs_znode_alloc(zsb, db, 0, obj_type, obj, sa_hdl);
678 ASSERT(*zpp != NULL);
680 err = zpl_xattr_security_init(ZTOI(*zpp), ZTOI(dzp));
681 ASSERT3S(err, ==, 0);
684 * If we are creating the root node, the "parent" we
685 * passed in is the znode for the root.
689 (*zpp)->z_sa_hdl = sa_hdl;
692 (*zpp)->z_pflags = pflags;
693 (*zpp)->z_mode = mode;
695 if (obj_type == DMU_OT_ZNODE ||
696 acl_ids->z_aclp->z_version < ZFS_ACL_VERSION_FUID) {
697 err = zfs_aclset_common(*zpp, acl_ids->z_aclp, cr, tx);
698 ASSERT3S(err, ==, 0);
700 kmem_free(sa_attrs, sizeof(sa_bulk_attr_t) * ZPL_END);
701 ZFS_OBJ_HOLD_EXIT(zsb, obj);
705 zfs_zget(zfs_sb_t *zsb, uint64_t obj_num, znode_t **zpp)
707 dmu_object_info_t doi;
715 ZFS_OBJ_HOLD_ENTER(zsb, obj_num);
717 err = sa_buf_hold(zsb->z_os, obj_num, NULL, &db);
719 ZFS_OBJ_HOLD_EXIT(zsb, obj_num);
723 dmu_object_info_from_db(db, &doi);
724 if (doi.doi_bonus_type != DMU_OT_SA &&
725 (doi.doi_bonus_type != DMU_OT_ZNODE ||
726 (doi.doi_bonus_type == DMU_OT_ZNODE &&
727 doi.doi_bonus_size < sizeof (znode_phys_t)))) {
728 sa_buf_rele(db, NULL);
729 ZFS_OBJ_HOLD_EXIT(zsb, obj_num);
733 hdl = dmu_buf_get_user(db);
735 zp = sa_get_userdata(hdl);
739 * Since "SA" does immediate eviction we
740 * should never find a sa handle that doesn't
741 * know about the znode.
744 ASSERT3P(zp, !=, NULL);
746 mutex_enter(&zp->z_lock);
747 ASSERT3U(zp->z_id, ==, obj_num);
748 if (zp->z_unlinked) {
755 sa_buf_rele(db, NULL);
756 mutex_exit(&zp->z_lock);
757 ZFS_OBJ_HOLD_EXIT(zsb, obj_num);
762 * Not found create new znode/vnode but only if file exists.
764 * There is a small window where zfs_vget() could
765 * find this object while a file create is still in
766 * progress. This is checked for in zfs_znode_alloc()
768 * if zfs_znode_alloc() fails it will drop the hold on the
771 zp = zfs_znode_alloc(zsb, db, doi.doi_data_block_size,
772 doi.doi_bonus_type, obj_num, NULL);
778 ZFS_OBJ_HOLD_EXIT(zsb, obj_num);
783 zfs_rezget(znode_t *zp)
785 zfs_sb_t *zsb = ZTOZSB(zp);
786 dmu_object_info_t doi;
788 uint64_t obj_num = zp->z_id;
790 sa_bulk_attr_t bulk[8];
795 ZFS_OBJ_HOLD_ENTER(zsb, obj_num);
797 mutex_enter(&zp->z_acl_lock);
798 if (zp->z_acl_cached) {
799 zfs_acl_free(zp->z_acl_cached);
800 zp->z_acl_cached = NULL;
803 mutex_exit(&zp->z_acl_lock);
804 ASSERT(zp->z_sa_hdl == NULL);
805 err = sa_buf_hold(zsb->z_os, obj_num, NULL, &db);
807 ZFS_OBJ_HOLD_EXIT(zsb, obj_num);
811 dmu_object_info_from_db(db, &doi);
812 if (doi.doi_bonus_type != DMU_OT_SA &&
813 (doi.doi_bonus_type != DMU_OT_ZNODE ||
814 (doi.doi_bonus_type == DMU_OT_ZNODE &&
815 doi.doi_bonus_size < sizeof (znode_phys_t)))) {
816 sa_buf_rele(db, NULL);
817 ZFS_OBJ_HOLD_EXIT(zsb, obj_num);
821 zfs_znode_sa_init(zsb, zp, db, doi.doi_bonus_type, NULL);
823 /* reload cached values */
824 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GEN(zsb), NULL,
826 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zsb), NULL,
827 &zp->z_size, sizeof (zp->z_size));
828 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zsb), NULL,
829 &zp->z_links, sizeof (zp->z_links));
830 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zsb), NULL,
831 &zp->z_pflags, sizeof (zp->z_pflags));
832 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zsb), NULL,
833 &zp->z_atime, sizeof (zp->z_atime));
834 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zsb), NULL,
835 &zp->z_uid, sizeof (zp->z_uid));
836 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zsb), NULL,
837 &zp->z_gid, sizeof (zp->z_gid));
838 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zsb), NULL,
839 &mode, sizeof (mode));
841 if (sa_bulk_lookup(zp->z_sa_hdl, bulk, count)) {
842 zfs_znode_dmu_fini(zp);
843 ZFS_OBJ_HOLD_EXIT(zsb, obj_num);
849 if (gen != zp->z_gen) {
850 zfs_znode_dmu_fini(zp);
851 ZFS_OBJ_HOLD_EXIT(zsb, obj_num);
855 zp->z_unlinked = (zp->z_links == 0);
856 zp->z_blksz = doi.doi_data_block_size;
858 ZFS_OBJ_HOLD_EXIT(zsb, obj_num);
864 zfs_znode_delete(znode_t *zp, dmu_tx_t *tx)
866 zfs_sb_t *zsb = ZTOZSB(zp);
867 objset_t *os = zsb->z_os;
868 uint64_t obj = zp->z_id;
869 uint64_t acl_obj = zfs_external_acl(zp);
871 ZFS_OBJ_HOLD_ENTER(zsb, obj);
873 VERIFY(!zp->z_is_sa);
874 VERIFY(0 == dmu_object_free(os, acl_obj, tx));
876 VERIFY(0 == dmu_object_free(os, obj, tx));
877 zfs_znode_dmu_fini(zp);
878 ZFS_OBJ_HOLD_EXIT(zsb, obj);
882 zfs_zinactive(znode_t *zp)
884 zfs_sb_t *zsb = ZTOZSB(zp);
885 uint64_t z_id = zp->z_id;
887 ASSERT(zp->z_sa_hdl);
890 * Don't allow a zfs_zget() while were trying to release this znode
892 ZFS_OBJ_HOLD_ENTER(zsb, z_id);
893 mutex_enter(&zp->z_lock);
896 * If this was the last reference to a file with no links,
897 * remove the file from the file system.
899 if (zp->z_unlinked) {
900 mutex_exit(&zp->z_lock);
901 ZFS_OBJ_HOLD_EXIT(zsb, z_id);
906 mutex_exit(&zp->z_lock);
907 zfs_znode_dmu_fini(zp);
908 ZFS_OBJ_HOLD_EXIT(zsb, z_id);
912 zfs_tstamp_update_setup(znode_t *zp, uint_t flag, uint64_t mtime[2],
913 uint64_t ctime[2], boolean_t have_tx)
919 if (have_tx) { /* will sa_bulk_update happen really soon? */
920 zp->z_atime_dirty = 0;
923 zp->z_atime_dirty = 1;
926 if (flag & ATTR_ATIME) {
927 ZFS_TIME_ENCODE(&now, zp->z_atime);
930 if (flag & ATTR_MTIME) {
931 ZFS_TIME_ENCODE(&now, mtime);
932 if (ZTOZSB(zp)->z_use_fuids) {
933 zp->z_pflags |= (ZFS_ARCHIVE |
938 if (flag & ATTR_CTIME) {
939 ZFS_TIME_ENCODE(&now, ctime);
940 if (ZTOZSB(zp)->z_use_fuids)
941 zp->z_pflags |= ZFS_ARCHIVE;
946 * Grow the block size for a file.
948 * IN: zp - znode of file to free data in.
949 * size - requested block size
950 * tx - open transaction.
952 * NOTE: this function assumes that the znode is write locked.
955 zfs_grow_blocksize(znode_t *zp, uint64_t size, dmu_tx_t *tx)
960 if (size <= zp->z_blksz)
963 * If the file size is already greater than the current blocksize,
964 * we will not grow. If there is more than one block in a file,
965 * the blocksize cannot change.
967 if (zp->z_blksz && zp->z_size > zp->z_blksz)
970 error = dmu_object_set_blocksize(ZTOZSB(zp)->z_os, zp->z_id,
973 if (error == ENOTSUP)
975 ASSERT3U(error, ==, 0);
977 /* What blocksize did we actually get? */
978 dmu_object_size_from_db(sa_get_db(zp->z_sa_hdl), &zp->z_blksz, &dummy);
983 * This is a dummy interface used when pvn_vplist_dirty() should *not*
984 * be calling back into the fs for a putpage(). E.g.: when truncating
985 * a file, the pages being "thrown away* don't need to be written out.
989 zfs_no_putpage(vnode_t *vp, page_t *pp, u_offset_t *offp, size_t *lenp,
990 int flags, cred_t *cr)
995 #endif /* HAVE_MMAP */
998 * Increase the file length
1000 * IN: zp - znode of file to free data in.
1001 * end - new end-of-file
1003 * RETURN: 0 if success
1004 * error code if failure
1007 zfs_extend(znode_t *zp, uint64_t end)
1009 zfs_sb_t *zsb = ZTOZSB(zp);
1016 * We will change zp_size, lock the whole file.
1018 rl = zfs_range_lock(zp, 0, UINT64_MAX, RL_WRITER);
1021 * Nothing to do if file already at desired length.
1023 if (end <= zp->z_size) {
1024 zfs_range_unlock(rl);
1028 tx = dmu_tx_create(zsb->z_os);
1029 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
1030 zfs_sa_upgrade_txholds(tx, zp);
1031 if (end > zp->z_blksz &&
1032 (!ISP2(zp->z_blksz) || zp->z_blksz < zsb->z_max_blksz)) {
1034 * We are growing the file past the current block size.
1036 if (zp->z_blksz > ZTOZSB(zp)->z_max_blksz) {
1037 ASSERT(!ISP2(zp->z_blksz));
1038 newblksz = MIN(end, SPA_MAXBLOCKSIZE);
1040 newblksz = MIN(end, ZTOZSB(zp)->z_max_blksz);
1042 dmu_tx_hold_write(tx, zp->z_id, 0, newblksz);
1047 error = dmu_tx_assign(tx, TXG_NOWAIT);
1049 if (error == ERESTART) {
1055 zfs_range_unlock(rl);
1060 zfs_grow_blocksize(zp, newblksz, tx);
1064 VERIFY(0 == sa_update(zp->z_sa_hdl, SA_ZPL_SIZE(ZTOZSB(zp)),
1065 &zp->z_size, sizeof (zp->z_size), tx));
1067 zfs_range_unlock(rl);
1075 * Free space in a file.
1077 * IN: zp - znode of file to free data in.
1078 * off - start of section to free.
1079 * len - length of section to free.
1081 * RETURN: 0 if success
1082 * error code if failure
1085 zfs_free_range(znode_t *zp, uint64_t off, uint64_t len)
1087 zfs_sb_t *zsb = ZTOZSB(zp);
1092 * Lock the range being freed.
1094 rl = zfs_range_lock(zp, off, len, RL_WRITER);
1097 * Nothing to do if file already at desired length.
1099 if (off >= zp->z_size) {
1100 zfs_range_unlock(rl);
1104 if (off + len > zp->z_size)
1105 len = zp->z_size - off;
1107 error = dmu_free_long_range(zsb->z_os, zp->z_id, off, len);
1109 zfs_range_unlock(rl);
1117 * IN: zp - znode of file to free data in.
1118 * end - new end-of-file.
1120 * RETURN: 0 if success
1121 * error code if failure
1124 zfs_trunc(znode_t *zp, uint64_t end)
1126 zfs_sb_t *zsb = ZTOZSB(zp);
1130 sa_bulk_attr_t bulk[2];
1134 * We will change zp_size, lock the whole file.
1136 rl = zfs_range_lock(zp, 0, UINT64_MAX, RL_WRITER);
1139 * Nothing to do if file already at desired length.
1141 if (end >= zp->z_size) {
1142 zfs_range_unlock(rl);
1146 error = dmu_free_long_range(zsb->z_os, zp->z_id, end, -1);
1148 zfs_range_unlock(rl);
1152 tx = dmu_tx_create(zsb->z_os);
1153 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
1154 zfs_sa_upgrade_txholds(tx, zp);
1155 error = dmu_tx_assign(tx, TXG_NOWAIT);
1157 if (error == ERESTART) {
1163 zfs_range_unlock(rl);
1168 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zsb),
1169 NULL, &zp->z_size, sizeof (zp->z_size));
1172 zp->z_pflags &= ~ZFS_SPARSE;
1173 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zsb),
1174 NULL, &zp->z_pflags, 8);
1176 VERIFY(sa_bulk_update(zp->z_sa_hdl, bulk, count, tx) == 0);
1180 zfs_range_unlock(rl);
1186 * Free space in a file
1188 * IN: zp - znode of file to free data in.
1189 * off - start of range
1190 * len - end of range (0 => EOF)
1191 * flag - current file open mode flags.
1192 * log - TRUE if this action should be logged
1194 * RETURN: 0 if success
1195 * error code if failure
1198 zfs_freesp(znode_t *zp, uint64_t off, uint64_t len, int flag, boolean_t log)
1200 #ifdef HAVE_MANDLOCKS
1201 struct inode *ip = ZTOI(zp);
1202 #endif /* HAVE_MANDLOCKS */
1204 zfs_sb_t *zsb = ZTOZSB(zp);
1205 zilog_t *zilog = zsb->z_log;
1207 uint64_t mtime[2], ctime[2];
1208 sa_bulk_attr_t bulk[3];
1212 if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_MODE(zsb), &mode,
1213 sizeof (mode))) != 0)
1216 if (off > zp->z_size) {
1217 error = zfs_extend(zp, off+len);
1218 if (error == 0 && log)
1224 #ifdef HAVE_MANDLOCKS
1226 * Check for any locks in the region to be freed.
1229 if (MANDLOCK(ip, (mode_t)mode)) {
1230 uint64_t length = (len ? len : zp->z_size - off);
1231 if (error = chklock(ip, FWRITE, off, length, flag, NULL))
1234 #endif /* HAVE_MANDLOCKS */
1237 error = zfs_trunc(zp, off);
1239 if ((error = zfs_free_range(zp, off, len)) == 0 &&
1240 off + len > zp->z_size)
1241 error = zfs_extend(zp, off+len);
1246 tx = dmu_tx_create(zsb->z_os);
1247 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
1248 zfs_sa_upgrade_txholds(tx, zp);
1249 error = dmu_tx_assign(tx, TXG_NOWAIT);
1251 if (error == ERESTART) {
1260 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zsb), NULL, mtime, 16);
1261 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zsb), NULL, ctime, 16);
1262 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zsb),
1263 NULL, &zp->z_pflags, 8);
1264 zfs_tstamp_update_setup(zp, CONTENT_MODIFIED, mtime, ctime, B_TRUE);
1265 error = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx);
1268 zfs_log_truncate(zilog, tx, TX_TRUNCATE, zp, off, len);
1271 zfs_inode_update(zp);
1276 zfs_create_fs(objset_t *os, cred_t *cr, nvlist_t *zplprops, dmu_tx_t *tx)
1278 uint64_t moid, obj, sa_obj, version;
1287 * First attempt to create master node.
1290 * In an empty objset, there are no blocks to read and thus
1291 * there can be no i/o errors (which we assert below).
1293 moid = MASTER_NODE_OBJ;
1294 error = zap_create_claim(os, moid, DMU_OT_MASTER_NODE,
1295 DMU_OT_NONE, 0, tx);
1299 * Set starting attributes.
1301 version = zfs_zpl_version_map(spa_version(dmu_objset_spa(os)));
1303 while ((elem = nvlist_next_nvpair(zplprops, elem)) != NULL) {
1304 /* For the moment we expect all zpl props to be uint64_ts */
1308 ASSERT(nvpair_type(elem) == DATA_TYPE_UINT64);
1309 VERIFY(nvpair_value_uint64(elem, &val) == 0);
1310 name = nvpair_name(elem);
1311 if (strcmp(name, zfs_prop_to_name(ZFS_PROP_VERSION)) == 0) {
1315 error = zap_update(os, moid, name, 8, 1, &val, tx);
1318 if (strcmp(name, zfs_prop_to_name(ZFS_PROP_NORMALIZE)) == 0)
1321 ASSERT(version != 0);
1322 error = zap_update(os, moid, ZPL_VERSION_STR, 8, 1, &version, tx);
1325 * Create zap object used for SA attribute registration
1328 if (version >= ZPL_VERSION_SA) {
1329 sa_obj = zap_create(os, DMU_OT_SA_MASTER_NODE,
1330 DMU_OT_NONE, 0, tx);
1331 error = zap_add(os, moid, ZFS_SA_ATTRS, 8, 1, &sa_obj, tx);
1337 * Create a delete queue.
1339 obj = zap_create(os, DMU_OT_UNLINKED_SET, DMU_OT_NONE, 0, tx);
1341 error = zap_add(os, moid, ZFS_UNLINKED_SET, 8, 1, &obj, tx);
1345 * Create root znode with code free of VFS dependencies. This
1346 * is important because without a registered filesystem and super
1347 * block all the required VFS hooks will be missing. The critical
1348 * thing is to just crete the required root znode.
1350 obj = zap_create_norm(os, norm, DMU_OT_DIRECTORY_CONTENTS,
1351 DMU_OT_ZNODE, sizeof (znode_phys_t), tx);
1353 VERIFY(0 == dmu_bonus_hold(os, obj, FTAG, &db));
1354 dmu_buf_will_dirty(db, tx);
1357 * Initialize the znode physical data to zero.
1359 ASSERT(db->db_size >= sizeof (znode_phys_t));
1360 bzero(db->db_data, db->db_size);
1363 if (USE_FUIDS(version, os))
1364 pzp->zp_flags = ZFS_ARCHIVE | ZFS_AV_MODIFIED;
1366 pzp->zp_size = 2; /* "." and ".." */
1368 pzp->zp_parent = obj;
1369 pzp->zp_gen = dmu_tx_get_txg(tx);
1370 pzp->zp_mode = S_IFDIR | 0755;
1371 pzp->zp_flags = ZFS_ACL_TRIVIAL;
1375 ZFS_TIME_ENCODE(&now, pzp->zp_crtime);
1376 ZFS_TIME_ENCODE(&now, pzp->zp_ctime);
1377 ZFS_TIME_ENCODE(&now, pzp->zp_atime);
1378 ZFS_TIME_ENCODE(&now, pzp->zp_mtime);
1380 error = zap_add(os, moid, ZFS_ROOT_OBJ, 8, 1, &obj, tx);
1383 dmu_buf_rele(db, FTAG);
1386 #endif /* _KERNEL */
1389 zfs_sa_setup(objset_t *osp, sa_attr_type_t **sa_table)
1391 uint64_t sa_obj = 0;
1394 error = zap_lookup(osp, MASTER_NODE_OBJ, ZFS_SA_ATTRS, 8, 1, &sa_obj);
1395 if (error != 0 && error != ENOENT)
1398 error = sa_setup(osp, sa_obj, zfs_attr_table, ZPL_END, sa_table);
1403 zfs_grab_sa_handle(objset_t *osp, uint64_t obj, sa_handle_t **hdlp,
1406 dmu_object_info_t doi;
1409 if ((error = sa_buf_hold(osp, obj, FTAG, db)) != 0)
1412 dmu_object_info_from_db(*db, &doi);
1413 if ((doi.doi_bonus_type != DMU_OT_SA &&
1414 doi.doi_bonus_type != DMU_OT_ZNODE) ||
1415 (doi.doi_bonus_type == DMU_OT_ZNODE &&
1416 doi.doi_bonus_size < sizeof (znode_phys_t))) {
1417 sa_buf_rele(*db, FTAG);
1421 error = sa_handle_get(osp, obj, NULL, SA_HDL_PRIVATE, hdlp);
1423 sa_buf_rele(*db, FTAG);
1431 zfs_release_sa_handle(sa_handle_t *hdl, dmu_buf_t *db)
1433 sa_handle_destroy(hdl);
1434 sa_buf_rele(db, FTAG);
1438 * Given an object number, return its parent object number and whether
1439 * or not the object is an extended attribute directory.
1442 zfs_obj_to_pobj(sa_handle_t *hdl, sa_attr_type_t *sa_table, uint64_t *pobjp,
1448 sa_bulk_attr_t bulk[3];
1452 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_PARENT], NULL,
1453 &parent, sizeof (parent));
1454 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_FLAGS], NULL,
1455 &pflags, sizeof (pflags));
1456 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_MODE], NULL,
1457 &mode, sizeof (mode));
1459 if ((error = sa_bulk_lookup(hdl, bulk, count)) != 0)
1463 *is_xattrdir = ((pflags & ZFS_XATTR) != 0) && S_ISDIR(mode);
1469 * Given an object number, return some zpl level statistics
1472 zfs_obj_to_stats_impl(sa_handle_t *hdl, sa_attr_type_t *sa_table,
1475 sa_bulk_attr_t bulk[4];
1478 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_MODE], NULL,
1479 &sb->zs_mode, sizeof (sb->zs_mode));
1480 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_GEN], NULL,
1481 &sb->zs_gen, sizeof (sb->zs_gen));
1482 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_LINKS], NULL,
1483 &sb->zs_links, sizeof (sb->zs_links));
1484 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_CTIME], NULL,
1485 &sb->zs_ctime, sizeof (sb->zs_ctime));
1487 return (sa_bulk_lookup(hdl, bulk, count));
1491 zfs_obj_to_path_impl(objset_t *osp, uint64_t obj, sa_handle_t *hdl,
1492 sa_attr_type_t *sa_table, char *buf, int len)
1494 sa_handle_t *sa_hdl;
1495 sa_handle_t *prevhdl = NULL;
1496 dmu_buf_t *prevdb = NULL;
1497 dmu_buf_t *sa_db = NULL;
1498 char *path = buf + len - 1;
1506 char component[MAXNAMELEN + 2];
1511 zfs_release_sa_handle(prevhdl, prevdb);
1513 if ((error = zfs_obj_to_pobj(sa_hdl, sa_table, &pobj,
1514 &is_xattrdir)) != 0)
1525 (void) sprintf(component + 1, "<xattrdir>");
1527 error = zap_value_search(osp, pobj, obj,
1528 ZFS_DIRENT_OBJ(-1ULL), component + 1);
1533 complen = strlen(component);
1535 ASSERT(path >= buf);
1536 bcopy(component, path, complen);
1539 if (sa_hdl != hdl) {
1543 error = zfs_grab_sa_handle(osp, obj, &sa_hdl, &sa_db);
1551 if (sa_hdl != NULL && sa_hdl != hdl) {
1552 ASSERT(sa_db != NULL);
1553 zfs_release_sa_handle(sa_hdl, sa_db);
1557 (void) memmove(buf, path, buf + len - path);
1563 zfs_obj_to_path(objset_t *osp, uint64_t obj, char *buf, int len)
1565 sa_attr_type_t *sa_table;
1570 error = zfs_sa_setup(osp, &sa_table);
1574 error = zfs_grab_sa_handle(osp, obj, &hdl, &db);
1578 error = zfs_obj_to_path_impl(osp, obj, hdl, sa_table, buf, len);
1580 zfs_release_sa_handle(hdl, db);
1585 zfs_obj_to_stats(objset_t *osp, uint64_t obj, zfs_stat_t *sb,
1588 char *path = buf + len - 1;
1589 sa_attr_type_t *sa_table;
1596 error = zfs_sa_setup(osp, &sa_table);
1600 error = zfs_grab_sa_handle(osp, obj, &hdl, &db);
1604 error = zfs_obj_to_stats_impl(hdl, sa_table, sb);
1606 zfs_release_sa_handle(hdl, db);
1610 error = zfs_obj_to_path_impl(osp, obj, hdl, sa_table, buf, len);
1612 zfs_release_sa_handle(hdl, db);
1616 #if defined(_KERNEL) && defined(HAVE_SPL)
1617 EXPORT_SYMBOL(zfs_create_fs);
1618 EXPORT_SYMBOL(zfs_obj_to_path);