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 2007 Sun Microsystems, Inc. All rights reserved.
23 * Use is subject to license terms.
26 /* Portions Copyright 2007 Jeremy Teo */
28 #pragma ident "%Z%%M% %I% %E% SMI"
31 #include <sys/types.h>
32 #include <sys/param.h>
34 #include <sys/systm.h>
35 #include <sys/sysmacros.h>
36 #include <sys/resource.h>
37 #include <sys/mntent.h>
39 #include <sys/vnode.h>
42 #include <sys/cmn_err.h>
43 #include <sys/errno.h>
44 #include <sys/unistd.h>
45 #include <sys/atomic.h>
46 #include <sys/zfs_dir.h>
47 #include <sys/zfs_acl.h>
48 #include <sys/zfs_ioctl.h>
49 #include <sys/zfs_rlock.h>
50 #include <sys/fs/zfs.h>
54 #include <sys/refcount.h>
57 #include <sys/zfs_znode.h>
58 #include <sys/refcount.h>
60 /* Used by fstat(1). */
61 SYSCTL_INT(_debug_sizeof, OID_AUTO, znode, CTLFLAG_RD, 0, sizeof(znode_t),
65 * Functions needed for userland (ie: libzpool) are not put under
66 * #ifdef_KERNEL; the rest of the functions have dependencies
67 * (such as VFS logic) that will not compile easily in userland.
70 struct kmem_cache *znode_cache = NULL;
74 znode_pageout_func(dmu_buf_t *dbuf, void *user_ptr)
76 znode_t *zp = user_ptr;
79 mutex_enter(&zp->z_lock);
82 mutex_exit(&zp->z_lock);
84 } else if (vp->v_count == 0) {
87 mutex_exit(&zp->z_lock);
88 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
89 vrecycle(vp, curthread);
94 /* signal force unmount that this znode can be freed */
96 mutex_exit(&zp->z_lock);
100 extern struct vop_vector zfs_vnodeops;
101 extern struct vop_vector zfs_fifoops;
104 * XXX: We cannot use this function as a cache constructor, because
105 * there is one global cache for all file systems and we need
106 * to pass vfsp here, which is not possible, because argument
107 * 'cdrarg' is defined at kmem_cache_create() time.
110 zfs_znode_cache_constructor(void *buf, void *cdrarg, int kmflags)
114 vfs_t *vfsp = cdrarg;
117 if (cdrarg != NULL) {
118 error = getnewvnode("zfs", vfsp, &zfs_vnodeops, &vp);
120 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
122 vp->v_data = (caddr_t)zp;
128 mutex_init(&zp->z_lock, NULL, MUTEX_DEFAULT, NULL);
129 rw_init(&zp->z_map_lock, NULL, RW_DEFAULT, NULL);
130 rw_init(&zp->z_parent_lock, NULL, RW_DEFAULT, NULL);
131 rw_init(&zp->z_name_lock, NULL, RW_DEFAULT, NULL);
132 mutex_init(&zp->z_acl_lock, NULL, MUTEX_DEFAULT, NULL);
134 mutex_init(&zp->z_range_lock, NULL, MUTEX_DEFAULT, NULL);
135 avl_create(&zp->z_range_avl, zfs_range_compare,
136 sizeof (rl_t), offsetof(rl_t, r_node));
145 zfs_znode_cache_destructor(void *buf, void *cdarg)
149 ASSERT(zp->z_dirlocks == 0);
150 mutex_destroy(&zp->z_lock);
151 rw_destroy(&zp->z_map_lock);
152 rw_destroy(&zp->z_parent_lock);
153 rw_destroy(&zp->z_name_lock);
154 mutex_destroy(&zp->z_acl_lock);
155 mutex_destroy(&zp->z_range_lock);
156 avl_destroy(&zp->z_range_avl);
158 ASSERT(zp->z_dbuf_held == 0);
167 ASSERT(znode_cache == NULL);
168 znode_cache = kmem_cache_create("zfs_znode_cache",
169 sizeof (znode_t), 0, /* zfs_znode_cache_constructor */ NULL,
170 zfs_znode_cache_destructor, NULL, NULL, NULL, 0);
180 kmem_cache_destroy(znode_cache);
185 * zfs_init_fs - Initialize the zfsvfs struct and the file system
186 * incore "master" object. Verify version compatibility.
189 zfs_init_fs(zfsvfs_t *zfsvfs, znode_t **zpp, cred_t *cr)
191 objset_t *os = zfsvfs->z_os;
192 uint64_t version = ZPL_VERSION;
194 dmu_object_info_t doi;
200 * XXX - hack to auto-create the pool root filesystem at
201 * the first attempted mount.
203 if (dmu_object_info(os, MASTER_NODE_OBJ, &doi) == ENOENT) {
204 dmu_tx_t *tx = dmu_tx_create(os);
206 dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, TRUE, NULL); /* master */
207 dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, TRUE, NULL); /* del queue */
208 dmu_tx_hold_bonus(tx, DMU_NEW_OBJECT); /* root node */
209 error = dmu_tx_assign(tx, TXG_WAIT);
210 ASSERT3U(error, ==, 0);
211 zfs_create_fs(os, cr, tx);
215 error = zap_lookup(os, MASTER_NODE_OBJ, ZPL_VERSION_OBJ, 8, 1,
219 } else if (version != ZPL_VERSION) {
220 (void) printf("Mismatched versions: File system "
221 "is version %lld on-disk format, which is "
222 "incompatible with this software version %lld!",
223 (u_longlong_t)version, ZPL_VERSION);
228 * The fsid is 64 bits, composed of an 8-bit fs type, which
229 * separates our fsid from any other filesystem types, and a
230 * 56-bit objset unique ID. The objset unique ID is unique to
231 * all objsets open on this system, provided by unique_create().
232 * The 8-bit fs type must be put in the low bits of fsid[1]
233 * because that's where other Solaris filesystems put it.
235 fsid_guid = dmu_objset_fsid_guid(os);
236 ASSERT((fsid_guid & ~((1ULL<<56)-1)) == 0);
237 zfsvfs->z_vfs->vfs_fsid.val[0] = fsid_guid;
238 zfsvfs->z_vfs->vfs_fsid.val[1] = ((fsid_guid>>32) << 8) |
239 zfsvfs->z_vfs->mnt_vfc->vfc_typenum & 0xFF;
241 error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_ROOT_OBJ, 8, 1,
245 ASSERT(zfsvfs->z_root != 0);
248 * Create the per mount vop tables.
252 * Initialize zget mutex's
254 for (i = 0; i != ZFS_OBJ_MTX_SZ; i++)
255 mutex_init(&zfsvfs->z_hold_mtx[i], NULL, MUTEX_DEFAULT, NULL);
257 error = zfs_zget(zfsvfs, zfsvfs->z_root, zpp);
260 ASSERT3U((*zpp)->z_id, ==, zfsvfs->z_root);
262 error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_UNLINKED_SET, 8, 1,
263 &zfsvfs->z_unlinkedobj);
271 * define a couple of values we need available
272 * for both 64 and 32 bit environments.
275 #define NBITSMINOR64 32
278 #define MAXMAJ64 0xffffffffUL
281 #define MAXMIN64 0xffffffffUL
285 * Create special expldev for ZFS private use.
286 * Can't use standard expldev since it doesn't do
287 * what we want. The standard expldev() takes a
288 * dev32_t in LP64 and expands it to a long dev_t.
289 * We need an interface that takes a dev32_t in ILP32
290 * and expands it to a long dev_t.
293 zfs_expldev(dev_t dev)
295 return (((uint64_t)umajor(dev) << NBITSMINOR64) | uminor(dev));
298 * Special cmpldev for ZFS private use.
299 * Can't use standard cmpldev since it takes
300 * a long dev_t and compresses it to dev32_t in
301 * LP64. We need to do a compaction of a long dev_t
302 * to a dev32_t in ILP32.
305 zfs_cmpldev(uint64_t dev)
307 return (makedev((dev >> NBITSMINOR64), (dev & MAXMIN64)));
311 * Construct a new znode/vnode and intialize.
313 * This does not do a call to dmu_set_user() that is
314 * up to the caller to do, in case you don't want to
318 zfs_znode_alloc(zfsvfs_t *zfsvfs, dmu_buf_t *db, uint64_t obj_num, int blksz)
324 zp = kmem_cache_alloc(znode_cache, KM_SLEEP);
325 zfs_znode_cache_constructor(zp, zfsvfs->z_vfs, 0);
327 ASSERT(zp->z_dirlocks == NULL);
329 zp->z_phys = db->db_data;
330 zp->z_zfsvfs = zfsvfs;
332 zp->z_atime_dirty = 0;
339 zp->z_seq = 0x7A4653;
342 mutex_enter(&zfsvfs->z_znodes_lock);
343 list_insert_tail(&zfsvfs->z_all_znodes, zp);
344 mutex_exit(&zfsvfs->z_znodes_lock);
350 vp->v_vflag |= VV_FORCEINSMQ;
351 error = insmntque(vp, zfsvfs->z_vfs);
352 vp->v_vflag &= ~VV_FORCEINSMQ;
353 KASSERT(error == 0, ("insmntque() failed: error %d", error));
355 vp->v_type = IFTOVT((mode_t)zp->z_phys->zp_mode);
356 switch (vp->v_type) {
358 zp->z_zn_prefetch = B_TRUE; /* z_prefetch default is enabled */
361 vp->v_op = &zfs_fifoops;
369 zfs_znode_dmu_init(znode_t *zp)
372 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
373 dmu_buf_t *db = zp->z_dbuf;
375 mutex_enter(&zp->z_lock);
377 nzp = dmu_buf_set_user_ie(db, zp, &zp->z_phys, znode_pageout_func);
381 * concurrent zgets on this object.
383 ASSERT3P(nzp, ==, NULL);
386 * Slap on VROOT if we are the root znode
388 if (zp->z_id == zfsvfs->z_root) {
389 ZTOV(zp)->v_flag |= VROOT;
392 ASSERT(zp->z_dbuf_held == 0);
394 VFS_HOLD(zfsvfs->z_vfs);
395 mutex_exit(&zp->z_lock);
399 * Create a new DMU object to hold a zfs znode.
401 * IN: dzp - parent directory for new znode
402 * vap - file attributes for new znode
403 * tx - dmu transaction id for zap operations
404 * cr - credentials of caller
406 * IS_ROOT_NODE - new object will be root
407 * IS_XATTR - new object is an attribute
408 * IS_REPLAY - intent log replay
410 * OUT: oid - ID of created object
414 zfs_mknode(znode_t *dzp, vattr_t *vap, uint64_t *oid, dmu_tx_t *tx, cred_t *cr,
415 uint_t flag, znode_t **zpp, int bonuslen)
420 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
425 ASSERT(vap && (vap->va_mask & (AT_TYPE|AT_MODE)) == (AT_TYPE|AT_MODE));
427 if (zfsvfs->z_assign >= TXG_INITIAL) { /* ZIL replay */
428 *oid = vap->va_nodeid;
430 now = vap->va_ctime; /* see zfs_replay_create() */
431 gen = vap->va_nblocks; /* ditto */
435 gen = dmu_tx_get_txg(tx);
439 * Create a new DMU object.
442 * There's currently no mechanism for pre-reading the blocks that will
443 * be to needed allocate a new object, so we accept the small chance
444 * that there will be an i/o error and we will fail one of the
447 if (vap->va_type == VDIR) {
448 if (flag & IS_REPLAY) {
449 err = zap_create_claim(zfsvfs->z_os, *oid,
450 DMU_OT_DIRECTORY_CONTENTS,
451 DMU_OT_ZNODE, sizeof (znode_phys_t) + bonuslen, tx);
452 ASSERT3U(err, ==, 0);
454 *oid = zap_create(zfsvfs->z_os,
455 DMU_OT_DIRECTORY_CONTENTS,
456 DMU_OT_ZNODE, sizeof (znode_phys_t) + bonuslen, tx);
459 if (flag & IS_REPLAY) {
460 err = dmu_object_claim(zfsvfs->z_os, *oid,
461 DMU_OT_PLAIN_FILE_CONTENTS, 0,
462 DMU_OT_ZNODE, sizeof (znode_phys_t) + bonuslen, tx);
463 ASSERT3U(err, ==, 0);
465 *oid = dmu_object_alloc(zfsvfs->z_os,
466 DMU_OT_PLAIN_FILE_CONTENTS, 0,
467 DMU_OT_ZNODE, sizeof (znode_phys_t) + bonuslen, tx);
470 VERIFY(0 == dmu_bonus_hold(zfsvfs->z_os, *oid, NULL, &dbp));
471 dmu_buf_will_dirty(dbp, tx);
474 * Initialize the znode physical data to zero.
476 ASSERT(dbp->db_size >= sizeof (znode_phys_t));
477 bzero(dbp->db_data, dbp->db_size);
481 * If this is the root, fix up the half-initialized parent pointer
482 * to reference the just-allocated physical data area.
484 if (flag & IS_ROOT_NODE) {
490 * If parent is an xattr, so am I.
492 if (dzp->z_phys->zp_flags & ZFS_XATTR)
495 if (vap->va_type == VBLK || vap->va_type == VCHR) {
496 pzp->zp_rdev = zfs_expldev(vap->va_rdev);
499 if (vap->va_type == VDIR) {
500 pzp->zp_size = 2; /* contents ("." and "..") */
501 pzp->zp_links = (flag & (IS_ROOT_NODE | IS_XATTR)) ? 2 : 1;
504 pzp->zp_parent = dzp->z_id;
506 pzp->zp_flags |= ZFS_XATTR;
510 ZFS_TIME_ENCODE(&now, pzp->zp_crtime);
511 ZFS_TIME_ENCODE(&now, pzp->zp_ctime);
513 if (vap->va_mask & AT_ATIME) {
514 ZFS_TIME_ENCODE(&vap->va_atime, pzp->zp_atime);
516 ZFS_TIME_ENCODE(&now, pzp->zp_atime);
519 if (vap->va_mask & AT_MTIME) {
520 ZFS_TIME_ENCODE(&vap->va_mtime, pzp->zp_mtime);
522 ZFS_TIME_ENCODE(&now, pzp->zp_mtime);
525 pzp->zp_mode = MAKEIMODE(vap->va_type, vap->va_mode);
526 zp = zfs_znode_alloc(zfsvfs, dbp, *oid, 0);
528 zfs_perm_init(zp, dzp, flag, vap, tx, cr);
531 kmutex_t *hash_mtx = ZFS_OBJ_MUTEX(zp);
533 mutex_enter(hash_mtx);
534 zfs_znode_dmu_init(zp);
535 mutex_exit(hash_mtx);
539 if (ZTOV(zp) != NULL) {
540 ZTOV(zp)->v_count = 0;
541 VOP_UNLOCK(ZTOV(zp), 0);
543 dmu_buf_rele(dbp, NULL);
549 zfs_zget(zfsvfs_t *zfsvfs, uint64_t obj_num, znode_t **zpp)
551 dmu_object_info_t doi;
559 ZFS_OBJ_HOLD_ENTER(zfsvfs, obj_num);
561 err = dmu_bonus_hold(zfsvfs->z_os, obj_num, NULL, &db);
563 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
567 dmu_object_info_from_db(db, &doi);
568 if (doi.doi_bonus_type != DMU_OT_ZNODE ||
569 doi.doi_bonus_size < sizeof (znode_phys_t)) {
570 dmu_buf_rele(db, NULL);
571 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
575 ASSERT(db->db_object == obj_num);
576 ASSERT(db->db_offset == -1);
577 ASSERT(db->db_data != NULL);
579 zp = dmu_buf_get_user(db);
582 mutex_enter(&zp->z_lock);
584 ASSERT3U(zp->z_id, ==, obj_num);
585 if (zp->z_unlinked) {
586 dmu_buf_rele(db, NULL);
587 mutex_exit(&zp->z_lock);
588 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
590 } else if (zp->z_dbuf_held) {
591 dmu_buf_rele(db, NULL);
594 VFS_HOLD(zfsvfs->z_vfs);
597 if (ZTOV(zp) != NULL)
600 err = getnewvnode("zfs", zfsvfs->z_vfs, &zfs_vnodeops,
604 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
605 vp->v_data = (caddr_t)zp;
608 vp->v_type = IFTOVT((mode_t)zp->z_phys->zp_mode);
609 if (vp->v_type == VDIR)
610 zp->z_zn_prefetch = B_TRUE; /* z_prefetch default is enabled */
611 vp->v_vflag |= VV_FORCEINSMQ;
612 err = insmntque(vp, zfsvfs->z_vfs);
613 vp->v_vflag &= ~VV_FORCEINSMQ;
614 KASSERT(err == 0, ("insmntque() failed: error %d", err));
617 mutex_exit(&zp->z_lock);
618 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
624 * Not found create new znode/vnode
626 zp = zfs_znode_alloc(zfsvfs, db, obj_num, doi.doi_data_block_size);
627 ASSERT3U(zp->z_id, ==, obj_num);
628 zfs_znode_dmu_init(zp);
629 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
631 if ((vp = ZTOV(zp)) != NULL)
637 zfs_znode_delete(znode_t *zp, dmu_tx_t *tx)
639 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
642 ZFS_OBJ_HOLD_ENTER(zfsvfs, zp->z_id);
643 if (zp->z_phys->zp_acl.z_acl_extern_obj) {
644 error = dmu_object_free(zfsvfs->z_os,
645 zp->z_phys->zp_acl.z_acl_extern_obj, tx);
646 ASSERT3U(error, ==, 0);
648 error = dmu_object_free(zfsvfs->z_os, zp->z_id, tx);
649 ASSERT3U(error, ==, 0);
651 ZFS_OBJ_HOLD_EXIT(zfsvfs, zp->z_id);
652 dmu_buf_rele(zp->z_dbuf, NULL);
656 zfs_zinactive(znode_t *zp)
658 vnode_t *vp = ZTOV(zp);
659 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
660 uint64_t z_id = zp->z_id;
662 ASSERT(zp->z_dbuf_held && zp->z_phys);
665 * Don't allow a zfs_zget() while were trying to release this znode
667 ZFS_OBJ_HOLD_ENTER(zfsvfs, z_id);
669 mutex_enter(&zp->z_lock);
671 if (vp->v_count > 0) {
673 * If the hold count is greater than zero, somebody has
674 * obtained a new reference on this znode while we were
675 * processing it here, so we are done.
678 mutex_exit(&zp->z_lock);
679 ZFS_OBJ_HOLD_EXIT(zfsvfs, z_id);
685 * If this was the last reference to a file with no links,
686 * remove the file from the file system.
688 if (zp->z_unlinked) {
690 mutex_exit(&zp->z_lock);
691 ZFS_OBJ_HOLD_EXIT(zfsvfs, z_id);
692 ASSERT(vp->v_count == 0);
693 vrecycle(vp, curthread);
695 VFS_RELE(zfsvfs->z_vfs);
699 ASSERT(zp->z_dbuf_held);
700 mutex_exit(&zp->z_lock);
701 ZFS_OBJ_HOLD_EXIT(zfsvfs, z_id);
705 zfs_znode_free(znode_t *zp)
707 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
709 mutex_enter(&zfsvfs->z_znodes_lock);
710 list_remove(&zfsvfs->z_all_znodes, zp);
711 mutex_exit(&zfsvfs->z_znodes_lock);
713 kmem_cache_free(znode_cache, zp);
717 zfs_time_stamper_locked(znode_t *zp, uint_t flag, dmu_tx_t *tx)
721 ASSERT(MUTEX_HELD(&zp->z_lock));
726 dmu_buf_will_dirty(zp->z_dbuf, tx);
727 zp->z_atime_dirty = 0;
730 zp->z_atime_dirty = 1;
734 ZFS_TIME_ENCODE(&now, zp->z_phys->zp_atime);
737 ZFS_TIME_ENCODE(&now, zp->z_phys->zp_mtime);
740 ZFS_TIME_ENCODE(&now, zp->z_phys->zp_ctime);
744 * Update the requested znode timestamps with the current time.
745 * If we are in a transaction, then go ahead and mark the znode
746 * dirty in the transaction so the timestamps will go to disk.
747 * Otherwise, we will get pushed next time the znode is updated
748 * in a transaction, or when this znode eventually goes inactive.
751 * 1 - Only the ACCESS time is ever updated outside of a transaction.
752 * 2 - Multiple consecutive updates will be collapsed into a single
753 * znode update by the transaction grouping semantics of the DMU.
756 zfs_time_stamper(znode_t *zp, uint_t flag, dmu_tx_t *tx)
758 mutex_enter(&zp->z_lock);
759 zfs_time_stamper_locked(zp, flag, tx);
760 mutex_exit(&zp->z_lock);
764 * Grow the block size for a file.
766 * IN: zp - znode of file to free data in.
767 * size - requested block size
768 * tx - open transaction.
770 * NOTE: this function assumes that the znode is write locked.
773 zfs_grow_blocksize(znode_t *zp, uint64_t size, dmu_tx_t *tx)
778 if (size <= zp->z_blksz)
781 * If the file size is already greater than the current blocksize,
782 * we will not grow. If there is more than one block in a file,
783 * the blocksize cannot change.
785 if (zp->z_blksz && zp->z_phys->zp_size > zp->z_blksz)
788 error = dmu_object_set_blocksize(zp->z_zfsvfs->z_os, zp->z_id,
790 if (error == ENOTSUP)
792 ASSERT3U(error, ==, 0);
794 /* What blocksize did we actually get? */
795 dmu_object_size_from_db(zp->z_dbuf, &zp->z_blksz, &dummy);
799 * Free space in a file.
801 * IN: zp - znode of file to free data in.
802 * off - start of section to free.
803 * len - length of section to free (0 => to EOF).
804 * flag - current file open mode flags.
806 * RETURN: 0 if success
807 * error code if failure
810 zfs_freesp(znode_t *zp, uint64_t off, uint64_t len, int flag, boolean_t log)
812 vnode_t *vp = ZTOV(zp);
814 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
815 zilog_t *zilog = zfsvfs->z_log;
817 uint64_t end = off + len;
818 uint64_t size, new_blksz;
821 if (ZTOV(zp)->v_type == VFIFO)
825 * If we will change zp_size then lock the whole file,
826 * otherwise just lock the range being freed.
828 if (len == 0 || off + len > zp->z_phys->zp_size) {
829 rl = zfs_range_lock(zp, 0, UINT64_MAX, RL_WRITER);
831 rl = zfs_range_lock(zp, off, len, RL_WRITER);
832 /* recheck, in case zp_size changed */
833 if (off + len > zp->z_phys->zp_size) {
834 /* lost race: file size changed, lock whole file */
835 zfs_range_unlock(rl);
836 rl = zfs_range_lock(zp, 0, UINT64_MAX, RL_WRITER);
841 * Nothing to do if file already at desired length.
843 size = zp->z_phys->zp_size;
844 if (len == 0 && size == off && off != 0) {
845 zfs_range_unlock(rl);
849 tx = dmu_tx_create(zfsvfs->z_os);
850 dmu_tx_hold_bonus(tx, zp->z_id);
853 (!ISP2(zp->z_blksz) || zp->z_blksz < zfsvfs->z_max_blksz)) {
855 * We are growing the file past the current block size.
857 if (zp->z_blksz > zp->z_zfsvfs->z_max_blksz) {
858 ASSERT(!ISP2(zp->z_blksz));
859 new_blksz = MIN(end, SPA_MAXBLOCKSIZE);
861 new_blksz = MIN(end, zp->z_zfsvfs->z_max_blksz);
863 dmu_tx_hold_write(tx, zp->z_id, 0, MIN(end, new_blksz));
864 } else if (off < size) {
866 * If len == 0, we are truncating the file.
868 dmu_tx_hold_free(tx, zp->z_id, off, len ? len : DMU_OBJECT_END);
871 error = dmu_tx_assign(tx, zfsvfs->z_assign);
873 if (error == ERESTART && zfsvfs->z_assign == TXG_NOWAIT)
876 zfs_range_unlock(rl);
881 zfs_grow_blocksize(zp, new_blksz, tx);
883 if (end > size || len == 0)
884 zp->z_phys->zp_size = end;
887 objset_t *os = zfsvfs->z_os;
894 VERIFY(0 == dmu_free_range(os, zp->z_id, off, rlen, tx));
898 zfs_time_stamper(zp, CONTENT_MODIFIED, tx);
899 zfs_log_truncate(zilog, tx, TX_TRUNCATE, zp, off, len);
902 zfs_range_unlock(rl);
907 * Clear any mapped pages in the truncated region. This has to
908 * happen outside of the transaction to avoid the possibility of
909 * a deadlock with someone trying to push a page that we are
910 * about to invalidate.
912 rw_enter(&zp->z_map_lock, RW_WRITER);
914 vnode_pager_setsize(vp, end);
917 error = vtruncbuf(vp, curthread->td_ucred, curthread, end, PAGE_SIZE);
919 error = vinvalbuf(vp, V_SAVE, 0, 0);
920 vnode_pager_setsize(vp, end);
923 rw_exit(&zp->z_map_lock);
929 zfs_create_fs(objset_t *os, cred_t *cr, dmu_tx_t *tx)
932 uint64_t moid, doid, roid = 0;
933 uint64_t version = ZPL_VERSION;
935 znode_t *rootzp = NULL;
939 * First attempt to create master node.
942 * In an empty objset, there are no blocks to read and thus
943 * there can be no i/o errors (which we assert below).
945 moid = MASTER_NODE_OBJ;
946 error = zap_create_claim(os, moid, DMU_OT_MASTER_NODE,
951 * Set starting attributes.
954 error = zap_update(os, moid, ZPL_VERSION_OBJ, 8, 1, &version, tx);
958 * Create a delete queue.
960 doid = zap_create(os, DMU_OT_UNLINKED_SET, DMU_OT_NONE, 0, tx);
962 error = zap_add(os, moid, ZFS_UNLINKED_SET, 8, 1, &doid, tx);
966 * Create root znode. Create minimal znode/vnode/zfsvfs
967 * to allow zfs_mknode to work.
969 vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE;
970 vattr.va_type = VDIR;
971 vattr.va_mode = S_IFDIR|0755;
972 vattr.va_uid = UID_ROOT;
973 vattr.va_gid = GID_WHEEL;
975 rootzp = kmem_cache_alloc(znode_cache, KM_SLEEP);
976 zfs_znode_cache_constructor(rootzp, NULL, 0);
977 rootzp->z_zfsvfs = &zfsvfs;
978 rootzp->z_unlinked = 0;
979 rootzp->z_atime_dirty = 0;
980 rootzp->z_dbuf_held = 0;
982 bzero(&zfsvfs, sizeof (zfsvfs_t));
985 zfsvfs.z_assign = TXG_NOWAIT;
986 zfsvfs.z_parent = &zfsvfs;
988 mutex_init(&zfsvfs.z_znodes_lock, NULL, MUTEX_DEFAULT, NULL);
989 list_create(&zfsvfs.z_all_znodes, sizeof (znode_t),
990 offsetof(znode_t, z_link_node));
992 zfs_mknode(rootzp, &vattr, &roid, tx, cr, IS_ROOT_NODE, NULL, 0);
993 ASSERT3U(rootzp->z_id, ==, roid);
994 error = zap_add(os, moid, ZFS_ROOT_OBJ, 8, 1, &roid, tx);
997 mutex_destroy(&zfsvfs.z_znodes_lock);
998 kmem_cache_free(znode_cache, rootzp);
1000 #endif /* _KERNEL */
1003 * Given an object number, return its parent object number and whether
1004 * or not the object is an extended attribute directory.
1007 zfs_obj_to_pobj(objset_t *osp, uint64_t obj, uint64_t *pobjp, int *is_xattrdir)
1010 dmu_object_info_t doi;
1014 if ((error = dmu_bonus_hold(osp, obj, FTAG, &db)) != 0)
1017 dmu_object_info_from_db(db, &doi);
1018 if (doi.doi_bonus_type != DMU_OT_ZNODE ||
1019 doi.doi_bonus_size < sizeof (znode_phys_t)) {
1020 dmu_buf_rele(db, FTAG);
1025 *pobjp = zp->zp_parent;
1026 *is_xattrdir = ((zp->zp_flags & ZFS_XATTR) != 0) &&
1027 S_ISDIR(zp->zp_mode);
1028 dmu_buf_rele(db, FTAG);
1034 zfs_obj_to_path(objset_t *osp, uint64_t obj, char *buf, int len)
1036 char *path = buf + len - 1;
1043 char component[MAXNAMELEN + 2];
1047 if ((error = zfs_obj_to_pobj(osp, obj, &pobj,
1048 &is_xattrdir)) != 0)
1059 (void) sprintf(component + 1, "<xattrdir>");
1061 error = zap_value_search(osp, pobj, obj, component + 1);
1066 complen = strlen(component);
1068 ASSERT(path >= buf);
1069 bcopy(component, path, complen);
1074 (void) memmove(buf, path, buf + len - path);