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]
23 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
24 * Copyright (c) 2012, 2015 by Delphix. All rights reserved.
25 * Copyright (c) 2014 Integros [integros.com]
26 * Copyright 2017 Nexenta Systems, Inc.
29 /* Portions Copyright 2007 Jeremy Teo */
30 /* Portions Copyright 2010 Robert Milkowski */
33 #include <sys/types.h>
34 #include <sys/param.h>
36 #include <sys/systm.h>
37 #include <sys/sysmacros.h>
38 #include <sys/resource.h>
40 #include <sys/endian.h>
42 #include <sys/vnode.h>
43 #if __FreeBSD_version >= 1300102
46 #include <sys/dirent.h>
50 #include <sys/taskq.h>
52 #include <sys/atomic.h>
53 #include <sys/namei.h>
55 #include <sys/cmn_err.h>
57 #include <sys/sysproto.h>
58 #include <sys/errno.h>
59 #include <sys/unistd.h>
60 #include <sys/zfs_dir.h>
61 #include <sys/zfs_ioctl.h>
62 #include <sys/fs/zfs.h>
64 #include <sys/dmu_objset.h>
70 #include <sys/policy.h>
71 #include <sys/sunddi.h>
72 #include <sys/filio.h>
74 #include <sys/zfs_ctldir.h>
75 #include <sys/zfs_fuid.h>
76 #include <sys/zfs_quota.h>
77 #include <sys/zfs_sa.h>
78 #include <sys/zfs_rlock.h>
79 #include <sys/extdirent.h>
82 #include <sys/sched.h>
84 #include <sys/vmmeter.h>
85 #include <vm/vm_param.h>
87 #include <sys/zfs_vnops.h>
89 #include <vm/vm_object.h>
91 #include <sys/extattr.h>
95 #define VN_OPEN_INVFS 0x0
100 #if __FreeBSD_version >= 1300047
101 #define vm_page_wire_lock(pp)
102 #define vm_page_wire_unlock(pp)
104 #define vm_page_wire_lock(pp) vm_page_lock(pp)
105 #define vm_page_wire_unlock(pp) vm_page_unlock(pp)
108 #ifdef DEBUG_VFS_LOCKS
109 #define VNCHECKREF(vp) \
110 VNASSERT((vp)->v_holdcnt > 0 && (vp)->v_usecount > 0, vp, \
111 ("%s: wrong ref counts", __func__));
113 #define VNCHECKREF(vp)
119 * Each vnode op performs some logical unit of work. To do this, the ZPL must
120 * properly lock its in-core state, create a DMU transaction, do the work,
121 * record this work in the intent log (ZIL), commit the DMU transaction,
122 * and wait for the intent log to commit if it is a synchronous operation.
123 * Moreover, the vnode ops must work in both normal and log replay context.
124 * The ordering of events is important to avoid deadlocks and references
125 * to freed memory. The example below illustrates the following Big Rules:
127 * (1) A check must be made in each zfs thread for a mounted file system.
128 * This is done avoiding races using ZFS_ENTER(zfsvfs).
129 * A ZFS_EXIT(zfsvfs) is needed before all returns. Any znodes
130 * must be checked with ZFS_VERIFY_ZP(zp). Both of these macros
131 * can return EIO from the calling function.
133 * (2) VN_RELE() should always be the last thing except for zil_commit()
134 * (if necessary) and ZFS_EXIT(). This is for 3 reasons:
135 * First, if it's the last reference, the vnode/znode
136 * can be freed, so the zp may point to freed memory. Second, the last
137 * reference will call zfs_zinactive(), which may induce a lot of work --
138 * pushing cached pages (which acquires range locks) and syncing out
139 * cached atime changes. Third, zfs_zinactive() may require a new tx,
140 * which could deadlock the system if you were already holding one.
141 * If you must call VN_RELE() within a tx then use VN_RELE_ASYNC().
143 * (3) All range locks must be grabbed before calling dmu_tx_assign(),
144 * as they can span dmu_tx_assign() calls.
146 * (4) If ZPL locks are held, pass TXG_NOWAIT as the second argument to
147 * dmu_tx_assign(). This is critical because we don't want to block
148 * while holding locks.
150 * If no ZPL locks are held (aside from ZFS_ENTER()), use TXG_WAIT. This
151 * reduces lock contention and CPU usage when we must wait (note that if
152 * throughput is constrained by the storage, nearly every transaction
155 * Note, in particular, that if a lock is sometimes acquired before
156 * the tx assigns, and sometimes after (e.g. z_lock), then failing
157 * to use a non-blocking assign can deadlock the system. The scenario:
159 * Thread A has grabbed a lock before calling dmu_tx_assign().
160 * Thread B is in an already-assigned tx, and blocks for this lock.
161 * Thread A calls dmu_tx_assign(TXG_WAIT) and blocks in txg_wait_open()
162 * forever, because the previous txg can't quiesce until B's tx commits.
164 * If dmu_tx_assign() returns ERESTART and zfsvfs->z_assign is TXG_NOWAIT,
165 * then drop all locks, call dmu_tx_wait(), and try again. On subsequent
166 * calls to dmu_tx_assign(), pass TXG_NOTHROTTLE in addition to TXG_NOWAIT,
167 * to indicate that this operation has already called dmu_tx_wait().
168 * This will ensure that we don't retry forever, waiting a short bit
171 * (5) If the operation succeeded, generate the intent log entry for it
172 * before dropping locks. This ensures that the ordering of events
173 * in the intent log matches the order in which they actually occurred.
174 * During ZIL replay the zfs_log_* functions will update the sequence
175 * number to indicate the zil transaction has replayed.
177 * (6) At the end of each vnode op, the DMU tx must always commit,
178 * regardless of whether there were any errors.
180 * (7) After dropping all locks, invoke zil_commit(zilog, foid)
181 * to ensure that synchronous semantics are provided when necessary.
183 * In general, this is how things should be ordered in each vnode op:
185 * ZFS_ENTER(zfsvfs); // exit if unmounted
187 * zfs_dirent_lookup(&dl, ...) // lock directory entry (may VN_HOLD())
188 * rw_enter(...); // grab any other locks you need
189 * tx = dmu_tx_create(...); // get DMU tx
190 * dmu_tx_hold_*(); // hold each object you might modify
191 * error = dmu_tx_assign(tx, (waited ? TXG_NOTHROTTLE : 0) | TXG_NOWAIT);
193 * rw_exit(...); // drop locks
194 * zfs_dirent_unlock(dl); // unlock directory entry
195 * VN_RELE(...); // release held vnodes
196 * if (error == ERESTART) {
202 * dmu_tx_abort(tx); // abort DMU tx
203 * ZFS_EXIT(zfsvfs); // finished in zfs
204 * return (error); // really out of space
206 * error = do_real_work(); // do whatever this VOP does
208 * zfs_log_*(...); // on success, make ZIL entry
209 * dmu_tx_commit(tx); // commit DMU tx -- error or not
210 * rw_exit(...); // drop locks
211 * zfs_dirent_unlock(dl); // unlock directory entry
212 * VN_RELE(...); // release held vnodes
213 * zil_commit(zilog, foid); // synchronous when necessary
214 * ZFS_EXIT(zfsvfs); // finished in zfs
215 * return (error); // done, report error
220 zfs_open(vnode_t **vpp, int flag, cred_t *cr)
222 znode_t *zp = VTOZ(*vpp);
223 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
228 if ((flag & FWRITE) && (zp->z_pflags & ZFS_APPENDONLY) &&
229 ((flag & FAPPEND) == 0)) {
231 return (SET_ERROR(EPERM));
234 if (!zfs_has_ctldir(zp) && zp->z_zfsvfs->z_vscan &&
235 ZTOV(zp)->v_type == VREG &&
236 !(zp->z_pflags & ZFS_AV_QUARANTINED) && zp->z_size > 0) {
237 if (fs_vscan(*vpp, cr, 0) != 0) {
239 return (SET_ERROR(EACCES));
243 /* Keep a count of the synchronous opens in the znode */
244 if (flag & (FSYNC | FDSYNC))
245 atomic_inc_32(&zp->z_sync_cnt);
253 zfs_close(vnode_t *vp, int flag, int count, offset_t offset, cred_t *cr)
255 znode_t *zp = VTOZ(vp);
256 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
261 /* Decrement the synchronous opens in the znode */
262 if ((flag & (FSYNC | FDSYNC)) && (count == 1))
263 atomic_dec_32(&zp->z_sync_cnt);
265 if (!zfs_has_ctldir(zp) && zp->z_zfsvfs->z_vscan &&
266 ZTOV(zp)->v_type == VREG &&
267 !(zp->z_pflags & ZFS_AV_QUARANTINED) && zp->z_size > 0)
268 VERIFY(fs_vscan(vp, cr, 1) == 0);
276 zfs_ioctl(vnode_t *vp, ulong_t com, intptr_t data, int flag, cred_t *cred,
288 * The following two ioctls are used by bfu. Faking out,
289 * necessary to avoid bfu errors.
301 off = *(offset_t *)data;
302 /* offset parameter is in/out */
303 error = zfs_holey(VTOZ(vp), com, &off);
306 *(offset_t *)data = off;
310 return (SET_ERROR(ENOTTY));
314 page_busy(vnode_t *vp, int64_t start, int64_t off, int64_t nbytes)
321 * At present vm_page_clear_dirty extends the cleared range to DEV_BSIZE
322 * aligned boundaries, if the range is not aligned. As a result a
323 * DEV_BSIZE subrange with partially dirty data may get marked as clean.
324 * It may happen that all DEV_BSIZE subranges are marked clean and thus
325 * the whole page would be considered clean despite have some
327 * For this reason we should shrink the range to DEV_BSIZE aligned
328 * boundaries before calling vm_page_clear_dirty.
330 end = rounddown2(off + nbytes, DEV_BSIZE);
331 off = roundup2(off, DEV_BSIZE);
335 zfs_vmobject_assert_wlocked_12(obj);
336 #if __FreeBSD_version < 1300050
338 if ((pp = vm_page_lookup(obj, OFF_TO_IDX(start))) != NULL &&
340 if (vm_page_xbusied(pp)) {
342 * Reference the page before unlocking and
343 * sleeping so that the page daemon is less
344 * likely to reclaim it.
346 vm_page_reference(pp);
348 zfs_vmobject_wunlock(obj);
349 vm_page_busy_sleep(pp, "zfsmwb", true);
350 zfs_vmobject_wlock(obj);
354 } else if (pp != NULL) {
359 ASSERT3U(pp->valid, ==, VM_PAGE_BITS_ALL);
360 vm_object_pip_add(obj, 1);
361 pmap_remove_write(pp);
363 vm_page_clear_dirty(pp, off, nbytes);
368 vm_page_grab_valid_unlocked(&pp, obj, OFF_TO_IDX(start),
369 VM_ALLOC_NOCREAT | VM_ALLOC_SBUSY | VM_ALLOC_NORMAL |
372 ASSERT3U(pp->valid, ==, VM_PAGE_BITS_ALL);
373 vm_object_pip_add(obj, 1);
374 pmap_remove_write(pp);
376 vm_page_clear_dirty(pp, off, nbytes);
383 page_unbusy(vm_page_t pp)
387 #if __FreeBSD_version >= 1300041
388 vm_object_pip_wakeup(pp->object);
390 vm_object_pip_subtract(pp->object, 1);
394 #if __FreeBSD_version > 1300051
396 page_hold(vnode_t *vp, int64_t start)
402 vm_page_grab_valid_unlocked(&m, obj, OFF_TO_IDX(start),
403 VM_ALLOC_NOCREAT | VM_ALLOC_WIRED | VM_ALLOC_IGN_SBUSY |
409 page_hold(vnode_t *vp, int64_t start)
415 zfs_vmobject_assert_wlocked(obj);
418 if ((pp = vm_page_lookup(obj, OFF_TO_IDX(start))) != NULL &&
420 if (vm_page_xbusied(pp)) {
422 * Reference the page before unlocking and
423 * sleeping so that the page daemon is less
424 * likely to reclaim it.
426 vm_page_reference(pp);
428 zfs_vmobject_wunlock(obj);
429 vm_page_busy_sleep(pp, "zfsmwb", true);
430 zfs_vmobject_wlock(obj);
434 ASSERT3U(pp->valid, ==, VM_PAGE_BITS_ALL);
435 vm_page_wire_lock(pp);
437 vm_page_wire_unlock(pp);
448 page_unhold(vm_page_t pp)
451 vm_page_wire_lock(pp);
452 #if __FreeBSD_version >= 1300035
453 vm_page_unwire(pp, PQ_ACTIVE);
457 vm_page_wire_unlock(pp);
461 * When a file is memory mapped, we must keep the IO data synchronized
462 * between the DMU cache and the memory mapped pages. What this means:
464 * On Write: If we find a memory mapped page, we write to *both*
465 * the page and the dmu buffer.
468 update_pages(znode_t *zp, int64_t start, int len, objset_t *os)
472 vnode_t *vp = ZTOV(zp);
476 ASSERT(vp->v_mount != NULL);
480 off = start & PAGEOFFSET;
481 zfs_vmobject_wlock_12(obj);
482 #if __FreeBSD_version >= 1300041
483 vm_object_pip_add(obj, 1);
485 for (start &= PAGEMASK; len > 0; start += PAGESIZE) {
487 int nbytes = imin(PAGESIZE - off, len);
489 if ((pp = page_busy(vp, start, off, nbytes)) != NULL) {
490 zfs_vmobject_wunlock_12(obj);
492 va = zfs_map_page(pp, &sf);
493 (void) dmu_read(os, zp->z_id, start + off, nbytes,
494 va + off, DMU_READ_PREFETCH);
497 zfs_vmobject_wlock_12(obj);
503 #if __FreeBSD_version >= 1300041
504 vm_object_pip_wakeup(obj);
506 vm_object_pip_wakeupn(obj, 0);
508 zfs_vmobject_wunlock_12(obj);
512 * Read with UIO_NOCOPY flag means that sendfile(2) requests
513 * ZFS to populate a range of page cache pages with data.
515 * NOTE: this function could be optimized to pre-allocate
516 * all pages in advance, drain exclusive busy on all of them,
517 * map them into contiguous KVA region and populate them
518 * in one single dmu_read() call.
521 mappedread_sf(znode_t *zp, int nbytes, zfs_uio_t *uio)
523 vnode_t *vp = ZTOV(zp);
524 objset_t *os = zp->z_zfsvfs->z_os;
533 ASSERT(zfs_uio_segflg(uio) == UIO_NOCOPY);
534 ASSERT(vp->v_mount != NULL);
537 ASSERT((zfs_uio_offset(uio) & PAGEOFFSET) == 0);
539 zfs_vmobject_wlock_12(obj);
540 for (start = zfs_uio_offset(uio); len > 0; start += PAGESIZE) {
541 int bytes = MIN(PAGESIZE, len);
543 pp = vm_page_grab_unlocked(obj, OFF_TO_IDX(start),
544 VM_ALLOC_SBUSY | VM_ALLOC_NORMAL | VM_ALLOC_IGN_SBUSY);
545 if (vm_page_none_valid(pp)) {
546 zfs_vmobject_wunlock_12(obj);
547 va = zfs_map_page(pp, &sf);
548 error = dmu_read(os, zp->z_id, start, bytes, va,
550 if (bytes != PAGESIZE && error == 0)
551 bzero(va + bytes, PAGESIZE - bytes);
553 zfs_vmobject_wlock_12(obj);
554 #if __FreeBSD_version >= 1300081
557 vm_page_activate(pp);
558 vm_page_do_sunbusy(pp);
560 zfs_vmobject_wlock(obj);
561 if (!vm_page_wired(pp) && pp->valid == 0 &&
562 vm_page_busy_tryupgrade(pp))
566 zfs_vmobject_wunlock(obj);
569 vm_page_do_sunbusy(pp);
572 if (pp->wire_count == 0 && pp->valid == 0 &&
576 pp->valid = VM_PAGE_BITS_ALL;
577 vm_page_activate(pp);
582 ASSERT3U(pp->valid, ==, VM_PAGE_BITS_ALL);
583 vm_page_do_sunbusy(pp);
587 zfs_uio_advance(uio, bytes);
590 zfs_vmobject_wunlock_12(obj);
595 * When a file is memory mapped, we must keep the IO data synchronized
596 * between the DMU cache and the memory mapped pages. What this means:
598 * On Read: We "read" preferentially from memory mapped pages,
599 * else we default from the dmu buffer.
601 * NOTE: We will always "break up" the IO into PAGESIZE uiomoves when
602 * the file is memory mapped.
605 mappedread(znode_t *zp, int nbytes, zfs_uio_t *uio)
607 vnode_t *vp = ZTOV(zp);
614 ASSERT(vp->v_mount != NULL);
618 start = zfs_uio_offset(uio);
619 off = start & PAGEOFFSET;
620 zfs_vmobject_wlock_12(obj);
621 for (start &= PAGEMASK; len > 0; start += PAGESIZE) {
623 uint64_t bytes = MIN(PAGESIZE - off, len);
625 if ((pp = page_hold(vp, start))) {
629 zfs_vmobject_wunlock_12(obj);
630 va = zfs_map_page(pp, &sf);
631 error = vn_io_fault_uiomove(va + off, bytes,
632 GET_UIO_STRUCT(uio));
634 zfs_vmobject_wlock_12(obj);
637 zfs_vmobject_wunlock_12(obj);
638 error = dmu_read_uio_dbuf(sa_get_db(zp->z_sa_hdl),
640 zfs_vmobject_wlock_12(obj);
647 zfs_vmobject_wunlock_12(obj);
652 zfs_write_simple(znode_t *zp, const void *data, size_t len,
653 loff_t pos, size_t *presid)
658 error = vn_rdwr(UIO_WRITE, ZTOV(zp), __DECONST(void *, data), len, pos,
659 UIO_SYSSPACE, IO_SYNC, kcred, NOCRED, &resid, curthread);
662 return (SET_ERROR(error));
663 } else if (presid == NULL) {
665 error = SET_ERROR(EIO);
674 zfs_zrele_async(znode_t *zp)
676 vnode_t *vp = ZTOV(zp);
677 objset_t *os = ITOZSB(vp)->z_os;
679 VN_RELE_ASYNC(vp, dsl_pool_zrele_taskq(dmu_objset_pool(os)));
683 zfs_dd_callback(struct mount *mp, void *arg, int lkflags, struct vnode **vpp)
688 error = vn_lock(*vpp, lkflags);
695 zfs_lookup_lock(vnode_t *dvp, vnode_t *vp, const char *name, int lkflags)
697 znode_t *zdp = VTOZ(dvp);
698 zfsvfs_t *zfsvfs __unused = zdp->z_zfsvfs;
702 if (zfsvfs->z_replay == B_FALSE)
703 ASSERT_VOP_LOCKED(dvp, __func__);
705 if ((zdp->z_pflags & ZFS_XATTR) == 0)
706 VERIFY(!RRM_LOCK_HELD(&zfsvfs->z_teardown_lock));
709 if (name[0] == 0 || (name[0] == '.' && name[1] == 0)) {
710 ASSERT3P(dvp, ==, vp);
712 ltype = lkflags & LK_TYPE_MASK;
713 if (ltype != VOP_ISLOCKED(dvp)) {
714 if (ltype == LK_EXCLUSIVE)
715 vn_lock(dvp, LK_UPGRADE | LK_RETRY);
716 else /* if (ltype == LK_SHARED) */
717 vn_lock(dvp, LK_DOWNGRADE | LK_RETRY);
720 * Relock for the "." case could leave us with
723 if (VN_IS_DOOMED(dvp)) {
725 return (SET_ERROR(ENOENT));
729 } else if (name[0] == '.' && name[1] == '.' && name[2] == 0) {
731 * Note that in this case, dvp is the child vnode, and we
732 * are looking up the parent vnode - exactly reverse from
733 * normal operation. Unlocking dvp requires some rather
734 * tricky unlock/relock dance to prevent mp from being freed;
735 * use vn_vget_ino_gen() which takes care of all that.
737 * XXX Note that there is a time window when both vnodes are
738 * unlocked. It is possible, although highly unlikely, that
739 * during that window the parent-child relationship between
740 * the vnodes may change, for example, get reversed.
741 * In that case we would have a wrong lock order for the vnodes.
742 * All other filesystems seem to ignore this problem, so we
744 * A potential solution could be implemented as follows:
745 * - using LK_NOWAIT when locking the second vnode and retrying
747 * - checking that the parent-child relationship still holds
748 * after locking both vnodes and retrying if it doesn't
750 error = vn_vget_ino_gen(dvp, zfs_dd_callback, vp, lkflags, &vp);
753 error = vn_lock(vp, lkflags);
761 * Lookup an entry in a directory, or an extended attribute directory.
762 * If it exists, return a held vnode reference for it.
764 * IN: dvp - vnode of directory to search.
765 * nm - name of entry to lookup.
766 * pnp - full pathname to lookup [UNUSED].
767 * flags - LOOKUP_XATTR set if looking for an attribute.
768 * rdir - root directory vnode [UNUSED].
769 * cr - credentials of caller.
770 * ct - caller context
772 * OUT: vpp - vnode of located entry, NULL if not found.
774 * RETURN: 0 on success, error code on failure.
781 zfs_lookup(vnode_t *dvp, const char *nm, vnode_t **vpp,
782 struct componentname *cnp, int nameiop, cred_t *cr, kthread_t *td,
783 int flags, boolean_t cached)
785 znode_t *zdp = VTOZ(dvp);
787 zfsvfs_t *zfsvfs = zdp->z_zfsvfs;
791 * Fast path lookup, however we must skip DNLC lookup
792 * for case folding or normalizing lookups because the
793 * DNLC code only stores the passed in name. This means
794 * creating 'a' and removing 'A' on a case insensitive
795 * file system would work, but DNLC still thinks 'a'
796 * exists and won't let you create it again on the next
797 * pass through fast path.
799 if (!(flags & LOOKUP_XATTR)) {
800 if (dvp->v_type != VDIR) {
801 return (SET_ERROR(ENOTDIR));
802 } else if (zdp->z_sa_hdl == NULL) {
803 return (SET_ERROR(EIO));
807 DTRACE_PROBE2(zfs__fastpath__lookup__miss, vnode_t *, dvp,
815 if (flags & LOOKUP_XATTR) {
817 * If the xattr property is off, refuse the lookup request.
819 if (!(zfsvfs->z_flags & ZSB_XATTR)) {
821 return (SET_ERROR(EOPNOTSUPP));
825 * We don't allow recursive attributes..
826 * Maybe someday we will.
828 if (zdp->z_pflags & ZFS_XATTR) {
830 return (SET_ERROR(EINVAL));
833 if ((error = zfs_get_xattrdir(VTOZ(dvp), &zp, cr, flags))) {
840 * Do we have permission to get into attribute directory?
842 error = zfs_zaccess(zp, ACE_EXECUTE, 0, B_FALSE, cr);
852 * Check accessibility of directory if we're not coming in via
857 if ((cnp->cn_flags & NOEXECCHECK) != 0) {
858 cnp->cn_flags &= ~NOEXECCHECK;
861 if ((error = zfs_zaccess(zdp, ACE_EXECUTE, 0, B_FALSE, cr))) {
867 if (zfsvfs->z_utf8 && u8_validate(nm, strlen(nm),
868 NULL, U8_VALIDATE_ENTIRE, &error) < 0) {
870 return (SET_ERROR(EILSEQ));
875 * First handle the special cases.
877 if ((cnp->cn_flags & ISDOTDOT) != 0) {
879 * If we are a snapshot mounted under .zfs, return
880 * the vp for the snapshot directory.
882 if (zdp->z_id == zfsvfs->z_root && zfsvfs->z_parent != zfsvfs) {
883 struct componentname cn;
888 ltype = VOP_ISLOCKED(dvp);
890 error = zfsctl_root(zfsvfs->z_parent, LK_SHARED,
893 cn.cn_nameptr = "snapshot";
894 cn.cn_namelen = strlen(cn.cn_nameptr);
895 cn.cn_nameiop = cnp->cn_nameiop;
896 cn.cn_flags = cnp->cn_flags & ~ISDOTDOT;
897 cn.cn_lkflags = cnp->cn_lkflags;
898 error = VOP_LOOKUP(zfsctl_vp, vpp, &cn);
901 vn_lock(dvp, ltype | LK_RETRY);
905 if (zfs_has_ctldir(zdp) && strcmp(nm, ZFS_CTLDIR_NAME) == 0) {
907 if ((cnp->cn_flags & ISLASTCN) != 0 && nameiop != LOOKUP)
908 return (SET_ERROR(ENOTSUP));
909 error = zfsctl_root(zfsvfs, cnp->cn_lkflags, vpp);
914 * The loop is retry the lookup if the parent-child relationship
915 * changes during the dot-dot locking complexities.
920 error = zfs_dirlook(zdp, nm, &zp);
928 error = zfs_lookup_lock(dvp, *vpp, nm, cnp->cn_lkflags);
931 * If we've got a locking error, then the vnode
932 * got reclaimed because of a force unmount.
933 * We never enter doomed vnodes into the name cache.
939 if ((cnp->cn_flags & ISDOTDOT) == 0)
943 if (zdp->z_sa_hdl == NULL) {
944 error = SET_ERROR(EIO);
946 error = sa_lookup(zdp->z_sa_hdl, SA_ZPL_PARENT(zfsvfs),
947 &parent, sizeof (parent));
954 if (zp->z_id == parent) {
964 /* Translate errors and add SAVENAME when needed. */
965 if (cnp->cn_flags & ISLASTCN) {
969 if (error == ENOENT) {
971 cnp->cn_flags |= SAVENAME;
977 cnp->cn_flags |= SAVENAME;
982 /* Insert name into cache (as non-existent) if appropriate. */
983 if (zfsvfs->z_use_namecache && !zfsvfs->z_replay &&
984 error == ENOENT && (cnp->cn_flags & MAKEENTRY) != 0)
985 cache_enter(dvp, NULL, cnp);
987 /* Insert name into cache if appropriate. */
988 if (zfsvfs->z_use_namecache && !zfsvfs->z_replay &&
989 error == 0 && (cnp->cn_flags & MAKEENTRY)) {
990 if (!(cnp->cn_flags & ISLASTCN) ||
991 (nameiop != DELETE && nameiop != RENAME)) {
992 cache_enter(dvp, *vpp, cnp);
1000 * Attempt to create a new entry in a directory. If the entry
1001 * already exists, truncate the file if permissible, else return
1002 * an error. Return the vp of the created or trunc'd file.
1004 * IN: dvp - vnode of directory to put new file entry in.
1005 * name - name of new file entry.
1006 * vap - attributes of new file.
1007 * excl - flag indicating exclusive or non-exclusive mode.
1008 * mode - mode to open file with.
1009 * cr - credentials of caller.
1010 * flag - large file flag [UNUSED].
1011 * ct - caller context
1012 * vsecp - ACL to be set
1014 * OUT: vpp - vnode of created or trunc'd entry.
1016 * RETURN: 0 on success, error code on failure.
1019 * dvp - ctime|mtime updated if new entry created
1020 * vp - ctime|mtime always, atime if new
1025 zfs_create(znode_t *dzp, const char *name, vattr_t *vap, int excl, int mode,
1026 znode_t **zpp, cred_t *cr, int flag, vsecattr_t *vsecp)
1029 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
1036 gid_t gid = crgetgid(cr);
1037 uint64_t projid = ZFS_DEFAULT_PROJID;
1038 zfs_acl_ids_t acl_ids;
1039 boolean_t fuid_dirtied;
1041 #ifdef DEBUG_VFS_LOCKS
1042 vnode_t *dvp = ZTOV(dzp);
1046 * If we have an ephemeral id, ACL, or XVATTR then
1047 * make sure file system is at proper version
1050 ksid = crgetsid(cr, KSID_OWNER);
1052 uid = ksid_getid(ksid);
1056 if (zfsvfs->z_use_fuids == B_FALSE &&
1057 (vsecp || (vap->va_mask & AT_XVATTR) ||
1058 IS_EPHEMERAL(uid) || IS_EPHEMERAL(gid)))
1059 return (SET_ERROR(EINVAL));
1064 zilog = zfsvfs->z_log;
1066 if (zfsvfs->z_utf8 && u8_validate(name, strlen(name),
1067 NULL, U8_VALIDATE_ENTIRE, &error) < 0) {
1069 return (SET_ERROR(EILSEQ));
1072 if (vap->va_mask & AT_XVATTR) {
1073 if ((error = secpolicy_xvattr(ZTOV(dzp), (xvattr_t *)vap,
1074 crgetuid(cr), cr, vap->va_type)) != 0) {
1082 if ((vap->va_mode & S_ISVTX) && secpolicy_vnode_stky_modify(cr))
1083 vap->va_mode &= ~S_ISVTX;
1085 error = zfs_dirent_lookup(dzp, name, &zp, ZNEW);
1090 ASSERT3P(zp, ==, NULL);
1093 * Create a new file object and update the directory
1096 if ((error = zfs_zaccess(dzp, ACE_ADD_FILE, 0, B_FALSE, cr))) {
1101 * We only support the creation of regular files in
1102 * extended attribute directories.
1105 if ((dzp->z_pflags & ZFS_XATTR) &&
1106 (vap->va_type != VREG)) {
1107 error = SET_ERROR(EINVAL);
1111 if ((error = zfs_acl_ids_create(dzp, 0, vap,
1112 cr, vsecp, &acl_ids)) != 0)
1115 if (S_ISREG(vap->va_mode) || S_ISDIR(vap->va_mode))
1116 projid = zfs_inherit_projid(dzp);
1117 if (zfs_acl_ids_overquota(zfsvfs, &acl_ids, projid)) {
1118 zfs_acl_ids_free(&acl_ids);
1119 error = SET_ERROR(EDQUOT);
1123 getnewvnode_reserve_();
1125 tx = dmu_tx_create(os);
1127 dmu_tx_hold_sa_create(tx, acl_ids.z_aclp->z_acl_bytes +
1128 ZFS_SA_BASE_ATTR_SIZE);
1130 fuid_dirtied = zfsvfs->z_fuid_dirty;
1132 zfs_fuid_txhold(zfsvfs, tx);
1133 dmu_tx_hold_zap(tx, dzp->z_id, TRUE, name);
1134 dmu_tx_hold_sa(tx, dzp->z_sa_hdl, B_FALSE);
1135 if (!zfsvfs->z_use_sa &&
1136 acl_ids.z_aclp->z_acl_bytes > ZFS_ACE_SPACE) {
1137 dmu_tx_hold_write(tx, DMU_NEW_OBJECT,
1138 0, acl_ids.z_aclp->z_acl_bytes);
1140 error = dmu_tx_assign(tx, TXG_WAIT);
1142 zfs_acl_ids_free(&acl_ids);
1144 getnewvnode_drop_reserve();
1148 zfs_mknode(dzp, vap, tx, cr, 0, &zp, &acl_ids);
1150 zfs_fuid_sync(zfsvfs, tx);
1152 (void) zfs_link_create(dzp, name, zp, tx, ZNEW);
1153 txtype = zfs_log_create_txtype(Z_FILE, vsecp, vap);
1154 zfs_log_create(zilog, tx, txtype, dzp, zp, name,
1155 vsecp, acl_ids.z_fuidp, vap);
1156 zfs_acl_ids_free(&acl_ids);
1159 getnewvnode_drop_reserve();
1167 if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS)
1168 zil_commit(zilog, 0);
1175 * Remove an entry from a directory.
1177 * IN: dvp - vnode of directory to remove entry from.
1178 * name - name of entry to remove.
1179 * cr - credentials of caller.
1180 * ct - caller context
1181 * flags - case flags
1183 * RETURN: 0 on success, error code on failure.
1187 * vp - ctime (if nlink > 0)
1192 zfs_remove_(vnode_t *dvp, vnode_t *vp, const char *name, cred_t *cr)
1194 znode_t *dzp = VTOZ(dvp);
1197 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
1211 zilog = zfsvfs->z_log;
1216 if ((error = zfs_zaccess_delete(dzp, zp, cr))) {
1221 * Need to use rmdir for removing directories.
1223 if (vp->v_type == VDIR) {
1224 error = SET_ERROR(EPERM);
1228 vnevent_remove(vp, dvp, name, ct);
1232 /* are there any extended attributes? */
1233 error = sa_lookup(zp->z_sa_hdl, SA_ZPL_XATTR(zfsvfs),
1234 &xattr_obj, sizeof (xattr_obj));
1235 if (error == 0 && xattr_obj) {
1236 error = zfs_zget(zfsvfs, xattr_obj, &xzp);
1241 * We may delete the znode now, or we may put it in the unlinked set;
1242 * it depends on whether we're the last link, and on whether there are
1243 * other holds on the vnode. So we dmu_tx_hold() the right things to
1244 * allow for either case.
1246 tx = dmu_tx_create(zfsvfs->z_os);
1247 dmu_tx_hold_zap(tx, dzp->z_id, FALSE, name);
1248 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
1249 zfs_sa_upgrade_txholds(tx, zp);
1250 zfs_sa_upgrade_txholds(tx, dzp);
1253 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE);
1254 dmu_tx_hold_sa(tx, xzp->z_sa_hdl, B_FALSE);
1257 /* charge as an update -- would be nice not to charge at all */
1258 dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL);
1261 * Mark this transaction as typically resulting in a net free of space
1263 dmu_tx_mark_netfree(tx);
1265 error = dmu_tx_assign(tx, TXG_WAIT);
1273 * Remove the directory entry.
1275 error = zfs_link_destroy(dzp, name, zp, tx, ZEXISTS, &unlinked);
1283 zfs_unlinked_add(zp, tx);
1284 vp->v_vflag |= VV_NOSYNC;
1286 /* XXX check changes to linux vnops */
1288 zfs_log_remove(zilog, tx, txtype, dzp, name, obj, unlinked);
1296 if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS)
1297 zil_commit(zilog, 0);
1306 zfs_lookup_internal(znode_t *dzp, const char *name, vnode_t **vpp,
1307 struct componentname *cnp, int nameiop)
1309 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
1312 cnp->cn_nameptr = __DECONST(char *, name);
1313 cnp->cn_namelen = strlen(name);
1314 cnp->cn_nameiop = nameiop;
1315 cnp->cn_flags = ISLASTCN | SAVENAME;
1316 cnp->cn_lkflags = LK_EXCLUSIVE | LK_RETRY;
1317 cnp->cn_cred = kcred;
1318 cnp->cn_thread = curthread;
1320 if (zfsvfs->z_use_namecache && !zfsvfs->z_replay) {
1321 struct vop_lookup_args a;
1323 a.a_gen.a_desc = &vop_lookup_desc;
1324 a.a_dvp = ZTOV(dzp);
1327 error = vfs_cache_lookup(&a);
1329 error = zfs_lookup(ZTOV(dzp), name, vpp, cnp, nameiop, kcred,
1330 curthread, 0, B_FALSE);
1334 printf("got error %d on name %s on op %d\n", error, name,
1343 zfs_remove(znode_t *dzp, const char *name, cred_t *cr, int flags)
1347 struct componentname cn;
1349 if ((error = zfs_lookup_internal(dzp, name, &vp, &cn, DELETE)))
1352 error = zfs_remove_(ZTOV(dzp), vp, name, cr);
1357 * Create a new directory and insert it into dvp using the name
1358 * provided. Return a pointer to the inserted directory.
1360 * IN: dvp - vnode of directory to add subdir to.
1361 * dirname - name of new directory.
1362 * vap - attributes of new directory.
1363 * cr - credentials of caller.
1364 * ct - caller context
1365 * flags - case flags
1366 * vsecp - ACL to be set
1368 * OUT: vpp - vnode of created directory.
1370 * RETURN: 0 on success, error code on failure.
1373 * dvp - ctime|mtime updated
1374 * vp - ctime|mtime|atime updated
1378 zfs_mkdir(znode_t *dzp, const char *dirname, vattr_t *vap, znode_t **zpp,
1379 cred_t *cr, int flags, vsecattr_t *vsecp)
1382 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
1389 gid_t gid = crgetgid(cr);
1390 zfs_acl_ids_t acl_ids;
1391 boolean_t fuid_dirtied;
1393 ASSERT(vap->va_type == VDIR);
1396 * If we have an ephemeral id, ACL, or XVATTR then
1397 * make sure file system is at proper version
1400 ksid = crgetsid(cr, KSID_OWNER);
1402 uid = ksid_getid(ksid);
1405 if (zfsvfs->z_use_fuids == B_FALSE &&
1406 ((vap->va_mask & AT_XVATTR) ||
1407 IS_EPHEMERAL(uid) || IS_EPHEMERAL(gid)))
1408 return (SET_ERROR(EINVAL));
1412 zilog = zfsvfs->z_log;
1414 if (dzp->z_pflags & ZFS_XATTR) {
1416 return (SET_ERROR(EINVAL));
1419 if (zfsvfs->z_utf8 && u8_validate(dirname,
1420 strlen(dirname), NULL, U8_VALIDATE_ENTIRE, &error) < 0) {
1422 return (SET_ERROR(EILSEQ));
1425 if (vap->va_mask & AT_XVATTR) {
1426 if ((error = secpolicy_xvattr(ZTOV(dzp), (xvattr_t *)vap,
1427 crgetuid(cr), cr, vap->va_type)) != 0) {
1433 if ((error = zfs_acl_ids_create(dzp, 0, vap, cr,
1434 NULL, &acl_ids)) != 0) {
1440 * First make sure the new directory doesn't exist.
1442 * Existence is checked first to make sure we don't return
1443 * EACCES instead of EEXIST which can cause some applications
1448 if ((error = zfs_dirent_lookup(dzp, dirname, &zp, ZNEW))) {
1449 zfs_acl_ids_free(&acl_ids);
1453 ASSERT3P(zp, ==, NULL);
1455 if ((error = zfs_zaccess(dzp, ACE_ADD_SUBDIRECTORY, 0, B_FALSE, cr))) {
1456 zfs_acl_ids_free(&acl_ids);
1461 if (zfs_acl_ids_overquota(zfsvfs, &acl_ids, zfs_inherit_projid(dzp))) {
1462 zfs_acl_ids_free(&acl_ids);
1464 return (SET_ERROR(EDQUOT));
1468 * Add a new entry to the directory.
1470 getnewvnode_reserve_();
1471 tx = dmu_tx_create(zfsvfs->z_os);
1472 dmu_tx_hold_zap(tx, dzp->z_id, TRUE, dirname);
1473 dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL);
1474 fuid_dirtied = zfsvfs->z_fuid_dirty;
1476 zfs_fuid_txhold(zfsvfs, tx);
1477 if (!zfsvfs->z_use_sa && acl_ids.z_aclp->z_acl_bytes > ZFS_ACE_SPACE) {
1478 dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0,
1479 acl_ids.z_aclp->z_acl_bytes);
1482 dmu_tx_hold_sa_create(tx, acl_ids.z_aclp->z_acl_bytes +
1483 ZFS_SA_BASE_ATTR_SIZE);
1485 error = dmu_tx_assign(tx, TXG_WAIT);
1487 zfs_acl_ids_free(&acl_ids);
1489 getnewvnode_drop_reserve();
1497 zfs_mknode(dzp, vap, tx, cr, 0, &zp, &acl_ids);
1500 zfs_fuid_sync(zfsvfs, tx);
1503 * Now put new name in parent dir.
1505 (void) zfs_link_create(dzp, dirname, zp, tx, ZNEW);
1509 txtype = zfs_log_create_txtype(Z_DIR, NULL, vap);
1510 zfs_log_create(zilog, tx, txtype, dzp, zp, dirname, NULL,
1511 acl_ids.z_fuidp, vap);
1513 zfs_acl_ids_free(&acl_ids);
1517 getnewvnode_drop_reserve();
1519 if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS)
1520 zil_commit(zilog, 0);
1526 #if __FreeBSD_version < 1300124
1528 cache_vop_rmdir(struct vnode *dvp, struct vnode *vp)
1537 * Remove a directory subdir entry. If the current working
1538 * directory is the same as the subdir to be removed, the
1541 * IN: dvp - vnode of directory to remove from.
1542 * name - name of directory to be removed.
1543 * cwd - vnode of current working directory.
1544 * cr - credentials of caller.
1545 * ct - caller context
1546 * flags - case flags
1548 * RETURN: 0 on success, error code on failure.
1551 * dvp - ctime|mtime updated
1555 zfs_rmdir_(vnode_t *dvp, vnode_t *vp, const char *name, cred_t *cr)
1557 znode_t *dzp = VTOZ(dvp);
1558 znode_t *zp = VTOZ(vp);
1559 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
1567 zilog = zfsvfs->z_log;
1570 if ((error = zfs_zaccess_delete(dzp, zp, cr))) {
1574 if (vp->v_type != VDIR) {
1575 error = SET_ERROR(ENOTDIR);
1579 vnevent_rmdir(vp, dvp, name, ct);
1581 tx = dmu_tx_create(zfsvfs->z_os);
1582 dmu_tx_hold_zap(tx, dzp->z_id, FALSE, name);
1583 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
1584 dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL);
1585 zfs_sa_upgrade_txholds(tx, zp);
1586 zfs_sa_upgrade_txholds(tx, dzp);
1587 dmu_tx_mark_netfree(tx);
1588 error = dmu_tx_assign(tx, TXG_WAIT);
1595 error = zfs_link_destroy(dzp, name, zp, tx, ZEXISTS, NULL);
1598 uint64_t txtype = TX_RMDIR;
1599 zfs_log_remove(zilog, tx, txtype, dzp, name,
1600 ZFS_NO_OBJECT, B_FALSE);
1605 cache_vop_rmdir(dvp, vp);
1607 if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS)
1608 zil_commit(zilog, 0);
1615 zfs_rmdir(znode_t *dzp, const char *name, znode_t *cwd, cred_t *cr, int flags)
1617 struct componentname cn;
1621 if ((error = zfs_lookup_internal(dzp, name, &vp, &cn, DELETE)))
1624 error = zfs_rmdir_(ZTOV(dzp), vp, name, cr);
1630 * Read as many directory entries as will fit into the provided
1631 * buffer from the given directory cursor position (specified in
1632 * the uio structure).
1634 * IN: vp - vnode of directory to read.
1635 * uio - structure supplying read location, range info,
1636 * and return buffer.
1637 * cr - credentials of caller.
1638 * ct - caller context
1639 * flags - case flags
1641 * OUT: uio - updated offset and range, buffer filled.
1642 * eofp - set to true if end-of-file detected.
1644 * RETURN: 0 on success, error code on failure.
1647 * vp - atime updated
1649 * Note that the low 4 bits of the cookie returned by zap is always zero.
1650 * This allows us to use the low range for "special" directory entries:
1651 * We use 0 for '.', and 1 for '..'. If this is the root of the filesystem,
1652 * we use the offset 2 for the '.zfs' directory.
1656 zfs_readdir(vnode_t *vp, zfs_uio_t *uio, cred_t *cr, int *eofp,
1657 int *ncookies, ulong_t **cookies)
1659 znode_t *zp = VTOZ(vp);
1663 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1668 zap_attribute_t zap;
1669 uint_t bytes_wanted;
1670 uint64_t offset; /* must be unsigned; checks for < 1 */
1676 boolean_t check_sysattrs;
1679 ulong_t *cooks = NULL;
1685 if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_PARENT(zfsvfs),
1686 &parent, sizeof (parent))) != 0) {
1692 * If we are not given an eof variable,
1699 * Check for valid iov_len.
1701 if (GET_UIO_STRUCT(uio)->uio_iov->iov_len <= 0) {
1703 return (SET_ERROR(EINVAL));
1707 * Quit if directory has been removed (posix)
1709 if ((*eofp = zp->z_unlinked) != 0) {
1716 offset = zfs_uio_offset(uio);
1717 prefetch = zp->z_zn_prefetch;
1720 * Initialize the iterator cursor.
1724 * Start iteration from the beginning of the directory.
1726 zap_cursor_init(&zc, os, zp->z_id);
1729 * The offset is a serialized cursor.
1731 zap_cursor_init_serialized(&zc, os, zp->z_id, offset);
1735 * Get space to change directory entries into fs independent format.
1737 iovp = GET_UIO_STRUCT(uio)->uio_iov;
1738 bytes_wanted = iovp->iov_len;
1739 if (zfs_uio_segflg(uio) != UIO_SYSSPACE || zfs_uio_iovcnt(uio) != 1) {
1740 bufsize = bytes_wanted;
1741 outbuf = kmem_alloc(bufsize, KM_SLEEP);
1742 odp = (struct dirent64 *)outbuf;
1744 bufsize = bytes_wanted;
1746 odp = (struct dirent64 *)iovp->iov_base;
1748 eodp = (struct edirent *)odp;
1750 if (ncookies != NULL) {
1752 * Minimum entry size is dirent size and 1 byte for a file name.
1754 ncooks = zfs_uio_resid(uio) / (sizeof (struct dirent) -
1755 sizeof (((struct dirent *)NULL)->d_name) + 1);
1756 cooks = malloc(ncooks * sizeof (ulong_t), M_TEMP, M_WAITOK);
1761 * If this VFS supports the system attribute view interface; and
1762 * we're looking at an extended attribute directory; and we care
1763 * about normalization conflicts on this vfs; then we must check
1764 * for normalization conflicts with the sysattr name space.
1767 check_sysattrs = vfs_has_feature(vp->v_vfsp, VFSFT_SYSATTR_VIEWS) &&
1768 (vp->v_flag & V_XATTRDIR) && zfsvfs->z_norm &&
1769 (flags & V_RDDIR_ENTFLAGS);
1775 * Transform to file-system independent format
1778 while (outcount < bytes_wanted) {
1781 off64_t *next = NULL;
1784 * Special case `.', `..', and `.zfs'.
1787 (void) strcpy(zap.za_name, ".");
1788 zap.za_normalization_conflict = 0;
1791 } else if (offset == 1) {
1792 (void) strcpy(zap.za_name, "..");
1793 zap.za_normalization_conflict = 0;
1796 } else if (offset == 2 && zfs_show_ctldir(zp)) {
1797 (void) strcpy(zap.za_name, ZFS_CTLDIR_NAME);
1798 zap.za_normalization_conflict = 0;
1799 objnum = ZFSCTL_INO_ROOT;
1805 if ((error = zap_cursor_retrieve(&zc, &zap))) {
1806 if ((*eofp = (error == ENOENT)) != 0)
1812 if (zap.za_integer_length != 8 ||
1813 zap.za_num_integers != 1) {
1814 cmn_err(CE_WARN, "zap_readdir: bad directory "
1815 "entry, obj = %lld, offset = %lld\n",
1816 (u_longlong_t)zp->z_id,
1817 (u_longlong_t)offset);
1818 error = SET_ERROR(ENXIO);
1822 objnum = ZFS_DIRENT_OBJ(zap.za_first_integer);
1824 * MacOS X can extract the object type here such as:
1825 * uint8_t type = ZFS_DIRENT_TYPE(zap.za_first_integer);
1827 type = ZFS_DIRENT_TYPE(zap.za_first_integer);
1829 if (check_sysattrs && !zap.za_normalization_conflict) {
1831 zap.za_normalization_conflict =
1832 xattr_sysattr_casechk(zap.za_name);
1834 panic("%s:%u: TODO", __func__, __LINE__);
1839 if (flags & V_RDDIR_ACCFILTER) {
1841 * If we have no access at all, don't include
1842 * this entry in the returned information
1845 if (zfs_zget(zp->z_zfsvfs, objnum, &ezp) != 0)
1847 if (!zfs_has_access(ezp, cr)) {
1854 if (flags & V_RDDIR_ENTFLAGS)
1855 reclen = EDIRENT_RECLEN(strlen(zap.za_name));
1857 reclen = DIRENT64_RECLEN(strlen(zap.za_name));
1860 * Will this entry fit in the buffer?
1862 if (outcount + reclen > bufsize) {
1864 * Did we manage to fit anything in the buffer?
1867 error = SET_ERROR(EINVAL);
1872 if (flags & V_RDDIR_ENTFLAGS) {
1874 * Add extended flag entry:
1876 eodp->ed_ino = objnum;
1877 eodp->ed_reclen = reclen;
1878 /* NOTE: ed_off is the offset for the *next* entry */
1879 next = &(eodp->ed_off);
1880 eodp->ed_eflags = zap.za_normalization_conflict ?
1881 ED_CASE_CONFLICT : 0;
1882 (void) strncpy(eodp->ed_name, zap.za_name,
1883 EDIRENT_NAMELEN(reclen));
1884 eodp = (edirent_t *)((intptr_t)eodp + reclen);
1889 odp->d_ino = objnum;
1890 odp->d_reclen = reclen;
1891 odp->d_namlen = strlen(zap.za_name);
1892 /* NOTE: d_off is the offset for the *next* entry. */
1894 strlcpy(odp->d_name, zap.za_name, odp->d_namlen + 1);
1896 dirent_terminate(odp);
1897 odp = (dirent64_t *)((intptr_t)odp + reclen);
1901 ASSERT(outcount <= bufsize);
1903 /* Prefetch znode */
1905 dmu_prefetch(os, objnum, 0, 0, 0,
1906 ZIO_PRIORITY_SYNC_READ);
1910 * Move to the next entry, fill in the previous offset.
1912 if (offset > 2 || (offset == 2 && !zfs_show_ctldir(zp))) {
1913 zap_cursor_advance(&zc);
1914 offset = zap_cursor_serialize(&zc);
1919 /* Fill the offset right after advancing the cursor. */
1922 if (cooks != NULL) {
1925 KASSERT(ncooks >= 0, ("ncookies=%d", ncooks));
1928 zp->z_zn_prefetch = B_FALSE; /* a lookup will re-enable pre-fetching */
1930 /* Subtract unused cookies */
1931 if (ncookies != NULL)
1932 *ncookies -= ncooks;
1934 if (zfs_uio_segflg(uio) == UIO_SYSSPACE && zfs_uio_iovcnt(uio) == 1) {
1935 iovp->iov_base += outcount;
1936 iovp->iov_len -= outcount;
1937 zfs_uio_resid(uio) -= outcount;
1939 zfs_uiomove(outbuf, (long)outcount, UIO_READ, uio))) {
1941 * Reset the pointer.
1943 offset = zfs_uio_offset(uio);
1947 zap_cursor_fini(&zc);
1948 if (zfs_uio_segflg(uio) != UIO_SYSSPACE || zfs_uio_iovcnt(uio) != 1)
1949 kmem_free(outbuf, bufsize);
1951 if (error == ENOENT)
1954 ZFS_ACCESSTIME_STAMP(zfsvfs, zp);
1956 zfs_uio_setoffset(uio, offset);
1958 if (error != 0 && cookies != NULL) {
1959 free(*cookies, M_TEMP);
1967 * Get the requested file attributes and place them in the provided
1970 * IN: vp - vnode of file.
1971 * vap - va_mask identifies requested attributes.
1972 * If AT_XVATTR set, then optional attrs are requested
1973 * flags - ATTR_NOACLCHECK (CIFS server context)
1974 * cr - credentials of caller.
1976 * OUT: vap - attribute values.
1978 * RETURN: 0 (always succeeds).
1982 zfs_getattr(vnode_t *vp, vattr_t *vap, int flags, cred_t *cr)
1984 znode_t *zp = VTOZ(vp);
1985 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1988 u_longlong_t nblocks;
1989 uint64_t mtime[2], ctime[2], crtime[2], rdev;
1990 xvattr_t *xvap = (xvattr_t *)vap; /* vap may be an xvattr_t * */
1991 xoptattr_t *xoap = NULL;
1992 boolean_t skipaclchk = (flags & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE;
1993 sa_bulk_attr_t bulk[4];
1999 zfs_fuid_map_ids(zp, cr, &vap->va_uid, &vap->va_gid);
2001 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL, &mtime, 16);
2002 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL, &ctime, 16);
2003 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CRTIME(zfsvfs), NULL, &crtime, 16);
2004 if (vp->v_type == VBLK || vp->v_type == VCHR)
2005 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_RDEV(zfsvfs), NULL,
2008 if ((error = sa_bulk_lookup(zp->z_sa_hdl, bulk, count)) != 0) {
2014 * If ACL is trivial don't bother looking for ACE_READ_ATTRIBUTES.
2015 * Also, if we are the owner don't bother, since owner should
2016 * always be allowed to read basic attributes of file.
2018 if (!(zp->z_pflags & ZFS_ACL_TRIVIAL) &&
2019 (vap->va_uid != crgetuid(cr))) {
2020 if ((error = zfs_zaccess(zp, ACE_READ_ATTRIBUTES, 0,
2028 * Return all attributes. It's cheaper to provide the answer
2029 * than to determine whether we were asked the question.
2032 vap->va_type = IFTOVT(zp->z_mode);
2033 vap->va_mode = zp->z_mode & ~S_IFMT;
2035 vap->va_nodeid = zp->z_id;
2036 vap->va_nlink = zp->z_links;
2037 if ((vp->v_flag & VROOT) && zfs_show_ctldir(zp) &&
2038 zp->z_links < ZFS_LINK_MAX)
2040 vap->va_size = zp->z_size;
2041 if (vp->v_type == VBLK || vp->v_type == VCHR)
2042 vap->va_rdev = zfs_cmpldev(rdev);
2043 vap->va_seq = zp->z_seq;
2044 vap->va_flags = 0; /* FreeBSD: Reset chflags(2) flags. */
2045 vap->va_filerev = zp->z_seq;
2048 * Add in any requested optional attributes and the create time.
2049 * Also set the corresponding bits in the returned attribute bitmap.
2051 if ((xoap = xva_getxoptattr(xvap)) != NULL && zfsvfs->z_use_fuids) {
2052 if (XVA_ISSET_REQ(xvap, XAT_ARCHIVE)) {
2054 ((zp->z_pflags & ZFS_ARCHIVE) != 0);
2055 XVA_SET_RTN(xvap, XAT_ARCHIVE);
2058 if (XVA_ISSET_REQ(xvap, XAT_READONLY)) {
2059 xoap->xoa_readonly =
2060 ((zp->z_pflags & ZFS_READONLY) != 0);
2061 XVA_SET_RTN(xvap, XAT_READONLY);
2064 if (XVA_ISSET_REQ(xvap, XAT_SYSTEM)) {
2066 ((zp->z_pflags & ZFS_SYSTEM) != 0);
2067 XVA_SET_RTN(xvap, XAT_SYSTEM);
2070 if (XVA_ISSET_REQ(xvap, XAT_HIDDEN)) {
2072 ((zp->z_pflags & ZFS_HIDDEN) != 0);
2073 XVA_SET_RTN(xvap, XAT_HIDDEN);
2076 if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK)) {
2077 xoap->xoa_nounlink =
2078 ((zp->z_pflags & ZFS_NOUNLINK) != 0);
2079 XVA_SET_RTN(xvap, XAT_NOUNLINK);
2082 if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE)) {
2083 xoap->xoa_immutable =
2084 ((zp->z_pflags & ZFS_IMMUTABLE) != 0);
2085 XVA_SET_RTN(xvap, XAT_IMMUTABLE);
2088 if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY)) {
2089 xoap->xoa_appendonly =
2090 ((zp->z_pflags & ZFS_APPENDONLY) != 0);
2091 XVA_SET_RTN(xvap, XAT_APPENDONLY);
2094 if (XVA_ISSET_REQ(xvap, XAT_NODUMP)) {
2096 ((zp->z_pflags & ZFS_NODUMP) != 0);
2097 XVA_SET_RTN(xvap, XAT_NODUMP);
2100 if (XVA_ISSET_REQ(xvap, XAT_OPAQUE)) {
2102 ((zp->z_pflags & ZFS_OPAQUE) != 0);
2103 XVA_SET_RTN(xvap, XAT_OPAQUE);
2106 if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED)) {
2107 xoap->xoa_av_quarantined =
2108 ((zp->z_pflags & ZFS_AV_QUARANTINED) != 0);
2109 XVA_SET_RTN(xvap, XAT_AV_QUARANTINED);
2112 if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED)) {
2113 xoap->xoa_av_modified =
2114 ((zp->z_pflags & ZFS_AV_MODIFIED) != 0);
2115 XVA_SET_RTN(xvap, XAT_AV_MODIFIED);
2118 if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP) &&
2119 vp->v_type == VREG) {
2120 zfs_sa_get_scanstamp(zp, xvap);
2123 if (XVA_ISSET_REQ(xvap, XAT_REPARSE)) {
2124 xoap->xoa_reparse = ((zp->z_pflags & ZFS_REPARSE) != 0);
2125 XVA_SET_RTN(xvap, XAT_REPARSE);
2127 if (XVA_ISSET_REQ(xvap, XAT_GEN)) {
2128 xoap->xoa_generation = zp->z_gen;
2129 XVA_SET_RTN(xvap, XAT_GEN);
2132 if (XVA_ISSET_REQ(xvap, XAT_OFFLINE)) {
2134 ((zp->z_pflags & ZFS_OFFLINE) != 0);
2135 XVA_SET_RTN(xvap, XAT_OFFLINE);
2138 if (XVA_ISSET_REQ(xvap, XAT_SPARSE)) {
2140 ((zp->z_pflags & ZFS_SPARSE) != 0);
2141 XVA_SET_RTN(xvap, XAT_SPARSE);
2144 if (XVA_ISSET_REQ(xvap, XAT_PROJINHERIT)) {
2145 xoap->xoa_projinherit =
2146 ((zp->z_pflags & ZFS_PROJINHERIT) != 0);
2147 XVA_SET_RTN(xvap, XAT_PROJINHERIT);
2150 if (XVA_ISSET_REQ(xvap, XAT_PROJID)) {
2151 xoap->xoa_projid = zp->z_projid;
2152 XVA_SET_RTN(xvap, XAT_PROJID);
2156 ZFS_TIME_DECODE(&vap->va_atime, zp->z_atime);
2157 ZFS_TIME_DECODE(&vap->va_mtime, mtime);
2158 ZFS_TIME_DECODE(&vap->va_ctime, ctime);
2159 ZFS_TIME_DECODE(&vap->va_birthtime, crtime);
2162 sa_object_size(zp->z_sa_hdl, &blksize, &nblocks);
2163 vap->va_blksize = blksize;
2164 vap->va_bytes = nblocks << 9; /* nblocks * 512 */
2166 if (zp->z_blksz == 0) {
2168 * Block size hasn't been set; suggest maximal I/O transfers.
2170 vap->va_blksize = zfsvfs->z_max_blksz;
2178 * Set the file attributes to the values contained in the
2181 * IN: zp - znode of file to be modified.
2182 * vap - new attribute values.
2183 * If AT_XVATTR set, then optional attrs are being set
2184 * flags - ATTR_UTIME set if non-default time values provided.
2185 * - ATTR_NOACLCHECK (CIFS context only).
2186 * cr - credentials of caller.
2187 * ct - caller context
2189 * RETURN: 0 on success, error code on failure.
2192 * vp - ctime updated, mtime updated if size changed.
2196 zfs_setattr(znode_t *zp, vattr_t *vap, int flags, cred_t *cr)
2198 vnode_t *vp = ZTOV(zp);
2199 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
2200 objset_t *os = zfsvfs->z_os;
2205 uint_t mask = vap->va_mask;
2206 uint_t saved_mask = 0;
2207 uint64_t saved_mode;
2210 uint64_t new_uid, new_gid;
2212 uint64_t mtime[2], ctime[2];
2213 uint64_t projid = ZFS_INVALID_PROJID;
2215 int need_policy = FALSE;
2217 zfs_fuid_info_t *fuidp = NULL;
2218 xvattr_t *xvap = (xvattr_t *)vap; /* vap may be an xvattr_t * */
2221 boolean_t skipaclchk = (flags & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE;
2222 boolean_t fuid_dirtied = B_FALSE;
2223 sa_bulk_attr_t bulk[7], xattr_bulk[7];
2224 int count = 0, xattr_count = 0;
2229 if (mask & AT_NOSET)
2230 return (SET_ERROR(EINVAL));
2235 zilog = zfsvfs->z_log;
2238 * Make sure that if we have ephemeral uid/gid or xvattr specified
2239 * that file system is at proper version level
2242 if (zfsvfs->z_use_fuids == B_FALSE &&
2243 (((mask & AT_UID) && IS_EPHEMERAL(vap->va_uid)) ||
2244 ((mask & AT_GID) && IS_EPHEMERAL(vap->va_gid)) ||
2245 (mask & AT_XVATTR))) {
2247 return (SET_ERROR(EINVAL));
2250 if (mask & AT_SIZE && vp->v_type == VDIR) {
2252 return (SET_ERROR(EISDIR));
2255 if (mask & AT_SIZE && vp->v_type != VREG && vp->v_type != VFIFO) {
2257 return (SET_ERROR(EINVAL));
2261 * If this is an xvattr_t, then get a pointer to the structure of
2262 * optional attributes. If this is NULL, then we have a vattr_t.
2264 xoap = xva_getxoptattr(xvap);
2266 xva_init(&tmpxvattr);
2269 * Immutable files can only alter immutable bit and atime
2271 if ((zp->z_pflags & ZFS_IMMUTABLE) &&
2272 ((mask & (AT_SIZE|AT_UID|AT_GID|AT_MTIME|AT_MODE)) ||
2273 ((mask & AT_XVATTR) && XVA_ISSET_REQ(xvap, XAT_CREATETIME)))) {
2275 return (SET_ERROR(EPERM));
2279 * Note: ZFS_READONLY is handled in zfs_zaccess_common.
2283 * Verify timestamps doesn't overflow 32 bits.
2284 * ZFS can handle large timestamps, but 32bit syscalls can't
2285 * handle times greater than 2039. This check should be removed
2286 * once large timestamps are fully supported.
2288 if (mask & (AT_ATIME | AT_MTIME)) {
2289 if (((mask & AT_ATIME) && TIMESPEC_OVERFLOW(&vap->va_atime)) ||
2290 ((mask & AT_MTIME) && TIMESPEC_OVERFLOW(&vap->va_mtime))) {
2292 return (SET_ERROR(EOVERFLOW));
2295 if (xoap != NULL && (mask & AT_XVATTR)) {
2296 if (XVA_ISSET_REQ(xvap, XAT_CREATETIME) &&
2297 TIMESPEC_OVERFLOW(&vap->va_birthtime)) {
2299 return (SET_ERROR(EOVERFLOW));
2302 if (XVA_ISSET_REQ(xvap, XAT_PROJID)) {
2303 if (!dmu_objset_projectquota_enabled(os) ||
2304 (!S_ISREG(zp->z_mode) && !S_ISDIR(zp->z_mode))) {
2306 return (SET_ERROR(EOPNOTSUPP));
2309 projid = xoap->xoa_projid;
2310 if (unlikely(projid == ZFS_INVALID_PROJID)) {
2312 return (SET_ERROR(EINVAL));
2315 if (projid == zp->z_projid && zp->z_pflags & ZFS_PROJID)
2316 projid = ZFS_INVALID_PROJID;
2321 if (XVA_ISSET_REQ(xvap, XAT_PROJINHERIT) &&
2322 (xoap->xoa_projinherit !=
2323 ((zp->z_pflags & ZFS_PROJINHERIT) != 0)) &&
2324 (!dmu_objset_projectquota_enabled(os) ||
2325 (!S_ISREG(zp->z_mode) && !S_ISDIR(zp->z_mode)))) {
2327 return (SET_ERROR(EOPNOTSUPP));
2334 if (zfsvfs->z_vfs->vfs_flag & VFS_RDONLY) {
2336 return (SET_ERROR(EROFS));
2340 * First validate permissions
2343 if (mask & AT_SIZE) {
2345 * XXX - Note, we are not providing any open
2346 * mode flags here (like FNDELAY), so we may
2347 * block if there are locks present... this
2348 * should be addressed in openat().
2350 /* XXX - would it be OK to generate a log record here? */
2351 err = zfs_freesp(zp, vap->va_size, 0, 0, FALSE);
2358 if (mask & (AT_ATIME|AT_MTIME) ||
2359 ((mask & AT_XVATTR) && (XVA_ISSET_REQ(xvap, XAT_HIDDEN) ||
2360 XVA_ISSET_REQ(xvap, XAT_READONLY) ||
2361 XVA_ISSET_REQ(xvap, XAT_ARCHIVE) ||
2362 XVA_ISSET_REQ(xvap, XAT_OFFLINE) ||
2363 XVA_ISSET_REQ(xvap, XAT_SPARSE) ||
2364 XVA_ISSET_REQ(xvap, XAT_CREATETIME) ||
2365 XVA_ISSET_REQ(xvap, XAT_SYSTEM)))) {
2366 need_policy = zfs_zaccess(zp, ACE_WRITE_ATTRIBUTES, 0,
2370 if (mask & (AT_UID|AT_GID)) {
2371 int idmask = (mask & (AT_UID|AT_GID));
2376 * NOTE: even if a new mode is being set,
2377 * we may clear S_ISUID/S_ISGID bits.
2380 if (!(mask & AT_MODE))
2381 vap->va_mode = zp->z_mode;
2384 * Take ownership or chgrp to group we are a member of
2387 take_owner = (mask & AT_UID) && (vap->va_uid == crgetuid(cr));
2388 take_group = (mask & AT_GID) &&
2389 zfs_groupmember(zfsvfs, vap->va_gid, cr);
2392 * If both AT_UID and AT_GID are set then take_owner and
2393 * take_group must both be set in order to allow taking
2396 * Otherwise, send the check through secpolicy_vnode_setattr()
2400 if (((idmask == (AT_UID|AT_GID)) && take_owner && take_group) ||
2401 ((idmask == AT_UID) && take_owner) ||
2402 ((idmask == AT_GID) && take_group)) {
2403 if (zfs_zaccess(zp, ACE_WRITE_OWNER, 0,
2404 skipaclchk, cr) == 0) {
2406 * Remove setuid/setgid for non-privileged users
2408 secpolicy_setid_clear(vap, vp, cr);
2409 trim_mask = (mask & (AT_UID|AT_GID));
2418 oldva.va_mode = zp->z_mode;
2419 zfs_fuid_map_ids(zp, cr, &oldva.va_uid, &oldva.va_gid);
2420 if (mask & AT_XVATTR) {
2422 * Update xvattr mask to include only those attributes
2423 * that are actually changing.
2425 * the bits will be restored prior to actually setting
2426 * the attributes so the caller thinks they were set.
2428 if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY)) {
2429 if (xoap->xoa_appendonly !=
2430 ((zp->z_pflags & ZFS_APPENDONLY) != 0)) {
2433 XVA_CLR_REQ(xvap, XAT_APPENDONLY);
2434 XVA_SET_REQ(&tmpxvattr, XAT_APPENDONLY);
2438 if (XVA_ISSET_REQ(xvap, XAT_PROJINHERIT)) {
2439 if (xoap->xoa_projinherit !=
2440 ((zp->z_pflags & ZFS_PROJINHERIT) != 0)) {
2443 XVA_CLR_REQ(xvap, XAT_PROJINHERIT);
2444 XVA_SET_REQ(&tmpxvattr, XAT_PROJINHERIT);
2448 if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK)) {
2449 if (xoap->xoa_nounlink !=
2450 ((zp->z_pflags & ZFS_NOUNLINK) != 0)) {
2453 XVA_CLR_REQ(xvap, XAT_NOUNLINK);
2454 XVA_SET_REQ(&tmpxvattr, XAT_NOUNLINK);
2458 if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE)) {
2459 if (xoap->xoa_immutable !=
2460 ((zp->z_pflags & ZFS_IMMUTABLE) != 0)) {
2463 XVA_CLR_REQ(xvap, XAT_IMMUTABLE);
2464 XVA_SET_REQ(&tmpxvattr, XAT_IMMUTABLE);
2468 if (XVA_ISSET_REQ(xvap, XAT_NODUMP)) {
2469 if (xoap->xoa_nodump !=
2470 ((zp->z_pflags & ZFS_NODUMP) != 0)) {
2473 XVA_CLR_REQ(xvap, XAT_NODUMP);
2474 XVA_SET_REQ(&tmpxvattr, XAT_NODUMP);
2478 if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED)) {
2479 if (xoap->xoa_av_modified !=
2480 ((zp->z_pflags & ZFS_AV_MODIFIED) != 0)) {
2483 XVA_CLR_REQ(xvap, XAT_AV_MODIFIED);
2484 XVA_SET_REQ(&tmpxvattr, XAT_AV_MODIFIED);
2488 if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED)) {
2489 if ((vp->v_type != VREG &&
2490 xoap->xoa_av_quarantined) ||
2491 xoap->xoa_av_quarantined !=
2492 ((zp->z_pflags & ZFS_AV_QUARANTINED) != 0)) {
2495 XVA_CLR_REQ(xvap, XAT_AV_QUARANTINED);
2496 XVA_SET_REQ(&tmpxvattr, XAT_AV_QUARANTINED);
2500 if (XVA_ISSET_REQ(xvap, XAT_REPARSE)) {
2502 return (SET_ERROR(EPERM));
2505 if (need_policy == FALSE &&
2506 (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP) ||
2507 XVA_ISSET_REQ(xvap, XAT_OPAQUE))) {
2512 if (mask & AT_MODE) {
2513 if (zfs_zaccess(zp, ACE_WRITE_ACL, 0, skipaclchk, cr) == 0) {
2514 err = secpolicy_setid_setsticky_clear(vp, vap,
2520 trim_mask |= AT_MODE;
2528 * If trim_mask is set then take ownership
2529 * has been granted or write_acl is present and user
2530 * has the ability to modify mode. In that case remove
2531 * UID|GID and or MODE from mask so that
2532 * secpolicy_vnode_setattr() doesn't revoke it.
2536 saved_mask = vap->va_mask;
2537 vap->va_mask &= ~trim_mask;
2538 if (trim_mask & AT_MODE) {
2540 * Save the mode, as secpolicy_vnode_setattr()
2541 * will overwrite it with ova.va_mode.
2543 saved_mode = vap->va_mode;
2546 err = secpolicy_vnode_setattr(cr, vp, vap, &oldva, flags,
2547 (int (*)(void *, int, cred_t *))zfs_zaccess_unix, zp);
2554 vap->va_mask |= saved_mask;
2555 if (trim_mask & AT_MODE) {
2557 * Recover the mode after
2558 * secpolicy_vnode_setattr().
2560 vap->va_mode = saved_mode;
2566 * secpolicy_vnode_setattr, or take ownership may have
2569 mask = vap->va_mask;
2571 if ((mask & (AT_UID | AT_GID)) || projid != ZFS_INVALID_PROJID) {
2572 err = sa_lookup(zp->z_sa_hdl, SA_ZPL_XATTR(zfsvfs),
2573 &xattr_obj, sizeof (xattr_obj));
2575 if (err == 0 && xattr_obj) {
2576 err = zfs_zget(zp->z_zfsvfs, xattr_obj, &attrzp);
2578 err = vn_lock(ZTOV(attrzp), LK_EXCLUSIVE);
2580 vrele(ZTOV(attrzp));
2585 if (mask & AT_UID) {
2586 new_uid = zfs_fuid_create(zfsvfs,
2587 (uint64_t)vap->va_uid, cr, ZFS_OWNER, &fuidp);
2588 if (new_uid != zp->z_uid &&
2589 zfs_id_overquota(zfsvfs, DMU_USERUSED_OBJECT,
2593 err = SET_ERROR(EDQUOT);
2598 if (mask & AT_GID) {
2599 new_gid = zfs_fuid_create(zfsvfs, (uint64_t)vap->va_gid,
2600 cr, ZFS_GROUP, &fuidp);
2601 if (new_gid != zp->z_gid &&
2602 zfs_id_overquota(zfsvfs, DMU_GROUPUSED_OBJECT,
2606 err = SET_ERROR(EDQUOT);
2611 if (projid != ZFS_INVALID_PROJID &&
2612 zfs_id_overquota(zfsvfs, DMU_PROJECTUSED_OBJECT, projid)) {
2615 err = SET_ERROR(EDQUOT);
2619 tx = dmu_tx_create(os);
2621 if (mask & AT_MODE) {
2622 uint64_t pmode = zp->z_mode;
2624 new_mode = (pmode & S_IFMT) | (vap->va_mode & ~S_IFMT);
2626 if (zp->z_zfsvfs->z_acl_mode == ZFS_ACL_RESTRICTED &&
2627 !(zp->z_pflags & ZFS_ACL_TRIVIAL)) {
2628 err = SET_ERROR(EPERM);
2632 if ((err = zfs_acl_chmod_setattr(zp, &aclp, new_mode)))
2635 if (!zp->z_is_sa && ((acl_obj = zfs_external_acl(zp)) != 0)) {
2637 * Are we upgrading ACL from old V0 format
2640 if (zfsvfs->z_version >= ZPL_VERSION_FUID &&
2641 zfs_znode_acl_version(zp) ==
2642 ZFS_ACL_VERSION_INITIAL) {
2643 dmu_tx_hold_free(tx, acl_obj, 0,
2645 dmu_tx_hold_write(tx, DMU_NEW_OBJECT,
2646 0, aclp->z_acl_bytes);
2648 dmu_tx_hold_write(tx, acl_obj, 0,
2651 } else if (!zp->z_is_sa && aclp->z_acl_bytes > ZFS_ACE_SPACE) {
2652 dmu_tx_hold_write(tx, DMU_NEW_OBJECT,
2653 0, aclp->z_acl_bytes);
2655 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE);
2657 if (((mask & AT_XVATTR) &&
2658 XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP)) ||
2659 (projid != ZFS_INVALID_PROJID &&
2660 !(zp->z_pflags & ZFS_PROJID)))
2661 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE);
2663 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
2667 dmu_tx_hold_sa(tx, attrzp->z_sa_hdl, B_FALSE);
2670 fuid_dirtied = zfsvfs->z_fuid_dirty;
2672 zfs_fuid_txhold(zfsvfs, tx);
2674 zfs_sa_upgrade_txholds(tx, zp);
2676 err = dmu_tx_assign(tx, TXG_WAIT);
2682 * Set each attribute requested.
2683 * We group settings according to the locks they need to acquire.
2685 * Note: you cannot set ctime directly, although it will be
2686 * updated as a side-effect of calling this function.
2689 if (projid != ZFS_INVALID_PROJID && !(zp->z_pflags & ZFS_PROJID)) {
2691 * For the existed object that is upgraded from old system,
2692 * its on-disk layout has no slot for the project ID attribute.
2693 * But quota accounting logic needs to access related slots by
2694 * offset directly. So we need to adjust old objects' layout
2695 * to make the project ID to some unified and fixed offset.
2698 err = sa_add_projid(attrzp->z_sa_hdl, tx, projid);
2700 err = sa_add_projid(zp->z_sa_hdl, tx, projid);
2702 if (unlikely(err == EEXIST))
2707 projid = ZFS_INVALID_PROJID;
2710 if (mask & (AT_UID|AT_GID|AT_MODE))
2711 mutex_enter(&zp->z_acl_lock);
2713 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL,
2714 &zp->z_pflags, sizeof (zp->z_pflags));
2717 if (mask & (AT_UID|AT_GID|AT_MODE))
2718 mutex_enter(&attrzp->z_acl_lock);
2719 SA_ADD_BULK_ATTR(xattr_bulk, xattr_count,
2720 SA_ZPL_FLAGS(zfsvfs), NULL, &attrzp->z_pflags,
2721 sizeof (attrzp->z_pflags));
2722 if (projid != ZFS_INVALID_PROJID) {
2723 attrzp->z_projid = projid;
2724 SA_ADD_BULK_ATTR(xattr_bulk, xattr_count,
2725 SA_ZPL_PROJID(zfsvfs), NULL, &attrzp->z_projid,
2726 sizeof (attrzp->z_projid));
2730 if (mask & (AT_UID|AT_GID)) {
2732 if (mask & AT_UID) {
2733 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zfsvfs), NULL,
2734 &new_uid, sizeof (new_uid));
2735 zp->z_uid = new_uid;
2737 SA_ADD_BULK_ATTR(xattr_bulk, xattr_count,
2738 SA_ZPL_UID(zfsvfs), NULL, &new_uid,
2740 attrzp->z_uid = new_uid;
2744 if (mask & AT_GID) {
2745 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zfsvfs),
2746 NULL, &new_gid, sizeof (new_gid));
2747 zp->z_gid = new_gid;
2749 SA_ADD_BULK_ATTR(xattr_bulk, xattr_count,
2750 SA_ZPL_GID(zfsvfs), NULL, &new_gid,
2752 attrzp->z_gid = new_gid;
2755 if (!(mask & AT_MODE)) {
2756 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs),
2757 NULL, &new_mode, sizeof (new_mode));
2758 new_mode = zp->z_mode;
2760 err = zfs_acl_chown_setattr(zp);
2763 err = zfs_acl_chown_setattr(attrzp);
2768 if (mask & AT_MODE) {
2769 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs), NULL,
2770 &new_mode, sizeof (new_mode));
2771 zp->z_mode = new_mode;
2772 ASSERT3U((uintptr_t)aclp, !=, 0);
2773 err = zfs_aclset_common(zp, aclp, cr, tx);
2775 if (zp->z_acl_cached)
2776 zfs_acl_free(zp->z_acl_cached);
2777 zp->z_acl_cached = aclp;
2782 if (mask & AT_ATIME) {
2783 ZFS_TIME_ENCODE(&vap->va_atime, zp->z_atime);
2784 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zfsvfs), NULL,
2785 &zp->z_atime, sizeof (zp->z_atime));
2788 if (mask & AT_MTIME) {
2789 ZFS_TIME_ENCODE(&vap->va_mtime, mtime);
2790 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL,
2791 mtime, sizeof (mtime));
2794 if (projid != ZFS_INVALID_PROJID) {
2795 zp->z_projid = projid;
2796 SA_ADD_BULK_ATTR(bulk, count,
2797 SA_ZPL_PROJID(zfsvfs), NULL, &zp->z_projid,
2798 sizeof (zp->z_projid));
2801 /* XXX - shouldn't this be done *before* the ATIME/MTIME checks? */
2802 if (mask & AT_SIZE && !(mask & AT_MTIME)) {
2803 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs),
2804 NULL, mtime, sizeof (mtime));
2805 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL,
2806 &ctime, sizeof (ctime));
2807 zfs_tstamp_update_setup(zp, CONTENT_MODIFIED, mtime, ctime);
2808 } else if (mask != 0) {
2809 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL,
2810 &ctime, sizeof (ctime));
2811 zfs_tstamp_update_setup(zp, STATE_CHANGED, mtime, ctime);
2813 SA_ADD_BULK_ATTR(xattr_bulk, xattr_count,
2814 SA_ZPL_CTIME(zfsvfs), NULL,
2815 &ctime, sizeof (ctime));
2816 zfs_tstamp_update_setup(attrzp, STATE_CHANGED,
2822 * Do this after setting timestamps to prevent timestamp
2823 * update from toggling bit
2826 if (xoap && (mask & AT_XVATTR)) {
2828 if (XVA_ISSET_REQ(xvap, XAT_CREATETIME))
2829 xoap->xoa_createtime = vap->va_birthtime;
2831 * restore trimmed off masks
2832 * so that return masks can be set for caller.
2835 if (XVA_ISSET_REQ(&tmpxvattr, XAT_APPENDONLY)) {
2836 XVA_SET_REQ(xvap, XAT_APPENDONLY);
2838 if (XVA_ISSET_REQ(&tmpxvattr, XAT_NOUNLINK)) {
2839 XVA_SET_REQ(xvap, XAT_NOUNLINK);
2841 if (XVA_ISSET_REQ(&tmpxvattr, XAT_IMMUTABLE)) {
2842 XVA_SET_REQ(xvap, XAT_IMMUTABLE);
2844 if (XVA_ISSET_REQ(&tmpxvattr, XAT_NODUMP)) {
2845 XVA_SET_REQ(xvap, XAT_NODUMP);
2847 if (XVA_ISSET_REQ(&tmpxvattr, XAT_AV_MODIFIED)) {
2848 XVA_SET_REQ(xvap, XAT_AV_MODIFIED);
2850 if (XVA_ISSET_REQ(&tmpxvattr, XAT_AV_QUARANTINED)) {
2851 XVA_SET_REQ(xvap, XAT_AV_QUARANTINED);
2853 if (XVA_ISSET_REQ(&tmpxvattr, XAT_PROJINHERIT)) {
2854 XVA_SET_REQ(xvap, XAT_PROJINHERIT);
2857 if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP))
2858 ASSERT(vp->v_type == VREG);
2860 zfs_xvattr_set(zp, xvap, tx);
2864 zfs_fuid_sync(zfsvfs, tx);
2867 zfs_log_setattr(zilog, tx, TX_SETATTR, zp, vap, mask, fuidp);
2869 if (mask & (AT_UID|AT_GID|AT_MODE))
2870 mutex_exit(&zp->z_acl_lock);
2873 if (mask & (AT_UID|AT_GID|AT_MODE))
2874 mutex_exit(&attrzp->z_acl_lock);
2877 if (err == 0 && attrzp) {
2878 err2 = sa_bulk_update(attrzp->z_sa_hdl, xattr_bulk,
2890 zfs_fuid_info_free(fuidp);
2897 err2 = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx);
2902 if (os->os_sync == ZFS_SYNC_ALWAYS)
2903 zil_commit(zilog, 0);
2910 * We acquire all but fdvp locks using non-blocking acquisitions. If we
2911 * fail to acquire any lock in the path we will drop all held locks,
2912 * acquire the new lock in a blocking fashion, and then release it and
2913 * restart the rename. This acquire/release step ensures that we do not
2914 * spin on a lock waiting for release. On error release all vnode locks
2915 * and decrement references the way tmpfs_rename() would do.
2918 zfs_rename_relock(struct vnode *sdvp, struct vnode **svpp,
2919 struct vnode *tdvp, struct vnode **tvpp,
2920 const struct componentname *scnp, const struct componentname *tcnp)
2923 struct vnode *nvp, *svp, *tvp;
2924 znode_t *sdzp, *tdzp, *szp, *tzp;
2925 const char *snm = scnp->cn_nameptr;
2926 const char *tnm = tcnp->cn_nameptr;
2930 if (*tvpp != NULL && *tvpp != tdvp)
2934 error = vn_lock(sdvp, LK_EXCLUSIVE);
2939 error = vn_lock(tdvp, LK_EXCLUSIVE | LK_NOWAIT);
2944 error = vn_lock(tdvp, LK_EXCLUSIVE);
2953 * Before using sdzp and tdzp we must ensure that they are live.
2954 * As a porting legacy from illumos we have two things to worry
2955 * about. One is typical for FreeBSD and it is that the vnode is
2956 * not reclaimed (doomed). The other is that the znode is live.
2957 * The current code can invalidate the znode without acquiring the
2958 * corresponding vnode lock if the object represented by the znode
2959 * and vnode is no longer valid after a rollback or receive operation.
2960 * z_teardown_lock hidden behind ZFS_ENTER and ZFS_EXIT is the lock
2961 * that protects the znodes from the invalidation.
2963 zfsvfs = sdzp->z_zfsvfs;
2964 ASSERT3P(zfsvfs, ==, tdzp->z_zfsvfs);
2968 * We can not use ZFS_VERIFY_ZP() here because it could directly return
2969 * bypassing the cleanup code in the case of an error.
2971 if (tdzp->z_sa_hdl == NULL || sdzp->z_sa_hdl == NULL) {
2975 error = SET_ERROR(EIO);
2980 * Re-resolve svp to be certain it still exists and fetch the
2983 error = zfs_dirent_lookup(sdzp, snm, &szp, ZEXISTS);
2985 /* Source entry invalid or not there. */
2989 if ((scnp->cn_flags & ISDOTDOT) != 0 ||
2990 (scnp->cn_namelen == 1 && scnp->cn_nameptr[0] == '.'))
2991 error = SET_ERROR(EINVAL);
2997 * Re-resolve tvp, if it disappeared we just carry on.
2999 error = zfs_dirent_lookup(tdzp, tnm, &tzp, 0);
3005 if ((tcnp->cn_flags & ISDOTDOT) != 0)
3006 error = SET_ERROR(EINVAL);
3015 * At present the vnode locks must be acquired before z_teardown_lock,
3016 * although it would be more logical to use the opposite order.
3021 * Now try acquire locks on svp and tvp.
3024 error = vn_lock(nvp, LK_EXCLUSIVE | LK_NOWAIT);
3030 if (error != EBUSY) {
3034 error = vn_lock(nvp, LK_EXCLUSIVE);
3041 * Concurrent rename race.
3046 error = SET_ERROR(EINVAL);
3061 error = vn_lock(nvp, LK_EXCLUSIVE | LK_NOWAIT);
3066 if (error != EBUSY) {
3070 error = vn_lock(nvp, LK_EXCLUSIVE);
3088 * Note that we must use VRELE_ASYNC in this function as it walks
3089 * up the directory tree and vrele may need to acquire an exclusive
3090 * lock if a last reference to a vnode is dropped.
3093 zfs_rename_check(znode_t *szp, znode_t *sdzp, znode_t *tdzp)
3100 zfsvfs = tdzp->z_zfsvfs;
3102 return (SET_ERROR(EINVAL));
3105 if (tdzp->z_id == zfsvfs->z_root)
3109 ASSERT(!zp->z_unlinked);
3110 if ((error = sa_lookup(zp->z_sa_hdl,
3111 SA_ZPL_PARENT(zfsvfs), &parent, sizeof (parent))) != 0)
3114 if (parent == szp->z_id) {
3115 error = SET_ERROR(EINVAL);
3118 if (parent == zfsvfs->z_root)
3120 if (parent == sdzp->z_id)
3123 error = zfs_zget(zfsvfs, parent, &zp1);
3128 VN_RELE_ASYNC(ZTOV(zp),
3129 dsl_pool_zrele_taskq(
3130 dmu_objset_pool(zfsvfs->z_os)));
3134 if (error == ENOTDIR)
3135 panic("checkpath: .. not a directory\n");
3137 VN_RELE_ASYNC(ZTOV(zp),
3138 dsl_pool_zrele_taskq(dmu_objset_pool(zfsvfs->z_os)));
3142 #if __FreeBSD_version < 1300124
3144 cache_vop_rename(struct vnode *fdvp, struct vnode *fvp, struct vnode *tdvp,
3145 struct vnode *tvp, struct componentname *fcnp, struct componentname *tcnp)
3151 cache_purge_negative(tdvp);
3156 * Move an entry from the provided source directory to the target
3157 * directory. Change the entry name as indicated.
3159 * IN: sdvp - Source directory containing the "old entry".
3160 * snm - Old entry name.
3161 * tdvp - Target directory to contain the "new entry".
3162 * tnm - New entry name.
3163 * cr - credentials of caller.
3164 * ct - caller context
3165 * flags - case flags
3167 * RETURN: 0 on success, error code on failure.
3170 * sdvp,tdvp - ctime|mtime updated
3174 zfs_rename_(vnode_t *sdvp, vnode_t **svpp, struct componentname *scnp,
3175 vnode_t *tdvp, vnode_t **tvpp, struct componentname *tcnp,
3176 cred_t *cr, int log)
3179 znode_t *sdzp, *tdzp, *szp, *tzp;
3180 zilog_t *zilog = NULL;
3182 const char *snm = scnp->cn_nameptr;
3183 const char *tnm = tcnp->cn_nameptr;
3185 bool want_seqc_end __maybe_unused = false;
3187 /* Reject renames across filesystems. */
3188 if ((*svpp)->v_mount != tdvp->v_mount ||
3189 ((*tvpp) != NULL && (*svpp)->v_mount != (*tvpp)->v_mount)) {
3190 error = SET_ERROR(EXDEV);
3194 if (zfsctl_is_node(tdvp)) {
3195 error = SET_ERROR(EXDEV);
3200 * Lock all four vnodes to ensure safety and semantics of renaming.
3202 error = zfs_rename_relock(sdvp, svpp, tdvp, tvpp, scnp, tcnp);
3204 /* no vnodes are locked in the case of error here */
3210 zfsvfs = tdzp->z_zfsvfs;
3211 zilog = zfsvfs->z_log;
3214 * After we re-enter ZFS_ENTER() we will have to revalidate all
3219 if (zfsvfs->z_utf8 && u8_validate(tnm,
3220 strlen(tnm), NULL, U8_VALIDATE_ENTIRE, &error) < 0) {
3221 error = SET_ERROR(EILSEQ);
3225 /* If source and target are the same file, there is nothing to do. */
3226 if ((*svpp) == (*tvpp)) {
3231 if (((*svpp)->v_type == VDIR && (*svpp)->v_mountedhere != NULL) ||
3232 ((*tvpp) != NULL && (*tvpp)->v_type == VDIR &&
3233 (*tvpp)->v_mountedhere != NULL)) {
3234 error = SET_ERROR(EXDEV);
3239 * We can not use ZFS_VERIFY_ZP() here because it could directly return
3240 * bypassing the cleanup code in the case of an error.
3242 if (tdzp->z_sa_hdl == NULL || sdzp->z_sa_hdl == NULL) {
3243 error = SET_ERROR(EIO);
3248 tzp = *tvpp == NULL ? NULL : VTOZ(*tvpp);
3249 if (szp->z_sa_hdl == NULL || (tzp != NULL && tzp->z_sa_hdl == NULL)) {
3250 error = SET_ERROR(EIO);
3255 * This is to prevent the creation of links into attribute space
3256 * by renaming a linked file into/outof an attribute directory.
3257 * See the comment in zfs_link() for why this is considered bad.
3259 if ((tdzp->z_pflags & ZFS_XATTR) != (sdzp->z_pflags & ZFS_XATTR)) {
3260 error = SET_ERROR(EINVAL);
3265 * If we are using project inheritance, means if the directory has
3266 * ZFS_PROJINHERIT set, then its descendant directories will inherit
3267 * not only the project ID, but also the ZFS_PROJINHERIT flag. Under
3268 * such case, we only allow renames into our tree when the project
3271 if (tdzp->z_pflags & ZFS_PROJINHERIT &&
3272 tdzp->z_projid != szp->z_projid) {
3273 error = SET_ERROR(EXDEV);
3278 * Must have write access at the source to remove the old entry
3279 * and write access at the target to create the new entry.
3280 * Note that if target and source are the same, this can be
3281 * done in a single check.
3283 if ((error = zfs_zaccess_rename(sdzp, szp, tdzp, tzp, cr)))
3286 if ((*svpp)->v_type == VDIR) {
3288 * Avoid ".", "..", and aliases of "." for obvious reasons.
3290 if ((scnp->cn_namelen == 1 && scnp->cn_nameptr[0] == '.') ||
3292 (scnp->cn_flags | tcnp->cn_flags) & ISDOTDOT) {
3298 * Check to make sure rename is valid.
3299 * Can't do a move like this: /usr/a/b to /usr/a/b/c/d
3301 if ((error = zfs_rename_check(szp, sdzp, tdzp)))
3306 * Does target exist?
3310 * Source and target must be the same type.
3312 if ((*svpp)->v_type == VDIR) {
3313 if ((*tvpp)->v_type != VDIR) {
3314 error = SET_ERROR(ENOTDIR);
3322 if ((*tvpp)->v_type == VDIR) {
3323 error = SET_ERROR(EISDIR);
3329 vn_seqc_write_begin(*svpp);
3330 vn_seqc_write_begin(sdvp);
3332 vn_seqc_write_begin(*tvpp);
3334 vn_seqc_write_begin(tdvp);
3335 #if __FreeBSD_version >= 1300102
3336 want_seqc_end = true;
3338 vnevent_rename_src(*svpp, sdvp, scnp->cn_nameptr, ct);
3340 vnevent_rename_dest(*tvpp, tdvp, tnm, ct);
3343 * notify the target directory if it is not the same
3344 * as source directory.
3347 vnevent_rename_dest_dir(tdvp, ct);
3350 tx = dmu_tx_create(zfsvfs->z_os);
3351 dmu_tx_hold_sa(tx, szp->z_sa_hdl, B_FALSE);
3352 dmu_tx_hold_sa(tx, sdzp->z_sa_hdl, B_FALSE);
3353 dmu_tx_hold_zap(tx, sdzp->z_id, FALSE, snm);
3354 dmu_tx_hold_zap(tx, tdzp->z_id, TRUE, tnm);
3356 dmu_tx_hold_sa(tx, tdzp->z_sa_hdl, B_FALSE);
3357 zfs_sa_upgrade_txholds(tx, tdzp);
3360 dmu_tx_hold_sa(tx, tzp->z_sa_hdl, B_FALSE);
3361 zfs_sa_upgrade_txholds(tx, tzp);
3364 zfs_sa_upgrade_txholds(tx, szp);
3365 dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL);
3366 error = dmu_tx_assign(tx, TXG_WAIT);
3373 if (tzp) /* Attempt to remove the existing target */
3374 error = zfs_link_destroy(tdzp, tnm, tzp, tx, 0, NULL);
3377 error = zfs_link_create(tdzp, tnm, szp, tx, ZRENAMING);
3379 szp->z_pflags |= ZFS_AV_MODIFIED;
3381 error = sa_update(szp->z_sa_hdl, SA_ZPL_FLAGS(zfsvfs),
3382 (void *)&szp->z_pflags, sizeof (uint64_t), tx);
3385 error = zfs_link_destroy(sdzp, snm, szp, tx, ZRENAMING,
3388 zfs_log_rename(zilog, tx, TX_RENAME, sdzp,
3389 snm, tdzp, tnm, szp);
3392 * Update path information for the target vnode
3394 vn_renamepath(tdvp, *svpp, tnm, strlen(tnm));
3397 * At this point, we have successfully created
3398 * the target name, but have failed to remove
3399 * the source name. Since the create was done
3400 * with the ZRENAMING flag, there are
3401 * complications; for one, the link count is
3402 * wrong. The easiest way to deal with this
3403 * is to remove the newly created target, and
3404 * return the original error. This must
3405 * succeed; fortunately, it is very unlikely to
3406 * fail, since we just created it.
3408 VERIFY3U(zfs_link_destroy(tdzp, tnm, szp, tx,
3409 ZRENAMING, NULL), ==, 0);
3413 cache_vop_rename(sdvp, *svpp, tdvp, *tvpp, scnp, tcnp);
3419 unlockout: /* all 4 vnodes are locked, ZFS_ENTER called */
3421 if (want_seqc_end) {
3422 vn_seqc_write_end(*svpp);
3423 vn_seqc_write_end(sdvp);
3425 vn_seqc_write_end(*tvpp);
3427 vn_seqc_write_end(tdvp);
3428 want_seqc_end = false;
3433 out: /* original two vnodes are locked */
3434 MPASS(!want_seqc_end);
3435 if (error == 0 && zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS)
3436 zil_commit(zilog, 0);
3446 zfs_rename(znode_t *sdzp, const char *sname, znode_t *tdzp, const char *tname,
3447 cred_t *cr, int flags)
3449 struct componentname scn, tcn;
3450 vnode_t *sdvp, *tdvp;
3457 error = zfs_lookup_internal(sdzp, sname, &svp, &scn, DELETE);
3458 if (sdzp->z_zfsvfs->z_replay == B_FALSE)
3464 vn_lock(tdvp, LK_EXCLUSIVE | LK_RETRY);
3465 error = zfs_lookup_internal(tdzp, tname, &tvp, &tcn, RENAME);
3466 if (error == EJUSTRETURN)
3468 else if (error != 0) {
3473 error = zfs_rename_(sdvp, &svp, &scn, tdvp, &tvp, &tcn, cr, 0);
3484 * Insert the indicated symbolic reference entry into the directory.
3486 * IN: dvp - Directory to contain new symbolic link.
3487 * link - Name for new symlink entry.
3488 * vap - Attributes of new entry.
3489 * cr - credentials of caller.
3490 * ct - caller context
3491 * flags - case flags
3493 * RETURN: 0 on success, error code on failure.
3496 * dvp - ctime|mtime updated
3500 zfs_symlink(znode_t *dzp, const char *name, vattr_t *vap,
3501 const char *link, znode_t **zpp, cred_t *cr, int flags)
3505 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
3507 uint64_t len = strlen(link);
3509 zfs_acl_ids_t acl_ids;
3510 boolean_t fuid_dirtied;
3511 uint64_t txtype = TX_SYMLINK;
3513 ASSERT(vap->va_type == VLNK);
3517 zilog = zfsvfs->z_log;
3519 if (zfsvfs->z_utf8 && u8_validate(name, strlen(name),
3520 NULL, U8_VALIDATE_ENTIRE, &error) < 0) {
3522 return (SET_ERROR(EILSEQ));
3525 if (len > MAXPATHLEN) {
3527 return (SET_ERROR(ENAMETOOLONG));
3530 if ((error = zfs_acl_ids_create(dzp, 0,
3531 vap, cr, NULL, &acl_ids)) != 0) {
3537 * Attempt to lock directory; fail if entry already exists.
3539 error = zfs_dirent_lookup(dzp, name, &zp, ZNEW);
3541 zfs_acl_ids_free(&acl_ids);
3546 if ((error = zfs_zaccess(dzp, ACE_ADD_FILE, 0, B_FALSE, cr))) {
3547 zfs_acl_ids_free(&acl_ids);
3552 if (zfs_acl_ids_overquota(zfsvfs, &acl_ids,
3554 zfs_acl_ids_free(&acl_ids);
3556 return (SET_ERROR(EDQUOT));
3559 getnewvnode_reserve_();
3560 tx = dmu_tx_create(zfsvfs->z_os);
3561 fuid_dirtied = zfsvfs->z_fuid_dirty;
3562 dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, MAX(1, len));
3563 dmu_tx_hold_zap(tx, dzp->z_id, TRUE, name);
3564 dmu_tx_hold_sa_create(tx, acl_ids.z_aclp->z_acl_bytes +
3565 ZFS_SA_BASE_ATTR_SIZE + len);
3566 dmu_tx_hold_sa(tx, dzp->z_sa_hdl, B_FALSE);
3567 if (!zfsvfs->z_use_sa && acl_ids.z_aclp->z_acl_bytes > ZFS_ACE_SPACE) {
3568 dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0,
3569 acl_ids.z_aclp->z_acl_bytes);
3572 zfs_fuid_txhold(zfsvfs, tx);
3573 error = dmu_tx_assign(tx, TXG_WAIT);
3575 zfs_acl_ids_free(&acl_ids);
3577 getnewvnode_drop_reserve();
3583 * Create a new object for the symlink.
3584 * for version 4 ZPL datsets the symlink will be an SA attribute
3586 zfs_mknode(dzp, vap, tx, cr, 0, &zp, &acl_ids);
3589 zfs_fuid_sync(zfsvfs, tx);
3592 error = sa_update(zp->z_sa_hdl, SA_ZPL_SYMLINK(zfsvfs),
3593 __DECONST(void *, link), len, tx);
3595 zfs_sa_symlink(zp, __DECONST(char *, link), len, tx);
3598 (void) sa_update(zp->z_sa_hdl, SA_ZPL_SIZE(zfsvfs),
3599 &zp->z_size, sizeof (zp->z_size), tx);
3601 * Insert the new object into the directory.
3603 (void) zfs_link_create(dzp, name, zp, tx, ZNEW);
3605 zfs_log_symlink(zilog, tx, txtype, dzp, zp, name, link);
3608 zfs_acl_ids_free(&acl_ids);
3612 getnewvnode_drop_reserve();
3614 if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS)
3615 zil_commit(zilog, 0);
3622 * Return, in the buffer contained in the provided uio structure,
3623 * the symbolic path referred to by vp.
3625 * IN: vp - vnode of symbolic link.
3626 * uio - structure to contain the link path.
3627 * cr - credentials of caller.
3628 * ct - caller context
3630 * OUT: uio - structure containing the link path.
3632 * RETURN: 0 on success, error code on failure.
3635 * vp - atime updated
3639 zfs_readlink(vnode_t *vp, zfs_uio_t *uio, cred_t *cr, caller_context_t *ct)
3641 znode_t *zp = VTOZ(vp);
3642 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
3649 error = sa_lookup_uio(zp->z_sa_hdl,
3650 SA_ZPL_SYMLINK(zfsvfs), uio);
3652 error = zfs_sa_readlink(zp, uio);
3654 ZFS_ACCESSTIME_STAMP(zfsvfs, zp);
3661 * Insert a new entry into directory tdvp referencing svp.
3663 * IN: tdvp - Directory to contain new entry.
3664 * svp - vnode of new entry.
3665 * name - name of new entry.
3666 * cr - credentials of caller.
3668 * RETURN: 0 on success, error code on failure.
3671 * tdvp - ctime|mtime updated
3672 * svp - ctime updated
3676 zfs_link(znode_t *tdzp, znode_t *szp, const char *name, cred_t *cr,
3680 zfsvfs_t *zfsvfs = tdzp->z_zfsvfs;
3687 ASSERT(ZTOV(tdzp)->v_type == VDIR);
3690 ZFS_VERIFY_ZP(tdzp);
3691 zilog = zfsvfs->z_log;
3694 * POSIX dictates that we return EPERM here.
3695 * Better choices include ENOTSUP or EISDIR.
3697 if (ZTOV(szp)->v_type == VDIR) {
3699 return (SET_ERROR(EPERM));
3705 * If we are using project inheritance, means if the directory has
3706 * ZFS_PROJINHERIT set, then its descendant directories will inherit
3707 * not only the project ID, but also the ZFS_PROJINHERIT flag. Under
3708 * such case, we only allow hard link creation in our tree when the
3709 * project IDs are the same.
3711 if (tdzp->z_pflags & ZFS_PROJINHERIT &&
3712 tdzp->z_projid != szp->z_projid) {
3714 return (SET_ERROR(EXDEV));
3717 if (szp->z_pflags & (ZFS_APPENDONLY |
3718 ZFS_IMMUTABLE | ZFS_READONLY)) {
3720 return (SET_ERROR(EPERM));
3723 /* Prevent links to .zfs/shares files */
3725 if ((error = sa_lookup(szp->z_sa_hdl, SA_ZPL_PARENT(zfsvfs),
3726 &parent, sizeof (uint64_t))) != 0) {
3730 if (parent == zfsvfs->z_shares_dir) {
3732 return (SET_ERROR(EPERM));
3735 if (zfsvfs->z_utf8 && u8_validate(name,
3736 strlen(name), NULL, U8_VALIDATE_ENTIRE, &error) < 0) {
3738 return (SET_ERROR(EILSEQ));
3742 * We do not support links between attributes and non-attributes
3743 * because of the potential security risk of creating links
3744 * into "normal" file space in order to circumvent restrictions
3745 * imposed in attribute space.
3747 if ((szp->z_pflags & ZFS_XATTR) != (tdzp->z_pflags & ZFS_XATTR)) {
3749 return (SET_ERROR(EINVAL));
3753 owner = zfs_fuid_map_id(zfsvfs, szp->z_uid, cr, ZFS_OWNER);
3754 if (owner != crgetuid(cr) && secpolicy_basic_link(ZTOV(szp), cr) != 0) {
3756 return (SET_ERROR(EPERM));
3759 if ((error = zfs_zaccess(tdzp, ACE_ADD_FILE, 0, B_FALSE, cr))) {
3765 * Attempt to lock directory; fail if entry already exists.
3767 error = zfs_dirent_lookup(tdzp, name, &tzp, ZNEW);
3773 tx = dmu_tx_create(zfsvfs->z_os);
3774 dmu_tx_hold_sa(tx, szp->z_sa_hdl, B_FALSE);
3775 dmu_tx_hold_zap(tx, tdzp->z_id, TRUE, name);
3776 zfs_sa_upgrade_txholds(tx, szp);
3777 zfs_sa_upgrade_txholds(tx, tdzp);
3778 error = dmu_tx_assign(tx, TXG_WAIT);
3785 error = zfs_link_create(tdzp, name, szp, tx, 0);
3788 uint64_t txtype = TX_LINK;
3789 zfs_log_link(zilog, tx, txtype, tdzp, szp, name);
3795 vnevent_link(ZTOV(szp), ct);
3798 if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS)
3799 zil_commit(zilog, 0);
3806 * Free or allocate space in a file. Currently, this function only
3807 * supports the `F_FREESP' command. However, this command is somewhat
3808 * misnamed, as its functionality includes the ability to allocate as
3809 * well as free space.
3811 * IN: ip - inode of file to free data in.
3812 * cmd - action to take (only F_FREESP supported).
3813 * bfp - section of file to free/alloc.
3814 * flag - current file open mode flags.
3815 * offset - current file offset.
3816 * cr - credentials of caller.
3818 * RETURN: 0 on success, error code on failure.
3821 * ip - ctime|mtime updated
3825 zfs_space(znode_t *zp, int cmd, flock64_t *bfp, int flag,
3826 offset_t offset, cred_t *cr)
3828 zfsvfs_t *zfsvfs = ZTOZSB(zp);
3835 if (cmd != F_FREESP) {
3837 return (SET_ERROR(EINVAL));
3841 * Callers might not be able to detect properly that we are read-only,
3842 * so check it explicitly here.
3844 if (zfs_is_readonly(zfsvfs)) {
3846 return (SET_ERROR(EROFS));
3849 if (bfp->l_len < 0) {
3851 return (SET_ERROR(EINVAL));
3855 * Permissions aren't checked on Solaris because on this OS
3856 * zfs_space() can only be called with an opened file handle.
3857 * On Linux we can get here through truncate_range() which
3858 * operates directly on inodes, so we need to check access rights.
3860 if ((error = zfs_zaccess(zp, ACE_WRITE_DATA, 0, B_FALSE, cr))) {
3866 len = bfp->l_len; /* 0 means from off to end of file */
3868 error = zfs_freesp(zp, off, len, flag, TRUE);
3876 zfs_inactive(vnode_t *vp, cred_t *cr, caller_context_t *ct)
3878 znode_t *zp = VTOZ(vp);
3879 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
3882 ZFS_RLOCK_TEARDOWN_INACTIVE(zfsvfs);
3883 if (zp->z_sa_hdl == NULL) {
3885 * The fs has been unmounted, or we did a
3886 * suspend/resume and this file no longer exists.
3888 ZFS_RUNLOCK_TEARDOWN_INACTIVE(zfsvfs);
3893 if (zp->z_unlinked) {
3895 * Fast path to recycle a vnode of a removed file.
3897 ZFS_RUNLOCK_TEARDOWN_INACTIVE(zfsvfs);
3902 if (zp->z_atime_dirty && zp->z_unlinked == 0) {
3903 dmu_tx_t *tx = dmu_tx_create(zfsvfs->z_os);
3905 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
3906 zfs_sa_upgrade_txholds(tx, zp);
3907 error = dmu_tx_assign(tx, TXG_WAIT);
3911 (void) sa_update(zp->z_sa_hdl, SA_ZPL_ATIME(zfsvfs),
3912 (void *)&zp->z_atime, sizeof (zp->z_atime), tx);
3913 zp->z_atime_dirty = 0;
3917 ZFS_RUNLOCK_TEARDOWN_INACTIVE(zfsvfs);
3921 CTASSERT(sizeof (struct zfid_short) <= sizeof (struct fid));
3922 CTASSERT(sizeof (struct zfid_long) <= sizeof (struct fid));
3926 zfs_fid(vnode_t *vp, fid_t *fidp, caller_context_t *ct)
3928 znode_t *zp = VTOZ(vp);
3929 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
3932 uint64_t object = zp->z_id;
3939 if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_GEN(zfsvfs),
3940 &gen64, sizeof (uint64_t))) != 0) {
3945 gen = (uint32_t)gen64;
3947 size = (zfsvfs->z_parent != zfsvfs) ? LONG_FID_LEN : SHORT_FID_LEN;
3948 fidp->fid_len = size;
3950 zfid = (zfid_short_t *)fidp;
3952 zfid->zf_len = size;
3954 for (i = 0; i < sizeof (zfid->zf_object); i++)
3955 zfid->zf_object[i] = (uint8_t)(object >> (8 * i));
3957 /* Must have a non-zero generation number to distinguish from .zfs */
3960 for (i = 0; i < sizeof (zfid->zf_gen); i++)
3961 zfid->zf_gen[i] = (uint8_t)(gen >> (8 * i));
3963 if (size == LONG_FID_LEN) {
3964 uint64_t objsetid = dmu_objset_id(zfsvfs->z_os);
3967 zlfid = (zfid_long_t *)fidp;
3969 for (i = 0; i < sizeof (zlfid->zf_setid); i++)
3970 zlfid->zf_setid[i] = (uint8_t)(objsetid >> (8 * i));
3972 /* XXX - this should be the generation number for the objset */
3973 for (i = 0; i < sizeof (zlfid->zf_setgen); i++)
3974 zlfid->zf_setgen[i] = 0;
3982 zfs_pathconf(vnode_t *vp, int cmd, ulong_t *valp, cred_t *cr,
3983 caller_context_t *ct)
3990 *valp = MIN(LONG_MAX, ZFS_LINK_MAX);
3993 case _PC_FILESIZEBITS:
3996 case _PC_MIN_HOLE_SIZE:
3997 *valp = (int)SPA_MINBLOCKSIZE;
3999 case _PC_ACL_EXTENDED:
4000 #if 0 /* POSIX ACLs are not implemented for ZFS on FreeBSD yet. */
4002 zfsvfs = zp->z_zfsvfs;
4005 *valp = zfsvfs->z_acl_type == ZFSACLTYPE_POSIX ? 1 : 0;
4014 zfsvfs = zp->z_zfsvfs;
4017 *valp = zfsvfs->z_acl_type == ZFS_ACLTYPE_NFSV4 ? 1 : 0;
4021 case _PC_ACL_PATH_MAX:
4022 *valp = ACL_MAX_ENTRIES;
4026 return (EOPNOTSUPP);
4031 zfs_getpages(struct vnode *vp, vm_page_t *ma, int count, int *rbehind,
4034 znode_t *zp = VTOZ(vp);
4035 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
4036 objset_t *os = zp->z_zfsvfs->z_os;
4037 zfs_locked_range_t *lr;
4039 off_t start, end, obj_size;
4041 int pgsin_b, pgsin_a;
4047 start = IDX_TO_OFF(ma[0]->pindex);
4048 end = IDX_TO_OFF(ma[count - 1]->pindex + 1);
4051 * Lock a range covering all required and optional pages.
4052 * Note that we need to handle the case of the block size growing.
4055 blksz = zp->z_blksz;
4056 lr = zfs_rangelock_tryenter(&zp->z_rangelock,
4057 rounddown(start, blksz),
4058 roundup(end, blksz) - rounddown(start, blksz), RL_READER);
4060 if (rahead != NULL) {
4064 if (rbehind != NULL) {
4070 if (blksz == zp->z_blksz)
4072 zfs_rangelock_exit(lr);
4075 object = ma[0]->object;
4076 zfs_vmobject_wlock(object);
4077 obj_size = object->un_pager.vnp.vnp_size;
4078 zfs_vmobject_wunlock(object);
4079 if (IDX_TO_OFF(ma[count - 1]->pindex) >= obj_size) {
4081 zfs_rangelock_exit(lr);
4083 return (zfs_vm_pagerret_bad);
4087 if (rbehind != NULL) {
4088 pgsin_b = OFF_TO_IDX(start - rounddown(start, blksz));
4089 pgsin_b = MIN(*rbehind, pgsin_b);
4093 if (rahead != NULL) {
4094 pgsin_a = OFF_TO_IDX(roundup(end, blksz) - end);
4095 if (end + IDX_TO_OFF(pgsin_a) >= obj_size)
4096 pgsin_a = OFF_TO_IDX(round_page(obj_size) - end);
4097 pgsin_a = MIN(*rahead, pgsin_a);
4101 * NB: we need to pass the exact byte size of the data that we expect
4102 * to read after accounting for the file size. This is required because
4103 * ZFS will panic if we request DMU to read beyond the end of the last
4106 error = dmu_read_pages(os, zp->z_id, ma, count, &pgsin_b, &pgsin_a,
4107 MIN(end, obj_size) - (end - PAGE_SIZE));
4110 zfs_rangelock_exit(lr);
4111 ZFS_ACCESSTIME_STAMP(zfsvfs, zp);
4115 return (zfs_vm_pagerret_error);
4117 VM_CNT_INC(v_vnodein);
4118 VM_CNT_ADD(v_vnodepgsin, count + pgsin_b + pgsin_a);
4119 if (rbehind != NULL)
4123 return (zfs_vm_pagerret_ok);
4126 #ifndef _SYS_SYSPROTO_H_
4127 struct vop_getpages_args {
4137 zfs_freebsd_getpages(struct vop_getpages_args *ap)
4140 return (zfs_getpages(ap->a_vp, ap->a_m, ap->a_count, ap->a_rbehind,
4145 zfs_putpages(struct vnode *vp, vm_page_t *ma, size_t len, int flags,
4148 znode_t *zp = VTOZ(vp);
4149 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
4150 zfs_locked_range_t *lr;
4158 vm_ooffset_t lo_off;
4169 object = vp->v_object;
4173 KASSERT(ma[0]->object == object, ("mismatching object"));
4174 KASSERT(len > 0 && (len & PAGE_MASK) == 0, ("unexpected length"));
4176 for (i = 0; i < pcount; i++)
4177 rtvals[i] = zfs_vm_pagerret_error;
4179 off = IDX_TO_OFF(ma[0]->pindex);
4180 blksz = zp->z_blksz;
4181 lo_off = rounddown(off, blksz);
4182 lo_len = roundup(len + (off - lo_off), blksz);
4183 lr = zfs_rangelock_enter(&zp->z_rangelock, lo_off, lo_len, RL_WRITER);
4185 zfs_vmobject_wlock(object);
4186 if (len + off > object->un_pager.vnp.vnp_size) {
4187 if (object->un_pager.vnp.vnp_size > off) {
4190 len = object->un_pager.vnp.vnp_size - off;
4192 if ((pgoff = (int)len & PAGE_MASK) != 0) {
4194 * If the object is locked and the following
4195 * conditions hold, then the page's dirty
4196 * field cannot be concurrently changed by a
4200 vm_page_assert_sbusied(m);
4201 KASSERT(!pmap_page_is_write_mapped(m),
4202 ("zfs_putpages: page %p is not read-only",
4204 vm_page_clear_dirty(m, pgoff, PAGE_SIZE -
4211 if (ncount < pcount) {
4212 for (i = ncount; i < pcount; i++) {
4213 rtvals[i] = zfs_vm_pagerret_bad;
4217 zfs_vmobject_wunlock(object);
4222 if (zfs_id_overblockquota(zfsvfs, DMU_USERUSED_OBJECT, zp->z_uid) ||
4223 zfs_id_overblockquota(zfsvfs, DMU_GROUPUSED_OBJECT, zp->z_gid) ||
4224 (zp->z_projid != ZFS_DEFAULT_PROJID &&
4225 zfs_id_overblockquota(zfsvfs, DMU_PROJECTUSED_OBJECT,
4230 tx = dmu_tx_create(zfsvfs->z_os);
4231 dmu_tx_hold_write(tx, zp->z_id, off, len);
4233 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
4234 zfs_sa_upgrade_txholds(tx, zp);
4235 err = dmu_tx_assign(tx, TXG_WAIT);
4241 if (zp->z_blksz < PAGE_SIZE) {
4242 for (i = 0; len > 0; off += tocopy, len -= tocopy, i++) {
4243 tocopy = len > PAGE_SIZE ? PAGE_SIZE : len;
4244 va = zfs_map_page(ma[i], &sf);
4245 dmu_write(zfsvfs->z_os, zp->z_id, off, tocopy, va, tx);
4249 err = dmu_write_pages(zfsvfs->z_os, zp->z_id, off, len, ma, tx);
4253 uint64_t mtime[2], ctime[2];
4254 sa_bulk_attr_t bulk[3];
4257 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL,
4259 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL,
4261 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL,
4263 zfs_tstamp_update_setup(zp, CONTENT_MODIFIED, mtime, ctime);
4264 err = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx);
4267 * XXX we should be passing a callback to undirty
4268 * but that would make the locking messier
4270 zfs_log_write(zfsvfs->z_log, tx, TX_WRITE, zp, off,
4271 len, 0, NULL, NULL);
4273 zfs_vmobject_wlock(object);
4274 for (i = 0; i < ncount; i++) {
4275 rtvals[i] = zfs_vm_pagerret_ok;
4276 vm_page_undirty(ma[i]);
4278 zfs_vmobject_wunlock(object);
4279 VM_CNT_INC(v_vnodeout);
4280 VM_CNT_ADD(v_vnodepgsout, ncount);
4285 zfs_rangelock_exit(lr);
4286 if ((flags & (zfs_vm_pagerput_sync | zfs_vm_pagerput_inval)) != 0 ||
4287 zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS)
4288 zil_commit(zfsvfs->z_log, zp->z_id);
4293 #ifndef _SYS_SYSPROTO_H_
4294 struct vop_putpages_args {
4304 zfs_freebsd_putpages(struct vop_putpages_args *ap)
4307 return (zfs_putpages(ap->a_vp, ap->a_m, ap->a_count, ap->a_sync,
4311 #ifndef _SYS_SYSPROTO_H_
4312 struct vop_bmap_args {
4315 struct bufobj **a_bop;
4323 zfs_freebsd_bmap(struct vop_bmap_args *ap)
4326 if (ap->a_bop != NULL)
4327 *ap->a_bop = &ap->a_vp->v_bufobj;
4328 if (ap->a_bnp != NULL)
4329 *ap->a_bnp = ap->a_bn;
4330 if (ap->a_runp != NULL)
4332 if (ap->a_runb != NULL)
4338 #ifndef _SYS_SYSPROTO_H_
4339 struct vop_open_args {
4342 struct ucred *a_cred;
4343 struct thread *a_td;
4348 zfs_freebsd_open(struct vop_open_args *ap)
4350 vnode_t *vp = ap->a_vp;
4351 znode_t *zp = VTOZ(vp);
4354 error = zfs_open(&vp, ap->a_mode, ap->a_cred);
4356 vnode_create_vobject(vp, zp->z_size, ap->a_td);
4360 #ifndef _SYS_SYSPROTO_H_
4361 struct vop_close_args {
4364 struct ucred *a_cred;
4365 struct thread *a_td;
4370 zfs_freebsd_close(struct vop_close_args *ap)
4373 return (zfs_close(ap->a_vp, ap->a_fflag, 1, 0, ap->a_cred));
4376 #ifndef _SYS_SYSPROTO_H_
4377 struct vop_ioctl_args {
4388 zfs_freebsd_ioctl(struct vop_ioctl_args *ap)
4391 return (zfs_ioctl(ap->a_vp, ap->a_command, (intptr_t)ap->a_data,
4392 ap->a_fflag, ap->a_cred, NULL));
4396 ioflags(int ioflags)
4400 if (ioflags & IO_APPEND)
4402 if (ioflags & IO_NDELAY)
4404 if (ioflags & IO_SYNC)
4405 flags |= (FSYNC | FDSYNC | FRSYNC);
4410 #ifndef _SYS_SYSPROTO_H_
4411 struct vop_read_args {
4415 struct ucred *a_cred;
4420 zfs_freebsd_read(struct vop_read_args *ap)
4423 zfs_uio_init(&uio, ap->a_uio);
4424 return (zfs_read(VTOZ(ap->a_vp), &uio, ioflags(ap->a_ioflag),
4428 #ifndef _SYS_SYSPROTO_H_
4429 struct vop_write_args {
4433 struct ucred *a_cred;
4438 zfs_freebsd_write(struct vop_write_args *ap)
4441 zfs_uio_init(&uio, ap->a_uio);
4442 return (zfs_write(VTOZ(ap->a_vp), &uio, ioflags(ap->a_ioflag),
4446 #if __FreeBSD_version >= 1300102
4448 * VOP_FPLOOKUP_VEXEC routines are subject to special circumstances, see
4449 * the comment above cache_fplookup for details.
4452 zfs_freebsd_fplookup_vexec(struct vop_fplookup_vexec_args *v)
4460 if (__predict_false(zp == NULL))
4462 pflags = atomic_load_64(&zp->z_pflags);
4463 if (pflags & ZFS_AV_QUARANTINED)
4465 if (pflags & ZFS_XATTR)
4467 if ((pflags & ZFS_NO_EXECS_DENIED) == 0)
4473 #ifndef _SYS_SYSPROTO_H_
4474 struct vop_access_args {
4476 accmode_t a_accmode;
4477 struct ucred *a_cred;
4478 struct thread *a_td;
4483 zfs_freebsd_access(struct vop_access_args *ap)
4485 vnode_t *vp = ap->a_vp;
4486 znode_t *zp = VTOZ(vp);
4491 if (ap->a_accmode == VEXEC) {
4492 if (zfs_fastaccesschk_execute(zp, ap->a_cred) == 0)
4497 * ZFS itself only knowns about VREAD, VWRITE, VEXEC and VAPPEND,
4499 accmode = ap->a_accmode & (VREAD|VWRITE|VEXEC|VAPPEND);
4501 error = zfs_access(zp, accmode, 0, ap->a_cred);
4504 * VADMIN has to be handled by vaccess().
4507 accmode = ap->a_accmode & ~(VREAD|VWRITE|VEXEC|VAPPEND);
4509 #if __FreeBSD_version >= 1300105
4510 error = vaccess(vp->v_type, zp->z_mode, zp->z_uid,
4511 zp->z_gid, accmode, ap->a_cred);
4513 error = vaccess(vp->v_type, zp->z_mode, zp->z_uid,
4514 zp->z_gid, accmode, ap->a_cred, NULL);
4520 * For VEXEC, ensure that at least one execute bit is set for
4523 if (error == 0 && (ap->a_accmode & VEXEC) != 0 && vp->v_type != VDIR &&
4524 (zp->z_mode & (S_IXUSR | S_IXGRP | S_IXOTH)) == 0) {
4531 #ifndef _SYS_SYSPROTO_H_
4532 struct vop_lookup_args {
4533 struct vnode *a_dvp;
4534 struct vnode **a_vpp;
4535 struct componentname *a_cnp;
4540 zfs_freebsd_lookup(struct vop_lookup_args *ap, boolean_t cached)
4542 struct componentname *cnp = ap->a_cnp;
4543 char nm[NAME_MAX + 1];
4545 ASSERT(cnp->cn_namelen < sizeof (nm));
4546 strlcpy(nm, cnp->cn_nameptr, MIN(cnp->cn_namelen + 1, sizeof (nm)));
4548 return (zfs_lookup(ap->a_dvp, nm, ap->a_vpp, cnp, cnp->cn_nameiop,
4549 cnp->cn_cred, cnp->cn_thread, 0, cached));
4553 zfs_freebsd_cachedlookup(struct vop_cachedlookup_args *ap)
4556 return (zfs_freebsd_lookup((struct vop_lookup_args *)ap, B_TRUE));
4559 #ifndef _SYS_SYSPROTO_H_
4560 struct vop_lookup_args {
4561 struct vnode *a_dvp;
4562 struct vnode **a_vpp;
4563 struct componentname *a_cnp;
4568 zfs_cache_lookup(struct vop_lookup_args *ap)
4572 zfsvfs = ap->a_dvp->v_mount->mnt_data;
4573 if (zfsvfs->z_use_namecache)
4574 return (vfs_cache_lookup(ap));
4576 return (zfs_freebsd_lookup(ap, B_FALSE));
4579 #ifndef _SYS_SYSPROTO_H_
4580 struct vop_create_args {
4581 struct vnode *a_dvp;
4582 struct vnode **a_vpp;
4583 struct componentname *a_cnp;
4584 struct vattr *a_vap;
4589 zfs_freebsd_create(struct vop_create_args *ap)
4592 struct componentname *cnp = ap->a_cnp;
4593 vattr_t *vap = ap->a_vap;
4597 ASSERT(cnp->cn_flags & SAVENAME);
4599 vattr_init_mask(vap);
4600 mode = vap->va_mode & ALLPERMS;
4601 zfsvfs = ap->a_dvp->v_mount->mnt_data;
4604 rc = zfs_create(VTOZ(ap->a_dvp), cnp->cn_nameptr, vap, !EXCL, mode,
4605 &zp, cnp->cn_cred, 0 /* flag */, NULL /* vsecattr */);
4607 *ap->a_vpp = ZTOV(zp);
4608 if (zfsvfs->z_use_namecache &&
4609 rc == 0 && (cnp->cn_flags & MAKEENTRY) != 0)
4610 cache_enter(ap->a_dvp, *ap->a_vpp, cnp);
4615 #ifndef _SYS_SYSPROTO_H_
4616 struct vop_remove_args {
4617 struct vnode *a_dvp;
4619 struct componentname *a_cnp;
4624 zfs_freebsd_remove(struct vop_remove_args *ap)
4627 ASSERT(ap->a_cnp->cn_flags & SAVENAME);
4629 return (zfs_remove_(ap->a_dvp, ap->a_vp, ap->a_cnp->cn_nameptr,
4630 ap->a_cnp->cn_cred));
4633 #ifndef _SYS_SYSPROTO_H_
4634 struct vop_mkdir_args {
4635 struct vnode *a_dvp;
4636 struct vnode **a_vpp;
4637 struct componentname *a_cnp;
4638 struct vattr *a_vap;
4643 zfs_freebsd_mkdir(struct vop_mkdir_args *ap)
4645 vattr_t *vap = ap->a_vap;
4649 ASSERT(ap->a_cnp->cn_flags & SAVENAME);
4651 vattr_init_mask(vap);
4654 rc = zfs_mkdir(VTOZ(ap->a_dvp), ap->a_cnp->cn_nameptr, vap, &zp,
4655 ap->a_cnp->cn_cred, 0, NULL);
4658 *ap->a_vpp = ZTOV(zp);
4662 #ifndef _SYS_SYSPROTO_H_
4663 struct vop_rmdir_args {
4664 struct vnode *a_dvp;
4666 struct componentname *a_cnp;
4671 zfs_freebsd_rmdir(struct vop_rmdir_args *ap)
4673 struct componentname *cnp = ap->a_cnp;
4675 ASSERT(cnp->cn_flags & SAVENAME);
4677 return (zfs_rmdir_(ap->a_dvp, ap->a_vp, cnp->cn_nameptr, cnp->cn_cred));
4680 #ifndef _SYS_SYSPROTO_H_
4681 struct vop_readdir_args {
4684 struct ucred *a_cred;
4687 ulong_t **a_cookies;
4692 zfs_freebsd_readdir(struct vop_readdir_args *ap)
4695 zfs_uio_init(&uio, ap->a_uio);
4696 return (zfs_readdir(ap->a_vp, &uio, ap->a_cred, ap->a_eofflag,
4697 ap->a_ncookies, ap->a_cookies));
4700 #ifndef _SYS_SYSPROTO_H_
4701 struct vop_fsync_args {
4704 struct thread *a_td;
4709 zfs_freebsd_fsync(struct vop_fsync_args *ap)
4713 return (zfs_fsync(VTOZ(ap->a_vp), 0, ap->a_td->td_ucred));
4716 #ifndef _SYS_SYSPROTO_H_
4717 struct vop_getattr_args {
4719 struct vattr *a_vap;
4720 struct ucred *a_cred;
4725 zfs_freebsd_getattr(struct vop_getattr_args *ap)
4727 vattr_t *vap = ap->a_vap;
4733 xvap.xva_vattr = *vap;
4734 xvap.xva_vattr.va_mask |= AT_XVATTR;
4736 /* Convert chflags into ZFS-type flags. */
4737 /* XXX: what about SF_SETTABLE?. */
4738 XVA_SET_REQ(&xvap, XAT_IMMUTABLE);
4739 XVA_SET_REQ(&xvap, XAT_APPENDONLY);
4740 XVA_SET_REQ(&xvap, XAT_NOUNLINK);
4741 XVA_SET_REQ(&xvap, XAT_NODUMP);
4742 XVA_SET_REQ(&xvap, XAT_READONLY);
4743 XVA_SET_REQ(&xvap, XAT_ARCHIVE);
4744 XVA_SET_REQ(&xvap, XAT_SYSTEM);
4745 XVA_SET_REQ(&xvap, XAT_HIDDEN);
4746 XVA_SET_REQ(&xvap, XAT_REPARSE);
4747 XVA_SET_REQ(&xvap, XAT_OFFLINE);
4748 XVA_SET_REQ(&xvap, XAT_SPARSE);
4750 error = zfs_getattr(ap->a_vp, (vattr_t *)&xvap, 0, ap->a_cred);
4754 /* Convert ZFS xattr into chflags. */
4755 #define FLAG_CHECK(fflag, xflag, xfield) do { \
4756 if (XVA_ISSET_RTN(&xvap, (xflag)) && (xfield) != 0) \
4757 fflags |= (fflag); \
4759 FLAG_CHECK(SF_IMMUTABLE, XAT_IMMUTABLE,
4760 xvap.xva_xoptattrs.xoa_immutable);
4761 FLAG_CHECK(SF_APPEND, XAT_APPENDONLY,
4762 xvap.xva_xoptattrs.xoa_appendonly);
4763 FLAG_CHECK(SF_NOUNLINK, XAT_NOUNLINK,
4764 xvap.xva_xoptattrs.xoa_nounlink);
4765 FLAG_CHECK(UF_ARCHIVE, XAT_ARCHIVE,
4766 xvap.xva_xoptattrs.xoa_archive);
4767 FLAG_CHECK(UF_NODUMP, XAT_NODUMP,
4768 xvap.xva_xoptattrs.xoa_nodump);
4769 FLAG_CHECK(UF_READONLY, XAT_READONLY,
4770 xvap.xva_xoptattrs.xoa_readonly);
4771 FLAG_CHECK(UF_SYSTEM, XAT_SYSTEM,
4772 xvap.xva_xoptattrs.xoa_system);
4773 FLAG_CHECK(UF_HIDDEN, XAT_HIDDEN,
4774 xvap.xva_xoptattrs.xoa_hidden);
4775 FLAG_CHECK(UF_REPARSE, XAT_REPARSE,
4776 xvap.xva_xoptattrs.xoa_reparse);
4777 FLAG_CHECK(UF_OFFLINE, XAT_OFFLINE,
4778 xvap.xva_xoptattrs.xoa_offline);
4779 FLAG_CHECK(UF_SPARSE, XAT_SPARSE,
4780 xvap.xva_xoptattrs.xoa_sparse);
4783 *vap = xvap.xva_vattr;
4784 vap->va_flags = fflags;
4788 #ifndef _SYS_SYSPROTO_H_
4789 struct vop_setattr_args {
4791 struct vattr *a_vap;
4792 struct ucred *a_cred;
4797 zfs_freebsd_setattr(struct vop_setattr_args *ap)
4799 vnode_t *vp = ap->a_vp;
4800 vattr_t *vap = ap->a_vap;
4801 cred_t *cred = ap->a_cred;
4806 vattr_init_mask(vap);
4807 vap->va_mask &= ~AT_NOSET;
4810 xvap.xva_vattr = *vap;
4812 zflags = VTOZ(vp)->z_pflags;
4814 if (vap->va_flags != VNOVAL) {
4815 zfsvfs_t *zfsvfs = VTOZ(vp)->z_zfsvfs;
4818 if (zfsvfs->z_use_fuids == B_FALSE)
4819 return (EOPNOTSUPP);
4821 fflags = vap->va_flags;
4824 * We need to figure out whether it makes sense to allow
4825 * UF_REPARSE through, since we don't really have other
4826 * facilities to handle reparse points and zfs_setattr()
4827 * doesn't currently allow setting that attribute anyway.
4829 if ((fflags & ~(SF_IMMUTABLE|SF_APPEND|SF_NOUNLINK|UF_ARCHIVE|
4830 UF_NODUMP|UF_SYSTEM|UF_HIDDEN|UF_READONLY|UF_REPARSE|
4831 UF_OFFLINE|UF_SPARSE)) != 0)
4832 return (EOPNOTSUPP);
4834 * Unprivileged processes are not permitted to unset system
4835 * flags, or modify flags if any system flags are set.
4836 * Privileged non-jail processes may not modify system flags
4837 * if securelevel > 0 and any existing system flags are set.
4838 * Privileged jail processes behave like privileged non-jail
4839 * processes if the PR_ALLOW_CHFLAGS permission bit is set;
4840 * otherwise, they behave like unprivileged processes.
4842 if (secpolicy_fs_owner(vp->v_mount, cred) == 0 ||
4843 spl_priv_check_cred(cred, PRIV_VFS_SYSFLAGS) == 0) {
4845 (ZFS_IMMUTABLE | ZFS_APPENDONLY | ZFS_NOUNLINK)) {
4846 error = securelevel_gt(cred, 0);
4852 * Callers may only modify the file flags on
4853 * objects they have VADMIN rights for.
4855 if ((error = VOP_ACCESS(vp, VADMIN, cred,
4859 (ZFS_IMMUTABLE | ZFS_APPENDONLY |
4864 (SF_IMMUTABLE | SF_APPEND | SF_NOUNLINK)) {
4869 #define FLAG_CHANGE(fflag, zflag, xflag, xfield) do { \
4870 if (((fflags & (fflag)) && !(zflags & (zflag))) || \
4871 ((zflags & (zflag)) && !(fflags & (fflag)))) { \
4872 XVA_SET_REQ(&xvap, (xflag)); \
4873 (xfield) = ((fflags & (fflag)) != 0); \
4876 /* Convert chflags into ZFS-type flags. */
4877 /* XXX: what about SF_SETTABLE?. */
4878 FLAG_CHANGE(SF_IMMUTABLE, ZFS_IMMUTABLE, XAT_IMMUTABLE,
4879 xvap.xva_xoptattrs.xoa_immutable);
4880 FLAG_CHANGE(SF_APPEND, ZFS_APPENDONLY, XAT_APPENDONLY,
4881 xvap.xva_xoptattrs.xoa_appendonly);
4882 FLAG_CHANGE(SF_NOUNLINK, ZFS_NOUNLINK, XAT_NOUNLINK,
4883 xvap.xva_xoptattrs.xoa_nounlink);
4884 FLAG_CHANGE(UF_ARCHIVE, ZFS_ARCHIVE, XAT_ARCHIVE,
4885 xvap.xva_xoptattrs.xoa_archive);
4886 FLAG_CHANGE(UF_NODUMP, ZFS_NODUMP, XAT_NODUMP,
4887 xvap.xva_xoptattrs.xoa_nodump);
4888 FLAG_CHANGE(UF_READONLY, ZFS_READONLY, XAT_READONLY,
4889 xvap.xva_xoptattrs.xoa_readonly);
4890 FLAG_CHANGE(UF_SYSTEM, ZFS_SYSTEM, XAT_SYSTEM,
4891 xvap.xva_xoptattrs.xoa_system);
4892 FLAG_CHANGE(UF_HIDDEN, ZFS_HIDDEN, XAT_HIDDEN,
4893 xvap.xva_xoptattrs.xoa_hidden);
4894 FLAG_CHANGE(UF_REPARSE, ZFS_REPARSE, XAT_REPARSE,
4895 xvap.xva_xoptattrs.xoa_reparse);
4896 FLAG_CHANGE(UF_OFFLINE, ZFS_OFFLINE, XAT_OFFLINE,
4897 xvap.xva_xoptattrs.xoa_offline);
4898 FLAG_CHANGE(UF_SPARSE, ZFS_SPARSE, XAT_SPARSE,
4899 xvap.xva_xoptattrs.xoa_sparse);
4902 if (vap->va_birthtime.tv_sec != VNOVAL) {
4903 xvap.xva_vattr.va_mask |= AT_XVATTR;
4904 XVA_SET_REQ(&xvap, XAT_CREATETIME);
4906 return (zfs_setattr(VTOZ(vp), (vattr_t *)&xvap, 0, cred));
4909 #ifndef _SYS_SYSPROTO_H_
4910 struct vop_rename_args {
4911 struct vnode *a_fdvp;
4912 struct vnode *a_fvp;
4913 struct componentname *a_fcnp;
4914 struct vnode *a_tdvp;
4915 struct vnode *a_tvp;
4916 struct componentname *a_tcnp;
4921 zfs_freebsd_rename(struct vop_rename_args *ap)
4923 vnode_t *fdvp = ap->a_fdvp;
4924 vnode_t *fvp = ap->a_fvp;
4925 vnode_t *tdvp = ap->a_tdvp;
4926 vnode_t *tvp = ap->a_tvp;
4929 ASSERT(ap->a_fcnp->cn_flags & (SAVENAME|SAVESTART));
4930 ASSERT(ap->a_tcnp->cn_flags & (SAVENAME|SAVESTART));
4932 error = zfs_rename_(fdvp, &fvp, ap->a_fcnp, tdvp, &tvp,
4933 ap->a_tcnp, ap->a_fcnp->cn_cred, 1);
4944 #ifndef _SYS_SYSPROTO_H_
4945 struct vop_symlink_args {
4946 struct vnode *a_dvp;
4947 struct vnode **a_vpp;
4948 struct componentname *a_cnp;
4949 struct vattr *a_vap;
4955 zfs_freebsd_symlink(struct vop_symlink_args *ap)
4957 struct componentname *cnp = ap->a_cnp;
4958 vattr_t *vap = ap->a_vap;
4962 ASSERT(cnp->cn_flags & SAVENAME);
4964 vap->va_type = VLNK; /* FreeBSD: Syscall only sets va_mode. */
4965 vattr_init_mask(vap);
4968 rc = zfs_symlink(VTOZ(ap->a_dvp), cnp->cn_nameptr, vap,
4969 ap->a_target, &zp, cnp->cn_cred, 0 /* flags */);
4971 *ap->a_vpp = ZTOV(zp);
4975 #ifndef _SYS_SYSPROTO_H_
4976 struct vop_readlink_args {
4979 struct ucred *a_cred;
4984 zfs_freebsd_readlink(struct vop_readlink_args *ap)
4987 zfs_uio_init(&uio, ap->a_uio);
4988 return (zfs_readlink(ap->a_vp, &uio, ap->a_cred, NULL));
4991 #ifndef _SYS_SYSPROTO_H_
4992 struct vop_link_args {
4993 struct vnode *a_tdvp;
4995 struct componentname *a_cnp;
5000 zfs_freebsd_link(struct vop_link_args *ap)
5002 struct componentname *cnp = ap->a_cnp;
5003 vnode_t *vp = ap->a_vp;
5004 vnode_t *tdvp = ap->a_tdvp;
5006 if (tdvp->v_mount != vp->v_mount)
5009 ASSERT(cnp->cn_flags & SAVENAME);
5011 return (zfs_link(VTOZ(tdvp), VTOZ(vp),
5012 cnp->cn_nameptr, cnp->cn_cred, 0));
5015 #ifndef _SYS_SYSPROTO_H_
5016 struct vop_inactive_args {
5018 struct thread *a_td;
5023 zfs_freebsd_inactive(struct vop_inactive_args *ap)
5025 vnode_t *vp = ap->a_vp;
5027 #if __FreeBSD_version >= 1300123
5028 zfs_inactive(vp, curthread->td_ucred, NULL);
5030 zfs_inactive(vp, ap->a_td->td_ucred, NULL);
5035 #if __FreeBSD_version >= 1300042
5036 #ifndef _SYS_SYSPROTO_H_
5037 struct vop_need_inactive_args {
5039 struct thread *a_td;
5044 zfs_freebsd_need_inactive(struct vop_need_inactive_args *ap)
5046 vnode_t *vp = ap->a_vp;
5047 znode_t *zp = VTOZ(vp);
5048 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
5051 if (vn_need_pageq_flush(vp))
5054 if (!ZFS_TRYRLOCK_TEARDOWN_INACTIVE(zfsvfs))
5056 need = (zp->z_sa_hdl == NULL || zp->z_unlinked || zp->z_atime_dirty);
5057 ZFS_RUNLOCK_TEARDOWN_INACTIVE(zfsvfs);
5063 #ifndef _SYS_SYSPROTO_H_
5064 struct vop_reclaim_args {
5066 struct thread *a_td;
5071 zfs_freebsd_reclaim(struct vop_reclaim_args *ap)
5073 vnode_t *vp = ap->a_vp;
5074 znode_t *zp = VTOZ(vp);
5075 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
5079 #if __FreeBSD_version < 1300042
5080 /* Destroy the vm object and flush associated pages. */
5081 vnode_destroy_vobject(vp);
5084 * z_teardown_inactive_lock protects from a race with
5085 * zfs_znode_dmu_fini in zfsvfs_teardown during
5088 ZFS_RLOCK_TEARDOWN_INACTIVE(zfsvfs);
5089 if (zp->z_sa_hdl == NULL)
5093 ZFS_RUNLOCK_TEARDOWN_INACTIVE(zfsvfs);
5099 #ifndef _SYS_SYSPROTO_H_
5100 struct vop_fid_args {
5107 zfs_freebsd_fid(struct vop_fid_args *ap)
5110 return (zfs_fid(ap->a_vp, (void *)ap->a_fid, NULL));
5114 #ifndef _SYS_SYSPROTO_H_
5115 struct vop_pathconf_args {
5118 register_t *a_retval;
5123 zfs_freebsd_pathconf(struct vop_pathconf_args *ap)
5128 error = zfs_pathconf(ap->a_vp, ap->a_name, &val,
5129 curthread->td_ucred, NULL);
5131 *ap->a_retval = val;
5134 if (error != EOPNOTSUPP)
5137 switch (ap->a_name) {
5139 *ap->a_retval = NAME_MAX;
5142 if (ap->a_vp->v_type == VDIR || ap->a_vp->v_type == VFIFO) {
5143 *ap->a_retval = PIPE_BUF;
5148 return (vop_stdpathconf(ap));
5153 * FreeBSD's extended attributes namespace defines file name prefix for ZFS'
5154 * extended attribute name:
5157 * system freebsd:system:
5158 * user (none, can be used to access ZFS fsattr(5) attributes
5159 * created on Solaris)
5162 zfs_create_attrname(int attrnamespace, const char *name, char *attrname,
5165 const char *namespace, *prefix, *suffix;
5167 /* We don't allow '/' character in attribute name. */
5168 if (strchr(name, '/') != NULL)
5170 /* We don't allow attribute names that start with "freebsd:" string. */
5171 if (strncmp(name, "freebsd:", 8) == 0)
5174 bzero(attrname, size);
5176 switch (attrnamespace) {
5177 case EXTATTR_NAMESPACE_USER:
5179 prefix = "freebsd:";
5180 namespace = EXTATTR_NAMESPACE_USER_STRING;
5184 * This is the default namespace by which we can access all
5185 * attributes created on Solaris.
5187 prefix = namespace = suffix = "";
5190 case EXTATTR_NAMESPACE_SYSTEM:
5191 prefix = "freebsd:";
5192 namespace = EXTATTR_NAMESPACE_SYSTEM_STRING;
5195 case EXTATTR_NAMESPACE_EMPTY:
5199 if (snprintf(attrname, size, "%s%s%s%s", prefix, namespace, suffix,
5201 return (ENAMETOOLONG);
5206 #ifndef _SYS_SYSPROTO_H_
5207 struct vop_getextattr {
5208 IN struct vnode *a_vp;
5209 IN int a_attrnamespace;
5210 IN const char *a_name;
5211 INOUT struct uio *a_uio;
5213 IN struct ucred *a_cred;
5214 IN struct thread *a_td;
5219 * Vnode operating to retrieve a named extended attribute.
5222 zfs_getextattr(struct vop_getextattr_args *ap)
5224 zfsvfs_t *zfsvfs = VTOZ(ap->a_vp)->z_zfsvfs;
5225 struct thread *td = ap->a_td;
5226 struct nameidata nd;
5229 vnode_t *xvp = NULL, *vp;
5233 * If the xattr property is off, refuse the request.
5235 if (!(zfsvfs->z_flags & ZSB_XATTR)) {
5236 return (SET_ERROR(EOPNOTSUPP));
5239 error = extattr_check_cred(ap->a_vp, ap->a_attrnamespace,
5240 ap->a_cred, ap->a_td, VREAD);
5244 error = zfs_create_attrname(ap->a_attrnamespace, ap->a_name, attrname,
5251 error = zfs_lookup(ap->a_vp, NULL, &xvp, NULL, 0, ap->a_cred, td,
5252 LOOKUP_XATTR, B_FALSE);
5259 NDINIT_ATVP(&nd, LOOKUP, NOFOLLOW, UIO_SYSSPACE, attrname,
5261 error = vn_open_cred(&nd, &flags, 0, VN_OPEN_INVFS, ap->a_cred, NULL);
5263 NDFREE(&nd, NDF_ONLY_PNBUF);
5266 if (error == ENOENT)
5271 if (ap->a_size != NULL) {
5272 error = VOP_GETATTR(vp, &va, ap->a_cred);
5274 *ap->a_size = (size_t)va.va_size;
5275 } else if (ap->a_uio != NULL)
5276 error = VOP_READ(vp, ap->a_uio, IO_UNIT, ap->a_cred);
5279 vn_close(vp, flags, ap->a_cred, td);
5284 #ifndef _SYS_SYSPROTO_H_
5285 struct vop_deleteextattr {
5286 IN struct vnode *a_vp;
5287 IN int a_attrnamespace;
5288 IN const char *a_name;
5289 IN struct ucred *a_cred;
5290 IN struct thread *a_td;
5295 * Vnode operation to remove a named attribute.
5298 zfs_deleteextattr(struct vop_deleteextattr_args *ap)
5300 zfsvfs_t *zfsvfs = VTOZ(ap->a_vp)->z_zfsvfs;
5301 struct thread *td = ap->a_td;
5302 struct nameidata nd;
5304 vnode_t *xvp = NULL, *vp;
5308 * If the xattr property is off, refuse the request.
5310 if (!(zfsvfs->z_flags & ZSB_XATTR)) {
5311 return (SET_ERROR(EOPNOTSUPP));
5314 error = extattr_check_cred(ap->a_vp, ap->a_attrnamespace,
5315 ap->a_cred, ap->a_td, VWRITE);
5319 error = zfs_create_attrname(ap->a_attrnamespace, ap->a_name, attrname,
5326 error = zfs_lookup(ap->a_vp, NULL, &xvp, NULL, 0, ap->a_cred, td,
5327 LOOKUP_XATTR, B_FALSE);
5333 NDINIT_ATVP(&nd, DELETE, NOFOLLOW | LOCKPARENT | LOCKLEAF,
5334 UIO_SYSSPACE, attrname, xvp, td);
5339 NDFREE(&nd, NDF_ONLY_PNBUF);
5340 if (error == ENOENT)
5345 error = VOP_REMOVE(nd.ni_dvp, vp, &nd.ni_cnd);
5346 NDFREE(&nd, NDF_ONLY_PNBUF);
5349 if (vp == nd.ni_dvp)
5358 #ifndef _SYS_SYSPROTO_H_
5359 struct vop_setextattr {
5360 IN struct vnode *a_vp;
5361 IN int a_attrnamespace;
5362 IN const char *a_name;
5363 INOUT struct uio *a_uio;
5364 IN struct ucred *a_cred;
5365 IN struct thread *a_td;
5370 * Vnode operation to set a named attribute.
5373 zfs_setextattr(struct vop_setextattr_args *ap)
5375 zfsvfs_t *zfsvfs = VTOZ(ap->a_vp)->z_zfsvfs;
5376 struct thread *td = ap->a_td;
5377 struct nameidata nd;
5380 vnode_t *xvp = NULL, *vp;
5384 * If the xattr property is off, refuse the request.
5386 if (!(zfsvfs->z_flags & ZSB_XATTR)) {
5387 return (SET_ERROR(EOPNOTSUPP));
5390 error = extattr_check_cred(ap->a_vp, ap->a_attrnamespace,
5391 ap->a_cred, ap->a_td, VWRITE);
5394 error = zfs_create_attrname(ap->a_attrnamespace, ap->a_name, attrname,
5401 error = zfs_lookup(ap->a_vp, NULL, &xvp, NULL, 0, ap->a_cred, td,
5402 LOOKUP_XATTR | CREATE_XATTR_DIR, B_FALSE);
5408 flags = FFLAGS(O_WRONLY | O_CREAT);
5409 NDINIT_ATVP(&nd, LOOKUP, NOFOLLOW, UIO_SYSSPACE, attrname,
5411 error = vn_open_cred(&nd, &flags, 0600, VN_OPEN_INVFS, ap->a_cred,
5414 NDFREE(&nd, NDF_ONLY_PNBUF);
5422 error = VOP_SETATTR(vp, &va, ap->a_cred);
5424 VOP_WRITE(vp, ap->a_uio, IO_UNIT, ap->a_cred);
5427 vn_close(vp, flags, ap->a_cred, td);
5432 #ifndef _SYS_SYSPROTO_H_
5433 struct vop_listextattr {
5434 IN struct vnode *a_vp;
5435 IN int a_attrnamespace;
5436 INOUT struct uio *a_uio;
5438 IN struct ucred *a_cred;
5439 IN struct thread *a_td;
5444 * Vnode operation to retrieve extended attributes on a vnode.
5447 zfs_listextattr(struct vop_listextattr_args *ap)
5449 zfsvfs_t *zfsvfs = VTOZ(ap->a_vp)->z_zfsvfs;
5450 struct thread *td = ap->a_td;
5451 struct nameidata nd;
5452 char attrprefix[16];
5453 uint8_t dirbuf[sizeof (struct dirent)];
5457 size_t *sizep = ap->a_size;
5459 vnode_t *xvp = NULL, *vp;
5460 int done, error, eof, pos;
5463 zfs_uio_init(&uio, ap->a_uio);
5466 * If the xattr property is off, refuse the request.
5468 if (!(zfsvfs->z_flags & ZSB_XATTR)) {
5469 return (SET_ERROR(EOPNOTSUPP));
5472 error = extattr_check_cred(ap->a_vp, ap->a_attrnamespace,
5473 ap->a_cred, ap->a_td, VREAD);
5477 error = zfs_create_attrname(ap->a_attrnamespace, "", attrprefix,
5478 sizeof (attrprefix));
5481 plen = strlen(attrprefix);
5488 error = zfs_lookup(ap->a_vp, NULL, &xvp, NULL, 0, ap->a_cred, td,
5489 LOOKUP_XATTR, B_FALSE);
5493 * ENOATTR means that the EA directory does not yet exist,
5494 * i.e. there are no extended attributes there.
5496 if (error == ENOATTR)
5501 NDINIT_ATVP(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | LOCKSHARED,
5502 UIO_SYSSPACE, ".", xvp, td);
5505 NDFREE(&nd, NDF_ONLY_PNBUF);
5511 auio.uio_iov = &aiov;
5512 auio.uio_iovcnt = 1;
5513 auio.uio_segflg = UIO_SYSSPACE;
5515 auio.uio_rw = UIO_READ;
5516 auio.uio_offset = 0;
5521 aiov.iov_base = (void *)dirbuf;
5522 aiov.iov_len = sizeof (dirbuf);
5523 auio.uio_resid = sizeof (dirbuf);
5524 error = VOP_READDIR(vp, &auio, ap->a_cred, &eof, NULL, NULL);
5525 done = sizeof (dirbuf) - auio.uio_resid;
5528 for (pos = 0; pos < done; ) {
5529 dp = (struct dirent *)(dirbuf + pos);
5530 pos += dp->d_reclen;
5532 * XXX: Temporarily we also accept DT_UNKNOWN, as this
5533 * is what we get when attribute was created on Solaris.
5535 if (dp->d_type != DT_REG && dp->d_type != DT_UNKNOWN)
5538 strncmp(dp->d_name, "freebsd:", 8) == 0)
5540 else if (strncmp(dp->d_name, attrprefix, plen) != 0)
5542 nlen = dp->d_namlen - plen;
5545 else if (GET_UIO_STRUCT(&uio) != NULL) {
5547 * Format of extattr name entry is one byte for
5548 * length and the rest for name.
5550 error = zfs_uiomove(&nlen, 1, zfs_uio_rw(&uio),
5553 error = zfs_uiomove(dp->d_name + plen,
5554 nlen, zfs_uio_rw(&uio), &uio);
5560 } while (!eof && error == 0);
5568 #ifndef _SYS_SYSPROTO_H_
5569 struct vop_getacl_args {
5579 zfs_freebsd_getacl(struct vop_getacl_args *ap)
5582 vsecattr_t vsecattr;
5584 if (ap->a_type != ACL_TYPE_NFS4)
5587 vsecattr.vsa_mask = VSA_ACE | VSA_ACECNT;
5588 if ((error = zfs_getsecattr(VTOZ(ap->a_vp),
5589 &vsecattr, 0, ap->a_cred)))
5592 error = acl_from_aces(ap->a_aclp, vsecattr.vsa_aclentp,
5593 vsecattr.vsa_aclcnt);
5594 if (vsecattr.vsa_aclentp != NULL)
5595 kmem_free(vsecattr.vsa_aclentp, vsecattr.vsa_aclentsz);
5600 #ifndef _SYS_SYSPROTO_H_
5601 struct vop_setacl_args {
5611 zfs_freebsd_setacl(struct vop_setacl_args *ap)
5614 vsecattr_t vsecattr;
5615 int aclbsize; /* size of acl list in bytes */
5618 if (ap->a_type != ACL_TYPE_NFS4)
5621 if (ap->a_aclp == NULL)
5624 if (ap->a_aclp->acl_cnt < 1 || ap->a_aclp->acl_cnt > MAX_ACL_ENTRIES)
5628 * With NFSv4 ACLs, chmod(2) may need to add additional entries,
5629 * splitting every entry into two and appending "canonical six"
5630 * entries at the end. Don't allow for setting an ACL that would
5631 * cause chmod(2) to run out of ACL entries.
5633 if (ap->a_aclp->acl_cnt * 2 + 6 > ACL_MAX_ENTRIES)
5636 error = acl_nfs4_check(ap->a_aclp, ap->a_vp->v_type == VDIR);
5640 vsecattr.vsa_mask = VSA_ACE;
5641 aclbsize = ap->a_aclp->acl_cnt * sizeof (ace_t);
5642 vsecattr.vsa_aclentp = kmem_alloc(aclbsize, KM_SLEEP);
5643 aaclp = vsecattr.vsa_aclentp;
5644 vsecattr.vsa_aclentsz = aclbsize;
5646 aces_from_acl(vsecattr.vsa_aclentp, &vsecattr.vsa_aclcnt, ap->a_aclp);
5647 error = zfs_setsecattr(VTOZ(ap->a_vp), &vsecattr, 0, ap->a_cred);
5648 kmem_free(aaclp, aclbsize);
5653 #ifndef _SYS_SYSPROTO_H_
5654 struct vop_aclcheck_args {
5664 zfs_freebsd_aclcheck(struct vop_aclcheck_args *ap)
5667 return (EOPNOTSUPP);
5671 zfs_vptocnp(struct vop_vptocnp_args *ap)
5673 vnode_t *covered_vp;
5674 vnode_t *vp = ap->a_vp;
5675 zfsvfs_t *zfsvfs = vp->v_vfsp->vfs_data;
5676 znode_t *zp = VTOZ(vp);
5684 * If we are a snapshot mounted under .zfs, run the operation
5685 * on the covered vnode.
5687 if (zp->z_id != zfsvfs->z_root || zfsvfs->z_parent == zfsvfs) {
5688 char name[MAXNAMLEN + 1];
5692 error = zfs_znode_parent_and_name(zp, &dzp, name);
5695 if (*ap->a_buflen < len)
5696 error = SET_ERROR(ENOMEM);
5699 *ap->a_buflen -= len;
5700 bcopy(name, ap->a_buf + *ap->a_buflen, len);
5701 *ap->a_vpp = ZTOV(dzp);
5708 covered_vp = vp->v_mount->mnt_vnodecovered;
5709 #if __FreeBSD_version >= 1300045
5710 enum vgetstate vs = vget_prep(covered_vp);
5714 ltype = VOP_ISLOCKED(vp);
5716 #if __FreeBSD_version >= 1300045
5717 error = vget_finish(covered_vp, LK_SHARED, vs);
5719 error = vget(covered_vp, LK_SHARED | LK_VNHELD, curthread);
5722 #if __FreeBSD_version >= 1300123
5723 error = VOP_VPTOCNP(covered_vp, ap->a_vpp, ap->a_buf,
5726 error = VOP_VPTOCNP(covered_vp, ap->a_vpp, ap->a_cred,
5727 ap->a_buf, ap->a_buflen);
5731 vn_lock(vp, ltype | LK_RETRY);
5732 if (VN_IS_DOOMED(vp))
5733 error = SET_ERROR(ENOENT);
5738 #ifndef _SYS_SYSPROTO_H_
5739 struct vop_lock1_args {
5748 zfs_lock(struct vop_lock1_args *ap)
5754 #if __FreeBSD_version >= 1300064
5757 err = vop_stdlock(ap);
5759 if (err == 0 && (ap->a_flags & LK_NOWAIT) == 0) {
5762 if (vp->v_mount != NULL && !VN_IS_DOOMED(vp) &&
5763 zp != NULL && (zp->z_pflags & ZFS_XATTR) == 0)
5764 VERIFY(!RRM_LOCK_HELD(&zp->z_zfsvfs->z_teardown_lock));
5770 struct vop_vector zfs_vnodeops;
5771 struct vop_vector zfs_fifoops;
5772 struct vop_vector zfs_shareops;
5774 struct vop_vector zfs_vnodeops = {
5775 .vop_default = &default_vnodeops,
5776 .vop_inactive = zfs_freebsd_inactive,
5777 #if __FreeBSD_version >= 1300042
5778 .vop_need_inactive = zfs_freebsd_need_inactive,
5780 .vop_reclaim = zfs_freebsd_reclaim,
5781 #if __FreeBSD_version >= 1300102
5782 .vop_fplookup_vexec = zfs_freebsd_fplookup_vexec,
5784 .vop_access = zfs_freebsd_access,
5785 .vop_allocate = VOP_EINVAL,
5786 .vop_lookup = zfs_cache_lookup,
5787 .vop_cachedlookup = zfs_freebsd_cachedlookup,
5788 .vop_getattr = zfs_freebsd_getattr,
5789 .vop_setattr = zfs_freebsd_setattr,
5790 .vop_create = zfs_freebsd_create,
5791 .vop_mknod = (vop_mknod_t *)zfs_freebsd_create,
5792 .vop_mkdir = zfs_freebsd_mkdir,
5793 .vop_readdir = zfs_freebsd_readdir,
5794 .vop_fsync = zfs_freebsd_fsync,
5795 .vop_open = zfs_freebsd_open,
5796 .vop_close = zfs_freebsd_close,
5797 .vop_rmdir = zfs_freebsd_rmdir,
5798 .vop_ioctl = zfs_freebsd_ioctl,
5799 .vop_link = zfs_freebsd_link,
5800 .vop_symlink = zfs_freebsd_symlink,
5801 .vop_readlink = zfs_freebsd_readlink,
5802 .vop_read = zfs_freebsd_read,
5803 .vop_write = zfs_freebsd_write,
5804 .vop_remove = zfs_freebsd_remove,
5805 .vop_rename = zfs_freebsd_rename,
5806 .vop_pathconf = zfs_freebsd_pathconf,
5807 .vop_bmap = zfs_freebsd_bmap,
5808 .vop_fid = zfs_freebsd_fid,
5809 .vop_getextattr = zfs_getextattr,
5810 .vop_deleteextattr = zfs_deleteextattr,
5811 .vop_setextattr = zfs_setextattr,
5812 .vop_listextattr = zfs_listextattr,
5813 .vop_getacl = zfs_freebsd_getacl,
5814 .vop_setacl = zfs_freebsd_setacl,
5815 .vop_aclcheck = zfs_freebsd_aclcheck,
5816 .vop_getpages = zfs_freebsd_getpages,
5817 .vop_putpages = zfs_freebsd_putpages,
5818 .vop_vptocnp = zfs_vptocnp,
5819 #if __FreeBSD_version >= 1300064
5821 .vop_lock1 = zfs_lock,
5823 .vop_lock1 = vop_lock,
5825 .vop_unlock = vop_unlock,
5826 .vop_islocked = vop_islocked,
5829 .vop_lock1 = zfs_lock,
5833 VFS_VOP_VECTOR_REGISTER(zfs_vnodeops);
5835 struct vop_vector zfs_fifoops = {
5836 .vop_default = &fifo_specops,
5837 .vop_fsync = zfs_freebsd_fsync,
5838 #if __FreeBSD_version >= 1300102
5839 .vop_fplookup_vexec = zfs_freebsd_fplookup_vexec,
5841 .vop_access = zfs_freebsd_access,
5842 .vop_getattr = zfs_freebsd_getattr,
5843 .vop_inactive = zfs_freebsd_inactive,
5844 .vop_read = VOP_PANIC,
5845 .vop_reclaim = zfs_freebsd_reclaim,
5846 .vop_setattr = zfs_freebsd_setattr,
5847 .vop_write = VOP_PANIC,
5848 .vop_pathconf = zfs_freebsd_pathconf,
5849 .vop_fid = zfs_freebsd_fid,
5850 .vop_getacl = zfs_freebsd_getacl,
5851 .vop_setacl = zfs_freebsd_setacl,
5852 .vop_aclcheck = zfs_freebsd_aclcheck,
5854 VFS_VOP_VECTOR_REGISTER(zfs_fifoops);
5857 * special share hidden files vnode operations template
5859 struct vop_vector zfs_shareops = {
5860 .vop_default = &default_vnodeops,
5861 #if __FreeBSD_version >= 1300121
5862 .vop_fplookup_vexec = VOP_EAGAIN,
5864 .vop_access = zfs_freebsd_access,
5865 .vop_inactive = zfs_freebsd_inactive,
5866 .vop_reclaim = zfs_freebsd_reclaim,
5867 .vop_fid = zfs_freebsd_fid,
5868 .vop_pathconf = zfs_freebsd_pathconf,
5870 VFS_VOP_VECTOR_REGISTER(zfs_shareops);