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, uio_t *uio)
523 vnode_t *vp = ZTOV(zp);
524 objset_t *os = zp->z_zfsvfs->z_os;
533 ASSERT(uio->uio_segflg == UIO_NOCOPY);
534 ASSERT(vp->v_mount != NULL);
537 ASSERT((uio->uio_loffset & PAGEOFFSET) == 0);
539 zfs_vmobject_wlock_12(obj);
540 for (start = uio->uio_loffset; 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 uio->uio_resid -= bytes;
588 uio->uio_offset += bytes;
591 zfs_vmobject_wunlock_12(obj);
596 * When a file is memory mapped, we must keep the IO data synchronized
597 * between the DMU cache and the memory mapped pages. What this means:
599 * On Read: We "read" preferentially from memory mapped pages,
600 * else we default from the dmu buffer.
602 * NOTE: We will always "break up" the IO into PAGESIZE uiomoves when
603 * the file is memory mapped.
606 mappedread(znode_t *zp, int nbytes, uio_t *uio)
608 vnode_t *vp = ZTOV(zp);
615 ASSERT(vp->v_mount != NULL);
619 start = uio->uio_loffset;
620 off = start & PAGEOFFSET;
621 zfs_vmobject_wlock_12(obj);
622 for (start &= PAGEMASK; len > 0; start += PAGESIZE) {
624 uint64_t bytes = MIN(PAGESIZE - off, len);
626 if ((pp = page_hold(vp, start))) {
630 zfs_vmobject_wunlock_12(obj);
631 va = zfs_map_page(pp, &sf);
632 error = vn_io_fault_uiomove(va + off, bytes, 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 (name[0] == 0 || (name[0] == '.' && name[1] == 0)) {
706 ASSERT3P(dvp, ==, vp);
708 ltype = lkflags & LK_TYPE_MASK;
709 if (ltype != VOP_ISLOCKED(dvp)) {
710 if (ltype == LK_EXCLUSIVE)
711 vn_lock(dvp, LK_UPGRADE | LK_RETRY);
712 else /* if (ltype == LK_SHARED) */
713 vn_lock(dvp, LK_DOWNGRADE | LK_RETRY);
716 * Relock for the "." case could leave us with
719 if (VN_IS_DOOMED(dvp)) {
721 return (SET_ERROR(ENOENT));
725 } else if (name[0] == '.' && name[1] == '.' && name[2] == 0) {
727 * Note that in this case, dvp is the child vnode, and we
728 * are looking up the parent vnode - exactly reverse from
729 * normal operation. Unlocking dvp requires some rather
730 * tricky unlock/relock dance to prevent mp from being freed;
731 * use vn_vget_ino_gen() which takes care of all that.
733 * XXX Note that there is a time window when both vnodes are
734 * unlocked. It is possible, although highly unlikely, that
735 * during that window the parent-child relationship between
736 * the vnodes may change, for example, get reversed.
737 * In that case we would have a wrong lock order for the vnodes.
738 * All other filesystems seem to ignore this problem, so we
740 * A potential solution could be implemented as follows:
741 * - using LK_NOWAIT when locking the second vnode and retrying
743 * - checking that the parent-child relationship still holds
744 * after locking both vnodes and retrying if it doesn't
746 error = vn_vget_ino_gen(dvp, zfs_dd_callback, vp, lkflags, &vp);
749 error = vn_lock(vp, lkflags);
757 * Lookup an entry in a directory, or an extended attribute directory.
758 * If it exists, return a held vnode reference for it.
760 * IN: dvp - vnode of directory to search.
761 * nm - name of entry to lookup.
762 * pnp - full pathname to lookup [UNUSED].
763 * flags - LOOKUP_XATTR set if looking for an attribute.
764 * rdir - root directory vnode [UNUSED].
765 * cr - credentials of caller.
766 * ct - caller context
768 * OUT: vpp - vnode of located entry, NULL if not found.
770 * RETURN: 0 on success, error code on failure.
777 zfs_lookup(vnode_t *dvp, const char *nm, vnode_t **vpp,
778 struct componentname *cnp, int nameiop, cred_t *cr, kthread_t *td,
779 int flags, boolean_t cached)
781 znode_t *zdp = VTOZ(dvp);
783 zfsvfs_t *zfsvfs = zdp->z_zfsvfs;
787 * Fast path lookup, however we must skip DNLC lookup
788 * for case folding or normalizing lookups because the
789 * DNLC code only stores the passed in name. This means
790 * creating 'a' and removing 'A' on a case insensitive
791 * file system would work, but DNLC still thinks 'a'
792 * exists and won't let you create it again on the next
793 * pass through fast path.
795 if (!(flags & LOOKUP_XATTR)) {
796 if (dvp->v_type != VDIR) {
797 return (SET_ERROR(ENOTDIR));
798 } else if (zdp->z_sa_hdl == NULL) {
799 return (SET_ERROR(EIO));
803 DTRACE_PROBE2(zfs__fastpath__lookup__miss, vnode_t *, dvp,
811 if (flags & LOOKUP_XATTR) {
813 * If the xattr property is off, refuse the lookup request.
815 if (!(zfsvfs->z_flags & ZSB_XATTR)) {
817 return (SET_ERROR(EOPNOTSUPP));
821 * We don't allow recursive attributes..
822 * Maybe someday we will.
824 if (zdp->z_pflags & ZFS_XATTR) {
826 return (SET_ERROR(EINVAL));
829 if ((error = zfs_get_xattrdir(VTOZ(dvp), &zp, cr, flags))) {
836 * Do we have permission to get into attribute directory?
838 error = zfs_zaccess(zp, ACE_EXECUTE, 0, B_FALSE, cr);
848 * Check accessibility of directory if we're not coming in via
853 if ((cnp->cn_flags & NOEXECCHECK) != 0) {
854 cnp->cn_flags &= ~NOEXECCHECK;
857 if ((error = zfs_zaccess(zdp, ACE_EXECUTE, 0, B_FALSE, cr))) {
863 if (zfsvfs->z_utf8 && u8_validate(nm, strlen(nm),
864 NULL, U8_VALIDATE_ENTIRE, &error) < 0) {
866 return (SET_ERROR(EILSEQ));
871 * First handle the special cases.
873 if ((cnp->cn_flags & ISDOTDOT) != 0) {
875 * If we are a snapshot mounted under .zfs, return
876 * the vp for the snapshot directory.
878 if (zdp->z_id == zfsvfs->z_root && zfsvfs->z_parent != zfsvfs) {
879 struct componentname cn;
884 ltype = VOP_ISLOCKED(dvp);
886 error = zfsctl_root(zfsvfs->z_parent, LK_SHARED,
889 cn.cn_nameptr = "snapshot";
890 cn.cn_namelen = strlen(cn.cn_nameptr);
891 cn.cn_nameiop = cnp->cn_nameiop;
892 cn.cn_flags = cnp->cn_flags & ~ISDOTDOT;
893 cn.cn_lkflags = cnp->cn_lkflags;
894 error = VOP_LOOKUP(zfsctl_vp, vpp, &cn);
897 vn_lock(dvp, ltype | LK_RETRY);
901 if (zfs_has_ctldir(zdp) && strcmp(nm, ZFS_CTLDIR_NAME) == 0) {
903 if ((cnp->cn_flags & ISLASTCN) != 0 && nameiop != LOOKUP)
904 return (SET_ERROR(ENOTSUP));
905 error = zfsctl_root(zfsvfs, cnp->cn_lkflags, vpp);
910 * The loop is retry the lookup if the parent-child relationship
911 * changes during the dot-dot locking complexities.
916 error = zfs_dirlook(zdp, nm, &zp);
924 error = zfs_lookup_lock(dvp, *vpp, nm, cnp->cn_lkflags);
927 * If we've got a locking error, then the vnode
928 * got reclaimed because of a force unmount.
929 * We never enter doomed vnodes into the name cache.
935 if ((cnp->cn_flags & ISDOTDOT) == 0)
939 if (zdp->z_sa_hdl == NULL) {
940 error = SET_ERROR(EIO);
942 error = sa_lookup(zdp->z_sa_hdl, SA_ZPL_PARENT(zfsvfs),
943 &parent, sizeof (parent));
950 if (zp->z_id == parent) {
960 /* Translate errors and add SAVENAME when needed. */
961 if (cnp->cn_flags & ISLASTCN) {
965 if (error == ENOENT) {
967 cnp->cn_flags |= SAVENAME;
973 cnp->cn_flags |= SAVENAME;
978 /* Insert name into cache (as non-existent) if appropriate. */
979 if (zfsvfs->z_use_namecache && !zfsvfs->z_replay &&
980 error == ENOENT && (cnp->cn_flags & MAKEENTRY) != 0)
981 cache_enter(dvp, NULL, cnp);
983 /* Insert name into cache if appropriate. */
984 if (zfsvfs->z_use_namecache && !zfsvfs->z_replay &&
985 error == 0 && (cnp->cn_flags & MAKEENTRY)) {
986 if (!(cnp->cn_flags & ISLASTCN) ||
987 (nameiop != DELETE && nameiop != RENAME)) {
988 cache_enter(dvp, *vpp, cnp);
996 * Attempt to create a new entry in a directory. If the entry
997 * already exists, truncate the file if permissible, else return
998 * an error. Return the vp of the created or trunc'd file.
1000 * IN: dvp - vnode of directory to put new file entry in.
1001 * name - name of new file entry.
1002 * vap - attributes of new file.
1003 * excl - flag indicating exclusive or non-exclusive mode.
1004 * mode - mode to open file with.
1005 * cr - credentials of caller.
1006 * flag - large file flag [UNUSED].
1007 * ct - caller context
1008 * vsecp - ACL to be set
1010 * OUT: vpp - vnode of created or trunc'd entry.
1012 * RETURN: 0 on success, error code on failure.
1015 * dvp - ctime|mtime updated if new entry created
1016 * vp - ctime|mtime always, atime if new
1021 zfs_create(znode_t *dzp, const char *name, vattr_t *vap, int excl, int mode,
1022 znode_t **zpp, cred_t *cr, int flag, vsecattr_t *vsecp)
1025 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
1032 gid_t gid = crgetgid(cr);
1033 uint64_t projid = ZFS_DEFAULT_PROJID;
1034 zfs_acl_ids_t acl_ids;
1035 boolean_t fuid_dirtied;
1037 #ifdef DEBUG_VFS_LOCKS
1038 vnode_t *dvp = ZTOV(dzp);
1042 * If we have an ephemeral id, ACL, or XVATTR then
1043 * make sure file system is at proper version
1046 ksid = crgetsid(cr, KSID_OWNER);
1048 uid = ksid_getid(ksid);
1052 if (zfsvfs->z_use_fuids == B_FALSE &&
1053 (vsecp || (vap->va_mask & AT_XVATTR) ||
1054 IS_EPHEMERAL(uid) || IS_EPHEMERAL(gid)))
1055 return (SET_ERROR(EINVAL));
1060 zilog = zfsvfs->z_log;
1062 if (zfsvfs->z_utf8 && u8_validate(name, strlen(name),
1063 NULL, U8_VALIDATE_ENTIRE, &error) < 0) {
1065 return (SET_ERROR(EILSEQ));
1068 if (vap->va_mask & AT_XVATTR) {
1069 if ((error = secpolicy_xvattr(ZTOV(dzp), (xvattr_t *)vap,
1070 crgetuid(cr), cr, vap->va_type)) != 0) {
1078 if ((vap->va_mode & S_ISVTX) && secpolicy_vnode_stky_modify(cr))
1079 vap->va_mode &= ~S_ISVTX;
1081 error = zfs_dirent_lookup(dzp, name, &zp, ZNEW);
1086 ASSERT3P(zp, ==, NULL);
1089 * Create a new file object and update the directory
1092 if ((error = zfs_zaccess(dzp, ACE_ADD_FILE, 0, B_FALSE, cr))) {
1097 * We only support the creation of regular files in
1098 * extended attribute directories.
1101 if ((dzp->z_pflags & ZFS_XATTR) &&
1102 (vap->va_type != VREG)) {
1103 error = SET_ERROR(EINVAL);
1107 if ((error = zfs_acl_ids_create(dzp, 0, vap,
1108 cr, vsecp, &acl_ids)) != 0)
1111 if (S_ISREG(vap->va_mode) || S_ISDIR(vap->va_mode))
1112 projid = zfs_inherit_projid(dzp);
1113 if (zfs_acl_ids_overquota(zfsvfs, &acl_ids, projid)) {
1114 zfs_acl_ids_free(&acl_ids);
1115 error = SET_ERROR(EDQUOT);
1119 getnewvnode_reserve_();
1121 tx = dmu_tx_create(os);
1123 dmu_tx_hold_sa_create(tx, acl_ids.z_aclp->z_acl_bytes +
1124 ZFS_SA_BASE_ATTR_SIZE);
1126 fuid_dirtied = zfsvfs->z_fuid_dirty;
1128 zfs_fuid_txhold(zfsvfs, tx);
1129 dmu_tx_hold_zap(tx, dzp->z_id, TRUE, name);
1130 dmu_tx_hold_sa(tx, dzp->z_sa_hdl, B_FALSE);
1131 if (!zfsvfs->z_use_sa &&
1132 acl_ids.z_aclp->z_acl_bytes > ZFS_ACE_SPACE) {
1133 dmu_tx_hold_write(tx, DMU_NEW_OBJECT,
1134 0, acl_ids.z_aclp->z_acl_bytes);
1136 error = dmu_tx_assign(tx, TXG_WAIT);
1138 zfs_acl_ids_free(&acl_ids);
1140 getnewvnode_drop_reserve();
1144 zfs_mknode(dzp, vap, tx, cr, 0, &zp, &acl_ids);
1146 zfs_fuid_sync(zfsvfs, tx);
1148 (void) zfs_link_create(dzp, name, zp, tx, ZNEW);
1149 txtype = zfs_log_create_txtype(Z_FILE, vsecp, vap);
1150 zfs_log_create(zilog, tx, txtype, dzp, zp, name,
1151 vsecp, acl_ids.z_fuidp, vap);
1152 zfs_acl_ids_free(&acl_ids);
1155 getnewvnode_drop_reserve();
1163 if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS)
1164 zil_commit(zilog, 0);
1171 * Remove an entry from a directory.
1173 * IN: dvp - vnode of directory to remove entry from.
1174 * name - name of entry to remove.
1175 * cr - credentials of caller.
1176 * ct - caller context
1177 * flags - case flags
1179 * RETURN: 0 on success, error code on failure.
1183 * vp - ctime (if nlink > 0)
1188 zfs_remove_(vnode_t *dvp, vnode_t *vp, const char *name, cred_t *cr)
1190 znode_t *dzp = VTOZ(dvp);
1193 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
1207 zilog = zfsvfs->z_log;
1212 if ((error = zfs_zaccess_delete(dzp, zp, cr))) {
1217 * Need to use rmdir for removing directories.
1219 if (vp->v_type == VDIR) {
1220 error = SET_ERROR(EPERM);
1224 vnevent_remove(vp, dvp, name, ct);
1228 /* are there any extended attributes? */
1229 error = sa_lookup(zp->z_sa_hdl, SA_ZPL_XATTR(zfsvfs),
1230 &xattr_obj, sizeof (xattr_obj));
1231 if (error == 0 && xattr_obj) {
1232 error = zfs_zget(zfsvfs, xattr_obj, &xzp);
1237 * We may delete the znode now, or we may put it in the unlinked set;
1238 * it depends on whether we're the last link, and on whether there are
1239 * other holds on the vnode. So we dmu_tx_hold() the right things to
1240 * allow for either case.
1242 tx = dmu_tx_create(zfsvfs->z_os);
1243 dmu_tx_hold_zap(tx, dzp->z_id, FALSE, name);
1244 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
1245 zfs_sa_upgrade_txholds(tx, zp);
1246 zfs_sa_upgrade_txholds(tx, dzp);
1249 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE);
1250 dmu_tx_hold_sa(tx, xzp->z_sa_hdl, B_FALSE);
1253 /* charge as an update -- would be nice not to charge at all */
1254 dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL);
1257 * Mark this transaction as typically resulting in a net free of space
1259 dmu_tx_mark_netfree(tx);
1261 error = dmu_tx_assign(tx, TXG_WAIT);
1269 * Remove the directory entry.
1271 error = zfs_link_destroy(dzp, name, zp, tx, ZEXISTS, &unlinked);
1279 zfs_unlinked_add(zp, tx);
1280 vp->v_vflag |= VV_NOSYNC;
1282 /* XXX check changes to linux vnops */
1284 zfs_log_remove(zilog, tx, txtype, dzp, name, obj, unlinked);
1292 if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS)
1293 zil_commit(zilog, 0);
1302 zfs_lookup_internal(znode_t *dzp, const char *name, vnode_t **vpp,
1303 struct componentname *cnp, int nameiop)
1305 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
1308 cnp->cn_nameptr = __DECONST(char *, name);
1309 cnp->cn_namelen = strlen(name);
1310 cnp->cn_nameiop = nameiop;
1311 cnp->cn_flags = ISLASTCN | SAVENAME;
1312 cnp->cn_lkflags = LK_EXCLUSIVE | LK_RETRY;
1313 cnp->cn_cred = kcred;
1314 cnp->cn_thread = curthread;
1316 if (zfsvfs->z_use_namecache && !zfsvfs->z_replay) {
1317 struct vop_lookup_args a;
1319 a.a_gen.a_desc = &vop_lookup_desc;
1320 a.a_dvp = ZTOV(dzp);
1323 error = vfs_cache_lookup(&a);
1325 error = zfs_lookup(ZTOV(dzp), name, vpp, cnp, nameiop, kcred,
1326 curthread, 0, B_FALSE);
1330 printf("got error %d on name %s on op %d\n", error, name,
1339 zfs_remove(znode_t *dzp, const char *name, cred_t *cr, int flags)
1343 struct componentname cn;
1345 if ((error = zfs_lookup_internal(dzp, name, &vp, &cn, DELETE)))
1348 error = zfs_remove_(ZTOV(dzp), vp, name, cr);
1353 * Create a new directory and insert it into dvp using the name
1354 * provided. Return a pointer to the inserted directory.
1356 * IN: dvp - vnode of directory to add subdir to.
1357 * dirname - name of new directory.
1358 * vap - attributes of new directory.
1359 * cr - credentials of caller.
1360 * ct - caller context
1361 * flags - case flags
1362 * vsecp - ACL to be set
1364 * OUT: vpp - vnode of created directory.
1366 * RETURN: 0 on success, error code on failure.
1369 * dvp - ctime|mtime updated
1370 * vp - ctime|mtime|atime updated
1374 zfs_mkdir(znode_t *dzp, const char *dirname, vattr_t *vap, znode_t **zpp,
1375 cred_t *cr, int flags, vsecattr_t *vsecp)
1378 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
1385 gid_t gid = crgetgid(cr);
1386 zfs_acl_ids_t acl_ids;
1387 boolean_t fuid_dirtied;
1389 ASSERT(vap->va_type == VDIR);
1392 * If we have an ephemeral id, ACL, or XVATTR then
1393 * make sure file system is at proper version
1396 ksid = crgetsid(cr, KSID_OWNER);
1398 uid = ksid_getid(ksid);
1401 if (zfsvfs->z_use_fuids == B_FALSE &&
1402 ((vap->va_mask & AT_XVATTR) ||
1403 IS_EPHEMERAL(uid) || IS_EPHEMERAL(gid)))
1404 return (SET_ERROR(EINVAL));
1408 zilog = zfsvfs->z_log;
1410 if (dzp->z_pflags & ZFS_XATTR) {
1412 return (SET_ERROR(EINVAL));
1415 if (zfsvfs->z_utf8 && u8_validate(dirname,
1416 strlen(dirname), NULL, U8_VALIDATE_ENTIRE, &error) < 0) {
1418 return (SET_ERROR(EILSEQ));
1421 if (vap->va_mask & AT_XVATTR) {
1422 if ((error = secpolicy_xvattr(ZTOV(dzp), (xvattr_t *)vap,
1423 crgetuid(cr), cr, vap->va_type)) != 0) {
1429 if ((error = zfs_acl_ids_create(dzp, 0, vap, cr,
1430 NULL, &acl_ids)) != 0) {
1436 * First make sure the new directory doesn't exist.
1438 * Existence is checked first to make sure we don't return
1439 * EACCES instead of EEXIST which can cause some applications
1444 if ((error = zfs_dirent_lookup(dzp, dirname, &zp, ZNEW))) {
1445 zfs_acl_ids_free(&acl_ids);
1449 ASSERT3P(zp, ==, NULL);
1451 if ((error = zfs_zaccess(dzp, ACE_ADD_SUBDIRECTORY, 0, B_FALSE, cr))) {
1452 zfs_acl_ids_free(&acl_ids);
1457 if (zfs_acl_ids_overquota(zfsvfs, &acl_ids, zfs_inherit_projid(dzp))) {
1458 zfs_acl_ids_free(&acl_ids);
1460 return (SET_ERROR(EDQUOT));
1464 * Add a new entry to the directory.
1466 getnewvnode_reserve_();
1467 tx = dmu_tx_create(zfsvfs->z_os);
1468 dmu_tx_hold_zap(tx, dzp->z_id, TRUE, dirname);
1469 dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL);
1470 fuid_dirtied = zfsvfs->z_fuid_dirty;
1472 zfs_fuid_txhold(zfsvfs, tx);
1473 if (!zfsvfs->z_use_sa && acl_ids.z_aclp->z_acl_bytes > ZFS_ACE_SPACE) {
1474 dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0,
1475 acl_ids.z_aclp->z_acl_bytes);
1478 dmu_tx_hold_sa_create(tx, acl_ids.z_aclp->z_acl_bytes +
1479 ZFS_SA_BASE_ATTR_SIZE);
1481 error = dmu_tx_assign(tx, TXG_WAIT);
1483 zfs_acl_ids_free(&acl_ids);
1485 getnewvnode_drop_reserve();
1493 zfs_mknode(dzp, vap, tx, cr, 0, &zp, &acl_ids);
1496 zfs_fuid_sync(zfsvfs, tx);
1499 * Now put new name in parent dir.
1501 (void) zfs_link_create(dzp, dirname, zp, tx, ZNEW);
1505 txtype = zfs_log_create_txtype(Z_DIR, NULL, vap);
1506 zfs_log_create(zilog, tx, txtype, dzp, zp, dirname, NULL,
1507 acl_ids.z_fuidp, vap);
1509 zfs_acl_ids_free(&acl_ids);
1513 getnewvnode_drop_reserve();
1515 if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS)
1516 zil_commit(zilog, 0);
1522 #if __FreeBSD_version < 1300124
1524 cache_vop_rmdir(struct vnode *dvp, struct vnode *vp)
1533 * Remove a directory subdir entry. If the current working
1534 * directory is the same as the subdir to be removed, the
1537 * IN: dvp - vnode of directory to remove from.
1538 * name - name of directory to be removed.
1539 * cwd - vnode of current working directory.
1540 * cr - credentials of caller.
1541 * ct - caller context
1542 * flags - case flags
1544 * RETURN: 0 on success, error code on failure.
1547 * dvp - ctime|mtime updated
1551 zfs_rmdir_(vnode_t *dvp, vnode_t *vp, const char *name, cred_t *cr)
1553 znode_t *dzp = VTOZ(dvp);
1554 znode_t *zp = VTOZ(vp);
1555 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
1563 zilog = zfsvfs->z_log;
1566 if ((error = zfs_zaccess_delete(dzp, zp, cr))) {
1570 if (vp->v_type != VDIR) {
1571 error = SET_ERROR(ENOTDIR);
1575 vnevent_rmdir(vp, dvp, name, ct);
1577 tx = dmu_tx_create(zfsvfs->z_os);
1578 dmu_tx_hold_zap(tx, dzp->z_id, FALSE, name);
1579 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
1580 dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL);
1581 zfs_sa_upgrade_txholds(tx, zp);
1582 zfs_sa_upgrade_txholds(tx, dzp);
1583 dmu_tx_mark_netfree(tx);
1584 error = dmu_tx_assign(tx, TXG_WAIT);
1591 error = zfs_link_destroy(dzp, name, zp, tx, ZEXISTS, NULL);
1594 uint64_t txtype = TX_RMDIR;
1595 zfs_log_remove(zilog, tx, txtype, dzp, name,
1596 ZFS_NO_OBJECT, B_FALSE);
1601 cache_vop_rmdir(dvp, vp);
1603 if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS)
1604 zil_commit(zilog, 0);
1611 zfs_rmdir(znode_t *dzp, const char *name, znode_t *cwd, cred_t *cr, int flags)
1613 struct componentname cn;
1617 if ((error = zfs_lookup_internal(dzp, name, &vp, &cn, DELETE)))
1620 error = zfs_rmdir_(ZTOV(dzp), vp, name, cr);
1626 * Read as many directory entries as will fit into the provided
1627 * buffer from the given directory cursor position (specified in
1628 * the uio structure).
1630 * IN: vp - vnode of directory to read.
1631 * uio - structure supplying read location, range info,
1632 * and return buffer.
1633 * cr - credentials of caller.
1634 * ct - caller context
1635 * flags - case flags
1637 * OUT: uio - updated offset and range, buffer filled.
1638 * eofp - set to true if end-of-file detected.
1640 * RETURN: 0 on success, error code on failure.
1643 * vp - atime updated
1645 * Note that the low 4 bits of the cookie returned by zap is always zero.
1646 * This allows us to use the low range for "special" directory entries:
1647 * We use 0 for '.', and 1 for '..'. If this is the root of the filesystem,
1648 * we use the offset 2 for the '.zfs' directory.
1652 zfs_readdir(vnode_t *vp, uio_t *uio, cred_t *cr, int *eofp,
1653 int *ncookies, ulong_t **cookies)
1655 znode_t *zp = VTOZ(vp);
1659 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1664 zap_attribute_t zap;
1665 uint_t bytes_wanted;
1666 uint64_t offset; /* must be unsigned; checks for < 1 */
1672 boolean_t check_sysattrs;
1675 ulong_t *cooks = NULL;
1681 if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_PARENT(zfsvfs),
1682 &parent, sizeof (parent))) != 0) {
1688 * If we are not given an eof variable,
1695 * Check for valid iov_len.
1697 if (uio->uio_iov->iov_len <= 0) {
1699 return (SET_ERROR(EINVAL));
1703 * Quit if directory has been removed (posix)
1705 if ((*eofp = zp->z_unlinked) != 0) {
1712 offset = uio->uio_loffset;
1713 prefetch = zp->z_zn_prefetch;
1716 * Initialize the iterator cursor.
1720 * Start iteration from the beginning of the directory.
1722 zap_cursor_init(&zc, os, zp->z_id);
1725 * The offset is a serialized cursor.
1727 zap_cursor_init_serialized(&zc, os, zp->z_id, offset);
1731 * Get space to change directory entries into fs independent format.
1733 iovp = uio->uio_iov;
1734 bytes_wanted = iovp->iov_len;
1735 if (uio->uio_segflg != UIO_SYSSPACE || uio->uio_iovcnt != 1) {
1736 bufsize = bytes_wanted;
1737 outbuf = kmem_alloc(bufsize, KM_SLEEP);
1738 odp = (struct dirent64 *)outbuf;
1740 bufsize = bytes_wanted;
1742 odp = (struct dirent64 *)iovp->iov_base;
1744 eodp = (struct edirent *)odp;
1746 if (ncookies != NULL) {
1748 * Minimum entry size is dirent size and 1 byte for a file name.
1750 ncooks = uio->uio_resid / (sizeof (struct dirent) -
1751 sizeof (((struct dirent *)NULL)->d_name) + 1);
1752 cooks = malloc(ncooks * sizeof (ulong_t), M_TEMP, M_WAITOK);
1757 * If this VFS supports the system attribute view interface; and
1758 * we're looking at an extended attribute directory; and we care
1759 * about normalization conflicts on this vfs; then we must check
1760 * for normalization conflicts with the sysattr name space.
1763 check_sysattrs = vfs_has_feature(vp->v_vfsp, VFSFT_SYSATTR_VIEWS) &&
1764 (vp->v_flag & V_XATTRDIR) && zfsvfs->z_norm &&
1765 (flags & V_RDDIR_ENTFLAGS);
1771 * Transform to file-system independent format
1774 while (outcount < bytes_wanted) {
1777 off64_t *next = NULL;
1780 * Special case `.', `..', and `.zfs'.
1783 (void) strcpy(zap.za_name, ".");
1784 zap.za_normalization_conflict = 0;
1787 } else if (offset == 1) {
1788 (void) strcpy(zap.za_name, "..");
1789 zap.za_normalization_conflict = 0;
1792 } else if (offset == 2 && zfs_show_ctldir(zp)) {
1793 (void) strcpy(zap.za_name, ZFS_CTLDIR_NAME);
1794 zap.za_normalization_conflict = 0;
1795 objnum = ZFSCTL_INO_ROOT;
1801 if ((error = zap_cursor_retrieve(&zc, &zap))) {
1802 if ((*eofp = (error == ENOENT)) != 0)
1808 if (zap.za_integer_length != 8 ||
1809 zap.za_num_integers != 1) {
1810 cmn_err(CE_WARN, "zap_readdir: bad directory "
1811 "entry, obj = %lld, offset = %lld\n",
1812 (u_longlong_t)zp->z_id,
1813 (u_longlong_t)offset);
1814 error = SET_ERROR(ENXIO);
1818 objnum = ZFS_DIRENT_OBJ(zap.za_first_integer);
1820 * MacOS X can extract the object type here such as:
1821 * uint8_t type = ZFS_DIRENT_TYPE(zap.za_first_integer);
1823 type = ZFS_DIRENT_TYPE(zap.za_first_integer);
1825 if (check_sysattrs && !zap.za_normalization_conflict) {
1827 zap.za_normalization_conflict =
1828 xattr_sysattr_casechk(zap.za_name);
1830 panic("%s:%u: TODO", __func__, __LINE__);
1835 if (flags & V_RDDIR_ACCFILTER) {
1837 * If we have no access at all, don't include
1838 * this entry in the returned information
1841 if (zfs_zget(zp->z_zfsvfs, objnum, &ezp) != 0)
1843 if (!zfs_has_access(ezp, cr)) {
1850 if (flags & V_RDDIR_ENTFLAGS)
1851 reclen = EDIRENT_RECLEN(strlen(zap.za_name));
1853 reclen = DIRENT64_RECLEN(strlen(zap.za_name));
1856 * Will this entry fit in the buffer?
1858 if (outcount + reclen > bufsize) {
1860 * Did we manage to fit anything in the buffer?
1863 error = SET_ERROR(EINVAL);
1868 if (flags & V_RDDIR_ENTFLAGS) {
1870 * Add extended flag entry:
1872 eodp->ed_ino = objnum;
1873 eodp->ed_reclen = reclen;
1874 /* NOTE: ed_off is the offset for the *next* entry */
1875 next = &(eodp->ed_off);
1876 eodp->ed_eflags = zap.za_normalization_conflict ?
1877 ED_CASE_CONFLICT : 0;
1878 (void) strncpy(eodp->ed_name, zap.za_name,
1879 EDIRENT_NAMELEN(reclen));
1880 eodp = (edirent_t *)((intptr_t)eodp + reclen);
1885 odp->d_ino = objnum;
1886 odp->d_reclen = reclen;
1887 odp->d_namlen = strlen(zap.za_name);
1888 /* NOTE: d_off is the offset for the *next* entry. */
1890 strlcpy(odp->d_name, zap.za_name, odp->d_namlen + 1);
1892 dirent_terminate(odp);
1893 odp = (dirent64_t *)((intptr_t)odp + reclen);
1897 ASSERT(outcount <= bufsize);
1899 /* Prefetch znode */
1901 dmu_prefetch(os, objnum, 0, 0, 0,
1902 ZIO_PRIORITY_SYNC_READ);
1906 * Move to the next entry, fill in the previous offset.
1908 if (offset > 2 || (offset == 2 && !zfs_show_ctldir(zp))) {
1909 zap_cursor_advance(&zc);
1910 offset = zap_cursor_serialize(&zc);
1915 /* Fill the offset right after advancing the cursor. */
1918 if (cooks != NULL) {
1921 KASSERT(ncooks >= 0, ("ncookies=%d", ncooks));
1924 zp->z_zn_prefetch = B_FALSE; /* a lookup will re-enable pre-fetching */
1926 /* Subtract unused cookies */
1927 if (ncookies != NULL)
1928 *ncookies -= ncooks;
1930 if (uio->uio_segflg == UIO_SYSSPACE && uio->uio_iovcnt == 1) {
1931 iovp->iov_base += outcount;
1932 iovp->iov_len -= outcount;
1933 uio->uio_resid -= outcount;
1934 } else if ((error = uiomove(outbuf, (long)outcount, UIO_READ, uio))) {
1936 * Reset the pointer.
1938 offset = uio->uio_loffset;
1942 zap_cursor_fini(&zc);
1943 if (uio->uio_segflg != UIO_SYSSPACE || uio->uio_iovcnt != 1)
1944 kmem_free(outbuf, bufsize);
1946 if (error == ENOENT)
1949 ZFS_ACCESSTIME_STAMP(zfsvfs, zp);
1951 uio->uio_loffset = offset;
1953 if (error != 0 && cookies != NULL) {
1954 free(*cookies, M_TEMP);
1962 * Get the requested file attributes and place them in the provided
1965 * IN: vp - vnode of file.
1966 * vap - va_mask identifies requested attributes.
1967 * If AT_XVATTR set, then optional attrs are requested
1968 * flags - ATTR_NOACLCHECK (CIFS server context)
1969 * cr - credentials of caller.
1971 * OUT: vap - attribute values.
1973 * RETURN: 0 (always succeeds).
1977 zfs_getattr(vnode_t *vp, vattr_t *vap, int flags, cred_t *cr)
1979 znode_t *zp = VTOZ(vp);
1980 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1983 u_longlong_t nblocks;
1984 uint64_t mtime[2], ctime[2], crtime[2], rdev;
1985 xvattr_t *xvap = (xvattr_t *)vap; /* vap may be an xvattr_t * */
1986 xoptattr_t *xoap = NULL;
1987 boolean_t skipaclchk = (flags & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE;
1988 sa_bulk_attr_t bulk[4];
1994 zfs_fuid_map_ids(zp, cr, &vap->va_uid, &vap->va_gid);
1996 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL, &mtime, 16);
1997 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL, &ctime, 16);
1998 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CRTIME(zfsvfs), NULL, &crtime, 16);
1999 if (vp->v_type == VBLK || vp->v_type == VCHR)
2000 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_RDEV(zfsvfs), NULL,
2003 if ((error = sa_bulk_lookup(zp->z_sa_hdl, bulk, count)) != 0) {
2009 * If ACL is trivial don't bother looking for ACE_READ_ATTRIBUTES.
2010 * Also, if we are the owner don't bother, since owner should
2011 * always be allowed to read basic attributes of file.
2013 if (!(zp->z_pflags & ZFS_ACL_TRIVIAL) &&
2014 (vap->va_uid != crgetuid(cr))) {
2015 if ((error = zfs_zaccess(zp, ACE_READ_ATTRIBUTES, 0,
2023 * Return all attributes. It's cheaper to provide the answer
2024 * than to determine whether we were asked the question.
2027 vap->va_type = IFTOVT(zp->z_mode);
2028 vap->va_mode = zp->z_mode & ~S_IFMT;
2030 vap->va_nodeid = zp->z_id;
2031 vap->va_nlink = zp->z_links;
2032 if ((vp->v_flag & VROOT) && zfs_show_ctldir(zp) &&
2033 zp->z_links < ZFS_LINK_MAX)
2035 vap->va_size = zp->z_size;
2036 if (vp->v_type == VBLK || vp->v_type == VCHR)
2037 vap->va_rdev = zfs_cmpldev(rdev);
2038 vap->va_seq = zp->z_seq;
2039 vap->va_flags = 0; /* FreeBSD: Reset chflags(2) flags. */
2040 vap->va_filerev = zp->z_seq;
2043 * Add in any requested optional attributes and the create time.
2044 * Also set the corresponding bits in the returned attribute bitmap.
2046 if ((xoap = xva_getxoptattr(xvap)) != NULL && zfsvfs->z_use_fuids) {
2047 if (XVA_ISSET_REQ(xvap, XAT_ARCHIVE)) {
2049 ((zp->z_pflags & ZFS_ARCHIVE) != 0);
2050 XVA_SET_RTN(xvap, XAT_ARCHIVE);
2053 if (XVA_ISSET_REQ(xvap, XAT_READONLY)) {
2054 xoap->xoa_readonly =
2055 ((zp->z_pflags & ZFS_READONLY) != 0);
2056 XVA_SET_RTN(xvap, XAT_READONLY);
2059 if (XVA_ISSET_REQ(xvap, XAT_SYSTEM)) {
2061 ((zp->z_pflags & ZFS_SYSTEM) != 0);
2062 XVA_SET_RTN(xvap, XAT_SYSTEM);
2065 if (XVA_ISSET_REQ(xvap, XAT_HIDDEN)) {
2067 ((zp->z_pflags & ZFS_HIDDEN) != 0);
2068 XVA_SET_RTN(xvap, XAT_HIDDEN);
2071 if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK)) {
2072 xoap->xoa_nounlink =
2073 ((zp->z_pflags & ZFS_NOUNLINK) != 0);
2074 XVA_SET_RTN(xvap, XAT_NOUNLINK);
2077 if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE)) {
2078 xoap->xoa_immutable =
2079 ((zp->z_pflags & ZFS_IMMUTABLE) != 0);
2080 XVA_SET_RTN(xvap, XAT_IMMUTABLE);
2083 if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY)) {
2084 xoap->xoa_appendonly =
2085 ((zp->z_pflags & ZFS_APPENDONLY) != 0);
2086 XVA_SET_RTN(xvap, XAT_APPENDONLY);
2089 if (XVA_ISSET_REQ(xvap, XAT_NODUMP)) {
2091 ((zp->z_pflags & ZFS_NODUMP) != 0);
2092 XVA_SET_RTN(xvap, XAT_NODUMP);
2095 if (XVA_ISSET_REQ(xvap, XAT_OPAQUE)) {
2097 ((zp->z_pflags & ZFS_OPAQUE) != 0);
2098 XVA_SET_RTN(xvap, XAT_OPAQUE);
2101 if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED)) {
2102 xoap->xoa_av_quarantined =
2103 ((zp->z_pflags & ZFS_AV_QUARANTINED) != 0);
2104 XVA_SET_RTN(xvap, XAT_AV_QUARANTINED);
2107 if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED)) {
2108 xoap->xoa_av_modified =
2109 ((zp->z_pflags & ZFS_AV_MODIFIED) != 0);
2110 XVA_SET_RTN(xvap, XAT_AV_MODIFIED);
2113 if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP) &&
2114 vp->v_type == VREG) {
2115 zfs_sa_get_scanstamp(zp, xvap);
2118 if (XVA_ISSET_REQ(xvap, XAT_REPARSE)) {
2119 xoap->xoa_reparse = ((zp->z_pflags & ZFS_REPARSE) != 0);
2120 XVA_SET_RTN(xvap, XAT_REPARSE);
2122 if (XVA_ISSET_REQ(xvap, XAT_GEN)) {
2123 xoap->xoa_generation = zp->z_gen;
2124 XVA_SET_RTN(xvap, XAT_GEN);
2127 if (XVA_ISSET_REQ(xvap, XAT_OFFLINE)) {
2129 ((zp->z_pflags & ZFS_OFFLINE) != 0);
2130 XVA_SET_RTN(xvap, XAT_OFFLINE);
2133 if (XVA_ISSET_REQ(xvap, XAT_SPARSE)) {
2135 ((zp->z_pflags & ZFS_SPARSE) != 0);
2136 XVA_SET_RTN(xvap, XAT_SPARSE);
2139 if (XVA_ISSET_REQ(xvap, XAT_PROJINHERIT)) {
2140 xoap->xoa_projinherit =
2141 ((zp->z_pflags & ZFS_PROJINHERIT) != 0);
2142 XVA_SET_RTN(xvap, XAT_PROJINHERIT);
2145 if (XVA_ISSET_REQ(xvap, XAT_PROJID)) {
2146 xoap->xoa_projid = zp->z_projid;
2147 XVA_SET_RTN(xvap, XAT_PROJID);
2151 ZFS_TIME_DECODE(&vap->va_atime, zp->z_atime);
2152 ZFS_TIME_DECODE(&vap->va_mtime, mtime);
2153 ZFS_TIME_DECODE(&vap->va_ctime, ctime);
2154 ZFS_TIME_DECODE(&vap->va_birthtime, crtime);
2157 sa_object_size(zp->z_sa_hdl, &blksize, &nblocks);
2158 vap->va_blksize = blksize;
2159 vap->va_bytes = nblocks << 9; /* nblocks * 512 */
2161 if (zp->z_blksz == 0) {
2163 * Block size hasn't been set; suggest maximal I/O transfers.
2165 vap->va_blksize = zfsvfs->z_max_blksz;
2173 * Set the file attributes to the values contained in the
2176 * IN: zp - znode of file to be modified.
2177 * vap - new attribute values.
2178 * If AT_XVATTR set, then optional attrs are being set
2179 * flags - ATTR_UTIME set if non-default time values provided.
2180 * - ATTR_NOACLCHECK (CIFS context only).
2181 * cr - credentials of caller.
2182 * ct - caller context
2184 * RETURN: 0 on success, error code on failure.
2187 * vp - ctime updated, mtime updated if size changed.
2191 zfs_setattr(znode_t *zp, vattr_t *vap, int flags, cred_t *cr)
2193 vnode_t *vp = ZTOV(zp);
2194 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
2195 objset_t *os = zfsvfs->z_os;
2200 uint_t mask = vap->va_mask;
2201 uint_t saved_mask = 0;
2202 uint64_t saved_mode;
2205 uint64_t new_uid, new_gid;
2207 uint64_t mtime[2], ctime[2];
2208 uint64_t projid = ZFS_INVALID_PROJID;
2210 int need_policy = FALSE;
2212 zfs_fuid_info_t *fuidp = NULL;
2213 xvattr_t *xvap = (xvattr_t *)vap; /* vap may be an xvattr_t * */
2216 boolean_t skipaclchk = (flags & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE;
2217 boolean_t fuid_dirtied = B_FALSE;
2218 sa_bulk_attr_t bulk[7], xattr_bulk[7];
2219 int count = 0, xattr_count = 0;
2224 if (mask & AT_NOSET)
2225 return (SET_ERROR(EINVAL));
2230 zilog = zfsvfs->z_log;
2233 * Make sure that if we have ephemeral uid/gid or xvattr specified
2234 * that file system is at proper version level
2237 if (zfsvfs->z_use_fuids == B_FALSE &&
2238 (((mask & AT_UID) && IS_EPHEMERAL(vap->va_uid)) ||
2239 ((mask & AT_GID) && IS_EPHEMERAL(vap->va_gid)) ||
2240 (mask & AT_XVATTR))) {
2242 return (SET_ERROR(EINVAL));
2245 if (mask & AT_SIZE && vp->v_type == VDIR) {
2247 return (SET_ERROR(EISDIR));
2250 if (mask & AT_SIZE && vp->v_type != VREG && vp->v_type != VFIFO) {
2252 return (SET_ERROR(EINVAL));
2256 * If this is an xvattr_t, then get a pointer to the structure of
2257 * optional attributes. If this is NULL, then we have a vattr_t.
2259 xoap = xva_getxoptattr(xvap);
2261 xva_init(&tmpxvattr);
2264 * Immutable files can only alter immutable bit and atime
2266 if ((zp->z_pflags & ZFS_IMMUTABLE) &&
2267 ((mask & (AT_SIZE|AT_UID|AT_GID|AT_MTIME|AT_MODE)) ||
2268 ((mask & AT_XVATTR) && XVA_ISSET_REQ(xvap, XAT_CREATETIME)))) {
2270 return (SET_ERROR(EPERM));
2274 * Note: ZFS_READONLY is handled in zfs_zaccess_common.
2278 * Verify timestamps doesn't overflow 32 bits.
2279 * ZFS can handle large timestamps, but 32bit syscalls can't
2280 * handle times greater than 2039. This check should be removed
2281 * once large timestamps are fully supported.
2283 if (mask & (AT_ATIME | AT_MTIME)) {
2284 if (((mask & AT_ATIME) && TIMESPEC_OVERFLOW(&vap->va_atime)) ||
2285 ((mask & AT_MTIME) && TIMESPEC_OVERFLOW(&vap->va_mtime))) {
2287 return (SET_ERROR(EOVERFLOW));
2290 if (xoap != NULL && (mask & AT_XVATTR)) {
2291 if (XVA_ISSET_REQ(xvap, XAT_CREATETIME) &&
2292 TIMESPEC_OVERFLOW(&vap->va_birthtime)) {
2294 return (SET_ERROR(EOVERFLOW));
2297 if (XVA_ISSET_REQ(xvap, XAT_PROJID)) {
2298 if (!dmu_objset_projectquota_enabled(os) ||
2299 (!S_ISREG(zp->z_mode) && !S_ISDIR(zp->z_mode))) {
2301 return (SET_ERROR(EOPNOTSUPP));
2304 projid = xoap->xoa_projid;
2305 if (unlikely(projid == ZFS_INVALID_PROJID)) {
2307 return (SET_ERROR(EINVAL));
2310 if (projid == zp->z_projid && zp->z_pflags & ZFS_PROJID)
2311 projid = ZFS_INVALID_PROJID;
2316 if (XVA_ISSET_REQ(xvap, XAT_PROJINHERIT) &&
2317 (xoap->xoa_projinherit !=
2318 ((zp->z_pflags & ZFS_PROJINHERIT) != 0)) &&
2319 (!dmu_objset_projectquota_enabled(os) ||
2320 (!S_ISREG(zp->z_mode) && !S_ISDIR(zp->z_mode)))) {
2322 return (SET_ERROR(EOPNOTSUPP));
2329 if (zfsvfs->z_vfs->vfs_flag & VFS_RDONLY) {
2331 return (SET_ERROR(EROFS));
2335 * First validate permissions
2338 if (mask & AT_SIZE) {
2340 * XXX - Note, we are not providing any open
2341 * mode flags here (like FNDELAY), so we may
2342 * block if there are locks present... this
2343 * should be addressed in openat().
2345 /* XXX - would it be OK to generate a log record here? */
2346 err = zfs_freesp(zp, vap->va_size, 0, 0, FALSE);
2353 if (mask & (AT_ATIME|AT_MTIME) ||
2354 ((mask & AT_XVATTR) && (XVA_ISSET_REQ(xvap, XAT_HIDDEN) ||
2355 XVA_ISSET_REQ(xvap, XAT_READONLY) ||
2356 XVA_ISSET_REQ(xvap, XAT_ARCHIVE) ||
2357 XVA_ISSET_REQ(xvap, XAT_OFFLINE) ||
2358 XVA_ISSET_REQ(xvap, XAT_SPARSE) ||
2359 XVA_ISSET_REQ(xvap, XAT_CREATETIME) ||
2360 XVA_ISSET_REQ(xvap, XAT_SYSTEM)))) {
2361 need_policy = zfs_zaccess(zp, ACE_WRITE_ATTRIBUTES, 0,
2365 if (mask & (AT_UID|AT_GID)) {
2366 int idmask = (mask & (AT_UID|AT_GID));
2371 * NOTE: even if a new mode is being set,
2372 * we may clear S_ISUID/S_ISGID bits.
2375 if (!(mask & AT_MODE))
2376 vap->va_mode = zp->z_mode;
2379 * Take ownership or chgrp to group we are a member of
2382 take_owner = (mask & AT_UID) && (vap->va_uid == crgetuid(cr));
2383 take_group = (mask & AT_GID) &&
2384 zfs_groupmember(zfsvfs, vap->va_gid, cr);
2387 * If both AT_UID and AT_GID are set then take_owner and
2388 * take_group must both be set in order to allow taking
2391 * Otherwise, send the check through secpolicy_vnode_setattr()
2395 if (((idmask == (AT_UID|AT_GID)) && take_owner && take_group) ||
2396 ((idmask == AT_UID) && take_owner) ||
2397 ((idmask == AT_GID) && take_group)) {
2398 if (zfs_zaccess(zp, ACE_WRITE_OWNER, 0,
2399 skipaclchk, cr) == 0) {
2401 * Remove setuid/setgid for non-privileged users
2403 secpolicy_setid_clear(vap, vp, cr);
2404 trim_mask = (mask & (AT_UID|AT_GID));
2413 oldva.va_mode = zp->z_mode;
2414 zfs_fuid_map_ids(zp, cr, &oldva.va_uid, &oldva.va_gid);
2415 if (mask & AT_XVATTR) {
2417 * Update xvattr mask to include only those attributes
2418 * that are actually changing.
2420 * the bits will be restored prior to actually setting
2421 * the attributes so the caller thinks they were set.
2423 if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY)) {
2424 if (xoap->xoa_appendonly !=
2425 ((zp->z_pflags & ZFS_APPENDONLY) != 0)) {
2428 XVA_CLR_REQ(xvap, XAT_APPENDONLY);
2429 XVA_SET_REQ(&tmpxvattr, XAT_APPENDONLY);
2433 if (XVA_ISSET_REQ(xvap, XAT_PROJINHERIT)) {
2434 if (xoap->xoa_projinherit !=
2435 ((zp->z_pflags & ZFS_PROJINHERIT) != 0)) {
2438 XVA_CLR_REQ(xvap, XAT_PROJINHERIT);
2439 XVA_SET_REQ(&tmpxvattr, XAT_PROJINHERIT);
2443 if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK)) {
2444 if (xoap->xoa_nounlink !=
2445 ((zp->z_pflags & ZFS_NOUNLINK) != 0)) {
2448 XVA_CLR_REQ(xvap, XAT_NOUNLINK);
2449 XVA_SET_REQ(&tmpxvattr, XAT_NOUNLINK);
2453 if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE)) {
2454 if (xoap->xoa_immutable !=
2455 ((zp->z_pflags & ZFS_IMMUTABLE) != 0)) {
2458 XVA_CLR_REQ(xvap, XAT_IMMUTABLE);
2459 XVA_SET_REQ(&tmpxvattr, XAT_IMMUTABLE);
2463 if (XVA_ISSET_REQ(xvap, XAT_NODUMP)) {
2464 if (xoap->xoa_nodump !=
2465 ((zp->z_pflags & ZFS_NODUMP) != 0)) {
2468 XVA_CLR_REQ(xvap, XAT_NODUMP);
2469 XVA_SET_REQ(&tmpxvattr, XAT_NODUMP);
2473 if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED)) {
2474 if (xoap->xoa_av_modified !=
2475 ((zp->z_pflags & ZFS_AV_MODIFIED) != 0)) {
2478 XVA_CLR_REQ(xvap, XAT_AV_MODIFIED);
2479 XVA_SET_REQ(&tmpxvattr, XAT_AV_MODIFIED);
2483 if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED)) {
2484 if ((vp->v_type != VREG &&
2485 xoap->xoa_av_quarantined) ||
2486 xoap->xoa_av_quarantined !=
2487 ((zp->z_pflags & ZFS_AV_QUARANTINED) != 0)) {
2490 XVA_CLR_REQ(xvap, XAT_AV_QUARANTINED);
2491 XVA_SET_REQ(&tmpxvattr, XAT_AV_QUARANTINED);
2495 if (XVA_ISSET_REQ(xvap, XAT_REPARSE)) {
2497 return (SET_ERROR(EPERM));
2500 if (need_policy == FALSE &&
2501 (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP) ||
2502 XVA_ISSET_REQ(xvap, XAT_OPAQUE))) {
2507 if (mask & AT_MODE) {
2508 if (zfs_zaccess(zp, ACE_WRITE_ACL, 0, skipaclchk, cr) == 0) {
2509 err = secpolicy_setid_setsticky_clear(vp, vap,
2515 trim_mask |= AT_MODE;
2523 * If trim_mask is set then take ownership
2524 * has been granted or write_acl is present and user
2525 * has the ability to modify mode. In that case remove
2526 * UID|GID and or MODE from mask so that
2527 * secpolicy_vnode_setattr() doesn't revoke it.
2531 saved_mask = vap->va_mask;
2532 vap->va_mask &= ~trim_mask;
2533 if (trim_mask & AT_MODE) {
2535 * Save the mode, as secpolicy_vnode_setattr()
2536 * will overwrite it with ova.va_mode.
2538 saved_mode = vap->va_mode;
2541 err = secpolicy_vnode_setattr(cr, vp, vap, &oldva, flags,
2542 (int (*)(void *, int, cred_t *))zfs_zaccess_unix, zp);
2549 vap->va_mask |= saved_mask;
2550 if (trim_mask & AT_MODE) {
2552 * Recover the mode after
2553 * secpolicy_vnode_setattr().
2555 vap->va_mode = saved_mode;
2561 * secpolicy_vnode_setattr, or take ownership may have
2564 mask = vap->va_mask;
2566 if ((mask & (AT_UID | AT_GID)) || projid != ZFS_INVALID_PROJID) {
2567 err = sa_lookup(zp->z_sa_hdl, SA_ZPL_XATTR(zfsvfs),
2568 &xattr_obj, sizeof (xattr_obj));
2570 if (err == 0 && xattr_obj) {
2571 err = zfs_zget(zp->z_zfsvfs, xattr_obj, &attrzp);
2573 err = vn_lock(ZTOV(attrzp), LK_EXCLUSIVE);
2575 vrele(ZTOV(attrzp));
2580 if (mask & AT_UID) {
2581 new_uid = zfs_fuid_create(zfsvfs,
2582 (uint64_t)vap->va_uid, cr, ZFS_OWNER, &fuidp);
2583 if (new_uid != zp->z_uid &&
2584 zfs_id_overquota(zfsvfs, DMU_USERUSED_OBJECT,
2588 err = SET_ERROR(EDQUOT);
2593 if (mask & AT_GID) {
2594 new_gid = zfs_fuid_create(zfsvfs, (uint64_t)vap->va_gid,
2595 cr, ZFS_GROUP, &fuidp);
2596 if (new_gid != zp->z_gid &&
2597 zfs_id_overquota(zfsvfs, DMU_GROUPUSED_OBJECT,
2601 err = SET_ERROR(EDQUOT);
2606 if (projid != ZFS_INVALID_PROJID &&
2607 zfs_id_overquota(zfsvfs, DMU_PROJECTUSED_OBJECT, projid)) {
2610 err = SET_ERROR(EDQUOT);
2614 tx = dmu_tx_create(os);
2616 if (mask & AT_MODE) {
2617 uint64_t pmode = zp->z_mode;
2619 new_mode = (pmode & S_IFMT) | (vap->va_mode & ~S_IFMT);
2621 if (zp->z_zfsvfs->z_acl_mode == ZFS_ACL_RESTRICTED &&
2622 !(zp->z_pflags & ZFS_ACL_TRIVIAL)) {
2623 err = SET_ERROR(EPERM);
2627 if ((err = zfs_acl_chmod_setattr(zp, &aclp, new_mode)))
2630 if (!zp->z_is_sa && ((acl_obj = zfs_external_acl(zp)) != 0)) {
2632 * Are we upgrading ACL from old V0 format
2635 if (zfsvfs->z_version >= ZPL_VERSION_FUID &&
2636 zfs_znode_acl_version(zp) ==
2637 ZFS_ACL_VERSION_INITIAL) {
2638 dmu_tx_hold_free(tx, acl_obj, 0,
2640 dmu_tx_hold_write(tx, DMU_NEW_OBJECT,
2641 0, aclp->z_acl_bytes);
2643 dmu_tx_hold_write(tx, acl_obj, 0,
2646 } else if (!zp->z_is_sa && aclp->z_acl_bytes > ZFS_ACE_SPACE) {
2647 dmu_tx_hold_write(tx, DMU_NEW_OBJECT,
2648 0, aclp->z_acl_bytes);
2650 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE);
2652 if (((mask & AT_XVATTR) &&
2653 XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP)) ||
2654 (projid != ZFS_INVALID_PROJID &&
2655 !(zp->z_pflags & ZFS_PROJID)))
2656 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE);
2658 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
2662 dmu_tx_hold_sa(tx, attrzp->z_sa_hdl, B_FALSE);
2665 fuid_dirtied = zfsvfs->z_fuid_dirty;
2667 zfs_fuid_txhold(zfsvfs, tx);
2669 zfs_sa_upgrade_txholds(tx, zp);
2671 err = dmu_tx_assign(tx, TXG_WAIT);
2677 * Set each attribute requested.
2678 * We group settings according to the locks they need to acquire.
2680 * Note: you cannot set ctime directly, although it will be
2681 * updated as a side-effect of calling this function.
2684 if (projid != ZFS_INVALID_PROJID && !(zp->z_pflags & ZFS_PROJID)) {
2686 * For the existed object that is upgraded from old system,
2687 * its on-disk layout has no slot for the project ID attribute.
2688 * But quota accounting logic needs to access related slots by
2689 * offset directly. So we need to adjust old objects' layout
2690 * to make the project ID to some unified and fixed offset.
2693 err = sa_add_projid(attrzp->z_sa_hdl, tx, projid);
2695 err = sa_add_projid(zp->z_sa_hdl, tx, projid);
2697 if (unlikely(err == EEXIST))
2702 projid = ZFS_INVALID_PROJID;
2705 if (mask & (AT_UID|AT_GID|AT_MODE))
2706 mutex_enter(&zp->z_acl_lock);
2708 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL,
2709 &zp->z_pflags, sizeof (zp->z_pflags));
2712 if (mask & (AT_UID|AT_GID|AT_MODE))
2713 mutex_enter(&attrzp->z_acl_lock);
2714 SA_ADD_BULK_ATTR(xattr_bulk, xattr_count,
2715 SA_ZPL_FLAGS(zfsvfs), NULL, &attrzp->z_pflags,
2716 sizeof (attrzp->z_pflags));
2717 if (projid != ZFS_INVALID_PROJID) {
2718 attrzp->z_projid = projid;
2719 SA_ADD_BULK_ATTR(xattr_bulk, xattr_count,
2720 SA_ZPL_PROJID(zfsvfs), NULL, &attrzp->z_projid,
2721 sizeof (attrzp->z_projid));
2725 if (mask & (AT_UID|AT_GID)) {
2727 if (mask & AT_UID) {
2728 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zfsvfs), NULL,
2729 &new_uid, sizeof (new_uid));
2730 zp->z_uid = new_uid;
2732 SA_ADD_BULK_ATTR(xattr_bulk, xattr_count,
2733 SA_ZPL_UID(zfsvfs), NULL, &new_uid,
2735 attrzp->z_uid = new_uid;
2739 if (mask & AT_GID) {
2740 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zfsvfs),
2741 NULL, &new_gid, sizeof (new_gid));
2742 zp->z_gid = new_gid;
2744 SA_ADD_BULK_ATTR(xattr_bulk, xattr_count,
2745 SA_ZPL_GID(zfsvfs), NULL, &new_gid,
2747 attrzp->z_gid = new_gid;
2750 if (!(mask & AT_MODE)) {
2751 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs),
2752 NULL, &new_mode, sizeof (new_mode));
2753 new_mode = zp->z_mode;
2755 err = zfs_acl_chown_setattr(zp);
2758 err = zfs_acl_chown_setattr(attrzp);
2763 if (mask & AT_MODE) {
2764 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs), NULL,
2765 &new_mode, sizeof (new_mode));
2766 zp->z_mode = new_mode;
2767 ASSERT3U((uintptr_t)aclp, !=, 0);
2768 err = zfs_aclset_common(zp, aclp, cr, tx);
2770 if (zp->z_acl_cached)
2771 zfs_acl_free(zp->z_acl_cached);
2772 zp->z_acl_cached = aclp;
2777 if (mask & AT_ATIME) {
2778 ZFS_TIME_ENCODE(&vap->va_atime, zp->z_atime);
2779 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zfsvfs), NULL,
2780 &zp->z_atime, sizeof (zp->z_atime));
2783 if (mask & AT_MTIME) {
2784 ZFS_TIME_ENCODE(&vap->va_mtime, mtime);
2785 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL,
2786 mtime, sizeof (mtime));
2789 if (projid != ZFS_INVALID_PROJID) {
2790 zp->z_projid = projid;
2791 SA_ADD_BULK_ATTR(bulk, count,
2792 SA_ZPL_PROJID(zfsvfs), NULL, &zp->z_projid,
2793 sizeof (zp->z_projid));
2796 /* XXX - shouldn't this be done *before* the ATIME/MTIME checks? */
2797 if (mask & AT_SIZE && !(mask & AT_MTIME)) {
2798 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs),
2799 NULL, mtime, sizeof (mtime));
2800 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL,
2801 &ctime, sizeof (ctime));
2802 zfs_tstamp_update_setup(zp, CONTENT_MODIFIED, mtime, ctime);
2803 } else if (mask != 0) {
2804 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL,
2805 &ctime, sizeof (ctime));
2806 zfs_tstamp_update_setup(zp, STATE_CHANGED, mtime, ctime);
2808 SA_ADD_BULK_ATTR(xattr_bulk, xattr_count,
2809 SA_ZPL_CTIME(zfsvfs), NULL,
2810 &ctime, sizeof (ctime));
2811 zfs_tstamp_update_setup(attrzp, STATE_CHANGED,
2817 * Do this after setting timestamps to prevent timestamp
2818 * update from toggling bit
2821 if (xoap && (mask & AT_XVATTR)) {
2823 if (XVA_ISSET_REQ(xvap, XAT_CREATETIME))
2824 xoap->xoa_createtime = vap->va_birthtime;
2826 * restore trimmed off masks
2827 * so that return masks can be set for caller.
2830 if (XVA_ISSET_REQ(&tmpxvattr, XAT_APPENDONLY)) {
2831 XVA_SET_REQ(xvap, XAT_APPENDONLY);
2833 if (XVA_ISSET_REQ(&tmpxvattr, XAT_NOUNLINK)) {
2834 XVA_SET_REQ(xvap, XAT_NOUNLINK);
2836 if (XVA_ISSET_REQ(&tmpxvattr, XAT_IMMUTABLE)) {
2837 XVA_SET_REQ(xvap, XAT_IMMUTABLE);
2839 if (XVA_ISSET_REQ(&tmpxvattr, XAT_NODUMP)) {
2840 XVA_SET_REQ(xvap, XAT_NODUMP);
2842 if (XVA_ISSET_REQ(&tmpxvattr, XAT_AV_MODIFIED)) {
2843 XVA_SET_REQ(xvap, XAT_AV_MODIFIED);
2845 if (XVA_ISSET_REQ(&tmpxvattr, XAT_AV_QUARANTINED)) {
2846 XVA_SET_REQ(xvap, XAT_AV_QUARANTINED);
2848 if (XVA_ISSET_REQ(&tmpxvattr, XAT_PROJINHERIT)) {
2849 XVA_SET_REQ(xvap, XAT_PROJINHERIT);
2852 if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP))
2853 ASSERT(vp->v_type == VREG);
2855 zfs_xvattr_set(zp, xvap, tx);
2859 zfs_fuid_sync(zfsvfs, tx);
2862 zfs_log_setattr(zilog, tx, TX_SETATTR, zp, vap, mask, fuidp);
2864 if (mask & (AT_UID|AT_GID|AT_MODE))
2865 mutex_exit(&zp->z_acl_lock);
2868 if (mask & (AT_UID|AT_GID|AT_MODE))
2869 mutex_exit(&attrzp->z_acl_lock);
2872 if (err == 0 && attrzp) {
2873 err2 = sa_bulk_update(attrzp->z_sa_hdl, xattr_bulk,
2885 zfs_fuid_info_free(fuidp);
2892 err2 = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx);
2897 if (os->os_sync == ZFS_SYNC_ALWAYS)
2898 zil_commit(zilog, 0);
2905 * We acquire all but fdvp locks using non-blocking acquisitions. If we
2906 * fail to acquire any lock in the path we will drop all held locks,
2907 * acquire the new lock in a blocking fashion, and then release it and
2908 * restart the rename. This acquire/release step ensures that we do not
2909 * spin on a lock waiting for release. On error release all vnode locks
2910 * and decrement references the way tmpfs_rename() would do.
2913 zfs_rename_relock(struct vnode *sdvp, struct vnode **svpp,
2914 struct vnode *tdvp, struct vnode **tvpp,
2915 const struct componentname *scnp, const struct componentname *tcnp)
2918 struct vnode *nvp, *svp, *tvp;
2919 znode_t *sdzp, *tdzp, *szp, *tzp;
2920 const char *snm = scnp->cn_nameptr;
2921 const char *tnm = tcnp->cn_nameptr;
2925 if (*tvpp != NULL && *tvpp != tdvp)
2929 error = vn_lock(sdvp, LK_EXCLUSIVE);
2934 error = vn_lock(tdvp, LK_EXCLUSIVE | LK_NOWAIT);
2939 error = vn_lock(tdvp, LK_EXCLUSIVE);
2948 * Before using sdzp and tdzp we must ensure that they are live.
2949 * As a porting legacy from illumos we have two things to worry
2950 * about. One is typical for FreeBSD and it is that the vnode is
2951 * not reclaimed (doomed). The other is that the znode is live.
2952 * The current code can invalidate the znode without acquiring the
2953 * corresponding vnode lock if the object represented by the znode
2954 * and vnode is no longer valid after a rollback or receive operation.
2955 * z_teardown_lock hidden behind ZFS_ENTER and ZFS_EXIT is the lock
2956 * that protects the znodes from the invalidation.
2958 zfsvfs = sdzp->z_zfsvfs;
2959 ASSERT3P(zfsvfs, ==, tdzp->z_zfsvfs);
2963 * We can not use ZFS_VERIFY_ZP() here because it could directly return
2964 * bypassing the cleanup code in the case of an error.
2966 if (tdzp->z_sa_hdl == NULL || sdzp->z_sa_hdl == NULL) {
2970 error = SET_ERROR(EIO);
2975 * Re-resolve svp to be certain it still exists and fetch the
2978 error = zfs_dirent_lookup(sdzp, snm, &szp, ZEXISTS);
2980 /* Source entry invalid or not there. */
2984 if ((scnp->cn_flags & ISDOTDOT) != 0 ||
2985 (scnp->cn_namelen == 1 && scnp->cn_nameptr[0] == '.'))
2986 error = SET_ERROR(EINVAL);
2992 * Re-resolve tvp, if it disappeared we just carry on.
2994 error = zfs_dirent_lookup(tdzp, tnm, &tzp, 0);
3000 if ((tcnp->cn_flags & ISDOTDOT) != 0)
3001 error = SET_ERROR(EINVAL);
3010 * At present the vnode locks must be acquired before z_teardown_lock,
3011 * although it would be more logical to use the opposite order.
3016 * Now try acquire locks on svp and tvp.
3019 error = vn_lock(nvp, LK_EXCLUSIVE | LK_NOWAIT);
3025 if (error != EBUSY) {
3029 error = vn_lock(nvp, LK_EXCLUSIVE);
3036 * Concurrent rename race.
3041 error = SET_ERROR(EINVAL);
3056 error = vn_lock(nvp, LK_EXCLUSIVE | LK_NOWAIT);
3061 if (error != EBUSY) {
3065 error = vn_lock(nvp, LK_EXCLUSIVE);
3083 * Note that we must use VRELE_ASYNC in this function as it walks
3084 * up the directory tree and vrele may need to acquire an exclusive
3085 * lock if a last reference to a vnode is dropped.
3088 zfs_rename_check(znode_t *szp, znode_t *sdzp, znode_t *tdzp)
3095 zfsvfs = tdzp->z_zfsvfs;
3097 return (SET_ERROR(EINVAL));
3100 if (tdzp->z_id == zfsvfs->z_root)
3104 ASSERT(!zp->z_unlinked);
3105 if ((error = sa_lookup(zp->z_sa_hdl,
3106 SA_ZPL_PARENT(zfsvfs), &parent, sizeof (parent))) != 0)
3109 if (parent == szp->z_id) {
3110 error = SET_ERROR(EINVAL);
3113 if (parent == zfsvfs->z_root)
3115 if (parent == sdzp->z_id)
3118 error = zfs_zget(zfsvfs, parent, &zp1);
3123 VN_RELE_ASYNC(ZTOV(zp),
3124 dsl_pool_zrele_taskq(
3125 dmu_objset_pool(zfsvfs->z_os)));
3129 if (error == ENOTDIR)
3130 panic("checkpath: .. not a directory\n");
3132 VN_RELE_ASYNC(ZTOV(zp),
3133 dsl_pool_zrele_taskq(dmu_objset_pool(zfsvfs->z_os)));
3137 #if __FreeBSD_version < 1300124
3139 cache_vop_rename(struct vnode *fdvp, struct vnode *fvp, struct vnode *tdvp,
3140 struct vnode *tvp, struct componentname *fcnp, struct componentname *tcnp)
3146 cache_purge_negative(tdvp);
3151 * Move an entry from the provided source directory to the target
3152 * directory. Change the entry name as indicated.
3154 * IN: sdvp - Source directory containing the "old entry".
3155 * snm - Old entry name.
3156 * tdvp - Target directory to contain the "new entry".
3157 * tnm - New entry name.
3158 * cr - credentials of caller.
3159 * ct - caller context
3160 * flags - case flags
3162 * RETURN: 0 on success, error code on failure.
3165 * sdvp,tdvp - ctime|mtime updated
3169 zfs_rename_(vnode_t *sdvp, vnode_t **svpp, struct componentname *scnp,
3170 vnode_t *tdvp, vnode_t **tvpp, struct componentname *tcnp,
3171 cred_t *cr, int log)
3174 znode_t *sdzp, *tdzp, *szp, *tzp;
3175 zilog_t *zilog = NULL;
3177 const char *snm = scnp->cn_nameptr;
3178 const char *tnm = tcnp->cn_nameptr;
3180 bool want_seqc_end __maybe_unused = false;
3182 /* Reject renames across filesystems. */
3183 if ((*svpp)->v_mount != tdvp->v_mount ||
3184 ((*tvpp) != NULL && (*svpp)->v_mount != (*tvpp)->v_mount)) {
3185 error = SET_ERROR(EXDEV);
3189 if (zfsctl_is_node(tdvp)) {
3190 error = SET_ERROR(EXDEV);
3195 * Lock all four vnodes to ensure safety and semantics of renaming.
3197 error = zfs_rename_relock(sdvp, svpp, tdvp, tvpp, scnp, tcnp);
3199 /* no vnodes are locked in the case of error here */
3205 zfsvfs = tdzp->z_zfsvfs;
3206 zilog = zfsvfs->z_log;
3209 * After we re-enter ZFS_ENTER() we will have to revalidate all
3214 if (zfsvfs->z_utf8 && u8_validate(tnm,
3215 strlen(tnm), NULL, U8_VALIDATE_ENTIRE, &error) < 0) {
3216 error = SET_ERROR(EILSEQ);
3220 /* If source and target are the same file, there is nothing to do. */
3221 if ((*svpp) == (*tvpp)) {
3226 if (((*svpp)->v_type == VDIR && (*svpp)->v_mountedhere != NULL) ||
3227 ((*tvpp) != NULL && (*tvpp)->v_type == VDIR &&
3228 (*tvpp)->v_mountedhere != NULL)) {
3229 error = SET_ERROR(EXDEV);
3234 * We can not use ZFS_VERIFY_ZP() here because it could directly return
3235 * bypassing the cleanup code in the case of an error.
3237 if (tdzp->z_sa_hdl == NULL || sdzp->z_sa_hdl == NULL) {
3238 error = SET_ERROR(EIO);
3243 tzp = *tvpp == NULL ? NULL : VTOZ(*tvpp);
3244 if (szp->z_sa_hdl == NULL || (tzp != NULL && tzp->z_sa_hdl == NULL)) {
3245 error = SET_ERROR(EIO);
3250 * This is to prevent the creation of links into attribute space
3251 * by renaming a linked file into/outof an attribute directory.
3252 * See the comment in zfs_link() for why this is considered bad.
3254 if ((tdzp->z_pflags & ZFS_XATTR) != (sdzp->z_pflags & ZFS_XATTR)) {
3255 error = SET_ERROR(EINVAL);
3260 * If we are using project inheritance, means if the directory has
3261 * ZFS_PROJINHERIT set, then its descendant directories will inherit
3262 * not only the project ID, but also the ZFS_PROJINHERIT flag. Under
3263 * such case, we only allow renames into our tree when the project
3266 if (tdzp->z_pflags & ZFS_PROJINHERIT &&
3267 tdzp->z_projid != szp->z_projid) {
3268 error = SET_ERROR(EXDEV);
3273 * Must have write access at the source to remove the old entry
3274 * and write access at the target to create the new entry.
3275 * Note that if target and source are the same, this can be
3276 * done in a single check.
3278 if ((error = zfs_zaccess_rename(sdzp, szp, tdzp, tzp, cr)))
3281 if ((*svpp)->v_type == VDIR) {
3283 * Avoid ".", "..", and aliases of "." for obvious reasons.
3285 if ((scnp->cn_namelen == 1 && scnp->cn_nameptr[0] == '.') ||
3287 (scnp->cn_flags | tcnp->cn_flags) & ISDOTDOT) {
3293 * Check to make sure rename is valid.
3294 * Can't do a move like this: /usr/a/b to /usr/a/b/c/d
3296 if ((error = zfs_rename_check(szp, sdzp, tdzp)))
3301 * Does target exist?
3305 * Source and target must be the same type.
3307 if ((*svpp)->v_type == VDIR) {
3308 if ((*tvpp)->v_type != VDIR) {
3309 error = SET_ERROR(ENOTDIR);
3317 if ((*tvpp)->v_type == VDIR) {
3318 error = SET_ERROR(EISDIR);
3324 vn_seqc_write_begin(*svpp);
3325 vn_seqc_write_begin(sdvp);
3327 vn_seqc_write_begin(*tvpp);
3329 vn_seqc_write_begin(tdvp);
3330 #if __FreeBSD_version >= 1300102
3331 want_seqc_end = true;
3333 vnevent_rename_src(*svpp, sdvp, scnp->cn_nameptr, ct);
3335 vnevent_rename_dest(*tvpp, tdvp, tnm, ct);
3338 * notify the target directory if it is not the same
3339 * as source directory.
3342 vnevent_rename_dest_dir(tdvp, ct);
3345 tx = dmu_tx_create(zfsvfs->z_os);
3346 dmu_tx_hold_sa(tx, szp->z_sa_hdl, B_FALSE);
3347 dmu_tx_hold_sa(tx, sdzp->z_sa_hdl, B_FALSE);
3348 dmu_tx_hold_zap(tx, sdzp->z_id, FALSE, snm);
3349 dmu_tx_hold_zap(tx, tdzp->z_id, TRUE, tnm);
3351 dmu_tx_hold_sa(tx, tdzp->z_sa_hdl, B_FALSE);
3352 zfs_sa_upgrade_txholds(tx, tdzp);
3355 dmu_tx_hold_sa(tx, tzp->z_sa_hdl, B_FALSE);
3356 zfs_sa_upgrade_txholds(tx, tzp);
3359 zfs_sa_upgrade_txholds(tx, szp);
3360 dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL);
3361 error = dmu_tx_assign(tx, TXG_WAIT);
3368 if (tzp) /* Attempt to remove the existing target */
3369 error = zfs_link_destroy(tdzp, tnm, tzp, tx, 0, NULL);
3372 error = zfs_link_create(tdzp, tnm, szp, tx, ZRENAMING);
3374 szp->z_pflags |= ZFS_AV_MODIFIED;
3376 error = sa_update(szp->z_sa_hdl, SA_ZPL_FLAGS(zfsvfs),
3377 (void *)&szp->z_pflags, sizeof (uint64_t), tx);
3380 error = zfs_link_destroy(sdzp, snm, szp, tx, ZRENAMING,
3383 zfs_log_rename(zilog, tx, TX_RENAME, sdzp,
3384 snm, tdzp, tnm, szp);
3387 * Update path information for the target vnode
3389 vn_renamepath(tdvp, *svpp, tnm, strlen(tnm));
3392 * At this point, we have successfully created
3393 * the target name, but have failed to remove
3394 * the source name. Since the create was done
3395 * with the ZRENAMING flag, there are
3396 * complications; for one, the link count is
3397 * wrong. The easiest way to deal with this
3398 * is to remove the newly created target, and
3399 * return the original error. This must
3400 * succeed; fortunately, it is very unlikely to
3401 * fail, since we just created it.
3403 VERIFY3U(zfs_link_destroy(tdzp, tnm, szp, tx,
3404 ZRENAMING, NULL), ==, 0);
3408 cache_vop_rename(sdvp, *svpp, tdvp, *tvpp, scnp, tcnp);
3414 unlockout: /* all 4 vnodes are locked, ZFS_ENTER called */
3416 if (want_seqc_end) {
3417 vn_seqc_write_end(*svpp);
3418 vn_seqc_write_end(sdvp);
3420 vn_seqc_write_end(*tvpp);
3422 vn_seqc_write_end(tdvp);
3423 want_seqc_end = false;
3428 out: /* original two vnodes are locked */
3429 MPASS(!want_seqc_end);
3430 if (error == 0 && zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS)
3431 zil_commit(zilog, 0);
3441 zfs_rename(znode_t *sdzp, const char *sname, znode_t *tdzp, const char *tname,
3442 cred_t *cr, int flags)
3444 struct componentname scn, tcn;
3445 vnode_t *sdvp, *tdvp;
3452 error = zfs_lookup_internal(sdzp, sname, &svp, &scn, DELETE);
3453 if (sdzp->z_zfsvfs->z_replay == B_FALSE)
3459 vn_lock(tdvp, LK_EXCLUSIVE | LK_RETRY);
3460 error = zfs_lookup_internal(tdzp, tname, &tvp, &tcn, RENAME);
3461 if (error == EJUSTRETURN)
3463 else if (error != 0) {
3468 error = zfs_rename_(sdvp, &svp, &scn, tdvp, &tvp, &tcn, cr, 0);
3479 * Insert the indicated symbolic reference entry into the directory.
3481 * IN: dvp - Directory to contain new symbolic link.
3482 * link - Name for new symlink entry.
3483 * vap - Attributes of new entry.
3484 * cr - credentials of caller.
3485 * ct - caller context
3486 * flags - case flags
3488 * RETURN: 0 on success, error code on failure.
3491 * dvp - ctime|mtime updated
3495 zfs_symlink(znode_t *dzp, const char *name, vattr_t *vap,
3496 const char *link, znode_t **zpp, cred_t *cr, int flags)
3500 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
3502 uint64_t len = strlen(link);
3504 zfs_acl_ids_t acl_ids;
3505 boolean_t fuid_dirtied;
3506 uint64_t txtype = TX_SYMLINK;
3508 ASSERT(vap->va_type == VLNK);
3512 zilog = zfsvfs->z_log;
3514 if (zfsvfs->z_utf8 && u8_validate(name, strlen(name),
3515 NULL, U8_VALIDATE_ENTIRE, &error) < 0) {
3517 return (SET_ERROR(EILSEQ));
3520 if (len > MAXPATHLEN) {
3522 return (SET_ERROR(ENAMETOOLONG));
3525 if ((error = zfs_acl_ids_create(dzp, 0,
3526 vap, cr, NULL, &acl_ids)) != 0) {
3532 * Attempt to lock directory; fail if entry already exists.
3534 error = zfs_dirent_lookup(dzp, name, &zp, ZNEW);
3536 zfs_acl_ids_free(&acl_ids);
3541 if ((error = zfs_zaccess(dzp, ACE_ADD_FILE, 0, B_FALSE, cr))) {
3542 zfs_acl_ids_free(&acl_ids);
3547 if (zfs_acl_ids_overquota(zfsvfs, &acl_ids,
3549 zfs_acl_ids_free(&acl_ids);
3551 return (SET_ERROR(EDQUOT));
3554 getnewvnode_reserve_();
3555 tx = dmu_tx_create(zfsvfs->z_os);
3556 fuid_dirtied = zfsvfs->z_fuid_dirty;
3557 dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, MAX(1, len));
3558 dmu_tx_hold_zap(tx, dzp->z_id, TRUE, name);
3559 dmu_tx_hold_sa_create(tx, acl_ids.z_aclp->z_acl_bytes +
3560 ZFS_SA_BASE_ATTR_SIZE + len);
3561 dmu_tx_hold_sa(tx, dzp->z_sa_hdl, B_FALSE);
3562 if (!zfsvfs->z_use_sa && acl_ids.z_aclp->z_acl_bytes > ZFS_ACE_SPACE) {
3563 dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0,
3564 acl_ids.z_aclp->z_acl_bytes);
3567 zfs_fuid_txhold(zfsvfs, tx);
3568 error = dmu_tx_assign(tx, TXG_WAIT);
3570 zfs_acl_ids_free(&acl_ids);
3572 getnewvnode_drop_reserve();
3578 * Create a new object for the symlink.
3579 * for version 4 ZPL datsets the symlink will be an SA attribute
3581 zfs_mknode(dzp, vap, tx, cr, 0, &zp, &acl_ids);
3584 zfs_fuid_sync(zfsvfs, tx);
3587 error = sa_update(zp->z_sa_hdl, SA_ZPL_SYMLINK(zfsvfs),
3588 __DECONST(void *, link), len, tx);
3590 zfs_sa_symlink(zp, __DECONST(char *, link), len, tx);
3593 (void) sa_update(zp->z_sa_hdl, SA_ZPL_SIZE(zfsvfs),
3594 &zp->z_size, sizeof (zp->z_size), tx);
3596 * Insert the new object into the directory.
3598 (void) zfs_link_create(dzp, name, zp, tx, ZNEW);
3600 zfs_log_symlink(zilog, tx, txtype, dzp, zp, name, link);
3603 zfs_acl_ids_free(&acl_ids);
3607 getnewvnode_drop_reserve();
3609 if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS)
3610 zil_commit(zilog, 0);
3617 * Return, in the buffer contained in the provided uio structure,
3618 * the symbolic path referred to by vp.
3620 * IN: vp - vnode of symbolic link.
3621 * uio - structure to contain the link path.
3622 * cr - credentials of caller.
3623 * ct - caller context
3625 * OUT: uio - structure containing the link path.
3627 * RETURN: 0 on success, error code on failure.
3630 * vp - atime updated
3634 zfs_readlink(vnode_t *vp, uio_t *uio, cred_t *cr, caller_context_t *ct)
3636 znode_t *zp = VTOZ(vp);
3637 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
3644 error = sa_lookup_uio(zp->z_sa_hdl,
3645 SA_ZPL_SYMLINK(zfsvfs), uio);
3647 error = zfs_sa_readlink(zp, uio);
3649 ZFS_ACCESSTIME_STAMP(zfsvfs, zp);
3656 * Insert a new entry into directory tdvp referencing svp.
3658 * IN: tdvp - Directory to contain new entry.
3659 * svp - vnode of new entry.
3660 * name - name of new entry.
3661 * cr - credentials of caller.
3663 * RETURN: 0 on success, error code on failure.
3666 * tdvp - ctime|mtime updated
3667 * svp - ctime updated
3671 zfs_link(znode_t *tdzp, znode_t *szp, const char *name, cred_t *cr,
3675 zfsvfs_t *zfsvfs = tdzp->z_zfsvfs;
3682 ASSERT(ZTOV(tdzp)->v_type == VDIR);
3685 ZFS_VERIFY_ZP(tdzp);
3686 zilog = zfsvfs->z_log;
3689 * POSIX dictates that we return EPERM here.
3690 * Better choices include ENOTSUP or EISDIR.
3692 if (ZTOV(szp)->v_type == VDIR) {
3694 return (SET_ERROR(EPERM));
3700 * If we are using project inheritance, means if the directory has
3701 * ZFS_PROJINHERIT set, then its descendant directories will inherit
3702 * not only the project ID, but also the ZFS_PROJINHERIT flag. Under
3703 * such case, we only allow hard link creation in our tree when the
3704 * project IDs are the same.
3706 if (tdzp->z_pflags & ZFS_PROJINHERIT &&
3707 tdzp->z_projid != szp->z_projid) {
3709 return (SET_ERROR(EXDEV));
3712 if (szp->z_pflags & (ZFS_APPENDONLY |
3713 ZFS_IMMUTABLE | ZFS_READONLY)) {
3715 return (SET_ERROR(EPERM));
3718 /* Prevent links to .zfs/shares files */
3720 if ((error = sa_lookup(szp->z_sa_hdl, SA_ZPL_PARENT(zfsvfs),
3721 &parent, sizeof (uint64_t))) != 0) {
3725 if (parent == zfsvfs->z_shares_dir) {
3727 return (SET_ERROR(EPERM));
3730 if (zfsvfs->z_utf8 && u8_validate(name,
3731 strlen(name), NULL, U8_VALIDATE_ENTIRE, &error) < 0) {
3733 return (SET_ERROR(EILSEQ));
3737 * We do not support links between attributes and non-attributes
3738 * because of the potential security risk of creating links
3739 * into "normal" file space in order to circumvent restrictions
3740 * imposed in attribute space.
3742 if ((szp->z_pflags & ZFS_XATTR) != (tdzp->z_pflags & ZFS_XATTR)) {
3744 return (SET_ERROR(EINVAL));
3748 owner = zfs_fuid_map_id(zfsvfs, szp->z_uid, cr, ZFS_OWNER);
3749 if (owner != crgetuid(cr) && secpolicy_basic_link(ZTOV(szp), cr) != 0) {
3751 return (SET_ERROR(EPERM));
3754 if ((error = zfs_zaccess(tdzp, ACE_ADD_FILE, 0, B_FALSE, cr))) {
3760 * Attempt to lock directory; fail if entry already exists.
3762 error = zfs_dirent_lookup(tdzp, name, &tzp, ZNEW);
3768 tx = dmu_tx_create(zfsvfs->z_os);
3769 dmu_tx_hold_sa(tx, szp->z_sa_hdl, B_FALSE);
3770 dmu_tx_hold_zap(tx, tdzp->z_id, TRUE, name);
3771 zfs_sa_upgrade_txholds(tx, szp);
3772 zfs_sa_upgrade_txholds(tx, tdzp);
3773 error = dmu_tx_assign(tx, TXG_WAIT);
3780 error = zfs_link_create(tdzp, name, szp, tx, 0);
3783 uint64_t txtype = TX_LINK;
3784 zfs_log_link(zilog, tx, txtype, tdzp, szp, name);
3790 vnevent_link(ZTOV(szp), ct);
3793 if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS)
3794 zil_commit(zilog, 0);
3801 * Free or allocate space in a file. Currently, this function only
3802 * supports the `F_FREESP' command. However, this command is somewhat
3803 * misnamed, as its functionality includes the ability to allocate as
3804 * well as free space.
3806 * IN: ip - inode of file to free data in.
3807 * cmd - action to take (only F_FREESP supported).
3808 * bfp - section of file to free/alloc.
3809 * flag - current file open mode flags.
3810 * offset - current file offset.
3811 * cr - credentials of caller.
3813 * RETURN: 0 on success, error code on failure.
3816 * ip - ctime|mtime updated
3820 zfs_space(znode_t *zp, int cmd, flock64_t *bfp, int flag,
3821 offset_t offset, cred_t *cr)
3823 zfsvfs_t *zfsvfs = ZTOZSB(zp);
3830 if (cmd != F_FREESP) {
3832 return (SET_ERROR(EINVAL));
3836 * Callers might not be able to detect properly that we are read-only,
3837 * so check it explicitly here.
3839 if (zfs_is_readonly(zfsvfs)) {
3841 return (SET_ERROR(EROFS));
3844 if (bfp->l_len < 0) {
3846 return (SET_ERROR(EINVAL));
3850 * Permissions aren't checked on Solaris because on this OS
3851 * zfs_space() can only be called with an opened file handle.
3852 * On Linux we can get here through truncate_range() which
3853 * operates directly on inodes, so we need to check access rights.
3855 if ((error = zfs_zaccess(zp, ACE_WRITE_DATA, 0, B_FALSE, cr))) {
3861 len = bfp->l_len; /* 0 means from off to end of file */
3863 error = zfs_freesp(zp, off, len, flag, TRUE);
3871 zfs_inactive(vnode_t *vp, cred_t *cr, caller_context_t *ct)
3873 znode_t *zp = VTOZ(vp);
3874 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
3877 ZFS_TEARDOWN_INACTIVE_ENTER_READ(zfsvfs);
3878 if (zp->z_sa_hdl == NULL) {
3880 * The fs has been unmounted, or we did a
3881 * suspend/resume and this file no longer exists.
3883 ZFS_TEARDOWN_INACTIVE_EXIT_READ(zfsvfs);
3888 if (zp->z_unlinked) {
3890 * Fast path to recycle a vnode of a removed file.
3892 ZFS_TEARDOWN_INACTIVE_EXIT_READ(zfsvfs);
3897 if (zp->z_atime_dirty && zp->z_unlinked == 0) {
3898 dmu_tx_t *tx = dmu_tx_create(zfsvfs->z_os);
3900 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
3901 zfs_sa_upgrade_txholds(tx, zp);
3902 error = dmu_tx_assign(tx, TXG_WAIT);
3906 (void) sa_update(zp->z_sa_hdl, SA_ZPL_ATIME(zfsvfs),
3907 (void *)&zp->z_atime, sizeof (zp->z_atime), tx);
3908 zp->z_atime_dirty = 0;
3912 ZFS_TEARDOWN_INACTIVE_EXIT_READ(zfsvfs);
3916 CTASSERT(sizeof (struct zfid_short) <= sizeof (struct fid));
3917 CTASSERT(sizeof (struct zfid_long) <= sizeof (struct fid));
3921 zfs_fid(vnode_t *vp, fid_t *fidp, caller_context_t *ct)
3923 znode_t *zp = VTOZ(vp);
3924 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
3927 uint64_t object = zp->z_id;
3934 if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_GEN(zfsvfs),
3935 &gen64, sizeof (uint64_t))) != 0) {
3940 gen = (uint32_t)gen64;
3942 size = (zfsvfs->z_parent != zfsvfs) ? LONG_FID_LEN : SHORT_FID_LEN;
3943 fidp->fid_len = size;
3945 zfid = (zfid_short_t *)fidp;
3947 zfid->zf_len = size;
3949 for (i = 0; i < sizeof (zfid->zf_object); i++)
3950 zfid->zf_object[i] = (uint8_t)(object >> (8 * i));
3952 /* Must have a non-zero generation number to distinguish from .zfs */
3955 for (i = 0; i < sizeof (zfid->zf_gen); i++)
3956 zfid->zf_gen[i] = (uint8_t)(gen >> (8 * i));
3958 if (size == LONG_FID_LEN) {
3959 uint64_t objsetid = dmu_objset_id(zfsvfs->z_os);
3962 zlfid = (zfid_long_t *)fidp;
3964 for (i = 0; i < sizeof (zlfid->zf_setid); i++)
3965 zlfid->zf_setid[i] = (uint8_t)(objsetid >> (8 * i));
3967 /* XXX - this should be the generation number for the objset */
3968 for (i = 0; i < sizeof (zlfid->zf_setgen); i++)
3969 zlfid->zf_setgen[i] = 0;
3977 zfs_pathconf(vnode_t *vp, int cmd, ulong_t *valp, cred_t *cr,
3978 caller_context_t *ct)
3985 *valp = MIN(LONG_MAX, ZFS_LINK_MAX);
3988 case _PC_FILESIZEBITS:
3991 case _PC_MIN_HOLE_SIZE:
3992 *valp = (int)SPA_MINBLOCKSIZE;
3994 case _PC_ACL_EXTENDED:
3995 #if 0 /* POSIX ACLs are not implemented for ZFS on FreeBSD yet. */
3997 zfsvfs = zp->z_zfsvfs;
4000 *valp = zfsvfs->z_acl_type == ZFSACLTYPE_POSIX ? 1 : 0;
4009 zfsvfs = zp->z_zfsvfs;
4012 *valp = zfsvfs->z_acl_type == ZFS_ACLTYPE_NFSV4 ? 1 : 0;
4016 case _PC_ACL_PATH_MAX:
4017 *valp = ACL_MAX_ENTRIES;
4021 return (EOPNOTSUPP);
4026 zfs_getpages(struct vnode *vp, vm_page_t *ma, int count, int *rbehind,
4029 znode_t *zp = VTOZ(vp);
4030 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
4031 objset_t *os = zp->z_zfsvfs->z_os;
4032 zfs_locked_range_t *lr;
4034 off_t start, end, obj_size;
4036 int pgsin_b, pgsin_a;
4042 start = IDX_TO_OFF(ma[0]->pindex);
4043 end = IDX_TO_OFF(ma[count - 1]->pindex + 1);
4046 * Lock a range covering all required and optional pages.
4047 * Note that we need to handle the case of the block size growing.
4050 blksz = zp->z_blksz;
4051 lr = zfs_rangelock_tryenter(&zp->z_rangelock,
4052 rounddown(start, blksz),
4053 roundup(end, blksz) - rounddown(start, blksz), RL_READER);
4055 if (rahead != NULL) {
4059 if (rbehind != NULL) {
4065 if (blksz == zp->z_blksz)
4067 zfs_rangelock_exit(lr);
4070 object = ma[0]->object;
4071 zfs_vmobject_wlock(object);
4072 obj_size = object->un_pager.vnp.vnp_size;
4073 zfs_vmobject_wunlock(object);
4074 if (IDX_TO_OFF(ma[count - 1]->pindex) >= obj_size) {
4076 zfs_rangelock_exit(lr);
4078 return (zfs_vm_pagerret_bad);
4082 if (rbehind != NULL) {
4083 pgsin_b = OFF_TO_IDX(start - rounddown(start, blksz));
4084 pgsin_b = MIN(*rbehind, pgsin_b);
4088 if (rahead != NULL) {
4089 pgsin_a = OFF_TO_IDX(roundup(end, blksz) - end);
4090 if (end + IDX_TO_OFF(pgsin_a) >= obj_size)
4091 pgsin_a = OFF_TO_IDX(round_page(obj_size) - end);
4092 pgsin_a = MIN(*rahead, pgsin_a);
4096 * NB: we need to pass the exact byte size of the data that we expect
4097 * to read after accounting for the file size. This is required because
4098 * ZFS will panic if we request DMU to read beyond the end of the last
4101 error = dmu_read_pages(os, zp->z_id, ma, count, &pgsin_b, &pgsin_a,
4102 MIN(end, obj_size) - (end - PAGE_SIZE));
4105 zfs_rangelock_exit(lr);
4106 ZFS_ACCESSTIME_STAMP(zfsvfs, zp);
4110 return (zfs_vm_pagerret_error);
4112 VM_CNT_INC(v_vnodein);
4113 VM_CNT_ADD(v_vnodepgsin, count + pgsin_b + pgsin_a);
4114 if (rbehind != NULL)
4118 return (zfs_vm_pagerret_ok);
4121 #ifndef _SYS_SYSPROTO_H_
4122 struct vop_getpages_args {
4132 zfs_freebsd_getpages(struct vop_getpages_args *ap)
4135 return (zfs_getpages(ap->a_vp, ap->a_m, ap->a_count, ap->a_rbehind,
4140 zfs_putpages(struct vnode *vp, vm_page_t *ma, size_t len, int flags,
4143 znode_t *zp = VTOZ(vp);
4144 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
4145 zfs_locked_range_t *lr;
4153 vm_ooffset_t lo_off;
4164 object = vp->v_object;
4168 KASSERT(ma[0]->object == object, ("mismatching object"));
4169 KASSERT(len > 0 && (len & PAGE_MASK) == 0, ("unexpected length"));
4171 for (i = 0; i < pcount; i++)
4172 rtvals[i] = zfs_vm_pagerret_error;
4174 off = IDX_TO_OFF(ma[0]->pindex);
4175 blksz = zp->z_blksz;
4176 lo_off = rounddown(off, blksz);
4177 lo_len = roundup(len + (off - lo_off), blksz);
4178 lr = zfs_rangelock_enter(&zp->z_rangelock, lo_off, lo_len, RL_WRITER);
4180 zfs_vmobject_wlock(object);
4181 if (len + off > object->un_pager.vnp.vnp_size) {
4182 if (object->un_pager.vnp.vnp_size > off) {
4185 len = object->un_pager.vnp.vnp_size - off;
4187 if ((pgoff = (int)len & PAGE_MASK) != 0) {
4189 * If the object is locked and the following
4190 * conditions hold, then the page's dirty
4191 * field cannot be concurrently changed by a
4195 vm_page_assert_sbusied(m);
4196 KASSERT(!pmap_page_is_write_mapped(m),
4197 ("zfs_putpages: page %p is not read-only",
4199 vm_page_clear_dirty(m, pgoff, PAGE_SIZE -
4206 if (ncount < pcount) {
4207 for (i = ncount; i < pcount; i++) {
4208 rtvals[i] = zfs_vm_pagerret_bad;
4212 zfs_vmobject_wunlock(object);
4217 if (zfs_id_overblockquota(zfsvfs, DMU_USERUSED_OBJECT, zp->z_uid) ||
4218 zfs_id_overblockquota(zfsvfs, DMU_GROUPUSED_OBJECT, zp->z_gid) ||
4219 (zp->z_projid != ZFS_DEFAULT_PROJID &&
4220 zfs_id_overblockquota(zfsvfs, DMU_PROJECTUSED_OBJECT,
4225 tx = dmu_tx_create(zfsvfs->z_os);
4226 dmu_tx_hold_write(tx, zp->z_id, off, len);
4228 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
4229 zfs_sa_upgrade_txholds(tx, zp);
4230 err = dmu_tx_assign(tx, TXG_WAIT);
4236 if (zp->z_blksz < PAGE_SIZE) {
4237 for (i = 0; len > 0; off += tocopy, len -= tocopy, i++) {
4238 tocopy = len > PAGE_SIZE ? PAGE_SIZE : len;
4239 va = zfs_map_page(ma[i], &sf);
4240 dmu_write(zfsvfs->z_os, zp->z_id, off, tocopy, va, tx);
4244 err = dmu_write_pages(zfsvfs->z_os, zp->z_id, off, len, ma, tx);
4248 uint64_t mtime[2], ctime[2];
4249 sa_bulk_attr_t bulk[3];
4252 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL,
4254 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL,
4256 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL,
4258 zfs_tstamp_update_setup(zp, CONTENT_MODIFIED, mtime, ctime);
4259 err = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx);
4262 * XXX we should be passing a callback to undirty
4263 * but that would make the locking messier
4265 zfs_log_write(zfsvfs->z_log, tx, TX_WRITE, zp, off,
4266 len, 0, NULL, NULL);
4268 zfs_vmobject_wlock(object);
4269 for (i = 0; i < ncount; i++) {
4270 rtvals[i] = zfs_vm_pagerret_ok;
4271 vm_page_undirty(ma[i]);
4273 zfs_vmobject_wunlock(object);
4274 VM_CNT_INC(v_vnodeout);
4275 VM_CNT_ADD(v_vnodepgsout, ncount);
4280 zfs_rangelock_exit(lr);
4281 if ((flags & (zfs_vm_pagerput_sync | zfs_vm_pagerput_inval)) != 0 ||
4282 zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS)
4283 zil_commit(zfsvfs->z_log, zp->z_id);
4288 #ifndef _SYS_SYSPROTO_H_
4289 struct vop_putpages_args {
4299 zfs_freebsd_putpages(struct vop_putpages_args *ap)
4302 return (zfs_putpages(ap->a_vp, ap->a_m, ap->a_count, ap->a_sync,
4306 #ifndef _SYS_SYSPROTO_H_
4307 struct vop_bmap_args {
4310 struct bufobj **a_bop;
4318 zfs_freebsd_bmap(struct vop_bmap_args *ap)
4321 if (ap->a_bop != NULL)
4322 *ap->a_bop = &ap->a_vp->v_bufobj;
4323 if (ap->a_bnp != NULL)
4324 *ap->a_bnp = ap->a_bn;
4325 if (ap->a_runp != NULL)
4327 if (ap->a_runb != NULL)
4333 #ifndef _SYS_SYSPROTO_H_
4334 struct vop_open_args {
4337 struct ucred *a_cred;
4338 struct thread *a_td;
4343 zfs_freebsd_open(struct vop_open_args *ap)
4345 vnode_t *vp = ap->a_vp;
4346 znode_t *zp = VTOZ(vp);
4349 error = zfs_open(&vp, ap->a_mode, ap->a_cred);
4351 vnode_create_vobject(vp, zp->z_size, ap->a_td);
4355 #ifndef _SYS_SYSPROTO_H_
4356 struct vop_close_args {
4359 struct ucred *a_cred;
4360 struct thread *a_td;
4365 zfs_freebsd_close(struct vop_close_args *ap)
4368 return (zfs_close(ap->a_vp, ap->a_fflag, 1, 0, ap->a_cred));
4371 #ifndef _SYS_SYSPROTO_H_
4372 struct vop_ioctl_args {
4383 zfs_freebsd_ioctl(struct vop_ioctl_args *ap)
4386 return (zfs_ioctl(ap->a_vp, ap->a_command, (intptr_t)ap->a_data,
4387 ap->a_fflag, ap->a_cred, NULL));
4391 ioflags(int ioflags)
4395 if (ioflags & IO_APPEND)
4397 if (ioflags & IO_NDELAY)
4399 if (ioflags & IO_SYNC)
4400 flags |= (FSYNC | FDSYNC | FRSYNC);
4405 #ifndef _SYS_SYSPROTO_H_
4406 struct vop_read_args {
4410 struct ucred *a_cred;
4415 zfs_freebsd_read(struct vop_read_args *ap)
4418 return (zfs_read(VTOZ(ap->a_vp), ap->a_uio, ioflags(ap->a_ioflag),
4422 #ifndef _SYS_SYSPROTO_H_
4423 struct vop_write_args {
4427 struct ucred *a_cred;
4432 zfs_freebsd_write(struct vop_write_args *ap)
4435 return (zfs_write(VTOZ(ap->a_vp), ap->a_uio, ioflags(ap->a_ioflag),
4439 #if __FreeBSD_version >= 1300102
4441 * VOP_FPLOOKUP_VEXEC routines are subject to special circumstances, see
4442 * the comment above cache_fplookup for details.
4445 zfs_freebsd_fplookup_vexec(struct vop_fplookup_vexec_args *v)
4453 if (__predict_false(zp == NULL))
4455 pflags = atomic_load_64(&zp->z_pflags);
4456 if (pflags & ZFS_AV_QUARANTINED)
4458 if (pflags & ZFS_XATTR)
4460 if ((pflags & ZFS_NO_EXECS_DENIED) == 0)
4466 #ifndef _SYS_SYSPROTO_H_
4467 struct vop_access_args {
4469 accmode_t a_accmode;
4470 struct ucred *a_cred;
4471 struct thread *a_td;
4476 zfs_freebsd_access(struct vop_access_args *ap)
4478 vnode_t *vp = ap->a_vp;
4479 znode_t *zp = VTOZ(vp);
4484 if (ap->a_accmode == VEXEC) {
4485 if (zfs_fastaccesschk_execute(zp, ap->a_cred) == 0)
4490 * ZFS itself only knowns about VREAD, VWRITE, VEXEC and VAPPEND,
4492 accmode = ap->a_accmode & (VREAD|VWRITE|VEXEC|VAPPEND);
4494 error = zfs_access(zp, accmode, 0, ap->a_cred);
4497 * VADMIN has to be handled by vaccess().
4500 accmode = ap->a_accmode & ~(VREAD|VWRITE|VEXEC|VAPPEND);
4502 #if __FreeBSD_version >= 1300105
4503 error = vaccess(vp->v_type, zp->z_mode, zp->z_uid,
4504 zp->z_gid, accmode, ap->a_cred);
4506 error = vaccess(vp->v_type, zp->z_mode, zp->z_uid,
4507 zp->z_gid, accmode, ap->a_cred, NULL);
4513 * For VEXEC, ensure that at least one execute bit is set for
4516 if (error == 0 && (ap->a_accmode & VEXEC) != 0 && vp->v_type != VDIR &&
4517 (zp->z_mode & (S_IXUSR | S_IXGRP | S_IXOTH)) == 0) {
4524 #ifndef _SYS_SYSPROTO_H_
4525 struct vop_lookup_args {
4526 struct vnode *a_dvp;
4527 struct vnode **a_vpp;
4528 struct componentname *a_cnp;
4533 zfs_freebsd_lookup(struct vop_lookup_args *ap, boolean_t cached)
4535 struct componentname *cnp = ap->a_cnp;
4536 char nm[NAME_MAX + 1];
4538 ASSERT(cnp->cn_namelen < sizeof (nm));
4539 strlcpy(nm, cnp->cn_nameptr, MIN(cnp->cn_namelen + 1, sizeof (nm)));
4541 return (zfs_lookup(ap->a_dvp, nm, ap->a_vpp, cnp, cnp->cn_nameiop,
4542 cnp->cn_cred, cnp->cn_thread, 0, cached));
4546 zfs_freebsd_cachedlookup(struct vop_cachedlookup_args *ap)
4549 return (zfs_freebsd_lookup((struct vop_lookup_args *)ap, B_TRUE));
4552 #ifndef _SYS_SYSPROTO_H_
4553 struct vop_lookup_args {
4554 struct vnode *a_dvp;
4555 struct vnode **a_vpp;
4556 struct componentname *a_cnp;
4561 zfs_cache_lookup(struct vop_lookup_args *ap)
4565 zfsvfs = ap->a_dvp->v_mount->mnt_data;
4566 if (zfsvfs->z_use_namecache)
4567 return (vfs_cache_lookup(ap));
4569 return (zfs_freebsd_lookup(ap, B_FALSE));
4572 #ifndef _SYS_SYSPROTO_H_
4573 struct vop_create_args {
4574 struct vnode *a_dvp;
4575 struct vnode **a_vpp;
4576 struct componentname *a_cnp;
4577 struct vattr *a_vap;
4582 zfs_freebsd_create(struct vop_create_args *ap)
4585 struct componentname *cnp = ap->a_cnp;
4586 vattr_t *vap = ap->a_vap;
4590 ASSERT(cnp->cn_flags & SAVENAME);
4592 vattr_init_mask(vap);
4593 mode = vap->va_mode & ALLPERMS;
4594 zfsvfs = ap->a_dvp->v_mount->mnt_data;
4597 rc = zfs_create(VTOZ(ap->a_dvp), cnp->cn_nameptr, vap, !EXCL, mode,
4598 &zp, cnp->cn_cred, 0 /* flag */, NULL /* vsecattr */);
4600 *ap->a_vpp = ZTOV(zp);
4601 if (zfsvfs->z_use_namecache &&
4602 rc == 0 && (cnp->cn_flags & MAKEENTRY) != 0)
4603 cache_enter(ap->a_dvp, *ap->a_vpp, cnp);
4608 #ifndef _SYS_SYSPROTO_H_
4609 struct vop_remove_args {
4610 struct vnode *a_dvp;
4612 struct componentname *a_cnp;
4617 zfs_freebsd_remove(struct vop_remove_args *ap)
4620 ASSERT(ap->a_cnp->cn_flags & SAVENAME);
4622 return (zfs_remove_(ap->a_dvp, ap->a_vp, ap->a_cnp->cn_nameptr,
4623 ap->a_cnp->cn_cred));
4626 #ifndef _SYS_SYSPROTO_H_
4627 struct vop_mkdir_args {
4628 struct vnode *a_dvp;
4629 struct vnode **a_vpp;
4630 struct componentname *a_cnp;
4631 struct vattr *a_vap;
4636 zfs_freebsd_mkdir(struct vop_mkdir_args *ap)
4638 vattr_t *vap = ap->a_vap;
4642 ASSERT(ap->a_cnp->cn_flags & SAVENAME);
4644 vattr_init_mask(vap);
4647 rc = zfs_mkdir(VTOZ(ap->a_dvp), ap->a_cnp->cn_nameptr, vap, &zp,
4648 ap->a_cnp->cn_cred, 0, NULL);
4651 *ap->a_vpp = ZTOV(zp);
4655 #ifndef _SYS_SYSPROTO_H_
4656 struct vop_rmdir_args {
4657 struct vnode *a_dvp;
4659 struct componentname *a_cnp;
4664 zfs_freebsd_rmdir(struct vop_rmdir_args *ap)
4666 struct componentname *cnp = ap->a_cnp;
4668 ASSERT(cnp->cn_flags & SAVENAME);
4670 return (zfs_rmdir_(ap->a_dvp, ap->a_vp, cnp->cn_nameptr, cnp->cn_cred));
4673 #ifndef _SYS_SYSPROTO_H_
4674 struct vop_readdir_args {
4677 struct ucred *a_cred;
4680 ulong_t **a_cookies;
4685 zfs_freebsd_readdir(struct vop_readdir_args *ap)
4688 return (zfs_readdir(ap->a_vp, ap->a_uio, ap->a_cred, ap->a_eofflag,
4689 ap->a_ncookies, ap->a_cookies));
4692 #ifndef _SYS_SYSPROTO_H_
4693 struct vop_fsync_args {
4696 struct thread *a_td;
4701 zfs_freebsd_fsync(struct vop_fsync_args *ap)
4705 return (zfs_fsync(VTOZ(ap->a_vp), 0, ap->a_td->td_ucred));
4708 #ifndef _SYS_SYSPROTO_H_
4709 struct vop_getattr_args {
4711 struct vattr *a_vap;
4712 struct ucred *a_cred;
4717 zfs_freebsd_getattr(struct vop_getattr_args *ap)
4719 vattr_t *vap = ap->a_vap;
4725 xvap.xva_vattr = *vap;
4726 xvap.xva_vattr.va_mask |= AT_XVATTR;
4728 /* Convert chflags into ZFS-type flags. */
4729 /* XXX: what about SF_SETTABLE?. */
4730 XVA_SET_REQ(&xvap, XAT_IMMUTABLE);
4731 XVA_SET_REQ(&xvap, XAT_APPENDONLY);
4732 XVA_SET_REQ(&xvap, XAT_NOUNLINK);
4733 XVA_SET_REQ(&xvap, XAT_NODUMP);
4734 XVA_SET_REQ(&xvap, XAT_READONLY);
4735 XVA_SET_REQ(&xvap, XAT_ARCHIVE);
4736 XVA_SET_REQ(&xvap, XAT_SYSTEM);
4737 XVA_SET_REQ(&xvap, XAT_HIDDEN);
4738 XVA_SET_REQ(&xvap, XAT_REPARSE);
4739 XVA_SET_REQ(&xvap, XAT_OFFLINE);
4740 XVA_SET_REQ(&xvap, XAT_SPARSE);
4742 error = zfs_getattr(ap->a_vp, (vattr_t *)&xvap, 0, ap->a_cred);
4746 /* Convert ZFS xattr into chflags. */
4747 #define FLAG_CHECK(fflag, xflag, xfield) do { \
4748 if (XVA_ISSET_RTN(&xvap, (xflag)) && (xfield) != 0) \
4749 fflags |= (fflag); \
4751 FLAG_CHECK(SF_IMMUTABLE, XAT_IMMUTABLE,
4752 xvap.xva_xoptattrs.xoa_immutable);
4753 FLAG_CHECK(SF_APPEND, XAT_APPENDONLY,
4754 xvap.xva_xoptattrs.xoa_appendonly);
4755 FLAG_CHECK(SF_NOUNLINK, XAT_NOUNLINK,
4756 xvap.xva_xoptattrs.xoa_nounlink);
4757 FLAG_CHECK(UF_ARCHIVE, XAT_ARCHIVE,
4758 xvap.xva_xoptattrs.xoa_archive);
4759 FLAG_CHECK(UF_NODUMP, XAT_NODUMP,
4760 xvap.xva_xoptattrs.xoa_nodump);
4761 FLAG_CHECK(UF_READONLY, XAT_READONLY,
4762 xvap.xva_xoptattrs.xoa_readonly);
4763 FLAG_CHECK(UF_SYSTEM, XAT_SYSTEM,
4764 xvap.xva_xoptattrs.xoa_system);
4765 FLAG_CHECK(UF_HIDDEN, XAT_HIDDEN,
4766 xvap.xva_xoptattrs.xoa_hidden);
4767 FLAG_CHECK(UF_REPARSE, XAT_REPARSE,
4768 xvap.xva_xoptattrs.xoa_reparse);
4769 FLAG_CHECK(UF_OFFLINE, XAT_OFFLINE,
4770 xvap.xva_xoptattrs.xoa_offline);
4771 FLAG_CHECK(UF_SPARSE, XAT_SPARSE,
4772 xvap.xva_xoptattrs.xoa_sparse);
4775 *vap = xvap.xva_vattr;
4776 vap->va_flags = fflags;
4780 #ifndef _SYS_SYSPROTO_H_
4781 struct vop_setattr_args {
4783 struct vattr *a_vap;
4784 struct ucred *a_cred;
4789 zfs_freebsd_setattr(struct vop_setattr_args *ap)
4791 vnode_t *vp = ap->a_vp;
4792 vattr_t *vap = ap->a_vap;
4793 cred_t *cred = ap->a_cred;
4798 vattr_init_mask(vap);
4799 vap->va_mask &= ~AT_NOSET;
4802 xvap.xva_vattr = *vap;
4804 zflags = VTOZ(vp)->z_pflags;
4806 if (vap->va_flags != VNOVAL) {
4807 zfsvfs_t *zfsvfs = VTOZ(vp)->z_zfsvfs;
4810 if (zfsvfs->z_use_fuids == B_FALSE)
4811 return (EOPNOTSUPP);
4813 fflags = vap->va_flags;
4816 * We need to figure out whether it makes sense to allow
4817 * UF_REPARSE through, since we don't really have other
4818 * facilities to handle reparse points and zfs_setattr()
4819 * doesn't currently allow setting that attribute anyway.
4821 if ((fflags & ~(SF_IMMUTABLE|SF_APPEND|SF_NOUNLINK|UF_ARCHIVE|
4822 UF_NODUMP|UF_SYSTEM|UF_HIDDEN|UF_READONLY|UF_REPARSE|
4823 UF_OFFLINE|UF_SPARSE)) != 0)
4824 return (EOPNOTSUPP);
4826 * Unprivileged processes are not permitted to unset system
4827 * flags, or modify flags if any system flags are set.
4828 * Privileged non-jail processes may not modify system flags
4829 * if securelevel > 0 and any existing system flags are set.
4830 * Privileged jail processes behave like privileged non-jail
4831 * processes if the PR_ALLOW_CHFLAGS permission bit is set;
4832 * otherwise, they behave like unprivileged processes.
4834 if (secpolicy_fs_owner(vp->v_mount, cred) == 0 ||
4835 spl_priv_check_cred(cred, PRIV_VFS_SYSFLAGS) == 0) {
4837 (ZFS_IMMUTABLE | ZFS_APPENDONLY | ZFS_NOUNLINK)) {
4838 error = securelevel_gt(cred, 0);
4844 * Callers may only modify the file flags on
4845 * objects they have VADMIN rights for.
4847 if ((error = VOP_ACCESS(vp, VADMIN, cred,
4851 (ZFS_IMMUTABLE | ZFS_APPENDONLY |
4856 (SF_IMMUTABLE | SF_APPEND | SF_NOUNLINK)) {
4861 #define FLAG_CHANGE(fflag, zflag, xflag, xfield) do { \
4862 if (((fflags & (fflag)) && !(zflags & (zflag))) || \
4863 ((zflags & (zflag)) && !(fflags & (fflag)))) { \
4864 XVA_SET_REQ(&xvap, (xflag)); \
4865 (xfield) = ((fflags & (fflag)) != 0); \
4868 /* Convert chflags into ZFS-type flags. */
4869 /* XXX: what about SF_SETTABLE?. */
4870 FLAG_CHANGE(SF_IMMUTABLE, ZFS_IMMUTABLE, XAT_IMMUTABLE,
4871 xvap.xva_xoptattrs.xoa_immutable);
4872 FLAG_CHANGE(SF_APPEND, ZFS_APPENDONLY, XAT_APPENDONLY,
4873 xvap.xva_xoptattrs.xoa_appendonly);
4874 FLAG_CHANGE(SF_NOUNLINK, ZFS_NOUNLINK, XAT_NOUNLINK,
4875 xvap.xva_xoptattrs.xoa_nounlink);
4876 FLAG_CHANGE(UF_ARCHIVE, ZFS_ARCHIVE, XAT_ARCHIVE,
4877 xvap.xva_xoptattrs.xoa_archive);
4878 FLAG_CHANGE(UF_NODUMP, ZFS_NODUMP, XAT_NODUMP,
4879 xvap.xva_xoptattrs.xoa_nodump);
4880 FLAG_CHANGE(UF_READONLY, ZFS_READONLY, XAT_READONLY,
4881 xvap.xva_xoptattrs.xoa_readonly);
4882 FLAG_CHANGE(UF_SYSTEM, ZFS_SYSTEM, XAT_SYSTEM,
4883 xvap.xva_xoptattrs.xoa_system);
4884 FLAG_CHANGE(UF_HIDDEN, ZFS_HIDDEN, XAT_HIDDEN,
4885 xvap.xva_xoptattrs.xoa_hidden);
4886 FLAG_CHANGE(UF_REPARSE, ZFS_REPARSE, XAT_REPARSE,
4887 xvap.xva_xoptattrs.xoa_reparse);
4888 FLAG_CHANGE(UF_OFFLINE, ZFS_OFFLINE, XAT_OFFLINE,
4889 xvap.xva_xoptattrs.xoa_offline);
4890 FLAG_CHANGE(UF_SPARSE, ZFS_SPARSE, XAT_SPARSE,
4891 xvap.xva_xoptattrs.xoa_sparse);
4894 if (vap->va_birthtime.tv_sec != VNOVAL) {
4895 xvap.xva_vattr.va_mask |= AT_XVATTR;
4896 XVA_SET_REQ(&xvap, XAT_CREATETIME);
4898 return (zfs_setattr(VTOZ(vp), (vattr_t *)&xvap, 0, cred));
4901 #ifndef _SYS_SYSPROTO_H_
4902 struct vop_rename_args {
4903 struct vnode *a_fdvp;
4904 struct vnode *a_fvp;
4905 struct componentname *a_fcnp;
4906 struct vnode *a_tdvp;
4907 struct vnode *a_tvp;
4908 struct componentname *a_tcnp;
4913 zfs_freebsd_rename(struct vop_rename_args *ap)
4915 vnode_t *fdvp = ap->a_fdvp;
4916 vnode_t *fvp = ap->a_fvp;
4917 vnode_t *tdvp = ap->a_tdvp;
4918 vnode_t *tvp = ap->a_tvp;
4921 ASSERT(ap->a_fcnp->cn_flags & (SAVENAME|SAVESTART));
4922 ASSERT(ap->a_tcnp->cn_flags & (SAVENAME|SAVESTART));
4924 error = zfs_rename_(fdvp, &fvp, ap->a_fcnp, tdvp, &tvp,
4925 ap->a_tcnp, ap->a_fcnp->cn_cred, 1);
4936 #ifndef _SYS_SYSPROTO_H_
4937 struct vop_symlink_args {
4938 struct vnode *a_dvp;
4939 struct vnode **a_vpp;
4940 struct componentname *a_cnp;
4941 struct vattr *a_vap;
4947 zfs_freebsd_symlink(struct vop_symlink_args *ap)
4949 struct componentname *cnp = ap->a_cnp;
4950 vattr_t *vap = ap->a_vap;
4954 ASSERT(cnp->cn_flags & SAVENAME);
4956 vap->va_type = VLNK; /* FreeBSD: Syscall only sets va_mode. */
4957 vattr_init_mask(vap);
4960 rc = zfs_symlink(VTOZ(ap->a_dvp), cnp->cn_nameptr, vap,
4961 ap->a_target, &zp, cnp->cn_cred, 0 /* flags */);
4963 *ap->a_vpp = ZTOV(zp);
4967 #ifndef _SYS_SYSPROTO_H_
4968 struct vop_readlink_args {
4971 struct ucred *a_cred;
4976 zfs_freebsd_readlink(struct vop_readlink_args *ap)
4979 return (zfs_readlink(ap->a_vp, ap->a_uio, ap->a_cred, NULL));
4982 #ifndef _SYS_SYSPROTO_H_
4983 struct vop_link_args {
4984 struct vnode *a_tdvp;
4986 struct componentname *a_cnp;
4991 zfs_freebsd_link(struct vop_link_args *ap)
4993 struct componentname *cnp = ap->a_cnp;
4994 vnode_t *vp = ap->a_vp;
4995 vnode_t *tdvp = ap->a_tdvp;
4997 if (tdvp->v_mount != vp->v_mount)
5000 ASSERT(cnp->cn_flags & SAVENAME);
5002 return (zfs_link(VTOZ(tdvp), VTOZ(vp),
5003 cnp->cn_nameptr, cnp->cn_cred, 0));
5006 #ifndef _SYS_SYSPROTO_H_
5007 struct vop_inactive_args {
5009 struct thread *a_td;
5014 zfs_freebsd_inactive(struct vop_inactive_args *ap)
5016 vnode_t *vp = ap->a_vp;
5018 #if __FreeBSD_version >= 1300123
5019 zfs_inactive(vp, curthread->td_ucred, NULL);
5021 zfs_inactive(vp, ap->a_td->td_ucred, NULL);
5026 #if __FreeBSD_version >= 1300042
5027 #ifndef _SYS_SYSPROTO_H_
5028 struct vop_need_inactive_args {
5034 zfs_freebsd_need_inactive(struct vop_need_inactive_args *ap)
5036 vnode_t *vp = ap->a_vp;
5037 znode_t *zp = VTOZ(vp);
5038 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
5041 if (vn_need_pageq_flush(vp))
5044 if (!ZFS_TEARDOWN_INACTIVE_TRY_ENTER_READ(zfsvfs))
5046 need = (zp->z_sa_hdl == NULL || zp->z_unlinked || zp->z_atime_dirty);
5047 ZFS_TEARDOWN_INACTIVE_EXIT_READ(zfsvfs);
5053 #ifndef _SYS_SYSPROTO_H_
5054 struct vop_reclaim_args {
5056 struct thread *a_td;
5061 zfs_freebsd_reclaim(struct vop_reclaim_args *ap)
5063 vnode_t *vp = ap->a_vp;
5064 znode_t *zp = VTOZ(vp);
5065 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
5069 #if __FreeBSD_version < 1300042
5070 /* Destroy the vm object and flush associated pages. */
5071 vnode_destroy_vobject(vp);
5074 * z_teardown_inactive_lock protects from a race with
5075 * zfs_znode_dmu_fini in zfsvfs_teardown during
5078 ZFS_TEARDOWN_INACTIVE_ENTER_READ(zfsvfs);
5079 if (zp->z_sa_hdl == NULL)
5083 ZFS_TEARDOWN_INACTIVE_EXIT_READ(zfsvfs);
5089 #ifndef _SYS_SYSPROTO_H_
5090 struct vop_fid_args {
5097 zfs_freebsd_fid(struct vop_fid_args *ap)
5100 return (zfs_fid(ap->a_vp, (void *)ap->a_fid, NULL));
5104 #ifndef _SYS_SYSPROTO_H_
5105 struct vop_pathconf_args {
5108 register_t *a_retval;
5113 zfs_freebsd_pathconf(struct vop_pathconf_args *ap)
5118 error = zfs_pathconf(ap->a_vp, ap->a_name, &val,
5119 curthread->td_ucred, NULL);
5121 *ap->a_retval = val;
5124 if (error != EOPNOTSUPP)
5127 switch (ap->a_name) {
5129 *ap->a_retval = NAME_MAX;
5132 if (ap->a_vp->v_type == VDIR || ap->a_vp->v_type == VFIFO) {
5133 *ap->a_retval = PIPE_BUF;
5138 return (vop_stdpathconf(ap));
5143 * FreeBSD's extended attributes namespace defines file name prefix for ZFS'
5144 * extended attribute name:
5147 * system freebsd:system:
5148 * user (none, can be used to access ZFS fsattr(5) attributes
5149 * created on Solaris)
5152 zfs_create_attrname(int attrnamespace, const char *name, char *attrname,
5155 const char *namespace, *prefix, *suffix;
5157 /* We don't allow '/' character in attribute name. */
5158 if (strchr(name, '/') != NULL)
5160 /* We don't allow attribute names that start with "freebsd:" string. */
5161 if (strncmp(name, "freebsd:", 8) == 0)
5164 bzero(attrname, size);
5166 switch (attrnamespace) {
5167 case EXTATTR_NAMESPACE_USER:
5169 prefix = "freebsd:";
5170 namespace = EXTATTR_NAMESPACE_USER_STRING;
5174 * This is the default namespace by which we can access all
5175 * attributes created on Solaris.
5177 prefix = namespace = suffix = "";
5180 case EXTATTR_NAMESPACE_SYSTEM:
5181 prefix = "freebsd:";
5182 namespace = EXTATTR_NAMESPACE_SYSTEM_STRING;
5185 case EXTATTR_NAMESPACE_EMPTY:
5189 if (snprintf(attrname, size, "%s%s%s%s", prefix, namespace, suffix,
5191 return (ENAMETOOLONG);
5196 #ifndef _SYS_SYSPROTO_H_
5197 struct vop_getextattr {
5198 IN struct vnode *a_vp;
5199 IN int a_attrnamespace;
5200 IN const char *a_name;
5201 INOUT struct uio *a_uio;
5203 IN struct ucred *a_cred;
5204 IN struct thread *a_td;
5209 * Vnode operating to retrieve a named extended attribute.
5212 zfs_getextattr(struct vop_getextattr_args *ap)
5214 zfsvfs_t *zfsvfs = VTOZ(ap->a_vp)->z_zfsvfs;
5215 struct thread *td = ap->a_td;
5216 struct nameidata nd;
5219 vnode_t *xvp = NULL, *vp;
5223 * If the xattr property is off, refuse the request.
5225 if (!(zfsvfs->z_flags & ZSB_XATTR)) {
5226 return (SET_ERROR(EOPNOTSUPP));
5229 error = extattr_check_cred(ap->a_vp, ap->a_attrnamespace,
5230 ap->a_cred, ap->a_td, VREAD);
5234 error = zfs_create_attrname(ap->a_attrnamespace, ap->a_name, attrname,
5241 error = zfs_lookup(ap->a_vp, NULL, &xvp, NULL, 0, ap->a_cred, td,
5242 LOOKUP_XATTR, B_FALSE);
5249 NDINIT_ATVP(&nd, LOOKUP, NOFOLLOW, UIO_SYSSPACE, attrname,
5251 error = vn_open_cred(&nd, &flags, 0, VN_OPEN_INVFS, ap->a_cred, NULL);
5253 NDFREE(&nd, NDF_ONLY_PNBUF);
5256 if (error == ENOENT)
5261 if (ap->a_size != NULL) {
5262 error = VOP_GETATTR(vp, &va, ap->a_cred);
5264 *ap->a_size = (size_t)va.va_size;
5265 } else if (ap->a_uio != NULL)
5266 error = VOP_READ(vp, ap->a_uio, IO_UNIT, ap->a_cred);
5269 vn_close(vp, flags, ap->a_cred, td);
5274 #ifndef _SYS_SYSPROTO_H_
5275 struct vop_deleteextattr {
5276 IN struct vnode *a_vp;
5277 IN int a_attrnamespace;
5278 IN const char *a_name;
5279 IN struct ucred *a_cred;
5280 IN struct thread *a_td;
5285 * Vnode operation to remove a named attribute.
5288 zfs_deleteextattr(struct vop_deleteextattr_args *ap)
5290 zfsvfs_t *zfsvfs = VTOZ(ap->a_vp)->z_zfsvfs;
5291 struct thread *td = ap->a_td;
5292 struct nameidata nd;
5294 vnode_t *xvp = NULL, *vp;
5298 * If the xattr property is off, refuse the request.
5300 if (!(zfsvfs->z_flags & ZSB_XATTR)) {
5301 return (SET_ERROR(EOPNOTSUPP));
5304 error = extattr_check_cred(ap->a_vp, ap->a_attrnamespace,
5305 ap->a_cred, ap->a_td, VWRITE);
5309 error = zfs_create_attrname(ap->a_attrnamespace, ap->a_name, attrname,
5316 error = zfs_lookup(ap->a_vp, NULL, &xvp, NULL, 0, ap->a_cred, td,
5317 LOOKUP_XATTR, B_FALSE);
5323 NDINIT_ATVP(&nd, DELETE, NOFOLLOW | LOCKPARENT | LOCKLEAF,
5324 UIO_SYSSPACE, attrname, xvp, td);
5329 NDFREE(&nd, NDF_ONLY_PNBUF);
5330 if (error == ENOENT)
5335 error = VOP_REMOVE(nd.ni_dvp, vp, &nd.ni_cnd);
5336 NDFREE(&nd, NDF_ONLY_PNBUF);
5339 if (vp == nd.ni_dvp)
5348 #ifndef _SYS_SYSPROTO_H_
5349 struct vop_setextattr {
5350 IN struct vnode *a_vp;
5351 IN int a_attrnamespace;
5352 IN const char *a_name;
5353 INOUT struct uio *a_uio;
5354 IN struct ucred *a_cred;
5355 IN struct thread *a_td;
5360 * Vnode operation to set a named attribute.
5363 zfs_setextattr(struct vop_setextattr_args *ap)
5365 zfsvfs_t *zfsvfs = VTOZ(ap->a_vp)->z_zfsvfs;
5366 struct thread *td = ap->a_td;
5367 struct nameidata nd;
5370 vnode_t *xvp = NULL, *vp;
5374 * If the xattr property is off, refuse the request.
5376 if (!(zfsvfs->z_flags & ZSB_XATTR)) {
5377 return (SET_ERROR(EOPNOTSUPP));
5380 error = extattr_check_cred(ap->a_vp, ap->a_attrnamespace,
5381 ap->a_cred, ap->a_td, VWRITE);
5384 error = zfs_create_attrname(ap->a_attrnamespace, ap->a_name, attrname,
5391 error = zfs_lookup(ap->a_vp, NULL, &xvp, NULL, 0, ap->a_cred, td,
5392 LOOKUP_XATTR | CREATE_XATTR_DIR, B_FALSE);
5398 flags = FFLAGS(O_WRONLY | O_CREAT);
5399 NDINIT_ATVP(&nd, LOOKUP, NOFOLLOW, UIO_SYSSPACE, attrname,
5401 error = vn_open_cred(&nd, &flags, 0600, VN_OPEN_INVFS, ap->a_cred,
5404 NDFREE(&nd, NDF_ONLY_PNBUF);
5412 error = VOP_SETATTR(vp, &va, ap->a_cred);
5414 VOP_WRITE(vp, ap->a_uio, IO_UNIT, ap->a_cred);
5417 vn_close(vp, flags, ap->a_cred, td);
5422 #ifndef _SYS_SYSPROTO_H_
5423 struct vop_listextattr {
5424 IN struct vnode *a_vp;
5425 IN int a_attrnamespace;
5426 INOUT struct uio *a_uio;
5428 IN struct ucred *a_cred;
5429 IN struct thread *a_td;
5434 * Vnode operation to retrieve extended attributes on a vnode.
5437 zfs_listextattr(struct vop_listextattr_args *ap)
5439 zfsvfs_t *zfsvfs = VTOZ(ap->a_vp)->z_zfsvfs;
5440 struct thread *td = ap->a_td;
5441 struct nameidata nd;
5442 char attrprefix[16];
5443 uint8_t dirbuf[sizeof (struct dirent)];
5446 struct uio auio, *uio = ap->a_uio;
5447 size_t *sizep = ap->a_size;
5449 vnode_t *xvp = NULL, *vp;
5450 int done, error, eof, pos;
5453 * If the xattr property is off, refuse the request.
5455 if (!(zfsvfs->z_flags & ZSB_XATTR)) {
5456 return (SET_ERROR(EOPNOTSUPP));
5459 error = extattr_check_cred(ap->a_vp, ap->a_attrnamespace,
5460 ap->a_cred, ap->a_td, VREAD);
5464 error = zfs_create_attrname(ap->a_attrnamespace, "", attrprefix,
5465 sizeof (attrprefix));
5468 plen = strlen(attrprefix);
5475 error = zfs_lookup(ap->a_vp, NULL, &xvp, NULL, 0, ap->a_cred, td,
5476 LOOKUP_XATTR, B_FALSE);
5480 * ENOATTR means that the EA directory does not yet exist,
5481 * i.e. there are no extended attributes there.
5483 if (error == ENOATTR)
5488 NDINIT_ATVP(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | LOCKSHARED,
5489 UIO_SYSSPACE, ".", xvp, td);
5492 NDFREE(&nd, NDF_ONLY_PNBUF);
5498 auio.uio_iov = &aiov;
5499 auio.uio_iovcnt = 1;
5500 auio.uio_segflg = UIO_SYSSPACE;
5502 auio.uio_rw = UIO_READ;
5503 auio.uio_offset = 0;
5508 aiov.iov_base = (void *)dirbuf;
5509 aiov.iov_len = sizeof (dirbuf);
5510 auio.uio_resid = sizeof (dirbuf);
5511 error = VOP_READDIR(vp, &auio, ap->a_cred, &eof, NULL, NULL);
5512 done = sizeof (dirbuf) - auio.uio_resid;
5515 for (pos = 0; pos < done; ) {
5516 dp = (struct dirent *)(dirbuf + pos);
5517 pos += dp->d_reclen;
5519 * XXX: Temporarily we also accept DT_UNKNOWN, as this
5520 * is what we get when attribute was created on Solaris.
5522 if (dp->d_type != DT_REG && dp->d_type != DT_UNKNOWN)
5525 strncmp(dp->d_name, "freebsd:", 8) == 0)
5527 else if (strncmp(dp->d_name, attrprefix, plen) != 0)
5529 nlen = dp->d_namlen - plen;
5532 else if (uio != NULL) {
5534 * Format of extattr name entry is one byte for
5535 * length and the rest for name.
5537 error = uiomove(&nlen, 1, uio->uio_rw, uio);
5539 error = uiomove(dp->d_name + plen, nlen,
5546 } while (!eof && error == 0);
5554 #ifndef _SYS_SYSPROTO_H_
5555 struct vop_getacl_args {
5565 zfs_freebsd_getacl(struct vop_getacl_args *ap)
5568 vsecattr_t vsecattr;
5570 if (ap->a_type != ACL_TYPE_NFS4)
5573 vsecattr.vsa_mask = VSA_ACE | VSA_ACECNT;
5574 if ((error = zfs_getsecattr(VTOZ(ap->a_vp),
5575 &vsecattr, 0, ap->a_cred)))
5578 error = acl_from_aces(ap->a_aclp, vsecattr.vsa_aclentp,
5579 vsecattr.vsa_aclcnt);
5580 if (vsecattr.vsa_aclentp != NULL)
5581 kmem_free(vsecattr.vsa_aclentp, vsecattr.vsa_aclentsz);
5586 #ifndef _SYS_SYSPROTO_H_
5587 struct vop_setacl_args {
5597 zfs_freebsd_setacl(struct vop_setacl_args *ap)
5600 vsecattr_t vsecattr;
5601 int aclbsize; /* size of acl list in bytes */
5604 if (ap->a_type != ACL_TYPE_NFS4)
5607 if (ap->a_aclp == NULL)
5610 if (ap->a_aclp->acl_cnt < 1 || ap->a_aclp->acl_cnt > MAX_ACL_ENTRIES)
5614 * With NFSv4 ACLs, chmod(2) may need to add additional entries,
5615 * splitting every entry into two and appending "canonical six"
5616 * entries at the end. Don't allow for setting an ACL that would
5617 * cause chmod(2) to run out of ACL entries.
5619 if (ap->a_aclp->acl_cnt * 2 + 6 > ACL_MAX_ENTRIES)
5622 error = acl_nfs4_check(ap->a_aclp, ap->a_vp->v_type == VDIR);
5626 vsecattr.vsa_mask = VSA_ACE;
5627 aclbsize = ap->a_aclp->acl_cnt * sizeof (ace_t);
5628 vsecattr.vsa_aclentp = kmem_alloc(aclbsize, KM_SLEEP);
5629 aaclp = vsecattr.vsa_aclentp;
5630 vsecattr.vsa_aclentsz = aclbsize;
5632 aces_from_acl(vsecattr.vsa_aclentp, &vsecattr.vsa_aclcnt, ap->a_aclp);
5633 error = zfs_setsecattr(VTOZ(ap->a_vp), &vsecattr, 0, ap->a_cred);
5634 kmem_free(aaclp, aclbsize);
5639 #ifndef _SYS_SYSPROTO_H_
5640 struct vop_aclcheck_args {
5650 zfs_freebsd_aclcheck(struct vop_aclcheck_args *ap)
5653 return (EOPNOTSUPP);
5657 zfs_vptocnp(struct vop_vptocnp_args *ap)
5659 vnode_t *covered_vp;
5660 vnode_t *vp = ap->a_vp;
5661 zfsvfs_t *zfsvfs = vp->v_vfsp->vfs_data;
5662 znode_t *zp = VTOZ(vp);
5670 * If we are a snapshot mounted under .zfs, run the operation
5671 * on the covered vnode.
5673 if (zp->z_id != zfsvfs->z_root || zfsvfs->z_parent == zfsvfs) {
5674 char name[MAXNAMLEN + 1];
5678 error = zfs_znode_parent_and_name(zp, &dzp, name);
5681 if (*ap->a_buflen < len)
5682 error = SET_ERROR(ENOMEM);
5685 *ap->a_buflen -= len;
5686 bcopy(name, ap->a_buf + *ap->a_buflen, len);
5687 *ap->a_vpp = ZTOV(dzp);
5694 covered_vp = vp->v_mount->mnt_vnodecovered;
5695 #if __FreeBSD_version >= 1300045
5696 enum vgetstate vs = vget_prep(covered_vp);
5700 ltype = VOP_ISLOCKED(vp);
5702 #if __FreeBSD_version >= 1300045
5703 error = vget_finish(covered_vp, LK_SHARED, vs);
5705 error = vget(covered_vp, LK_SHARED | LK_VNHELD, curthread);
5708 #if __FreeBSD_version >= 1300123
5709 error = VOP_VPTOCNP(covered_vp, ap->a_vpp, ap->a_buf,
5712 error = VOP_VPTOCNP(covered_vp, ap->a_vpp, ap->a_cred,
5713 ap->a_buf, ap->a_buflen);
5717 vn_lock(vp, ltype | LK_RETRY);
5718 if (VN_IS_DOOMED(vp))
5719 error = SET_ERROR(ENOENT);
5723 struct vop_vector zfs_vnodeops;
5724 struct vop_vector zfs_fifoops;
5725 struct vop_vector zfs_shareops;
5727 struct vop_vector zfs_vnodeops = {
5728 .vop_default = &default_vnodeops,
5729 .vop_inactive = zfs_freebsd_inactive,
5730 #if __FreeBSD_version >= 1300042
5731 .vop_need_inactive = zfs_freebsd_need_inactive,
5733 .vop_reclaim = zfs_freebsd_reclaim,
5734 #if __FreeBSD_version >= 1300102
5735 .vop_fplookup_vexec = zfs_freebsd_fplookup_vexec,
5737 .vop_access = zfs_freebsd_access,
5738 .vop_allocate = VOP_EINVAL,
5739 .vop_lookup = zfs_cache_lookup,
5740 .vop_cachedlookup = zfs_freebsd_cachedlookup,
5741 .vop_getattr = zfs_freebsd_getattr,
5742 .vop_setattr = zfs_freebsd_setattr,
5743 .vop_create = zfs_freebsd_create,
5744 .vop_mknod = (vop_mknod_t *)zfs_freebsd_create,
5745 .vop_mkdir = zfs_freebsd_mkdir,
5746 .vop_readdir = zfs_freebsd_readdir,
5747 .vop_fsync = zfs_freebsd_fsync,
5748 .vop_open = zfs_freebsd_open,
5749 .vop_close = zfs_freebsd_close,
5750 .vop_rmdir = zfs_freebsd_rmdir,
5751 .vop_ioctl = zfs_freebsd_ioctl,
5752 .vop_link = zfs_freebsd_link,
5753 .vop_symlink = zfs_freebsd_symlink,
5754 .vop_readlink = zfs_freebsd_readlink,
5755 .vop_read = zfs_freebsd_read,
5756 .vop_write = zfs_freebsd_write,
5757 .vop_remove = zfs_freebsd_remove,
5758 .vop_rename = zfs_freebsd_rename,
5759 .vop_pathconf = zfs_freebsd_pathconf,
5760 .vop_bmap = zfs_freebsd_bmap,
5761 .vop_fid = zfs_freebsd_fid,
5762 .vop_getextattr = zfs_getextattr,
5763 .vop_deleteextattr = zfs_deleteextattr,
5764 .vop_setextattr = zfs_setextattr,
5765 .vop_listextattr = zfs_listextattr,
5766 .vop_getacl = zfs_freebsd_getacl,
5767 .vop_setacl = zfs_freebsd_setacl,
5768 .vop_aclcheck = zfs_freebsd_aclcheck,
5769 .vop_getpages = zfs_freebsd_getpages,
5770 .vop_putpages = zfs_freebsd_putpages,
5771 .vop_vptocnp = zfs_vptocnp,
5772 #if __FreeBSD_version >= 1300064
5773 .vop_lock1 = vop_lock,
5774 .vop_unlock = vop_unlock,
5775 .vop_islocked = vop_islocked,
5778 VFS_VOP_VECTOR_REGISTER(zfs_vnodeops);
5780 struct vop_vector zfs_fifoops = {
5781 .vop_default = &fifo_specops,
5782 .vop_fsync = zfs_freebsd_fsync,
5783 #if __FreeBSD_version >= 1300102
5784 .vop_fplookup_vexec = zfs_freebsd_fplookup_vexec,
5786 .vop_access = zfs_freebsd_access,
5787 .vop_getattr = zfs_freebsd_getattr,
5788 .vop_inactive = zfs_freebsd_inactive,
5789 .vop_read = VOP_PANIC,
5790 .vop_reclaim = zfs_freebsd_reclaim,
5791 .vop_setattr = zfs_freebsd_setattr,
5792 .vop_write = VOP_PANIC,
5793 .vop_pathconf = zfs_freebsd_pathconf,
5794 .vop_fid = zfs_freebsd_fid,
5795 .vop_getacl = zfs_freebsd_getacl,
5796 .vop_setacl = zfs_freebsd_setacl,
5797 .vop_aclcheck = zfs_freebsd_aclcheck,
5799 VFS_VOP_VECTOR_REGISTER(zfs_fifoops);
5802 * special share hidden files vnode operations template
5804 struct vop_vector zfs_shareops = {
5805 .vop_default = &default_vnodeops,
5806 #if __FreeBSD_version >= 1300121
5807 .vop_fplookup_vexec = VOP_EAGAIN,
5809 .vop_access = zfs_freebsd_access,
5810 .vop_inactive = zfs_freebsd_inactive,
5811 .vop_reclaim = zfs_freebsd_reclaim,
5812 .vop_fid = zfs_freebsd_fid,
5813 .vop_pathconf = zfs_freebsd_pathconf,
5815 VFS_VOP_VECTOR_REGISTER(zfs_shareops);