4 * The contents of this file are subject to the terms of the
5 * Common Development and Distribution License (the "License").
6 * You may not use this file except in compliance with the License.
8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9 * or http://www.opensolaris.org/os/licensing.
10 * See the License for the specific language governing permissions
11 * and limitations under the License.
13 * When distributing Covered Code, include this CDDL HEADER in each
14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15 * If applicable, add the following below this CDDL HEADER, with the
16 * fields enclosed by brackets "[]" replaced with your own identifying
17 * information: Portions Copyright [yyyy] [name of copyright owner]
22 * Copyright 2008 Sun Microsystems, Inc. All rights reserved.
23 * Use is subject to license terms.
26 /* Portions Copyright 2007 Jeremy Teo */
28 #include <sys/types.h>
29 #include <sys/param.h>
31 #include <sys/systm.h>
32 #include <sys/sysmacros.h>
33 #include <sys/resource.h>
35 #include <sys/vnode.h>
39 #include <sys/taskq.h>
41 #include <sys/atomic.h>
42 #include <sys/namei.h>
44 #include <sys/cmn_err.h>
45 #include <sys/errno.h>
46 #include <sys/unistd.h>
47 #include <sys/zfs_dir.h>
48 #include <sys/zfs_ioctl.h>
49 #include <sys/fs/zfs.h>
55 #include <sys/dirent.h>
56 #include <sys/policy.h>
57 #include <sys/sunddi.h>
58 #include <sys/filio.h>
59 #include <sys/zfs_ctldir.h>
60 #include <sys/zfs_fuid.h>
62 #include <sys/zfs_rlock.h>
63 #include <sys/extdirent.h>
64 #include <sys/kidmap.h>
67 #include <sys/sf_buf.h>
68 #include <sys/sched.h>
74 * Each vnode op performs some logical unit of work. To do this, the ZPL must
75 * properly lock its in-core state, create a DMU transaction, do the work,
76 * record this work in the intent log (ZIL), commit the DMU transaction,
77 * and wait for the intent log to commit if it is a synchronous operation.
78 * Moreover, the vnode ops must work in both normal and log replay context.
79 * The ordering of events is important to avoid deadlocks and references
80 * to freed memory. The example below illustrates the following Big Rules:
82 * (1) A check must be made in each zfs thread for a mounted file system.
83 * This is done avoiding races using ZFS_ENTER(zfsvfs).
84 * A ZFS_EXIT(zfsvfs) is needed before all returns. Any znodes
85 * must be checked with ZFS_VERIFY_ZP(zp). Both of these macros
86 * can return EIO from the calling function.
88 * (2) VN_RELE() should always be the last thing except for zil_commit()
89 * (if necessary) and ZFS_EXIT(). This is for 3 reasons:
90 * First, if it's the last reference, the vnode/znode
91 * can be freed, so the zp may point to freed memory. Second, the last
92 * reference will call zfs_zinactive(), which may induce a lot of work --
93 * pushing cached pages (which acquires range locks) and syncing out
94 * cached atime changes. Third, zfs_zinactive() may require a new tx,
95 * which could deadlock the system if you were already holding one.
96 * If you must call VN_RELE() within a tx then use VN_RELE_ASYNC().
98 * (3) All range locks must be grabbed before calling dmu_tx_assign(),
99 * as they can span dmu_tx_assign() calls.
101 * (4) Always pass zfsvfs->z_assign as the second argument to dmu_tx_assign().
102 * In normal operation, this will be TXG_NOWAIT. During ZIL replay,
103 * it will be a specific txg. Either way, dmu_tx_assign() never blocks.
104 * This is critical because we don't want to block while holding locks.
105 * Note, in particular, that if a lock is sometimes acquired before
106 * the tx assigns, and sometimes after (e.g. z_lock), then failing to
107 * use a non-blocking assign can deadlock the system. The scenario:
109 * Thread A has grabbed a lock before calling dmu_tx_assign().
110 * Thread B is in an already-assigned tx, and blocks for this lock.
111 * Thread A calls dmu_tx_assign(TXG_WAIT) and blocks in txg_wait_open()
112 * forever, because the previous txg can't quiesce until B's tx commits.
114 * If dmu_tx_assign() returns ERESTART and zfsvfs->z_assign is TXG_NOWAIT,
115 * then drop all locks, call dmu_tx_wait(), and try again.
117 * (5) If the operation succeeded, generate the intent log entry for it
118 * before dropping locks. This ensures that the ordering of events
119 * in the intent log matches the order in which they actually occurred.
121 * (6) At the end of each vnode op, the DMU tx must always commit,
122 * regardless of whether there were any errors.
124 * (7) After dropping all locks, invoke zil_commit(zilog, seq, foid)
125 * to ensure that synchronous semantics are provided when necessary.
127 * In general, this is how things should be ordered in each vnode op:
129 * ZFS_ENTER(zfsvfs); // exit if unmounted
131 * zfs_dirent_lock(&dl, ...) // lock directory entry (may VN_HOLD())
132 * rw_enter(...); // grab any other locks you need
133 * tx = dmu_tx_create(...); // get DMU tx
134 * dmu_tx_hold_*(); // hold each object you might modify
135 * error = dmu_tx_assign(tx, zfsvfs->z_assign); // try to assign
137 * rw_exit(...); // drop locks
138 * zfs_dirent_unlock(dl); // unlock directory entry
139 * VN_RELE(...); // release held vnodes
140 * if (error == ERESTART && zfsvfs->z_assign == TXG_NOWAIT) {
145 * dmu_tx_abort(tx); // abort DMU tx
146 * ZFS_EXIT(zfsvfs); // finished in zfs
147 * return (error); // really out of space
149 * error = do_real_work(); // do whatever this VOP does
151 * zfs_log_*(...); // on success, make ZIL entry
152 * dmu_tx_commit(tx); // commit DMU tx -- error or not
153 * rw_exit(...); // drop locks
154 * zfs_dirent_unlock(dl); // unlock directory entry
155 * VN_RELE(...); // release held vnodes
156 * zil_commit(zilog, seq, foid); // synchronous when necessary
157 * ZFS_EXIT(zfsvfs); // finished in zfs
158 * return (error); // done, report error
163 zfs_open(vnode_t **vpp, int flag, cred_t *cr, caller_context_t *ct)
165 znode_t *zp = VTOZ(*vpp);
167 if ((flag & FWRITE) && (zp->z_phys->zp_flags & ZFS_APPENDONLY) &&
168 ((flag & FAPPEND) == 0)) {
172 if (!zfs_has_ctldir(zp) && zp->z_zfsvfs->z_vscan &&
173 ZTOV(zp)->v_type == VREG &&
174 !(zp->z_phys->zp_flags & ZFS_AV_QUARANTINED) &&
175 zp->z_phys->zp_size > 0)
176 if (fs_vscan(*vpp, cr, 0) != 0)
179 /* Keep a count of the synchronous opens in the znode */
180 if (flag & (FSYNC | FDSYNC))
181 atomic_inc_32(&zp->z_sync_cnt);
188 zfs_close(vnode_t *vp, int flag, int count, offset_t offset, cred_t *cr,
189 caller_context_t *ct)
191 znode_t *zp = VTOZ(vp);
193 /* Decrement the synchronous opens in the znode */
194 if ((flag & (FSYNC | FDSYNC)) && (count == 1))
195 atomic_dec_32(&zp->z_sync_cnt);
198 * Clean up any locks held by this process on the vp.
200 cleanlocks(vp, ddi_get_pid(), 0);
201 cleanshares(vp, ddi_get_pid());
203 if (!zfs_has_ctldir(zp) && zp->z_zfsvfs->z_vscan &&
204 ZTOV(zp)->v_type == VREG &&
205 !(zp->z_phys->zp_flags & ZFS_AV_QUARANTINED) &&
206 zp->z_phys->zp_size > 0)
207 VERIFY(fs_vscan(vp, cr, 1) == 0);
213 * Lseek support for finding holes (cmd == _FIO_SEEK_HOLE) and
214 * data (cmd == _FIO_SEEK_DATA). "off" is an in/out parameter.
217 zfs_holey(vnode_t *vp, u_long cmd, offset_t *off)
219 znode_t *zp = VTOZ(vp);
220 uint64_t noff = (uint64_t)*off; /* new offset */
225 file_sz = zp->z_phys->zp_size;
226 if (noff >= file_sz) {
230 if (cmd == _FIO_SEEK_HOLE)
235 error = dmu_offset_next(zp->z_zfsvfs->z_os, zp->z_id, hole, &noff);
238 if ((error == ESRCH) || (noff > file_sz)) {
240 * Handle the virtual hole at the end of file.
257 zfs_ioctl(vnode_t *vp, u_long com, intptr_t data, int flag, cred_t *cred,
258 int *rvalp, caller_context_t *ct)
270 * The following two ioctls are used by bfu. Faking out,
271 * necessary to avoid bfu errors.
279 if (ddi_copyin((void *)data, &off, sizeof (off), flag))
283 zfsvfs = zp->z_zfsvfs;
287 /* offset parameter is in/out */
288 error = zfs_holey(vp, com, &off);
292 if (ddi_copyout(&off, (void *)data, sizeof (off), flag))
300 * When a file is memory mapped, we must keep the IO data synchronized
301 * between the DMU cache and the memory mapped pages. What this means:
303 * On Write: If we find a memory mapped page, we write to *both*
304 * the page and the dmu buffer.
306 * NOTE: We will always "break up" the IO into PAGESIZE uiomoves when
307 * the file is memory mapped.
310 mappedwrite(vnode_t *vp, int nbytes, uio_t *uio, dmu_tx_t *tx)
312 znode_t *zp = VTOZ(vp);
313 objset_t *os = zp->z_zfsvfs->z_os;
322 ASSERT(vp->v_mount != NULL);
326 start = uio->uio_loffset;
327 off = start & PAGEOFFSET;
330 for (start &= PAGEMASK; len > 0; start += PAGESIZE) {
331 uint64_t bytes = MIN(PAGESIZE - off, len);
335 if ((m = vm_page_lookup(obj, OFF_TO_IDX(start))) != NULL &&
336 vm_page_is_valid(m, (vm_offset_t)off, bytes)) {
340 if (vm_page_sleep_if_busy(m, FALSE, "zfsmwb"))
342 fsize = obj->un_pager.vnp.vnp_size;
344 vm_page_lock_queues();
346 vm_page_unlock_queues();
347 VM_OBJECT_UNLOCK(obj);
349 error = dmu_write_uio(os, zp->z_id, uio,
355 sf = sf_buf_alloc(m, SFB_CPUPRIVATE);
356 va = (caddr_t)sf_buf_kva(sf);
357 woff = uio->uio_loffset - off;
358 error = uiomove(va + off, bytes, UIO_WRITE, uio);
360 * The uiomove() above could have been partially
361 * successful, that's why we call dmu_write()
362 * below unconditionally. The page was marked
363 * non-dirty above and we would lose the changes
364 * without doing so. If the uiomove() failed
365 * entirely, well, we just write what we got
366 * before one more time.
368 dmu_write(os, zp->z_id, woff,
369 MIN(PAGESIZE, fsize - woff), va, tx);
376 if (__predict_false(obj->cache != NULL)) {
377 vm_page_cache_free(obj, OFF_TO_IDX(start),
378 OFF_TO_IDX(start) + 1);
387 VM_OBJECT_UNLOCK(obj);
388 if (error == 0 && dirbytes > 0)
389 error = dmu_write_uio(os, zp->z_id, uio, dirbytes, tx);
394 * When a file is memory mapped, we must keep the IO data synchronized
395 * between the DMU cache and the memory mapped pages. What this means:
397 * On Read: We "read" preferentially from memory mapped pages,
398 * else we default from the dmu buffer.
400 * NOTE: We will always "break up" the IO into PAGESIZE uiomoves when
401 * the file is memory mapped.
404 mappedread(vnode_t *vp, int nbytes, uio_t *uio)
406 znode_t *zp = VTOZ(vp);
407 objset_t *os = zp->z_zfsvfs->z_os;
417 ASSERT(vp->v_mount != NULL);
421 start = uio->uio_loffset;
422 off = start & PAGEOFFSET;
425 for (start &= PAGEMASK; len > 0; start += PAGESIZE) {
426 uint64_t bytes = MIN(PAGESIZE - off, len);
429 if ((m = vm_page_lookup(obj, OFF_TO_IDX(start))) != NULL &&
430 vm_page_is_valid(m, (vm_offset_t)off, bytes)) {
431 if (vm_page_sleep_if_busy(m, FALSE, "zfsmrb"))
434 VM_OBJECT_UNLOCK(obj);
436 error = dmu_read_uio(os, zp->z_id, uio,
442 sf = sf_buf_alloc(m, SFB_CPUPRIVATE);
443 va = (caddr_t)sf_buf_kva(sf);
444 error = uiomove(va + off, bytes, UIO_READ, uio);
450 } else if (m != NULL && uio->uio_segflg == UIO_NOCOPY) {
452 * The code below is here to make sendfile(2) work
453 * correctly with ZFS. As pointed out by ups@
454 * sendfile(2) should be changed to use VOP_GETPAGES(),
455 * but it pessimize performance of sendfile/UFS, that's
456 * why I handle this special case in ZFS code.
458 if (vm_page_sleep_if_busy(m, FALSE, "zfsmrb"))
461 VM_OBJECT_UNLOCK(obj);
463 error = dmu_read_uio(os, zp->z_id, uio,
469 sf = sf_buf_alloc(m, SFB_CPUPRIVATE);
470 va = (caddr_t)sf_buf_kva(sf);
471 error = dmu_read(os, zp->z_id, start + off,
472 bytes, (void *)(va + off));
479 uio->uio_resid -= bytes;
488 VM_OBJECT_UNLOCK(obj);
489 if (error == 0 && dirbytes > 0)
490 error = dmu_read_uio(os, zp->z_id, uio, dirbytes);
494 offset_t zfs_read_chunk_size = 1024 * 1024; /* Tunable */
497 * Read bytes from specified file into supplied buffer.
499 * IN: vp - vnode of file to be read from.
500 * uio - structure supplying read location, range info,
502 * ioflag - SYNC flags; used to provide FRSYNC semantics.
503 * cr - credentials of caller.
504 * ct - caller context
506 * OUT: uio - updated offset and range, buffer filled.
508 * RETURN: 0 if success
509 * error code if failure
512 * vp - atime updated if byte count > 0
516 zfs_read(vnode_t *vp, uio_t *uio, int ioflag, cred_t *cr, caller_context_t *ct)
518 znode_t *zp = VTOZ(vp);
519 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
529 if (zp->z_phys->zp_flags & ZFS_AV_QUARANTINED) {
535 * Validate file offset
537 if (uio->uio_loffset < (offset_t)0) {
543 * Fasttrack empty reads
545 if (uio->uio_resid == 0) {
551 * Check for mandatory locks
553 if (MANDMODE((mode_t)zp->z_phys->zp_mode)) {
554 if (error = chklock(vp, FREAD,
555 uio->uio_loffset, uio->uio_resid, uio->uio_fmode, ct)) {
562 * If we're in FRSYNC mode, sync out this znode before reading it.
565 zil_commit(zfsvfs->z_log, zp->z_last_itx, zp->z_id);
568 * Lock the range against changes.
570 rl = zfs_range_lock(zp, uio->uio_loffset, uio->uio_resid, RL_READER);
573 * If we are reading past end-of-file we can skip
574 * to the end; but we might still need to set atime.
576 if (uio->uio_loffset >= zp->z_phys->zp_size) {
581 ASSERT(uio->uio_loffset < zp->z_phys->zp_size);
582 n = MIN(uio->uio_resid, zp->z_phys->zp_size - uio->uio_loffset);
585 nbytes = MIN(n, zfs_read_chunk_size -
586 P2PHASE(uio->uio_loffset, zfs_read_chunk_size));
588 if (vn_has_cached_data(vp))
589 error = mappedread(vp, nbytes, uio);
591 error = dmu_read_uio(os, zp->z_id, uio, nbytes);
593 /* convert checksum errors into IO errors */
603 zfs_range_unlock(rl);
605 ZFS_ACCESSTIME_STAMP(zfsvfs, zp);
611 * Fault in the pages of the first n bytes specified by the uio structure.
612 * 1 byte in each page is touched and the uio struct is unmodified.
613 * Any error will exit this routine as this is only a best
614 * attempt to get the pages resident. This is a copy of ufs_trans_touch().
617 zfs_prefault_write(ssize_t n, struct uio *uio)
623 if (uio->uio_segflg != UIO_USERSPACE)
629 cnt = MIN(iov->iov_len, n);
631 /* empty iov entry */
637 * touch each page in this segment.
643 incr = MIN(cnt, PAGESIZE);
648 * touch the last byte in case it straddles a page.
658 * Write the bytes to a file.
660 * IN: vp - vnode of file to be written to.
661 * uio - structure supplying write location, range info,
663 * ioflag - IO_APPEND flag set if in append mode.
664 * cr - credentials of caller.
665 * ct - caller context (NFS/CIFS fem monitor only)
667 * OUT: uio - updated offset and range.
669 * RETURN: 0 if success
670 * error code if failure
673 * vp - ctime|mtime updated if byte count > 0
677 zfs_write(vnode_t *vp, uio_t *uio, int ioflag, cred_t *cr, caller_context_t *ct)
679 znode_t *zp = VTOZ(vp);
680 rlim64_t limit = MAXOFFSET_T;
681 ssize_t start_resid = uio->uio_resid;
685 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
690 int max_blksz = zfsvfs->z_max_blksz;
695 * Fasttrack empty write
701 if (limit == RLIM64_INFINITY || limit > MAXOFFSET_T)
708 * If immutable or not appending then return EPERM
710 pflags = zp->z_phys->zp_flags;
711 if ((pflags & (ZFS_IMMUTABLE | ZFS_READONLY)) ||
712 ((pflags & ZFS_APPENDONLY) && !(ioflag & FAPPEND) &&
713 (uio->uio_loffset < zp->z_phys->zp_size))) {
718 zilog = zfsvfs->z_log;
721 * Pre-fault the pages to ensure slow (eg NFS) pages
724 zfs_prefault_write(n, uio);
727 * If in append mode, set the io offset pointer to eof.
729 if (ioflag & IO_APPEND) {
731 * Range lock for a file append:
732 * The value for the start of range will be determined by
733 * zfs_range_lock() (to guarantee append semantics).
734 * If this write will cause the block size to increase,
735 * zfs_range_lock() will lock the entire file, so we must
736 * later reduce the range after we grow the block size.
738 rl = zfs_range_lock(zp, 0, n, RL_APPEND);
739 if (rl->r_len == UINT64_MAX) {
740 /* overlocked, zp_size can't change */
741 woff = uio->uio_loffset = zp->z_phys->zp_size;
743 woff = uio->uio_loffset = rl->r_off;
746 woff = uio->uio_loffset;
748 * Validate file offset
756 * If we need to grow the block size then zfs_range_lock()
757 * will lock a wider range than we request here.
758 * Later after growing the block size we reduce the range.
760 rl = zfs_range_lock(zp, woff, n, RL_WRITER);
764 zfs_range_unlock(rl);
769 if ((woff + n) > limit || woff > (limit - n))
773 * Check for mandatory locks
775 if (MANDMODE((mode_t)zp->z_phys->zp_mode) &&
776 (error = chklock(vp, FWRITE, woff, n, uio->uio_fmode, ct)) != 0) {
777 zfs_range_unlock(rl);
781 end_size = MAX(zp->z_phys->zp_size, woff + n);
784 * Write the file in reasonable size chunks. Each chunk is written
785 * in a separate transaction; this keeps the intent log records small
786 * and allows us to do more fine-grained space accounting.
790 * Start a transaction.
792 woff = uio->uio_loffset;
793 tx = dmu_tx_create(zfsvfs->z_os);
794 dmu_tx_hold_bonus(tx, zp->z_id);
795 dmu_tx_hold_write(tx, zp->z_id, woff, MIN(n, max_blksz));
796 error = dmu_tx_assign(tx, zfsvfs->z_assign);
798 if (error == ERESTART &&
799 zfsvfs->z_assign == TXG_NOWAIT) {
809 * If zfs_range_lock() over-locked we grow the blocksize
810 * and then reduce the lock range. This will only happen
811 * on the first iteration since zfs_range_reduce() will
812 * shrink down r_len to the appropriate size.
814 if (rl->r_len == UINT64_MAX) {
817 if (zp->z_blksz > max_blksz) {
818 ASSERT(!ISP2(zp->z_blksz));
819 new_blksz = MIN(end_size, SPA_MAXBLOCKSIZE);
821 new_blksz = MIN(end_size, max_blksz);
823 zfs_grow_blocksize(zp, new_blksz, tx);
824 zfs_range_reduce(rl, woff, n);
828 * XXX - should we really limit each write to z_max_blksz?
829 * Perhaps we should use SPA_MAXBLOCKSIZE chunks?
831 nbytes = MIN(n, max_blksz - P2PHASE(woff, max_blksz));
833 if (woff + nbytes > zp->z_phys->zp_size)
834 vnode_pager_setsize(vp, woff + nbytes);
836 rw_enter(&zp->z_map_lock, RW_READER);
838 tx_bytes = uio->uio_resid;
839 if (vn_has_cached_data(vp)) {
840 rw_exit(&zp->z_map_lock);
841 error = mappedwrite(vp, nbytes, uio, tx);
843 error = dmu_write_uio(zfsvfs->z_os, zp->z_id,
845 rw_exit(&zp->z_map_lock);
847 tx_bytes -= uio->uio_resid;
850 * If we made no progress, we're done. If we made even
851 * partial progress, update the znode and ZIL accordingly.
860 * Clear Set-UID/Set-GID bits on successful write if not
861 * privileged and at least one of the excute bits is set.
863 * It would be nice to to this after all writes have
864 * been done, but that would still expose the ISUID/ISGID
865 * to another app after the partial write is committed.
867 * Note: we don't call zfs_fuid_map_id() here because
868 * user 0 is not an ephemeral uid.
870 mutex_enter(&zp->z_acl_lock);
871 if ((zp->z_phys->zp_mode & (S_IXUSR | (S_IXUSR >> 3) |
872 (S_IXUSR >> 6))) != 0 &&
873 (zp->z_phys->zp_mode & (S_ISUID | S_ISGID)) != 0 &&
874 secpolicy_vnode_setid_retain(vp, cr,
875 (zp->z_phys->zp_mode & S_ISUID) != 0 &&
876 zp->z_phys->zp_uid == 0) != 0) {
877 zp->z_phys->zp_mode &= ~(S_ISUID | S_ISGID);
879 mutex_exit(&zp->z_acl_lock);
882 * Update time stamp. NOTE: This marks the bonus buffer as
883 * dirty, so we don't have to do it again for zp_size.
885 zfs_time_stamper(zp, CONTENT_MODIFIED, tx);
888 * Update the file size (zp_size) if it has changed;
889 * account for possible concurrent updates.
891 while ((end_size = zp->z_phys->zp_size) < uio->uio_loffset)
892 (void) atomic_cas_64(&zp->z_phys->zp_size, end_size,
894 zfs_log_write(zilog, tx, TX_WRITE, zp, woff, tx_bytes, ioflag);
899 ASSERT(tx_bytes == nbytes);
903 zfs_range_unlock(rl);
906 * If we're in replay mode, or we made no progress, return error.
907 * Otherwise, it's at least a partial write, so it's successful.
909 if (zfsvfs->z_assign >= TXG_INITIAL || uio->uio_resid == start_resid) {
914 if (ioflag & (FSYNC | FDSYNC))
915 zil_commit(zilog, zp->z_last_itx, zp->z_id);
922 zfs_get_done(dmu_buf_t *db, void *vzgd)
924 zgd_t *zgd = (zgd_t *)vzgd;
925 rl_t *rl = zgd->zgd_rl;
926 vnode_t *vp = ZTOV(rl->r_zp);
929 vfslocked = VFS_LOCK_GIANT(vp->v_vfsp);
930 dmu_buf_rele(db, vzgd);
931 zfs_range_unlock(rl);
933 * Release the vnode asynchronously as we currently have the
934 * txg stopped from syncing.
936 VN_RELE_ASYNC(vp, NULL);
937 zil_add_block(zgd->zgd_zilog, zgd->zgd_bp);
938 kmem_free(zgd, sizeof (zgd_t));
939 VFS_UNLOCK_GIANT(vfslocked);
943 * Get data to generate a TX_WRITE intent log record.
946 zfs_get_data(void *arg, lr_write_t *lr, char *buf, zio_t *zio)
948 zfsvfs_t *zfsvfs = arg;
949 objset_t *os = zfsvfs->z_os;
951 uint64_t off = lr->lr_offset;
955 int dlen = lr->lr_length; /* length of user data */
962 * Nothing to do if the file has been removed
964 if (zfs_zget(zfsvfs, lr->lr_foid, &zp) != 0)
966 if (zp->z_unlinked) {
968 * Release the vnode asynchronously as we currently have the
969 * txg stopped from syncing.
971 VN_RELE_ASYNC(ZTOV(zp), NULL);
977 * Write records come in two flavors: immediate and indirect.
978 * For small writes it's cheaper to store the data with the
979 * log record (immediate); for large writes it's cheaper to
980 * sync the data and get a pointer to it (indirect) so that
981 * we don't have to write the data twice.
983 if (buf != NULL) { /* immediate write */
984 rl = zfs_range_lock(zp, off, dlen, RL_READER);
985 /* test for truncation needs to be done while range locked */
986 if (off >= zp->z_phys->zp_size) {
990 VERIFY(0 == dmu_read(os, lr->lr_foid, off, dlen, buf));
991 } else { /* indirect write */
992 uint64_t boff; /* block starting offset */
995 * Have to lock the whole block to ensure when it's
996 * written out and it's checksum is being calculated
997 * that no one can change the data. We need to re-check
998 * blocksize after we get the lock in case it's changed!
1001 if (ISP2(zp->z_blksz)) {
1002 boff = P2ALIGN_TYPED(off, zp->z_blksz,
1008 rl = zfs_range_lock(zp, boff, dlen, RL_READER);
1009 if (zp->z_blksz == dlen)
1011 zfs_range_unlock(rl);
1013 /* test for truncation needs to be done while range locked */
1014 if (off >= zp->z_phys->zp_size) {
1018 zgd = (zgd_t *)kmem_alloc(sizeof (zgd_t), KM_SLEEP);
1020 zgd->zgd_zilog = zfsvfs->z_log;
1021 zgd->zgd_bp = &lr->lr_blkptr;
1022 VERIFY(0 == dmu_buf_hold(os, lr->lr_foid, boff, zgd, &db));
1023 ASSERT(boff == db->db_offset);
1024 lr->lr_blkoff = off - boff;
1025 error = dmu_sync(zio, db, &lr->lr_blkptr,
1026 lr->lr_common.lrc_txg, zfs_get_done, zgd);
1027 ASSERT((error && error != EINPROGRESS) ||
1028 lr->lr_length <= zp->z_blksz);
1030 zil_add_block(zfsvfs->z_log, &lr->lr_blkptr);
1032 * If we get EINPROGRESS, then we need to wait for a
1033 * write IO initiated by dmu_sync() to complete before
1034 * we can release this dbuf. We will finish everything
1035 * up in the zfs_get_done() callback.
1037 if (error == EINPROGRESS)
1039 dmu_buf_rele(db, zgd);
1040 kmem_free(zgd, sizeof (zgd_t));
1043 zfs_range_unlock(rl);
1045 * Release the vnode asynchronously as we currently have the
1046 * txg stopped from syncing.
1048 VN_RELE_ASYNC(ZTOV(zp), NULL);
1054 zfs_access(vnode_t *vp, int mode, int flag, cred_t *cr,
1055 caller_context_t *ct)
1057 znode_t *zp = VTOZ(vp);
1058 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1064 if (flag & V_ACE_MASK)
1065 error = zfs_zaccess(zp, mode, flag, B_FALSE, cr);
1067 error = zfs_zaccess_rwx(zp, mode, flag, cr);
1074 * Lookup an entry in a directory, or an extended attribute directory.
1075 * If it exists, return a held vnode reference for it.
1077 * IN: dvp - vnode of directory to search.
1078 * nm - name of entry to lookup.
1079 * pnp - full pathname to lookup [UNUSED].
1080 * flags - LOOKUP_XATTR set if looking for an attribute.
1081 * rdir - root directory vnode [UNUSED].
1082 * cr - credentials of caller.
1083 * ct - caller context
1084 * direntflags - directory lookup flags
1085 * realpnp - returned pathname.
1087 * OUT: vpp - vnode of located entry, NULL if not found.
1089 * RETURN: 0 if success
1090 * error code if failure
1097 zfs_lookup(vnode_t *dvp, char *nm, vnode_t **vpp, struct componentname *cnp,
1098 int nameiop, cred_t *cr, kthread_t *td, int flags)
1100 znode_t *zdp = VTOZ(dvp);
1101 zfsvfs_t *zfsvfs = zdp->z_zfsvfs;
1103 int *direntflags = NULL;
1104 void *realpnp = NULL;
1111 if (flags & LOOKUP_XATTR) {
1114 * If the xattr property is off, refuse the lookup request.
1116 if (!(zfsvfs->z_vfs->vfs_flag & VFS_XATTR)) {
1123 * We don't allow recursive attributes..
1124 * Maybe someday we will.
1126 if (zdp->z_phys->zp_flags & ZFS_XATTR) {
1131 if (error = zfs_get_xattrdir(VTOZ(dvp), vpp, cr, flags)) {
1137 * Do we have permission to get into attribute directory?
1140 if (error = zfs_zaccess(VTOZ(*vpp), ACE_EXECUTE, 0,
1150 if (dvp->v_type != VDIR) {
1156 * Check accessibility of directory.
1159 if (error = zfs_zaccess(zdp, ACE_EXECUTE, 0, B_FALSE, cr)) {
1164 if (zfsvfs->z_utf8 && u8_validate(nm, strlen(nm),
1165 NULL, U8_VALIDATE_ENTIRE, &error) < 0) {
1170 error = zfs_dirlook(zdp, nm, vpp, flags, direntflags, realpnp);
1173 * Convert device special files
1175 if (IS_DEVVP(*vpp)) {
1178 svp = specvp(*vpp, (*vpp)->v_rdev, (*vpp)->v_type, cr);
1189 /* Translate errors and add SAVENAME when needed. */
1190 if (cnp->cn_flags & ISLASTCN) {
1194 if (error == ENOENT) {
1195 error = EJUSTRETURN;
1196 cnp->cn_flags |= SAVENAME;
1202 cnp->cn_flags |= SAVENAME;
1206 if (error == 0 && (nm[0] != '.' || nm[1] != '\0')) {
1209 if (cnp->cn_flags & ISDOTDOT) {
1210 ltype = VOP_ISLOCKED(dvp);
1213 error = vn_lock(*vpp, cnp->cn_lkflags);
1214 if (cnp->cn_flags & ISDOTDOT)
1215 vn_lock(dvp, ltype | LK_RETRY);
1223 #ifdef FREEBSD_NAMECACHE
1225 * Insert name into cache (as non-existent) if appropriate.
1227 if (error == ENOENT && (cnp->cn_flags & MAKEENTRY) && nameiop != CREATE)
1228 cache_enter(dvp, *vpp, cnp);
1230 * Insert name into cache if appropriate.
1232 if (error == 0 && (cnp->cn_flags & MAKEENTRY)) {
1233 if (!(cnp->cn_flags & ISLASTCN) ||
1234 (nameiop != DELETE && nameiop != RENAME)) {
1235 cache_enter(dvp, *vpp, cnp);
1244 * Attempt to create a new entry in a directory. If the entry
1245 * already exists, truncate the file if permissible, else return
1246 * an error. Return the vp of the created or trunc'd file.
1248 * IN: dvp - vnode of directory to put new file entry in.
1249 * name - name of new file entry.
1250 * vap - attributes of new file.
1251 * excl - flag indicating exclusive or non-exclusive mode.
1252 * mode - mode to open file with.
1253 * cr - credentials of caller.
1254 * flag - large file flag [UNUSED].
1255 * ct - caller context
1256 * vsecp - ACL to be set
1258 * OUT: vpp - vnode of created or trunc'd entry.
1260 * RETURN: 0 if success
1261 * error code if failure
1264 * dvp - ctime|mtime updated if new entry created
1265 * vp - ctime|mtime always, atime if new
1270 zfs_create(vnode_t *dvp, char *name, vattr_t *vap, int excl, int mode,
1271 vnode_t **vpp, cred_t *cr, kthread_t *td)
1273 znode_t *zp, *dzp = VTOZ(dvp);
1274 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
1280 zfs_acl_t *aclp = NULL;
1281 zfs_fuid_info_t *fuidp = NULL;
1286 * If we have an ephemeral id, ACL, or XVATTR then
1287 * make sure file system is at proper version
1290 if (zfsvfs->z_use_fuids == B_FALSE &&
1291 (vsecp || (vap->va_mask & AT_XVATTR) ||
1292 IS_EPHEMERAL(crgetuid(cr)) || IS_EPHEMERAL(crgetgid(cr))))
1298 zilog = zfsvfs->z_log;
1300 if (zfsvfs->z_utf8 && u8_validate(name, strlen(name),
1301 NULL, U8_VALIDATE_ENTIRE, &error) < 0) {
1306 if (vap->va_mask & AT_XVATTR) {
1307 if ((error = secpolicy_xvattr((xvattr_t *)vap,
1308 crgetuid(cr), cr, vap->va_type)) != 0) {
1316 if ((vap->va_mode & S_ISVTX) && secpolicy_vnode_stky_modify(cr))
1317 vap->va_mode &= ~S_ISVTX;
1319 if (*name == '\0') {
1321 * Null component name refers to the directory itself.
1328 /* possible VN_HOLD(zp) */
1331 if (flag & FIGNORECASE)
1334 error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg,
1337 if (strcmp(name, "..") == 0)
1345 if (vsecp && aclp == NULL) {
1346 error = zfs_vsec_2_aclp(zfsvfs, vap->va_type, vsecp, &aclp);
1350 zfs_dirent_unlock(dl);
1359 * Create a new file object and update the directory
1362 if (error = zfs_zaccess(dzp, ACE_ADD_FILE, 0, B_FALSE, cr)) {
1367 * We only support the creation of regular files in
1368 * extended attribute directories.
1370 if ((dzp->z_phys->zp_flags & ZFS_XATTR) &&
1371 (vap->va_type != VREG)) {
1376 tx = dmu_tx_create(os);
1377 dmu_tx_hold_bonus(tx, DMU_NEW_OBJECT);
1378 if ((aclp && aclp->z_has_fuids) || IS_EPHEMERAL(crgetuid(cr)) ||
1379 IS_EPHEMERAL(crgetgid(cr))) {
1380 if (zfsvfs->z_fuid_obj == 0) {
1381 dmu_tx_hold_bonus(tx, DMU_NEW_OBJECT);
1382 dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0,
1383 FUID_SIZE_ESTIMATE(zfsvfs));
1384 dmu_tx_hold_zap(tx, MASTER_NODE_OBJ,
1387 dmu_tx_hold_bonus(tx, zfsvfs->z_fuid_obj);
1388 dmu_tx_hold_write(tx, zfsvfs->z_fuid_obj, 0,
1389 FUID_SIZE_ESTIMATE(zfsvfs));
1392 dmu_tx_hold_bonus(tx, dzp->z_id);
1393 dmu_tx_hold_zap(tx, dzp->z_id, TRUE, name);
1394 if ((dzp->z_phys->zp_flags & ZFS_INHERIT_ACE) || aclp) {
1395 dmu_tx_hold_write(tx, DMU_NEW_OBJECT,
1396 0, SPA_MAXBLOCKSIZE);
1398 error = dmu_tx_assign(tx, zfsvfs->z_assign);
1400 zfs_dirent_unlock(dl);
1401 if (error == ERESTART &&
1402 zfsvfs->z_assign == TXG_NOWAIT) {
1413 zfs_mknode(dzp, vap, tx, cr, 0, &zp, 0, aclp, &fuidp);
1414 (void) zfs_link_create(dl, zp, tx, ZNEW);
1415 txtype = zfs_log_create_txtype(Z_FILE, vsecp, vap);
1416 if (flag & FIGNORECASE)
1418 zfs_log_create(zilog, tx, txtype, dzp, zp, name,
1421 zfs_fuid_info_free(fuidp);
1424 int aflags = (flag & FAPPEND) ? V_APPEND : 0;
1427 * A directory entry already exists for this name.
1430 * Can't truncate an existing file if in exclusive mode.
1437 * Can't open a directory for writing.
1439 if ((ZTOV(zp)->v_type == VDIR) && (mode & S_IWRITE)) {
1444 * Verify requested access to file.
1446 if (mode && (error = zfs_zaccess_rwx(zp, mode, aflags, cr))) {
1450 mutex_enter(&dzp->z_lock);
1452 mutex_exit(&dzp->z_lock);
1455 * Truncate regular files if requested.
1457 if ((ZTOV(zp)->v_type == VREG) &&
1458 (vap->va_mask & AT_SIZE) && (vap->va_size == 0)) {
1459 /* we can't hold any locks when calling zfs_freesp() */
1460 zfs_dirent_unlock(dl);
1462 error = zfs_freesp(zp, 0, 0, mode, TRUE);
1464 vnevent_create(ZTOV(zp), ct);
1470 zfs_dirent_unlock(dl);
1478 * If vnode is for a device return a specfs vnode instead.
1480 if (IS_DEVVP(*vpp)) {
1483 svp = specvp(*vpp, (*vpp)->v_rdev, (*vpp)->v_type, cr);
1499 * Remove an entry from a directory.
1501 * IN: dvp - vnode of directory to remove entry from.
1502 * name - name of entry to remove.
1503 * cr - credentials of caller.
1504 * ct - caller context
1505 * flags - case flags
1507 * RETURN: 0 if success
1508 * error code if failure
1512 * vp - ctime (if nlink > 0)
1516 zfs_remove(vnode_t *dvp, char *name, cred_t *cr, caller_context_t *ct,
1519 znode_t *zp, *dzp = VTOZ(dvp);
1520 znode_t *xzp = NULL;
1522 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
1524 uint64_t acl_obj, xattr_obj;
1527 boolean_t may_delete_now, delete_now = FALSE;
1528 boolean_t unlinked, toobig = FALSE;
1530 pathname_t *realnmp = NULL;
1537 zilog = zfsvfs->z_log;
1539 if (flags & FIGNORECASE) {
1547 * Attempt to lock directory; fail if entry doesn't exist.
1549 if (error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg,
1559 if (error = zfs_zaccess_delete(dzp, zp, cr)) {
1564 * Need to use rmdir for removing directories.
1566 if (vp->v_type == VDIR) {
1571 vnevent_remove(vp, dvp, name, ct);
1574 dnlc_remove(dvp, realnmp->pn_buf);
1576 dnlc_remove(dvp, name);
1578 may_delete_now = FALSE;
1581 * We may delete the znode now, or we may put it in the unlinked set;
1582 * it depends on whether we're the last link, and on whether there are
1583 * other holds on the vnode. So we dmu_tx_hold() the right things to
1584 * allow for either case.
1586 tx = dmu_tx_create(zfsvfs->z_os);
1587 dmu_tx_hold_zap(tx, dzp->z_id, FALSE, name);
1588 dmu_tx_hold_bonus(tx, zp->z_id);
1589 if (may_delete_now) {
1591 zp->z_phys->zp_size > zp->z_blksz * DMU_MAX_DELETEBLKCNT;
1592 /* if the file is too big, only hold_free a token amount */
1593 dmu_tx_hold_free(tx, zp->z_id, 0,
1594 (toobig ? DMU_MAX_ACCESS : DMU_OBJECT_END));
1597 /* are there any extended attributes? */
1598 if ((xattr_obj = zp->z_phys->zp_xattr) != 0) {
1599 /* XXX - do we need this if we are deleting? */
1600 dmu_tx_hold_bonus(tx, xattr_obj);
1603 /* are there any additional acls */
1604 if ((acl_obj = zp->z_phys->zp_acl.z_acl_extern_obj) != 0 &&
1606 dmu_tx_hold_free(tx, acl_obj, 0, DMU_OBJECT_END);
1608 /* charge as an update -- would be nice not to charge at all */
1609 dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL);
1611 error = dmu_tx_assign(tx, zfsvfs->z_assign);
1613 zfs_dirent_unlock(dl);
1615 if (error == ERESTART && zfsvfs->z_assign == TXG_NOWAIT) {
1628 * Remove the directory entry.
1630 error = zfs_link_destroy(dl, zp, tx, zflg, &unlinked);
1637 if (0 && unlinked) {
1639 delete_now = may_delete_now && !toobig &&
1640 vp->v_count == 1 && !vn_has_cached_data(vp) &&
1641 zp->z_phys->zp_xattr == xattr_obj &&
1642 zp->z_phys->zp_acl.z_acl_extern_obj == acl_obj;
1647 if (zp->z_phys->zp_xattr) {
1648 error = zfs_zget(zfsvfs, zp->z_phys->zp_xattr, &xzp);
1649 ASSERT3U(error, ==, 0);
1650 ASSERT3U(xzp->z_phys->zp_links, ==, 2);
1651 dmu_buf_will_dirty(xzp->z_dbuf, tx);
1652 mutex_enter(&xzp->z_lock);
1653 xzp->z_unlinked = 1;
1654 xzp->z_phys->zp_links = 0;
1655 mutex_exit(&xzp->z_lock);
1656 zfs_unlinked_add(xzp, tx);
1657 zp->z_phys->zp_xattr = 0; /* probably unnecessary */
1659 mutex_enter(&zp->z_lock);
1662 ASSERT3U(vp->v_count, ==, 0);
1664 mutex_exit(&zp->z_lock);
1665 zfs_znode_delete(zp, tx);
1666 } else if (unlinked) {
1667 zfs_unlinked_add(zp, tx);
1671 if (flags & FIGNORECASE)
1673 zfs_log_remove(zilog, tx, txtype, dzp, name);
1680 zfs_dirent_unlock(dl);
1685 /* this rele is delayed to prevent nesting transactions */
1694 * Create a new directory and insert it into dvp using the name
1695 * provided. Return a pointer to the inserted directory.
1697 * IN: dvp - vnode of directory to add subdir to.
1698 * dirname - name of new directory.
1699 * vap - attributes of new directory.
1700 * cr - credentials of caller.
1701 * ct - caller context
1702 * vsecp - ACL to be set
1704 * OUT: vpp - vnode of created directory.
1706 * RETURN: 0 if success
1707 * error code if failure
1710 * dvp - ctime|mtime updated
1711 * vp - ctime|mtime|atime updated
1715 zfs_mkdir(vnode_t *dvp, char *dirname, vattr_t *vap, vnode_t **vpp, cred_t *cr,
1716 caller_context_t *ct, int flags, vsecattr_t *vsecp)
1718 znode_t *zp, *dzp = VTOZ(dvp);
1719 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
1725 zfs_acl_t *aclp = NULL;
1726 zfs_fuid_info_t *fuidp = NULL;
1729 ASSERT(vap->va_type == VDIR);
1732 * If we have an ephemeral id, ACL, or XVATTR then
1733 * make sure file system is at proper version
1736 if (zfsvfs->z_use_fuids == B_FALSE &&
1737 (vsecp || (vap->va_mask & AT_XVATTR) || IS_EPHEMERAL(crgetuid(cr))||
1738 IS_EPHEMERAL(crgetgid(cr))))
1743 zilog = zfsvfs->z_log;
1745 if (dzp->z_phys->zp_flags & ZFS_XATTR) {
1750 if (zfsvfs->z_utf8 && u8_validate(dirname,
1751 strlen(dirname), NULL, U8_VALIDATE_ENTIRE, &error) < 0) {
1755 if (flags & FIGNORECASE)
1758 if (vap->va_mask & AT_XVATTR)
1759 if ((error = secpolicy_xvattr((xvattr_t *)vap,
1760 crgetuid(cr), cr, vap->va_type)) != 0) {
1766 * First make sure the new directory doesn't exist.
1771 if (error = zfs_dirent_lock(&dl, dzp, dirname, &zp, zf,
1777 if (error = zfs_zaccess(dzp, ACE_ADD_SUBDIRECTORY, 0, B_FALSE, cr)) {
1778 zfs_dirent_unlock(dl);
1783 if (vsecp && aclp == NULL) {
1784 error = zfs_vsec_2_aclp(zfsvfs, vap->va_type, vsecp, &aclp);
1786 zfs_dirent_unlock(dl);
1792 * Add a new entry to the directory.
1794 tx = dmu_tx_create(zfsvfs->z_os);
1795 dmu_tx_hold_zap(tx, dzp->z_id, TRUE, dirname);
1796 dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL);
1797 if ((aclp && aclp->z_has_fuids) || IS_EPHEMERAL(crgetuid(cr)) ||
1798 IS_EPHEMERAL(crgetgid(cr))) {
1799 if (zfsvfs->z_fuid_obj == 0) {
1800 dmu_tx_hold_bonus(tx, DMU_NEW_OBJECT);
1801 dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0,
1802 FUID_SIZE_ESTIMATE(zfsvfs));
1803 dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, FALSE, NULL);
1805 dmu_tx_hold_bonus(tx, zfsvfs->z_fuid_obj);
1806 dmu_tx_hold_write(tx, zfsvfs->z_fuid_obj, 0,
1807 FUID_SIZE_ESTIMATE(zfsvfs));
1810 if ((dzp->z_phys->zp_flags & ZFS_INHERIT_ACE) || aclp)
1811 dmu_tx_hold_write(tx, DMU_NEW_OBJECT,
1812 0, SPA_MAXBLOCKSIZE);
1813 error = dmu_tx_assign(tx, zfsvfs->z_assign);
1815 zfs_dirent_unlock(dl);
1816 if (error == ERESTART && zfsvfs->z_assign == TXG_NOWAIT) {
1831 zfs_mknode(dzp, vap, tx, cr, 0, &zp, 0, aclp, &fuidp);
1837 * Now put new name in parent dir.
1839 (void) zfs_link_create(dl, zp, tx, ZNEW);
1843 txtype = zfs_log_create_txtype(Z_DIR, vsecp, vap);
1844 if (flags & FIGNORECASE)
1846 zfs_log_create(zilog, tx, txtype, dzp, zp, dirname, vsecp, fuidp, vap);
1849 zfs_fuid_info_free(fuidp);
1852 zfs_dirent_unlock(dl);
1859 * Remove a directory subdir entry. If the current working
1860 * directory is the same as the subdir to be removed, the
1863 * IN: dvp - vnode of directory to remove from.
1864 * name - name of directory to be removed.
1865 * cwd - vnode of current working directory.
1866 * cr - credentials of caller.
1867 * ct - caller context
1868 * flags - case flags
1870 * RETURN: 0 if success
1871 * error code if failure
1874 * dvp - ctime|mtime updated
1878 zfs_rmdir(vnode_t *dvp, char *name, vnode_t *cwd, cred_t *cr,
1879 caller_context_t *ct, int flags)
1881 znode_t *dzp = VTOZ(dvp);
1884 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
1893 zilog = zfsvfs->z_log;
1895 if (flags & FIGNORECASE)
1901 * Attempt to lock directory; fail if entry doesn't exist.
1903 if (error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg,
1911 if (error = zfs_zaccess_delete(dzp, zp, cr)) {
1915 if (vp->v_type != VDIR) {
1925 vnevent_rmdir(vp, dvp, name, ct);
1928 * Grab a lock on the directory to make sure that noone is
1929 * trying to add (or lookup) entries while we are removing it.
1931 rw_enter(&zp->z_name_lock, RW_WRITER);
1934 * Grab a lock on the parent pointer to make sure we play well
1935 * with the treewalk and directory rename code.
1937 rw_enter(&zp->z_parent_lock, RW_WRITER);
1939 tx = dmu_tx_create(zfsvfs->z_os);
1940 dmu_tx_hold_zap(tx, dzp->z_id, FALSE, name);
1941 dmu_tx_hold_bonus(tx, zp->z_id);
1942 dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL);
1943 error = dmu_tx_assign(tx, zfsvfs->z_assign);
1945 rw_exit(&zp->z_parent_lock);
1946 rw_exit(&zp->z_name_lock);
1947 zfs_dirent_unlock(dl);
1949 if (error == ERESTART && zfsvfs->z_assign == TXG_NOWAIT) {
1959 #ifdef FREEBSD_NAMECACHE
1963 error = zfs_link_destroy(dl, zp, tx, zflg, NULL);
1966 uint64_t txtype = TX_RMDIR;
1967 if (flags & FIGNORECASE)
1969 zfs_log_remove(zilog, tx, txtype, dzp, name);
1974 rw_exit(&zp->z_parent_lock);
1975 rw_exit(&zp->z_name_lock);
1976 #ifdef FREEBSD_NAMECACHE
1980 zfs_dirent_unlock(dl);
1989 * Read as many directory entries as will fit into the provided
1990 * buffer from the given directory cursor position (specified in
1991 * the uio structure.
1993 * IN: vp - vnode of directory to read.
1994 * uio - structure supplying read location, range info,
1995 * and return buffer.
1996 * cr - credentials of caller.
1997 * ct - caller context
1998 * flags - case flags
2000 * OUT: uio - updated offset and range, buffer filled.
2001 * eofp - set to true if end-of-file detected.
2003 * RETURN: 0 if success
2004 * error code if failure
2007 * vp - atime updated
2009 * Note that the low 4 bits of the cookie returned by zap is always zero.
2010 * This allows us to use the low range for "special" directory entries:
2011 * We use 0 for '.', and 1 for '..'. If this is the root of the filesystem,
2012 * we use the offset 2 for the '.zfs' directory.
2016 zfs_readdir(vnode_t *vp, uio_t *uio, cred_t *cr, int *eofp, int *ncookies, u_long **cookies)
2018 znode_t *zp = VTOZ(vp);
2022 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
2027 zap_attribute_t zap;
2028 uint_t bytes_wanted;
2029 uint64_t offset; /* must be unsigned; checks for < 1 */
2034 boolean_t check_sysattrs;
2037 u_long *cooks = NULL;
2044 * If we are not given an eof variable,
2051 * Check for valid iov_len.
2053 if (uio->uio_iov->iov_len <= 0) {
2059 * Quit if directory has been removed (posix)
2061 if ((*eofp = zp->z_unlinked) != 0) {
2068 offset = uio->uio_loffset;
2069 prefetch = zp->z_zn_prefetch;
2072 * Initialize the iterator cursor.
2076 * Start iteration from the beginning of the directory.
2078 zap_cursor_init(&zc, os, zp->z_id);
2081 * The offset is a serialized cursor.
2083 zap_cursor_init_serialized(&zc, os, zp->z_id, offset);
2087 * Get space to change directory entries into fs independent format.
2089 iovp = uio->uio_iov;
2090 bytes_wanted = iovp->iov_len;
2091 if (uio->uio_segflg != UIO_SYSSPACE || uio->uio_iovcnt != 1) {
2092 bufsize = bytes_wanted;
2093 outbuf = kmem_alloc(bufsize, KM_SLEEP);
2094 odp = (struct dirent64 *)outbuf;
2096 bufsize = bytes_wanted;
2097 odp = (struct dirent64 *)iovp->iov_base;
2099 eodp = (struct edirent *)odp;
2101 if (ncookies != NULL) {
2103 * Minimum entry size is dirent size and 1 byte for a file name.
2105 ncooks = uio->uio_resid / (sizeof(struct dirent) - sizeof(((struct dirent *)NULL)->d_name) + 1);
2106 cooks = malloc(ncooks * sizeof(u_long), M_TEMP, M_WAITOK);
2111 * If this VFS supports the system attribute view interface; and
2112 * we're looking at an extended attribute directory; and we care
2113 * about normalization conflicts on this vfs; then we must check
2114 * for normalization conflicts with the sysattr name space.
2117 check_sysattrs = vfs_has_feature(vp->v_vfsp, VFSFT_SYSATTR_VIEWS) &&
2118 (vp->v_flag & V_XATTRDIR) && zfsvfs->z_norm &&
2119 (flags & V_RDDIR_ENTFLAGS);
2125 * Transform to file-system independent format
2128 while (outcount < bytes_wanted) {
2134 * Special case `.', `..', and `.zfs'.
2137 (void) strcpy(zap.za_name, ".");
2138 zap.za_normalization_conflict = 0;
2141 } else if (offset == 1) {
2142 (void) strcpy(zap.za_name, "..");
2143 zap.za_normalization_conflict = 0;
2144 objnum = zp->z_phys->zp_parent;
2146 } else if (offset == 2 && zfs_show_ctldir(zp)) {
2147 (void) strcpy(zap.za_name, ZFS_CTLDIR_NAME);
2148 zap.za_normalization_conflict = 0;
2149 objnum = ZFSCTL_INO_ROOT;
2155 if (error = zap_cursor_retrieve(&zc, &zap)) {
2156 if ((*eofp = (error == ENOENT)) != 0)
2162 if (zap.za_integer_length != 8 ||
2163 zap.za_num_integers != 1) {
2164 cmn_err(CE_WARN, "zap_readdir: bad directory "
2165 "entry, obj = %lld, offset = %lld\n",
2166 (u_longlong_t)zp->z_id,
2167 (u_longlong_t)offset);
2172 objnum = ZFS_DIRENT_OBJ(zap.za_first_integer);
2174 * MacOS X can extract the object type here such as:
2175 * uint8_t type = ZFS_DIRENT_TYPE(zap.za_first_integer);
2177 type = ZFS_DIRENT_TYPE(zap.za_first_integer);
2179 if (check_sysattrs && !zap.za_normalization_conflict) {
2181 zap.za_normalization_conflict =
2182 xattr_sysattr_casechk(zap.za_name);
2184 panic("%s:%u: TODO", __func__, __LINE__);
2189 if (flags & V_RDDIR_ENTFLAGS)
2190 reclen = EDIRENT_RECLEN(strlen(zap.za_name));
2192 reclen = DIRENT64_RECLEN(strlen(zap.za_name));
2195 * Will this entry fit in the buffer?
2197 if (outcount + reclen > bufsize) {
2199 * Did we manage to fit anything in the buffer?
2207 if (flags & V_RDDIR_ENTFLAGS) {
2209 * Add extended flag entry:
2211 eodp->ed_ino = objnum;
2212 eodp->ed_reclen = reclen;
2213 /* NOTE: ed_off is the offset for the *next* entry */
2214 next = &(eodp->ed_off);
2215 eodp->ed_eflags = zap.za_normalization_conflict ?
2216 ED_CASE_CONFLICT : 0;
2217 (void) strncpy(eodp->ed_name, zap.za_name,
2218 EDIRENT_NAMELEN(reclen));
2219 eodp = (edirent_t *)((intptr_t)eodp + reclen);
2224 odp->d_ino = objnum;
2225 odp->d_reclen = reclen;
2226 odp->d_namlen = strlen(zap.za_name);
2227 (void) strlcpy(odp->d_name, zap.za_name, odp->d_namlen + 1);
2229 odp = (dirent64_t *)((intptr_t)odp + reclen);
2233 ASSERT(outcount <= bufsize);
2235 /* Prefetch znode */
2237 dmu_prefetch(os, objnum, 0, 0);
2240 * Move to the next entry, fill in the previous offset.
2242 if (offset > 2 || (offset == 2 && !zfs_show_ctldir(zp))) {
2243 zap_cursor_advance(&zc);
2244 offset = zap_cursor_serialize(&zc);
2249 if (cooks != NULL) {
2252 KASSERT(ncooks >= 0, ("ncookies=%d", ncooks));
2255 zp->z_zn_prefetch = B_FALSE; /* a lookup will re-enable pre-fetching */
2257 /* Subtract unused cookies */
2258 if (ncookies != NULL)
2259 *ncookies -= ncooks;
2261 if (uio->uio_segflg == UIO_SYSSPACE && uio->uio_iovcnt == 1) {
2262 iovp->iov_base += outcount;
2263 iovp->iov_len -= outcount;
2264 uio->uio_resid -= outcount;
2265 } else if (error = uiomove(outbuf, (long)outcount, UIO_READ, uio)) {
2267 * Reset the pointer.
2269 offset = uio->uio_loffset;
2273 zap_cursor_fini(&zc);
2274 if (uio->uio_segflg != UIO_SYSSPACE || uio->uio_iovcnt != 1)
2275 kmem_free(outbuf, bufsize);
2277 if (error == ENOENT)
2280 ZFS_ACCESSTIME_STAMP(zfsvfs, zp);
2282 uio->uio_loffset = offset;
2284 if (error != 0 && cookies != NULL) {
2285 free(*cookies, M_TEMP);
2292 ulong_t zfs_fsync_sync_cnt = 4;
2295 zfs_fsync(vnode_t *vp, int syncflag, cred_t *cr, caller_context_t *ct)
2297 znode_t *zp = VTOZ(vp);
2298 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
2300 (void) tsd_set(zfs_fsyncer_key, (void *)zfs_fsync_sync_cnt);
2304 zil_commit(zfsvfs->z_log, zp->z_last_itx, zp->z_id);
2311 * Get the requested file attributes and place them in the provided
2314 * IN: vp - vnode of file.
2315 * vap - va_mask identifies requested attributes.
2316 * If AT_XVATTR set, then optional attrs are requested
2317 * flags - ATTR_NOACLCHECK (CIFS server context)
2318 * cr - credentials of caller.
2319 * ct - caller context
2321 * OUT: vap - attribute values.
2323 * RETURN: 0 (always succeeds)
2327 zfs_getattr(vnode_t *vp, vattr_t *vap, int flags, cred_t *cr,
2328 caller_context_t *ct)
2330 znode_t *zp = VTOZ(vp);
2331 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
2335 u_longlong_t nblocks;
2337 xvattr_t *xvap = (xvattr_t *)vap; /* vap may be an xvattr_t * */
2338 xoptattr_t *xoap = NULL;
2339 boolean_t skipaclchk = (flags & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE;
2345 mutex_enter(&zp->z_lock);
2348 * If ACL is trivial don't bother looking for ACE_READ_ATTRIBUTES.
2349 * Also, if we are the owner don't bother, since owner should
2350 * always be allowed to read basic attributes of file.
2352 if (!(pzp->zp_flags & ZFS_ACL_TRIVIAL) &&
2353 (pzp->zp_uid != crgetuid(cr))) {
2354 if (error = zfs_zaccess(zp, ACE_READ_ATTRIBUTES, 0,
2356 mutex_exit(&zp->z_lock);
2363 * Return all attributes. It's cheaper to provide the answer
2364 * than to determine whether we were asked the question.
2367 vap->va_type = IFTOVT(pzp->zp_mode);
2368 vap->va_mode = pzp->zp_mode & ~S_IFMT;
2369 zfs_fuid_map_ids(zp, cr, &vap->va_uid, &vap->va_gid);
2370 // vap->va_fsid = zp->z_zfsvfs->z_vfs->vfs_dev;
2371 vap->va_nodeid = zp->z_id;
2372 if ((vp->v_flag & VROOT) && zfs_show_ctldir(zp))
2373 links = pzp->zp_links + 1;
2375 links = pzp->zp_links;
2376 vap->va_nlink = MIN(links, UINT32_MAX); /* nlink_t limit! */
2377 vap->va_size = pzp->zp_size;
2378 vap->va_fsid = vp->v_mount->mnt_stat.f_fsid.val[0];
2379 vap->va_rdev = zfs_cmpldev(pzp->zp_rdev);
2380 vap->va_seq = zp->z_seq;
2381 vap->va_flags = 0; /* FreeBSD: Reset chflags(2) flags. */
2384 * Add in any requested optional attributes and the create time.
2385 * Also set the corresponding bits in the returned attribute bitmap.
2387 if ((xoap = xva_getxoptattr(xvap)) != NULL && zfsvfs->z_use_fuids) {
2388 if (XVA_ISSET_REQ(xvap, XAT_ARCHIVE)) {
2390 ((pzp->zp_flags & ZFS_ARCHIVE) != 0);
2391 XVA_SET_RTN(xvap, XAT_ARCHIVE);
2394 if (XVA_ISSET_REQ(xvap, XAT_READONLY)) {
2395 xoap->xoa_readonly =
2396 ((pzp->zp_flags & ZFS_READONLY) != 0);
2397 XVA_SET_RTN(xvap, XAT_READONLY);
2400 if (XVA_ISSET_REQ(xvap, XAT_SYSTEM)) {
2402 ((pzp->zp_flags & ZFS_SYSTEM) != 0);
2403 XVA_SET_RTN(xvap, XAT_SYSTEM);
2406 if (XVA_ISSET_REQ(xvap, XAT_HIDDEN)) {
2408 ((pzp->zp_flags & ZFS_HIDDEN) != 0);
2409 XVA_SET_RTN(xvap, XAT_HIDDEN);
2412 if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK)) {
2413 xoap->xoa_nounlink =
2414 ((pzp->zp_flags & ZFS_NOUNLINK) != 0);
2415 XVA_SET_RTN(xvap, XAT_NOUNLINK);
2418 if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE)) {
2419 xoap->xoa_immutable =
2420 ((pzp->zp_flags & ZFS_IMMUTABLE) != 0);
2421 XVA_SET_RTN(xvap, XAT_IMMUTABLE);
2424 if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY)) {
2425 xoap->xoa_appendonly =
2426 ((pzp->zp_flags & ZFS_APPENDONLY) != 0);
2427 XVA_SET_RTN(xvap, XAT_APPENDONLY);
2430 if (XVA_ISSET_REQ(xvap, XAT_NODUMP)) {
2432 ((pzp->zp_flags & ZFS_NODUMP) != 0);
2433 XVA_SET_RTN(xvap, XAT_NODUMP);
2436 if (XVA_ISSET_REQ(xvap, XAT_OPAQUE)) {
2438 ((pzp->zp_flags & ZFS_OPAQUE) != 0);
2439 XVA_SET_RTN(xvap, XAT_OPAQUE);
2442 if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED)) {
2443 xoap->xoa_av_quarantined =
2444 ((pzp->zp_flags & ZFS_AV_QUARANTINED) != 0);
2445 XVA_SET_RTN(xvap, XAT_AV_QUARANTINED);
2448 if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED)) {
2449 xoap->xoa_av_modified =
2450 ((pzp->zp_flags & ZFS_AV_MODIFIED) != 0);
2451 XVA_SET_RTN(xvap, XAT_AV_MODIFIED);
2454 if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP) &&
2455 vp->v_type == VREG &&
2456 (pzp->zp_flags & ZFS_BONUS_SCANSTAMP)) {
2458 dmu_object_info_t doi;
2461 * Only VREG files have anti-virus scanstamps, so we
2462 * won't conflict with symlinks in the bonus buffer.
2464 dmu_object_info_from_db(zp->z_dbuf, &doi);
2465 len = sizeof (xoap->xoa_av_scanstamp) +
2466 sizeof (znode_phys_t);
2467 if (len <= doi.doi_bonus_size) {
2469 * pzp points to the start of the
2470 * znode_phys_t. pzp + 1 points to the
2471 * first byte after the znode_phys_t.
2473 (void) memcpy(xoap->xoa_av_scanstamp,
2475 sizeof (xoap->xoa_av_scanstamp));
2476 XVA_SET_RTN(xvap, XAT_AV_SCANSTAMP);
2480 if (XVA_ISSET_REQ(xvap, XAT_CREATETIME)) {
2481 ZFS_TIME_DECODE(&xoap->xoa_createtime, pzp->zp_crtime);
2482 XVA_SET_RTN(xvap, XAT_CREATETIME);
2486 ZFS_TIME_DECODE(&vap->va_atime, pzp->zp_atime);
2487 ZFS_TIME_DECODE(&vap->va_mtime, pzp->zp_mtime);
2488 ZFS_TIME_DECODE(&vap->va_ctime, pzp->zp_ctime);
2489 ZFS_TIME_DECODE(&vap->va_birthtime, pzp->zp_crtime);
2491 mutex_exit(&zp->z_lock);
2493 dmu_object_size_from_db(zp->z_dbuf, &blksize, &nblocks);
2494 vap->va_blksize = blksize;
2495 vap->va_bytes = nblocks << 9; /* nblocks * 512 */
2497 if (zp->z_blksz == 0) {
2499 * Block size hasn't been set; suggest maximal I/O transfers.
2501 vap->va_blksize = zfsvfs->z_max_blksz;
2509 * Set the file attributes to the values contained in the
2512 * IN: vp - vnode of file to be modified.
2513 * vap - new attribute values.
2514 * If AT_XVATTR set, then optional attrs are being set
2515 * flags - ATTR_UTIME set if non-default time values provided.
2516 * - ATTR_NOACLCHECK (CIFS context only).
2517 * cr - credentials of caller.
2518 * ct - caller context
2520 * RETURN: 0 if success
2521 * error code if failure
2524 * vp - ctime updated, mtime updated if size changed.
2528 zfs_setattr(vnode_t *vp, vattr_t *vap, int flags, cred_t *cr,
2529 caller_context_t *ct)
2531 znode_t *zp = VTOZ(vp);
2533 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
2537 uint_t mask = vap->va_mask;
2542 int need_policy = FALSE;
2544 zfs_fuid_info_t *fuidp = NULL;
2545 xvattr_t *xvap = (xvattr_t *)vap; /* vap may be an xvattr_t * */
2547 zfs_acl_t *aclp = NULL;
2548 boolean_t skipaclchk = (flags & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE;
2553 if (mask & AT_NOSET)
2560 zilog = zfsvfs->z_log;
2563 * Make sure that if we have ephemeral uid/gid or xvattr specified
2564 * that file system is at proper version level
2567 if (zfsvfs->z_use_fuids == B_FALSE &&
2568 (((mask & AT_UID) && IS_EPHEMERAL(vap->va_uid)) ||
2569 ((mask & AT_GID) && IS_EPHEMERAL(vap->va_gid)) ||
2570 (mask & AT_XVATTR))) {
2575 if (mask & AT_SIZE && vp->v_type == VDIR) {
2580 if (mask & AT_SIZE && vp->v_type != VREG && vp->v_type != VFIFO) {
2586 * If this is an xvattr_t, then get a pointer to the structure of
2587 * optional attributes. If this is NULL, then we have a vattr_t.
2589 xoap = xva_getxoptattr(xvap);
2592 * Immutable files can only alter immutable bit and atime
2594 if ((pzp->zp_flags & ZFS_IMMUTABLE) &&
2595 ((mask & (AT_SIZE|AT_UID|AT_GID|AT_MTIME|AT_MODE)) ||
2596 ((mask & AT_XVATTR) && XVA_ISSET_REQ(xvap, XAT_CREATETIME)))) {
2601 if ((mask & AT_SIZE) && (pzp->zp_flags & ZFS_READONLY)) {
2607 * Verify timestamps doesn't overflow 32 bits.
2608 * ZFS can handle large timestamps, but 32bit syscalls can't
2609 * handle times greater than 2039. This check should be removed
2610 * once large timestamps are fully supported.
2612 if (mask & (AT_ATIME | AT_MTIME)) {
2613 if (((mask & AT_ATIME) && TIMESPEC_OVERFLOW(&vap->va_atime)) ||
2614 ((mask & AT_MTIME) && TIMESPEC_OVERFLOW(&vap->va_mtime))) {
2623 if (zfsvfs->z_vfs->vfs_flag & VFS_RDONLY) {
2629 * First validate permissions
2632 if (mask & AT_SIZE) {
2633 err = zfs_zaccess(zp, ACE_WRITE_DATA, 0, skipaclchk, cr);
2639 * XXX - Note, we are not providing any open
2640 * mode flags here (like FNDELAY), so we may
2641 * block if there are locks present... this
2642 * should be addressed in openat().
2644 /* XXX - would it be OK to generate a log record here? */
2645 err = zfs_freesp(zp, vap->va_size, 0, 0, FALSE);
2652 if (mask & (AT_ATIME|AT_MTIME) ||
2653 ((mask & AT_XVATTR) && (XVA_ISSET_REQ(xvap, XAT_HIDDEN) ||
2654 XVA_ISSET_REQ(xvap, XAT_READONLY) ||
2655 XVA_ISSET_REQ(xvap, XAT_ARCHIVE) ||
2656 XVA_ISSET_REQ(xvap, XAT_CREATETIME) ||
2657 XVA_ISSET_REQ(xvap, XAT_SYSTEM))))
2658 need_policy = zfs_zaccess(zp, ACE_WRITE_ATTRIBUTES, 0,
2661 if (mask & (AT_UID|AT_GID)) {
2662 int idmask = (mask & (AT_UID|AT_GID));
2667 * NOTE: even if a new mode is being set,
2668 * we may clear S_ISUID/S_ISGID bits.
2671 if (!(mask & AT_MODE))
2672 vap->va_mode = pzp->zp_mode;
2675 * Take ownership or chgrp to group we are a member of
2678 take_owner = (mask & AT_UID) && (vap->va_uid == crgetuid(cr));
2679 take_group = (mask & AT_GID) &&
2680 zfs_groupmember(zfsvfs, vap->va_gid, cr);
2683 * If both AT_UID and AT_GID are set then take_owner and
2684 * take_group must both be set in order to allow taking
2687 * Otherwise, send the check through secpolicy_vnode_setattr()
2691 if (((idmask == (AT_UID|AT_GID)) && take_owner && take_group) ||
2692 ((idmask == AT_UID) && take_owner) ||
2693 ((idmask == AT_GID) && take_group)) {
2694 if (zfs_zaccess(zp, ACE_WRITE_OWNER, 0,
2695 skipaclchk, cr) == 0) {
2697 * Remove setuid/setgid for non-privileged users
2699 secpolicy_setid_clear(vap, vp, cr);
2700 trim_mask = (mask & (AT_UID|AT_GID));
2709 mutex_enter(&zp->z_lock);
2710 oldva.va_mode = pzp->zp_mode;
2711 zfs_fuid_map_ids(zp, cr, &oldva.va_uid, &oldva.va_gid);
2712 if (mask & AT_XVATTR) {
2713 if ((need_policy == FALSE) &&
2714 (XVA_ISSET_REQ(xvap, XAT_APPENDONLY) &&
2715 xoap->xoa_appendonly !=
2716 ((pzp->zp_flags & ZFS_APPENDONLY) != 0)) ||
2717 (XVA_ISSET_REQ(xvap, XAT_NOUNLINK) &&
2718 xoap->xoa_nounlink !=
2719 ((pzp->zp_flags & ZFS_NOUNLINK) != 0)) ||
2720 (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE) &&
2721 xoap->xoa_immutable !=
2722 ((pzp->zp_flags & ZFS_IMMUTABLE) != 0)) ||
2723 (XVA_ISSET_REQ(xvap, XAT_NODUMP) &&
2725 ((pzp->zp_flags & ZFS_NODUMP) != 0)) ||
2726 (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED) &&
2727 xoap->xoa_av_modified !=
2728 ((pzp->zp_flags & ZFS_AV_MODIFIED) != 0)) ||
2729 ((XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED) &&
2730 ((vp->v_type != VREG && xoap->xoa_av_quarantined) ||
2731 xoap->xoa_av_quarantined !=
2732 ((pzp->zp_flags & ZFS_AV_QUARANTINED) != 0)))) ||
2733 (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP)) ||
2734 (XVA_ISSET_REQ(xvap, XAT_OPAQUE))) {
2739 mutex_exit(&zp->z_lock);
2741 if (mask & AT_MODE) {
2742 if (zfs_zaccess(zp, ACE_WRITE_ACL, 0, skipaclchk, cr) == 0) {
2743 err = secpolicy_setid_setsticky_clear(vp, vap,
2749 trim_mask |= AT_MODE;
2757 * If trim_mask is set then take ownership
2758 * has been granted or write_acl is present and user
2759 * has the ability to modify mode. In that case remove
2760 * UID|GID and or MODE from mask so that
2761 * secpolicy_vnode_setattr() doesn't revoke it.
2765 saved_mask = vap->va_mask;
2766 vap->va_mask &= ~trim_mask;
2768 err = secpolicy_vnode_setattr(cr, vp, vap, &oldva, flags,
2769 (int (*)(void *, int, cred_t *))zfs_zaccess_unix, zp);
2776 vap->va_mask |= saved_mask;
2780 * secpolicy_vnode_setattr, or take ownership may have
2783 mask = vap->va_mask;
2785 tx = dmu_tx_create(zfsvfs->z_os);
2786 dmu_tx_hold_bonus(tx, zp->z_id);
2787 if (((mask & AT_UID) && IS_EPHEMERAL(vap->va_uid)) ||
2788 ((mask & AT_GID) && IS_EPHEMERAL(vap->va_gid))) {
2789 if (zfsvfs->z_fuid_obj == 0) {
2790 dmu_tx_hold_bonus(tx, DMU_NEW_OBJECT);
2791 dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0,
2792 FUID_SIZE_ESTIMATE(zfsvfs));
2793 dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, FALSE, NULL);
2795 dmu_tx_hold_bonus(tx, zfsvfs->z_fuid_obj);
2796 dmu_tx_hold_write(tx, zfsvfs->z_fuid_obj, 0,
2797 FUID_SIZE_ESTIMATE(zfsvfs));
2801 if (mask & AT_MODE) {
2802 uint64_t pmode = pzp->zp_mode;
2804 new_mode = (pmode & S_IFMT) | (vap->va_mode & ~S_IFMT);
2806 if (err = zfs_acl_chmod_setattr(zp, &aclp, new_mode)) {
2811 if (pzp->zp_acl.z_acl_extern_obj) {
2812 /* Are we upgrading ACL from old V0 format to new V1 */
2813 if (zfsvfs->z_version <= ZPL_VERSION_FUID &&
2814 pzp->zp_acl.z_acl_version ==
2815 ZFS_ACL_VERSION_INITIAL) {
2816 dmu_tx_hold_free(tx,
2817 pzp->zp_acl.z_acl_extern_obj, 0,
2819 dmu_tx_hold_write(tx, DMU_NEW_OBJECT,
2820 0, aclp->z_acl_bytes);
2822 dmu_tx_hold_write(tx,
2823 pzp->zp_acl.z_acl_extern_obj, 0,
2826 } else if (aclp->z_acl_bytes > ZFS_ACE_SPACE) {
2827 dmu_tx_hold_write(tx, DMU_NEW_OBJECT,
2828 0, aclp->z_acl_bytes);
2832 if ((mask & (AT_UID | AT_GID)) && pzp->zp_xattr != 0) {
2833 err = zfs_zget(zp->z_zfsvfs, pzp->zp_xattr, &attrzp);
2841 dmu_tx_hold_bonus(tx, attrzp->z_id);
2844 err = dmu_tx_assign(tx, zfsvfs->z_assign);
2847 VN_RELE(ZTOV(attrzp));
2854 if (err == ERESTART && zfsvfs->z_assign == TXG_NOWAIT) {
2864 dmu_buf_will_dirty(zp->z_dbuf, tx);
2867 * Set each attribute requested.
2868 * We group settings according to the locks they need to acquire.
2870 * Note: you cannot set ctime directly, although it will be
2871 * updated as a side-effect of calling this function.
2874 mutex_enter(&zp->z_lock);
2876 if (mask & AT_MODE) {
2877 mutex_enter(&zp->z_acl_lock);
2878 zp->z_phys->zp_mode = new_mode;
2879 err = zfs_aclset_common(zp, aclp, cr, &fuidp, tx);
2880 ASSERT3U(err, ==, 0);
2881 mutex_exit(&zp->z_acl_lock);
2885 mutex_enter(&attrzp->z_lock);
2887 if (mask & AT_UID) {
2888 pzp->zp_uid = zfs_fuid_create(zfsvfs,
2889 vap->va_uid, cr, ZFS_OWNER, tx, &fuidp);
2891 attrzp->z_phys->zp_uid = zfs_fuid_create(zfsvfs,
2892 vap->va_uid, cr, ZFS_OWNER, tx, &fuidp);
2896 if (mask & AT_GID) {
2897 pzp->zp_gid = zfs_fuid_create(zfsvfs, vap->va_gid,
2898 cr, ZFS_GROUP, tx, &fuidp);
2900 attrzp->z_phys->zp_gid = zfs_fuid_create(zfsvfs,
2901 vap->va_gid, cr, ZFS_GROUP, tx, &fuidp);
2908 mutex_exit(&attrzp->z_lock);
2910 if (mask & AT_ATIME)
2911 ZFS_TIME_ENCODE(&vap->va_atime, pzp->zp_atime);
2913 if (mask & AT_MTIME)
2914 ZFS_TIME_ENCODE(&vap->va_mtime, pzp->zp_mtime);
2916 /* XXX - shouldn't this be done *before* the ATIME/MTIME checks? */
2918 zfs_time_stamper_locked(zp, CONTENT_MODIFIED, tx);
2920 zfs_time_stamper_locked(zp, STATE_CHANGED, tx);
2922 * Do this after setting timestamps to prevent timestamp
2923 * update from toggling bit
2926 if (xoap && (mask & AT_XVATTR)) {
2927 if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP)) {
2929 dmu_object_info_t doi;
2931 ASSERT(vp->v_type == VREG);
2933 /* Grow the bonus buffer if necessary. */
2934 dmu_object_info_from_db(zp->z_dbuf, &doi);
2935 len = sizeof (xoap->xoa_av_scanstamp) +
2936 sizeof (znode_phys_t);
2937 if (len > doi.doi_bonus_size)
2938 VERIFY(dmu_set_bonus(zp->z_dbuf, len, tx) == 0);
2940 zfs_xvattr_set(zp, xvap);
2944 zfs_log_setattr(zilog, tx, TX_SETATTR, zp, vap, mask, fuidp);
2947 zfs_fuid_info_free(fuidp);
2948 mutex_exit(&zp->z_lock);
2951 VN_RELE(ZTOV(attrzp));
2959 typedef struct zfs_zlock {
2960 krwlock_t *zl_rwlock; /* lock we acquired */
2961 znode_t *zl_znode; /* znode we held */
2962 struct zfs_zlock *zl_next; /* next in list */
2966 * Drop locks and release vnodes that were held by zfs_rename_lock().
2969 zfs_rename_unlock(zfs_zlock_t **zlpp)
2973 while ((zl = *zlpp) != NULL) {
2974 if (zl->zl_znode != NULL)
2975 VN_RELE(ZTOV(zl->zl_znode));
2976 rw_exit(zl->zl_rwlock);
2977 *zlpp = zl->zl_next;
2978 kmem_free(zl, sizeof (*zl));
2983 * Search back through the directory tree, using the ".." entries.
2984 * Lock each directory in the chain to prevent concurrent renames.
2985 * Fail any attempt to move a directory into one of its own descendants.
2986 * XXX - z_parent_lock can overlap with map or grow locks
2989 zfs_rename_lock(znode_t *szp, znode_t *tdzp, znode_t *sdzp, zfs_zlock_t **zlpp)
2993 uint64_t rootid = zp->z_zfsvfs->z_root;
2994 uint64_t *oidp = &zp->z_id;
2995 krwlock_t *rwlp = &szp->z_parent_lock;
2996 krw_t rw = RW_WRITER;
2999 * First pass write-locks szp and compares to zp->z_id.
3000 * Later passes read-lock zp and compare to zp->z_parent.
3003 if (!rw_tryenter(rwlp, rw)) {
3005 * Another thread is renaming in this path.
3006 * Note that if we are a WRITER, we don't have any
3007 * parent_locks held yet.
3009 if (rw == RW_READER && zp->z_id > szp->z_id) {
3011 * Drop our locks and restart
3013 zfs_rename_unlock(&zl);
3017 rwlp = &szp->z_parent_lock;
3022 * Wait for other thread to drop its locks
3028 zl = kmem_alloc(sizeof (*zl), KM_SLEEP);
3029 zl->zl_rwlock = rwlp;
3030 zl->zl_znode = NULL;
3031 zl->zl_next = *zlpp;
3034 if (*oidp == szp->z_id) /* We're a descendant of szp */
3037 if (*oidp == rootid) /* We've hit the top */
3040 if (rw == RW_READER) { /* i.e. not the first pass */
3041 int error = zfs_zget(zp->z_zfsvfs, *oidp, &zp);
3046 oidp = &zp->z_phys->zp_parent;
3047 rwlp = &zp->z_parent_lock;
3050 } while (zp->z_id != sdzp->z_id);
3056 * Move an entry from the provided source directory to the target
3057 * directory. Change the entry name as indicated.
3059 * IN: sdvp - Source directory containing the "old entry".
3060 * snm - Old entry name.
3061 * tdvp - Target directory to contain the "new entry".
3062 * tnm - New entry name.
3063 * cr - credentials of caller.
3064 * ct - caller context
3065 * flags - case flags
3067 * RETURN: 0 if success
3068 * error code if failure
3071 * sdvp,tdvp - ctime|mtime updated
3075 zfs_rename(vnode_t *sdvp, char *snm, vnode_t *tdvp, char *tnm, cred_t *cr,
3076 caller_context_t *ct, int flags)
3078 znode_t *tdzp, *szp, *tzp;
3079 znode_t *sdzp = VTOZ(sdvp);
3080 zfsvfs_t *zfsvfs = sdzp->z_zfsvfs;
3083 zfs_dirlock_t *sdl, *tdl;
3086 int cmp, serr, terr;
3091 ZFS_VERIFY_ZP(sdzp);
3092 zilog = zfsvfs->z_log;
3095 * Make sure we have the real vp for the target directory.
3097 if (VOP_REALVP(tdvp, &realvp, ct) == 0)
3100 if (tdvp->v_vfsp != sdvp->v_vfsp) {
3106 ZFS_VERIFY_ZP(tdzp);
3107 if (zfsvfs->z_utf8 && u8_validate(tnm,
3108 strlen(tnm), NULL, U8_VALIDATE_ENTIRE, &error) < 0) {
3113 if (flags & FIGNORECASE)
3122 * This is to prevent the creation of links into attribute space
3123 * by renaming a linked file into/outof an attribute directory.
3124 * See the comment in zfs_link() for why this is considered bad.
3126 if ((tdzp->z_phys->zp_flags & ZFS_XATTR) !=
3127 (sdzp->z_phys->zp_flags & ZFS_XATTR)) {
3133 * Lock source and target directory entries. To prevent deadlock,
3134 * a lock ordering must be defined. We lock the directory with
3135 * the smallest object id first, or if it's a tie, the one with
3136 * the lexically first name.
3138 if (sdzp->z_id < tdzp->z_id) {
3140 } else if (sdzp->z_id > tdzp->z_id) {
3144 * First compare the two name arguments without
3145 * considering any case folding.
3147 int nofold = (zfsvfs->z_norm & ~U8_TEXTPREP_TOUPPER);
3149 cmp = u8_strcmp(snm, tnm, 0, nofold, U8_UNICODE_LATEST, &error);
3150 ASSERT(error == 0 || !zfsvfs->z_utf8);
3153 * POSIX: "If the old argument and the new argument
3154 * both refer to links to the same existing file,
3155 * the rename() function shall return successfully
3156 * and perform no other action."
3162 * If the file system is case-folding, then we may
3163 * have some more checking to do. A case-folding file
3164 * system is either supporting mixed case sensitivity
3165 * access or is completely case-insensitive. Note
3166 * that the file system is always case preserving.
3168 * In mixed sensitivity mode case sensitive behavior
3169 * is the default. FIGNORECASE must be used to
3170 * explicitly request case insensitive behavior.
3172 * If the source and target names provided differ only
3173 * by case (e.g., a request to rename 'tim' to 'Tim'),
3174 * we will treat this as a special case in the
3175 * case-insensitive mode: as long as the source name
3176 * is an exact match, we will allow this to proceed as
3177 * a name-change request.
3179 if ((zfsvfs->z_case == ZFS_CASE_INSENSITIVE ||
3180 (zfsvfs->z_case == ZFS_CASE_MIXED &&
3181 flags & FIGNORECASE)) &&
3182 u8_strcmp(snm, tnm, 0, zfsvfs->z_norm, U8_UNICODE_LATEST,
3185 * case preserving rename request, require exact
3194 serr = zfs_dirent_lock(&sdl, sdzp, snm, &szp,
3195 ZEXISTS | zflg, NULL, NULL);
3196 terr = zfs_dirent_lock(&tdl,
3197 tdzp, tnm, &tzp, ZRENAMING | zflg, NULL, NULL);
3199 terr = zfs_dirent_lock(&tdl,
3200 tdzp, tnm, &tzp, zflg, NULL, NULL);
3201 serr = zfs_dirent_lock(&sdl,
3202 sdzp, snm, &szp, ZEXISTS | ZRENAMING | zflg,
3208 * Source entry invalid or not there.
3211 zfs_dirent_unlock(tdl);
3215 if (strcmp(snm, ".") == 0 || strcmp(snm, "..") == 0)
3221 zfs_dirent_unlock(sdl);
3223 if (strcmp(tnm, "..") == 0)
3230 * Must have write access at the source to remove the old entry
3231 * and write access at the target to create the new entry.
3232 * Note that if target and source are the same, this can be
3233 * done in a single check.
3236 if (error = zfs_zaccess_rename(sdzp, szp, tdzp, tzp, cr))
3239 if (ZTOV(szp)->v_type == VDIR) {
3241 * Check to make sure rename is valid.
3242 * Can't do a move like this: /usr/a/b to /usr/a/b/c/d
3244 if (error = zfs_rename_lock(szp, tdzp, sdzp, &zl))
3249 * Does target exist?
3253 * Source and target must be the same type.
3255 if (ZTOV(szp)->v_type == VDIR) {
3256 if (ZTOV(tzp)->v_type != VDIR) {
3261 if (ZTOV(tzp)->v_type == VDIR) {
3267 * POSIX dictates that when the source and target
3268 * entries refer to the same file object, rename
3269 * must do nothing and exit without error.
3271 if (szp->z_id == tzp->z_id) {
3277 vnevent_rename_src(ZTOV(szp), sdvp, snm, ct);
3279 vnevent_rename_dest(ZTOV(tzp), tdvp, tnm, ct);
3282 * notify the target directory if it is not the same
3283 * as source directory.
3286 vnevent_rename_dest_dir(tdvp, ct);
3289 tx = dmu_tx_create(zfsvfs->z_os);
3290 dmu_tx_hold_bonus(tx, szp->z_id); /* nlink changes */
3291 dmu_tx_hold_bonus(tx, sdzp->z_id); /* nlink changes */
3292 dmu_tx_hold_zap(tx, sdzp->z_id, FALSE, snm);
3293 dmu_tx_hold_zap(tx, tdzp->z_id, TRUE, tnm);
3295 dmu_tx_hold_bonus(tx, tdzp->z_id); /* nlink changes */
3297 dmu_tx_hold_bonus(tx, tzp->z_id); /* parent changes */
3298 dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL);
3299 error = dmu_tx_assign(tx, zfsvfs->z_assign);
3302 zfs_rename_unlock(&zl);
3303 zfs_dirent_unlock(sdl);
3304 zfs_dirent_unlock(tdl);
3308 if (error == ERESTART && zfsvfs->z_assign == TXG_NOWAIT) {
3318 if (tzp) /* Attempt to remove the existing target */
3319 error = zfs_link_destroy(tdl, tzp, tx, zflg, NULL);
3322 error = zfs_link_create(tdl, szp, tx, ZRENAMING);
3324 szp->z_phys->zp_flags |= ZFS_AV_MODIFIED;
3326 error = zfs_link_destroy(sdl, szp, tx, ZRENAMING, NULL);
3329 zfs_log_rename(zilog, tx,
3330 TX_RENAME | (flags & FIGNORECASE ? TX_CI : 0),
3331 sdzp, sdl->dl_name, tdzp, tdl->dl_name, szp);
3333 /* Update path information for the target vnode */
3334 vn_renamepath(tdvp, ZTOV(szp), tnm, strlen(tnm));
3336 #ifdef FREEBSD_NAMECACHE
3347 zfs_rename_unlock(&zl);
3349 zfs_dirent_unlock(sdl);
3350 zfs_dirent_unlock(tdl);
3362 * Insert the indicated symbolic reference entry into the directory.
3364 * IN: dvp - Directory to contain new symbolic link.
3365 * link - Name for new symlink entry.
3366 * vap - Attributes of new entry.
3367 * target - Target path of new symlink.
3368 * cr - credentials of caller.
3369 * ct - caller context
3370 * flags - case flags
3372 * RETURN: 0 if success
3373 * error code if failure
3376 * dvp - ctime|mtime updated
3380 zfs_symlink(vnode_t *dvp, vnode_t **vpp, char *name, vattr_t *vap, char *link,
3381 cred_t *cr, kthread_t *td)
3383 znode_t *zp, *dzp = VTOZ(dvp);
3386 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
3388 int len = strlen(link);
3391 zfs_fuid_info_t *fuidp = NULL;
3394 ASSERT(vap->va_type == VLNK);
3398 zilog = zfsvfs->z_log;
3400 if (zfsvfs->z_utf8 && u8_validate(name, strlen(name),
3401 NULL, U8_VALIDATE_ENTIRE, &error) < 0) {
3405 if (flags & FIGNORECASE)
3408 if (error = zfs_zaccess(dzp, ACE_ADD_FILE, 0, B_FALSE, cr)) {
3413 if (len > MAXPATHLEN) {
3415 return (ENAMETOOLONG);
3419 * Attempt to lock directory; fail if entry already exists.
3421 error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg, NULL, NULL);
3427 tx = dmu_tx_create(zfsvfs->z_os);
3428 dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, MAX(1, len));
3429 dmu_tx_hold_bonus(tx, dzp->z_id);
3430 dmu_tx_hold_zap(tx, dzp->z_id, TRUE, name);
3431 if (dzp->z_phys->zp_flags & ZFS_INHERIT_ACE)
3432 dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, SPA_MAXBLOCKSIZE);
3433 if (IS_EPHEMERAL(crgetuid(cr)) || IS_EPHEMERAL(crgetgid(cr))) {
3434 if (zfsvfs->z_fuid_obj == 0) {
3435 dmu_tx_hold_bonus(tx, DMU_NEW_OBJECT);
3436 dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0,
3437 FUID_SIZE_ESTIMATE(zfsvfs));
3438 dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, FALSE, NULL);
3440 dmu_tx_hold_bonus(tx, zfsvfs->z_fuid_obj);
3441 dmu_tx_hold_write(tx, zfsvfs->z_fuid_obj, 0,
3442 FUID_SIZE_ESTIMATE(zfsvfs));
3445 error = dmu_tx_assign(tx, zfsvfs->z_assign);
3447 zfs_dirent_unlock(dl);
3448 if (error == ERESTART && zfsvfs->z_assign == TXG_NOWAIT) {
3458 dmu_buf_will_dirty(dzp->z_dbuf, tx);
3461 * Create a new object for the symlink.
3462 * Put the link content into bonus buffer if it will fit;
3463 * otherwise, store it just like any other file data.
3465 if (sizeof (znode_phys_t) + len <= dmu_bonus_max()) {
3466 zfs_mknode(dzp, vap, tx, cr, 0, &zp, len, NULL, &fuidp);
3468 bcopy(link, zp->z_phys + 1, len);
3472 zfs_mknode(dzp, vap, tx, cr, 0, &zp, 0, NULL, &fuidp);
3474 * Nothing can access the znode yet so no locking needed
3475 * for growing the znode's blocksize.
3477 zfs_grow_blocksize(zp, len, tx);
3479 VERIFY(0 == dmu_buf_hold(zfsvfs->z_os,
3480 zp->z_id, 0, FTAG, &dbp));
3481 dmu_buf_will_dirty(dbp, tx);
3483 ASSERT3U(len, <=, dbp->db_size);
3484 bcopy(link, dbp->db_data, len);
3485 dmu_buf_rele(dbp, FTAG);
3487 zp->z_phys->zp_size = len;
3490 * Insert the new object into the directory.
3492 (void) zfs_link_create(dl, zp, tx, ZNEW);
3495 uint64_t txtype = TX_SYMLINK;
3496 if (flags & FIGNORECASE)
3498 zfs_log_symlink(zilog, tx, txtype, dzp, zp, name, link);
3502 zfs_fuid_info_free(fuidp);
3506 zfs_dirent_unlock(dl);
3513 * Return, in the buffer contained in the provided uio structure,
3514 * the symbolic path referred to by vp.
3516 * IN: vp - vnode of symbolic link.
3517 * uoip - structure to contain the link path.
3518 * cr - credentials of caller.
3519 * ct - caller context
3521 * OUT: uio - structure to contain the link path.
3523 * RETURN: 0 if success
3524 * error code if failure
3527 * vp - atime updated
3531 zfs_readlink(vnode_t *vp, uio_t *uio, cred_t *cr, caller_context_t *ct)
3533 znode_t *zp = VTOZ(vp);
3534 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
3541 bufsz = (size_t)zp->z_phys->zp_size;
3542 if (bufsz + sizeof (znode_phys_t) <= zp->z_dbuf->db_size) {
3543 error = uiomove(zp->z_phys + 1,
3544 MIN((size_t)bufsz, uio->uio_resid), UIO_READ, uio);
3547 error = dmu_buf_hold(zfsvfs->z_os, zp->z_id, 0, FTAG, &dbp);
3552 error = uiomove(dbp->db_data,
3553 MIN((size_t)bufsz, uio->uio_resid), UIO_READ, uio);
3554 dmu_buf_rele(dbp, FTAG);
3557 ZFS_ACCESSTIME_STAMP(zfsvfs, zp);
3563 * Insert a new entry into directory tdvp referencing svp.
3565 * IN: tdvp - Directory to contain new entry.
3566 * svp - vnode of new entry.
3567 * name - name of new entry.
3568 * cr - credentials of caller.
3569 * ct - caller context
3571 * RETURN: 0 if success
3572 * error code if failure
3575 * tdvp - ctime|mtime updated
3576 * svp - ctime updated
3580 zfs_link(vnode_t *tdvp, vnode_t *svp, char *name, cred_t *cr,
3581 caller_context_t *ct, int flags)
3583 znode_t *dzp = VTOZ(tdvp);
3585 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
3594 ASSERT(tdvp->v_type == VDIR);
3598 zilog = zfsvfs->z_log;
3600 if (VOP_REALVP(svp, &realvp, ct) == 0)
3603 if (svp->v_vfsp != tdvp->v_vfsp) {
3610 if (zfsvfs->z_utf8 && u8_validate(name,
3611 strlen(name), NULL, U8_VALIDATE_ENTIRE, &error) < 0) {
3615 if (flags & FIGNORECASE)
3620 * We do not support links between attributes and non-attributes
3621 * because of the potential security risk of creating links
3622 * into "normal" file space in order to circumvent restrictions
3623 * imposed in attribute space.
3625 if ((szp->z_phys->zp_flags & ZFS_XATTR) !=
3626 (dzp->z_phys->zp_flags & ZFS_XATTR)) {
3632 * POSIX dictates that we return EPERM here.
3633 * Better choices include ENOTSUP or EISDIR.
3635 if (svp->v_type == VDIR) {
3640 owner = zfs_fuid_map_id(zfsvfs, szp->z_phys->zp_uid, cr, ZFS_OWNER);
3641 if (owner != crgetuid(cr) &&
3642 secpolicy_basic_link(svp, cr) != 0) {
3647 if (error = zfs_zaccess(dzp, ACE_ADD_FILE, 0, B_FALSE, cr)) {
3653 * Attempt to lock directory; fail if entry already exists.
3655 error = zfs_dirent_lock(&dl, dzp, name, &tzp, zf, NULL, NULL);
3661 tx = dmu_tx_create(zfsvfs->z_os);
3662 dmu_tx_hold_bonus(tx, szp->z_id);
3663 dmu_tx_hold_zap(tx, dzp->z_id, TRUE, name);
3664 error = dmu_tx_assign(tx, zfsvfs->z_assign);
3666 zfs_dirent_unlock(dl);
3667 if (error == ERESTART && zfsvfs->z_assign == TXG_NOWAIT) {
3677 error = zfs_link_create(dl, szp, tx, 0);
3680 uint64_t txtype = TX_LINK;
3681 if (flags & FIGNORECASE)
3683 zfs_log_link(zilog, tx, txtype, dzp, szp, name);
3688 zfs_dirent_unlock(dl);
3691 vnevent_link(svp, ct);
3700 zfs_inactive(vnode_t *vp, cred_t *cr, caller_context_t *ct)
3702 znode_t *zp = VTOZ(vp);
3703 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
3706 rw_enter(&zfsvfs->z_teardown_inactive_lock, RW_READER);
3707 if (zp->z_dbuf == NULL) {
3709 * The fs has been unmounted, or we did a
3710 * suspend/resume and this file no longer exists.
3712 mutex_enter(&zp->z_lock);
3714 vp->v_count = 0; /* count arrives as 1 */
3715 mutex_exit(&zp->z_lock);
3716 rw_exit(&zfsvfs->z_teardown_inactive_lock);
3721 if (zp->z_atime_dirty && zp->z_unlinked == 0) {
3722 dmu_tx_t *tx = dmu_tx_create(zfsvfs->z_os);
3724 dmu_tx_hold_bonus(tx, zp->z_id);
3725 error = dmu_tx_assign(tx, TXG_WAIT);
3729 dmu_buf_will_dirty(zp->z_dbuf, tx);
3730 mutex_enter(&zp->z_lock);
3731 zp->z_atime_dirty = 0;
3732 mutex_exit(&zp->z_lock);
3738 rw_exit(&zfsvfs->z_teardown_inactive_lock);
3741 CTASSERT(sizeof(struct zfid_short) <= sizeof(struct fid));
3742 CTASSERT(sizeof(struct zfid_long) <= sizeof(struct fid));
3746 zfs_fid(vnode_t *vp, fid_t *fidp, caller_context_t *ct)
3748 znode_t *zp = VTOZ(vp);
3749 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
3751 uint64_t object = zp->z_id;
3757 gen = (uint32_t)zp->z_gen;
3759 size = (zfsvfs->z_parent != zfsvfs) ? LONG_FID_LEN : SHORT_FID_LEN;
3760 fidp->fid_len = size;
3762 zfid = (zfid_short_t *)fidp;
3764 zfid->zf_len = size;
3766 for (i = 0; i < sizeof (zfid->zf_object); i++)
3767 zfid->zf_object[i] = (uint8_t)(object >> (8 * i));
3769 /* Must have a non-zero generation number to distinguish from .zfs */
3772 for (i = 0; i < sizeof (zfid->zf_gen); i++)
3773 zfid->zf_gen[i] = (uint8_t)(gen >> (8 * i));
3775 if (size == LONG_FID_LEN) {
3776 uint64_t objsetid = dmu_objset_id(zfsvfs->z_os);
3779 zlfid = (zfid_long_t *)fidp;
3781 for (i = 0; i < sizeof (zlfid->zf_setid); i++)
3782 zlfid->zf_setid[i] = (uint8_t)(objsetid >> (8 * i));
3784 /* XXX - this should be the generation number for the objset */
3785 for (i = 0; i < sizeof (zlfid->zf_setgen); i++)
3786 zlfid->zf_setgen[i] = 0;
3794 zfs_pathconf(vnode_t *vp, int cmd, ulong_t *valp, cred_t *cr,
3795 caller_context_t *ct)
3807 case _PC_FILESIZEBITS:
3812 case _PC_XATTR_EXISTS:
3814 zfsvfs = zp->z_zfsvfs;
3818 error = zfs_dirent_lock(&dl, zp, "", &xzp,
3819 ZXATTR | ZEXISTS | ZSHARED, NULL, NULL);
3821 zfs_dirent_unlock(dl);
3822 if (!zfs_dirempty(xzp))
3825 } else if (error == ENOENT) {
3827 * If there aren't extended attributes, it's the
3828 * same as having zero of them.
3836 case _PC_ACL_EXTENDED:
3837 *valp = 0; /* TODO */
3840 case _PC_MIN_HOLE_SIZE:
3841 *valp = (int)SPA_MINBLOCKSIZE;
3845 return (EOPNOTSUPP);
3851 zfs_getsecattr(vnode_t *vp, vsecattr_t *vsecp, int flag, cred_t *cr,
3852 caller_context_t *ct)
3854 znode_t *zp = VTOZ(vp);
3855 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
3857 boolean_t skipaclchk = (flag & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE;
3861 error = zfs_getacl(zp, vsecp, skipaclchk, cr);
3869 zfs_setsecattr(vnode_t *vp, vsecattr_t *vsecp, int flag, cred_t *cr,
3870 caller_context_t *ct)
3872 znode_t *zp = VTOZ(vp);
3873 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
3875 boolean_t skipaclchk = (flag & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE;
3879 error = zfs_setacl(zp, vsecp, skipaclchk, cr);
3885 zfs_freebsd_open(ap)
3886 struct vop_open_args /* {
3889 struct ucred *a_cred;
3890 struct thread *a_td;
3893 vnode_t *vp = ap->a_vp;
3894 znode_t *zp = VTOZ(vp);
3897 error = zfs_open(&vp, ap->a_mode, ap->a_cred, NULL);
3899 vnode_create_vobject(vp, zp->z_phys->zp_size, ap->a_td);
3904 zfs_freebsd_close(ap)
3905 struct vop_close_args /* {
3908 struct ucred *a_cred;
3909 struct thread *a_td;
3913 return (zfs_close(ap->a_vp, ap->a_fflag, 0, 0, ap->a_cred, NULL));
3917 zfs_freebsd_ioctl(ap)
3918 struct vop_ioctl_args /* {
3928 return (zfs_ioctl(ap->a_vp, ap->a_command, (intptr_t)ap->a_data,
3929 ap->a_fflag, ap->a_cred, NULL, NULL));
3933 zfs_freebsd_read(ap)
3934 struct vop_read_args /* {
3938 struct ucred *a_cred;
3942 return (zfs_read(ap->a_vp, ap->a_uio, ap->a_ioflag, ap->a_cred, NULL));
3946 zfs_freebsd_write(ap)
3947 struct vop_write_args /* {
3951 struct ucred *a_cred;
3955 return (zfs_write(ap->a_vp, ap->a_uio, ap->a_ioflag, ap->a_cred, NULL));
3959 zfs_freebsd_access(ap)
3960 struct vop_access_args /* {
3962 accmode_t a_accmode;
3963 struct ucred *a_cred;
3964 struct thread *a_td;
3969 * ZFS itself only knowns about VREAD, VWRITE and VEXEC, the rest
3970 * we have to handle by calling vaccess().
3972 if ((ap->a_accmode & ~(VREAD|VWRITE|VEXEC)) != 0) {
3973 vnode_t *vp = ap->a_vp;
3974 znode_t *zp = VTOZ(vp);
3975 znode_phys_t *zphys = zp->z_phys;
3977 return (vaccess(vp->v_type, zphys->zp_mode, zphys->zp_uid,
3978 zphys->zp_gid, ap->a_accmode, ap->a_cred, NULL));
3981 return (zfs_access(ap->a_vp, ap->a_accmode, 0, ap->a_cred, NULL));
3985 zfs_freebsd_lookup(ap)
3986 struct vop_lookup_args /* {
3987 struct vnode *a_dvp;
3988 struct vnode **a_vpp;
3989 struct componentname *a_cnp;
3992 struct componentname *cnp = ap->a_cnp;
3993 char nm[NAME_MAX + 1];
3995 ASSERT(cnp->cn_namelen < sizeof(nm));
3996 strlcpy(nm, cnp->cn_nameptr, MIN(cnp->cn_namelen + 1, sizeof(nm)));
3998 return (zfs_lookup(ap->a_dvp, nm, ap->a_vpp, cnp, cnp->cn_nameiop,
3999 cnp->cn_cred, cnp->cn_thread, 0));
4003 zfs_freebsd_create(ap)
4004 struct vop_create_args /* {
4005 struct vnode *a_dvp;
4006 struct vnode **a_vpp;
4007 struct componentname *a_cnp;
4008 struct vattr *a_vap;
4011 struct componentname *cnp = ap->a_cnp;
4012 vattr_t *vap = ap->a_vap;
4015 ASSERT(cnp->cn_flags & SAVENAME);
4017 vattr_init_mask(vap);
4018 mode = vap->va_mode & ALLPERMS;
4020 return (zfs_create(ap->a_dvp, cnp->cn_nameptr, vap, !EXCL, mode,
4021 ap->a_vpp, cnp->cn_cred, cnp->cn_thread));
4025 zfs_freebsd_remove(ap)
4026 struct vop_remove_args /* {
4027 struct vnode *a_dvp;
4029 struct componentname *a_cnp;
4033 ASSERT(ap->a_cnp->cn_flags & SAVENAME);
4035 return (zfs_remove(ap->a_dvp, ap->a_cnp->cn_nameptr,
4036 ap->a_cnp->cn_cred, NULL, 0));
4040 zfs_freebsd_mkdir(ap)
4041 struct vop_mkdir_args /* {
4042 struct vnode *a_dvp;
4043 struct vnode **a_vpp;
4044 struct componentname *a_cnp;
4045 struct vattr *a_vap;
4048 vattr_t *vap = ap->a_vap;
4050 ASSERT(ap->a_cnp->cn_flags & SAVENAME);
4052 vattr_init_mask(vap);
4054 return (zfs_mkdir(ap->a_dvp, ap->a_cnp->cn_nameptr, vap, ap->a_vpp,
4055 ap->a_cnp->cn_cred, NULL, 0, NULL));
4059 zfs_freebsd_rmdir(ap)
4060 struct vop_rmdir_args /* {
4061 struct vnode *a_dvp;
4063 struct componentname *a_cnp;
4066 struct componentname *cnp = ap->a_cnp;
4068 ASSERT(cnp->cn_flags & SAVENAME);
4070 return (zfs_rmdir(ap->a_dvp, cnp->cn_nameptr, NULL, cnp->cn_cred, NULL, 0));
4074 zfs_freebsd_readdir(ap)
4075 struct vop_readdir_args /* {
4078 struct ucred *a_cred;
4085 return (zfs_readdir(ap->a_vp, ap->a_uio, ap->a_cred, ap->a_eofflag,
4086 ap->a_ncookies, ap->a_cookies));
4090 zfs_freebsd_fsync(ap)
4091 struct vop_fsync_args /* {
4094 struct thread *a_td;
4099 return (zfs_fsync(ap->a_vp, 0, ap->a_td->td_ucred, NULL));
4103 zfs_freebsd_getattr(ap)
4104 struct vop_getattr_args /* {
4106 struct vattr *a_vap;
4107 struct ucred *a_cred;
4108 struct thread *a_td;
4111 vattr_t *vap = ap->a_vap;
4117 xvap.xva_vattr = *vap;
4118 xvap.xva_vattr.va_mask |= AT_XVATTR;
4120 /* Convert chflags into ZFS-type flags. */
4121 /* XXX: what about SF_SETTABLE?. */
4122 XVA_SET_REQ(&xvap, XAT_IMMUTABLE);
4123 XVA_SET_REQ(&xvap, XAT_APPENDONLY);
4124 XVA_SET_REQ(&xvap, XAT_NOUNLINK);
4125 XVA_SET_REQ(&xvap, XAT_NODUMP);
4126 error = zfs_getattr(ap->a_vp, (vattr_t *)&xvap, 0, ap->a_cred, NULL);
4130 /* Convert ZFS xattr into chflags. */
4131 #define FLAG_CHECK(fflag, xflag, xfield) do { \
4132 if (XVA_ISSET_RTN(&xvap, (xflag)) && (xfield) != 0) \
4133 fflags |= (fflag); \
4135 FLAG_CHECK(SF_IMMUTABLE, XAT_IMMUTABLE,
4136 xvap.xva_xoptattrs.xoa_immutable);
4137 FLAG_CHECK(SF_APPEND, XAT_APPENDONLY,
4138 xvap.xva_xoptattrs.xoa_appendonly);
4139 FLAG_CHECK(SF_NOUNLINK, XAT_NOUNLINK,
4140 xvap.xva_xoptattrs.xoa_nounlink);
4141 FLAG_CHECK(UF_NODUMP, XAT_NODUMP,
4142 xvap.xva_xoptattrs.xoa_nodump);
4144 *vap = xvap.xva_vattr;
4145 vap->va_flags = fflags;
4150 zfs_freebsd_setattr(ap)
4151 struct vop_setattr_args /* {
4153 struct vattr *a_vap;
4154 struct ucred *a_cred;
4155 struct thread *a_td;
4158 vnode_t *vp = ap->a_vp;
4159 vattr_t *vap = ap->a_vap;
4160 cred_t *cred = ap->a_cred;
4165 vattr_init_mask(vap);
4166 vap->va_mask &= ~AT_NOSET;
4169 xvap.xva_vattr = *vap;
4171 zflags = VTOZ(vp)->z_phys->zp_flags;
4173 if (vap->va_flags != VNOVAL) {
4176 fflags = vap->va_flags;
4177 if ((fflags & ~(SF_IMMUTABLE|SF_APPEND|SF_NOUNLINK|UF_NODUMP)) != 0)
4178 return (EOPNOTSUPP);
4180 * Callers may only modify the file flags on objects they
4181 * have VADMIN rights for.
4183 if ((error = VOP_ACCESS(vp, VADMIN, cred, curthread)) != 0)
4186 * Unprivileged processes are not permitted to unset system
4187 * flags, or modify flags if any system flags are set.
4188 * Privileged non-jail processes may not modify system flags
4189 * if securelevel > 0 and any existing system flags are set.
4190 * Privileged jail processes behave like privileged non-jail
4191 * processes if the security.jail.chflags_allowed sysctl is
4192 * is non-zero; otherwise, they behave like unprivileged
4195 if (priv_check_cred(cred, PRIV_VFS_SYSFLAGS, 0) == 0) {
4197 (ZFS_IMMUTABLE | ZFS_APPENDONLY | ZFS_NOUNLINK)) {
4198 error = securelevel_gt(cred, 0);
4204 (ZFS_IMMUTABLE | ZFS_APPENDONLY | ZFS_NOUNLINK)) {
4208 (SF_IMMUTABLE | SF_APPEND | SF_NOUNLINK)) {
4213 #define FLAG_CHANGE(fflag, zflag, xflag, xfield) do { \
4214 if (((fflags & (fflag)) && !(zflags & (zflag))) || \
4215 ((zflags & (zflag)) && !(fflags & (fflag)))) { \
4216 XVA_SET_REQ(&xvap, (xflag)); \
4217 (xfield) = ((fflags & (fflag)) != 0); \
4220 /* Convert chflags into ZFS-type flags. */
4221 /* XXX: what about SF_SETTABLE?. */
4222 FLAG_CHANGE(SF_IMMUTABLE, ZFS_IMMUTABLE, XAT_IMMUTABLE,
4223 xvap.xva_xoptattrs.xoa_immutable);
4224 FLAG_CHANGE(SF_APPEND, ZFS_APPENDONLY, XAT_APPENDONLY,
4225 xvap.xva_xoptattrs.xoa_appendonly);
4226 FLAG_CHANGE(SF_NOUNLINK, ZFS_NOUNLINK, XAT_NOUNLINK,
4227 xvap.xva_xoptattrs.xoa_nounlink);
4228 FLAG_CHANGE(UF_NODUMP, ZFS_NODUMP, XAT_NODUMP,
4229 xvap.xva_xoptattrs.xoa_nodump);
4232 return (zfs_setattr(vp, (vattr_t *)&xvap, 0, cred, NULL));
4236 zfs_freebsd_rename(ap)
4237 struct vop_rename_args /* {
4238 struct vnode *a_fdvp;
4239 struct vnode *a_fvp;
4240 struct componentname *a_fcnp;
4241 struct vnode *a_tdvp;
4242 struct vnode *a_tvp;
4243 struct componentname *a_tcnp;
4246 vnode_t *fdvp = ap->a_fdvp;
4247 vnode_t *fvp = ap->a_fvp;
4248 vnode_t *tdvp = ap->a_tdvp;
4249 vnode_t *tvp = ap->a_tvp;
4252 ASSERT(ap->a_fcnp->cn_flags & (SAVENAME|SAVESTART));
4253 ASSERT(ap->a_tcnp->cn_flags & (SAVENAME|SAVESTART));
4255 error = zfs_rename(fdvp, ap->a_fcnp->cn_nameptr, tdvp,
4256 ap->a_tcnp->cn_nameptr, ap->a_fcnp->cn_cred, NULL, 0);
4271 zfs_freebsd_symlink(ap)
4272 struct vop_symlink_args /* {
4273 struct vnode *a_dvp;
4274 struct vnode **a_vpp;
4275 struct componentname *a_cnp;
4276 struct vattr *a_vap;
4280 struct componentname *cnp = ap->a_cnp;
4281 vattr_t *vap = ap->a_vap;
4283 ASSERT(cnp->cn_flags & SAVENAME);
4285 vap->va_type = VLNK; /* FreeBSD: Syscall only sets va_mode. */
4286 vattr_init_mask(vap);
4288 return (zfs_symlink(ap->a_dvp, ap->a_vpp, cnp->cn_nameptr, vap,
4289 ap->a_target, cnp->cn_cred, cnp->cn_thread));
4293 zfs_freebsd_readlink(ap)
4294 struct vop_readlink_args /* {
4297 struct ucred *a_cred;
4301 return (zfs_readlink(ap->a_vp, ap->a_uio, ap->a_cred, NULL));
4305 zfs_freebsd_link(ap)
4306 struct vop_link_args /* {
4307 struct vnode *a_tdvp;
4309 struct componentname *a_cnp;
4312 struct componentname *cnp = ap->a_cnp;
4314 ASSERT(cnp->cn_flags & SAVENAME);
4316 return (zfs_link(ap->a_tdvp, ap->a_vp, cnp->cn_nameptr, cnp->cn_cred, NULL, 0));
4320 zfs_freebsd_inactive(ap)
4321 struct vop_inactive_args /* {
4323 struct thread *a_td;
4326 vnode_t *vp = ap->a_vp;
4328 zfs_inactive(vp, ap->a_td->td_ucred, NULL);
4333 zfs_reclaim_complete(void *arg, int pending)
4336 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
4338 ZFS_LOG(1, "zp=%p", zp);
4339 ZFS_OBJ_HOLD_ENTER(zfsvfs, zp->z_id);
4340 zfs_znode_dmu_fini(zp);
4341 ZFS_OBJ_HOLD_EXIT(zfsvfs, zp->z_id);
4346 zfs_freebsd_reclaim(ap)
4347 struct vop_reclaim_args /* {
4349 struct thread *a_td;
4352 vnode_t *vp = ap->a_vp;
4353 znode_t *zp = VTOZ(vp);
4359 * Destroy the vm object and flush associated pages.
4361 vnode_destroy_vobject(vp);
4363 mutex_enter(&zp->z_lock);
4366 if (!zp->z_unlinked) {
4369 zfsvfs = zp->z_zfsvfs;
4370 mutex_exit(&zp->z_lock);
4371 locked = MUTEX_HELD(ZFS_OBJ_MUTEX(zfsvfs, zp->z_id)) ? 2 :
4372 ZFS_OBJ_HOLD_TRYENTER(zfsvfs, zp->z_id);
4375 * Lock can't be obtained due to deadlock possibility,
4376 * so defer znode destruction.
4378 TASK_INIT(&zp->z_task, 0, zfs_reclaim_complete, zp);
4379 taskqueue_enqueue(taskqueue_thread, &zp->z_task);
4381 zfs_znode_dmu_fini(zp);
4383 ZFS_OBJ_HOLD_EXIT(zfsvfs, zp->z_id);
4387 mutex_exit(&zp->z_lock);
4391 ASSERT(vp->v_holdcnt >= 1);
4398 struct vop_fid_args /* {
4404 return (zfs_fid(ap->a_vp, (void *)ap->a_fid, NULL));
4408 zfs_freebsd_pathconf(ap)
4409 struct vop_pathconf_args /* {
4412 register_t *a_retval;
4418 error = zfs_pathconf(ap->a_vp, ap->a_name, &val, curthread->td_ucred, NULL);
4420 *ap->a_retval = val;
4421 else if (error == EOPNOTSUPP)
4422 error = vop_stdpathconf(ap);
4427 * FreeBSD's extended attributes namespace defines file name prefix for ZFS'
4428 * extended attribute name:
4431 * system freebsd:system:
4432 * user (none, can be used to access ZFS fsattr(5) attributes
4433 * created on Solaris)
4436 zfs_create_attrname(int attrnamespace, const char *name, char *attrname,
4439 const char *namespace, *prefix, *suffix;
4441 /* We don't allow '/' character in attribute name. */
4442 if (strchr(name, '/') != NULL)
4444 /* We don't allow attribute names that start with "freebsd:" string. */
4445 if (strncmp(name, "freebsd:", 8) == 0)
4448 bzero(attrname, size);
4450 switch (attrnamespace) {
4451 case EXTATTR_NAMESPACE_USER:
4453 prefix = "freebsd:";
4454 namespace = EXTATTR_NAMESPACE_USER_STRING;
4458 * This is the default namespace by which we can access all
4459 * attributes created on Solaris.
4461 prefix = namespace = suffix = "";
4464 case EXTATTR_NAMESPACE_SYSTEM:
4465 prefix = "freebsd:";
4466 namespace = EXTATTR_NAMESPACE_SYSTEM_STRING;
4469 case EXTATTR_NAMESPACE_EMPTY:
4473 if (snprintf(attrname, size, "%s%s%s%s", prefix, namespace, suffix,
4475 return (ENAMETOOLONG);
4481 * Vnode operating to retrieve a named extended attribute.
4484 zfs_getextattr(struct vop_getextattr_args *ap)
4487 IN struct vnode *a_vp;
4488 IN int a_attrnamespace;
4489 IN const char *a_name;
4490 INOUT struct uio *a_uio;
4492 IN struct ucred *a_cred;
4493 IN struct thread *a_td;
4497 zfsvfs_t *zfsvfs = VTOZ(ap->a_vp)->z_zfsvfs;
4498 struct thread *td = ap->a_td;
4499 struct nameidata nd;
4502 vnode_t *xvp = NULL, *vp;
4505 error = extattr_check_cred(ap->a_vp, ap->a_attrnamespace,
4506 ap->a_cred, ap->a_td, VREAD);
4510 error = zfs_create_attrname(ap->a_attrnamespace, ap->a_name, attrname,
4517 error = zfs_lookup(ap->a_vp, NULL, &xvp, NULL, 0, ap->a_cred, td,
4525 NDINIT_ATVP(&nd, LOOKUP, NOFOLLOW | MPSAFE, UIO_SYSSPACE, attrname,
4527 error = vn_open_cred(&nd, &flags, 0, 0, ap->a_cred, NULL);
4529 NDFREE(&nd, NDF_ONLY_PNBUF);
4531 if (error == ENOENT)
4537 if (ap->a_size != NULL) {
4538 error = VOP_GETATTR(vp, &va, ap->a_cred);
4540 *ap->a_size = (size_t)va.va_size;
4541 } else if (ap->a_uio != NULL)
4542 error = VOP_READ(vp, ap->a_uio, IO_UNIT | IO_SYNC, ap->a_cred);
4545 vn_close(vp, flags, ap->a_cred, td);
4552 * Vnode operation to remove a named attribute.
4555 zfs_deleteextattr(struct vop_deleteextattr_args *ap)
4558 IN struct vnode *a_vp;
4559 IN int a_attrnamespace;
4560 IN const char *a_name;
4561 IN struct ucred *a_cred;
4562 IN struct thread *a_td;
4566 zfsvfs_t *zfsvfs = VTOZ(ap->a_vp)->z_zfsvfs;
4567 struct thread *td = ap->a_td;
4568 struct nameidata nd;
4571 vnode_t *xvp = NULL, *vp;
4574 error = extattr_check_cred(ap->a_vp, ap->a_attrnamespace,
4575 ap->a_cred, ap->a_td, VWRITE);
4579 error = zfs_create_attrname(ap->a_attrnamespace, ap->a_name, attrname,
4586 error = zfs_lookup(ap->a_vp, NULL, &xvp, NULL, 0, ap->a_cred, td,
4593 NDINIT_ATVP(&nd, DELETE, NOFOLLOW | LOCKPARENT | LOCKLEAF | MPSAFE,
4594 UIO_SYSSPACE, attrname, xvp, td);
4597 NDFREE(&nd, NDF_ONLY_PNBUF);
4599 if (error == ENOENT)
4604 error = VOP_REMOVE(nd.ni_dvp, vp, &nd.ni_cnd);
4607 if (vp == nd.ni_dvp)
4617 * Vnode operation to set a named attribute.
4620 zfs_setextattr(struct vop_setextattr_args *ap)
4623 IN struct vnode *a_vp;
4624 IN int a_attrnamespace;
4625 IN const char *a_name;
4626 INOUT struct uio *a_uio;
4627 IN struct ucred *a_cred;
4628 IN struct thread *a_td;
4632 zfsvfs_t *zfsvfs = VTOZ(ap->a_vp)->z_zfsvfs;
4633 struct thread *td = ap->a_td;
4634 struct nameidata nd;
4637 vnode_t *xvp = NULL, *vp;
4640 error = extattr_check_cred(ap->a_vp, ap->a_attrnamespace,
4641 ap->a_cred, ap->a_td, VWRITE);
4645 error = zfs_create_attrname(ap->a_attrnamespace, ap->a_name, attrname,
4652 error = zfs_lookup(ap->a_vp, NULL, &xvp, NULL, 0, ap->a_cred, td,
4653 LOOKUP_XATTR | CREATE_XATTR_DIR);
4659 flags = FFLAGS(O_WRONLY | O_CREAT);
4660 NDINIT_ATVP(&nd, LOOKUP, NOFOLLOW | MPSAFE, UIO_SYSSPACE, attrname,
4662 error = vn_open_cred(&nd, &flags, 0600, 0, ap->a_cred, NULL);
4664 NDFREE(&nd, NDF_ONLY_PNBUF);
4672 error = VOP_SETATTR(vp, &va, ap->a_cred);
4674 VOP_WRITE(vp, ap->a_uio, IO_UNIT | IO_SYNC, ap->a_cred);
4677 vn_close(vp, flags, ap->a_cred, td);
4684 * Vnode operation to retrieve extended attributes on a vnode.
4687 zfs_listextattr(struct vop_listextattr_args *ap)
4690 IN struct vnode *a_vp;
4691 IN int a_attrnamespace;
4692 INOUT struct uio *a_uio;
4694 IN struct ucred *a_cred;
4695 IN struct thread *a_td;
4699 zfsvfs_t *zfsvfs = VTOZ(ap->a_vp)->z_zfsvfs;
4700 struct thread *td = ap->a_td;
4701 struct nameidata nd;
4702 char attrprefix[16];
4703 u_char dirbuf[sizeof(struct dirent)];
4706 struct uio auio, *uio = ap->a_uio;
4707 size_t *sizep = ap->a_size;
4709 vnode_t *xvp = NULL, *vp;
4710 int done, error, eof, pos;
4712 error = extattr_check_cred(ap->a_vp, ap->a_attrnamespace,
4713 ap->a_cred, ap->a_td, VREAD);
4717 error = zfs_create_attrname(ap->a_attrnamespace, "", attrprefix,
4718 sizeof(attrprefix));
4721 plen = strlen(attrprefix);
4728 error = zfs_lookup(ap->a_vp, NULL, &xvp, NULL, 0, ap->a_cred, td,
4732 * ENOATTR means that the EA directory does not yet exist,
4733 * i.e. there are no extended attributes there.
4735 if (error == ENOATTR)
4741 NDINIT_ATVP(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | LOCKSHARED | MPSAFE,
4742 UIO_SYSSPACE, ".", xvp, td);
4745 NDFREE(&nd, NDF_ONLY_PNBUF);
4751 auio.uio_iov = &aiov;
4752 auio.uio_iovcnt = 1;
4753 auio.uio_segflg = UIO_SYSSPACE;
4755 auio.uio_rw = UIO_READ;
4756 auio.uio_offset = 0;
4761 aiov.iov_base = (void *)dirbuf;
4762 aiov.iov_len = sizeof(dirbuf);
4763 auio.uio_resid = sizeof(dirbuf);
4764 error = VOP_READDIR(vp, &auio, ap->a_cred, &eof, NULL, NULL);
4765 done = sizeof(dirbuf) - auio.uio_resid;
4768 for (pos = 0; pos < done;) {
4769 dp = (struct dirent *)(dirbuf + pos);
4770 pos += dp->d_reclen;
4772 * XXX: Temporarily we also accept DT_UNKNOWN, as this
4773 * is what we get when attribute was created on Solaris.
4775 if (dp->d_type != DT_REG && dp->d_type != DT_UNKNOWN)
4777 if (plen == 0 && strncmp(dp->d_name, "freebsd:", 8) == 0)
4779 else if (strncmp(dp->d_name, attrprefix, plen) != 0)
4781 nlen = dp->d_namlen - plen;
4784 else if (uio != NULL) {
4786 * Format of extattr name entry is one byte for
4787 * length and the rest for name.
4789 error = uiomove(&nlen, 1, uio->uio_rw, uio);
4791 error = uiomove(dp->d_name + plen, nlen,
4798 } while (!eof && error == 0);
4807 zfs_freebsd_getacl(ap)
4808 struct vop_getacl_args /* {
4817 vsecattr_t vsecattr;
4819 if (ap->a_type != ACL_TYPE_NFS4)
4820 return (EOPNOTSUPP);
4822 vsecattr.vsa_mask = VSA_ACE | VSA_ACECNT;
4823 if (error = zfs_getsecattr(ap->a_vp, &vsecattr, 0, ap->a_cred, NULL))
4826 error = acl_from_aces(ap->a_aclp, vsecattr.vsa_aclentp, vsecattr.vsa_aclcnt);
4827 if (vsecattr.vsa_aclentp != NULL)
4828 kmem_free(vsecattr.vsa_aclentp, vsecattr.vsa_aclentsz);
4834 zfs_freebsd_setacl(ap)
4835 struct vop_setacl_args /* {
4844 vsecattr_t vsecattr;
4845 int aclbsize; /* size of acl list in bytes */
4848 if (ap->a_type != ACL_TYPE_NFS4)
4849 return (EOPNOTSUPP);
4851 if (ap->a_aclp->acl_cnt < 1 || ap->a_aclp->acl_cnt > MAX_ACL_ENTRIES)
4855 * With NFS4 ACLs, chmod(2) may need to add additional entries,
4856 * splitting every entry into two and appending "canonical six"
4857 * entries at the end. Don't allow for setting an ACL that would
4858 * cause chmod(2) to run out of ACL entries.
4860 if (ap->a_aclp->acl_cnt * 2 + 6 > ACL_MAX_ENTRIES)
4863 vsecattr.vsa_mask = VSA_ACE;
4864 aclbsize = ap->a_aclp->acl_cnt * sizeof(ace_t);
4865 vsecattr.vsa_aclentp = kmem_alloc(aclbsize, KM_SLEEP);
4866 aaclp = vsecattr.vsa_aclentp;
4867 vsecattr.vsa_aclentsz = aclbsize;
4869 aces_from_acl(vsecattr.vsa_aclentp, &vsecattr.vsa_aclcnt, ap->a_aclp);
4870 error = zfs_setsecattr(ap->a_vp, &vsecattr, 0, ap->a_cred, NULL);
4871 kmem_free(aaclp, aclbsize);
4877 zfs_freebsd_aclcheck(ap)
4878 struct vop_aclcheck_args /* {
4887 return (EOPNOTSUPP);
4890 struct vop_vector zfs_vnodeops;
4891 struct vop_vector zfs_fifoops;
4893 struct vop_vector zfs_vnodeops = {
4894 .vop_default = &default_vnodeops,
4895 .vop_inactive = zfs_freebsd_inactive,
4896 .vop_reclaim = zfs_freebsd_reclaim,
4897 .vop_access = zfs_freebsd_access,
4898 #ifdef FREEBSD_NAMECACHE
4899 .vop_lookup = vfs_cache_lookup,
4900 .vop_cachedlookup = zfs_freebsd_lookup,
4902 .vop_lookup = zfs_freebsd_lookup,
4904 .vop_getattr = zfs_freebsd_getattr,
4905 .vop_setattr = zfs_freebsd_setattr,
4906 .vop_create = zfs_freebsd_create,
4907 .vop_mknod = zfs_freebsd_create,
4908 .vop_mkdir = zfs_freebsd_mkdir,
4909 .vop_readdir = zfs_freebsd_readdir,
4910 .vop_fsync = zfs_freebsd_fsync,
4911 .vop_open = zfs_freebsd_open,
4912 .vop_close = zfs_freebsd_close,
4913 .vop_rmdir = zfs_freebsd_rmdir,
4914 .vop_ioctl = zfs_freebsd_ioctl,
4915 .vop_link = zfs_freebsd_link,
4916 .vop_symlink = zfs_freebsd_symlink,
4917 .vop_readlink = zfs_freebsd_readlink,
4918 .vop_read = zfs_freebsd_read,
4919 .vop_write = zfs_freebsd_write,
4920 .vop_remove = zfs_freebsd_remove,
4921 .vop_rename = zfs_freebsd_rename,
4922 .vop_pathconf = zfs_freebsd_pathconf,
4923 .vop_bmap = VOP_EOPNOTSUPP,
4924 .vop_fid = zfs_freebsd_fid,
4925 .vop_getextattr = zfs_getextattr,
4926 .vop_deleteextattr = zfs_deleteextattr,
4927 .vop_setextattr = zfs_setextattr,
4928 .vop_listextattr = zfs_listextattr,
4930 .vop_getacl = zfs_freebsd_getacl,
4931 .vop_setacl = zfs_freebsd_setacl,
4932 .vop_aclcheck = zfs_freebsd_aclcheck,
4936 struct vop_vector zfs_fifoops = {
4937 .vop_default = &fifo_specops,
4938 .vop_fsync = VOP_PANIC,
4939 .vop_access = zfs_freebsd_access,
4940 .vop_getattr = zfs_freebsd_getattr,
4941 .vop_inactive = zfs_freebsd_inactive,
4942 .vop_read = VOP_PANIC,
4943 .vop_reclaim = zfs_freebsd_reclaim,
4944 .vop_setattr = zfs_freebsd_setattr,
4945 .vop_write = VOP_PANIC,
4946 .vop_fid = zfs_freebsd_fid,
4948 .vop_getacl = zfs_freebsd_getacl,
4949 .vop_setacl = zfs_freebsd_setacl,
4950 .vop_aclcheck = zfs_freebsd_aclcheck,