4 * The contents of this file are subject to the terms of the
5 * Common Development and Distribution License (the "License").
6 * You may not use this file except in compliance with the License.
8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9 * or http://www.opensolaris.org/os/licensing.
10 * See the License for the specific language governing permissions
11 * and limitations under the License.
13 * When distributing Covered Code, include this CDDL HEADER in each
14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15 * If applicable, add the following below this CDDL HEADER, with the
16 * fields enclosed by brackets "[]" replaced with your own identifying
17 * information: Portions Copyright [yyyy] [name of copyright owner]
22 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
25 /* Portions Copyright 2007 Jeremy Teo */
27 #include <sys/types.h>
28 #include <sys/param.h>
30 #include <sys/systm.h>
31 #include <sys/sysmacros.h>
32 #include <sys/resource.h>
34 #include <sys/vnode.h>
38 #include <sys/taskq.h>
40 #include <sys/atomic.h>
41 #include <sys/namei.h>
43 #include <sys/cmn_err.h>
44 #include <sys/errno.h>
45 #include <sys/unistd.h>
46 #include <sys/zfs_dir.h>
47 #include <sys/zfs_ioctl.h>
48 #include <sys/fs/zfs.h>
54 #include <sys/dirent.h>
55 #include <sys/policy.h>
56 #include <sys/sunddi.h>
57 #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 TXG_NOWAIT as the second argument to dmu_tx_assign().
102 * This is critical because we don't want to block while holding locks.
103 * Note, in particular, that if a lock is sometimes acquired before
104 * the tx assigns, and sometimes after (e.g. z_lock), then failing to
105 * use a non-blocking assign can deadlock the system. The scenario:
107 * Thread A has grabbed a lock before calling dmu_tx_assign().
108 * Thread B is in an already-assigned tx, and blocks for this lock.
109 * Thread A calls dmu_tx_assign(TXG_WAIT) and blocks in txg_wait_open()
110 * forever, because the previous txg can't quiesce until B's tx commits.
112 * If dmu_tx_assign() returns ERESTART and zfsvfs->z_assign is TXG_NOWAIT,
113 * then drop all locks, call dmu_tx_wait(), and try again.
115 * (5) If the operation succeeded, generate the intent log entry for it
116 * before dropping locks. This ensures that the ordering of events
117 * in the intent log matches the order in which they actually occurred.
118 * During ZIL replay the zfs_log_* functions will update the sequence
119 * number to indicate the zil transaction has replayed.
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, TXG_NOWAIT); // 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) {
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);
166 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
171 if ((flag & FWRITE) && (zp->z_phys->zp_flags & ZFS_APPENDONLY) &&
172 ((flag & FAPPEND) == 0)) {
177 if (!zfs_has_ctldir(zp) && zp->z_zfsvfs->z_vscan &&
178 ZTOV(zp)->v_type == VREG &&
179 !(zp->z_phys->zp_flags & ZFS_AV_QUARANTINED) &&
180 zp->z_phys->zp_size > 0) {
181 if (fs_vscan(*vpp, cr, 0) != 0) {
187 /* Keep a count of the synchronous opens in the znode */
188 if (flag & (FSYNC | FDSYNC))
189 atomic_inc_32(&zp->z_sync_cnt);
197 zfs_close(vnode_t *vp, int flag, int count, offset_t offset, cred_t *cr,
198 caller_context_t *ct)
200 znode_t *zp = VTOZ(vp);
201 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
204 * Clean up any locks held by this process on the vp.
206 cleanlocks(vp, ddi_get_pid(), 0);
207 cleanshares(vp, ddi_get_pid());
212 /* Decrement the synchronous opens in the znode */
213 if ((flag & (FSYNC | FDSYNC)) && (count == 1))
214 atomic_dec_32(&zp->z_sync_cnt);
216 if (!zfs_has_ctldir(zp) && zp->z_zfsvfs->z_vscan &&
217 ZTOV(zp)->v_type == VREG &&
218 !(zp->z_phys->zp_flags & ZFS_AV_QUARANTINED) &&
219 zp->z_phys->zp_size > 0)
220 VERIFY(fs_vscan(vp, cr, 1) == 0);
227 * Lseek support for finding holes (cmd == _FIO_SEEK_HOLE) and
228 * data (cmd == _FIO_SEEK_DATA). "off" is an in/out parameter.
231 zfs_holey(vnode_t *vp, u_long cmd, offset_t *off)
233 znode_t *zp = VTOZ(vp);
234 uint64_t noff = (uint64_t)*off; /* new offset */
239 file_sz = zp->z_phys->zp_size;
240 if (noff >= file_sz) {
244 if (cmd == _FIO_SEEK_HOLE)
249 error = dmu_offset_next(zp->z_zfsvfs->z_os, zp->z_id, hole, &noff);
252 if ((error == ESRCH) || (noff > file_sz)) {
254 * Handle the virtual hole at the end of file.
271 zfs_ioctl(vnode_t *vp, u_long com, intptr_t data, int flag, cred_t *cred,
272 int *rvalp, caller_context_t *ct)
284 * The following two ioctls are used by bfu. Faking out,
285 * necessary to avoid bfu errors.
293 if (ddi_copyin((void *)data, &off, sizeof (off), flag))
297 zfsvfs = zp->z_zfsvfs;
301 /* offset parameter is in/out */
302 error = zfs_holey(vp, com, &off);
306 if (ddi_copyout(&off, (void *)data, sizeof (off), flag))
314 page_lookup(vnode_t *vp, int64_t start, int64_t off, int64_t nbytes)
320 VM_OBJECT_LOCK_ASSERT(obj, MA_OWNED);
323 if ((pp = vm_page_lookup(obj, OFF_TO_IDX(start))) != NULL &&
324 vm_page_is_valid(pp, (vm_offset_t)off, nbytes)) {
325 if (vm_page_sleep_if_busy(pp, FALSE, "zfsmwb"))
328 vm_page_lock_queues();
330 vm_page_unlock_queues();
332 if (__predict_false(obj->cache != NULL)) {
333 vm_page_cache_free(obj, OFF_TO_IDX(start),
334 OFF_TO_IDX(start) + 1);
344 page_unlock(vm_page_t pp)
351 zfs_map_page(vm_page_t pp, struct sf_buf **sfp)
354 *sfp = sf_buf_alloc(pp, 0);
355 return ((caddr_t)sf_buf_kva(*sfp));
359 zfs_unmap_page(struct sf_buf *sf)
367 * When a file is memory mapped, we must keep the IO data synchronized
368 * between the DMU cache and the memory mapped pages. What this means:
370 * On Write: If we find a memory mapped page, we write to *both*
371 * the page and the dmu buffer.
375 update_pages(vnode_t *vp, int64_t start, int len, objset_t *os, uint64_t oid,
376 int segflg, dmu_tx_t *tx)
382 ASSERT(vp->v_mount != NULL);
386 off = start & PAGEOFFSET;
388 for (start &= PAGEMASK; len > 0; start += PAGESIZE) {
390 int nbytes = MIN(PAGESIZE - off, len);
392 if ((pp = page_lookup(vp, start, off, nbytes)) != NULL) {
395 VM_OBJECT_UNLOCK(obj);
396 va = zfs_map_page(pp, &sf);
397 if (segflg == UIO_NOCOPY) {
398 (void) dmu_write(os, oid, start+off, nbytes,
401 (void) dmu_read(os, oid, start+off, nbytes,
402 va+off, DMU_READ_PREFETCH);;
412 VM_OBJECT_UNLOCK(obj);
416 * When a file is memory mapped, we must keep the IO data synchronized
417 * between the DMU cache and the memory mapped pages. What this means:
419 * On Read: We "read" preferentially from memory mapped pages,
420 * else we default from the dmu buffer.
422 * NOTE: We will always "break up" the IO into PAGESIZE uiomoves when
423 * the file is memory mapped.
426 mappedread(vnode_t *vp, int nbytes, uio_t *uio)
428 znode_t *zp = VTOZ(vp);
429 objset_t *os = zp->z_zfsvfs->z_os;
440 ASSERT(vp->v_mount != NULL);
444 start = uio->uio_loffset;
445 off = start & PAGEOFFSET;
448 for (start &= PAGEMASK; len > 0; start += PAGESIZE) {
449 int bytes = MIN(PAGESIZE - off, len);
452 if ((m = vm_page_lookup(obj, OFF_TO_IDX(start))) != NULL &&
453 vm_page_is_valid(m, off, bytes)) {
454 if (vm_page_sleep_if_busy(m, FALSE, "zfsmrb"))
457 VM_OBJECT_UNLOCK(obj);
459 error = dmu_read_uio(os, zp->z_id, uio,
464 uiomove_fromphys(&m, off, bytes, uio);
467 } else if (m != NULL && uio->uio_segflg == UIO_NOCOPY) {
469 * The code below is here to make sendfile(2) work
470 * correctly with ZFS. As pointed out by ups@
471 * sendfile(2) should be changed to use VOP_GETPAGES(),
472 * but it pessimize performance of sendfile/UFS, that's
473 * why I handle this special case in ZFS code.
476 ("unexpected offset in mappedread for sendfile"));
477 if (vm_page_sleep_if_busy(m, FALSE, "zfsmrb"))
480 VM_OBJECT_UNLOCK(obj);
482 error = dmu_read_uio(os, zp->z_id, uio,
487 va = zfs_map_page(m, &sf);
488 error = dmu_read(os, zp->z_id, start, bytes, va,
490 if (bytes != PAGE_SIZE)
491 bzero(va + bytes, PAGE_SIZE - bytes);
496 m->valid = VM_PAGE_BITS_ALL;
499 uio->uio_resid -= bytes;
500 uio->uio_offset += bytes;
510 VM_OBJECT_UNLOCK(obj);
511 if (error == 0 && dirbytes > 0)
512 error = dmu_read_uio(os, zp->z_id, uio, dirbytes);
516 offset_t zfs_read_chunk_size = 1024 * 1024; /* Tunable */
519 * Read bytes from specified file into supplied buffer.
521 * IN: vp - vnode of file to be read from.
522 * uio - structure supplying read location, range info,
524 * ioflag - SYNC flags; used to provide FRSYNC semantics.
525 * cr - credentials of caller.
526 * ct - caller context
528 * OUT: uio - updated offset and range, buffer filled.
530 * RETURN: 0 if success
531 * error code if failure
534 * vp - atime updated if byte count > 0
538 zfs_read(vnode_t *vp, uio_t *uio, int ioflag, cred_t *cr, caller_context_t *ct)
540 znode_t *zp = VTOZ(vp);
541 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
551 if (zp->z_phys->zp_flags & ZFS_AV_QUARANTINED) {
557 * Validate file offset
559 if (uio->uio_loffset < (offset_t)0) {
565 * Fasttrack empty reads
567 if (uio->uio_resid == 0) {
573 * Check for mandatory locks
575 if (MANDMODE((mode_t)zp->z_phys->zp_mode)) {
576 if (error = chklock(vp, FREAD,
577 uio->uio_loffset, uio->uio_resid, uio->uio_fmode, ct)) {
584 * If we're in FRSYNC mode, sync out this znode before reading it.
587 zil_commit(zfsvfs->z_log, zp->z_last_itx, zp->z_id);
590 * Lock the range against changes.
592 rl = zfs_range_lock(zp, uio->uio_loffset, uio->uio_resid, RL_READER);
595 * If we are reading past end-of-file we can skip
596 * to the end; but we might still need to set atime.
598 if (uio->uio_loffset >= zp->z_phys->zp_size) {
603 ASSERT(uio->uio_loffset < zp->z_phys->zp_size);
604 n = MIN(uio->uio_resid, zp->z_phys->zp_size - uio->uio_loffset);
607 nbytes = MIN(n, zfs_read_chunk_size -
608 P2PHASE(uio->uio_loffset, zfs_read_chunk_size));
610 if (vn_has_cached_data(vp))
611 error = mappedread(vp, nbytes, uio);
613 error = dmu_read_uio(os, zp->z_id, uio, nbytes);
615 /* convert checksum errors into IO errors */
625 zfs_range_unlock(rl);
627 ZFS_ACCESSTIME_STAMP(zfsvfs, zp);
633 * Fault in the pages of the first n bytes specified by the uio structure.
634 * 1 byte in each page is touched and the uio struct is unmodified.
635 * Any error will exit this routine as this is only a best
636 * attempt to get the pages resident. This is a copy of ufs_trans_touch().
639 zfs_prefault_write(ssize_t n, struct uio *uio)
645 if (uio->uio_segflg != UIO_USERSPACE)
651 cnt = MIN(iov->iov_len, n);
653 /* empty iov entry */
659 * touch each page in this segment.
665 incr = MIN(cnt, PAGESIZE);
670 * touch the last byte in case it straddles a page.
680 * Write the bytes to a file.
682 * IN: vp - vnode of file to be written to.
683 * uio - structure supplying write location, range info,
685 * ioflag - IO_APPEND flag set if in append mode.
686 * cr - credentials of caller.
687 * ct - caller context (NFS/CIFS fem monitor only)
689 * OUT: uio - updated offset and range.
691 * RETURN: 0 if success
692 * error code if failure
695 * vp - ctime|mtime updated if byte count > 0
699 zfs_write(vnode_t *vp, uio_t *uio, int ioflag, cred_t *cr, caller_context_t *ct)
701 znode_t *zp = VTOZ(vp);
702 rlim64_t limit = MAXOFFSET_T;
703 ssize_t start_resid = uio->uio_resid;
707 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
712 int max_blksz = zfsvfs->z_max_blksz;
718 * Fasttrack empty write
724 if (limit == RLIM64_INFINITY || limit > MAXOFFSET_T)
731 * If immutable or not appending then return EPERM
733 pflags = zp->z_phys->zp_flags;
734 if ((pflags & (ZFS_IMMUTABLE | ZFS_READONLY)) ||
735 ((pflags & ZFS_APPENDONLY) && !(ioflag & FAPPEND) &&
736 (uio->uio_loffset < zp->z_phys->zp_size))) {
741 zilog = zfsvfs->z_log;
744 * Pre-fault the pages to ensure slow (eg NFS) pages
747 zfs_prefault_write(n, uio);
750 * If in append mode, set the io offset pointer to eof.
752 if (ioflag & IO_APPEND) {
754 * Range lock for a file append:
755 * The value for the start of range will be determined by
756 * zfs_range_lock() (to guarantee append semantics).
757 * If this write will cause the block size to increase,
758 * zfs_range_lock() will lock the entire file, so we must
759 * later reduce the range after we grow the block size.
761 rl = zfs_range_lock(zp, 0, n, RL_APPEND);
762 if (rl->r_len == UINT64_MAX) {
763 /* overlocked, zp_size can't change */
764 woff = uio->uio_loffset = zp->z_phys->zp_size;
766 woff = uio->uio_loffset = rl->r_off;
769 woff = uio->uio_loffset;
771 * Validate file offset
779 * If we need to grow the block size then zfs_range_lock()
780 * will lock a wider range than we request here.
781 * Later after growing the block size we reduce the range.
783 rl = zfs_range_lock(zp, woff, n, RL_WRITER);
787 zfs_range_unlock(rl);
792 if ((woff + n) > limit || woff > (limit - n))
796 * Check for mandatory locks
798 if (MANDMODE((mode_t)zp->z_phys->zp_mode) &&
799 (error = chklock(vp, FWRITE, woff, n, uio->uio_fmode, ct)) != 0) {
800 zfs_range_unlock(rl);
804 end_size = MAX(zp->z_phys->zp_size, woff + n);
807 * Write the file in reasonable size chunks. Each chunk is written
808 * in a separate transaction; this keeps the intent log records small
809 * and allows us to do more fine-grained space accounting.
813 woff = uio->uio_loffset;
816 if (zfs_usergroup_overquota(zfsvfs,
817 B_FALSE, zp->z_phys->zp_uid) ||
818 zfs_usergroup_overquota(zfsvfs,
819 B_TRUE, zp->z_phys->zp_gid)) {
821 dmu_return_arcbuf(abuf);
827 * If dmu_assign_arcbuf() is expected to execute with minimum
828 * overhead loan an arc buffer and copy user data to it before
829 * we enter a txg. This avoids holding a txg forever while we
830 * pagefault on a hanging NFS server mapping.
832 if (abuf == NULL && n >= max_blksz &&
833 woff >= zp->z_phys->zp_size &&
834 P2PHASE(woff, max_blksz) == 0 &&
835 zp->z_blksz == max_blksz) {
838 abuf = dmu_request_arcbuf(zp->z_dbuf, max_blksz);
839 ASSERT(abuf != NULL);
840 ASSERT(arc_buf_size(abuf) == max_blksz);
841 if (error = uiocopy(abuf->b_data, max_blksz,
842 UIO_WRITE, uio, &cbytes)) {
843 dmu_return_arcbuf(abuf);
846 ASSERT(cbytes == max_blksz);
850 * Start a transaction.
852 tx = dmu_tx_create(zfsvfs->z_os);
853 dmu_tx_hold_bonus(tx, zp->z_id);
854 dmu_tx_hold_write(tx, zp->z_id, woff, MIN(n, max_blksz));
855 error = dmu_tx_assign(tx, TXG_NOWAIT);
857 if (error == ERESTART) {
864 dmu_return_arcbuf(abuf);
869 * If zfs_range_lock() over-locked we grow the blocksize
870 * and then reduce the lock range. This will only happen
871 * on the first iteration since zfs_range_reduce() will
872 * shrink down r_len to the appropriate size.
874 if (rl->r_len == UINT64_MAX) {
877 if (zp->z_blksz > max_blksz) {
878 ASSERT(!ISP2(zp->z_blksz));
879 new_blksz = MIN(end_size, SPA_MAXBLOCKSIZE);
881 new_blksz = MIN(end_size, max_blksz);
883 zfs_grow_blocksize(zp, new_blksz, tx);
884 zfs_range_reduce(rl, woff, n);
888 * XXX - should we really limit each write to z_max_blksz?
889 * Perhaps we should use SPA_MAXBLOCKSIZE chunks?
891 nbytes = MIN(n, max_blksz - P2PHASE(woff, max_blksz));
893 if (woff + nbytes > zp->z_phys->zp_size)
894 vnode_pager_setsize(vp, woff + nbytes);
897 tx_bytes = uio->uio_resid;
898 error = dmu_write_uio(zfsvfs->z_os, zp->z_id, uio,
900 tx_bytes -= uio->uio_resid;
903 ASSERT(tx_bytes == max_blksz);
904 dmu_assign_arcbuf(zp->z_dbuf, woff, abuf, tx);
905 ASSERT(tx_bytes <= uio->uio_resid);
906 uioskip(uio, tx_bytes);
909 if (tx_bytes && vn_has_cached_data(vp)) {
910 update_pages(vp, woff, tx_bytes, zfsvfs->z_os,
911 zp->z_id, uio->uio_segflg, tx);
915 * If we made no progress, we're done. If we made even
916 * partial progress, update the znode and ZIL accordingly.
925 * Clear Set-UID/Set-GID bits on successful write if not
926 * privileged and at least one of the excute bits is set.
928 * It would be nice to to this after all writes have
929 * been done, but that would still expose the ISUID/ISGID
930 * to another app after the partial write is committed.
932 * Note: we don't call zfs_fuid_map_id() here because
933 * user 0 is not an ephemeral uid.
935 mutex_enter(&zp->z_acl_lock);
936 if ((zp->z_phys->zp_mode & (S_IXUSR | (S_IXUSR >> 3) |
937 (S_IXUSR >> 6))) != 0 &&
938 (zp->z_phys->zp_mode & (S_ISUID | S_ISGID)) != 0 &&
939 secpolicy_vnode_setid_retain(vp, cr,
940 (zp->z_phys->zp_mode & S_ISUID) != 0 &&
941 zp->z_phys->zp_uid == 0) != 0) {
942 zp->z_phys->zp_mode &= ~(S_ISUID | S_ISGID);
944 mutex_exit(&zp->z_acl_lock);
947 * Update time stamp. NOTE: This marks the bonus buffer as
948 * dirty, so we don't have to do it again for zp_size.
950 zfs_time_stamper(zp, CONTENT_MODIFIED, tx);
953 * Update the file size (zp_size) if it has changed;
954 * account for possible concurrent updates.
956 while ((end_size = zp->z_phys->zp_size) < uio->uio_loffset)
957 (void) atomic_cas_64(&zp->z_phys->zp_size, end_size,
959 zfs_log_write(zilog, tx, TX_WRITE, zp, woff, tx_bytes, ioflag);
964 ASSERT(tx_bytes == nbytes);
968 zfs_range_unlock(rl);
971 * If we're in replay mode, or we made no progress, return error.
972 * Otherwise, it's at least a partial write, so it's successful.
974 if (zfsvfs->z_replay || uio->uio_resid == start_resid) {
979 if (ioflag & (FSYNC | FDSYNC))
980 zil_commit(zilog, zp->z_last_itx, zp->z_id);
987 zfs_get_done(dmu_buf_t *db, void *vzgd)
989 zgd_t *zgd = (zgd_t *)vzgd;
990 rl_t *rl = zgd->zgd_rl;
991 vnode_t *vp = ZTOV(rl->r_zp);
992 objset_t *os = rl->r_zp->z_zfsvfs->z_os;
995 vfslocked = VFS_LOCK_GIANT(vp->v_vfsp);
996 dmu_buf_rele(db, vzgd);
997 zfs_range_unlock(rl);
999 * Release the vnode asynchronously as we currently have the
1000 * txg stopped from syncing.
1002 VN_RELE_ASYNC(vp, dsl_pool_vnrele_taskq(dmu_objset_pool(os)));
1003 zil_add_block(zgd->zgd_zilog, zgd->zgd_bp);
1004 kmem_free(zgd, sizeof (zgd_t));
1005 VFS_UNLOCK_GIANT(vfslocked);
1009 * Get data to generate a TX_WRITE intent log record.
1012 zfs_get_data(void *arg, lr_write_t *lr, char *buf, zio_t *zio)
1014 zfsvfs_t *zfsvfs = arg;
1015 objset_t *os = zfsvfs->z_os;
1017 uint64_t off = lr->lr_offset;
1021 int dlen = lr->lr_length; /* length of user data */
1028 * Nothing to do if the file has been removed
1030 if (zfs_zget(zfsvfs, lr->lr_foid, &zp) != 0)
1032 if (zp->z_unlinked) {
1034 * Release the vnode asynchronously as we currently have the
1035 * txg stopped from syncing.
1037 VN_RELE_ASYNC(ZTOV(zp),
1038 dsl_pool_vnrele_taskq(dmu_objset_pool(os)));
1043 * Write records come in two flavors: immediate and indirect.
1044 * For small writes it's cheaper to store the data with the
1045 * log record (immediate); for large writes it's cheaper to
1046 * sync the data and get a pointer to it (indirect) so that
1047 * we don't have to write the data twice.
1049 if (buf != NULL) { /* immediate write */
1050 rl = zfs_range_lock(zp, off, dlen, RL_READER);
1051 /* test for truncation needs to be done while range locked */
1052 if (off >= zp->z_phys->zp_size) {
1056 VERIFY(0 == dmu_read(os, lr->lr_foid, off, dlen, buf,
1057 DMU_READ_NO_PREFETCH));
1058 } else { /* indirect write */
1059 uint64_t boff; /* block starting offset */
1062 * Have to lock the whole block to ensure when it's
1063 * written out and it's checksum is being calculated
1064 * that no one can change the data. We need to re-check
1065 * blocksize after we get the lock in case it's changed!
1068 if (ISP2(zp->z_blksz)) {
1069 boff = P2ALIGN_TYPED(off, zp->z_blksz,
1075 rl = zfs_range_lock(zp, boff, dlen, RL_READER);
1076 if (zp->z_blksz == dlen)
1078 zfs_range_unlock(rl);
1080 /* test for truncation needs to be done while range locked */
1081 if (off >= zp->z_phys->zp_size) {
1085 zgd = (zgd_t *)kmem_alloc(sizeof (zgd_t), KM_SLEEP);
1087 zgd->zgd_zilog = zfsvfs->z_log;
1088 zgd->zgd_bp = &lr->lr_blkptr;
1089 VERIFY(0 == dmu_buf_hold(os, lr->lr_foid, boff, zgd, &db));
1090 ASSERT(boff == db->db_offset);
1091 lr->lr_blkoff = off - boff;
1092 error = dmu_sync(zio, db, &lr->lr_blkptr,
1093 lr->lr_common.lrc_txg, zfs_get_done, zgd);
1094 ASSERT((error && error != EINPROGRESS) ||
1095 lr->lr_length <= zp->z_blksz);
1098 * dmu_sync() can compress a block of zeros to a null
1099 * blkptr but the block size still needs to be passed
1100 * through to replay.
1102 BP_SET_LSIZE(&lr->lr_blkptr, db->db_size);
1103 zil_add_block(zfsvfs->z_log, &lr->lr_blkptr);
1107 * If we get EINPROGRESS, then we need to wait for a
1108 * write IO initiated by dmu_sync() to complete before
1109 * we can release this dbuf. We will finish everything
1110 * up in the zfs_get_done() callback.
1112 if (error == EINPROGRESS) {
1114 } else if (error == EALREADY) {
1115 lr->lr_common.lrc_txtype = TX_WRITE2;
1118 dmu_buf_rele(db, zgd);
1119 kmem_free(zgd, sizeof (zgd_t));
1122 zfs_range_unlock(rl);
1124 * Release the vnode asynchronously as we currently have the
1125 * txg stopped from syncing.
1127 VN_RELE_ASYNC(ZTOV(zp), dsl_pool_vnrele_taskq(dmu_objset_pool(os)));
1133 zfs_access(vnode_t *vp, int mode, int flag, cred_t *cr,
1134 caller_context_t *ct)
1136 znode_t *zp = VTOZ(vp);
1137 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1143 if (flag & V_ACE_MASK)
1144 error = zfs_zaccess(zp, mode, flag, B_FALSE, cr);
1146 error = zfs_zaccess_rwx(zp, mode, flag, cr);
1153 * If vnode is for a device return a specfs vnode instead.
1156 specvp_check(vnode_t **vpp, cred_t *cr)
1160 if (IS_DEVVP(*vpp)) {
1163 svp = specvp(*vpp, (*vpp)->v_rdev, (*vpp)->v_type, cr);
1174 * Lookup an entry in a directory, or an extended attribute directory.
1175 * If it exists, return a held vnode reference for it.
1177 * IN: dvp - vnode of directory to search.
1178 * nm - name of entry to lookup.
1179 * pnp - full pathname to lookup [UNUSED].
1180 * flags - LOOKUP_XATTR set if looking for an attribute.
1181 * rdir - root directory vnode [UNUSED].
1182 * cr - credentials of caller.
1183 * ct - caller context
1184 * direntflags - directory lookup flags
1185 * realpnp - returned pathname.
1187 * OUT: vpp - vnode of located entry, NULL if not found.
1189 * RETURN: 0 if success
1190 * error code if failure
1197 zfs_lookup(vnode_t *dvp, char *nm, vnode_t **vpp, struct componentname *cnp,
1198 int nameiop, cred_t *cr, kthread_t *td, int flags)
1200 znode_t *zdp = VTOZ(dvp);
1201 zfsvfs_t *zfsvfs = zdp->z_zfsvfs;
1203 int *direntflags = NULL;
1204 void *realpnp = NULL;
1207 if (!(flags & (LOOKUP_XATTR | FIGNORECASE))) {
1209 if (dvp->v_type != VDIR) {
1211 } else if (zdp->z_dbuf == NULL) {
1215 if (nm[0] == 0 || (nm[0] == '.' && nm[1] == '\0')) {
1216 error = zfs_fastaccesschk_execute(zdp, cr);
1224 vnode_t *tvp = dnlc_lookup(dvp, nm);
1227 error = zfs_fastaccesschk_execute(zdp, cr);
1232 if (tvp == DNLC_NO_VNODE) {
1237 return (specvp_check(vpp, cr));
1243 DTRACE_PROBE2(zfs__fastpath__lookup__miss, vnode_t *, dvp, char *, nm);
1250 if (flags & LOOKUP_XATTR) {
1253 * If the xattr property is off, refuse the lookup request.
1255 if (!(zfsvfs->z_vfs->vfs_flag & VFS_XATTR)) {
1262 * We don't allow recursive attributes..
1263 * Maybe someday we will.
1265 if (zdp->z_phys->zp_flags & ZFS_XATTR) {
1270 if (error = zfs_get_xattrdir(VTOZ(dvp), vpp, cr, flags)) {
1276 * Do we have permission to get into attribute directory?
1279 if (error = zfs_zaccess(VTOZ(*vpp), ACE_EXECUTE, 0,
1289 if (dvp->v_type != VDIR) {
1295 * Check accessibility of directory.
1298 if (error = zfs_zaccess(zdp, ACE_EXECUTE, 0, B_FALSE, cr)) {
1303 if (zfsvfs->z_utf8 && u8_validate(nm, strlen(nm),
1304 NULL, U8_VALIDATE_ENTIRE, &error) < 0) {
1309 error = zfs_dirlook(zdp, nm, vpp, flags, direntflags, realpnp);
1311 error = specvp_check(vpp, cr);
1313 /* Translate errors and add SAVENAME when needed. */
1314 if (cnp->cn_flags & ISLASTCN) {
1318 if (error == ENOENT) {
1319 error = EJUSTRETURN;
1320 cnp->cn_flags |= SAVENAME;
1326 cnp->cn_flags |= SAVENAME;
1330 if (error == 0 && (nm[0] != '.' || nm[1] != '\0')) {
1333 if (cnp->cn_flags & ISDOTDOT) {
1334 ltype = VOP_ISLOCKED(dvp);
1338 error = vn_lock(*vpp, cnp->cn_lkflags);
1339 if (cnp->cn_flags & ISDOTDOT)
1340 vn_lock(dvp, ltype | LK_RETRY);
1350 #ifdef FREEBSD_NAMECACHE
1352 * Insert name into cache (as non-existent) if appropriate.
1354 if (error == ENOENT && (cnp->cn_flags & MAKEENTRY) && nameiop != CREATE)
1355 cache_enter(dvp, *vpp, cnp);
1357 * Insert name into cache if appropriate.
1359 if (error == 0 && (cnp->cn_flags & MAKEENTRY)) {
1360 if (!(cnp->cn_flags & ISLASTCN) ||
1361 (nameiop != DELETE && nameiop != RENAME)) {
1362 cache_enter(dvp, *vpp, cnp);
1371 * Attempt to create a new entry in a directory. If the entry
1372 * already exists, truncate the file if permissible, else return
1373 * an error. Return the vp of the created or trunc'd file.
1375 * IN: dvp - vnode of directory to put new file entry in.
1376 * name - name of new file entry.
1377 * vap - attributes of new file.
1378 * excl - flag indicating exclusive or non-exclusive mode.
1379 * mode - mode to open file with.
1380 * cr - credentials of caller.
1381 * flag - large file flag [UNUSED].
1382 * ct - caller context
1383 * vsecp - ACL to be set
1385 * OUT: vpp - vnode of created or trunc'd entry.
1387 * RETURN: 0 if success
1388 * error code if failure
1391 * dvp - ctime|mtime updated if new entry created
1392 * vp - ctime|mtime always, atime if new
1397 zfs_create(vnode_t *dvp, char *name, vattr_t *vap, int excl, int mode,
1398 vnode_t **vpp, cred_t *cr, kthread_t *td)
1400 znode_t *zp, *dzp = VTOZ(dvp);
1401 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
1409 gid_t gid = crgetgid(cr);
1410 zfs_acl_ids_t acl_ids;
1411 boolean_t fuid_dirtied;
1416 * If we have an ephemeral id, ACL, or XVATTR then
1417 * make sure file system is at proper version
1420 ksid = crgetsid(cr, KSID_OWNER);
1422 uid = ksid_getid(ksid);
1425 if (zfsvfs->z_use_fuids == B_FALSE &&
1426 (vsecp || (vap->va_mask & AT_XVATTR) ||
1427 IS_EPHEMERAL(crgetuid(cr)) || IS_EPHEMERAL(crgetgid(cr))))
1433 zilog = zfsvfs->z_log;
1435 if (zfsvfs->z_utf8 && u8_validate(name, strlen(name),
1436 NULL, U8_VALIDATE_ENTIRE, &error) < 0) {
1441 if (vap->va_mask & AT_XVATTR) {
1442 if ((error = secpolicy_xvattr(dvp, (xvattr_t *)vap,
1443 crgetuid(cr), cr, vap->va_type)) != 0) {
1451 if ((vap->va_mode & S_ISVTX) && secpolicy_vnode_stky_modify(cr))
1452 vap->va_mode &= ~S_ISVTX;
1454 if (*name == '\0') {
1456 * Null component name refers to the directory itself.
1463 /* possible VN_HOLD(zp) */
1466 if (flag & FIGNORECASE)
1469 error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg,
1472 if (strcmp(name, "..") == 0)
1482 * Create a new file object and update the directory
1485 if (error = zfs_zaccess(dzp, ACE_ADD_FILE, 0, B_FALSE, cr)) {
1490 * We only support the creation of regular files in
1491 * extended attribute directories.
1493 if ((dzp->z_phys->zp_flags & ZFS_XATTR) &&
1494 (vap->va_type != VREG)) {
1500 if ((error = zfs_acl_ids_create(dzp, 0, vap, cr, vsecp,
1503 if (zfs_acl_ids_overquota(zfsvfs, &acl_ids)) {
1504 zfs_acl_ids_free(&acl_ids);
1509 tx = dmu_tx_create(os);
1510 dmu_tx_hold_bonus(tx, DMU_NEW_OBJECT);
1511 fuid_dirtied = zfsvfs->z_fuid_dirty;
1513 zfs_fuid_txhold(zfsvfs, tx);
1514 dmu_tx_hold_bonus(tx, dzp->z_id);
1515 dmu_tx_hold_zap(tx, dzp->z_id, TRUE, name);
1516 if (acl_ids.z_aclp->z_acl_bytes > ZFS_ACE_SPACE) {
1517 dmu_tx_hold_write(tx, DMU_NEW_OBJECT,
1518 0, SPA_MAXBLOCKSIZE);
1520 error = dmu_tx_assign(tx, TXG_NOWAIT);
1522 zfs_acl_ids_free(&acl_ids);
1523 zfs_dirent_unlock(dl);
1524 if (error == ERESTART) {
1533 zfs_mknode(dzp, vap, tx, cr, 0, &zp, 0, &acl_ids);
1536 zfs_fuid_sync(zfsvfs, tx);
1538 (void) zfs_link_create(dl, zp, tx, ZNEW);
1540 txtype = zfs_log_create_txtype(Z_FILE, vsecp, vap);
1541 if (flag & FIGNORECASE)
1543 zfs_log_create(zilog, tx, txtype, dzp, zp, name,
1544 vsecp, acl_ids.z_fuidp, vap);
1545 zfs_acl_ids_free(&acl_ids);
1548 int aflags = (flag & FAPPEND) ? V_APPEND : 0;
1551 * A directory entry already exists for this name.
1554 * Can't truncate an existing file if in exclusive mode.
1561 * Can't open a directory for writing.
1563 if ((ZTOV(zp)->v_type == VDIR) && (mode & S_IWRITE)) {
1568 * Verify requested access to file.
1570 if (mode && (error = zfs_zaccess_rwx(zp, mode, aflags, cr))) {
1574 mutex_enter(&dzp->z_lock);
1576 mutex_exit(&dzp->z_lock);
1579 * Truncate regular files if requested.
1581 if ((ZTOV(zp)->v_type == VREG) &&
1582 (vap->va_mask & AT_SIZE) && (vap->va_size == 0)) {
1583 /* we can't hold any locks when calling zfs_freesp() */
1584 zfs_dirent_unlock(dl);
1586 error = zfs_freesp(zp, 0, 0, mode, TRUE);
1588 vnevent_create(ZTOV(zp), ct);
1594 zfs_dirent_unlock(dl);
1601 error = specvp_check(vpp, cr);
1609 * Remove an entry from a directory.
1611 * IN: dvp - vnode of directory to remove entry from.
1612 * name - name of entry to remove.
1613 * cr - credentials of caller.
1614 * ct - caller context
1615 * flags - case flags
1617 * RETURN: 0 if success
1618 * error code if failure
1622 * vp - ctime (if nlink > 0)
1626 zfs_remove(vnode_t *dvp, char *name, cred_t *cr, caller_context_t *ct,
1629 znode_t *zp, *dzp = VTOZ(dvp);
1630 znode_t *xzp = NULL;
1632 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
1634 uint64_t acl_obj, xattr_obj;
1637 boolean_t may_delete_now, delete_now = FALSE;
1638 boolean_t unlinked, toobig = FALSE;
1640 pathname_t *realnmp = NULL;
1647 zilog = zfsvfs->z_log;
1649 if (flags & FIGNORECASE) {
1657 * Attempt to lock directory; fail if entry doesn't exist.
1659 if (error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg,
1669 if (error = zfs_zaccess_delete(dzp, zp, cr)) {
1674 * Need to use rmdir for removing directories.
1676 if (vp->v_type == VDIR) {
1681 vnevent_remove(vp, dvp, name, ct);
1684 dnlc_remove(dvp, realnmp->pn_buf);
1686 dnlc_remove(dvp, name);
1688 may_delete_now = FALSE;
1691 * We may delete the znode now, or we may put it in the unlinked set;
1692 * it depends on whether we're the last link, and on whether there are
1693 * other holds on the vnode. So we dmu_tx_hold() the right things to
1694 * allow for either case.
1696 tx = dmu_tx_create(zfsvfs->z_os);
1697 dmu_tx_hold_zap(tx, dzp->z_id, FALSE, name);
1698 dmu_tx_hold_bonus(tx, zp->z_id);
1699 if (may_delete_now) {
1701 zp->z_phys->zp_size > zp->z_blksz * DMU_MAX_DELETEBLKCNT;
1702 /* if the file is too big, only hold_free a token amount */
1703 dmu_tx_hold_free(tx, zp->z_id, 0,
1704 (toobig ? DMU_MAX_ACCESS : DMU_OBJECT_END));
1707 /* are there any extended attributes? */
1708 if ((xattr_obj = zp->z_phys->zp_xattr) != 0) {
1709 /* XXX - do we need this if we are deleting? */
1710 dmu_tx_hold_bonus(tx, xattr_obj);
1713 /* are there any additional acls */
1714 if ((acl_obj = zp->z_phys->zp_acl.z_acl_extern_obj) != 0 &&
1716 dmu_tx_hold_free(tx, acl_obj, 0, DMU_OBJECT_END);
1718 /* charge as an update -- would be nice not to charge at all */
1719 dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL);
1721 error = dmu_tx_assign(tx, TXG_NOWAIT);
1723 zfs_dirent_unlock(dl);
1725 if (error == ERESTART) {
1738 * Remove the directory entry.
1740 error = zfs_link_destroy(dl, zp, tx, zflg, &unlinked);
1747 if (0 && unlinked) {
1749 delete_now = may_delete_now && !toobig &&
1750 vp->v_count == 1 && !vn_has_cached_data(vp) &&
1751 zp->z_phys->zp_xattr == xattr_obj &&
1752 zp->z_phys->zp_acl.z_acl_extern_obj == acl_obj;
1757 if (zp->z_phys->zp_xattr) {
1758 error = zfs_zget(zfsvfs, zp->z_phys->zp_xattr, &xzp);
1759 ASSERT3U(error, ==, 0);
1760 ASSERT3U(xzp->z_phys->zp_links, ==, 2);
1761 dmu_buf_will_dirty(xzp->z_dbuf, tx);
1762 mutex_enter(&xzp->z_lock);
1763 xzp->z_unlinked = 1;
1764 xzp->z_phys->zp_links = 0;
1765 mutex_exit(&xzp->z_lock);
1766 zfs_unlinked_add(xzp, tx);
1767 zp->z_phys->zp_xattr = 0; /* probably unnecessary */
1769 mutex_enter(&zp->z_lock);
1772 ASSERT3U(vp->v_count, ==, 0);
1774 mutex_exit(&zp->z_lock);
1775 zfs_znode_delete(zp, tx);
1776 } else if (unlinked) {
1777 zfs_unlinked_add(zp, tx);
1781 if (flags & FIGNORECASE)
1783 zfs_log_remove(zilog, tx, txtype, dzp, name);
1790 zfs_dirent_unlock(dl);
1795 /* this rele is delayed to prevent nesting transactions */
1804 * Create a new directory and insert it into dvp using the name
1805 * provided. Return a pointer to the inserted directory.
1807 * IN: dvp - vnode of directory to add subdir to.
1808 * dirname - name of new directory.
1809 * vap - attributes of new directory.
1810 * cr - credentials of caller.
1811 * ct - caller context
1812 * vsecp - ACL to be set
1814 * OUT: vpp - vnode of created directory.
1816 * RETURN: 0 if success
1817 * error code if failure
1820 * dvp - ctime|mtime updated
1821 * vp - ctime|mtime|atime updated
1825 zfs_mkdir(vnode_t *dvp, char *dirname, vattr_t *vap, vnode_t **vpp, cred_t *cr,
1826 caller_context_t *ct, int flags, vsecattr_t *vsecp)
1828 znode_t *zp, *dzp = VTOZ(dvp);
1829 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
1838 gid_t gid = crgetgid(cr);
1839 zfs_acl_ids_t acl_ids;
1840 boolean_t fuid_dirtied;
1842 ASSERT(vap->va_type == VDIR);
1845 * If we have an ephemeral id, ACL, or XVATTR then
1846 * make sure file system is at proper version
1849 ksid = crgetsid(cr, KSID_OWNER);
1851 uid = ksid_getid(ksid);
1854 if (zfsvfs->z_use_fuids == B_FALSE &&
1855 (vsecp || (vap->va_mask & AT_XVATTR) || IS_EPHEMERAL(crgetuid(cr))||
1856 IS_EPHEMERAL(crgetgid(cr))))
1861 zilog = zfsvfs->z_log;
1863 if (dzp->z_phys->zp_flags & ZFS_XATTR) {
1868 if (zfsvfs->z_utf8 && u8_validate(dirname,
1869 strlen(dirname), NULL, U8_VALIDATE_ENTIRE, &error) < 0) {
1873 if (flags & FIGNORECASE)
1876 if (vap->va_mask & AT_XVATTR)
1877 if ((error = secpolicy_xvattr(dvp, (xvattr_t *)vap,
1878 crgetuid(cr), cr, vap->va_type)) != 0) {
1884 * First make sure the new directory doesn't exist.
1889 if (error = zfs_dirent_lock(&dl, dzp, dirname, &zp, zf,
1895 if (error = zfs_zaccess(dzp, ACE_ADD_SUBDIRECTORY, 0, B_FALSE, cr)) {
1896 zfs_dirent_unlock(dl);
1901 if ((error = zfs_acl_ids_create(dzp, 0, vap, cr, vsecp,
1903 zfs_dirent_unlock(dl);
1907 if (zfs_acl_ids_overquota(zfsvfs, &acl_ids)) {
1908 zfs_acl_ids_free(&acl_ids);
1909 zfs_dirent_unlock(dl);
1915 * Add a new entry to the directory.
1917 tx = dmu_tx_create(zfsvfs->z_os);
1918 dmu_tx_hold_zap(tx, dzp->z_id, TRUE, dirname);
1919 dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL);
1920 fuid_dirtied = zfsvfs->z_fuid_dirty;
1922 zfs_fuid_txhold(zfsvfs, tx);
1923 if (acl_ids.z_aclp->z_acl_bytes > ZFS_ACE_SPACE)
1924 dmu_tx_hold_write(tx, DMU_NEW_OBJECT,
1925 0, SPA_MAXBLOCKSIZE);
1926 error = dmu_tx_assign(tx, TXG_NOWAIT);
1928 zfs_acl_ids_free(&acl_ids);
1929 zfs_dirent_unlock(dl);
1930 if (error == ERESTART) {
1943 zfs_mknode(dzp, vap, tx, cr, 0, &zp, 0, &acl_ids);
1946 zfs_fuid_sync(zfsvfs, tx);
1948 * Now put new name in parent dir.
1950 (void) zfs_link_create(dl, zp, tx, ZNEW);
1954 txtype = zfs_log_create_txtype(Z_DIR, vsecp, vap);
1955 if (flags & FIGNORECASE)
1957 zfs_log_create(zilog, tx, txtype, dzp, zp, dirname, vsecp,
1958 acl_ids.z_fuidp, vap);
1960 zfs_acl_ids_free(&acl_ids);
1963 zfs_dirent_unlock(dl);
1970 * Remove a directory subdir entry. If the current working
1971 * directory is the same as the subdir to be removed, the
1974 * IN: dvp - vnode of directory to remove from.
1975 * name - name of directory to be removed.
1976 * cwd - vnode of current working directory.
1977 * cr - credentials of caller.
1978 * ct - caller context
1979 * flags - case flags
1981 * RETURN: 0 if success
1982 * error code if failure
1985 * dvp - ctime|mtime updated
1989 zfs_rmdir(vnode_t *dvp, char *name, vnode_t *cwd, cred_t *cr,
1990 caller_context_t *ct, int flags)
1992 znode_t *dzp = VTOZ(dvp);
1995 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
2004 zilog = zfsvfs->z_log;
2006 if (flags & FIGNORECASE)
2012 * Attempt to lock directory; fail if entry doesn't exist.
2014 if (error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg,
2022 if (error = zfs_zaccess_delete(dzp, zp, cr)) {
2026 if (vp->v_type != VDIR) {
2036 vnevent_rmdir(vp, dvp, name, ct);
2039 * Grab a lock on the directory to make sure that noone is
2040 * trying to add (or lookup) entries while we are removing it.
2042 rw_enter(&zp->z_name_lock, RW_WRITER);
2045 * Grab a lock on the parent pointer to make sure we play well
2046 * with the treewalk and directory rename code.
2048 rw_enter(&zp->z_parent_lock, RW_WRITER);
2050 tx = dmu_tx_create(zfsvfs->z_os);
2051 dmu_tx_hold_zap(tx, dzp->z_id, FALSE, name);
2052 dmu_tx_hold_bonus(tx, zp->z_id);
2053 dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL);
2054 error = dmu_tx_assign(tx, TXG_NOWAIT);
2056 rw_exit(&zp->z_parent_lock);
2057 rw_exit(&zp->z_name_lock);
2058 zfs_dirent_unlock(dl);
2060 if (error == ERESTART) {
2070 #ifdef FREEBSD_NAMECACHE
2074 error = zfs_link_destroy(dl, zp, tx, zflg, NULL);
2077 uint64_t txtype = TX_RMDIR;
2078 if (flags & FIGNORECASE)
2080 zfs_log_remove(zilog, tx, txtype, dzp, name);
2085 rw_exit(&zp->z_parent_lock);
2086 rw_exit(&zp->z_name_lock);
2087 #ifdef FREEBSD_NAMECACHE
2091 zfs_dirent_unlock(dl);
2100 * Read as many directory entries as will fit into the provided
2101 * buffer from the given directory cursor position (specified in
2102 * the uio structure.
2104 * IN: vp - vnode of directory to read.
2105 * uio - structure supplying read location, range info,
2106 * and return buffer.
2107 * cr - credentials of caller.
2108 * ct - caller context
2109 * flags - case flags
2111 * OUT: uio - updated offset and range, buffer filled.
2112 * eofp - set to true if end-of-file detected.
2114 * RETURN: 0 if success
2115 * error code if failure
2118 * vp - atime updated
2120 * Note that the low 4 bits of the cookie returned by zap is always zero.
2121 * This allows us to use the low range for "special" directory entries:
2122 * We use 0 for '.', and 1 for '..'. If this is the root of the filesystem,
2123 * we use the offset 2 for the '.zfs' directory.
2127 zfs_readdir(vnode_t *vp, uio_t *uio, cred_t *cr, int *eofp, int *ncookies, u_long **cookies)
2129 znode_t *zp = VTOZ(vp);
2133 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
2138 zap_attribute_t zap;
2139 uint_t bytes_wanted;
2140 uint64_t offset; /* must be unsigned; checks for < 1 */
2145 boolean_t check_sysattrs;
2148 u_long *cooks = NULL;
2155 * If we are not given an eof variable,
2162 * Check for valid iov_len.
2164 if (uio->uio_iov->iov_len <= 0) {
2170 * Quit if directory has been removed (posix)
2172 if ((*eofp = zp->z_unlinked) != 0) {
2179 offset = uio->uio_loffset;
2180 prefetch = zp->z_zn_prefetch;
2183 * Initialize the iterator cursor.
2187 * Start iteration from the beginning of the directory.
2189 zap_cursor_init(&zc, os, zp->z_id);
2192 * The offset is a serialized cursor.
2194 zap_cursor_init_serialized(&zc, os, zp->z_id, offset);
2198 * Get space to change directory entries into fs independent format.
2200 iovp = uio->uio_iov;
2201 bytes_wanted = iovp->iov_len;
2202 if (uio->uio_segflg != UIO_SYSSPACE || uio->uio_iovcnt != 1) {
2203 bufsize = bytes_wanted;
2204 outbuf = kmem_alloc(bufsize, KM_SLEEP);
2205 odp = (struct dirent64 *)outbuf;
2207 bufsize = bytes_wanted;
2208 odp = (struct dirent64 *)iovp->iov_base;
2210 eodp = (struct edirent *)odp;
2212 if (ncookies != NULL) {
2214 * Minimum entry size is dirent size and 1 byte for a file name.
2216 ncooks = uio->uio_resid / (sizeof(struct dirent) - sizeof(((struct dirent *)NULL)->d_name) + 1);
2217 cooks = malloc(ncooks * sizeof(u_long), M_TEMP, M_WAITOK);
2222 * If this VFS supports the system attribute view interface; and
2223 * we're looking at an extended attribute directory; and we care
2224 * about normalization conflicts on this vfs; then we must check
2225 * for normalization conflicts with the sysattr name space.
2228 check_sysattrs = vfs_has_feature(vp->v_vfsp, VFSFT_SYSATTR_VIEWS) &&
2229 (vp->v_flag & V_XATTRDIR) && zfsvfs->z_norm &&
2230 (flags & V_RDDIR_ENTFLAGS);
2236 * Transform to file-system independent format
2239 while (outcount < bytes_wanted) {
2245 * Special case `.', `..', and `.zfs'.
2248 (void) strcpy(zap.za_name, ".");
2249 zap.za_normalization_conflict = 0;
2252 } else if (offset == 1) {
2253 (void) strcpy(zap.za_name, "..");
2254 zap.za_normalization_conflict = 0;
2255 objnum = zp->z_phys->zp_parent;
2257 } else if (offset == 2 && zfs_show_ctldir(zp)) {
2258 (void) strcpy(zap.za_name, ZFS_CTLDIR_NAME);
2259 zap.za_normalization_conflict = 0;
2260 objnum = ZFSCTL_INO_ROOT;
2266 if (error = zap_cursor_retrieve(&zc, &zap)) {
2267 if ((*eofp = (error == ENOENT)) != 0)
2273 if (zap.za_integer_length != 8 ||
2274 zap.za_num_integers != 1) {
2275 cmn_err(CE_WARN, "zap_readdir: bad directory "
2276 "entry, obj = %lld, offset = %lld\n",
2277 (u_longlong_t)zp->z_id,
2278 (u_longlong_t)offset);
2283 objnum = ZFS_DIRENT_OBJ(zap.za_first_integer);
2285 * MacOS X can extract the object type here such as:
2286 * uint8_t type = ZFS_DIRENT_TYPE(zap.za_first_integer);
2288 type = ZFS_DIRENT_TYPE(zap.za_first_integer);
2290 if (check_sysattrs && !zap.za_normalization_conflict) {
2292 zap.za_normalization_conflict =
2293 xattr_sysattr_casechk(zap.za_name);
2295 panic("%s:%u: TODO", __func__, __LINE__);
2300 if (flags & V_RDDIR_ACCFILTER) {
2302 * If we have no access at all, don't include
2303 * this entry in the returned information
2306 if (zfs_zget(zp->z_zfsvfs, objnum, &ezp) != 0)
2308 if (!zfs_has_access(ezp, cr)) {
2315 if (flags & V_RDDIR_ENTFLAGS)
2316 reclen = EDIRENT_RECLEN(strlen(zap.za_name));
2318 reclen = DIRENT64_RECLEN(strlen(zap.za_name));
2321 * Will this entry fit in the buffer?
2323 if (outcount + reclen > bufsize) {
2325 * Did we manage to fit anything in the buffer?
2333 if (flags & V_RDDIR_ENTFLAGS) {
2335 * Add extended flag entry:
2337 eodp->ed_ino = objnum;
2338 eodp->ed_reclen = reclen;
2339 /* NOTE: ed_off is the offset for the *next* entry */
2340 next = &(eodp->ed_off);
2341 eodp->ed_eflags = zap.za_normalization_conflict ?
2342 ED_CASE_CONFLICT : 0;
2343 (void) strncpy(eodp->ed_name, zap.za_name,
2344 EDIRENT_NAMELEN(reclen));
2345 eodp = (edirent_t *)((intptr_t)eodp + reclen);
2350 odp->d_ino = objnum;
2351 odp->d_reclen = reclen;
2352 odp->d_namlen = strlen(zap.za_name);
2353 (void) strlcpy(odp->d_name, zap.za_name, odp->d_namlen + 1);
2355 odp = (dirent64_t *)((intptr_t)odp + reclen);
2359 ASSERT(outcount <= bufsize);
2361 /* Prefetch znode */
2363 dmu_prefetch(os, objnum, 0, 0);
2367 * Move to the next entry, fill in the previous offset.
2369 if (offset > 2 || (offset == 2 && !zfs_show_ctldir(zp))) {
2370 zap_cursor_advance(&zc);
2371 offset = zap_cursor_serialize(&zc);
2376 if (cooks != NULL) {
2379 KASSERT(ncooks >= 0, ("ncookies=%d", ncooks));
2382 zp->z_zn_prefetch = B_FALSE; /* a lookup will re-enable pre-fetching */
2384 /* Subtract unused cookies */
2385 if (ncookies != NULL)
2386 *ncookies -= ncooks;
2388 if (uio->uio_segflg == UIO_SYSSPACE && uio->uio_iovcnt == 1) {
2389 iovp->iov_base += outcount;
2390 iovp->iov_len -= outcount;
2391 uio->uio_resid -= outcount;
2392 } else if (error = uiomove(outbuf, (long)outcount, UIO_READ, uio)) {
2394 * Reset the pointer.
2396 offset = uio->uio_loffset;
2400 zap_cursor_fini(&zc);
2401 if (uio->uio_segflg != UIO_SYSSPACE || uio->uio_iovcnt != 1)
2402 kmem_free(outbuf, bufsize);
2404 if (error == ENOENT)
2407 ZFS_ACCESSTIME_STAMP(zfsvfs, zp);
2409 uio->uio_loffset = offset;
2411 if (error != 0 && cookies != NULL) {
2412 free(*cookies, M_TEMP);
2419 ulong_t zfs_fsync_sync_cnt = 4;
2422 zfs_fsync(vnode_t *vp, int syncflag, cred_t *cr, caller_context_t *ct)
2424 znode_t *zp = VTOZ(vp);
2425 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
2427 (void) tsd_set(zfs_fsyncer_key, (void *)zfs_fsync_sync_cnt);
2431 zil_commit(zfsvfs->z_log, zp->z_last_itx, zp->z_id);
2438 * Get the requested file attributes and place them in the provided
2441 * IN: vp - vnode of file.
2442 * vap - va_mask identifies requested attributes.
2443 * If AT_XVATTR set, then optional attrs are requested
2444 * flags - ATTR_NOACLCHECK (CIFS server context)
2445 * cr - credentials of caller.
2446 * ct - caller context
2448 * OUT: vap - attribute values.
2450 * RETURN: 0 (always succeeds)
2454 zfs_getattr(vnode_t *vp, vattr_t *vap, int flags, cred_t *cr,
2455 caller_context_t *ct)
2457 znode_t *zp = VTOZ(vp);
2458 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
2462 u_longlong_t nblocks;
2464 xvattr_t *xvap = (xvattr_t *)vap; /* vap may be an xvattr_t * */
2465 xoptattr_t *xoap = NULL;
2466 boolean_t skipaclchk = (flags & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE;
2473 * If ACL is trivial don't bother looking for ACE_READ_ATTRIBUTES.
2474 * Also, if we are the owner don't bother, since owner should
2475 * always be allowed to read basic attributes of file.
2477 if (!(pzp->zp_flags & ZFS_ACL_TRIVIAL) &&
2478 (pzp->zp_uid != crgetuid(cr))) {
2479 if (error = zfs_zaccess(zp, ACE_READ_ATTRIBUTES, 0,
2487 * Return all attributes. It's cheaper to provide the answer
2488 * than to determine whether we were asked the question.
2491 mutex_enter(&zp->z_lock);
2492 vap->va_type = IFTOVT(pzp->zp_mode);
2493 vap->va_mode = pzp->zp_mode & ~S_IFMT;
2494 zfs_fuid_map_ids(zp, cr, &vap->va_uid, &vap->va_gid);
2495 // vap->va_fsid = zp->z_zfsvfs->z_vfs->vfs_dev;
2496 vap->va_nodeid = zp->z_id;
2497 if ((vp->v_flag & VROOT) && zfs_show_ctldir(zp))
2498 links = pzp->zp_links + 1;
2500 links = pzp->zp_links;
2501 vap->va_nlink = MIN(links, UINT32_MAX); /* nlink_t limit! */
2502 vap->va_size = pzp->zp_size;
2503 vap->va_fsid = vp->v_mount->mnt_stat.f_fsid.val[0];
2504 vap->va_rdev = zfs_cmpldev(pzp->zp_rdev);
2505 vap->va_seq = zp->z_seq;
2506 vap->va_flags = 0; /* FreeBSD: Reset chflags(2) flags. */
2509 * Add in any requested optional attributes and the create time.
2510 * Also set the corresponding bits in the returned attribute bitmap.
2512 if ((xoap = xva_getxoptattr(xvap)) != NULL && zfsvfs->z_use_fuids) {
2513 if (XVA_ISSET_REQ(xvap, XAT_ARCHIVE)) {
2515 ((pzp->zp_flags & ZFS_ARCHIVE) != 0);
2516 XVA_SET_RTN(xvap, XAT_ARCHIVE);
2519 if (XVA_ISSET_REQ(xvap, XAT_READONLY)) {
2520 xoap->xoa_readonly =
2521 ((pzp->zp_flags & ZFS_READONLY) != 0);
2522 XVA_SET_RTN(xvap, XAT_READONLY);
2525 if (XVA_ISSET_REQ(xvap, XAT_SYSTEM)) {
2527 ((pzp->zp_flags & ZFS_SYSTEM) != 0);
2528 XVA_SET_RTN(xvap, XAT_SYSTEM);
2531 if (XVA_ISSET_REQ(xvap, XAT_HIDDEN)) {
2533 ((pzp->zp_flags & ZFS_HIDDEN) != 0);
2534 XVA_SET_RTN(xvap, XAT_HIDDEN);
2537 if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK)) {
2538 xoap->xoa_nounlink =
2539 ((pzp->zp_flags & ZFS_NOUNLINK) != 0);
2540 XVA_SET_RTN(xvap, XAT_NOUNLINK);
2543 if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE)) {
2544 xoap->xoa_immutable =
2545 ((pzp->zp_flags & ZFS_IMMUTABLE) != 0);
2546 XVA_SET_RTN(xvap, XAT_IMMUTABLE);
2549 if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY)) {
2550 xoap->xoa_appendonly =
2551 ((pzp->zp_flags & ZFS_APPENDONLY) != 0);
2552 XVA_SET_RTN(xvap, XAT_APPENDONLY);
2555 if (XVA_ISSET_REQ(xvap, XAT_NODUMP)) {
2557 ((pzp->zp_flags & ZFS_NODUMP) != 0);
2558 XVA_SET_RTN(xvap, XAT_NODUMP);
2561 if (XVA_ISSET_REQ(xvap, XAT_OPAQUE)) {
2563 ((pzp->zp_flags & ZFS_OPAQUE) != 0);
2564 XVA_SET_RTN(xvap, XAT_OPAQUE);
2567 if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED)) {
2568 xoap->xoa_av_quarantined =
2569 ((pzp->zp_flags & ZFS_AV_QUARANTINED) != 0);
2570 XVA_SET_RTN(xvap, XAT_AV_QUARANTINED);
2573 if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED)) {
2574 xoap->xoa_av_modified =
2575 ((pzp->zp_flags & ZFS_AV_MODIFIED) != 0);
2576 XVA_SET_RTN(xvap, XAT_AV_MODIFIED);
2579 if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP) &&
2580 vp->v_type == VREG &&
2581 (pzp->zp_flags & ZFS_BONUS_SCANSTAMP)) {
2583 dmu_object_info_t doi;
2586 * Only VREG files have anti-virus scanstamps, so we
2587 * won't conflict with symlinks in the bonus buffer.
2589 dmu_object_info_from_db(zp->z_dbuf, &doi);
2590 len = sizeof (xoap->xoa_av_scanstamp) +
2591 sizeof (znode_phys_t);
2592 if (len <= doi.doi_bonus_size) {
2594 * pzp points to the start of the
2595 * znode_phys_t. pzp + 1 points to the
2596 * first byte after the znode_phys_t.
2598 (void) memcpy(xoap->xoa_av_scanstamp,
2600 sizeof (xoap->xoa_av_scanstamp));
2601 XVA_SET_RTN(xvap, XAT_AV_SCANSTAMP);
2605 if (XVA_ISSET_REQ(xvap, XAT_CREATETIME)) {
2606 ZFS_TIME_DECODE(&xoap->xoa_createtime, pzp->zp_crtime);
2607 XVA_SET_RTN(xvap, XAT_CREATETIME);
2611 ZFS_TIME_DECODE(&vap->va_atime, pzp->zp_atime);
2612 ZFS_TIME_DECODE(&vap->va_mtime, pzp->zp_mtime);
2613 ZFS_TIME_DECODE(&vap->va_ctime, pzp->zp_ctime);
2614 ZFS_TIME_DECODE(&vap->va_birthtime, pzp->zp_crtime);
2616 mutex_exit(&zp->z_lock);
2618 dmu_object_size_from_db(zp->z_dbuf, &blksize, &nblocks);
2619 vap->va_blksize = blksize;
2620 vap->va_bytes = nblocks << 9; /* nblocks * 512 */
2622 if (zp->z_blksz == 0) {
2624 * Block size hasn't been set; suggest maximal I/O transfers.
2626 vap->va_blksize = zfsvfs->z_max_blksz;
2634 * Set the file attributes to the values contained in the
2637 * IN: vp - vnode of file to be modified.
2638 * vap - new attribute values.
2639 * If AT_XVATTR set, then optional attrs are being set
2640 * flags - ATTR_UTIME set if non-default time values provided.
2641 * - ATTR_NOACLCHECK (CIFS context only).
2642 * cr - credentials of caller.
2643 * ct - caller context
2645 * RETURN: 0 if success
2646 * error code if failure
2649 * vp - ctime updated, mtime updated if size changed.
2653 zfs_setattr(vnode_t *vp, vattr_t *vap, int flags, cred_t *cr,
2654 caller_context_t *ct)
2656 znode_t *zp = VTOZ(vp);
2658 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
2663 uint_t mask = vap->va_mask;
2665 uint64_t saved_mode;
2668 uint64_t new_uid, new_gid;
2670 int need_policy = FALSE;
2672 zfs_fuid_info_t *fuidp = NULL;
2673 xvattr_t *xvap = (xvattr_t *)vap; /* vap may be an xvattr_t * */
2675 zfs_acl_t *aclp = NULL;
2676 boolean_t skipaclchk = (flags & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE;
2677 boolean_t fuid_dirtied = B_FALSE;
2682 if (mask & AT_NOSET)
2689 zilog = zfsvfs->z_log;
2692 * Make sure that if we have ephemeral uid/gid or xvattr specified
2693 * that file system is at proper version level
2696 if (zfsvfs->z_use_fuids == B_FALSE &&
2697 (((mask & AT_UID) && IS_EPHEMERAL(vap->va_uid)) ||
2698 ((mask & AT_GID) && IS_EPHEMERAL(vap->va_gid)) ||
2699 (mask & AT_XVATTR))) {
2704 if (mask & AT_SIZE && vp->v_type == VDIR) {
2709 if (mask & AT_SIZE && vp->v_type != VREG && vp->v_type != VFIFO) {
2715 * If this is an xvattr_t, then get a pointer to the structure of
2716 * optional attributes. If this is NULL, then we have a vattr_t.
2718 xoap = xva_getxoptattr(xvap);
2720 xva_init(&tmpxvattr);
2723 * Immutable files can only alter immutable bit and atime
2725 if ((pzp->zp_flags & ZFS_IMMUTABLE) &&
2726 ((mask & (AT_SIZE|AT_UID|AT_GID|AT_MTIME|AT_MODE)) ||
2727 ((mask & AT_XVATTR) && XVA_ISSET_REQ(xvap, XAT_CREATETIME)))) {
2732 if ((mask & AT_SIZE) && (pzp->zp_flags & ZFS_READONLY)) {
2738 * Verify timestamps doesn't overflow 32 bits.
2739 * ZFS can handle large timestamps, but 32bit syscalls can't
2740 * handle times greater than 2039. This check should be removed
2741 * once large timestamps are fully supported.
2743 if (mask & (AT_ATIME | AT_MTIME)) {
2744 if (((mask & AT_ATIME) && TIMESPEC_OVERFLOW(&vap->va_atime)) ||
2745 ((mask & AT_MTIME) && TIMESPEC_OVERFLOW(&vap->va_mtime))) {
2754 /* Can this be moved to before the top label? */
2755 if (zfsvfs->z_vfs->vfs_flag & VFS_RDONLY) {
2761 * First validate permissions
2764 if (mask & AT_SIZE) {
2765 err = zfs_zaccess(zp, ACE_WRITE_DATA, 0, skipaclchk, cr);
2771 * XXX - Note, we are not providing any open
2772 * mode flags here (like FNDELAY), so we may
2773 * block if there are locks present... this
2774 * should be addressed in openat().
2776 /* XXX - would it be OK to generate a log record here? */
2777 err = zfs_freesp(zp, vap->va_size, 0, 0, FALSE);
2784 if (mask & (AT_ATIME|AT_MTIME) ||
2785 ((mask & AT_XVATTR) && (XVA_ISSET_REQ(xvap, XAT_HIDDEN) ||
2786 XVA_ISSET_REQ(xvap, XAT_READONLY) ||
2787 XVA_ISSET_REQ(xvap, XAT_ARCHIVE) ||
2788 XVA_ISSET_REQ(xvap, XAT_CREATETIME) ||
2789 XVA_ISSET_REQ(xvap, XAT_SYSTEM))))
2790 need_policy = zfs_zaccess(zp, ACE_WRITE_ATTRIBUTES, 0,
2793 if (mask & (AT_UID|AT_GID)) {
2794 int idmask = (mask & (AT_UID|AT_GID));
2799 * NOTE: even if a new mode is being set,
2800 * we may clear S_ISUID/S_ISGID bits.
2803 if (!(mask & AT_MODE))
2804 vap->va_mode = pzp->zp_mode;
2807 * Take ownership or chgrp to group we are a member of
2810 take_owner = (mask & AT_UID) && (vap->va_uid == crgetuid(cr));
2811 take_group = (mask & AT_GID) &&
2812 zfs_groupmember(zfsvfs, vap->va_gid, cr);
2815 * If both AT_UID and AT_GID are set then take_owner and
2816 * take_group must both be set in order to allow taking
2819 * Otherwise, send the check through secpolicy_vnode_setattr()
2823 if (((idmask == (AT_UID|AT_GID)) && take_owner && take_group) ||
2824 ((idmask == AT_UID) && take_owner) ||
2825 ((idmask == AT_GID) && take_group)) {
2826 if (zfs_zaccess(zp, ACE_WRITE_OWNER, 0,
2827 skipaclchk, cr) == 0) {
2829 * Remove setuid/setgid for non-privileged users
2831 secpolicy_setid_clear(vap, vp, cr);
2832 trim_mask = (mask & (AT_UID|AT_GID));
2841 mutex_enter(&zp->z_lock);
2842 oldva.va_mode = pzp->zp_mode;
2843 zfs_fuid_map_ids(zp, cr, &oldva.va_uid, &oldva.va_gid);
2844 if (mask & AT_XVATTR) {
2846 * Update xvattr mask to include only those attributes
2847 * that are actually changing.
2849 * the bits will be restored prior to actually setting
2850 * the attributes so the caller thinks they were set.
2852 if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY)) {
2853 if (xoap->xoa_appendonly !=
2854 ((pzp->zp_flags & ZFS_APPENDONLY) != 0)) {
2857 XVA_CLR_REQ(xvap, XAT_APPENDONLY);
2858 XVA_SET_REQ(&tmpxvattr, XAT_APPENDONLY);
2862 if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK)) {
2863 if (xoap->xoa_nounlink !=
2864 ((pzp->zp_flags & ZFS_NOUNLINK) != 0)) {
2867 XVA_CLR_REQ(xvap, XAT_NOUNLINK);
2868 XVA_SET_REQ(&tmpxvattr, XAT_NOUNLINK);
2872 if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE)) {
2873 if (xoap->xoa_immutable !=
2874 ((pzp->zp_flags & ZFS_IMMUTABLE) != 0)) {
2877 XVA_CLR_REQ(xvap, XAT_IMMUTABLE);
2878 XVA_SET_REQ(&tmpxvattr, XAT_IMMUTABLE);
2882 if (XVA_ISSET_REQ(xvap, XAT_NODUMP)) {
2883 if (xoap->xoa_nodump !=
2884 ((pzp->zp_flags & ZFS_NODUMP) != 0)) {
2887 XVA_CLR_REQ(xvap, XAT_NODUMP);
2888 XVA_SET_REQ(&tmpxvattr, XAT_NODUMP);
2892 if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED)) {
2893 if (xoap->xoa_av_modified !=
2894 ((pzp->zp_flags & ZFS_AV_MODIFIED) != 0)) {
2897 XVA_CLR_REQ(xvap, XAT_AV_MODIFIED);
2898 XVA_SET_REQ(&tmpxvattr, XAT_AV_MODIFIED);
2902 if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED)) {
2903 if ((vp->v_type != VREG &&
2904 xoap->xoa_av_quarantined) ||
2905 xoap->xoa_av_quarantined !=
2906 ((pzp->zp_flags & ZFS_AV_QUARANTINED) != 0)) {
2909 XVA_CLR_REQ(xvap, XAT_AV_QUARANTINED);
2910 XVA_SET_REQ(&tmpxvattr, XAT_AV_QUARANTINED);
2914 if (need_policy == FALSE &&
2915 (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP) ||
2916 XVA_ISSET_REQ(xvap, XAT_OPAQUE))) {
2921 mutex_exit(&zp->z_lock);
2923 if (mask & AT_MODE) {
2924 if (zfs_zaccess(zp, ACE_WRITE_ACL, 0, skipaclchk, cr) == 0) {
2925 err = secpolicy_setid_setsticky_clear(vp, vap,
2931 trim_mask |= AT_MODE;
2939 * If trim_mask is set then take ownership
2940 * has been granted or write_acl is present and user
2941 * has the ability to modify mode. In that case remove
2942 * UID|GID and or MODE from mask so that
2943 * secpolicy_vnode_setattr() doesn't revoke it.
2947 saved_mask = vap->va_mask;
2948 vap->va_mask &= ~trim_mask;
2949 if (trim_mask & AT_MODE) {
2951 * Save the mode, as secpolicy_vnode_setattr()
2952 * will overwrite it with ova.va_mode.
2954 saved_mode = vap->va_mode;
2957 err = secpolicy_vnode_setattr(cr, vp, vap, &oldva, flags,
2958 (int (*)(void *, int, cred_t *))zfs_zaccess_unix, zp);
2965 vap->va_mask |= saved_mask;
2966 if (trim_mask & AT_MODE) {
2968 * Recover the mode after
2969 * secpolicy_vnode_setattr().
2971 vap->va_mode = saved_mode;
2977 * secpolicy_vnode_setattr, or take ownership may have
2980 mask = vap->va_mask;
2982 tx = dmu_tx_create(zfsvfs->z_os);
2983 dmu_tx_hold_bonus(tx, zp->z_id);
2985 if (mask & AT_MODE) {
2986 uint64_t pmode = pzp->zp_mode;
2988 new_mode = (pmode & S_IFMT) | (vap->va_mode & ~S_IFMT);
2990 if (err = zfs_acl_chmod_setattr(zp, &aclp, new_mode))
2992 if (pzp->zp_acl.z_acl_extern_obj) {
2993 /* Are we upgrading ACL from old V0 format to new V1 */
2994 if (zfsvfs->z_version <= ZPL_VERSION_FUID &&
2995 pzp->zp_acl.z_acl_version ==
2996 ZFS_ACL_VERSION_INITIAL) {
2997 dmu_tx_hold_free(tx,
2998 pzp->zp_acl.z_acl_extern_obj, 0,
3000 dmu_tx_hold_write(tx, DMU_NEW_OBJECT,
3001 0, aclp->z_acl_bytes);
3003 dmu_tx_hold_write(tx,
3004 pzp->zp_acl.z_acl_extern_obj, 0,
3007 } else if (aclp->z_acl_bytes > ZFS_ACE_SPACE) {
3008 dmu_tx_hold_write(tx, DMU_NEW_OBJECT,
3009 0, aclp->z_acl_bytes);
3013 if (mask & (AT_UID | AT_GID)) {
3014 if (pzp->zp_xattr) {
3015 err = zfs_zget(zp->z_zfsvfs, pzp->zp_xattr, &attrzp);
3018 dmu_tx_hold_bonus(tx, attrzp->z_id);
3020 if (mask & AT_UID) {
3021 new_uid = zfs_fuid_create(zfsvfs,
3022 (uint64_t)vap->va_uid, cr, ZFS_OWNER, &fuidp);
3023 if (new_uid != pzp->zp_uid &&
3024 zfs_usergroup_overquota(zfsvfs, B_FALSE, new_uid)) {
3030 if (mask & AT_GID) {
3031 new_gid = zfs_fuid_create(zfsvfs, (uint64_t)vap->va_gid,
3032 cr, ZFS_GROUP, &fuidp);
3033 if (new_gid != pzp->zp_gid &&
3034 zfs_usergroup_overquota(zfsvfs, B_TRUE, new_gid)) {
3039 fuid_dirtied = zfsvfs->z_fuid_dirty;
3041 if (zfsvfs->z_fuid_obj == 0) {
3042 dmu_tx_hold_bonus(tx, DMU_NEW_OBJECT);
3043 dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0,
3044 FUID_SIZE_ESTIMATE(zfsvfs));
3045 dmu_tx_hold_zap(tx, MASTER_NODE_OBJ,
3048 dmu_tx_hold_bonus(tx, zfsvfs->z_fuid_obj);
3049 dmu_tx_hold_write(tx, zfsvfs->z_fuid_obj, 0,
3050 FUID_SIZE_ESTIMATE(zfsvfs));
3055 err = dmu_tx_assign(tx, TXG_NOWAIT);
3057 if (err == ERESTART)
3062 dmu_buf_will_dirty(zp->z_dbuf, tx);
3065 * Set each attribute requested.
3066 * We group settings according to the locks they need to acquire.
3068 * Note: you cannot set ctime directly, although it will be
3069 * updated as a side-effect of calling this function.
3072 mutex_enter(&zp->z_lock);
3074 if (mask & AT_MODE) {
3075 mutex_enter(&zp->z_acl_lock);
3076 zp->z_phys->zp_mode = new_mode;
3077 err = zfs_aclset_common(zp, aclp, cr, tx);
3078 ASSERT3U(err, ==, 0);
3079 zp->z_acl_cached = aclp;
3081 mutex_exit(&zp->z_acl_lock);
3085 mutex_enter(&attrzp->z_lock);
3087 if (mask & AT_UID) {
3088 pzp->zp_uid = new_uid;
3090 attrzp->z_phys->zp_uid = new_uid;
3093 if (mask & AT_GID) {
3094 pzp->zp_gid = new_gid;
3096 attrzp->z_phys->zp_gid = new_gid;
3100 mutex_exit(&attrzp->z_lock);
3102 if (mask & AT_ATIME)
3103 ZFS_TIME_ENCODE(&vap->va_atime, pzp->zp_atime);
3105 if (mask & AT_MTIME)
3106 ZFS_TIME_ENCODE(&vap->va_mtime, pzp->zp_mtime);
3108 /* XXX - shouldn't this be done *before* the ATIME/MTIME checks? */
3110 zfs_time_stamper_locked(zp, CONTENT_MODIFIED, tx);
3112 zfs_time_stamper_locked(zp, STATE_CHANGED, tx);
3114 * Do this after setting timestamps to prevent timestamp
3115 * update from toggling bit
3118 if (xoap && (mask & AT_XVATTR)) {
3121 * restore trimmed off masks
3122 * so that return masks can be set for caller.
3125 if (XVA_ISSET_REQ(&tmpxvattr, XAT_APPENDONLY)) {
3126 XVA_SET_REQ(xvap, XAT_APPENDONLY);
3128 if (XVA_ISSET_REQ(&tmpxvattr, XAT_NOUNLINK)) {
3129 XVA_SET_REQ(xvap, XAT_NOUNLINK);
3131 if (XVA_ISSET_REQ(&tmpxvattr, XAT_IMMUTABLE)) {
3132 XVA_SET_REQ(xvap, XAT_IMMUTABLE);
3134 if (XVA_ISSET_REQ(&tmpxvattr, XAT_NODUMP)) {
3135 XVA_SET_REQ(xvap, XAT_NODUMP);
3137 if (XVA_ISSET_REQ(&tmpxvattr, XAT_AV_MODIFIED)) {
3138 XVA_SET_REQ(xvap, XAT_AV_MODIFIED);
3140 if (XVA_ISSET_REQ(&tmpxvattr, XAT_AV_QUARANTINED)) {
3141 XVA_SET_REQ(xvap, XAT_AV_QUARANTINED);
3144 if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP)) {
3146 dmu_object_info_t doi;
3148 ASSERT(vp->v_type == VREG);
3150 /* Grow the bonus buffer if necessary. */
3151 dmu_object_info_from_db(zp->z_dbuf, &doi);
3152 len = sizeof (xoap->xoa_av_scanstamp) +
3153 sizeof (znode_phys_t);
3154 if (len > doi.doi_bonus_size)
3155 VERIFY(dmu_set_bonus(zp->z_dbuf, len, tx) == 0);
3157 zfs_xvattr_set(zp, xvap);
3161 zfs_fuid_sync(zfsvfs, tx);
3164 zfs_log_setattr(zilog, tx, TX_SETATTR, zp, vap, mask, fuidp);
3166 mutex_exit(&zp->z_lock);
3170 VN_RELE(ZTOV(attrzp));
3176 zfs_fuid_info_free(fuidp);
3185 if (err == ERESTART)
3192 typedef struct zfs_zlock {
3193 krwlock_t *zl_rwlock; /* lock we acquired */
3194 znode_t *zl_znode; /* znode we held */
3195 struct zfs_zlock *zl_next; /* next in list */
3199 * Drop locks and release vnodes that were held by zfs_rename_lock().
3202 zfs_rename_unlock(zfs_zlock_t **zlpp)
3206 while ((zl = *zlpp) != NULL) {
3207 if (zl->zl_znode != NULL)
3208 VN_RELE(ZTOV(zl->zl_znode));
3209 rw_exit(zl->zl_rwlock);
3210 *zlpp = zl->zl_next;
3211 kmem_free(zl, sizeof (*zl));
3216 * Search back through the directory tree, using the ".." entries.
3217 * Lock each directory in the chain to prevent concurrent renames.
3218 * Fail any attempt to move a directory into one of its own descendants.
3219 * XXX - z_parent_lock can overlap with map or grow locks
3222 zfs_rename_lock(znode_t *szp, znode_t *tdzp, znode_t *sdzp, zfs_zlock_t **zlpp)
3226 uint64_t rootid = zp->z_zfsvfs->z_root;
3227 uint64_t *oidp = &zp->z_id;
3228 krwlock_t *rwlp = &szp->z_parent_lock;
3229 krw_t rw = RW_WRITER;
3232 * First pass write-locks szp and compares to zp->z_id.
3233 * Later passes read-lock zp and compare to zp->z_parent.
3236 if (!rw_tryenter(rwlp, rw)) {
3238 * Another thread is renaming in this path.
3239 * Note that if we are a WRITER, we don't have any
3240 * parent_locks held yet.
3242 if (rw == RW_READER && zp->z_id > szp->z_id) {
3244 * Drop our locks and restart
3246 zfs_rename_unlock(&zl);
3250 rwlp = &szp->z_parent_lock;
3255 * Wait for other thread to drop its locks
3261 zl = kmem_alloc(sizeof (*zl), KM_SLEEP);
3262 zl->zl_rwlock = rwlp;
3263 zl->zl_znode = NULL;
3264 zl->zl_next = *zlpp;
3267 if (*oidp == szp->z_id) /* We're a descendant of szp */
3270 if (*oidp == rootid) /* We've hit the top */
3273 if (rw == RW_READER) { /* i.e. not the first pass */
3274 int error = zfs_zget(zp->z_zfsvfs, *oidp, &zp);
3279 oidp = &zp->z_phys->zp_parent;
3280 rwlp = &zp->z_parent_lock;
3283 } while (zp->z_id != sdzp->z_id);
3289 * Move an entry from the provided source directory to the target
3290 * directory. Change the entry name as indicated.
3292 * IN: sdvp - Source directory containing the "old entry".
3293 * snm - Old entry name.
3294 * tdvp - Target directory to contain the "new entry".
3295 * tnm - New entry name.
3296 * cr - credentials of caller.
3297 * ct - caller context
3298 * flags - case flags
3300 * RETURN: 0 if success
3301 * error code if failure
3304 * sdvp,tdvp - ctime|mtime updated
3308 zfs_rename(vnode_t *sdvp, char *snm, vnode_t *tdvp, char *tnm, cred_t *cr,
3309 caller_context_t *ct, int flags)
3311 znode_t *tdzp, *szp, *tzp;
3312 znode_t *sdzp = VTOZ(sdvp);
3313 zfsvfs_t *zfsvfs = sdzp->z_zfsvfs;
3316 zfs_dirlock_t *sdl, *tdl;
3319 int cmp, serr, terr;
3324 ZFS_VERIFY_ZP(sdzp);
3325 zilog = zfsvfs->z_log;
3328 * Make sure we have the real vp for the target directory.
3330 if (VOP_REALVP(tdvp, &realvp, ct) == 0)
3333 if (tdvp->v_vfsp != sdvp->v_vfsp || zfsctl_is_node(tdvp)) {
3339 ZFS_VERIFY_ZP(tdzp);
3340 if (zfsvfs->z_utf8 && u8_validate(tnm,
3341 strlen(tnm), NULL, U8_VALIDATE_ENTIRE, &error) < 0) {
3346 if (flags & FIGNORECASE)
3355 * This is to prevent the creation of links into attribute space
3356 * by renaming a linked file into/outof an attribute directory.
3357 * See the comment in zfs_link() for why this is considered bad.
3359 if ((tdzp->z_phys->zp_flags & ZFS_XATTR) !=
3360 (sdzp->z_phys->zp_flags & ZFS_XATTR)) {
3366 * Lock source and target directory entries. To prevent deadlock,
3367 * a lock ordering must be defined. We lock the directory with
3368 * the smallest object id first, or if it's a tie, the one with
3369 * the lexically first name.
3371 if (sdzp->z_id < tdzp->z_id) {
3373 } else if (sdzp->z_id > tdzp->z_id) {
3377 * First compare the two name arguments without
3378 * considering any case folding.
3380 int nofold = (zfsvfs->z_norm & ~U8_TEXTPREP_TOUPPER);
3382 cmp = u8_strcmp(snm, tnm, 0, nofold, U8_UNICODE_LATEST, &error);
3383 ASSERT(error == 0 || !zfsvfs->z_utf8);
3386 * POSIX: "If the old argument and the new argument
3387 * both refer to links to the same existing file,
3388 * the rename() function shall return successfully
3389 * and perform no other action."
3395 * If the file system is case-folding, then we may
3396 * have some more checking to do. A case-folding file
3397 * system is either supporting mixed case sensitivity
3398 * access or is completely case-insensitive. Note
3399 * that the file system is always case preserving.
3401 * In mixed sensitivity mode case sensitive behavior
3402 * is the default. FIGNORECASE must be used to
3403 * explicitly request case insensitive behavior.
3405 * If the source and target names provided differ only
3406 * by case (e.g., a request to rename 'tim' to 'Tim'),
3407 * we will treat this as a special case in the
3408 * case-insensitive mode: as long as the source name
3409 * is an exact match, we will allow this to proceed as
3410 * a name-change request.
3412 if ((zfsvfs->z_case == ZFS_CASE_INSENSITIVE ||
3413 (zfsvfs->z_case == ZFS_CASE_MIXED &&
3414 flags & FIGNORECASE)) &&
3415 u8_strcmp(snm, tnm, 0, zfsvfs->z_norm, U8_UNICODE_LATEST,
3418 * case preserving rename request, require exact
3427 * If the source and destination directories are the same, we should
3428 * grab the z_name_lock of that directory only once.
3432 rw_enter(&sdzp->z_name_lock, RW_READER);
3436 serr = zfs_dirent_lock(&sdl, sdzp, snm, &szp,
3437 ZEXISTS | zflg, NULL, NULL);
3438 terr = zfs_dirent_lock(&tdl,
3439 tdzp, tnm, &tzp, ZRENAMING | zflg, NULL, NULL);
3441 terr = zfs_dirent_lock(&tdl,
3442 tdzp, tnm, &tzp, zflg, NULL, NULL);
3443 serr = zfs_dirent_lock(&sdl,
3444 sdzp, snm, &szp, ZEXISTS | ZRENAMING | zflg,
3450 * Source entry invalid or not there.
3453 zfs_dirent_unlock(tdl);
3459 rw_exit(&sdzp->z_name_lock);
3461 if (strcmp(snm, ".") == 0 || strcmp(snm, "..") == 0)
3467 zfs_dirent_unlock(sdl);
3471 rw_exit(&sdzp->z_name_lock);
3473 if (strcmp(tnm, "..") == 0)
3480 * Must have write access at the source to remove the old entry
3481 * and write access at the target to create the new entry.
3482 * Note that if target and source are the same, this can be
3483 * done in a single check.
3486 if (error = zfs_zaccess_rename(sdzp, szp, tdzp, tzp, cr))
3489 if (ZTOV(szp)->v_type == VDIR) {
3491 * Check to make sure rename is valid.
3492 * Can't do a move like this: /usr/a/b to /usr/a/b/c/d
3494 if (error = zfs_rename_lock(szp, tdzp, sdzp, &zl))
3499 * Does target exist?
3503 * Source and target must be the same type.
3505 if (ZTOV(szp)->v_type == VDIR) {
3506 if (ZTOV(tzp)->v_type != VDIR) {
3511 if (ZTOV(tzp)->v_type == VDIR) {
3517 * POSIX dictates that when the source and target
3518 * entries refer to the same file object, rename
3519 * must do nothing and exit without error.
3521 if (szp->z_id == tzp->z_id) {
3527 vnevent_rename_src(ZTOV(szp), sdvp, snm, ct);
3529 vnevent_rename_dest(ZTOV(tzp), tdvp, tnm, ct);
3532 * notify the target directory if it is not the same
3533 * as source directory.
3536 vnevent_rename_dest_dir(tdvp, ct);
3539 tx = dmu_tx_create(zfsvfs->z_os);
3540 dmu_tx_hold_bonus(tx, szp->z_id); /* nlink changes */
3541 dmu_tx_hold_bonus(tx, sdzp->z_id); /* nlink changes */
3542 dmu_tx_hold_zap(tx, sdzp->z_id, FALSE, snm);
3543 dmu_tx_hold_zap(tx, tdzp->z_id, TRUE, tnm);
3545 dmu_tx_hold_bonus(tx, tdzp->z_id); /* nlink changes */
3547 dmu_tx_hold_bonus(tx, tzp->z_id); /* parent changes */
3548 dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL);
3549 error = dmu_tx_assign(tx, TXG_NOWAIT);
3552 zfs_rename_unlock(&zl);
3553 zfs_dirent_unlock(sdl);
3554 zfs_dirent_unlock(tdl);
3557 rw_exit(&sdzp->z_name_lock);
3562 if (error == ERESTART) {
3572 if (tzp) /* Attempt to remove the existing target */
3573 error = zfs_link_destroy(tdl, tzp, tx, zflg, NULL);
3576 error = zfs_link_create(tdl, szp, tx, ZRENAMING);
3578 szp->z_phys->zp_flags |= ZFS_AV_MODIFIED;
3580 error = zfs_link_destroy(sdl, szp, tx, ZRENAMING, NULL);
3583 zfs_log_rename(zilog, tx,
3584 TX_RENAME | (flags & FIGNORECASE ? TX_CI : 0),
3585 sdzp, sdl->dl_name, tdzp, tdl->dl_name, szp);
3587 /* Update path information for the target vnode */
3588 vn_renamepath(tdvp, ZTOV(szp), tnm, strlen(tnm));
3590 #ifdef FREEBSD_NAMECACHE
3601 zfs_rename_unlock(&zl);
3603 zfs_dirent_unlock(sdl);
3604 zfs_dirent_unlock(tdl);
3607 rw_exit(&sdzp->z_name_lock);
3619 * Insert the indicated symbolic reference entry into the directory.
3621 * IN: dvp - Directory to contain new symbolic link.
3622 * link - Name for new symlink entry.
3623 * vap - Attributes of new entry.
3624 * target - Target path of new symlink.
3625 * cr - credentials of caller.
3626 * ct - caller context
3627 * flags - case flags
3629 * RETURN: 0 if success
3630 * error code if failure
3633 * dvp - ctime|mtime updated
3637 zfs_symlink(vnode_t *dvp, vnode_t **vpp, char *name, vattr_t *vap, char *link,
3638 cred_t *cr, kthread_t *td)
3640 znode_t *zp, *dzp = VTOZ(dvp);
3643 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
3645 int len = strlen(link);
3648 zfs_acl_ids_t acl_ids;
3649 boolean_t fuid_dirtied;
3652 ASSERT(vap->va_type == VLNK);
3656 zilog = zfsvfs->z_log;
3658 if (zfsvfs->z_utf8 && u8_validate(name, strlen(name),
3659 NULL, U8_VALIDATE_ENTIRE, &error) < 0) {
3663 if (flags & FIGNORECASE)
3666 if (error = zfs_zaccess(dzp, ACE_ADD_FILE, 0, B_FALSE, cr)) {
3671 if (len > MAXPATHLEN) {
3673 return (ENAMETOOLONG);
3677 * Attempt to lock directory; fail if entry already exists.
3679 error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg, NULL, NULL);
3685 VERIFY(0 == zfs_acl_ids_create(dzp, 0, vap, cr, NULL, &acl_ids));
3686 if (zfs_acl_ids_overquota(zfsvfs, &acl_ids)) {
3687 zfs_acl_ids_free(&acl_ids);
3688 zfs_dirent_unlock(dl);
3692 tx = dmu_tx_create(zfsvfs->z_os);
3693 fuid_dirtied = zfsvfs->z_fuid_dirty;
3694 dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, MAX(1, len));
3695 dmu_tx_hold_bonus(tx, dzp->z_id);
3696 dmu_tx_hold_zap(tx, dzp->z_id, TRUE, name);
3697 if (acl_ids.z_aclp->z_acl_bytes > ZFS_ACE_SPACE)
3698 dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, SPA_MAXBLOCKSIZE);
3700 zfs_fuid_txhold(zfsvfs, tx);
3701 error = dmu_tx_assign(tx, TXG_NOWAIT);
3703 zfs_acl_ids_free(&acl_ids);
3704 zfs_dirent_unlock(dl);
3705 if (error == ERESTART) {
3715 dmu_buf_will_dirty(dzp->z_dbuf, tx);
3718 * Create a new object for the symlink.
3719 * Put the link content into bonus buffer if it will fit;
3720 * otherwise, store it just like any other file data.
3722 if (sizeof (znode_phys_t) + len <= dmu_bonus_max()) {
3723 zfs_mknode(dzp, vap, tx, cr, 0, &zp, len, &acl_ids);
3725 bcopy(link, zp->z_phys + 1, len);
3729 zfs_mknode(dzp, vap, tx, cr, 0, &zp, 0, &acl_ids);
3732 zfs_fuid_sync(zfsvfs, tx);
3734 * Nothing can access the znode yet so no locking needed
3735 * for growing the znode's blocksize.
3737 zfs_grow_blocksize(zp, len, tx);
3739 VERIFY(0 == dmu_buf_hold(zfsvfs->z_os,
3740 zp->z_id, 0, FTAG, &dbp));
3741 dmu_buf_will_dirty(dbp, tx);
3743 ASSERT3U(len, <=, dbp->db_size);
3744 bcopy(link, dbp->db_data, len);
3745 dmu_buf_rele(dbp, FTAG);
3747 zp->z_phys->zp_size = len;
3750 * Insert the new object into the directory.
3752 (void) zfs_link_create(dl, zp, tx, ZNEW);
3754 uint64_t txtype = TX_SYMLINK;
3755 if (flags & FIGNORECASE)
3757 zfs_log_symlink(zilog, tx, txtype, dzp, zp, name, link);
3761 zfs_acl_ids_free(&acl_ids);
3765 zfs_dirent_unlock(dl);
3772 * Return, in the buffer contained in the provided uio structure,
3773 * the symbolic path referred to by vp.
3775 * IN: vp - vnode of symbolic link.
3776 * uoip - structure to contain the link path.
3777 * cr - credentials of caller.
3778 * ct - caller context
3780 * OUT: uio - structure to contain the link path.
3782 * RETURN: 0 if success
3783 * error code if failure
3786 * vp - atime updated
3790 zfs_readlink(vnode_t *vp, uio_t *uio, cred_t *cr, caller_context_t *ct)
3792 znode_t *zp = VTOZ(vp);
3793 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
3800 bufsz = (size_t)zp->z_phys->zp_size;
3801 if (bufsz + sizeof (znode_phys_t) <= zp->z_dbuf->db_size) {
3802 error = uiomove(zp->z_phys + 1,
3803 MIN((size_t)bufsz, uio->uio_resid), UIO_READ, uio);
3806 error = dmu_buf_hold(zfsvfs->z_os, zp->z_id, 0, FTAG, &dbp);
3811 error = uiomove(dbp->db_data,
3812 MIN((size_t)bufsz, uio->uio_resid), UIO_READ, uio);
3813 dmu_buf_rele(dbp, FTAG);
3816 ZFS_ACCESSTIME_STAMP(zfsvfs, zp);
3822 * Insert a new entry into directory tdvp referencing svp.
3824 * IN: tdvp - Directory to contain new entry.
3825 * svp - vnode of new entry.
3826 * name - name of new entry.
3827 * cr - credentials of caller.
3828 * ct - caller context
3830 * RETURN: 0 if success
3831 * error code if failure
3834 * tdvp - ctime|mtime updated
3835 * svp - ctime updated
3839 zfs_link(vnode_t *tdvp, vnode_t *svp, char *name, cred_t *cr,
3840 caller_context_t *ct, int flags)
3842 znode_t *dzp = VTOZ(tdvp);
3844 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
3854 ASSERT(tdvp->v_type == VDIR);
3858 zilog = zfsvfs->z_log;
3860 if (VOP_REALVP(svp, &realvp, ct) == 0)
3864 * POSIX dictates that we return EPERM here.
3865 * Better choices include ENOTSUP or EISDIR.
3867 if (svp->v_type == VDIR) {
3872 if (svp->v_vfsp != tdvp->v_vfsp || zfsctl_is_node(svp)) {
3880 /* Prevent links to .zfs/shares files */
3882 if (szp->z_phys->zp_parent == zfsvfs->z_shares_dir) {
3887 if (zfsvfs->z_utf8 && u8_validate(name,
3888 strlen(name), NULL, U8_VALIDATE_ENTIRE, &error) < 0) {
3892 if (flags & FIGNORECASE)
3896 * We do not support links between attributes and non-attributes
3897 * because of the potential security risk of creating links
3898 * into "normal" file space in order to circumvent restrictions
3899 * imposed in attribute space.
3901 if ((szp->z_phys->zp_flags & ZFS_XATTR) !=
3902 (dzp->z_phys->zp_flags & ZFS_XATTR)) {
3908 owner = zfs_fuid_map_id(zfsvfs, szp->z_phys->zp_uid, cr, ZFS_OWNER);
3909 if (owner != crgetuid(cr) &&
3910 secpolicy_basic_link(svp, cr) != 0) {
3915 if (error = zfs_zaccess(dzp, ACE_ADD_FILE, 0, B_FALSE, cr)) {
3922 * Attempt to lock directory; fail if entry already exists.
3924 error = zfs_dirent_lock(&dl, dzp, name, &tzp, zf, NULL, NULL);
3930 tx = dmu_tx_create(zfsvfs->z_os);
3931 dmu_tx_hold_bonus(tx, szp->z_id);
3932 dmu_tx_hold_zap(tx, dzp->z_id, TRUE, name);
3933 error = dmu_tx_assign(tx, TXG_NOWAIT);
3935 zfs_dirent_unlock(dl);
3936 if (error == ERESTART) {
3946 error = zfs_link_create(dl, szp, tx, 0);
3949 uint64_t txtype = TX_LINK;
3950 if (flags & FIGNORECASE)
3952 zfs_log_link(zilog, tx, txtype, dzp, szp, name);
3957 zfs_dirent_unlock(dl);
3960 vnevent_link(svp, ct);
3969 zfs_inactive(vnode_t *vp, cred_t *cr, caller_context_t *ct)
3971 znode_t *zp = VTOZ(vp);
3972 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
3975 rw_enter(&zfsvfs->z_teardown_inactive_lock, RW_READER);
3976 if (zp->z_dbuf == NULL) {
3978 * The fs has been unmounted, or we did a
3979 * suspend/resume and this file no longer exists.
3982 vp->v_count = 0; /* count arrives as 1 */
3984 vrecycle(vp, curthread);
3985 rw_exit(&zfsvfs->z_teardown_inactive_lock);
3989 if (zp->z_atime_dirty && zp->z_unlinked == 0) {
3990 dmu_tx_t *tx = dmu_tx_create(zfsvfs->z_os);
3992 dmu_tx_hold_bonus(tx, zp->z_id);
3993 error = dmu_tx_assign(tx, TXG_WAIT);
3997 dmu_buf_will_dirty(zp->z_dbuf, tx);
3998 mutex_enter(&zp->z_lock);
3999 zp->z_atime_dirty = 0;
4000 mutex_exit(&zp->z_lock);
4006 rw_exit(&zfsvfs->z_teardown_inactive_lock);
4009 CTASSERT(sizeof(struct zfid_short) <= sizeof(struct fid));
4010 CTASSERT(sizeof(struct zfid_long) <= sizeof(struct fid));
4014 zfs_fid(vnode_t *vp, fid_t *fidp, caller_context_t *ct)
4016 znode_t *zp = VTOZ(vp);
4017 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
4019 uint64_t object = zp->z_id;
4025 gen = (uint32_t)zp->z_gen;
4027 size = (zfsvfs->z_parent != zfsvfs) ? LONG_FID_LEN : SHORT_FID_LEN;
4028 fidp->fid_len = size;
4030 zfid = (zfid_short_t *)fidp;
4032 zfid->zf_len = size;
4034 for (i = 0; i < sizeof (zfid->zf_object); i++)
4035 zfid->zf_object[i] = (uint8_t)(object >> (8 * i));
4037 /* Must have a non-zero generation number to distinguish from .zfs */
4040 for (i = 0; i < sizeof (zfid->zf_gen); i++)
4041 zfid->zf_gen[i] = (uint8_t)(gen >> (8 * i));
4043 if (size == LONG_FID_LEN) {
4044 uint64_t objsetid = dmu_objset_id(zfsvfs->z_os);
4047 zlfid = (zfid_long_t *)fidp;
4049 for (i = 0; i < sizeof (zlfid->zf_setid); i++)
4050 zlfid->zf_setid[i] = (uint8_t)(objsetid >> (8 * i));
4052 /* XXX - this should be the generation number for the objset */
4053 for (i = 0; i < sizeof (zlfid->zf_setgen); i++)
4054 zlfid->zf_setgen[i] = 0;
4062 zfs_pathconf(vnode_t *vp, int cmd, ulong_t *valp, cred_t *cr,
4063 caller_context_t *ct)
4075 case _PC_FILESIZEBITS:
4080 case _PC_XATTR_EXISTS:
4082 zfsvfs = zp->z_zfsvfs;
4086 error = zfs_dirent_lock(&dl, zp, "", &xzp,
4087 ZXATTR | ZEXISTS | ZSHARED, NULL, NULL);
4089 zfs_dirent_unlock(dl);
4090 if (!zfs_dirempty(xzp))
4093 } else if (error == ENOENT) {
4095 * If there aren't extended attributes, it's the
4096 * same as having zero of them.
4104 case _PC_ACL_EXTENDED:
4112 case _PC_ACL_PATH_MAX:
4113 *valp = ACL_MAX_ENTRIES;
4116 case _PC_MIN_HOLE_SIZE:
4117 *valp = (int)SPA_MINBLOCKSIZE;
4121 return (EOPNOTSUPP);
4127 zfs_getsecattr(vnode_t *vp, vsecattr_t *vsecp, int flag, cred_t *cr,
4128 caller_context_t *ct)
4130 znode_t *zp = VTOZ(vp);
4131 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
4133 boolean_t skipaclchk = (flag & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE;
4137 error = zfs_getacl(zp, vsecp, skipaclchk, cr);
4145 zfs_setsecattr(vnode_t *vp, vsecattr_t *vsecp, int flag, cred_t *cr,
4146 caller_context_t *ct)
4148 znode_t *zp = VTOZ(vp);
4149 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
4151 boolean_t skipaclchk = (flag & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE;
4155 error = zfs_setacl(zp, vsecp, skipaclchk, cr);
4161 zfs_getpages(struct vnode *vp, vm_page_t *m, int count, int reqpage)
4163 znode_t *zp = VTOZ(vp);
4164 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
4165 objset_t *os = zp->z_zfsvfs->z_os;
4176 pcount = round_page(count) / PAGE_SIZE;
4178 object = mreq->object;
4181 KASSERT(vp->v_object == object, ("mismatching object"));
4183 VM_OBJECT_LOCK(object);
4185 for (i = 0; i < pcount; i++) {
4189 vm_page_unlock(m[i]);
4194 if (mreq->valid != VM_PAGE_BITS_ALL)
4195 vm_page_zero_invalid(mreq, TRUE);
4196 VM_OBJECT_UNLOCK(object);
4198 return (VM_PAGER_OK);
4201 PCPU_INC(cnt.v_vnodein);
4202 PCPU_INC(cnt.v_vnodepgsin);
4204 if (IDX_TO_OFF(mreq->pindex) >= object->un_pager.vnp.vnp_size) {
4205 VM_OBJECT_UNLOCK(object);
4207 return (VM_PAGER_BAD);
4211 if (IDX_TO_OFF(mreq->pindex) + size > object->un_pager.vnp.vnp_size)
4212 size = object->un_pager.vnp.vnp_size - IDX_TO_OFF(mreq->pindex);
4214 VM_OBJECT_UNLOCK(object);
4216 va = zfs_map_page(mreq, &sf);
4217 error = dmu_read(os, zp->z_id, IDX_TO_OFF(mreq->pindex),
4218 size, va, DMU_READ_PREFETCH);
4219 if (size != PAGE_SIZE)
4220 bzero(va + size, PAGE_SIZE - size);
4223 VM_OBJECT_LOCK(object);
4226 mreq->valid = VM_PAGE_BITS_ALL;
4227 KASSERT(mreq->dirty == 0, ("zfs_getpages: page %p is dirty", mreq));
4229 VM_OBJECT_UNLOCK(object);
4231 ZFS_ACCESSTIME_STAMP(zfsvfs, zp);
4233 return (error ? VM_PAGER_ERROR : VM_PAGER_OK);
4237 zfs_freebsd_getpages(ap)
4238 struct vop_getpages_args /* {
4243 vm_ooffset_t a_offset;
4247 return (zfs_getpages(ap->a_vp, ap->a_m, ap->a_count, ap->a_reqpage));
4251 zfs_freebsd_open(ap)
4252 struct vop_open_args /* {
4255 struct ucred *a_cred;
4256 struct thread *a_td;
4259 vnode_t *vp = ap->a_vp;
4260 znode_t *zp = VTOZ(vp);
4263 error = zfs_open(&vp, ap->a_mode, ap->a_cred, NULL);
4265 vnode_create_vobject(vp, zp->z_phys->zp_size, ap->a_td);
4270 zfs_freebsd_close(ap)
4271 struct vop_close_args /* {
4274 struct ucred *a_cred;
4275 struct thread *a_td;
4279 return (zfs_close(ap->a_vp, ap->a_fflag, 0, 0, ap->a_cred, NULL));
4283 zfs_freebsd_ioctl(ap)
4284 struct vop_ioctl_args /* {
4294 return (zfs_ioctl(ap->a_vp, ap->a_command, (intptr_t)ap->a_data,
4295 ap->a_fflag, ap->a_cred, NULL, NULL));
4299 zfs_freebsd_read(ap)
4300 struct vop_read_args /* {
4304 struct ucred *a_cred;
4308 return (zfs_read(ap->a_vp, ap->a_uio, ap->a_ioflag, ap->a_cred, NULL));
4312 zfs_freebsd_write(ap)
4313 struct vop_write_args /* {
4317 struct ucred *a_cred;
4321 return (zfs_write(ap->a_vp, ap->a_uio, ap->a_ioflag, ap->a_cred, NULL));
4325 zfs_freebsd_access(ap)
4326 struct vop_access_args /* {
4328 accmode_t a_accmode;
4329 struct ucred *a_cred;
4330 struct thread *a_td;
4337 * ZFS itself only knowns about VREAD, VWRITE, VEXEC and VAPPEND,
4339 accmode = ap->a_accmode & (VREAD|VWRITE|VEXEC|VAPPEND);
4341 error = zfs_access(ap->a_vp, accmode, 0, ap->a_cred, NULL);
4344 * VADMIN has to be handled by vaccess().
4347 accmode = ap->a_accmode & ~(VREAD|VWRITE|VEXEC|VAPPEND);
4349 vnode_t *vp = ap->a_vp;
4350 znode_t *zp = VTOZ(vp);
4351 znode_phys_t *zphys = zp->z_phys;
4353 error = vaccess(vp->v_type, zphys->zp_mode,
4354 zphys->zp_uid, zphys->zp_gid, accmode, ap->a_cred,
4363 zfs_freebsd_lookup(ap)
4364 struct vop_lookup_args /* {
4365 struct vnode *a_dvp;
4366 struct vnode **a_vpp;
4367 struct componentname *a_cnp;
4370 struct componentname *cnp = ap->a_cnp;
4371 char nm[NAME_MAX + 1];
4373 ASSERT(cnp->cn_namelen < sizeof(nm));
4374 strlcpy(nm, cnp->cn_nameptr, MIN(cnp->cn_namelen + 1, sizeof(nm)));
4376 return (zfs_lookup(ap->a_dvp, nm, ap->a_vpp, cnp, cnp->cn_nameiop,
4377 cnp->cn_cred, cnp->cn_thread, 0));
4381 zfs_freebsd_create(ap)
4382 struct vop_create_args /* {
4383 struct vnode *a_dvp;
4384 struct vnode **a_vpp;
4385 struct componentname *a_cnp;
4386 struct vattr *a_vap;
4389 struct componentname *cnp = ap->a_cnp;
4390 vattr_t *vap = ap->a_vap;
4393 ASSERT(cnp->cn_flags & SAVENAME);
4395 vattr_init_mask(vap);
4396 mode = vap->va_mode & ALLPERMS;
4398 return (zfs_create(ap->a_dvp, cnp->cn_nameptr, vap, !EXCL, mode,
4399 ap->a_vpp, cnp->cn_cred, cnp->cn_thread));
4403 zfs_freebsd_remove(ap)
4404 struct vop_remove_args /* {
4405 struct vnode *a_dvp;
4407 struct componentname *a_cnp;
4411 ASSERT(ap->a_cnp->cn_flags & SAVENAME);
4413 return (zfs_remove(ap->a_dvp, ap->a_cnp->cn_nameptr,
4414 ap->a_cnp->cn_cred, NULL, 0));
4418 zfs_freebsd_mkdir(ap)
4419 struct vop_mkdir_args /* {
4420 struct vnode *a_dvp;
4421 struct vnode **a_vpp;
4422 struct componentname *a_cnp;
4423 struct vattr *a_vap;
4426 vattr_t *vap = ap->a_vap;
4428 ASSERT(ap->a_cnp->cn_flags & SAVENAME);
4430 vattr_init_mask(vap);
4432 return (zfs_mkdir(ap->a_dvp, ap->a_cnp->cn_nameptr, vap, ap->a_vpp,
4433 ap->a_cnp->cn_cred, NULL, 0, NULL));
4437 zfs_freebsd_rmdir(ap)
4438 struct vop_rmdir_args /* {
4439 struct vnode *a_dvp;
4441 struct componentname *a_cnp;
4444 struct componentname *cnp = ap->a_cnp;
4446 ASSERT(cnp->cn_flags & SAVENAME);
4448 return (zfs_rmdir(ap->a_dvp, cnp->cn_nameptr, NULL, cnp->cn_cred, NULL, 0));
4452 zfs_freebsd_readdir(ap)
4453 struct vop_readdir_args /* {
4456 struct ucred *a_cred;
4463 return (zfs_readdir(ap->a_vp, ap->a_uio, ap->a_cred, ap->a_eofflag,
4464 ap->a_ncookies, ap->a_cookies));
4468 zfs_freebsd_fsync(ap)
4469 struct vop_fsync_args /* {
4472 struct thread *a_td;
4477 return (zfs_fsync(ap->a_vp, 0, ap->a_td->td_ucred, NULL));
4481 zfs_freebsd_getattr(ap)
4482 struct vop_getattr_args /* {
4484 struct vattr *a_vap;
4485 struct ucred *a_cred;
4486 struct thread *a_td;
4489 vattr_t *vap = ap->a_vap;
4495 xvap.xva_vattr = *vap;
4496 xvap.xva_vattr.va_mask |= AT_XVATTR;
4498 /* Convert chflags into ZFS-type flags. */
4499 /* XXX: what about SF_SETTABLE?. */
4500 XVA_SET_REQ(&xvap, XAT_IMMUTABLE);
4501 XVA_SET_REQ(&xvap, XAT_APPENDONLY);
4502 XVA_SET_REQ(&xvap, XAT_NOUNLINK);
4503 XVA_SET_REQ(&xvap, XAT_NODUMP);
4504 error = zfs_getattr(ap->a_vp, (vattr_t *)&xvap, 0, ap->a_cred, NULL);
4508 /* Convert ZFS xattr into chflags. */
4509 #define FLAG_CHECK(fflag, xflag, xfield) do { \
4510 if (XVA_ISSET_RTN(&xvap, (xflag)) && (xfield) != 0) \
4511 fflags |= (fflag); \
4513 FLAG_CHECK(SF_IMMUTABLE, XAT_IMMUTABLE,
4514 xvap.xva_xoptattrs.xoa_immutable);
4515 FLAG_CHECK(SF_APPEND, XAT_APPENDONLY,
4516 xvap.xva_xoptattrs.xoa_appendonly);
4517 FLAG_CHECK(SF_NOUNLINK, XAT_NOUNLINK,
4518 xvap.xva_xoptattrs.xoa_nounlink);
4519 FLAG_CHECK(UF_NODUMP, XAT_NODUMP,
4520 xvap.xva_xoptattrs.xoa_nodump);
4522 *vap = xvap.xva_vattr;
4523 vap->va_flags = fflags;
4528 zfs_freebsd_setattr(ap)
4529 struct vop_setattr_args /* {
4531 struct vattr *a_vap;
4532 struct ucred *a_cred;
4533 struct thread *a_td;
4536 vnode_t *vp = ap->a_vp;
4537 vattr_t *vap = ap->a_vap;
4538 cred_t *cred = ap->a_cred;
4543 vattr_init_mask(vap);
4544 vap->va_mask &= ~AT_NOSET;
4547 xvap.xva_vattr = *vap;
4549 zflags = VTOZ(vp)->z_phys->zp_flags;
4551 if (vap->va_flags != VNOVAL) {
4552 zfsvfs_t *zfsvfs = VTOZ(vp)->z_zfsvfs;
4555 if (zfsvfs->z_use_fuids == B_FALSE)
4556 return (EOPNOTSUPP);
4558 fflags = vap->va_flags;
4559 if ((fflags & ~(SF_IMMUTABLE|SF_APPEND|SF_NOUNLINK|UF_NODUMP)) != 0)
4560 return (EOPNOTSUPP);
4562 * Unprivileged processes are not permitted to unset system
4563 * flags, or modify flags if any system flags are set.
4564 * Privileged non-jail processes may not modify system flags
4565 * if securelevel > 0 and any existing system flags are set.
4566 * Privileged jail processes behave like privileged non-jail
4567 * processes if the security.jail.chflags_allowed sysctl is
4568 * is non-zero; otherwise, they behave like unprivileged
4571 if (secpolicy_fs_owner(vp->v_mount, cred) == 0 ||
4572 priv_check_cred(cred, PRIV_VFS_SYSFLAGS, 0) == 0) {
4574 (ZFS_IMMUTABLE | ZFS_APPENDONLY | ZFS_NOUNLINK)) {
4575 error = securelevel_gt(cred, 0);
4581 * Callers may only modify the file flags on objects they
4582 * have VADMIN rights for.
4584 if ((error = VOP_ACCESS(vp, VADMIN, cred, curthread)) != 0)
4587 (ZFS_IMMUTABLE | ZFS_APPENDONLY | ZFS_NOUNLINK)) {
4591 (SF_IMMUTABLE | SF_APPEND | SF_NOUNLINK)) {
4596 #define FLAG_CHANGE(fflag, zflag, xflag, xfield) do { \
4597 if (((fflags & (fflag)) && !(zflags & (zflag))) || \
4598 ((zflags & (zflag)) && !(fflags & (fflag)))) { \
4599 XVA_SET_REQ(&xvap, (xflag)); \
4600 (xfield) = ((fflags & (fflag)) != 0); \
4603 /* Convert chflags into ZFS-type flags. */
4604 /* XXX: what about SF_SETTABLE?. */
4605 FLAG_CHANGE(SF_IMMUTABLE, ZFS_IMMUTABLE, XAT_IMMUTABLE,
4606 xvap.xva_xoptattrs.xoa_immutable);
4607 FLAG_CHANGE(SF_APPEND, ZFS_APPENDONLY, XAT_APPENDONLY,
4608 xvap.xva_xoptattrs.xoa_appendonly);
4609 FLAG_CHANGE(SF_NOUNLINK, ZFS_NOUNLINK, XAT_NOUNLINK,
4610 xvap.xva_xoptattrs.xoa_nounlink);
4611 FLAG_CHANGE(UF_NODUMP, ZFS_NODUMP, XAT_NODUMP,
4612 xvap.xva_xoptattrs.xoa_nodump);
4615 return (zfs_setattr(vp, (vattr_t *)&xvap, 0, cred, NULL));
4619 zfs_freebsd_rename(ap)
4620 struct vop_rename_args /* {
4621 struct vnode *a_fdvp;
4622 struct vnode *a_fvp;
4623 struct componentname *a_fcnp;
4624 struct vnode *a_tdvp;
4625 struct vnode *a_tvp;
4626 struct componentname *a_tcnp;
4629 vnode_t *fdvp = ap->a_fdvp;
4630 vnode_t *fvp = ap->a_fvp;
4631 vnode_t *tdvp = ap->a_tdvp;
4632 vnode_t *tvp = ap->a_tvp;
4635 ASSERT(ap->a_fcnp->cn_flags & (SAVENAME|SAVESTART));
4636 ASSERT(ap->a_tcnp->cn_flags & (SAVENAME|SAVESTART));
4638 error = zfs_rename(fdvp, ap->a_fcnp->cn_nameptr, tdvp,
4639 ap->a_tcnp->cn_nameptr, ap->a_fcnp->cn_cred, NULL, 0);
4654 zfs_freebsd_symlink(ap)
4655 struct vop_symlink_args /* {
4656 struct vnode *a_dvp;
4657 struct vnode **a_vpp;
4658 struct componentname *a_cnp;
4659 struct vattr *a_vap;
4663 struct componentname *cnp = ap->a_cnp;
4664 vattr_t *vap = ap->a_vap;
4666 ASSERT(cnp->cn_flags & SAVENAME);
4668 vap->va_type = VLNK; /* FreeBSD: Syscall only sets va_mode. */
4669 vattr_init_mask(vap);
4671 return (zfs_symlink(ap->a_dvp, ap->a_vpp, cnp->cn_nameptr, vap,
4672 ap->a_target, cnp->cn_cred, cnp->cn_thread));
4676 zfs_freebsd_readlink(ap)
4677 struct vop_readlink_args /* {
4680 struct ucred *a_cred;
4684 return (zfs_readlink(ap->a_vp, ap->a_uio, ap->a_cred, NULL));
4688 zfs_freebsd_link(ap)
4689 struct vop_link_args /* {
4690 struct vnode *a_tdvp;
4692 struct componentname *a_cnp;
4695 struct componentname *cnp = ap->a_cnp;
4697 ASSERT(cnp->cn_flags & SAVENAME);
4699 return (zfs_link(ap->a_tdvp, ap->a_vp, cnp->cn_nameptr, cnp->cn_cred, NULL, 0));
4703 zfs_freebsd_inactive(ap)
4704 struct vop_inactive_args /* {
4706 struct thread *a_td;
4709 vnode_t *vp = ap->a_vp;
4711 zfs_inactive(vp, ap->a_td->td_ucred, NULL);
4716 zfs_reclaim_complete(void *arg, int pending)
4719 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
4721 rw_enter(&zfsvfs->z_teardown_inactive_lock, RW_READER);
4722 if (zp->z_dbuf != NULL) {
4723 ZFS_OBJ_HOLD_ENTER(zfsvfs, zp->z_id);
4724 zfs_znode_dmu_fini(zp);
4725 ZFS_OBJ_HOLD_EXIT(zfsvfs, zp->z_id);
4728 rw_exit(&zfsvfs->z_teardown_inactive_lock);
4730 * If the file system is being unmounted, there is a process waiting
4731 * for us, wake it up.
4733 if (zfsvfs->z_unmounted)
4738 zfs_freebsd_reclaim(ap)
4739 struct vop_reclaim_args /* {
4741 struct thread *a_td;
4744 vnode_t *vp = ap->a_vp;
4745 znode_t *zp = VTOZ(vp);
4746 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
4748 rw_enter(&zfsvfs->z_teardown_inactive_lock, RW_READER);
4753 * Destroy the vm object and flush associated pages.
4755 vnode_destroy_vobject(vp);
4757 mutex_enter(&zp->z_lock);
4758 ASSERT(zp->z_phys != NULL);
4760 mutex_exit(&zp->z_lock);
4764 else if (zp->z_dbuf == NULL)
4766 else /* if (!zp->z_unlinked && zp->z_dbuf != NULL) */ {
4769 locked = MUTEX_HELD(ZFS_OBJ_MUTEX(zfsvfs, zp->z_id)) ? 2 :
4770 ZFS_OBJ_HOLD_TRYENTER(zfsvfs, zp->z_id);
4773 * Lock can't be obtained due to deadlock possibility,
4774 * so defer znode destruction.
4776 TASK_INIT(&zp->z_task, 0, zfs_reclaim_complete, zp);
4777 taskqueue_enqueue(taskqueue_thread, &zp->z_task);
4779 zfs_znode_dmu_fini(zp);
4781 ZFS_OBJ_HOLD_EXIT(zfsvfs, zp->z_id);
4787 ASSERT(vp->v_holdcnt >= 1);
4789 rw_exit(&zfsvfs->z_teardown_inactive_lock);
4795 struct vop_fid_args /* {
4801 return (zfs_fid(ap->a_vp, (void *)ap->a_fid, NULL));
4805 zfs_freebsd_pathconf(ap)
4806 struct vop_pathconf_args /* {
4809 register_t *a_retval;
4815 error = zfs_pathconf(ap->a_vp, ap->a_name, &val, curthread->td_ucred, NULL);
4817 *ap->a_retval = val;
4818 else if (error == EOPNOTSUPP)
4819 error = vop_stdpathconf(ap);
4824 zfs_freebsd_fifo_pathconf(ap)
4825 struct vop_pathconf_args /* {
4828 register_t *a_retval;
4832 switch (ap->a_name) {
4833 case _PC_ACL_EXTENDED:
4835 case _PC_ACL_PATH_MAX:
4836 case _PC_MAC_PRESENT:
4837 return (zfs_freebsd_pathconf(ap));
4839 return (fifo_specops.vop_pathconf(ap));
4844 * FreeBSD's extended attributes namespace defines file name prefix for ZFS'
4845 * extended attribute name:
4848 * system freebsd:system:
4849 * user (none, can be used to access ZFS fsattr(5) attributes
4850 * created on Solaris)
4853 zfs_create_attrname(int attrnamespace, const char *name, char *attrname,
4856 const char *namespace, *prefix, *suffix;
4858 /* We don't allow '/' character in attribute name. */
4859 if (strchr(name, '/') != NULL)
4861 /* We don't allow attribute names that start with "freebsd:" string. */
4862 if (strncmp(name, "freebsd:", 8) == 0)
4865 bzero(attrname, size);
4867 switch (attrnamespace) {
4868 case EXTATTR_NAMESPACE_USER:
4870 prefix = "freebsd:";
4871 namespace = EXTATTR_NAMESPACE_USER_STRING;
4875 * This is the default namespace by which we can access all
4876 * attributes created on Solaris.
4878 prefix = namespace = suffix = "";
4881 case EXTATTR_NAMESPACE_SYSTEM:
4882 prefix = "freebsd:";
4883 namespace = EXTATTR_NAMESPACE_SYSTEM_STRING;
4886 case EXTATTR_NAMESPACE_EMPTY:
4890 if (snprintf(attrname, size, "%s%s%s%s", prefix, namespace, suffix,
4892 return (ENAMETOOLONG);
4898 * Vnode operating to retrieve a named extended attribute.
4901 zfs_getextattr(struct vop_getextattr_args *ap)
4904 IN struct vnode *a_vp;
4905 IN int a_attrnamespace;
4906 IN const char *a_name;
4907 INOUT struct uio *a_uio;
4909 IN struct ucred *a_cred;
4910 IN struct thread *a_td;
4914 zfsvfs_t *zfsvfs = VTOZ(ap->a_vp)->z_zfsvfs;
4915 struct thread *td = ap->a_td;
4916 struct nameidata nd;
4919 vnode_t *xvp = NULL, *vp;
4922 error = extattr_check_cred(ap->a_vp, ap->a_attrnamespace,
4923 ap->a_cred, ap->a_td, VREAD);
4927 error = zfs_create_attrname(ap->a_attrnamespace, ap->a_name, attrname,
4934 error = zfs_lookup(ap->a_vp, NULL, &xvp, NULL, 0, ap->a_cred, td,
4942 NDINIT_ATVP(&nd, LOOKUP, NOFOLLOW | MPSAFE, UIO_SYSSPACE, attrname,
4944 error = vn_open_cred(&nd, &flags, 0, 0, ap->a_cred, NULL);
4946 NDFREE(&nd, NDF_ONLY_PNBUF);
4949 if (error == ENOENT)
4954 if (ap->a_size != NULL) {
4955 error = VOP_GETATTR(vp, &va, ap->a_cred);
4957 *ap->a_size = (size_t)va.va_size;
4958 } else if (ap->a_uio != NULL)
4959 error = VOP_READ(vp, ap->a_uio, IO_UNIT | IO_SYNC, ap->a_cred);
4962 vn_close(vp, flags, ap->a_cred, td);
4969 * Vnode operation to remove a named attribute.
4972 zfs_deleteextattr(struct vop_deleteextattr_args *ap)
4975 IN struct vnode *a_vp;
4976 IN int a_attrnamespace;
4977 IN const char *a_name;
4978 IN struct ucred *a_cred;
4979 IN struct thread *a_td;
4983 zfsvfs_t *zfsvfs = VTOZ(ap->a_vp)->z_zfsvfs;
4984 struct thread *td = ap->a_td;
4985 struct nameidata nd;
4988 vnode_t *xvp = NULL, *vp;
4991 error = extattr_check_cred(ap->a_vp, ap->a_attrnamespace,
4992 ap->a_cred, ap->a_td, VWRITE);
4996 error = zfs_create_attrname(ap->a_attrnamespace, ap->a_name, attrname,
5003 error = zfs_lookup(ap->a_vp, NULL, &xvp, NULL, 0, ap->a_cred, td,
5010 NDINIT_ATVP(&nd, DELETE, NOFOLLOW | LOCKPARENT | LOCKLEAF | MPSAFE,
5011 UIO_SYSSPACE, attrname, xvp, td);
5014 NDFREE(&nd, NDF_ONLY_PNBUF);
5017 if (error == ENOENT)
5021 error = VOP_REMOVE(nd.ni_dvp, vp, &nd.ni_cnd);
5024 if (vp == nd.ni_dvp)
5034 * Vnode operation to set a named attribute.
5037 zfs_setextattr(struct vop_setextattr_args *ap)
5040 IN struct vnode *a_vp;
5041 IN int a_attrnamespace;
5042 IN const char *a_name;
5043 INOUT struct uio *a_uio;
5044 IN struct ucred *a_cred;
5045 IN struct thread *a_td;
5049 zfsvfs_t *zfsvfs = VTOZ(ap->a_vp)->z_zfsvfs;
5050 struct thread *td = ap->a_td;
5051 struct nameidata nd;
5054 vnode_t *xvp = NULL, *vp;
5057 error = extattr_check_cred(ap->a_vp, ap->a_attrnamespace,
5058 ap->a_cred, ap->a_td, VWRITE);
5062 error = zfs_create_attrname(ap->a_attrnamespace, ap->a_name, attrname,
5069 error = zfs_lookup(ap->a_vp, NULL, &xvp, NULL, 0, ap->a_cred, td,
5070 LOOKUP_XATTR | CREATE_XATTR_DIR);
5076 flags = FFLAGS(O_WRONLY | O_CREAT);
5077 NDINIT_ATVP(&nd, LOOKUP, NOFOLLOW | MPSAFE, UIO_SYSSPACE, attrname,
5079 error = vn_open_cred(&nd, &flags, 0600, 0, ap->a_cred, NULL);
5081 NDFREE(&nd, NDF_ONLY_PNBUF);
5089 error = VOP_SETATTR(vp, &va, ap->a_cred);
5091 VOP_WRITE(vp, ap->a_uio, IO_UNIT | IO_SYNC, ap->a_cred);
5094 vn_close(vp, flags, ap->a_cred, td);
5101 * Vnode operation to retrieve extended attributes on a vnode.
5104 zfs_listextattr(struct vop_listextattr_args *ap)
5107 IN struct vnode *a_vp;
5108 IN int a_attrnamespace;
5109 INOUT struct uio *a_uio;
5111 IN struct ucred *a_cred;
5112 IN struct thread *a_td;
5116 zfsvfs_t *zfsvfs = VTOZ(ap->a_vp)->z_zfsvfs;
5117 struct thread *td = ap->a_td;
5118 struct nameidata nd;
5119 char attrprefix[16];
5120 u_char dirbuf[sizeof(struct dirent)];
5123 struct uio auio, *uio = ap->a_uio;
5124 size_t *sizep = ap->a_size;
5126 vnode_t *xvp = NULL, *vp;
5127 int done, error, eof, pos;
5129 error = extattr_check_cred(ap->a_vp, ap->a_attrnamespace,
5130 ap->a_cred, ap->a_td, VREAD);
5134 error = zfs_create_attrname(ap->a_attrnamespace, "", attrprefix,
5135 sizeof(attrprefix));
5138 plen = strlen(attrprefix);
5145 error = zfs_lookup(ap->a_vp, NULL, &xvp, NULL, 0, ap->a_cred, td,
5150 * ENOATTR means that the EA directory does not yet exist,
5151 * i.e. there are no extended attributes there.
5153 if (error == ENOATTR)
5158 NDINIT_ATVP(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | LOCKSHARED | MPSAFE,
5159 UIO_SYSSPACE, ".", xvp, td);
5162 NDFREE(&nd, NDF_ONLY_PNBUF);
5168 auio.uio_iov = &aiov;
5169 auio.uio_iovcnt = 1;
5170 auio.uio_segflg = UIO_SYSSPACE;
5172 auio.uio_rw = UIO_READ;
5173 auio.uio_offset = 0;
5178 aiov.iov_base = (void *)dirbuf;
5179 aiov.iov_len = sizeof(dirbuf);
5180 auio.uio_resid = sizeof(dirbuf);
5181 error = VOP_READDIR(vp, &auio, ap->a_cred, &eof, NULL, NULL);
5182 done = sizeof(dirbuf) - auio.uio_resid;
5185 for (pos = 0; pos < done;) {
5186 dp = (struct dirent *)(dirbuf + pos);
5187 pos += dp->d_reclen;
5189 * XXX: Temporarily we also accept DT_UNKNOWN, as this
5190 * is what we get when attribute was created on Solaris.
5192 if (dp->d_type != DT_REG && dp->d_type != DT_UNKNOWN)
5194 if (plen == 0 && strncmp(dp->d_name, "freebsd:", 8) == 0)
5196 else if (strncmp(dp->d_name, attrprefix, plen) != 0)
5198 nlen = dp->d_namlen - plen;
5201 else if (uio != NULL) {
5203 * Format of extattr name entry is one byte for
5204 * length and the rest for name.
5206 error = uiomove(&nlen, 1, uio->uio_rw, uio);
5208 error = uiomove(dp->d_name + plen, nlen,
5215 } while (!eof && error == 0);
5224 zfs_freebsd_getacl(ap)
5225 struct vop_getacl_args /* {
5234 vsecattr_t vsecattr;
5236 if (ap->a_type != ACL_TYPE_NFS4)
5239 vsecattr.vsa_mask = VSA_ACE | VSA_ACECNT;
5240 if (error = zfs_getsecattr(ap->a_vp, &vsecattr, 0, ap->a_cred, NULL))
5243 error = acl_from_aces(ap->a_aclp, vsecattr.vsa_aclentp, vsecattr.vsa_aclcnt);
5244 if (vsecattr.vsa_aclentp != NULL)
5245 kmem_free(vsecattr.vsa_aclentp, vsecattr.vsa_aclentsz);
5251 zfs_freebsd_setacl(ap)
5252 struct vop_setacl_args /* {
5261 vsecattr_t vsecattr;
5262 int aclbsize; /* size of acl list in bytes */
5265 if (ap->a_type != ACL_TYPE_NFS4)
5268 if (ap->a_aclp->acl_cnt < 1 || ap->a_aclp->acl_cnt > MAX_ACL_ENTRIES)
5272 * With NFSv4 ACLs, chmod(2) may need to add additional entries,
5273 * splitting every entry into two and appending "canonical six"
5274 * entries at the end. Don't allow for setting an ACL that would
5275 * cause chmod(2) to run out of ACL entries.
5277 if (ap->a_aclp->acl_cnt * 2 + 6 > ACL_MAX_ENTRIES)
5280 error = acl_nfs4_check(ap->a_aclp, ap->a_vp->v_type == VDIR);
5284 vsecattr.vsa_mask = VSA_ACE;
5285 aclbsize = ap->a_aclp->acl_cnt * sizeof(ace_t);
5286 vsecattr.vsa_aclentp = kmem_alloc(aclbsize, KM_SLEEP);
5287 aaclp = vsecattr.vsa_aclentp;
5288 vsecattr.vsa_aclentsz = aclbsize;
5290 aces_from_acl(vsecattr.vsa_aclentp, &vsecattr.vsa_aclcnt, ap->a_aclp);
5291 error = zfs_setsecattr(ap->a_vp, &vsecattr, 0, ap->a_cred, NULL);
5292 kmem_free(aaclp, aclbsize);
5298 zfs_freebsd_aclcheck(ap)
5299 struct vop_aclcheck_args /* {
5308 return (EOPNOTSUPP);
5311 struct vop_vector zfs_vnodeops;
5312 struct vop_vector zfs_fifoops;
5313 struct vop_vector zfs_shareops;
5315 struct vop_vector zfs_vnodeops = {
5316 .vop_default = &default_vnodeops,
5317 .vop_inactive = zfs_freebsd_inactive,
5318 .vop_reclaim = zfs_freebsd_reclaim,
5319 .vop_access = zfs_freebsd_access,
5320 #ifdef FREEBSD_NAMECACHE
5321 .vop_lookup = vfs_cache_lookup,
5322 .vop_cachedlookup = zfs_freebsd_lookup,
5324 .vop_lookup = zfs_freebsd_lookup,
5326 .vop_getattr = zfs_freebsd_getattr,
5327 .vop_setattr = zfs_freebsd_setattr,
5328 .vop_create = zfs_freebsd_create,
5329 .vop_mknod = zfs_freebsd_create,
5330 .vop_mkdir = zfs_freebsd_mkdir,
5331 .vop_readdir = zfs_freebsd_readdir,
5332 .vop_fsync = zfs_freebsd_fsync,
5333 .vop_open = zfs_freebsd_open,
5334 .vop_close = zfs_freebsd_close,
5335 .vop_rmdir = zfs_freebsd_rmdir,
5336 .vop_ioctl = zfs_freebsd_ioctl,
5337 .vop_link = zfs_freebsd_link,
5338 .vop_symlink = zfs_freebsd_symlink,
5339 .vop_readlink = zfs_freebsd_readlink,
5340 .vop_read = zfs_freebsd_read,
5341 .vop_write = zfs_freebsd_write,
5342 .vop_remove = zfs_freebsd_remove,
5343 .vop_rename = zfs_freebsd_rename,
5344 .vop_pathconf = zfs_freebsd_pathconf,
5345 .vop_bmap = VOP_EOPNOTSUPP,
5346 .vop_fid = zfs_freebsd_fid,
5347 .vop_getextattr = zfs_getextattr,
5348 .vop_deleteextattr = zfs_deleteextattr,
5349 .vop_setextattr = zfs_setextattr,
5350 .vop_listextattr = zfs_listextattr,
5351 .vop_getacl = zfs_freebsd_getacl,
5352 .vop_setacl = zfs_freebsd_setacl,
5353 .vop_aclcheck = zfs_freebsd_aclcheck,
5354 .vop_getpages = zfs_freebsd_getpages,
5357 struct vop_vector zfs_fifoops = {
5358 .vop_default = &fifo_specops,
5359 .vop_fsync = zfs_freebsd_fsync,
5360 .vop_access = zfs_freebsd_access,
5361 .vop_getattr = zfs_freebsd_getattr,
5362 .vop_inactive = zfs_freebsd_inactive,
5363 .vop_read = VOP_PANIC,
5364 .vop_reclaim = zfs_freebsd_reclaim,
5365 .vop_setattr = zfs_freebsd_setattr,
5366 .vop_write = VOP_PANIC,
5367 .vop_pathconf = zfs_freebsd_fifo_pathconf,
5368 .vop_fid = zfs_freebsd_fid,
5369 .vop_getacl = zfs_freebsd_getacl,
5370 .vop_setacl = zfs_freebsd_setacl,
5371 .vop_aclcheck = zfs_freebsd_aclcheck,
5375 * special share hidden files vnode operations template
5377 struct vop_vector zfs_shareops = {
5378 .vop_default = &default_vnodeops,
5379 .vop_access = zfs_freebsd_access,
5380 .vop_inactive = zfs_freebsd_inactive,
5381 .vop_reclaim = zfs_freebsd_reclaim,
5382 .vop_fid = zfs_freebsd_fid,
5383 .vop_pathconf = zfs_freebsd_pathconf,