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 static int zil_fault_io = 0;
1013 * Get data to generate a TX_WRITE intent log record.
1016 zfs_get_data(void *arg, lr_write_t *lr, char *buf, zio_t *zio)
1018 zfsvfs_t *zfsvfs = arg;
1019 objset_t *os = zfsvfs->z_os;
1021 uint64_t off = lr->lr_offset;
1025 int dlen = lr->lr_length; /* length of user data */
1032 * Nothing to do if the file has been removed
1034 if (zfs_zget(zfsvfs, lr->lr_foid, &zp) != 0)
1036 if (zp->z_unlinked) {
1038 * Release the vnode asynchronously as we currently have the
1039 * txg stopped from syncing.
1041 VN_RELE_ASYNC(ZTOV(zp),
1042 dsl_pool_vnrele_taskq(dmu_objset_pool(os)));
1047 * Write records come in two flavors: immediate and indirect.
1048 * For small writes it's cheaper to store the data with the
1049 * log record (immediate); for large writes it's cheaper to
1050 * sync the data and get a pointer to it (indirect) so that
1051 * we don't have to write the data twice.
1053 if (buf != NULL) { /* immediate write */
1054 rl = zfs_range_lock(zp, off, dlen, RL_READER);
1055 /* test for truncation needs to be done while range locked */
1056 if (off >= zp->z_phys->zp_size) {
1060 VERIFY(0 == dmu_read(os, lr->lr_foid, off, dlen, buf,
1061 DMU_READ_NO_PREFETCH));
1062 } else { /* indirect write */
1063 uint64_t boff; /* block starting offset */
1066 * Have to lock the whole block to ensure when it's
1067 * written out and it's checksum is being calculated
1068 * that no one can change the data. We need to re-check
1069 * blocksize after we get the lock in case it's changed!
1072 if (ISP2(zp->z_blksz)) {
1073 boff = P2ALIGN_TYPED(off, zp->z_blksz,
1079 rl = zfs_range_lock(zp, boff, dlen, RL_READER);
1080 if (zp->z_blksz == dlen)
1082 zfs_range_unlock(rl);
1084 /* test for truncation needs to be done while range locked */
1085 if (off >= zp->z_phys->zp_size) {
1089 zgd = (zgd_t *)kmem_alloc(sizeof (zgd_t), KM_SLEEP);
1091 zgd->zgd_zilog = zfsvfs->z_log;
1092 zgd->zgd_bp = &lr->lr_blkptr;
1098 error = dmu_buf_hold(os, lr->lr_foid, boff, zgd, &db);
1101 error = dmu_buf_hold(os, lr->lr_foid, boff, zgd, &db);
1104 kmem_free(zgd, sizeof (zgd_t));
1108 ASSERT(boff == db->db_offset);
1109 lr->lr_blkoff = off - boff;
1110 error = dmu_sync(zio, db, &lr->lr_blkptr,
1111 lr->lr_common.lrc_txg, zfs_get_done, zgd);
1112 ASSERT((error && error != EINPROGRESS) ||
1113 lr->lr_length <= zp->z_blksz);
1116 * dmu_sync() can compress a block of zeros to a null
1117 * blkptr but the block size still needs to be passed
1118 * through to replay.
1120 BP_SET_LSIZE(&lr->lr_blkptr, db->db_size);
1121 zil_add_block(zfsvfs->z_log, &lr->lr_blkptr);
1125 * If we get EINPROGRESS, then we need to wait for a
1126 * write IO initiated by dmu_sync() to complete before
1127 * we can release this dbuf. We will finish everything
1128 * up in the zfs_get_done() callback.
1130 if (error == EINPROGRESS) {
1132 } else if (error == EALREADY) {
1133 lr->lr_common.lrc_txtype = TX_WRITE2;
1136 dmu_buf_rele(db, zgd);
1137 kmem_free(zgd, sizeof (zgd_t));
1140 zfs_range_unlock(rl);
1142 * Release the vnode asynchronously as we currently have the
1143 * txg stopped from syncing.
1145 VN_RELE_ASYNC(ZTOV(zp), dsl_pool_vnrele_taskq(dmu_objset_pool(os)));
1151 zfs_access(vnode_t *vp, int mode, int flag, cred_t *cr,
1152 caller_context_t *ct)
1154 znode_t *zp = VTOZ(vp);
1155 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1161 if (flag & V_ACE_MASK)
1162 error = zfs_zaccess(zp, mode, flag, B_FALSE, cr);
1164 error = zfs_zaccess_rwx(zp, mode, flag, cr);
1171 * If vnode is for a device return a specfs vnode instead.
1174 specvp_check(vnode_t **vpp, cred_t *cr)
1178 if (IS_DEVVP(*vpp)) {
1181 svp = specvp(*vpp, (*vpp)->v_rdev, (*vpp)->v_type, cr);
1192 * Lookup an entry in a directory, or an extended attribute directory.
1193 * If it exists, return a held vnode reference for it.
1195 * IN: dvp - vnode of directory to search.
1196 * nm - name of entry to lookup.
1197 * pnp - full pathname to lookup [UNUSED].
1198 * flags - LOOKUP_XATTR set if looking for an attribute.
1199 * rdir - root directory vnode [UNUSED].
1200 * cr - credentials of caller.
1201 * ct - caller context
1202 * direntflags - directory lookup flags
1203 * realpnp - returned pathname.
1205 * OUT: vpp - vnode of located entry, NULL if not found.
1207 * RETURN: 0 if success
1208 * error code if failure
1215 zfs_lookup(vnode_t *dvp, char *nm, vnode_t **vpp, struct componentname *cnp,
1216 int nameiop, cred_t *cr, kthread_t *td, int flags)
1218 znode_t *zdp = VTOZ(dvp);
1219 zfsvfs_t *zfsvfs = zdp->z_zfsvfs;
1221 int *direntflags = NULL;
1222 void *realpnp = NULL;
1225 if (!(flags & (LOOKUP_XATTR | FIGNORECASE))) {
1227 if (dvp->v_type != VDIR) {
1229 } else if (zdp->z_dbuf == NULL) {
1233 if (nm[0] == 0 || (nm[0] == '.' && nm[1] == '\0')) {
1234 error = zfs_fastaccesschk_execute(zdp, cr);
1242 vnode_t *tvp = dnlc_lookup(dvp, nm);
1245 error = zfs_fastaccesschk_execute(zdp, cr);
1250 if (tvp == DNLC_NO_VNODE) {
1255 return (specvp_check(vpp, cr));
1261 DTRACE_PROBE2(zfs__fastpath__lookup__miss, vnode_t *, dvp, char *, nm);
1268 if (flags & LOOKUP_XATTR) {
1271 * If the xattr property is off, refuse the lookup request.
1273 if (!(zfsvfs->z_vfs->vfs_flag & VFS_XATTR)) {
1280 * We don't allow recursive attributes..
1281 * Maybe someday we will.
1283 if (zdp->z_phys->zp_flags & ZFS_XATTR) {
1288 if (error = zfs_get_xattrdir(VTOZ(dvp), vpp, cr, flags)) {
1294 * Do we have permission to get into attribute directory?
1297 if (error = zfs_zaccess(VTOZ(*vpp), ACE_EXECUTE, 0,
1307 if (dvp->v_type != VDIR) {
1313 * Check accessibility of directory.
1316 if (error = zfs_zaccess(zdp, ACE_EXECUTE, 0, B_FALSE, cr)) {
1321 if (zfsvfs->z_utf8 && u8_validate(nm, strlen(nm),
1322 NULL, U8_VALIDATE_ENTIRE, &error) < 0) {
1327 error = zfs_dirlook(zdp, nm, vpp, flags, direntflags, realpnp);
1329 error = specvp_check(vpp, cr);
1331 /* Translate errors and add SAVENAME when needed. */
1332 if (cnp->cn_flags & ISLASTCN) {
1336 if (error == ENOENT) {
1337 error = EJUSTRETURN;
1338 cnp->cn_flags |= SAVENAME;
1344 cnp->cn_flags |= SAVENAME;
1348 if (error == 0 && (nm[0] != '.' || nm[1] != '\0')) {
1351 if (cnp->cn_flags & ISDOTDOT) {
1352 ltype = VOP_ISLOCKED(dvp);
1356 error = vn_lock(*vpp, cnp->cn_lkflags);
1357 if (cnp->cn_flags & ISDOTDOT)
1358 vn_lock(dvp, ltype | LK_RETRY);
1368 #ifdef FREEBSD_NAMECACHE
1370 * Insert name into cache (as non-existent) if appropriate.
1372 if (error == ENOENT && (cnp->cn_flags & MAKEENTRY) && nameiop != CREATE)
1373 cache_enter(dvp, *vpp, cnp);
1375 * Insert name into cache if appropriate.
1377 if (error == 0 && (cnp->cn_flags & MAKEENTRY)) {
1378 if (!(cnp->cn_flags & ISLASTCN) ||
1379 (nameiop != DELETE && nameiop != RENAME)) {
1380 cache_enter(dvp, *vpp, cnp);
1389 * Attempt to create a new entry in a directory. If the entry
1390 * already exists, truncate the file if permissible, else return
1391 * an error. Return the vp of the created or trunc'd file.
1393 * IN: dvp - vnode of directory to put new file entry in.
1394 * name - name of new file entry.
1395 * vap - attributes of new file.
1396 * excl - flag indicating exclusive or non-exclusive mode.
1397 * mode - mode to open file with.
1398 * cr - credentials of caller.
1399 * flag - large file flag [UNUSED].
1400 * ct - caller context
1401 * vsecp - ACL to be set
1403 * OUT: vpp - vnode of created or trunc'd entry.
1405 * RETURN: 0 if success
1406 * error code if failure
1409 * dvp - ctime|mtime updated if new entry created
1410 * vp - ctime|mtime always, atime if new
1415 zfs_create(vnode_t *dvp, char *name, vattr_t *vap, int excl, int mode,
1416 vnode_t **vpp, cred_t *cr, kthread_t *td)
1418 znode_t *zp, *dzp = VTOZ(dvp);
1419 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
1427 gid_t gid = crgetgid(cr);
1428 zfs_acl_ids_t acl_ids;
1429 boolean_t fuid_dirtied;
1434 * If we have an ephemeral id, ACL, or XVATTR then
1435 * make sure file system is at proper version
1438 ksid = crgetsid(cr, KSID_OWNER);
1440 uid = ksid_getid(ksid);
1443 if (zfsvfs->z_use_fuids == B_FALSE &&
1444 (vsecp || (vap->va_mask & AT_XVATTR) ||
1445 IS_EPHEMERAL(crgetuid(cr)) || IS_EPHEMERAL(crgetgid(cr))))
1451 zilog = zfsvfs->z_log;
1453 if (zfsvfs->z_utf8 && u8_validate(name, strlen(name),
1454 NULL, U8_VALIDATE_ENTIRE, &error) < 0) {
1459 if (vap->va_mask & AT_XVATTR) {
1460 if ((error = secpolicy_xvattr(dvp, (xvattr_t *)vap,
1461 crgetuid(cr), cr, vap->va_type)) != 0) {
1469 if ((vap->va_mode & S_ISVTX) && secpolicy_vnode_stky_modify(cr))
1470 vap->va_mode &= ~S_ISVTX;
1472 if (*name == '\0') {
1474 * Null component name refers to the directory itself.
1481 /* possible VN_HOLD(zp) */
1484 if (flag & FIGNORECASE)
1487 error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg,
1490 if (strcmp(name, "..") == 0)
1500 * Create a new file object and update the directory
1503 if (error = zfs_zaccess(dzp, ACE_ADD_FILE, 0, B_FALSE, cr)) {
1508 * We only support the creation of regular files in
1509 * extended attribute directories.
1511 if ((dzp->z_phys->zp_flags & ZFS_XATTR) &&
1512 (vap->va_type != VREG)) {
1518 if ((error = zfs_acl_ids_create(dzp, 0, vap, cr, vsecp,
1521 if (zfs_acl_ids_overquota(zfsvfs, &acl_ids)) {
1522 zfs_acl_ids_free(&acl_ids);
1527 tx = dmu_tx_create(os);
1528 dmu_tx_hold_bonus(tx, DMU_NEW_OBJECT);
1529 fuid_dirtied = zfsvfs->z_fuid_dirty;
1531 zfs_fuid_txhold(zfsvfs, tx);
1532 dmu_tx_hold_bonus(tx, dzp->z_id);
1533 dmu_tx_hold_zap(tx, dzp->z_id, TRUE, name);
1534 if (acl_ids.z_aclp->z_acl_bytes > ZFS_ACE_SPACE) {
1535 dmu_tx_hold_write(tx, DMU_NEW_OBJECT,
1536 0, SPA_MAXBLOCKSIZE);
1538 error = dmu_tx_assign(tx, TXG_NOWAIT);
1540 zfs_acl_ids_free(&acl_ids);
1541 zfs_dirent_unlock(dl);
1542 if (error == ERESTART) {
1551 zfs_mknode(dzp, vap, tx, cr, 0, &zp, 0, &acl_ids);
1554 zfs_fuid_sync(zfsvfs, tx);
1556 (void) zfs_link_create(dl, zp, tx, ZNEW);
1558 txtype = zfs_log_create_txtype(Z_FILE, vsecp, vap);
1559 if (flag & FIGNORECASE)
1561 zfs_log_create(zilog, tx, txtype, dzp, zp, name,
1562 vsecp, acl_ids.z_fuidp, vap);
1563 zfs_acl_ids_free(&acl_ids);
1566 int aflags = (flag & FAPPEND) ? V_APPEND : 0;
1569 * A directory entry already exists for this name.
1572 * Can't truncate an existing file if in exclusive mode.
1579 * Can't open a directory for writing.
1581 if ((ZTOV(zp)->v_type == VDIR) && (mode & S_IWRITE)) {
1586 * Verify requested access to file.
1588 if (mode && (error = zfs_zaccess_rwx(zp, mode, aflags, cr))) {
1592 mutex_enter(&dzp->z_lock);
1594 mutex_exit(&dzp->z_lock);
1597 * Truncate regular files if requested.
1599 if ((ZTOV(zp)->v_type == VREG) &&
1600 (vap->va_mask & AT_SIZE) && (vap->va_size == 0)) {
1601 /* we can't hold any locks when calling zfs_freesp() */
1602 zfs_dirent_unlock(dl);
1604 error = zfs_freesp(zp, 0, 0, mode, TRUE);
1606 vnevent_create(ZTOV(zp), ct);
1612 zfs_dirent_unlock(dl);
1619 error = specvp_check(vpp, cr);
1627 * Remove an entry from a directory.
1629 * IN: dvp - vnode of directory to remove entry from.
1630 * name - name of entry to remove.
1631 * cr - credentials of caller.
1632 * ct - caller context
1633 * flags - case flags
1635 * RETURN: 0 if success
1636 * error code if failure
1640 * vp - ctime (if nlink > 0)
1644 zfs_remove(vnode_t *dvp, char *name, cred_t *cr, caller_context_t *ct,
1647 znode_t *zp, *dzp = VTOZ(dvp);
1648 znode_t *xzp = NULL;
1650 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
1652 uint64_t acl_obj, xattr_obj;
1655 boolean_t may_delete_now, delete_now = FALSE;
1656 boolean_t unlinked, toobig = FALSE;
1658 pathname_t *realnmp = NULL;
1665 zilog = zfsvfs->z_log;
1667 if (flags & FIGNORECASE) {
1675 * Attempt to lock directory; fail if entry doesn't exist.
1677 if (error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg,
1687 if (error = zfs_zaccess_delete(dzp, zp, cr)) {
1692 * Need to use rmdir for removing directories.
1694 if (vp->v_type == VDIR) {
1699 vnevent_remove(vp, dvp, name, ct);
1702 dnlc_remove(dvp, realnmp->pn_buf);
1704 dnlc_remove(dvp, name);
1706 may_delete_now = FALSE;
1709 * We may delete the znode now, or we may put it in the unlinked set;
1710 * it depends on whether we're the last link, and on whether there are
1711 * other holds on the vnode. So we dmu_tx_hold() the right things to
1712 * allow for either case.
1714 tx = dmu_tx_create(zfsvfs->z_os);
1715 dmu_tx_hold_zap(tx, dzp->z_id, FALSE, name);
1716 dmu_tx_hold_bonus(tx, zp->z_id);
1717 if (may_delete_now) {
1719 zp->z_phys->zp_size > zp->z_blksz * DMU_MAX_DELETEBLKCNT;
1720 /* if the file is too big, only hold_free a token amount */
1721 dmu_tx_hold_free(tx, zp->z_id, 0,
1722 (toobig ? DMU_MAX_ACCESS : DMU_OBJECT_END));
1725 /* are there any extended attributes? */
1726 if ((xattr_obj = zp->z_phys->zp_xattr) != 0) {
1727 /* XXX - do we need this if we are deleting? */
1728 dmu_tx_hold_bonus(tx, xattr_obj);
1731 /* are there any additional acls */
1732 if ((acl_obj = zp->z_phys->zp_acl.z_acl_extern_obj) != 0 &&
1734 dmu_tx_hold_free(tx, acl_obj, 0, DMU_OBJECT_END);
1736 /* charge as an update -- would be nice not to charge at all */
1737 dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL);
1739 error = dmu_tx_assign(tx, TXG_NOWAIT);
1741 zfs_dirent_unlock(dl);
1743 if (error == ERESTART) {
1756 * Remove the directory entry.
1758 error = zfs_link_destroy(dl, zp, tx, zflg, &unlinked);
1765 if (0 && unlinked) {
1767 delete_now = may_delete_now && !toobig &&
1768 vp->v_count == 1 && !vn_has_cached_data(vp) &&
1769 zp->z_phys->zp_xattr == xattr_obj &&
1770 zp->z_phys->zp_acl.z_acl_extern_obj == acl_obj;
1775 if (zp->z_phys->zp_xattr) {
1776 error = zfs_zget(zfsvfs, zp->z_phys->zp_xattr, &xzp);
1777 ASSERT3U(error, ==, 0);
1778 ASSERT3U(xzp->z_phys->zp_links, ==, 2);
1779 dmu_buf_will_dirty(xzp->z_dbuf, tx);
1780 mutex_enter(&xzp->z_lock);
1781 xzp->z_unlinked = 1;
1782 xzp->z_phys->zp_links = 0;
1783 mutex_exit(&xzp->z_lock);
1784 zfs_unlinked_add(xzp, tx);
1785 zp->z_phys->zp_xattr = 0; /* probably unnecessary */
1787 mutex_enter(&zp->z_lock);
1790 ASSERT3U(vp->v_count, ==, 0);
1792 mutex_exit(&zp->z_lock);
1793 zfs_znode_delete(zp, tx);
1794 } else if (unlinked) {
1795 zfs_unlinked_add(zp, tx);
1799 if (flags & FIGNORECASE)
1801 zfs_log_remove(zilog, tx, txtype, dzp, name);
1808 zfs_dirent_unlock(dl);
1813 /* this rele is delayed to prevent nesting transactions */
1822 * Create a new directory and insert it into dvp using the name
1823 * provided. Return a pointer to the inserted directory.
1825 * IN: dvp - vnode of directory to add subdir to.
1826 * dirname - name of new directory.
1827 * vap - attributes of new directory.
1828 * cr - credentials of caller.
1829 * ct - caller context
1830 * vsecp - ACL to be set
1832 * OUT: vpp - vnode of created directory.
1834 * RETURN: 0 if success
1835 * error code if failure
1838 * dvp - ctime|mtime updated
1839 * vp - ctime|mtime|atime updated
1843 zfs_mkdir(vnode_t *dvp, char *dirname, vattr_t *vap, vnode_t **vpp, cred_t *cr,
1844 caller_context_t *ct, int flags, vsecattr_t *vsecp)
1846 znode_t *zp, *dzp = VTOZ(dvp);
1847 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
1856 gid_t gid = crgetgid(cr);
1857 zfs_acl_ids_t acl_ids;
1858 boolean_t fuid_dirtied;
1860 ASSERT(vap->va_type == VDIR);
1863 * If we have an ephemeral id, ACL, or XVATTR then
1864 * make sure file system is at proper version
1867 ksid = crgetsid(cr, KSID_OWNER);
1869 uid = ksid_getid(ksid);
1872 if (zfsvfs->z_use_fuids == B_FALSE &&
1873 (vsecp || (vap->va_mask & AT_XVATTR) || IS_EPHEMERAL(crgetuid(cr))||
1874 IS_EPHEMERAL(crgetgid(cr))))
1879 zilog = zfsvfs->z_log;
1881 if (dzp->z_phys->zp_flags & ZFS_XATTR) {
1886 if (zfsvfs->z_utf8 && u8_validate(dirname,
1887 strlen(dirname), NULL, U8_VALIDATE_ENTIRE, &error) < 0) {
1891 if (flags & FIGNORECASE)
1894 if (vap->va_mask & AT_XVATTR)
1895 if ((error = secpolicy_xvattr(dvp, (xvattr_t *)vap,
1896 crgetuid(cr), cr, vap->va_type)) != 0) {
1902 * First make sure the new directory doesn't exist.
1907 if (error = zfs_dirent_lock(&dl, dzp, dirname, &zp, zf,
1913 if (error = zfs_zaccess(dzp, ACE_ADD_SUBDIRECTORY, 0, B_FALSE, cr)) {
1914 zfs_dirent_unlock(dl);
1919 if ((error = zfs_acl_ids_create(dzp, 0, vap, cr, vsecp,
1921 zfs_dirent_unlock(dl);
1925 if (zfs_acl_ids_overquota(zfsvfs, &acl_ids)) {
1926 zfs_acl_ids_free(&acl_ids);
1927 zfs_dirent_unlock(dl);
1933 * Add a new entry to the directory.
1935 tx = dmu_tx_create(zfsvfs->z_os);
1936 dmu_tx_hold_zap(tx, dzp->z_id, TRUE, dirname);
1937 dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL);
1938 fuid_dirtied = zfsvfs->z_fuid_dirty;
1940 zfs_fuid_txhold(zfsvfs, tx);
1941 if (acl_ids.z_aclp->z_acl_bytes > ZFS_ACE_SPACE)
1942 dmu_tx_hold_write(tx, DMU_NEW_OBJECT,
1943 0, SPA_MAXBLOCKSIZE);
1944 error = dmu_tx_assign(tx, TXG_NOWAIT);
1946 zfs_acl_ids_free(&acl_ids);
1947 zfs_dirent_unlock(dl);
1948 if (error == ERESTART) {
1961 zfs_mknode(dzp, vap, tx, cr, 0, &zp, 0, &acl_ids);
1964 zfs_fuid_sync(zfsvfs, tx);
1966 * Now put new name in parent dir.
1968 (void) zfs_link_create(dl, zp, tx, ZNEW);
1972 txtype = zfs_log_create_txtype(Z_DIR, vsecp, vap);
1973 if (flags & FIGNORECASE)
1975 zfs_log_create(zilog, tx, txtype, dzp, zp, dirname, vsecp,
1976 acl_ids.z_fuidp, vap);
1978 zfs_acl_ids_free(&acl_ids);
1981 zfs_dirent_unlock(dl);
1988 * Remove a directory subdir entry. If the current working
1989 * directory is the same as the subdir to be removed, the
1992 * IN: dvp - vnode of directory to remove from.
1993 * name - name of directory to be removed.
1994 * cwd - vnode of current working directory.
1995 * cr - credentials of caller.
1996 * ct - caller context
1997 * flags - case flags
1999 * RETURN: 0 if success
2000 * error code if failure
2003 * dvp - ctime|mtime updated
2007 zfs_rmdir(vnode_t *dvp, char *name, vnode_t *cwd, cred_t *cr,
2008 caller_context_t *ct, int flags)
2010 znode_t *dzp = VTOZ(dvp);
2013 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
2022 zilog = zfsvfs->z_log;
2024 if (flags & FIGNORECASE)
2030 * Attempt to lock directory; fail if entry doesn't exist.
2032 if (error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg,
2040 if (error = zfs_zaccess_delete(dzp, zp, cr)) {
2044 if (vp->v_type != VDIR) {
2054 vnevent_rmdir(vp, dvp, name, ct);
2057 * Grab a lock on the directory to make sure that noone is
2058 * trying to add (or lookup) entries while we are removing it.
2060 rw_enter(&zp->z_name_lock, RW_WRITER);
2063 * Grab a lock on the parent pointer to make sure we play well
2064 * with the treewalk and directory rename code.
2066 rw_enter(&zp->z_parent_lock, RW_WRITER);
2068 tx = dmu_tx_create(zfsvfs->z_os);
2069 dmu_tx_hold_zap(tx, dzp->z_id, FALSE, name);
2070 dmu_tx_hold_bonus(tx, zp->z_id);
2071 dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL);
2072 error = dmu_tx_assign(tx, TXG_NOWAIT);
2074 rw_exit(&zp->z_parent_lock);
2075 rw_exit(&zp->z_name_lock);
2076 zfs_dirent_unlock(dl);
2078 if (error == ERESTART) {
2088 #ifdef FREEBSD_NAMECACHE
2092 error = zfs_link_destroy(dl, zp, tx, zflg, NULL);
2095 uint64_t txtype = TX_RMDIR;
2096 if (flags & FIGNORECASE)
2098 zfs_log_remove(zilog, tx, txtype, dzp, name);
2103 rw_exit(&zp->z_parent_lock);
2104 rw_exit(&zp->z_name_lock);
2105 #ifdef FREEBSD_NAMECACHE
2109 zfs_dirent_unlock(dl);
2118 * Read as many directory entries as will fit into the provided
2119 * buffer from the given directory cursor position (specified in
2120 * the uio structure.
2122 * IN: vp - vnode of directory to read.
2123 * uio - structure supplying read location, range info,
2124 * and return buffer.
2125 * cr - credentials of caller.
2126 * ct - caller context
2127 * flags - case flags
2129 * OUT: uio - updated offset and range, buffer filled.
2130 * eofp - set to true if end-of-file detected.
2132 * RETURN: 0 if success
2133 * error code if failure
2136 * vp - atime updated
2138 * Note that the low 4 bits of the cookie returned by zap is always zero.
2139 * This allows us to use the low range for "special" directory entries:
2140 * We use 0 for '.', and 1 for '..'. If this is the root of the filesystem,
2141 * we use the offset 2 for the '.zfs' directory.
2145 zfs_readdir(vnode_t *vp, uio_t *uio, cred_t *cr, int *eofp, int *ncookies, u_long **cookies)
2147 znode_t *zp = VTOZ(vp);
2151 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
2156 zap_attribute_t zap;
2157 uint_t bytes_wanted;
2158 uint64_t offset; /* must be unsigned; checks for < 1 */
2163 boolean_t check_sysattrs;
2166 u_long *cooks = NULL;
2173 * If we are not given an eof variable,
2180 * Check for valid iov_len.
2182 if (uio->uio_iov->iov_len <= 0) {
2188 * Quit if directory has been removed (posix)
2190 if ((*eofp = zp->z_unlinked) != 0) {
2197 offset = uio->uio_loffset;
2198 prefetch = zp->z_zn_prefetch;
2201 * Initialize the iterator cursor.
2205 * Start iteration from the beginning of the directory.
2207 zap_cursor_init(&zc, os, zp->z_id);
2210 * The offset is a serialized cursor.
2212 zap_cursor_init_serialized(&zc, os, zp->z_id, offset);
2216 * Get space to change directory entries into fs independent format.
2218 iovp = uio->uio_iov;
2219 bytes_wanted = iovp->iov_len;
2220 if (uio->uio_segflg != UIO_SYSSPACE || uio->uio_iovcnt != 1) {
2221 bufsize = bytes_wanted;
2222 outbuf = kmem_alloc(bufsize, KM_SLEEP);
2223 odp = (struct dirent64 *)outbuf;
2225 bufsize = bytes_wanted;
2226 odp = (struct dirent64 *)iovp->iov_base;
2228 eodp = (struct edirent *)odp;
2230 if (ncookies != NULL) {
2232 * Minimum entry size is dirent size and 1 byte for a file name.
2234 ncooks = uio->uio_resid / (sizeof(struct dirent) - sizeof(((struct dirent *)NULL)->d_name) + 1);
2235 cooks = malloc(ncooks * sizeof(u_long), M_TEMP, M_WAITOK);
2240 * If this VFS supports the system attribute view interface; and
2241 * we're looking at an extended attribute directory; and we care
2242 * about normalization conflicts on this vfs; then we must check
2243 * for normalization conflicts with the sysattr name space.
2246 check_sysattrs = vfs_has_feature(vp->v_vfsp, VFSFT_SYSATTR_VIEWS) &&
2247 (vp->v_flag & V_XATTRDIR) && zfsvfs->z_norm &&
2248 (flags & V_RDDIR_ENTFLAGS);
2254 * Transform to file-system independent format
2257 while (outcount < bytes_wanted) {
2263 * Special case `.', `..', and `.zfs'.
2266 (void) strcpy(zap.za_name, ".");
2267 zap.za_normalization_conflict = 0;
2270 } else if (offset == 1) {
2271 (void) strcpy(zap.za_name, "..");
2272 zap.za_normalization_conflict = 0;
2273 objnum = zp->z_phys->zp_parent;
2275 } else if (offset == 2 && zfs_show_ctldir(zp)) {
2276 (void) strcpy(zap.za_name, ZFS_CTLDIR_NAME);
2277 zap.za_normalization_conflict = 0;
2278 objnum = ZFSCTL_INO_ROOT;
2284 if (error = zap_cursor_retrieve(&zc, &zap)) {
2285 if ((*eofp = (error == ENOENT)) != 0)
2291 if (zap.za_integer_length != 8 ||
2292 zap.za_num_integers != 1) {
2293 cmn_err(CE_WARN, "zap_readdir: bad directory "
2294 "entry, obj = %lld, offset = %lld\n",
2295 (u_longlong_t)zp->z_id,
2296 (u_longlong_t)offset);
2301 objnum = ZFS_DIRENT_OBJ(zap.za_first_integer);
2303 * MacOS X can extract the object type here such as:
2304 * uint8_t type = ZFS_DIRENT_TYPE(zap.za_first_integer);
2306 type = ZFS_DIRENT_TYPE(zap.za_first_integer);
2308 if (check_sysattrs && !zap.za_normalization_conflict) {
2310 zap.za_normalization_conflict =
2311 xattr_sysattr_casechk(zap.za_name);
2313 panic("%s:%u: TODO", __func__, __LINE__);
2318 if (flags & V_RDDIR_ACCFILTER) {
2320 * If we have no access at all, don't include
2321 * this entry in the returned information
2324 if (zfs_zget(zp->z_zfsvfs, objnum, &ezp) != 0)
2326 if (!zfs_has_access(ezp, cr)) {
2333 if (flags & V_RDDIR_ENTFLAGS)
2334 reclen = EDIRENT_RECLEN(strlen(zap.za_name));
2336 reclen = DIRENT64_RECLEN(strlen(zap.za_name));
2339 * Will this entry fit in the buffer?
2341 if (outcount + reclen > bufsize) {
2343 * Did we manage to fit anything in the buffer?
2351 if (flags & V_RDDIR_ENTFLAGS) {
2353 * Add extended flag entry:
2355 eodp->ed_ino = objnum;
2356 eodp->ed_reclen = reclen;
2357 /* NOTE: ed_off is the offset for the *next* entry */
2358 next = &(eodp->ed_off);
2359 eodp->ed_eflags = zap.za_normalization_conflict ?
2360 ED_CASE_CONFLICT : 0;
2361 (void) strncpy(eodp->ed_name, zap.za_name,
2362 EDIRENT_NAMELEN(reclen));
2363 eodp = (edirent_t *)((intptr_t)eodp + reclen);
2368 odp->d_ino = objnum;
2369 odp->d_reclen = reclen;
2370 odp->d_namlen = strlen(zap.za_name);
2371 (void) strlcpy(odp->d_name, zap.za_name, odp->d_namlen + 1);
2373 odp = (dirent64_t *)((intptr_t)odp + reclen);
2377 ASSERT(outcount <= bufsize);
2379 /* Prefetch znode */
2381 dmu_prefetch(os, objnum, 0, 0);
2385 * Move to the next entry, fill in the previous offset.
2387 if (offset > 2 || (offset == 2 && !zfs_show_ctldir(zp))) {
2388 zap_cursor_advance(&zc);
2389 offset = zap_cursor_serialize(&zc);
2394 if (cooks != NULL) {
2397 KASSERT(ncooks >= 0, ("ncookies=%d", ncooks));
2400 zp->z_zn_prefetch = B_FALSE; /* a lookup will re-enable pre-fetching */
2402 /* Subtract unused cookies */
2403 if (ncookies != NULL)
2404 *ncookies -= ncooks;
2406 if (uio->uio_segflg == UIO_SYSSPACE && uio->uio_iovcnt == 1) {
2407 iovp->iov_base += outcount;
2408 iovp->iov_len -= outcount;
2409 uio->uio_resid -= outcount;
2410 } else if (error = uiomove(outbuf, (long)outcount, UIO_READ, uio)) {
2412 * Reset the pointer.
2414 offset = uio->uio_loffset;
2418 zap_cursor_fini(&zc);
2419 if (uio->uio_segflg != UIO_SYSSPACE || uio->uio_iovcnt != 1)
2420 kmem_free(outbuf, bufsize);
2422 if (error == ENOENT)
2425 ZFS_ACCESSTIME_STAMP(zfsvfs, zp);
2427 uio->uio_loffset = offset;
2429 if (error != 0 && cookies != NULL) {
2430 free(*cookies, M_TEMP);
2437 ulong_t zfs_fsync_sync_cnt = 4;
2440 zfs_fsync(vnode_t *vp, int syncflag, cred_t *cr, caller_context_t *ct)
2442 znode_t *zp = VTOZ(vp);
2443 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
2445 (void) tsd_set(zfs_fsyncer_key, (void *)zfs_fsync_sync_cnt);
2449 zil_commit(zfsvfs->z_log, zp->z_last_itx, zp->z_id);
2456 * Get the requested file attributes and place them in the provided
2459 * IN: vp - vnode of file.
2460 * vap - va_mask identifies requested attributes.
2461 * If AT_XVATTR set, then optional attrs are requested
2462 * flags - ATTR_NOACLCHECK (CIFS server context)
2463 * cr - credentials of caller.
2464 * ct - caller context
2466 * OUT: vap - attribute values.
2468 * RETURN: 0 (always succeeds)
2472 zfs_getattr(vnode_t *vp, vattr_t *vap, int flags, cred_t *cr,
2473 caller_context_t *ct)
2475 znode_t *zp = VTOZ(vp);
2476 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
2480 u_longlong_t nblocks;
2482 xvattr_t *xvap = (xvattr_t *)vap; /* vap may be an xvattr_t * */
2483 xoptattr_t *xoap = NULL;
2484 boolean_t skipaclchk = (flags & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE;
2491 * If ACL is trivial don't bother looking for ACE_READ_ATTRIBUTES.
2492 * Also, if we are the owner don't bother, since owner should
2493 * always be allowed to read basic attributes of file.
2495 if (!(pzp->zp_flags & ZFS_ACL_TRIVIAL) &&
2496 (pzp->zp_uid != crgetuid(cr))) {
2497 if (error = zfs_zaccess(zp, ACE_READ_ATTRIBUTES, 0,
2505 * Return all attributes. It's cheaper to provide the answer
2506 * than to determine whether we were asked the question.
2509 mutex_enter(&zp->z_lock);
2510 vap->va_type = IFTOVT(pzp->zp_mode);
2511 vap->va_mode = pzp->zp_mode & ~S_IFMT;
2512 zfs_fuid_map_ids(zp, cr, &vap->va_uid, &vap->va_gid);
2513 // vap->va_fsid = zp->z_zfsvfs->z_vfs->vfs_dev;
2514 vap->va_nodeid = zp->z_id;
2515 if ((vp->v_flag & VROOT) && zfs_show_ctldir(zp))
2516 links = pzp->zp_links + 1;
2518 links = pzp->zp_links;
2519 vap->va_nlink = MIN(links, UINT32_MAX); /* nlink_t limit! */
2520 vap->va_size = pzp->zp_size;
2521 vap->va_fsid = vp->v_mount->mnt_stat.f_fsid.val[0];
2522 vap->va_rdev = zfs_cmpldev(pzp->zp_rdev);
2523 vap->va_seq = zp->z_seq;
2524 vap->va_flags = 0; /* FreeBSD: Reset chflags(2) flags. */
2527 * Add in any requested optional attributes and the create time.
2528 * Also set the corresponding bits in the returned attribute bitmap.
2530 if ((xoap = xva_getxoptattr(xvap)) != NULL && zfsvfs->z_use_fuids) {
2531 if (XVA_ISSET_REQ(xvap, XAT_ARCHIVE)) {
2533 ((pzp->zp_flags & ZFS_ARCHIVE) != 0);
2534 XVA_SET_RTN(xvap, XAT_ARCHIVE);
2537 if (XVA_ISSET_REQ(xvap, XAT_READONLY)) {
2538 xoap->xoa_readonly =
2539 ((pzp->zp_flags & ZFS_READONLY) != 0);
2540 XVA_SET_RTN(xvap, XAT_READONLY);
2543 if (XVA_ISSET_REQ(xvap, XAT_SYSTEM)) {
2545 ((pzp->zp_flags & ZFS_SYSTEM) != 0);
2546 XVA_SET_RTN(xvap, XAT_SYSTEM);
2549 if (XVA_ISSET_REQ(xvap, XAT_HIDDEN)) {
2551 ((pzp->zp_flags & ZFS_HIDDEN) != 0);
2552 XVA_SET_RTN(xvap, XAT_HIDDEN);
2555 if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK)) {
2556 xoap->xoa_nounlink =
2557 ((pzp->zp_flags & ZFS_NOUNLINK) != 0);
2558 XVA_SET_RTN(xvap, XAT_NOUNLINK);
2561 if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE)) {
2562 xoap->xoa_immutable =
2563 ((pzp->zp_flags & ZFS_IMMUTABLE) != 0);
2564 XVA_SET_RTN(xvap, XAT_IMMUTABLE);
2567 if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY)) {
2568 xoap->xoa_appendonly =
2569 ((pzp->zp_flags & ZFS_APPENDONLY) != 0);
2570 XVA_SET_RTN(xvap, XAT_APPENDONLY);
2573 if (XVA_ISSET_REQ(xvap, XAT_NODUMP)) {
2575 ((pzp->zp_flags & ZFS_NODUMP) != 0);
2576 XVA_SET_RTN(xvap, XAT_NODUMP);
2579 if (XVA_ISSET_REQ(xvap, XAT_OPAQUE)) {
2581 ((pzp->zp_flags & ZFS_OPAQUE) != 0);
2582 XVA_SET_RTN(xvap, XAT_OPAQUE);
2585 if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED)) {
2586 xoap->xoa_av_quarantined =
2587 ((pzp->zp_flags & ZFS_AV_QUARANTINED) != 0);
2588 XVA_SET_RTN(xvap, XAT_AV_QUARANTINED);
2591 if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED)) {
2592 xoap->xoa_av_modified =
2593 ((pzp->zp_flags & ZFS_AV_MODIFIED) != 0);
2594 XVA_SET_RTN(xvap, XAT_AV_MODIFIED);
2597 if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP) &&
2598 vp->v_type == VREG &&
2599 (pzp->zp_flags & ZFS_BONUS_SCANSTAMP)) {
2601 dmu_object_info_t doi;
2604 * Only VREG files have anti-virus scanstamps, so we
2605 * won't conflict with symlinks in the bonus buffer.
2607 dmu_object_info_from_db(zp->z_dbuf, &doi);
2608 len = sizeof (xoap->xoa_av_scanstamp) +
2609 sizeof (znode_phys_t);
2610 if (len <= doi.doi_bonus_size) {
2612 * pzp points to the start of the
2613 * znode_phys_t. pzp + 1 points to the
2614 * first byte after the znode_phys_t.
2616 (void) memcpy(xoap->xoa_av_scanstamp,
2618 sizeof (xoap->xoa_av_scanstamp));
2619 XVA_SET_RTN(xvap, XAT_AV_SCANSTAMP);
2623 if (XVA_ISSET_REQ(xvap, XAT_CREATETIME)) {
2624 ZFS_TIME_DECODE(&xoap->xoa_createtime, pzp->zp_crtime);
2625 XVA_SET_RTN(xvap, XAT_CREATETIME);
2629 ZFS_TIME_DECODE(&vap->va_atime, pzp->zp_atime);
2630 ZFS_TIME_DECODE(&vap->va_mtime, pzp->zp_mtime);
2631 ZFS_TIME_DECODE(&vap->va_ctime, pzp->zp_ctime);
2632 ZFS_TIME_DECODE(&vap->va_birthtime, pzp->zp_crtime);
2634 mutex_exit(&zp->z_lock);
2636 dmu_object_size_from_db(zp->z_dbuf, &blksize, &nblocks);
2637 vap->va_blksize = blksize;
2638 vap->va_bytes = nblocks << 9; /* nblocks * 512 */
2640 if (zp->z_blksz == 0) {
2642 * Block size hasn't been set; suggest maximal I/O transfers.
2644 vap->va_blksize = zfsvfs->z_max_blksz;
2652 * Set the file attributes to the values contained in the
2655 * IN: vp - vnode of file to be modified.
2656 * vap - new attribute values.
2657 * If AT_XVATTR set, then optional attrs are being set
2658 * flags - ATTR_UTIME set if non-default time values provided.
2659 * - ATTR_NOACLCHECK (CIFS context only).
2660 * cr - credentials of caller.
2661 * ct - caller context
2663 * RETURN: 0 if success
2664 * error code if failure
2667 * vp - ctime updated, mtime updated if size changed.
2671 zfs_setattr(vnode_t *vp, vattr_t *vap, int flags, cred_t *cr,
2672 caller_context_t *ct)
2674 znode_t *zp = VTOZ(vp);
2676 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
2681 uint_t mask = vap->va_mask;
2683 uint64_t saved_mode;
2686 uint64_t new_uid, new_gid;
2688 int need_policy = FALSE;
2690 zfs_fuid_info_t *fuidp = NULL;
2691 xvattr_t *xvap = (xvattr_t *)vap; /* vap may be an xvattr_t * */
2693 zfs_acl_t *aclp = NULL;
2694 boolean_t skipaclchk = (flags & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE;
2695 boolean_t fuid_dirtied = B_FALSE;
2700 if (mask & AT_NOSET)
2707 zilog = zfsvfs->z_log;
2710 * Make sure that if we have ephemeral uid/gid or xvattr specified
2711 * that file system is at proper version level
2714 if (zfsvfs->z_use_fuids == B_FALSE &&
2715 (((mask & AT_UID) && IS_EPHEMERAL(vap->va_uid)) ||
2716 ((mask & AT_GID) && IS_EPHEMERAL(vap->va_gid)) ||
2717 (mask & AT_XVATTR))) {
2722 if (mask & AT_SIZE && vp->v_type == VDIR) {
2727 if (mask & AT_SIZE && vp->v_type != VREG && vp->v_type != VFIFO) {
2733 * If this is an xvattr_t, then get a pointer to the structure of
2734 * optional attributes. If this is NULL, then we have a vattr_t.
2736 xoap = xva_getxoptattr(xvap);
2738 xva_init(&tmpxvattr);
2741 * Immutable files can only alter immutable bit and atime
2743 if ((pzp->zp_flags & ZFS_IMMUTABLE) &&
2744 ((mask & (AT_SIZE|AT_UID|AT_GID|AT_MTIME|AT_MODE)) ||
2745 ((mask & AT_XVATTR) && XVA_ISSET_REQ(xvap, XAT_CREATETIME)))) {
2750 if ((mask & AT_SIZE) && (pzp->zp_flags & ZFS_READONLY)) {
2756 * Verify timestamps doesn't overflow 32 bits.
2757 * ZFS can handle large timestamps, but 32bit syscalls can't
2758 * handle times greater than 2039. This check should be removed
2759 * once large timestamps are fully supported.
2761 if (mask & (AT_ATIME | AT_MTIME)) {
2762 if (((mask & AT_ATIME) && TIMESPEC_OVERFLOW(&vap->va_atime)) ||
2763 ((mask & AT_MTIME) && TIMESPEC_OVERFLOW(&vap->va_mtime))) {
2772 /* Can this be moved to before the top label? */
2773 if (zfsvfs->z_vfs->vfs_flag & VFS_RDONLY) {
2779 * First validate permissions
2782 if (mask & AT_SIZE) {
2783 err = zfs_zaccess(zp, ACE_WRITE_DATA, 0, skipaclchk, cr);
2789 * XXX - Note, we are not providing any open
2790 * mode flags here (like FNDELAY), so we may
2791 * block if there are locks present... this
2792 * should be addressed in openat().
2794 /* XXX - would it be OK to generate a log record here? */
2795 err = zfs_freesp(zp, vap->va_size, 0, 0, FALSE);
2802 if (mask & (AT_ATIME|AT_MTIME) ||
2803 ((mask & AT_XVATTR) && (XVA_ISSET_REQ(xvap, XAT_HIDDEN) ||
2804 XVA_ISSET_REQ(xvap, XAT_READONLY) ||
2805 XVA_ISSET_REQ(xvap, XAT_ARCHIVE) ||
2806 XVA_ISSET_REQ(xvap, XAT_CREATETIME) ||
2807 XVA_ISSET_REQ(xvap, XAT_SYSTEM))))
2808 need_policy = zfs_zaccess(zp, ACE_WRITE_ATTRIBUTES, 0,
2811 if (mask & (AT_UID|AT_GID)) {
2812 int idmask = (mask & (AT_UID|AT_GID));
2817 * NOTE: even if a new mode is being set,
2818 * we may clear S_ISUID/S_ISGID bits.
2821 if (!(mask & AT_MODE))
2822 vap->va_mode = pzp->zp_mode;
2825 * Take ownership or chgrp to group we are a member of
2828 take_owner = (mask & AT_UID) && (vap->va_uid == crgetuid(cr));
2829 take_group = (mask & AT_GID) &&
2830 zfs_groupmember(zfsvfs, vap->va_gid, cr);
2833 * If both AT_UID and AT_GID are set then take_owner and
2834 * take_group must both be set in order to allow taking
2837 * Otherwise, send the check through secpolicy_vnode_setattr()
2841 if (((idmask == (AT_UID|AT_GID)) && take_owner && take_group) ||
2842 ((idmask == AT_UID) && take_owner) ||
2843 ((idmask == AT_GID) && take_group)) {
2844 if (zfs_zaccess(zp, ACE_WRITE_OWNER, 0,
2845 skipaclchk, cr) == 0) {
2847 * Remove setuid/setgid for non-privileged users
2849 secpolicy_setid_clear(vap, vp, cr);
2850 trim_mask = (mask & (AT_UID|AT_GID));
2859 mutex_enter(&zp->z_lock);
2860 oldva.va_mode = pzp->zp_mode;
2861 zfs_fuid_map_ids(zp, cr, &oldva.va_uid, &oldva.va_gid);
2862 if (mask & AT_XVATTR) {
2864 * Update xvattr mask to include only those attributes
2865 * that are actually changing.
2867 * the bits will be restored prior to actually setting
2868 * the attributes so the caller thinks they were set.
2870 if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY)) {
2871 if (xoap->xoa_appendonly !=
2872 ((pzp->zp_flags & ZFS_APPENDONLY) != 0)) {
2875 XVA_CLR_REQ(xvap, XAT_APPENDONLY);
2876 XVA_SET_REQ(&tmpxvattr, XAT_APPENDONLY);
2880 if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK)) {
2881 if (xoap->xoa_nounlink !=
2882 ((pzp->zp_flags & ZFS_NOUNLINK) != 0)) {
2885 XVA_CLR_REQ(xvap, XAT_NOUNLINK);
2886 XVA_SET_REQ(&tmpxvattr, XAT_NOUNLINK);
2890 if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE)) {
2891 if (xoap->xoa_immutable !=
2892 ((pzp->zp_flags & ZFS_IMMUTABLE) != 0)) {
2895 XVA_CLR_REQ(xvap, XAT_IMMUTABLE);
2896 XVA_SET_REQ(&tmpxvattr, XAT_IMMUTABLE);
2900 if (XVA_ISSET_REQ(xvap, XAT_NODUMP)) {
2901 if (xoap->xoa_nodump !=
2902 ((pzp->zp_flags & ZFS_NODUMP) != 0)) {
2905 XVA_CLR_REQ(xvap, XAT_NODUMP);
2906 XVA_SET_REQ(&tmpxvattr, XAT_NODUMP);
2910 if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED)) {
2911 if (xoap->xoa_av_modified !=
2912 ((pzp->zp_flags & ZFS_AV_MODIFIED) != 0)) {
2915 XVA_CLR_REQ(xvap, XAT_AV_MODIFIED);
2916 XVA_SET_REQ(&tmpxvattr, XAT_AV_MODIFIED);
2920 if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED)) {
2921 if ((vp->v_type != VREG &&
2922 xoap->xoa_av_quarantined) ||
2923 xoap->xoa_av_quarantined !=
2924 ((pzp->zp_flags & ZFS_AV_QUARANTINED) != 0)) {
2927 XVA_CLR_REQ(xvap, XAT_AV_QUARANTINED);
2928 XVA_SET_REQ(&tmpxvattr, XAT_AV_QUARANTINED);
2932 if (need_policy == FALSE &&
2933 (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP) ||
2934 XVA_ISSET_REQ(xvap, XAT_OPAQUE))) {
2939 mutex_exit(&zp->z_lock);
2941 if (mask & AT_MODE) {
2942 if (zfs_zaccess(zp, ACE_WRITE_ACL, 0, skipaclchk, cr) == 0) {
2943 err = secpolicy_setid_setsticky_clear(vp, vap,
2949 trim_mask |= AT_MODE;
2957 * If trim_mask is set then take ownership
2958 * has been granted or write_acl is present and user
2959 * has the ability to modify mode. In that case remove
2960 * UID|GID and or MODE from mask so that
2961 * secpolicy_vnode_setattr() doesn't revoke it.
2965 saved_mask = vap->va_mask;
2966 vap->va_mask &= ~trim_mask;
2967 if (trim_mask & AT_MODE) {
2969 * Save the mode, as secpolicy_vnode_setattr()
2970 * will overwrite it with ova.va_mode.
2972 saved_mode = vap->va_mode;
2975 err = secpolicy_vnode_setattr(cr, vp, vap, &oldva, flags,
2976 (int (*)(void *, int, cred_t *))zfs_zaccess_unix, zp);
2983 vap->va_mask |= saved_mask;
2984 if (trim_mask & AT_MODE) {
2986 * Recover the mode after
2987 * secpolicy_vnode_setattr().
2989 vap->va_mode = saved_mode;
2995 * secpolicy_vnode_setattr, or take ownership may have
2998 mask = vap->va_mask;
3000 tx = dmu_tx_create(zfsvfs->z_os);
3001 dmu_tx_hold_bonus(tx, zp->z_id);
3003 if (mask & AT_MODE) {
3004 uint64_t pmode = pzp->zp_mode;
3006 new_mode = (pmode & S_IFMT) | (vap->va_mode & ~S_IFMT);
3008 if (err = zfs_acl_chmod_setattr(zp, &aclp, new_mode))
3010 if (pzp->zp_acl.z_acl_extern_obj) {
3011 /* Are we upgrading ACL from old V0 format to new V1 */
3012 if (zfsvfs->z_version <= ZPL_VERSION_FUID &&
3013 pzp->zp_acl.z_acl_version ==
3014 ZFS_ACL_VERSION_INITIAL) {
3015 dmu_tx_hold_free(tx,
3016 pzp->zp_acl.z_acl_extern_obj, 0,
3018 dmu_tx_hold_write(tx, DMU_NEW_OBJECT,
3019 0, aclp->z_acl_bytes);
3021 dmu_tx_hold_write(tx,
3022 pzp->zp_acl.z_acl_extern_obj, 0,
3025 } else if (aclp->z_acl_bytes > ZFS_ACE_SPACE) {
3026 dmu_tx_hold_write(tx, DMU_NEW_OBJECT,
3027 0, aclp->z_acl_bytes);
3031 if (mask & (AT_UID | AT_GID)) {
3032 if (pzp->zp_xattr) {
3033 err = zfs_zget(zp->z_zfsvfs, pzp->zp_xattr, &attrzp);
3036 dmu_tx_hold_bonus(tx, attrzp->z_id);
3038 if (mask & AT_UID) {
3039 new_uid = zfs_fuid_create(zfsvfs,
3040 (uint64_t)vap->va_uid, cr, ZFS_OWNER, &fuidp);
3041 if (new_uid != pzp->zp_uid &&
3042 zfs_usergroup_overquota(zfsvfs, B_FALSE, new_uid)) {
3048 if (mask & AT_GID) {
3049 new_gid = zfs_fuid_create(zfsvfs, (uint64_t)vap->va_gid,
3050 cr, ZFS_GROUP, &fuidp);
3051 if (new_gid != pzp->zp_gid &&
3052 zfs_usergroup_overquota(zfsvfs, B_TRUE, new_gid)) {
3057 fuid_dirtied = zfsvfs->z_fuid_dirty;
3059 if (zfsvfs->z_fuid_obj == 0) {
3060 dmu_tx_hold_bonus(tx, DMU_NEW_OBJECT);
3061 dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0,
3062 FUID_SIZE_ESTIMATE(zfsvfs));
3063 dmu_tx_hold_zap(tx, MASTER_NODE_OBJ,
3066 dmu_tx_hold_bonus(tx, zfsvfs->z_fuid_obj);
3067 dmu_tx_hold_write(tx, zfsvfs->z_fuid_obj, 0,
3068 FUID_SIZE_ESTIMATE(zfsvfs));
3073 err = dmu_tx_assign(tx, TXG_NOWAIT);
3075 if (err == ERESTART)
3080 dmu_buf_will_dirty(zp->z_dbuf, tx);
3083 * Set each attribute requested.
3084 * We group settings according to the locks they need to acquire.
3086 * Note: you cannot set ctime directly, although it will be
3087 * updated as a side-effect of calling this function.
3090 mutex_enter(&zp->z_lock);
3092 if (mask & AT_MODE) {
3093 mutex_enter(&zp->z_acl_lock);
3094 zp->z_phys->zp_mode = new_mode;
3095 err = zfs_aclset_common(zp, aclp, cr, tx);
3096 ASSERT3U(err, ==, 0);
3097 zp->z_acl_cached = aclp;
3099 mutex_exit(&zp->z_acl_lock);
3103 mutex_enter(&attrzp->z_lock);
3105 if (mask & AT_UID) {
3106 pzp->zp_uid = new_uid;
3108 attrzp->z_phys->zp_uid = new_uid;
3111 if (mask & AT_GID) {
3112 pzp->zp_gid = new_gid;
3114 attrzp->z_phys->zp_gid = new_gid;
3118 mutex_exit(&attrzp->z_lock);
3120 if (mask & AT_ATIME)
3121 ZFS_TIME_ENCODE(&vap->va_atime, pzp->zp_atime);
3123 if (mask & AT_MTIME)
3124 ZFS_TIME_ENCODE(&vap->va_mtime, pzp->zp_mtime);
3126 /* XXX - shouldn't this be done *before* the ATIME/MTIME checks? */
3128 zfs_time_stamper_locked(zp, CONTENT_MODIFIED, tx);
3130 zfs_time_stamper_locked(zp, STATE_CHANGED, tx);
3132 * Do this after setting timestamps to prevent timestamp
3133 * update from toggling bit
3136 if (xoap && (mask & AT_XVATTR)) {
3139 * restore trimmed off masks
3140 * so that return masks can be set for caller.
3143 if (XVA_ISSET_REQ(&tmpxvattr, XAT_APPENDONLY)) {
3144 XVA_SET_REQ(xvap, XAT_APPENDONLY);
3146 if (XVA_ISSET_REQ(&tmpxvattr, XAT_NOUNLINK)) {
3147 XVA_SET_REQ(xvap, XAT_NOUNLINK);
3149 if (XVA_ISSET_REQ(&tmpxvattr, XAT_IMMUTABLE)) {
3150 XVA_SET_REQ(xvap, XAT_IMMUTABLE);
3152 if (XVA_ISSET_REQ(&tmpxvattr, XAT_NODUMP)) {
3153 XVA_SET_REQ(xvap, XAT_NODUMP);
3155 if (XVA_ISSET_REQ(&tmpxvattr, XAT_AV_MODIFIED)) {
3156 XVA_SET_REQ(xvap, XAT_AV_MODIFIED);
3158 if (XVA_ISSET_REQ(&tmpxvattr, XAT_AV_QUARANTINED)) {
3159 XVA_SET_REQ(xvap, XAT_AV_QUARANTINED);
3162 if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP)) {
3164 dmu_object_info_t doi;
3166 ASSERT(vp->v_type == VREG);
3168 /* Grow the bonus buffer if necessary. */
3169 dmu_object_info_from_db(zp->z_dbuf, &doi);
3170 len = sizeof (xoap->xoa_av_scanstamp) +
3171 sizeof (znode_phys_t);
3172 if (len > doi.doi_bonus_size)
3173 VERIFY(dmu_set_bonus(zp->z_dbuf, len, tx) == 0);
3175 zfs_xvattr_set(zp, xvap);
3179 zfs_fuid_sync(zfsvfs, tx);
3182 zfs_log_setattr(zilog, tx, TX_SETATTR, zp, vap, mask, fuidp);
3184 mutex_exit(&zp->z_lock);
3188 VN_RELE(ZTOV(attrzp));
3194 zfs_fuid_info_free(fuidp);
3203 if (err == ERESTART)
3210 typedef struct zfs_zlock {
3211 krwlock_t *zl_rwlock; /* lock we acquired */
3212 znode_t *zl_znode; /* znode we held */
3213 struct zfs_zlock *zl_next; /* next in list */
3217 * Drop locks and release vnodes that were held by zfs_rename_lock().
3220 zfs_rename_unlock(zfs_zlock_t **zlpp)
3224 while ((zl = *zlpp) != NULL) {
3225 if (zl->zl_znode != NULL)
3226 VN_RELE(ZTOV(zl->zl_znode));
3227 rw_exit(zl->zl_rwlock);
3228 *zlpp = zl->zl_next;
3229 kmem_free(zl, sizeof (*zl));
3234 * Search back through the directory tree, using the ".." entries.
3235 * Lock each directory in the chain to prevent concurrent renames.
3236 * Fail any attempt to move a directory into one of its own descendants.
3237 * XXX - z_parent_lock can overlap with map or grow locks
3240 zfs_rename_lock(znode_t *szp, znode_t *tdzp, znode_t *sdzp, zfs_zlock_t **zlpp)
3244 uint64_t rootid = zp->z_zfsvfs->z_root;
3245 uint64_t *oidp = &zp->z_id;
3246 krwlock_t *rwlp = &szp->z_parent_lock;
3247 krw_t rw = RW_WRITER;
3250 * First pass write-locks szp and compares to zp->z_id.
3251 * Later passes read-lock zp and compare to zp->z_parent.
3254 if (!rw_tryenter(rwlp, rw)) {
3256 * Another thread is renaming in this path.
3257 * Note that if we are a WRITER, we don't have any
3258 * parent_locks held yet.
3260 if (rw == RW_READER && zp->z_id > szp->z_id) {
3262 * Drop our locks and restart
3264 zfs_rename_unlock(&zl);
3268 rwlp = &szp->z_parent_lock;
3273 * Wait for other thread to drop its locks
3279 zl = kmem_alloc(sizeof (*zl), KM_SLEEP);
3280 zl->zl_rwlock = rwlp;
3281 zl->zl_znode = NULL;
3282 zl->zl_next = *zlpp;
3285 if (*oidp == szp->z_id) /* We're a descendant of szp */
3288 if (*oidp == rootid) /* We've hit the top */
3291 if (rw == RW_READER) { /* i.e. not the first pass */
3292 int error = zfs_zget(zp->z_zfsvfs, *oidp, &zp);
3297 oidp = &zp->z_phys->zp_parent;
3298 rwlp = &zp->z_parent_lock;
3301 } while (zp->z_id != sdzp->z_id);
3307 * Move an entry from the provided source directory to the target
3308 * directory. Change the entry name as indicated.
3310 * IN: sdvp - Source directory containing the "old entry".
3311 * snm - Old entry name.
3312 * tdvp - Target directory to contain the "new entry".
3313 * tnm - New entry name.
3314 * cr - credentials of caller.
3315 * ct - caller context
3316 * flags - case flags
3318 * RETURN: 0 if success
3319 * error code if failure
3322 * sdvp,tdvp - ctime|mtime updated
3326 zfs_rename(vnode_t *sdvp, char *snm, vnode_t *tdvp, char *tnm, cred_t *cr,
3327 caller_context_t *ct, int flags)
3329 znode_t *tdzp, *szp, *tzp;
3330 znode_t *sdzp = VTOZ(sdvp);
3331 zfsvfs_t *zfsvfs = sdzp->z_zfsvfs;
3334 zfs_dirlock_t *sdl, *tdl;
3337 int cmp, serr, terr;
3342 ZFS_VERIFY_ZP(sdzp);
3343 zilog = zfsvfs->z_log;
3346 * Make sure we have the real vp for the target directory.
3348 if (VOP_REALVP(tdvp, &realvp, ct) == 0)
3351 if (tdvp->v_vfsp != sdvp->v_vfsp || zfsctl_is_node(tdvp)) {
3357 ZFS_VERIFY_ZP(tdzp);
3358 if (zfsvfs->z_utf8 && u8_validate(tnm,
3359 strlen(tnm), NULL, U8_VALIDATE_ENTIRE, &error) < 0) {
3364 if (flags & FIGNORECASE)
3373 * This is to prevent the creation of links into attribute space
3374 * by renaming a linked file into/outof an attribute directory.
3375 * See the comment in zfs_link() for why this is considered bad.
3377 if ((tdzp->z_phys->zp_flags & ZFS_XATTR) !=
3378 (sdzp->z_phys->zp_flags & ZFS_XATTR)) {
3384 * Lock source and target directory entries. To prevent deadlock,
3385 * a lock ordering must be defined. We lock the directory with
3386 * the smallest object id first, or if it's a tie, the one with
3387 * the lexically first name.
3389 if (sdzp->z_id < tdzp->z_id) {
3391 } else if (sdzp->z_id > tdzp->z_id) {
3395 * First compare the two name arguments without
3396 * considering any case folding.
3398 int nofold = (zfsvfs->z_norm & ~U8_TEXTPREP_TOUPPER);
3400 cmp = u8_strcmp(snm, tnm, 0, nofold, U8_UNICODE_LATEST, &error);
3401 ASSERT(error == 0 || !zfsvfs->z_utf8);
3404 * POSIX: "If the old argument and the new argument
3405 * both refer to links to the same existing file,
3406 * the rename() function shall return successfully
3407 * and perform no other action."
3413 * If the file system is case-folding, then we may
3414 * have some more checking to do. A case-folding file
3415 * system is either supporting mixed case sensitivity
3416 * access or is completely case-insensitive. Note
3417 * that the file system is always case preserving.
3419 * In mixed sensitivity mode case sensitive behavior
3420 * is the default. FIGNORECASE must be used to
3421 * explicitly request case insensitive behavior.
3423 * If the source and target names provided differ only
3424 * by case (e.g., a request to rename 'tim' to 'Tim'),
3425 * we will treat this as a special case in the
3426 * case-insensitive mode: as long as the source name
3427 * is an exact match, we will allow this to proceed as
3428 * a name-change request.
3430 if ((zfsvfs->z_case == ZFS_CASE_INSENSITIVE ||
3431 (zfsvfs->z_case == ZFS_CASE_MIXED &&
3432 flags & FIGNORECASE)) &&
3433 u8_strcmp(snm, tnm, 0, zfsvfs->z_norm, U8_UNICODE_LATEST,
3436 * case preserving rename request, require exact
3445 * If the source and destination directories are the same, we should
3446 * grab the z_name_lock of that directory only once.
3450 rw_enter(&sdzp->z_name_lock, RW_READER);
3454 serr = zfs_dirent_lock(&sdl, sdzp, snm, &szp,
3455 ZEXISTS | zflg, NULL, NULL);
3456 terr = zfs_dirent_lock(&tdl,
3457 tdzp, tnm, &tzp, ZRENAMING | zflg, NULL, NULL);
3459 terr = zfs_dirent_lock(&tdl,
3460 tdzp, tnm, &tzp, zflg, NULL, NULL);
3461 serr = zfs_dirent_lock(&sdl,
3462 sdzp, snm, &szp, ZEXISTS | ZRENAMING | zflg,
3468 * Source entry invalid or not there.
3471 zfs_dirent_unlock(tdl);
3477 rw_exit(&sdzp->z_name_lock);
3479 if (strcmp(snm, ".") == 0 || strcmp(snm, "..") == 0)
3485 zfs_dirent_unlock(sdl);
3489 rw_exit(&sdzp->z_name_lock);
3491 if (strcmp(tnm, "..") == 0)
3498 * Must have write access at the source to remove the old entry
3499 * and write access at the target to create the new entry.
3500 * Note that if target and source are the same, this can be
3501 * done in a single check.
3504 if (error = zfs_zaccess_rename(sdzp, szp, tdzp, tzp, cr))
3507 if (ZTOV(szp)->v_type == VDIR) {
3509 * Check to make sure rename is valid.
3510 * Can't do a move like this: /usr/a/b to /usr/a/b/c/d
3512 if (error = zfs_rename_lock(szp, tdzp, sdzp, &zl))
3517 * Does target exist?
3521 * Source and target must be the same type.
3523 if (ZTOV(szp)->v_type == VDIR) {
3524 if (ZTOV(tzp)->v_type != VDIR) {
3529 if (ZTOV(tzp)->v_type == VDIR) {
3535 * POSIX dictates that when the source and target
3536 * entries refer to the same file object, rename
3537 * must do nothing and exit without error.
3539 if (szp->z_id == tzp->z_id) {
3545 vnevent_rename_src(ZTOV(szp), sdvp, snm, ct);
3547 vnevent_rename_dest(ZTOV(tzp), tdvp, tnm, ct);
3550 * notify the target directory if it is not the same
3551 * as source directory.
3554 vnevent_rename_dest_dir(tdvp, ct);
3557 tx = dmu_tx_create(zfsvfs->z_os);
3558 dmu_tx_hold_bonus(tx, szp->z_id); /* nlink changes */
3559 dmu_tx_hold_bonus(tx, sdzp->z_id); /* nlink changes */
3560 dmu_tx_hold_zap(tx, sdzp->z_id, FALSE, snm);
3561 dmu_tx_hold_zap(tx, tdzp->z_id, TRUE, tnm);
3563 dmu_tx_hold_bonus(tx, tdzp->z_id); /* nlink changes */
3565 dmu_tx_hold_bonus(tx, tzp->z_id); /* parent changes */
3566 dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL);
3567 error = dmu_tx_assign(tx, TXG_NOWAIT);
3570 zfs_rename_unlock(&zl);
3571 zfs_dirent_unlock(sdl);
3572 zfs_dirent_unlock(tdl);
3575 rw_exit(&sdzp->z_name_lock);
3580 if (error == ERESTART) {
3590 if (tzp) /* Attempt to remove the existing target */
3591 error = zfs_link_destroy(tdl, tzp, tx, zflg, NULL);
3594 error = zfs_link_create(tdl, szp, tx, ZRENAMING);
3596 szp->z_phys->zp_flags |= ZFS_AV_MODIFIED;
3598 error = zfs_link_destroy(sdl, szp, tx, ZRENAMING, NULL);
3601 zfs_log_rename(zilog, tx,
3602 TX_RENAME | (flags & FIGNORECASE ? TX_CI : 0),
3603 sdzp, sdl->dl_name, tdzp, tdl->dl_name, szp);
3605 /* Update path information for the target vnode */
3606 vn_renamepath(tdvp, ZTOV(szp), tnm, strlen(tnm));
3608 #ifdef FREEBSD_NAMECACHE
3619 zfs_rename_unlock(&zl);
3621 zfs_dirent_unlock(sdl);
3622 zfs_dirent_unlock(tdl);
3625 rw_exit(&sdzp->z_name_lock);
3637 * Insert the indicated symbolic reference entry into the directory.
3639 * IN: dvp - Directory to contain new symbolic link.
3640 * link - Name for new symlink entry.
3641 * vap - Attributes of new entry.
3642 * target - Target path of new symlink.
3643 * cr - credentials of caller.
3644 * ct - caller context
3645 * flags - case flags
3647 * RETURN: 0 if success
3648 * error code if failure
3651 * dvp - ctime|mtime updated
3655 zfs_symlink(vnode_t *dvp, vnode_t **vpp, char *name, vattr_t *vap, char *link,
3656 cred_t *cr, kthread_t *td)
3658 znode_t *zp, *dzp = VTOZ(dvp);
3661 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
3663 int len = strlen(link);
3666 zfs_acl_ids_t acl_ids;
3667 boolean_t fuid_dirtied;
3670 ASSERT(vap->va_type == VLNK);
3674 zilog = zfsvfs->z_log;
3676 if (zfsvfs->z_utf8 && u8_validate(name, strlen(name),
3677 NULL, U8_VALIDATE_ENTIRE, &error) < 0) {
3681 if (flags & FIGNORECASE)
3684 if (error = zfs_zaccess(dzp, ACE_ADD_FILE, 0, B_FALSE, cr)) {
3689 if (len > MAXPATHLEN) {
3691 return (ENAMETOOLONG);
3695 * Attempt to lock directory; fail if entry already exists.
3697 error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg, NULL, NULL);
3703 VERIFY(0 == zfs_acl_ids_create(dzp, 0, vap, cr, NULL, &acl_ids));
3704 if (zfs_acl_ids_overquota(zfsvfs, &acl_ids)) {
3705 zfs_acl_ids_free(&acl_ids);
3706 zfs_dirent_unlock(dl);
3710 tx = dmu_tx_create(zfsvfs->z_os);
3711 fuid_dirtied = zfsvfs->z_fuid_dirty;
3712 dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, MAX(1, len));
3713 dmu_tx_hold_bonus(tx, dzp->z_id);
3714 dmu_tx_hold_zap(tx, dzp->z_id, TRUE, name);
3715 if (acl_ids.z_aclp->z_acl_bytes > ZFS_ACE_SPACE)
3716 dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, SPA_MAXBLOCKSIZE);
3718 zfs_fuid_txhold(zfsvfs, tx);
3719 error = dmu_tx_assign(tx, TXG_NOWAIT);
3721 zfs_acl_ids_free(&acl_ids);
3722 zfs_dirent_unlock(dl);
3723 if (error == ERESTART) {
3733 dmu_buf_will_dirty(dzp->z_dbuf, tx);
3736 * Create a new object for the symlink.
3737 * Put the link content into bonus buffer if it will fit;
3738 * otherwise, store it just like any other file data.
3740 if (sizeof (znode_phys_t) + len <= dmu_bonus_max()) {
3741 zfs_mknode(dzp, vap, tx, cr, 0, &zp, len, &acl_ids);
3743 bcopy(link, zp->z_phys + 1, len);
3747 zfs_mknode(dzp, vap, tx, cr, 0, &zp, 0, &acl_ids);
3750 zfs_fuid_sync(zfsvfs, tx);
3752 * Nothing can access the znode yet so no locking needed
3753 * for growing the znode's blocksize.
3755 zfs_grow_blocksize(zp, len, tx);
3757 VERIFY(0 == dmu_buf_hold(zfsvfs->z_os,
3758 zp->z_id, 0, FTAG, &dbp));
3759 dmu_buf_will_dirty(dbp, tx);
3761 ASSERT3U(len, <=, dbp->db_size);
3762 bcopy(link, dbp->db_data, len);
3763 dmu_buf_rele(dbp, FTAG);
3765 zp->z_phys->zp_size = len;
3768 * Insert the new object into the directory.
3770 (void) zfs_link_create(dl, zp, tx, ZNEW);
3772 uint64_t txtype = TX_SYMLINK;
3773 if (flags & FIGNORECASE)
3775 zfs_log_symlink(zilog, tx, txtype, dzp, zp, name, link);
3779 zfs_acl_ids_free(&acl_ids);
3783 zfs_dirent_unlock(dl);
3790 * Return, in the buffer contained in the provided uio structure,
3791 * the symbolic path referred to by vp.
3793 * IN: vp - vnode of symbolic link.
3794 * uoip - structure to contain the link path.
3795 * cr - credentials of caller.
3796 * ct - caller context
3798 * OUT: uio - structure to contain the link path.
3800 * RETURN: 0 if success
3801 * error code if failure
3804 * vp - atime updated
3808 zfs_readlink(vnode_t *vp, uio_t *uio, cred_t *cr, caller_context_t *ct)
3810 znode_t *zp = VTOZ(vp);
3811 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
3818 bufsz = (size_t)zp->z_phys->zp_size;
3819 if (bufsz + sizeof (znode_phys_t) <= zp->z_dbuf->db_size) {
3820 error = uiomove(zp->z_phys + 1,
3821 MIN((size_t)bufsz, uio->uio_resid), UIO_READ, uio);
3824 error = dmu_buf_hold(zfsvfs->z_os, zp->z_id, 0, FTAG, &dbp);
3829 error = uiomove(dbp->db_data,
3830 MIN((size_t)bufsz, uio->uio_resid), UIO_READ, uio);
3831 dmu_buf_rele(dbp, FTAG);
3834 ZFS_ACCESSTIME_STAMP(zfsvfs, zp);
3840 * Insert a new entry into directory tdvp referencing svp.
3842 * IN: tdvp - Directory to contain new entry.
3843 * svp - vnode of new entry.
3844 * name - name of new entry.
3845 * cr - credentials of caller.
3846 * ct - caller context
3848 * RETURN: 0 if success
3849 * error code if failure
3852 * tdvp - ctime|mtime updated
3853 * svp - ctime updated
3857 zfs_link(vnode_t *tdvp, vnode_t *svp, char *name, cred_t *cr,
3858 caller_context_t *ct, int flags)
3860 znode_t *dzp = VTOZ(tdvp);
3862 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
3872 ASSERT(tdvp->v_type == VDIR);
3876 zilog = zfsvfs->z_log;
3878 if (VOP_REALVP(svp, &realvp, ct) == 0)
3882 * POSIX dictates that we return EPERM here.
3883 * Better choices include ENOTSUP or EISDIR.
3885 if (svp->v_type == VDIR) {
3890 if (svp->v_vfsp != tdvp->v_vfsp || zfsctl_is_node(svp)) {
3898 /* Prevent links to .zfs/shares files */
3900 if (szp->z_phys->zp_parent == zfsvfs->z_shares_dir) {
3905 if (zfsvfs->z_utf8 && u8_validate(name,
3906 strlen(name), NULL, U8_VALIDATE_ENTIRE, &error) < 0) {
3910 if (flags & FIGNORECASE)
3914 * We do not support links between attributes and non-attributes
3915 * because of the potential security risk of creating links
3916 * into "normal" file space in order to circumvent restrictions
3917 * imposed in attribute space.
3919 if ((szp->z_phys->zp_flags & ZFS_XATTR) !=
3920 (dzp->z_phys->zp_flags & ZFS_XATTR)) {
3926 owner = zfs_fuid_map_id(zfsvfs, szp->z_phys->zp_uid, cr, ZFS_OWNER);
3927 if (owner != crgetuid(cr) &&
3928 secpolicy_basic_link(svp, cr) != 0) {
3933 if (error = zfs_zaccess(dzp, ACE_ADD_FILE, 0, B_FALSE, cr)) {
3940 * Attempt to lock directory; fail if entry already exists.
3942 error = zfs_dirent_lock(&dl, dzp, name, &tzp, zf, NULL, NULL);
3948 tx = dmu_tx_create(zfsvfs->z_os);
3949 dmu_tx_hold_bonus(tx, szp->z_id);
3950 dmu_tx_hold_zap(tx, dzp->z_id, TRUE, name);
3951 error = dmu_tx_assign(tx, TXG_NOWAIT);
3953 zfs_dirent_unlock(dl);
3954 if (error == ERESTART) {
3964 error = zfs_link_create(dl, szp, tx, 0);
3967 uint64_t txtype = TX_LINK;
3968 if (flags & FIGNORECASE)
3970 zfs_log_link(zilog, tx, txtype, dzp, szp, name);
3975 zfs_dirent_unlock(dl);
3978 vnevent_link(svp, ct);
3987 zfs_inactive(vnode_t *vp, cred_t *cr, caller_context_t *ct)
3989 znode_t *zp = VTOZ(vp);
3990 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
3993 rw_enter(&zfsvfs->z_teardown_inactive_lock, RW_READER);
3994 if (zp->z_dbuf == NULL) {
3996 * The fs has been unmounted, or we did a
3997 * suspend/resume and this file no longer exists.
4000 vp->v_count = 0; /* count arrives as 1 */
4002 vrecycle(vp, curthread);
4003 rw_exit(&zfsvfs->z_teardown_inactive_lock);
4007 if (zp->z_atime_dirty && zp->z_unlinked == 0) {
4008 dmu_tx_t *tx = dmu_tx_create(zfsvfs->z_os);
4010 dmu_tx_hold_bonus(tx, zp->z_id);
4011 error = dmu_tx_assign(tx, TXG_WAIT);
4015 dmu_buf_will_dirty(zp->z_dbuf, tx);
4016 mutex_enter(&zp->z_lock);
4017 zp->z_atime_dirty = 0;
4018 mutex_exit(&zp->z_lock);
4024 rw_exit(&zfsvfs->z_teardown_inactive_lock);
4027 CTASSERT(sizeof(struct zfid_short) <= sizeof(struct fid));
4028 CTASSERT(sizeof(struct zfid_long) <= sizeof(struct fid));
4032 zfs_fid(vnode_t *vp, fid_t *fidp, caller_context_t *ct)
4034 znode_t *zp = VTOZ(vp);
4035 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
4037 uint64_t object = zp->z_id;
4043 gen = (uint32_t)zp->z_gen;
4045 size = (zfsvfs->z_parent != zfsvfs) ? LONG_FID_LEN : SHORT_FID_LEN;
4046 fidp->fid_len = size;
4048 zfid = (zfid_short_t *)fidp;
4050 zfid->zf_len = size;
4052 for (i = 0; i < sizeof (zfid->zf_object); i++)
4053 zfid->zf_object[i] = (uint8_t)(object >> (8 * i));
4055 /* Must have a non-zero generation number to distinguish from .zfs */
4058 for (i = 0; i < sizeof (zfid->zf_gen); i++)
4059 zfid->zf_gen[i] = (uint8_t)(gen >> (8 * i));
4061 if (size == LONG_FID_LEN) {
4062 uint64_t objsetid = dmu_objset_id(zfsvfs->z_os);
4065 zlfid = (zfid_long_t *)fidp;
4067 for (i = 0; i < sizeof (zlfid->zf_setid); i++)
4068 zlfid->zf_setid[i] = (uint8_t)(objsetid >> (8 * i));
4070 /* XXX - this should be the generation number for the objset */
4071 for (i = 0; i < sizeof (zlfid->zf_setgen); i++)
4072 zlfid->zf_setgen[i] = 0;
4080 zfs_pathconf(vnode_t *vp, int cmd, ulong_t *valp, cred_t *cr,
4081 caller_context_t *ct)
4093 case _PC_FILESIZEBITS:
4098 case _PC_XATTR_EXISTS:
4100 zfsvfs = zp->z_zfsvfs;
4104 error = zfs_dirent_lock(&dl, zp, "", &xzp,
4105 ZXATTR | ZEXISTS | ZSHARED, NULL, NULL);
4107 zfs_dirent_unlock(dl);
4108 if (!zfs_dirempty(xzp))
4111 } else if (error == ENOENT) {
4113 * If there aren't extended attributes, it's the
4114 * same as having zero of them.
4122 case _PC_ACL_EXTENDED:
4130 case _PC_ACL_PATH_MAX:
4131 *valp = ACL_MAX_ENTRIES;
4134 case _PC_MIN_HOLE_SIZE:
4135 *valp = (int)SPA_MINBLOCKSIZE;
4139 return (EOPNOTSUPP);
4145 zfs_getsecattr(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_getacl(zp, vsecp, skipaclchk, cr);
4163 zfs_setsecattr(vnode_t *vp, vsecattr_t *vsecp, int flag, cred_t *cr,
4164 caller_context_t *ct)
4166 znode_t *zp = VTOZ(vp);
4167 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
4169 boolean_t skipaclchk = (flag & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE;
4173 error = zfs_setacl(zp, vsecp, skipaclchk, cr);
4179 zfs_getpages(struct vnode *vp, vm_page_t *m, int count, int reqpage)
4181 znode_t *zp = VTOZ(vp);
4182 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
4183 objset_t *os = zp->z_zfsvfs->z_os;
4194 pcount = round_page(count) / PAGE_SIZE;
4196 object = mreq->object;
4199 KASSERT(vp->v_object == object, ("mismatching object"));
4201 VM_OBJECT_LOCK(object);
4202 vm_page_lock_queues();
4203 for (i = 0; i < pcount; i++) {
4208 vm_page_unlock_queues();
4211 if (mreq->valid != VM_PAGE_BITS_ALL)
4212 vm_page_zero_invalid(mreq, TRUE);
4213 VM_OBJECT_UNLOCK(object);
4215 return (VM_PAGER_OK);
4218 PCPU_INC(cnt.v_vnodein);
4219 PCPU_INC(cnt.v_vnodepgsin);
4221 if (IDX_TO_OFF(mreq->pindex) >= object->un_pager.vnp.vnp_size) {
4222 VM_OBJECT_UNLOCK(object);
4224 return (VM_PAGER_BAD);
4228 if (IDX_TO_OFF(mreq->pindex) + size > object->un_pager.vnp.vnp_size)
4229 size = object->un_pager.vnp.vnp_size - IDX_TO_OFF(mreq->pindex);
4231 VM_OBJECT_UNLOCK(object);
4233 va = zfs_map_page(mreq, &sf);
4234 error = dmu_read(os, zp->z_id, IDX_TO_OFF(mreq->pindex),
4235 size, va, DMU_READ_PREFETCH);
4236 if (size != PAGE_SIZE)
4237 bzero(va + size, PAGE_SIZE - size);
4240 VM_OBJECT_LOCK(object);
4243 mreq->valid = VM_PAGE_BITS_ALL;
4244 KASSERT(mreq->dirty == 0, ("zfs_getpages: page %p is dirty", mreq));
4246 VM_OBJECT_UNLOCK(object);
4248 ZFS_ACCESSTIME_STAMP(zfsvfs, zp);
4250 return (error ? VM_PAGER_ERROR : VM_PAGER_OK);
4254 zfs_freebsd_getpages(ap)
4255 struct vop_getpages_args /* {
4260 vm_ooffset_t a_offset;
4264 return (zfs_getpages(ap->a_vp, ap->a_m, ap->a_count, ap->a_reqpage));
4268 zfs_freebsd_open(ap)
4269 struct vop_open_args /* {
4272 struct ucred *a_cred;
4273 struct thread *a_td;
4276 vnode_t *vp = ap->a_vp;
4277 znode_t *zp = VTOZ(vp);
4280 error = zfs_open(&vp, ap->a_mode, ap->a_cred, NULL);
4282 vnode_create_vobject(vp, zp->z_phys->zp_size, ap->a_td);
4287 zfs_freebsd_close(ap)
4288 struct vop_close_args /* {
4291 struct ucred *a_cred;
4292 struct thread *a_td;
4296 return (zfs_close(ap->a_vp, ap->a_fflag, 0, 0, ap->a_cred, NULL));
4300 zfs_freebsd_ioctl(ap)
4301 struct vop_ioctl_args /* {
4311 return (zfs_ioctl(ap->a_vp, ap->a_command, (intptr_t)ap->a_data,
4312 ap->a_fflag, ap->a_cred, NULL, NULL));
4316 zfs_freebsd_read(ap)
4317 struct vop_read_args /* {
4321 struct ucred *a_cred;
4325 return (zfs_read(ap->a_vp, ap->a_uio, ap->a_ioflag, ap->a_cred, NULL));
4329 zfs_freebsd_write(ap)
4330 struct vop_write_args /* {
4334 struct ucred *a_cred;
4338 return (zfs_write(ap->a_vp, ap->a_uio, ap->a_ioflag, ap->a_cred, NULL));
4342 zfs_freebsd_access(ap)
4343 struct vop_access_args /* {
4345 accmode_t a_accmode;
4346 struct ucred *a_cred;
4347 struct thread *a_td;
4354 * ZFS itself only knowns about VREAD, VWRITE, VEXEC and VAPPEND,
4356 accmode = ap->a_accmode & (VREAD|VWRITE|VEXEC|VAPPEND);
4358 error = zfs_access(ap->a_vp, accmode, 0, ap->a_cred, NULL);
4361 * VADMIN has to be handled by vaccess().
4364 accmode = ap->a_accmode & ~(VREAD|VWRITE|VEXEC|VAPPEND);
4366 vnode_t *vp = ap->a_vp;
4367 znode_t *zp = VTOZ(vp);
4368 znode_phys_t *zphys = zp->z_phys;
4370 error = vaccess(vp->v_type, zphys->zp_mode,
4371 zphys->zp_uid, zphys->zp_gid, accmode, ap->a_cred,
4380 zfs_freebsd_lookup(ap)
4381 struct vop_lookup_args /* {
4382 struct vnode *a_dvp;
4383 struct vnode **a_vpp;
4384 struct componentname *a_cnp;
4387 struct componentname *cnp = ap->a_cnp;
4388 char nm[NAME_MAX + 1];
4390 ASSERT(cnp->cn_namelen < sizeof(nm));
4391 strlcpy(nm, cnp->cn_nameptr, MIN(cnp->cn_namelen + 1, sizeof(nm)));
4393 return (zfs_lookup(ap->a_dvp, nm, ap->a_vpp, cnp, cnp->cn_nameiop,
4394 cnp->cn_cred, cnp->cn_thread, 0));
4398 zfs_freebsd_create(ap)
4399 struct vop_create_args /* {
4400 struct vnode *a_dvp;
4401 struct vnode **a_vpp;
4402 struct componentname *a_cnp;
4403 struct vattr *a_vap;
4406 struct componentname *cnp = ap->a_cnp;
4407 vattr_t *vap = ap->a_vap;
4410 ASSERT(cnp->cn_flags & SAVENAME);
4412 vattr_init_mask(vap);
4413 mode = vap->va_mode & ALLPERMS;
4415 return (zfs_create(ap->a_dvp, cnp->cn_nameptr, vap, !EXCL, mode,
4416 ap->a_vpp, cnp->cn_cred, cnp->cn_thread));
4420 zfs_freebsd_remove(ap)
4421 struct vop_remove_args /* {
4422 struct vnode *a_dvp;
4424 struct componentname *a_cnp;
4428 ASSERT(ap->a_cnp->cn_flags & SAVENAME);
4430 return (zfs_remove(ap->a_dvp, ap->a_cnp->cn_nameptr,
4431 ap->a_cnp->cn_cred, NULL, 0));
4435 zfs_freebsd_mkdir(ap)
4436 struct vop_mkdir_args /* {
4437 struct vnode *a_dvp;
4438 struct vnode **a_vpp;
4439 struct componentname *a_cnp;
4440 struct vattr *a_vap;
4443 vattr_t *vap = ap->a_vap;
4445 ASSERT(ap->a_cnp->cn_flags & SAVENAME);
4447 vattr_init_mask(vap);
4449 return (zfs_mkdir(ap->a_dvp, ap->a_cnp->cn_nameptr, vap, ap->a_vpp,
4450 ap->a_cnp->cn_cred, NULL, 0, NULL));
4454 zfs_freebsd_rmdir(ap)
4455 struct vop_rmdir_args /* {
4456 struct vnode *a_dvp;
4458 struct componentname *a_cnp;
4461 struct componentname *cnp = ap->a_cnp;
4463 ASSERT(cnp->cn_flags & SAVENAME);
4465 return (zfs_rmdir(ap->a_dvp, cnp->cn_nameptr, NULL, cnp->cn_cred, NULL, 0));
4469 zfs_freebsd_readdir(ap)
4470 struct vop_readdir_args /* {
4473 struct ucred *a_cred;
4480 return (zfs_readdir(ap->a_vp, ap->a_uio, ap->a_cred, ap->a_eofflag,
4481 ap->a_ncookies, ap->a_cookies));
4485 zfs_freebsd_fsync(ap)
4486 struct vop_fsync_args /* {
4489 struct thread *a_td;
4494 return (zfs_fsync(ap->a_vp, 0, ap->a_td->td_ucred, NULL));
4498 zfs_freebsd_getattr(ap)
4499 struct vop_getattr_args /* {
4501 struct vattr *a_vap;
4502 struct ucred *a_cred;
4503 struct thread *a_td;
4506 vattr_t *vap = ap->a_vap;
4512 xvap.xva_vattr = *vap;
4513 xvap.xva_vattr.va_mask |= AT_XVATTR;
4515 /* Convert chflags into ZFS-type flags. */
4516 /* XXX: what about SF_SETTABLE?. */
4517 XVA_SET_REQ(&xvap, XAT_IMMUTABLE);
4518 XVA_SET_REQ(&xvap, XAT_APPENDONLY);
4519 XVA_SET_REQ(&xvap, XAT_NOUNLINK);
4520 XVA_SET_REQ(&xvap, XAT_NODUMP);
4521 error = zfs_getattr(ap->a_vp, (vattr_t *)&xvap, 0, ap->a_cred, NULL);
4525 /* Convert ZFS xattr into chflags. */
4526 #define FLAG_CHECK(fflag, xflag, xfield) do { \
4527 if (XVA_ISSET_RTN(&xvap, (xflag)) && (xfield) != 0) \
4528 fflags |= (fflag); \
4530 FLAG_CHECK(SF_IMMUTABLE, XAT_IMMUTABLE,
4531 xvap.xva_xoptattrs.xoa_immutable);
4532 FLAG_CHECK(SF_APPEND, XAT_APPENDONLY,
4533 xvap.xva_xoptattrs.xoa_appendonly);
4534 FLAG_CHECK(SF_NOUNLINK, XAT_NOUNLINK,
4535 xvap.xva_xoptattrs.xoa_nounlink);
4536 FLAG_CHECK(UF_NODUMP, XAT_NODUMP,
4537 xvap.xva_xoptattrs.xoa_nodump);
4539 *vap = xvap.xva_vattr;
4540 vap->va_flags = fflags;
4545 zfs_freebsd_setattr(ap)
4546 struct vop_setattr_args /* {
4548 struct vattr *a_vap;
4549 struct ucred *a_cred;
4550 struct thread *a_td;
4553 vnode_t *vp = ap->a_vp;
4554 vattr_t *vap = ap->a_vap;
4555 cred_t *cred = ap->a_cred;
4560 vattr_init_mask(vap);
4561 vap->va_mask &= ~AT_NOSET;
4564 xvap.xva_vattr = *vap;
4566 zflags = VTOZ(vp)->z_phys->zp_flags;
4568 if (vap->va_flags != VNOVAL) {
4569 zfsvfs_t *zfsvfs = VTOZ(vp)->z_zfsvfs;
4572 if (zfsvfs->z_use_fuids == B_FALSE)
4573 return (EOPNOTSUPP);
4575 fflags = vap->va_flags;
4576 if ((fflags & ~(SF_IMMUTABLE|SF_APPEND|SF_NOUNLINK|UF_NODUMP)) != 0)
4577 return (EOPNOTSUPP);
4579 * Unprivileged processes are not permitted to unset system
4580 * flags, or modify flags if any system flags are set.
4581 * Privileged non-jail processes may not modify system flags
4582 * if securelevel > 0 and any existing system flags are set.
4583 * Privileged jail processes behave like privileged non-jail
4584 * processes if the security.jail.chflags_allowed sysctl is
4585 * is non-zero; otherwise, they behave like unprivileged
4588 if (secpolicy_fs_owner(vp->v_mount, cred) == 0 ||
4589 priv_check_cred(cred, PRIV_VFS_SYSFLAGS, 0) == 0) {
4591 (ZFS_IMMUTABLE | ZFS_APPENDONLY | ZFS_NOUNLINK)) {
4592 error = securelevel_gt(cred, 0);
4598 * Callers may only modify the file flags on objects they
4599 * have VADMIN rights for.
4601 if ((error = VOP_ACCESS(vp, VADMIN, cred, curthread)) != 0)
4604 (ZFS_IMMUTABLE | ZFS_APPENDONLY | ZFS_NOUNLINK)) {
4608 (SF_IMMUTABLE | SF_APPEND | SF_NOUNLINK)) {
4613 #define FLAG_CHANGE(fflag, zflag, xflag, xfield) do { \
4614 if (((fflags & (fflag)) && !(zflags & (zflag))) || \
4615 ((zflags & (zflag)) && !(fflags & (fflag)))) { \
4616 XVA_SET_REQ(&xvap, (xflag)); \
4617 (xfield) = ((fflags & (fflag)) != 0); \
4620 /* Convert chflags into ZFS-type flags. */
4621 /* XXX: what about SF_SETTABLE?. */
4622 FLAG_CHANGE(SF_IMMUTABLE, ZFS_IMMUTABLE, XAT_IMMUTABLE,
4623 xvap.xva_xoptattrs.xoa_immutable);
4624 FLAG_CHANGE(SF_APPEND, ZFS_APPENDONLY, XAT_APPENDONLY,
4625 xvap.xva_xoptattrs.xoa_appendonly);
4626 FLAG_CHANGE(SF_NOUNLINK, ZFS_NOUNLINK, XAT_NOUNLINK,
4627 xvap.xva_xoptattrs.xoa_nounlink);
4628 FLAG_CHANGE(UF_NODUMP, ZFS_NODUMP, XAT_NODUMP,
4629 xvap.xva_xoptattrs.xoa_nodump);
4632 return (zfs_setattr(vp, (vattr_t *)&xvap, 0, cred, NULL));
4636 zfs_freebsd_rename(ap)
4637 struct vop_rename_args /* {
4638 struct vnode *a_fdvp;
4639 struct vnode *a_fvp;
4640 struct componentname *a_fcnp;
4641 struct vnode *a_tdvp;
4642 struct vnode *a_tvp;
4643 struct componentname *a_tcnp;
4646 vnode_t *fdvp = ap->a_fdvp;
4647 vnode_t *fvp = ap->a_fvp;
4648 vnode_t *tdvp = ap->a_tdvp;
4649 vnode_t *tvp = ap->a_tvp;
4652 ASSERT(ap->a_fcnp->cn_flags & (SAVENAME|SAVESTART));
4653 ASSERT(ap->a_tcnp->cn_flags & (SAVENAME|SAVESTART));
4655 error = zfs_rename(fdvp, ap->a_fcnp->cn_nameptr, tdvp,
4656 ap->a_tcnp->cn_nameptr, ap->a_fcnp->cn_cred, NULL, 0);
4671 zfs_freebsd_symlink(ap)
4672 struct vop_symlink_args /* {
4673 struct vnode *a_dvp;
4674 struct vnode **a_vpp;
4675 struct componentname *a_cnp;
4676 struct vattr *a_vap;
4680 struct componentname *cnp = ap->a_cnp;
4681 vattr_t *vap = ap->a_vap;
4683 ASSERT(cnp->cn_flags & SAVENAME);
4685 vap->va_type = VLNK; /* FreeBSD: Syscall only sets va_mode. */
4686 vattr_init_mask(vap);
4688 return (zfs_symlink(ap->a_dvp, ap->a_vpp, cnp->cn_nameptr, vap,
4689 ap->a_target, cnp->cn_cred, cnp->cn_thread));
4693 zfs_freebsd_readlink(ap)
4694 struct vop_readlink_args /* {
4697 struct ucred *a_cred;
4701 return (zfs_readlink(ap->a_vp, ap->a_uio, ap->a_cred, NULL));
4705 zfs_freebsd_link(ap)
4706 struct vop_link_args /* {
4707 struct vnode *a_tdvp;
4709 struct componentname *a_cnp;
4712 struct componentname *cnp = ap->a_cnp;
4714 ASSERT(cnp->cn_flags & SAVENAME);
4716 return (zfs_link(ap->a_tdvp, ap->a_vp, cnp->cn_nameptr, cnp->cn_cred, NULL, 0));
4720 zfs_freebsd_inactive(ap)
4721 struct vop_inactive_args /* {
4723 struct thread *a_td;
4726 vnode_t *vp = ap->a_vp;
4728 zfs_inactive(vp, ap->a_td->td_ucred, NULL);
4733 zfs_reclaim_complete(void *arg, int pending)
4736 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
4738 rw_enter(&zfsvfs->z_teardown_inactive_lock, RW_READER);
4739 if (zp->z_dbuf != NULL) {
4740 ZFS_OBJ_HOLD_ENTER(zfsvfs, zp->z_id);
4741 zfs_znode_dmu_fini(zp);
4742 ZFS_OBJ_HOLD_EXIT(zfsvfs, zp->z_id);
4745 rw_exit(&zfsvfs->z_teardown_inactive_lock);
4747 * If the file system is being unmounted, there is a process waiting
4748 * for us, wake it up.
4750 if (zfsvfs->z_unmounted)
4755 zfs_freebsd_reclaim(ap)
4756 struct vop_reclaim_args /* {
4758 struct thread *a_td;
4761 vnode_t *vp = ap->a_vp;
4762 znode_t *zp = VTOZ(vp);
4763 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
4765 rw_enter(&zfsvfs->z_teardown_inactive_lock, RW_READER);
4770 * Destroy the vm object and flush associated pages.
4772 vnode_destroy_vobject(vp);
4774 mutex_enter(&zp->z_lock);
4775 ASSERT(zp->z_phys != NULL);
4777 mutex_exit(&zp->z_lock);
4781 else if (zp->z_dbuf == NULL)
4783 else /* if (!zp->z_unlinked && zp->z_dbuf != NULL) */ {
4786 locked = MUTEX_HELD(ZFS_OBJ_MUTEX(zfsvfs, zp->z_id)) ? 2 :
4787 ZFS_OBJ_HOLD_TRYENTER(zfsvfs, zp->z_id);
4790 * Lock can't be obtained due to deadlock possibility,
4791 * so defer znode destruction.
4793 TASK_INIT(&zp->z_task, 0, zfs_reclaim_complete, zp);
4794 taskqueue_enqueue(taskqueue_thread, &zp->z_task);
4796 zfs_znode_dmu_fini(zp);
4798 ZFS_OBJ_HOLD_EXIT(zfsvfs, zp->z_id);
4804 ASSERT(vp->v_holdcnt >= 1);
4806 rw_exit(&zfsvfs->z_teardown_inactive_lock);
4812 struct vop_fid_args /* {
4818 return (zfs_fid(ap->a_vp, (void *)ap->a_fid, NULL));
4822 zfs_freebsd_pathconf(ap)
4823 struct vop_pathconf_args /* {
4826 register_t *a_retval;
4832 error = zfs_pathconf(ap->a_vp, ap->a_name, &val, curthread->td_ucred, NULL);
4834 *ap->a_retval = val;
4835 else if (error == EOPNOTSUPP)
4836 error = vop_stdpathconf(ap);
4841 zfs_freebsd_fifo_pathconf(ap)
4842 struct vop_pathconf_args /* {
4845 register_t *a_retval;
4849 switch (ap->a_name) {
4850 case _PC_ACL_EXTENDED:
4852 case _PC_ACL_PATH_MAX:
4853 case _PC_MAC_PRESENT:
4854 return (zfs_freebsd_pathconf(ap));
4856 return (fifo_specops.vop_pathconf(ap));
4861 * FreeBSD's extended attributes namespace defines file name prefix for ZFS'
4862 * extended attribute name:
4865 * system freebsd:system:
4866 * user (none, can be used to access ZFS fsattr(5) attributes
4867 * created on Solaris)
4870 zfs_create_attrname(int attrnamespace, const char *name, char *attrname,
4873 const char *namespace, *prefix, *suffix;
4875 /* We don't allow '/' character in attribute name. */
4876 if (strchr(name, '/') != NULL)
4878 /* We don't allow attribute names that start with "freebsd:" string. */
4879 if (strncmp(name, "freebsd:", 8) == 0)
4882 bzero(attrname, size);
4884 switch (attrnamespace) {
4885 case EXTATTR_NAMESPACE_USER:
4887 prefix = "freebsd:";
4888 namespace = EXTATTR_NAMESPACE_USER_STRING;
4892 * This is the default namespace by which we can access all
4893 * attributes created on Solaris.
4895 prefix = namespace = suffix = "";
4898 case EXTATTR_NAMESPACE_SYSTEM:
4899 prefix = "freebsd:";
4900 namespace = EXTATTR_NAMESPACE_SYSTEM_STRING;
4903 case EXTATTR_NAMESPACE_EMPTY:
4907 if (snprintf(attrname, size, "%s%s%s%s", prefix, namespace, suffix,
4909 return (ENAMETOOLONG);
4915 * Vnode operating to retrieve a named extended attribute.
4918 zfs_getextattr(struct vop_getextattr_args *ap)
4921 IN struct vnode *a_vp;
4922 IN int a_attrnamespace;
4923 IN const char *a_name;
4924 INOUT struct uio *a_uio;
4926 IN struct ucred *a_cred;
4927 IN struct thread *a_td;
4931 zfsvfs_t *zfsvfs = VTOZ(ap->a_vp)->z_zfsvfs;
4932 struct thread *td = ap->a_td;
4933 struct nameidata nd;
4936 vnode_t *xvp = NULL, *vp;
4939 error = extattr_check_cred(ap->a_vp, ap->a_attrnamespace,
4940 ap->a_cred, ap->a_td, VREAD);
4944 error = zfs_create_attrname(ap->a_attrnamespace, ap->a_name, attrname,
4951 error = zfs_lookup(ap->a_vp, NULL, &xvp, NULL, 0, ap->a_cred, td,
4959 NDINIT_ATVP(&nd, LOOKUP, NOFOLLOW | MPSAFE, UIO_SYSSPACE, attrname,
4961 error = vn_open_cred(&nd, &flags, 0, 0, ap->a_cred, NULL);
4963 NDFREE(&nd, NDF_ONLY_PNBUF);
4966 if (error == ENOENT)
4971 if (ap->a_size != NULL) {
4972 error = VOP_GETATTR(vp, &va, ap->a_cred);
4974 *ap->a_size = (size_t)va.va_size;
4975 } else if (ap->a_uio != NULL)
4976 error = VOP_READ(vp, ap->a_uio, IO_UNIT | IO_SYNC, ap->a_cred);
4979 vn_close(vp, flags, ap->a_cred, td);
4986 * Vnode operation to remove a named attribute.
4989 zfs_deleteextattr(struct vop_deleteextattr_args *ap)
4992 IN struct vnode *a_vp;
4993 IN int a_attrnamespace;
4994 IN const char *a_name;
4995 IN struct ucred *a_cred;
4996 IN struct thread *a_td;
5000 zfsvfs_t *zfsvfs = VTOZ(ap->a_vp)->z_zfsvfs;
5001 struct thread *td = ap->a_td;
5002 struct nameidata nd;
5005 vnode_t *xvp = NULL, *vp;
5008 error = extattr_check_cred(ap->a_vp, ap->a_attrnamespace,
5009 ap->a_cred, ap->a_td, VWRITE);
5013 error = zfs_create_attrname(ap->a_attrnamespace, ap->a_name, attrname,
5020 error = zfs_lookup(ap->a_vp, NULL, &xvp, NULL, 0, ap->a_cred, td,
5027 NDINIT_ATVP(&nd, DELETE, NOFOLLOW | LOCKPARENT | LOCKLEAF | MPSAFE,
5028 UIO_SYSSPACE, attrname, xvp, td);
5031 NDFREE(&nd, NDF_ONLY_PNBUF);
5034 if (error == ENOENT)
5038 error = VOP_REMOVE(nd.ni_dvp, vp, &nd.ni_cnd);
5041 if (vp == nd.ni_dvp)
5051 * Vnode operation to set a named attribute.
5054 zfs_setextattr(struct vop_setextattr_args *ap)
5057 IN struct vnode *a_vp;
5058 IN int a_attrnamespace;
5059 IN const char *a_name;
5060 INOUT struct uio *a_uio;
5061 IN struct ucred *a_cred;
5062 IN struct thread *a_td;
5066 zfsvfs_t *zfsvfs = VTOZ(ap->a_vp)->z_zfsvfs;
5067 struct thread *td = ap->a_td;
5068 struct nameidata nd;
5071 vnode_t *xvp = NULL, *vp;
5074 error = extattr_check_cred(ap->a_vp, ap->a_attrnamespace,
5075 ap->a_cred, ap->a_td, VWRITE);
5079 error = zfs_create_attrname(ap->a_attrnamespace, ap->a_name, attrname,
5086 error = zfs_lookup(ap->a_vp, NULL, &xvp, NULL, 0, ap->a_cred, td,
5087 LOOKUP_XATTR | CREATE_XATTR_DIR);
5093 flags = FFLAGS(O_WRONLY | O_CREAT);
5094 NDINIT_ATVP(&nd, LOOKUP, NOFOLLOW | MPSAFE, UIO_SYSSPACE, attrname,
5096 error = vn_open_cred(&nd, &flags, 0600, 0, ap->a_cred, NULL);
5098 NDFREE(&nd, NDF_ONLY_PNBUF);
5106 error = VOP_SETATTR(vp, &va, ap->a_cred);
5108 VOP_WRITE(vp, ap->a_uio, IO_UNIT | IO_SYNC, ap->a_cred);
5111 vn_close(vp, flags, ap->a_cred, td);
5118 * Vnode operation to retrieve extended attributes on a vnode.
5121 zfs_listextattr(struct vop_listextattr_args *ap)
5124 IN struct vnode *a_vp;
5125 IN int a_attrnamespace;
5126 INOUT struct uio *a_uio;
5128 IN struct ucred *a_cred;
5129 IN struct thread *a_td;
5133 zfsvfs_t *zfsvfs = VTOZ(ap->a_vp)->z_zfsvfs;
5134 struct thread *td = ap->a_td;
5135 struct nameidata nd;
5136 char attrprefix[16];
5137 u_char dirbuf[sizeof(struct dirent)];
5140 struct uio auio, *uio = ap->a_uio;
5141 size_t *sizep = ap->a_size;
5143 vnode_t *xvp = NULL, *vp;
5144 int done, error, eof, pos;
5146 error = extattr_check_cred(ap->a_vp, ap->a_attrnamespace,
5147 ap->a_cred, ap->a_td, VREAD);
5151 error = zfs_create_attrname(ap->a_attrnamespace, "", attrprefix,
5152 sizeof(attrprefix));
5155 plen = strlen(attrprefix);
5162 error = zfs_lookup(ap->a_vp, NULL, &xvp, NULL, 0, ap->a_cred, td,
5167 * ENOATTR means that the EA directory does not yet exist,
5168 * i.e. there are no extended attributes there.
5170 if (error == ENOATTR)
5175 NDINIT_ATVP(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | LOCKSHARED | MPSAFE,
5176 UIO_SYSSPACE, ".", xvp, td);
5179 NDFREE(&nd, NDF_ONLY_PNBUF);
5185 auio.uio_iov = &aiov;
5186 auio.uio_iovcnt = 1;
5187 auio.uio_segflg = UIO_SYSSPACE;
5189 auio.uio_rw = UIO_READ;
5190 auio.uio_offset = 0;
5195 aiov.iov_base = (void *)dirbuf;
5196 aiov.iov_len = sizeof(dirbuf);
5197 auio.uio_resid = sizeof(dirbuf);
5198 error = VOP_READDIR(vp, &auio, ap->a_cred, &eof, NULL, NULL);
5199 done = sizeof(dirbuf) - auio.uio_resid;
5202 for (pos = 0; pos < done;) {
5203 dp = (struct dirent *)(dirbuf + pos);
5204 pos += dp->d_reclen;
5206 * XXX: Temporarily we also accept DT_UNKNOWN, as this
5207 * is what we get when attribute was created on Solaris.
5209 if (dp->d_type != DT_REG && dp->d_type != DT_UNKNOWN)
5211 if (plen == 0 && strncmp(dp->d_name, "freebsd:", 8) == 0)
5213 else if (strncmp(dp->d_name, attrprefix, plen) != 0)
5215 nlen = dp->d_namlen - plen;
5218 else if (uio != NULL) {
5220 * Format of extattr name entry is one byte for
5221 * length and the rest for name.
5223 error = uiomove(&nlen, 1, uio->uio_rw, uio);
5225 error = uiomove(dp->d_name + plen, nlen,
5232 } while (!eof && error == 0);
5241 zfs_freebsd_getacl(ap)
5242 struct vop_getacl_args /* {
5251 vsecattr_t vsecattr;
5253 if (ap->a_type != ACL_TYPE_NFS4)
5256 vsecattr.vsa_mask = VSA_ACE | VSA_ACECNT;
5257 if (error = zfs_getsecattr(ap->a_vp, &vsecattr, 0, ap->a_cred, NULL))
5260 error = acl_from_aces(ap->a_aclp, vsecattr.vsa_aclentp, vsecattr.vsa_aclcnt);
5261 if (vsecattr.vsa_aclentp != NULL)
5262 kmem_free(vsecattr.vsa_aclentp, vsecattr.vsa_aclentsz);
5268 zfs_freebsd_setacl(ap)
5269 struct vop_setacl_args /* {
5278 vsecattr_t vsecattr;
5279 int aclbsize; /* size of acl list in bytes */
5282 if (ap->a_type != ACL_TYPE_NFS4)
5285 if (ap->a_aclp->acl_cnt < 1 || ap->a_aclp->acl_cnt > MAX_ACL_ENTRIES)
5289 * With NFSv4 ACLs, chmod(2) may need to add additional entries,
5290 * splitting every entry into two and appending "canonical six"
5291 * entries at the end. Don't allow for setting an ACL that would
5292 * cause chmod(2) to run out of ACL entries.
5294 if (ap->a_aclp->acl_cnt * 2 + 6 > ACL_MAX_ENTRIES)
5297 error = acl_nfs4_check(ap->a_aclp, ap->a_vp->v_type == VDIR);
5301 vsecattr.vsa_mask = VSA_ACE;
5302 aclbsize = ap->a_aclp->acl_cnt * sizeof(ace_t);
5303 vsecattr.vsa_aclentp = kmem_alloc(aclbsize, KM_SLEEP);
5304 aaclp = vsecattr.vsa_aclentp;
5305 vsecattr.vsa_aclentsz = aclbsize;
5307 aces_from_acl(vsecattr.vsa_aclentp, &vsecattr.vsa_aclcnt, ap->a_aclp);
5308 error = zfs_setsecattr(ap->a_vp, &vsecattr, 0, ap->a_cred, NULL);
5309 kmem_free(aaclp, aclbsize);
5315 zfs_freebsd_aclcheck(ap)
5316 struct vop_aclcheck_args /* {
5325 return (EOPNOTSUPP);
5328 struct vop_vector zfs_vnodeops;
5329 struct vop_vector zfs_fifoops;
5330 struct vop_vector zfs_shareops;
5332 struct vop_vector zfs_vnodeops = {
5333 .vop_default = &default_vnodeops,
5334 .vop_inactive = zfs_freebsd_inactive,
5335 .vop_reclaim = zfs_freebsd_reclaim,
5336 .vop_access = zfs_freebsd_access,
5337 #ifdef FREEBSD_NAMECACHE
5338 .vop_lookup = vfs_cache_lookup,
5339 .vop_cachedlookup = zfs_freebsd_lookup,
5341 .vop_lookup = zfs_freebsd_lookup,
5343 .vop_getattr = zfs_freebsd_getattr,
5344 .vop_setattr = zfs_freebsd_setattr,
5345 .vop_create = zfs_freebsd_create,
5346 .vop_mknod = zfs_freebsd_create,
5347 .vop_mkdir = zfs_freebsd_mkdir,
5348 .vop_readdir = zfs_freebsd_readdir,
5349 .vop_fsync = zfs_freebsd_fsync,
5350 .vop_open = zfs_freebsd_open,
5351 .vop_close = zfs_freebsd_close,
5352 .vop_rmdir = zfs_freebsd_rmdir,
5353 .vop_ioctl = zfs_freebsd_ioctl,
5354 .vop_link = zfs_freebsd_link,
5355 .vop_symlink = zfs_freebsd_symlink,
5356 .vop_readlink = zfs_freebsd_readlink,
5357 .vop_read = zfs_freebsd_read,
5358 .vop_write = zfs_freebsd_write,
5359 .vop_remove = zfs_freebsd_remove,
5360 .vop_rename = zfs_freebsd_rename,
5361 .vop_pathconf = zfs_freebsd_pathconf,
5362 .vop_bmap = VOP_EOPNOTSUPP,
5363 .vop_fid = zfs_freebsd_fid,
5364 .vop_getextattr = zfs_getextattr,
5365 .vop_deleteextattr = zfs_deleteextattr,
5366 .vop_setextattr = zfs_setextattr,
5367 .vop_listextattr = zfs_listextattr,
5368 .vop_getacl = zfs_freebsd_getacl,
5369 .vop_setacl = zfs_freebsd_setacl,
5370 .vop_aclcheck = zfs_freebsd_aclcheck,
5371 .vop_getpages = zfs_freebsd_getpages,
5374 struct vop_vector zfs_fifoops = {
5375 .vop_default = &fifo_specops,
5376 .vop_fsync = zfs_freebsd_fsync,
5377 .vop_access = zfs_freebsd_access,
5378 .vop_getattr = zfs_freebsd_getattr,
5379 .vop_inactive = zfs_freebsd_inactive,
5380 .vop_read = VOP_PANIC,
5381 .vop_reclaim = zfs_freebsd_reclaim,
5382 .vop_setattr = zfs_freebsd_setattr,
5383 .vop_write = VOP_PANIC,
5384 .vop_pathconf = zfs_freebsd_fifo_pathconf,
5385 .vop_fid = zfs_freebsd_fid,
5386 .vop_getacl = zfs_freebsd_getacl,
5387 .vop_setacl = zfs_freebsd_setacl,
5388 .vop_aclcheck = zfs_freebsd_aclcheck,
5392 * special share hidden files vnode operations template
5394 struct vop_vector zfs_shareops = {
5395 .vop_default = &default_vnodeops,
5396 .vop_access = zfs_freebsd_access,
5397 .vop_inactive = zfs_freebsd_inactive,
5398 .vop_reclaim = zfs_freebsd_reclaim,
5399 .vop_fid = zfs_freebsd_fid,
5400 .vop_pathconf = zfs_freebsd_pathconf,