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>
70 #include <vm/vm_pageout.h>
75 * Each vnode op performs some logical unit of work. To do this, the ZPL must
76 * properly lock its in-core state, create a DMU transaction, do the work,
77 * record this work in the intent log (ZIL), commit the DMU transaction,
78 * and wait for the intent log to commit if it is a synchronous operation.
79 * Moreover, the vnode ops must work in both normal and log replay context.
80 * The ordering of events is important to avoid deadlocks and references
81 * to freed memory. The example below illustrates the following Big Rules:
83 * (1) A check must be made in each zfs thread for a mounted file system.
84 * This is done avoiding races using ZFS_ENTER(zfsvfs).
85 * A ZFS_EXIT(zfsvfs) is needed before all returns. Any znodes
86 * must be checked with ZFS_VERIFY_ZP(zp). Both of these macros
87 * can return EIO from the calling function.
89 * (2) VN_RELE() should always be the last thing except for zil_commit()
90 * (if necessary) and ZFS_EXIT(). This is for 3 reasons:
91 * First, if it's the last reference, the vnode/znode
92 * can be freed, so the zp may point to freed memory. Second, the last
93 * reference will call zfs_zinactive(), which may induce a lot of work --
94 * pushing cached pages (which acquires range locks) and syncing out
95 * cached atime changes. Third, zfs_zinactive() may require a new tx,
96 * which could deadlock the system if you were already holding one.
97 * If you must call VN_RELE() within a tx then use VN_RELE_ASYNC().
99 * (3) All range locks must be grabbed before calling dmu_tx_assign(),
100 * as they can span dmu_tx_assign() calls.
102 * (4) Always pass TXG_NOWAIT as the second argument to dmu_tx_assign().
103 * This is critical because we don't want to block while holding locks.
104 * Note, in particular, that if a lock is sometimes acquired before
105 * the tx assigns, and sometimes after (e.g. z_lock), then failing to
106 * use a non-blocking assign can deadlock the system. The scenario:
108 * Thread A has grabbed a lock before calling dmu_tx_assign().
109 * Thread B is in an already-assigned tx, and blocks for this lock.
110 * Thread A calls dmu_tx_assign(TXG_WAIT) and blocks in txg_wait_open()
111 * forever, because the previous txg can't quiesce until B's tx commits.
113 * If dmu_tx_assign() returns ERESTART and zfsvfs->z_assign is TXG_NOWAIT,
114 * then drop all locks, call dmu_tx_wait(), and try again.
116 * (5) If the operation succeeded, generate the intent log entry for it
117 * before dropping locks. This ensures that the ordering of events
118 * in the intent log matches the order in which they actually occurred.
119 * During ZIL replay the zfs_log_* functions will update the sequence
120 * number to indicate the zil transaction has replayed.
122 * (6) At the end of each vnode op, the DMU tx must always commit,
123 * regardless of whether there were any errors.
125 * (7) After dropping all locks, invoke zil_commit(zilog, seq, foid)
126 * to ensure that synchronous semantics are provided when necessary.
128 * In general, this is how things should be ordered in each vnode op:
130 * ZFS_ENTER(zfsvfs); // exit if unmounted
132 * zfs_dirent_lock(&dl, ...) // lock directory entry (may VN_HOLD())
133 * rw_enter(...); // grab any other locks you need
134 * tx = dmu_tx_create(...); // get DMU tx
135 * dmu_tx_hold_*(); // hold each object you might modify
136 * error = dmu_tx_assign(tx, TXG_NOWAIT); // try to assign
138 * rw_exit(...); // drop locks
139 * zfs_dirent_unlock(dl); // unlock directory entry
140 * VN_RELE(...); // release held vnodes
141 * if (error == ERESTART) {
146 * dmu_tx_abort(tx); // abort DMU tx
147 * ZFS_EXIT(zfsvfs); // finished in zfs
148 * return (error); // really out of space
150 * error = do_real_work(); // do whatever this VOP does
152 * zfs_log_*(...); // on success, make ZIL entry
153 * dmu_tx_commit(tx); // commit DMU tx -- error or not
154 * rw_exit(...); // drop locks
155 * zfs_dirent_unlock(dl); // unlock directory entry
156 * VN_RELE(...); // release held vnodes
157 * zil_commit(zilog, seq, foid); // synchronous when necessary
158 * ZFS_EXIT(zfsvfs); // finished in zfs
159 * return (error); // done, report error
164 zfs_open(vnode_t **vpp, int flag, cred_t *cr, caller_context_t *ct)
166 znode_t *zp = VTOZ(*vpp);
167 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
172 if ((flag & FWRITE) && (zp->z_phys->zp_flags & ZFS_APPENDONLY) &&
173 ((flag & FAPPEND) == 0)) {
178 if (!zfs_has_ctldir(zp) && zp->z_zfsvfs->z_vscan &&
179 ZTOV(zp)->v_type == VREG &&
180 !(zp->z_phys->zp_flags & ZFS_AV_QUARANTINED) &&
181 zp->z_phys->zp_size > 0) {
182 if (fs_vscan(*vpp, cr, 0) != 0) {
188 /* Keep a count of the synchronous opens in the znode */
189 if (flag & (FSYNC | FDSYNC))
190 atomic_inc_32(&zp->z_sync_cnt);
198 zfs_close(vnode_t *vp, int flag, int count, offset_t offset, cred_t *cr,
199 caller_context_t *ct)
201 znode_t *zp = VTOZ(vp);
202 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
205 * Clean up any locks held by this process on the vp.
207 cleanlocks(vp, ddi_get_pid(), 0);
208 cleanshares(vp, ddi_get_pid());
213 /* Decrement the synchronous opens in the znode */
214 if ((flag & (FSYNC | FDSYNC)) && (count == 1))
215 atomic_dec_32(&zp->z_sync_cnt);
217 if (!zfs_has_ctldir(zp) && zp->z_zfsvfs->z_vscan &&
218 ZTOV(zp)->v_type == VREG &&
219 !(zp->z_phys->zp_flags & ZFS_AV_QUARANTINED) &&
220 zp->z_phys->zp_size > 0)
221 VERIFY(fs_vscan(vp, cr, 1) == 0);
228 * Lseek support for finding holes (cmd == _FIO_SEEK_HOLE) and
229 * data (cmd == _FIO_SEEK_DATA). "off" is an in/out parameter.
232 zfs_holey(vnode_t *vp, u_long cmd, offset_t *off)
234 znode_t *zp = VTOZ(vp);
235 uint64_t noff = (uint64_t)*off; /* new offset */
240 file_sz = zp->z_phys->zp_size;
241 if (noff >= file_sz) {
245 if (cmd == _FIO_SEEK_HOLE)
250 error = dmu_offset_next(zp->z_zfsvfs->z_os, zp->z_id, hole, &noff);
253 if ((error == ESRCH) || (noff > file_sz)) {
255 * Handle the virtual hole at the end of file.
272 zfs_ioctl(vnode_t *vp, u_long com, intptr_t data, int flag, cred_t *cred,
273 int *rvalp, caller_context_t *ct)
285 * The following two ioctls are used by bfu. Faking out,
286 * necessary to avoid bfu errors.
294 if (ddi_copyin((void *)data, &off, sizeof (off), flag))
298 zfsvfs = zp->z_zfsvfs;
302 /* offset parameter is in/out */
303 error = zfs_holey(vp, com, &off);
307 if (ddi_copyout(&off, (void *)data, sizeof (off), flag))
315 page_lookup(vnode_t *vp, int64_t start, int64_t off, int64_t nbytes)
321 VM_OBJECT_LOCK_ASSERT(obj, MA_OWNED);
324 if ((pp = vm_page_lookup(obj, OFF_TO_IDX(start))) != NULL &&
325 vm_page_is_valid(pp, (vm_offset_t)off, nbytes)) {
326 if (vm_page_sleep_if_busy(pp, FALSE, "zfsmwb"))
329 vm_page_lock_queues();
331 vm_page_unlock_queues();
333 if (__predict_false(obj->cache != NULL)) {
334 vm_page_cache_free(obj, OFF_TO_IDX(start),
335 OFF_TO_IDX(start) + 1);
345 page_unlock(vm_page_t pp)
352 zfs_map_page(vm_page_t pp, struct sf_buf **sfp)
355 *sfp = sf_buf_alloc(pp, 0);
356 return ((caddr_t)sf_buf_kva(*sfp));
360 zfs_unmap_page(struct sf_buf *sf)
368 * When a file is memory mapped, we must keep the IO data synchronized
369 * between the DMU cache and the memory mapped pages. What this means:
371 * On Write: If we find a memory mapped page, we write to *both*
372 * the page and the dmu buffer.
376 update_pages(vnode_t *vp, int64_t start, int len, objset_t *os, uint64_t oid,
377 int segflg, dmu_tx_t *tx)
383 ASSERT(vp->v_mount != NULL);
387 off = start & PAGEOFFSET;
389 for (start &= PAGEMASK; len > 0; start += PAGESIZE) {
391 int nbytes = MIN(PAGESIZE - off, len);
393 if ((pp = page_lookup(vp, start, off, nbytes)) != NULL) {
396 VM_OBJECT_UNLOCK(obj);
397 va = zfs_map_page(pp, &sf);
398 if (segflg == UIO_NOCOPY) {
399 (void) dmu_write(os, oid, start+off, nbytes,
402 (void) dmu_read(os, oid, start+off, nbytes,
403 va+off, DMU_READ_PREFETCH);;
413 VM_OBJECT_UNLOCK(obj);
417 * When a file is memory mapped, we must keep the IO data synchronized
418 * between the DMU cache and the memory mapped pages. What this means:
420 * On Read: We "read" preferentially from memory mapped pages,
421 * else we default from the dmu buffer.
423 * NOTE: We will always "break up" the IO into PAGESIZE uiomoves when
424 * the file is memory mapped.
427 mappedread(vnode_t *vp, int nbytes, uio_t *uio)
429 znode_t *zp = VTOZ(vp);
430 objset_t *os = zp->z_zfsvfs->z_os;
441 ASSERT(vp->v_mount != NULL);
445 start = uio->uio_loffset;
446 off = start & PAGEOFFSET;
449 for (start &= PAGEMASK; len > 0; start += PAGESIZE) {
450 int bytes = MIN(PAGESIZE - off, len);
453 if ((m = vm_page_lookup(obj, OFF_TO_IDX(start))) != NULL &&
454 vm_page_is_valid(m, off, bytes)) {
455 if (vm_page_sleep_if_busy(m, FALSE, "zfsmrb"))
458 VM_OBJECT_UNLOCK(obj);
460 error = dmu_read_uio(os, zp->z_id, uio,
465 uiomove_fromphys(&m, off, bytes, uio);
468 } else if (uio->uio_segflg == UIO_NOCOPY) {
470 * The code below is here to make sendfile(2) work
471 * correctly with ZFS. As pointed out by ups@
472 * sendfile(2) should be changed to use VOP_GETPAGES(),
473 * but it pessimize performance of sendfile/UFS, that's
474 * why I handle this special case in ZFS code.
477 ("unexpected offset in mappedread for sendfile"));
478 if (m != NULL && vm_page_sleep_if_busy(m, FALSE, "zfsmrb"))
481 m = vm_page_alloc(obj, OFF_TO_IDX(start),
482 VM_ALLOC_NOBUSY | VM_ALLOC_NORMAL);
484 VM_OBJECT_UNLOCK(obj);
491 VM_OBJECT_UNLOCK(obj);
493 error = dmu_read_uio(os, zp->z_id, uio,
498 va = zfs_map_page(m, &sf);
499 error = dmu_read(os, zp->z_id, start, bytes, va,
501 if (bytes != PAGE_SIZE)
502 bzero(va + bytes, PAGE_SIZE - bytes);
507 m->valid = VM_PAGE_BITS_ALL;
508 vm_page_io_finish(m);
510 uio->uio_resid -= bytes;
511 uio->uio_offset += bytes;
521 VM_OBJECT_UNLOCK(obj);
522 if (error == 0 && dirbytes > 0)
523 error = dmu_read_uio(os, zp->z_id, uio, dirbytes);
527 offset_t zfs_read_chunk_size = 1024 * 1024; /* Tunable */
530 * Read bytes from specified file into supplied buffer.
532 * IN: vp - vnode of file to be read from.
533 * uio - structure supplying read location, range info,
535 * ioflag - SYNC flags; used to provide FRSYNC semantics.
536 * cr - credentials of caller.
537 * ct - caller context
539 * OUT: uio - updated offset and range, buffer filled.
541 * RETURN: 0 if success
542 * error code if failure
545 * vp - atime updated if byte count > 0
549 zfs_read(vnode_t *vp, uio_t *uio, int ioflag, cred_t *cr, caller_context_t *ct)
551 znode_t *zp = VTOZ(vp);
552 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
562 if (zp->z_phys->zp_flags & ZFS_AV_QUARANTINED) {
568 * Validate file offset
570 if (uio->uio_loffset < (offset_t)0) {
576 * Fasttrack empty reads
578 if (uio->uio_resid == 0) {
584 * Check for mandatory locks
586 if (MANDMODE((mode_t)zp->z_phys->zp_mode)) {
587 if (error = chklock(vp, FREAD,
588 uio->uio_loffset, uio->uio_resid, uio->uio_fmode, ct)) {
595 * If we're in FRSYNC mode, sync out this znode before reading it.
598 zil_commit(zfsvfs->z_log, zp->z_last_itx, zp->z_id);
601 * Lock the range against changes.
603 rl = zfs_range_lock(zp, uio->uio_loffset, uio->uio_resid, RL_READER);
606 * If we are reading past end-of-file we can skip
607 * to the end; but we might still need to set atime.
609 if (uio->uio_loffset >= zp->z_phys->zp_size) {
614 ASSERT(uio->uio_loffset < zp->z_phys->zp_size);
615 n = MIN(uio->uio_resid, zp->z_phys->zp_size - uio->uio_loffset);
618 nbytes = MIN(n, zfs_read_chunk_size -
619 P2PHASE(uio->uio_loffset, zfs_read_chunk_size));
621 if (vn_has_cached_data(vp))
622 error = mappedread(vp, nbytes, uio);
624 error = dmu_read_uio(os, zp->z_id, uio, nbytes);
626 /* convert checksum errors into IO errors */
636 zfs_range_unlock(rl);
638 ZFS_ACCESSTIME_STAMP(zfsvfs, zp);
644 * Fault in the pages of the first n bytes specified by the uio structure.
645 * 1 byte in each page is touched and the uio struct is unmodified.
646 * Any error will exit this routine as this is only a best
647 * attempt to get the pages resident. This is a copy of ufs_trans_touch().
650 zfs_prefault_write(ssize_t n, struct uio *uio)
656 if (uio->uio_segflg != UIO_USERSPACE)
662 cnt = MIN(iov->iov_len, n);
664 /* empty iov entry */
670 * touch each page in this segment.
676 incr = MIN(cnt, PAGESIZE);
681 * touch the last byte in case it straddles a page.
691 * Write the bytes to a file.
693 * IN: vp - vnode of file to be written to.
694 * uio - structure supplying write location, range info,
696 * ioflag - FAPPEND flag set if in append mode.
697 * cr - credentials of caller.
698 * ct - caller context (NFS/CIFS fem monitor only)
700 * OUT: uio - updated offset and range.
702 * RETURN: 0 if success
703 * error code if failure
706 * vp - ctime|mtime updated if byte count > 0
710 zfs_write(vnode_t *vp, uio_t *uio, int ioflag, cred_t *cr, caller_context_t *ct)
712 znode_t *zp = VTOZ(vp);
713 rlim64_t limit = MAXOFFSET_T;
714 ssize_t start_resid = uio->uio_resid;
718 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
723 int max_blksz = zfsvfs->z_max_blksz;
729 * Fasttrack empty write
735 if (limit == RLIM64_INFINITY || limit > MAXOFFSET_T)
742 * If immutable or not appending then return EPERM
744 pflags = zp->z_phys->zp_flags;
745 if ((pflags & (ZFS_IMMUTABLE | ZFS_READONLY)) ||
746 ((pflags & ZFS_APPENDONLY) && !(ioflag & FAPPEND) &&
747 (uio->uio_loffset < zp->z_phys->zp_size))) {
752 zilog = zfsvfs->z_log;
755 * Pre-fault the pages to ensure slow (eg NFS) pages
758 zfs_prefault_write(n, uio);
761 * If in append mode, set the io offset pointer to eof.
763 if (ioflag & FAPPEND) {
765 * Range lock for a file append:
766 * The value for the start of range will be determined by
767 * zfs_range_lock() (to guarantee append semantics).
768 * If this write will cause the block size to increase,
769 * zfs_range_lock() will lock the entire file, so we must
770 * later reduce the range after we grow the block size.
772 rl = zfs_range_lock(zp, 0, n, RL_APPEND);
773 if (rl->r_len == UINT64_MAX) {
774 /* overlocked, zp_size can't change */
775 woff = uio->uio_loffset = zp->z_phys->zp_size;
777 woff = uio->uio_loffset = rl->r_off;
780 woff = uio->uio_loffset;
782 * Validate file offset
790 * If we need to grow the block size then zfs_range_lock()
791 * will lock a wider range than we request here.
792 * Later after growing the block size we reduce the range.
794 rl = zfs_range_lock(zp, woff, n, RL_WRITER);
798 zfs_range_unlock(rl);
803 if ((woff + n) > limit || woff > (limit - n))
807 * Check for mandatory locks
809 if (MANDMODE((mode_t)zp->z_phys->zp_mode) &&
810 (error = chklock(vp, FWRITE, woff, n, uio->uio_fmode, ct)) != 0) {
811 zfs_range_unlock(rl);
815 end_size = MAX(zp->z_phys->zp_size, woff + n);
818 * Write the file in reasonable size chunks. Each chunk is written
819 * in a separate transaction; this keeps the intent log records small
820 * and allows us to do more fine-grained space accounting.
824 woff = uio->uio_loffset;
827 if (zfs_usergroup_overquota(zfsvfs,
828 B_FALSE, zp->z_phys->zp_uid) ||
829 zfs_usergroup_overquota(zfsvfs,
830 B_TRUE, zp->z_phys->zp_gid)) {
832 dmu_return_arcbuf(abuf);
838 * If dmu_assign_arcbuf() is expected to execute with minimum
839 * overhead loan an arc buffer and copy user data to it before
840 * we enter a txg. This avoids holding a txg forever while we
841 * pagefault on a hanging NFS server mapping.
843 if (abuf == NULL && n >= max_blksz &&
844 woff >= zp->z_phys->zp_size &&
845 P2PHASE(woff, max_blksz) == 0 &&
846 zp->z_blksz == max_blksz) {
849 abuf = dmu_request_arcbuf(zp->z_dbuf, max_blksz);
850 ASSERT(abuf != NULL);
851 ASSERT(arc_buf_size(abuf) == max_blksz);
852 if (error = uiocopy(abuf->b_data, max_blksz,
853 UIO_WRITE, uio, &cbytes)) {
854 dmu_return_arcbuf(abuf);
857 ASSERT(cbytes == max_blksz);
861 * Start a transaction.
863 tx = dmu_tx_create(zfsvfs->z_os);
864 dmu_tx_hold_bonus(tx, zp->z_id);
865 dmu_tx_hold_write(tx, zp->z_id, woff, MIN(n, max_blksz));
866 error = dmu_tx_assign(tx, TXG_NOWAIT);
868 if (error == ERESTART) {
875 dmu_return_arcbuf(abuf);
880 * If zfs_range_lock() over-locked we grow the blocksize
881 * and then reduce the lock range. This will only happen
882 * on the first iteration since zfs_range_reduce() will
883 * shrink down r_len to the appropriate size.
885 if (rl->r_len == UINT64_MAX) {
888 if (zp->z_blksz > max_blksz) {
889 ASSERT(!ISP2(zp->z_blksz));
890 new_blksz = MIN(end_size, SPA_MAXBLOCKSIZE);
892 new_blksz = MIN(end_size, max_blksz);
894 zfs_grow_blocksize(zp, new_blksz, tx);
895 zfs_range_reduce(rl, woff, n);
899 * XXX - should we really limit each write to z_max_blksz?
900 * Perhaps we should use SPA_MAXBLOCKSIZE chunks?
902 nbytes = MIN(n, max_blksz - P2PHASE(woff, max_blksz));
904 if (woff + nbytes > zp->z_phys->zp_size)
905 vnode_pager_setsize(vp, woff + nbytes);
908 tx_bytes = uio->uio_resid;
909 error = dmu_write_uio(zfsvfs->z_os, zp->z_id, uio,
911 tx_bytes -= uio->uio_resid;
914 ASSERT(tx_bytes == max_blksz);
915 dmu_assign_arcbuf(zp->z_dbuf, woff, abuf, tx);
916 ASSERT(tx_bytes <= uio->uio_resid);
917 uioskip(uio, tx_bytes);
920 if (tx_bytes && vn_has_cached_data(vp)) {
921 update_pages(vp, woff, tx_bytes, zfsvfs->z_os,
922 zp->z_id, uio->uio_segflg, tx);
926 * If we made no progress, we're done. If we made even
927 * partial progress, update the znode and ZIL accordingly.
936 * Clear Set-UID/Set-GID bits on successful write if not
937 * privileged and at least one of the excute bits is set.
939 * It would be nice to to this after all writes have
940 * been done, but that would still expose the ISUID/ISGID
941 * to another app after the partial write is committed.
943 * Note: we don't call zfs_fuid_map_id() here because
944 * user 0 is not an ephemeral uid.
946 mutex_enter(&zp->z_acl_lock);
947 if ((zp->z_phys->zp_mode & (S_IXUSR | (S_IXUSR >> 3) |
948 (S_IXUSR >> 6))) != 0 &&
949 (zp->z_phys->zp_mode & (S_ISUID | S_ISGID)) != 0 &&
950 secpolicy_vnode_setid_retain(vp, cr,
951 (zp->z_phys->zp_mode & S_ISUID) != 0 &&
952 zp->z_phys->zp_uid == 0) != 0) {
953 zp->z_phys->zp_mode &= ~(S_ISUID | S_ISGID);
955 mutex_exit(&zp->z_acl_lock);
958 * Update time stamp. NOTE: This marks the bonus buffer as
959 * dirty, so we don't have to do it again for zp_size.
961 zfs_time_stamper(zp, CONTENT_MODIFIED, tx);
964 * Update the file size (zp_size) if it has changed;
965 * account for possible concurrent updates.
967 while ((end_size = zp->z_phys->zp_size) < uio->uio_loffset)
968 (void) atomic_cas_64(&zp->z_phys->zp_size, end_size,
970 zfs_log_write(zilog, tx, TX_WRITE, zp, woff, tx_bytes, ioflag);
975 ASSERT(tx_bytes == nbytes);
979 zfs_range_unlock(rl);
982 * If we're in replay mode, or we made no progress, return error.
983 * Otherwise, it's at least a partial write, so it's successful.
985 if (zfsvfs->z_replay || uio->uio_resid == start_resid) {
990 if (ioflag & (FSYNC | FDSYNC))
991 zil_commit(zilog, zp->z_last_itx, zp->z_id);
998 zfs_get_done(dmu_buf_t *db, void *vzgd)
1000 zgd_t *zgd = (zgd_t *)vzgd;
1001 rl_t *rl = zgd->zgd_rl;
1002 vnode_t *vp = ZTOV(rl->r_zp);
1003 objset_t *os = rl->r_zp->z_zfsvfs->z_os;
1006 vfslocked = VFS_LOCK_GIANT(vp->v_vfsp);
1007 dmu_buf_rele(db, vzgd);
1008 zfs_range_unlock(rl);
1010 * Release the vnode asynchronously as we currently have the
1011 * txg stopped from syncing.
1013 VN_RELE_ASYNC(vp, dsl_pool_vnrele_taskq(dmu_objset_pool(os)));
1014 zil_add_block(zgd->zgd_zilog, zgd->zgd_bp);
1015 kmem_free(zgd, sizeof (zgd_t));
1016 VFS_UNLOCK_GIANT(vfslocked);
1020 static int zil_fault_io = 0;
1024 * Get data to generate a TX_WRITE intent log record.
1027 zfs_get_data(void *arg, lr_write_t *lr, char *buf, zio_t *zio)
1029 zfsvfs_t *zfsvfs = arg;
1030 objset_t *os = zfsvfs->z_os;
1032 uint64_t off = lr->lr_offset;
1036 int dlen = lr->lr_length; /* length of user data */
1043 * Nothing to do if the file has been removed
1045 if (zfs_zget(zfsvfs, lr->lr_foid, &zp) != 0)
1047 if (zp->z_unlinked) {
1049 * Release the vnode asynchronously as we currently have the
1050 * txg stopped from syncing.
1052 VN_RELE_ASYNC(ZTOV(zp),
1053 dsl_pool_vnrele_taskq(dmu_objset_pool(os)));
1058 * Write records come in two flavors: immediate and indirect.
1059 * For small writes it's cheaper to store the data with the
1060 * log record (immediate); for large writes it's cheaper to
1061 * sync the data and get a pointer to it (indirect) so that
1062 * we don't have to write the data twice.
1064 if (buf != NULL) { /* immediate write */
1065 rl = zfs_range_lock(zp, off, dlen, RL_READER);
1066 /* test for truncation needs to be done while range locked */
1067 if (off >= zp->z_phys->zp_size) {
1071 VERIFY(0 == dmu_read(os, lr->lr_foid, off, dlen, buf,
1072 DMU_READ_NO_PREFETCH));
1073 } else { /* indirect write */
1074 uint64_t boff; /* block starting offset */
1077 * Have to lock the whole block to ensure when it's
1078 * written out and it's checksum is being calculated
1079 * that no one can change the data. We need to re-check
1080 * blocksize after we get the lock in case it's changed!
1083 if (ISP2(zp->z_blksz)) {
1084 boff = P2ALIGN_TYPED(off, zp->z_blksz,
1090 rl = zfs_range_lock(zp, boff, dlen, RL_READER);
1091 if (zp->z_blksz == dlen)
1093 zfs_range_unlock(rl);
1095 /* test for truncation needs to be done while range locked */
1096 if (off >= zp->z_phys->zp_size) {
1100 zgd = (zgd_t *)kmem_alloc(sizeof (zgd_t), KM_SLEEP);
1102 zgd->zgd_zilog = zfsvfs->z_log;
1103 zgd->zgd_bp = &lr->lr_blkptr;
1109 error = dmu_buf_hold(os, lr->lr_foid, boff, zgd, &db);
1112 error = dmu_buf_hold(os, lr->lr_foid, boff, zgd, &db);
1115 kmem_free(zgd, sizeof (zgd_t));
1119 ASSERT(boff == db->db_offset);
1120 lr->lr_blkoff = off - boff;
1121 error = dmu_sync(zio, db, &lr->lr_blkptr,
1122 lr->lr_common.lrc_txg, zfs_get_done, zgd);
1123 ASSERT((error && error != EINPROGRESS) ||
1124 lr->lr_length <= zp->z_blksz);
1127 * dmu_sync() can compress a block of zeros to a null
1128 * blkptr but the block size still needs to be passed
1129 * through to replay.
1131 BP_SET_LSIZE(&lr->lr_blkptr, db->db_size);
1132 zil_add_block(zfsvfs->z_log, &lr->lr_blkptr);
1136 * If we get EINPROGRESS, then we need to wait for a
1137 * write IO initiated by dmu_sync() to complete before
1138 * we can release this dbuf. We will finish everything
1139 * up in the zfs_get_done() callback.
1141 if (error == EINPROGRESS) {
1143 } else if (error == EALREADY) {
1144 lr->lr_common.lrc_txtype = TX_WRITE2;
1147 dmu_buf_rele(db, zgd);
1148 kmem_free(zgd, sizeof (zgd_t));
1151 zfs_range_unlock(rl);
1153 * Release the vnode asynchronously as we currently have the
1154 * txg stopped from syncing.
1156 VN_RELE_ASYNC(ZTOV(zp), dsl_pool_vnrele_taskq(dmu_objset_pool(os)));
1162 zfs_access(vnode_t *vp, int mode, int flag, cred_t *cr,
1163 caller_context_t *ct)
1165 znode_t *zp = VTOZ(vp);
1166 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1172 if (flag & V_ACE_MASK)
1173 error = zfs_zaccess(zp, mode, flag, B_FALSE, cr);
1175 error = zfs_zaccess_rwx(zp, mode, flag, cr);
1182 * If vnode is for a device return a specfs vnode instead.
1185 specvp_check(vnode_t **vpp, cred_t *cr)
1189 if (IS_DEVVP(*vpp)) {
1192 svp = specvp(*vpp, (*vpp)->v_rdev, (*vpp)->v_type, cr);
1203 * Lookup an entry in a directory, or an extended attribute directory.
1204 * If it exists, return a held vnode reference for it.
1206 * IN: dvp - vnode of directory to search.
1207 * nm - name of entry to lookup.
1208 * pnp - full pathname to lookup [UNUSED].
1209 * flags - LOOKUP_XATTR set if looking for an attribute.
1210 * rdir - root directory vnode [UNUSED].
1211 * cr - credentials of caller.
1212 * ct - caller context
1213 * direntflags - directory lookup flags
1214 * realpnp - returned pathname.
1216 * OUT: vpp - vnode of located entry, NULL if not found.
1218 * RETURN: 0 if success
1219 * error code if failure
1226 zfs_lookup(vnode_t *dvp, char *nm, vnode_t **vpp, struct componentname *cnp,
1227 int nameiop, cred_t *cr, kthread_t *td, int flags)
1229 znode_t *zdp = VTOZ(dvp);
1230 zfsvfs_t *zfsvfs = zdp->z_zfsvfs;
1232 int *direntflags = NULL;
1233 void *realpnp = NULL;
1236 if (!(flags & (LOOKUP_XATTR | FIGNORECASE))) {
1238 if (dvp->v_type != VDIR) {
1240 } else if (zdp->z_dbuf == NULL) {
1244 if (nm[0] == 0 || (nm[0] == '.' && nm[1] == '\0')) {
1245 error = zfs_fastaccesschk_execute(zdp, cr);
1253 vnode_t *tvp = dnlc_lookup(dvp, nm);
1256 error = zfs_fastaccesschk_execute(zdp, cr);
1261 if (tvp == DNLC_NO_VNODE) {
1266 return (specvp_check(vpp, cr));
1272 DTRACE_PROBE2(zfs__fastpath__lookup__miss, vnode_t *, dvp, char *, nm);
1279 if (flags & LOOKUP_XATTR) {
1282 * If the xattr property is off, refuse the lookup request.
1284 if (!(zfsvfs->z_vfs->vfs_flag & VFS_XATTR)) {
1291 * We don't allow recursive attributes..
1292 * Maybe someday we will.
1294 if (zdp->z_phys->zp_flags & ZFS_XATTR) {
1299 if (error = zfs_get_xattrdir(VTOZ(dvp), vpp, cr, flags)) {
1305 * Do we have permission to get into attribute directory?
1308 if (error = zfs_zaccess(VTOZ(*vpp), ACE_EXECUTE, 0,
1318 if (dvp->v_type != VDIR) {
1324 * Check accessibility of directory.
1327 if (error = zfs_zaccess(zdp, ACE_EXECUTE, 0, B_FALSE, cr)) {
1332 if (zfsvfs->z_utf8 && u8_validate(nm, strlen(nm),
1333 NULL, U8_VALIDATE_ENTIRE, &error) < 0) {
1338 error = zfs_dirlook(zdp, nm, vpp, flags, direntflags, realpnp);
1340 error = specvp_check(vpp, cr);
1342 /* Translate errors and add SAVENAME when needed. */
1343 if (cnp->cn_flags & ISLASTCN) {
1347 if (error == ENOENT) {
1348 error = EJUSTRETURN;
1349 cnp->cn_flags |= SAVENAME;
1355 cnp->cn_flags |= SAVENAME;
1359 if (error == 0 && (nm[0] != '.' || nm[1] != '\0')) {
1362 if (cnp->cn_flags & ISDOTDOT) {
1363 ltype = VOP_ISLOCKED(dvp);
1367 error = vn_lock(*vpp, cnp->cn_lkflags);
1368 if (cnp->cn_flags & ISDOTDOT)
1369 vn_lock(dvp, ltype | LK_RETRY);
1379 #ifdef FREEBSD_NAMECACHE
1381 * Insert name into cache (as non-existent) if appropriate.
1383 if (error == ENOENT && (cnp->cn_flags & MAKEENTRY) && nameiop != CREATE)
1384 cache_enter(dvp, *vpp, cnp);
1386 * Insert name into cache if appropriate.
1388 if (error == 0 && (cnp->cn_flags & MAKEENTRY)) {
1389 if (!(cnp->cn_flags & ISLASTCN) ||
1390 (nameiop != DELETE && nameiop != RENAME)) {
1391 cache_enter(dvp, *vpp, cnp);
1400 * Attempt to create a new entry in a directory. If the entry
1401 * already exists, truncate the file if permissible, else return
1402 * an error. Return the vp of the created or trunc'd file.
1404 * IN: dvp - vnode of directory to put new file entry in.
1405 * name - name of new file entry.
1406 * vap - attributes of new file.
1407 * excl - flag indicating exclusive or non-exclusive mode.
1408 * mode - mode to open file with.
1409 * cr - credentials of caller.
1410 * flag - large file flag [UNUSED].
1411 * ct - caller context
1412 * vsecp - ACL to be set
1414 * OUT: vpp - vnode of created or trunc'd entry.
1416 * RETURN: 0 if success
1417 * error code if failure
1420 * dvp - ctime|mtime updated if new entry created
1421 * vp - ctime|mtime always, atime if new
1426 zfs_create(vnode_t *dvp, char *name, vattr_t *vap, int excl, int mode,
1427 vnode_t **vpp, cred_t *cr, kthread_t *td)
1429 znode_t *zp, *dzp = VTOZ(dvp);
1430 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
1438 gid_t gid = crgetgid(cr);
1439 zfs_acl_ids_t acl_ids;
1440 boolean_t fuid_dirtied;
1445 * If we have an ephemeral id, ACL, or XVATTR then
1446 * make sure file system is at proper version
1449 ksid = crgetsid(cr, KSID_OWNER);
1451 uid = ksid_getid(ksid);
1454 if (zfsvfs->z_use_fuids == B_FALSE &&
1455 (vsecp || (vap->va_mask & AT_XVATTR) ||
1456 IS_EPHEMERAL(crgetuid(cr)) || IS_EPHEMERAL(crgetgid(cr))))
1462 zilog = zfsvfs->z_log;
1464 if (zfsvfs->z_utf8 && u8_validate(name, strlen(name),
1465 NULL, U8_VALIDATE_ENTIRE, &error) < 0) {
1470 if (vap->va_mask & AT_XVATTR) {
1471 if ((error = secpolicy_xvattr(dvp, (xvattr_t *)vap,
1472 crgetuid(cr), cr, vap->va_type)) != 0) {
1480 if ((vap->va_mode & S_ISVTX) && secpolicy_vnode_stky_modify(cr))
1481 vap->va_mode &= ~S_ISVTX;
1483 if (*name == '\0') {
1485 * Null component name refers to the directory itself.
1492 /* possible VN_HOLD(zp) */
1495 if (flag & FIGNORECASE)
1498 error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg,
1501 if (strcmp(name, "..") == 0)
1511 * Create a new file object and update the directory
1514 if (error = zfs_zaccess(dzp, ACE_ADD_FILE, 0, B_FALSE, cr)) {
1519 * We only support the creation of regular files in
1520 * extended attribute directories.
1522 if ((dzp->z_phys->zp_flags & ZFS_XATTR) &&
1523 (vap->va_type != VREG)) {
1529 if ((error = zfs_acl_ids_create(dzp, 0, vap, cr, vsecp,
1532 if (zfs_acl_ids_overquota(zfsvfs, &acl_ids)) {
1533 zfs_acl_ids_free(&acl_ids);
1538 tx = dmu_tx_create(os);
1539 dmu_tx_hold_bonus(tx, DMU_NEW_OBJECT);
1540 fuid_dirtied = zfsvfs->z_fuid_dirty;
1542 zfs_fuid_txhold(zfsvfs, tx);
1543 dmu_tx_hold_bonus(tx, dzp->z_id);
1544 dmu_tx_hold_zap(tx, dzp->z_id, TRUE, name);
1545 if (acl_ids.z_aclp->z_acl_bytes > ZFS_ACE_SPACE) {
1546 dmu_tx_hold_write(tx, DMU_NEW_OBJECT,
1547 0, SPA_MAXBLOCKSIZE);
1549 error = dmu_tx_assign(tx, TXG_NOWAIT);
1551 zfs_acl_ids_free(&acl_ids);
1552 zfs_dirent_unlock(dl);
1553 if (error == ERESTART) {
1562 zfs_mknode(dzp, vap, tx, cr, 0, &zp, 0, &acl_ids);
1565 zfs_fuid_sync(zfsvfs, tx);
1567 (void) zfs_link_create(dl, zp, tx, ZNEW);
1569 txtype = zfs_log_create_txtype(Z_FILE, vsecp, vap);
1570 if (flag & FIGNORECASE)
1572 zfs_log_create(zilog, tx, txtype, dzp, zp, name,
1573 vsecp, acl_ids.z_fuidp, vap);
1574 zfs_acl_ids_free(&acl_ids);
1577 int aflags = (flag & FAPPEND) ? V_APPEND : 0;
1580 * A directory entry already exists for this name.
1583 * Can't truncate an existing file if in exclusive mode.
1590 * Can't open a directory for writing.
1592 if ((ZTOV(zp)->v_type == VDIR) && (mode & S_IWRITE)) {
1597 * Verify requested access to file.
1599 if (mode && (error = zfs_zaccess_rwx(zp, mode, aflags, cr))) {
1603 mutex_enter(&dzp->z_lock);
1605 mutex_exit(&dzp->z_lock);
1608 * Truncate regular files if requested.
1610 if ((ZTOV(zp)->v_type == VREG) &&
1611 (vap->va_mask & AT_SIZE) && (vap->va_size == 0)) {
1612 /* we can't hold any locks when calling zfs_freesp() */
1613 zfs_dirent_unlock(dl);
1615 error = zfs_freesp(zp, 0, 0, mode, TRUE);
1617 vnevent_create(ZTOV(zp), ct);
1623 zfs_dirent_unlock(dl);
1630 error = specvp_check(vpp, cr);
1638 * Remove an entry from a directory.
1640 * IN: dvp - vnode of directory to remove entry from.
1641 * name - name of entry to remove.
1642 * cr - credentials of caller.
1643 * ct - caller context
1644 * flags - case flags
1646 * RETURN: 0 if success
1647 * error code if failure
1651 * vp - ctime (if nlink > 0)
1655 zfs_remove(vnode_t *dvp, char *name, cred_t *cr, caller_context_t *ct,
1658 znode_t *zp, *dzp = VTOZ(dvp);
1659 znode_t *xzp = NULL;
1661 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
1663 uint64_t acl_obj, xattr_obj;
1666 boolean_t may_delete_now, delete_now = FALSE;
1667 boolean_t unlinked, toobig = FALSE;
1669 pathname_t *realnmp = NULL;
1676 zilog = zfsvfs->z_log;
1678 if (flags & FIGNORECASE) {
1686 * Attempt to lock directory; fail if entry doesn't exist.
1688 if (error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg,
1698 if (error = zfs_zaccess_delete(dzp, zp, cr)) {
1703 * Need to use rmdir for removing directories.
1705 if (vp->v_type == VDIR) {
1710 vnevent_remove(vp, dvp, name, ct);
1713 dnlc_remove(dvp, realnmp->pn_buf);
1715 dnlc_remove(dvp, name);
1717 may_delete_now = FALSE;
1720 * We may delete the znode now, or we may put it in the unlinked set;
1721 * it depends on whether we're the last link, and on whether there are
1722 * other holds on the vnode. So we dmu_tx_hold() the right things to
1723 * allow for either case.
1725 tx = dmu_tx_create(zfsvfs->z_os);
1726 dmu_tx_hold_zap(tx, dzp->z_id, FALSE, name);
1727 dmu_tx_hold_bonus(tx, zp->z_id);
1728 if (may_delete_now) {
1730 zp->z_phys->zp_size > zp->z_blksz * DMU_MAX_DELETEBLKCNT;
1731 /* if the file is too big, only hold_free a token amount */
1732 dmu_tx_hold_free(tx, zp->z_id, 0,
1733 (toobig ? DMU_MAX_ACCESS : DMU_OBJECT_END));
1736 /* are there any extended attributes? */
1737 if ((xattr_obj = zp->z_phys->zp_xattr) != 0) {
1738 /* XXX - do we need this if we are deleting? */
1739 dmu_tx_hold_bonus(tx, xattr_obj);
1742 /* are there any additional acls */
1743 if ((acl_obj = zp->z_phys->zp_acl.z_acl_extern_obj) != 0 &&
1745 dmu_tx_hold_free(tx, acl_obj, 0, DMU_OBJECT_END);
1747 /* charge as an update -- would be nice not to charge at all */
1748 dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL);
1750 error = dmu_tx_assign(tx, TXG_NOWAIT);
1752 zfs_dirent_unlock(dl);
1754 if (error == ERESTART) {
1767 * Remove the directory entry.
1769 error = zfs_link_destroy(dl, zp, tx, zflg, &unlinked);
1776 if (0 && unlinked) {
1778 delete_now = may_delete_now && !toobig &&
1779 vp->v_count == 1 && !vn_has_cached_data(vp) &&
1780 zp->z_phys->zp_xattr == xattr_obj &&
1781 zp->z_phys->zp_acl.z_acl_extern_obj == acl_obj;
1786 if (zp->z_phys->zp_xattr) {
1787 error = zfs_zget(zfsvfs, zp->z_phys->zp_xattr, &xzp);
1788 ASSERT3U(error, ==, 0);
1789 ASSERT3U(xzp->z_phys->zp_links, ==, 2);
1790 dmu_buf_will_dirty(xzp->z_dbuf, tx);
1791 mutex_enter(&xzp->z_lock);
1792 xzp->z_unlinked = 1;
1793 xzp->z_phys->zp_links = 0;
1794 mutex_exit(&xzp->z_lock);
1795 zfs_unlinked_add(xzp, tx);
1796 zp->z_phys->zp_xattr = 0; /* probably unnecessary */
1798 mutex_enter(&zp->z_lock);
1801 ASSERT3U(vp->v_count, ==, 0);
1803 mutex_exit(&zp->z_lock);
1804 zfs_znode_delete(zp, tx);
1805 } else if (unlinked) {
1806 zfs_unlinked_add(zp, tx);
1810 if (flags & FIGNORECASE)
1812 zfs_log_remove(zilog, tx, txtype, dzp, name);
1819 zfs_dirent_unlock(dl);
1824 /* this rele is delayed to prevent nesting transactions */
1833 * Create a new directory and insert it into dvp using the name
1834 * provided. Return a pointer to the inserted directory.
1836 * IN: dvp - vnode of directory to add subdir to.
1837 * dirname - name of new directory.
1838 * vap - attributes of new directory.
1839 * cr - credentials of caller.
1840 * ct - caller context
1841 * vsecp - ACL to be set
1843 * OUT: vpp - vnode of created directory.
1845 * RETURN: 0 if success
1846 * error code if failure
1849 * dvp - ctime|mtime updated
1850 * vp - ctime|mtime|atime updated
1854 zfs_mkdir(vnode_t *dvp, char *dirname, vattr_t *vap, vnode_t **vpp, cred_t *cr,
1855 caller_context_t *ct, int flags, vsecattr_t *vsecp)
1857 znode_t *zp, *dzp = VTOZ(dvp);
1858 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
1867 gid_t gid = crgetgid(cr);
1868 zfs_acl_ids_t acl_ids;
1869 boolean_t fuid_dirtied;
1871 ASSERT(vap->va_type == VDIR);
1874 * If we have an ephemeral id, ACL, or XVATTR then
1875 * make sure file system is at proper version
1878 ksid = crgetsid(cr, KSID_OWNER);
1880 uid = ksid_getid(ksid);
1883 if (zfsvfs->z_use_fuids == B_FALSE &&
1884 (vsecp || (vap->va_mask & AT_XVATTR) || IS_EPHEMERAL(crgetuid(cr))||
1885 IS_EPHEMERAL(crgetgid(cr))))
1890 zilog = zfsvfs->z_log;
1892 if (dzp->z_phys->zp_flags & ZFS_XATTR) {
1897 if (zfsvfs->z_utf8 && u8_validate(dirname,
1898 strlen(dirname), NULL, U8_VALIDATE_ENTIRE, &error) < 0) {
1902 if (flags & FIGNORECASE)
1905 if (vap->va_mask & AT_XVATTR)
1906 if ((error = secpolicy_xvattr(dvp, (xvattr_t *)vap,
1907 crgetuid(cr), cr, vap->va_type)) != 0) {
1913 * First make sure the new directory doesn't exist.
1918 if (error = zfs_dirent_lock(&dl, dzp, dirname, &zp, zf,
1924 if (error = zfs_zaccess(dzp, ACE_ADD_SUBDIRECTORY, 0, B_FALSE, cr)) {
1925 zfs_dirent_unlock(dl);
1930 if ((error = zfs_acl_ids_create(dzp, 0, vap, cr, vsecp,
1932 zfs_dirent_unlock(dl);
1936 if (zfs_acl_ids_overquota(zfsvfs, &acl_ids)) {
1937 zfs_acl_ids_free(&acl_ids);
1938 zfs_dirent_unlock(dl);
1944 * Add a new entry to the directory.
1946 tx = dmu_tx_create(zfsvfs->z_os);
1947 dmu_tx_hold_zap(tx, dzp->z_id, TRUE, dirname);
1948 dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL);
1949 fuid_dirtied = zfsvfs->z_fuid_dirty;
1951 zfs_fuid_txhold(zfsvfs, tx);
1952 if (acl_ids.z_aclp->z_acl_bytes > ZFS_ACE_SPACE)
1953 dmu_tx_hold_write(tx, DMU_NEW_OBJECT,
1954 0, SPA_MAXBLOCKSIZE);
1955 error = dmu_tx_assign(tx, TXG_NOWAIT);
1957 zfs_acl_ids_free(&acl_ids);
1958 zfs_dirent_unlock(dl);
1959 if (error == ERESTART) {
1972 zfs_mknode(dzp, vap, tx, cr, 0, &zp, 0, &acl_ids);
1975 zfs_fuid_sync(zfsvfs, tx);
1977 * Now put new name in parent dir.
1979 (void) zfs_link_create(dl, zp, tx, ZNEW);
1983 txtype = zfs_log_create_txtype(Z_DIR, vsecp, vap);
1984 if (flags & FIGNORECASE)
1986 zfs_log_create(zilog, tx, txtype, dzp, zp, dirname, vsecp,
1987 acl_ids.z_fuidp, vap);
1989 zfs_acl_ids_free(&acl_ids);
1992 zfs_dirent_unlock(dl);
1999 * Remove a directory subdir entry. If the current working
2000 * directory is the same as the subdir to be removed, the
2003 * IN: dvp - vnode of directory to remove from.
2004 * name - name of directory to be removed.
2005 * cwd - vnode of current working directory.
2006 * cr - credentials of caller.
2007 * ct - caller context
2008 * flags - case flags
2010 * RETURN: 0 if success
2011 * error code if failure
2014 * dvp - ctime|mtime updated
2018 zfs_rmdir(vnode_t *dvp, char *name, vnode_t *cwd, cred_t *cr,
2019 caller_context_t *ct, int flags)
2021 znode_t *dzp = VTOZ(dvp);
2024 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
2033 zilog = zfsvfs->z_log;
2035 if (flags & FIGNORECASE)
2041 * Attempt to lock directory; fail if entry doesn't exist.
2043 if (error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg,
2051 if (error = zfs_zaccess_delete(dzp, zp, cr)) {
2055 if (vp->v_type != VDIR) {
2065 vnevent_rmdir(vp, dvp, name, ct);
2068 * Grab a lock on the directory to make sure that noone is
2069 * trying to add (or lookup) entries while we are removing it.
2071 rw_enter(&zp->z_name_lock, RW_WRITER);
2074 * Grab a lock on the parent pointer to make sure we play well
2075 * with the treewalk and directory rename code.
2077 rw_enter(&zp->z_parent_lock, RW_WRITER);
2079 tx = dmu_tx_create(zfsvfs->z_os);
2080 dmu_tx_hold_zap(tx, dzp->z_id, FALSE, name);
2081 dmu_tx_hold_bonus(tx, zp->z_id);
2082 dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL);
2083 error = dmu_tx_assign(tx, TXG_NOWAIT);
2085 rw_exit(&zp->z_parent_lock);
2086 rw_exit(&zp->z_name_lock);
2087 zfs_dirent_unlock(dl);
2089 if (error == ERESTART) {
2099 #ifdef FREEBSD_NAMECACHE
2103 error = zfs_link_destroy(dl, zp, tx, zflg, NULL);
2106 uint64_t txtype = TX_RMDIR;
2107 if (flags & FIGNORECASE)
2109 zfs_log_remove(zilog, tx, txtype, dzp, name);
2114 rw_exit(&zp->z_parent_lock);
2115 rw_exit(&zp->z_name_lock);
2116 #ifdef FREEBSD_NAMECACHE
2120 zfs_dirent_unlock(dl);
2129 * Read as many directory entries as will fit into the provided
2130 * buffer from the given directory cursor position (specified in
2131 * the uio structure.
2133 * IN: vp - vnode of directory to read.
2134 * uio - structure supplying read location, range info,
2135 * and return buffer.
2136 * cr - credentials of caller.
2137 * ct - caller context
2138 * flags - case flags
2140 * OUT: uio - updated offset and range, buffer filled.
2141 * eofp - set to true if end-of-file detected.
2143 * RETURN: 0 if success
2144 * error code if failure
2147 * vp - atime updated
2149 * Note that the low 4 bits of the cookie returned by zap is always zero.
2150 * This allows us to use the low range for "special" directory entries:
2151 * We use 0 for '.', and 1 for '..'. If this is the root of the filesystem,
2152 * we use the offset 2 for the '.zfs' directory.
2156 zfs_readdir(vnode_t *vp, uio_t *uio, cred_t *cr, int *eofp, int *ncookies, u_long **cookies)
2158 znode_t *zp = VTOZ(vp);
2162 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
2167 zap_attribute_t zap;
2168 uint_t bytes_wanted;
2169 uint64_t offset; /* must be unsigned; checks for < 1 */
2174 boolean_t check_sysattrs;
2177 u_long *cooks = NULL;
2184 * If we are not given an eof variable,
2191 * Check for valid iov_len.
2193 if (uio->uio_iov->iov_len <= 0) {
2199 * Quit if directory has been removed (posix)
2201 if ((*eofp = zp->z_unlinked) != 0) {
2208 offset = uio->uio_loffset;
2209 prefetch = zp->z_zn_prefetch;
2212 * Initialize the iterator cursor.
2216 * Start iteration from the beginning of the directory.
2218 zap_cursor_init(&zc, os, zp->z_id);
2221 * The offset is a serialized cursor.
2223 zap_cursor_init_serialized(&zc, os, zp->z_id, offset);
2227 * Get space to change directory entries into fs independent format.
2229 iovp = uio->uio_iov;
2230 bytes_wanted = iovp->iov_len;
2231 if (uio->uio_segflg != UIO_SYSSPACE || uio->uio_iovcnt != 1) {
2232 bufsize = bytes_wanted;
2233 outbuf = kmem_alloc(bufsize, KM_SLEEP);
2234 odp = (struct dirent64 *)outbuf;
2236 bufsize = bytes_wanted;
2237 odp = (struct dirent64 *)iovp->iov_base;
2239 eodp = (struct edirent *)odp;
2241 if (ncookies != NULL) {
2243 * Minimum entry size is dirent size and 1 byte for a file name.
2245 ncooks = uio->uio_resid / (sizeof(struct dirent) - sizeof(((struct dirent *)NULL)->d_name) + 1);
2246 cooks = malloc(ncooks * sizeof(u_long), M_TEMP, M_WAITOK);
2251 * If this VFS supports the system attribute view interface; and
2252 * we're looking at an extended attribute directory; and we care
2253 * about normalization conflicts on this vfs; then we must check
2254 * for normalization conflicts with the sysattr name space.
2257 check_sysattrs = vfs_has_feature(vp->v_vfsp, VFSFT_SYSATTR_VIEWS) &&
2258 (vp->v_flag & V_XATTRDIR) && zfsvfs->z_norm &&
2259 (flags & V_RDDIR_ENTFLAGS);
2265 * Transform to file-system independent format
2268 while (outcount < bytes_wanted) {
2274 * Special case `.', `..', and `.zfs'.
2277 (void) strcpy(zap.za_name, ".");
2278 zap.za_normalization_conflict = 0;
2281 } else if (offset == 1) {
2282 (void) strcpy(zap.za_name, "..");
2283 zap.za_normalization_conflict = 0;
2284 objnum = zp->z_phys->zp_parent;
2286 } else if (offset == 2 && zfs_show_ctldir(zp)) {
2287 (void) strcpy(zap.za_name, ZFS_CTLDIR_NAME);
2288 zap.za_normalization_conflict = 0;
2289 objnum = ZFSCTL_INO_ROOT;
2295 if (error = zap_cursor_retrieve(&zc, &zap)) {
2296 if ((*eofp = (error == ENOENT)) != 0)
2302 if (zap.za_integer_length != 8 ||
2303 zap.za_num_integers != 1) {
2304 cmn_err(CE_WARN, "zap_readdir: bad directory "
2305 "entry, obj = %lld, offset = %lld\n",
2306 (u_longlong_t)zp->z_id,
2307 (u_longlong_t)offset);
2312 objnum = ZFS_DIRENT_OBJ(zap.za_first_integer);
2314 * MacOS X can extract the object type here such as:
2315 * uint8_t type = ZFS_DIRENT_TYPE(zap.za_first_integer);
2317 type = ZFS_DIRENT_TYPE(zap.za_first_integer);
2319 if (check_sysattrs && !zap.za_normalization_conflict) {
2321 zap.za_normalization_conflict =
2322 xattr_sysattr_casechk(zap.za_name);
2324 panic("%s:%u: TODO", __func__, __LINE__);
2329 if (flags & V_RDDIR_ACCFILTER) {
2331 * If we have no access at all, don't include
2332 * this entry in the returned information
2335 if (zfs_zget(zp->z_zfsvfs, objnum, &ezp) != 0)
2337 if (!zfs_has_access(ezp, cr)) {
2344 if (flags & V_RDDIR_ENTFLAGS)
2345 reclen = EDIRENT_RECLEN(strlen(zap.za_name));
2347 reclen = DIRENT64_RECLEN(strlen(zap.za_name));
2350 * Will this entry fit in the buffer?
2352 if (outcount + reclen > bufsize) {
2354 * Did we manage to fit anything in the buffer?
2362 if (flags & V_RDDIR_ENTFLAGS) {
2364 * Add extended flag entry:
2366 eodp->ed_ino = objnum;
2367 eodp->ed_reclen = reclen;
2368 /* NOTE: ed_off is the offset for the *next* entry */
2369 next = &(eodp->ed_off);
2370 eodp->ed_eflags = zap.za_normalization_conflict ?
2371 ED_CASE_CONFLICT : 0;
2372 (void) strncpy(eodp->ed_name, zap.za_name,
2373 EDIRENT_NAMELEN(reclen));
2374 eodp = (edirent_t *)((intptr_t)eodp + reclen);
2379 odp->d_ino = objnum;
2380 odp->d_reclen = reclen;
2381 odp->d_namlen = strlen(zap.za_name);
2382 (void) strlcpy(odp->d_name, zap.za_name, odp->d_namlen + 1);
2384 odp = (dirent64_t *)((intptr_t)odp + reclen);
2388 ASSERT(outcount <= bufsize);
2390 /* Prefetch znode */
2392 dmu_prefetch(os, objnum, 0, 0);
2396 * Move to the next entry, fill in the previous offset.
2398 if (offset > 2 || (offset == 2 && !zfs_show_ctldir(zp))) {
2399 zap_cursor_advance(&zc);
2400 offset = zap_cursor_serialize(&zc);
2405 if (cooks != NULL) {
2408 KASSERT(ncooks >= 0, ("ncookies=%d", ncooks));
2411 zp->z_zn_prefetch = B_FALSE; /* a lookup will re-enable pre-fetching */
2413 /* Subtract unused cookies */
2414 if (ncookies != NULL)
2415 *ncookies -= ncooks;
2417 if (uio->uio_segflg == UIO_SYSSPACE && uio->uio_iovcnt == 1) {
2418 iovp->iov_base += outcount;
2419 iovp->iov_len -= outcount;
2420 uio->uio_resid -= outcount;
2421 } else if (error = uiomove(outbuf, (long)outcount, UIO_READ, uio)) {
2423 * Reset the pointer.
2425 offset = uio->uio_loffset;
2429 zap_cursor_fini(&zc);
2430 if (uio->uio_segflg != UIO_SYSSPACE || uio->uio_iovcnt != 1)
2431 kmem_free(outbuf, bufsize);
2433 if (error == ENOENT)
2436 ZFS_ACCESSTIME_STAMP(zfsvfs, zp);
2438 uio->uio_loffset = offset;
2440 if (error != 0 && cookies != NULL) {
2441 free(*cookies, M_TEMP);
2448 ulong_t zfs_fsync_sync_cnt = 4;
2451 zfs_fsync(vnode_t *vp, int syncflag, cred_t *cr, caller_context_t *ct)
2453 znode_t *zp = VTOZ(vp);
2454 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
2456 (void) tsd_set(zfs_fsyncer_key, (void *)zfs_fsync_sync_cnt);
2460 zil_commit(zfsvfs->z_log, zp->z_last_itx, zp->z_id);
2467 * Get the requested file attributes and place them in the provided
2470 * IN: vp - vnode of file.
2471 * vap - va_mask identifies requested attributes.
2472 * If AT_XVATTR set, then optional attrs are requested
2473 * flags - ATTR_NOACLCHECK (CIFS server context)
2474 * cr - credentials of caller.
2475 * ct - caller context
2477 * OUT: vap - attribute values.
2479 * RETURN: 0 (always succeeds)
2483 zfs_getattr(vnode_t *vp, vattr_t *vap, int flags, cred_t *cr,
2484 caller_context_t *ct)
2486 znode_t *zp = VTOZ(vp);
2487 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
2491 u_longlong_t nblocks;
2493 xvattr_t *xvap = (xvattr_t *)vap; /* vap may be an xvattr_t * */
2494 xoptattr_t *xoap = NULL;
2495 boolean_t skipaclchk = (flags & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE;
2502 * If ACL is trivial don't bother looking for ACE_READ_ATTRIBUTES.
2503 * Also, if we are the owner don't bother, since owner should
2504 * always be allowed to read basic attributes of file.
2506 if (!(pzp->zp_flags & ZFS_ACL_TRIVIAL) &&
2507 (pzp->zp_uid != crgetuid(cr))) {
2508 if (error = zfs_zaccess(zp, ACE_READ_ATTRIBUTES, 0,
2516 * Return all attributes. It's cheaper to provide the answer
2517 * than to determine whether we were asked the question.
2520 mutex_enter(&zp->z_lock);
2521 vap->va_type = IFTOVT(pzp->zp_mode);
2522 vap->va_mode = pzp->zp_mode & ~S_IFMT;
2523 zfs_fuid_map_ids(zp, cr, &vap->va_uid, &vap->va_gid);
2524 // vap->va_fsid = zp->z_zfsvfs->z_vfs->vfs_dev;
2525 vap->va_nodeid = zp->z_id;
2526 if ((vp->v_flag & VROOT) && zfs_show_ctldir(zp))
2527 links = pzp->zp_links + 1;
2529 links = pzp->zp_links;
2530 vap->va_nlink = MIN(links, UINT32_MAX); /* nlink_t limit! */
2531 vap->va_size = pzp->zp_size;
2532 vap->va_fsid = vp->v_mount->mnt_stat.f_fsid.val[0];
2533 vap->va_rdev = zfs_cmpldev(pzp->zp_rdev);
2534 vap->va_seq = zp->z_seq;
2535 vap->va_flags = 0; /* FreeBSD: Reset chflags(2) flags. */
2538 * Add in any requested optional attributes and the create time.
2539 * Also set the corresponding bits in the returned attribute bitmap.
2541 if ((xoap = xva_getxoptattr(xvap)) != NULL && zfsvfs->z_use_fuids) {
2542 if (XVA_ISSET_REQ(xvap, XAT_ARCHIVE)) {
2544 ((pzp->zp_flags & ZFS_ARCHIVE) != 0);
2545 XVA_SET_RTN(xvap, XAT_ARCHIVE);
2548 if (XVA_ISSET_REQ(xvap, XAT_READONLY)) {
2549 xoap->xoa_readonly =
2550 ((pzp->zp_flags & ZFS_READONLY) != 0);
2551 XVA_SET_RTN(xvap, XAT_READONLY);
2554 if (XVA_ISSET_REQ(xvap, XAT_SYSTEM)) {
2556 ((pzp->zp_flags & ZFS_SYSTEM) != 0);
2557 XVA_SET_RTN(xvap, XAT_SYSTEM);
2560 if (XVA_ISSET_REQ(xvap, XAT_HIDDEN)) {
2562 ((pzp->zp_flags & ZFS_HIDDEN) != 0);
2563 XVA_SET_RTN(xvap, XAT_HIDDEN);
2566 if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK)) {
2567 xoap->xoa_nounlink =
2568 ((pzp->zp_flags & ZFS_NOUNLINK) != 0);
2569 XVA_SET_RTN(xvap, XAT_NOUNLINK);
2572 if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE)) {
2573 xoap->xoa_immutable =
2574 ((pzp->zp_flags & ZFS_IMMUTABLE) != 0);
2575 XVA_SET_RTN(xvap, XAT_IMMUTABLE);
2578 if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY)) {
2579 xoap->xoa_appendonly =
2580 ((pzp->zp_flags & ZFS_APPENDONLY) != 0);
2581 XVA_SET_RTN(xvap, XAT_APPENDONLY);
2584 if (XVA_ISSET_REQ(xvap, XAT_NODUMP)) {
2586 ((pzp->zp_flags & ZFS_NODUMP) != 0);
2587 XVA_SET_RTN(xvap, XAT_NODUMP);
2590 if (XVA_ISSET_REQ(xvap, XAT_OPAQUE)) {
2592 ((pzp->zp_flags & ZFS_OPAQUE) != 0);
2593 XVA_SET_RTN(xvap, XAT_OPAQUE);
2596 if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED)) {
2597 xoap->xoa_av_quarantined =
2598 ((pzp->zp_flags & ZFS_AV_QUARANTINED) != 0);
2599 XVA_SET_RTN(xvap, XAT_AV_QUARANTINED);
2602 if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED)) {
2603 xoap->xoa_av_modified =
2604 ((pzp->zp_flags & ZFS_AV_MODIFIED) != 0);
2605 XVA_SET_RTN(xvap, XAT_AV_MODIFIED);
2608 if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP) &&
2609 vp->v_type == VREG &&
2610 (pzp->zp_flags & ZFS_BONUS_SCANSTAMP)) {
2612 dmu_object_info_t doi;
2615 * Only VREG files have anti-virus scanstamps, so we
2616 * won't conflict with symlinks in the bonus buffer.
2618 dmu_object_info_from_db(zp->z_dbuf, &doi);
2619 len = sizeof (xoap->xoa_av_scanstamp) +
2620 sizeof (znode_phys_t);
2621 if (len <= doi.doi_bonus_size) {
2623 * pzp points to the start of the
2624 * znode_phys_t. pzp + 1 points to the
2625 * first byte after the znode_phys_t.
2627 (void) memcpy(xoap->xoa_av_scanstamp,
2629 sizeof (xoap->xoa_av_scanstamp));
2630 XVA_SET_RTN(xvap, XAT_AV_SCANSTAMP);
2634 if (XVA_ISSET_REQ(xvap, XAT_CREATETIME)) {
2635 ZFS_TIME_DECODE(&xoap->xoa_createtime, pzp->zp_crtime);
2636 XVA_SET_RTN(xvap, XAT_CREATETIME);
2640 ZFS_TIME_DECODE(&vap->va_atime, pzp->zp_atime);
2641 ZFS_TIME_DECODE(&vap->va_mtime, pzp->zp_mtime);
2642 ZFS_TIME_DECODE(&vap->va_ctime, pzp->zp_ctime);
2643 ZFS_TIME_DECODE(&vap->va_birthtime, pzp->zp_crtime);
2645 mutex_exit(&zp->z_lock);
2647 dmu_object_size_from_db(zp->z_dbuf, &blksize, &nblocks);
2648 vap->va_blksize = blksize;
2649 vap->va_bytes = nblocks << 9; /* nblocks * 512 */
2651 if (zp->z_blksz == 0) {
2653 * Block size hasn't been set; suggest maximal I/O transfers.
2655 vap->va_blksize = zfsvfs->z_max_blksz;
2663 * Set the file attributes to the values contained in the
2666 * IN: vp - vnode of file to be modified.
2667 * vap - new attribute values.
2668 * If AT_XVATTR set, then optional attrs are being set
2669 * flags - ATTR_UTIME set if non-default time values provided.
2670 * - ATTR_NOACLCHECK (CIFS context only).
2671 * cr - credentials of caller.
2672 * ct - caller context
2674 * RETURN: 0 if success
2675 * error code if failure
2678 * vp - ctime updated, mtime updated if size changed.
2682 zfs_setattr(vnode_t *vp, vattr_t *vap, int flags, cred_t *cr,
2683 caller_context_t *ct)
2685 znode_t *zp = VTOZ(vp);
2687 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
2692 uint_t mask = vap->va_mask;
2694 uint64_t saved_mode;
2697 uint64_t new_uid, new_gid;
2699 int need_policy = FALSE;
2701 zfs_fuid_info_t *fuidp = NULL;
2702 xvattr_t *xvap = (xvattr_t *)vap; /* vap may be an xvattr_t * */
2704 zfs_acl_t *aclp = NULL;
2705 boolean_t skipaclchk = (flags & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE;
2706 boolean_t fuid_dirtied = B_FALSE;
2711 if (mask & AT_NOSET)
2718 zilog = zfsvfs->z_log;
2721 * Make sure that if we have ephemeral uid/gid or xvattr specified
2722 * that file system is at proper version level
2725 if (zfsvfs->z_use_fuids == B_FALSE &&
2726 (((mask & AT_UID) && IS_EPHEMERAL(vap->va_uid)) ||
2727 ((mask & AT_GID) && IS_EPHEMERAL(vap->va_gid)) ||
2728 (mask & AT_XVATTR))) {
2733 if (mask & AT_SIZE && vp->v_type == VDIR) {
2738 if (mask & AT_SIZE && vp->v_type != VREG && vp->v_type != VFIFO) {
2744 * If this is an xvattr_t, then get a pointer to the structure of
2745 * optional attributes. If this is NULL, then we have a vattr_t.
2747 xoap = xva_getxoptattr(xvap);
2749 xva_init(&tmpxvattr);
2752 * Immutable files can only alter immutable bit and atime
2754 if ((pzp->zp_flags & ZFS_IMMUTABLE) &&
2755 ((mask & (AT_SIZE|AT_UID|AT_GID|AT_MTIME|AT_MODE)) ||
2756 ((mask & AT_XVATTR) && XVA_ISSET_REQ(xvap, XAT_CREATETIME)))) {
2761 if ((mask & AT_SIZE) && (pzp->zp_flags & ZFS_READONLY)) {
2767 * Verify timestamps doesn't overflow 32 bits.
2768 * ZFS can handle large timestamps, but 32bit syscalls can't
2769 * handle times greater than 2039. This check should be removed
2770 * once large timestamps are fully supported.
2772 if (mask & (AT_ATIME | AT_MTIME)) {
2773 if (((mask & AT_ATIME) && TIMESPEC_OVERFLOW(&vap->va_atime)) ||
2774 ((mask & AT_MTIME) && TIMESPEC_OVERFLOW(&vap->va_mtime))) {
2783 /* Can this be moved to before the top label? */
2784 if (zfsvfs->z_vfs->vfs_flag & VFS_RDONLY) {
2790 * First validate permissions
2793 if (mask & AT_SIZE) {
2794 err = zfs_zaccess(zp, ACE_WRITE_DATA, 0, skipaclchk, cr);
2800 * XXX - Note, we are not providing any open
2801 * mode flags here (like FNDELAY), so we may
2802 * block if there are locks present... this
2803 * should be addressed in openat().
2805 /* XXX - would it be OK to generate a log record here? */
2806 err = zfs_freesp(zp, vap->va_size, 0, 0, FALSE);
2813 if (mask & (AT_ATIME|AT_MTIME) ||
2814 ((mask & AT_XVATTR) && (XVA_ISSET_REQ(xvap, XAT_HIDDEN) ||
2815 XVA_ISSET_REQ(xvap, XAT_READONLY) ||
2816 XVA_ISSET_REQ(xvap, XAT_ARCHIVE) ||
2817 XVA_ISSET_REQ(xvap, XAT_CREATETIME) ||
2818 XVA_ISSET_REQ(xvap, XAT_SYSTEM))))
2819 need_policy = zfs_zaccess(zp, ACE_WRITE_ATTRIBUTES, 0,
2822 if (mask & (AT_UID|AT_GID)) {
2823 int idmask = (mask & (AT_UID|AT_GID));
2828 * NOTE: even if a new mode is being set,
2829 * we may clear S_ISUID/S_ISGID bits.
2832 if (!(mask & AT_MODE))
2833 vap->va_mode = pzp->zp_mode;
2836 * Take ownership or chgrp to group we are a member of
2839 take_owner = (mask & AT_UID) && (vap->va_uid == crgetuid(cr));
2840 take_group = (mask & AT_GID) &&
2841 zfs_groupmember(zfsvfs, vap->va_gid, cr);
2844 * If both AT_UID and AT_GID are set then take_owner and
2845 * take_group must both be set in order to allow taking
2848 * Otherwise, send the check through secpolicy_vnode_setattr()
2852 if (((idmask == (AT_UID|AT_GID)) && take_owner && take_group) ||
2853 ((idmask == AT_UID) && take_owner) ||
2854 ((idmask == AT_GID) && take_group)) {
2855 if (zfs_zaccess(zp, ACE_WRITE_OWNER, 0,
2856 skipaclchk, cr) == 0) {
2858 * Remove setuid/setgid for non-privileged users
2860 secpolicy_setid_clear(vap, vp, cr);
2861 trim_mask = (mask & (AT_UID|AT_GID));
2870 mutex_enter(&zp->z_lock);
2871 oldva.va_mode = pzp->zp_mode;
2872 zfs_fuid_map_ids(zp, cr, &oldva.va_uid, &oldva.va_gid);
2873 if (mask & AT_XVATTR) {
2875 * Update xvattr mask to include only those attributes
2876 * that are actually changing.
2878 * the bits will be restored prior to actually setting
2879 * the attributes so the caller thinks they were set.
2881 if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY)) {
2882 if (xoap->xoa_appendonly !=
2883 ((pzp->zp_flags & ZFS_APPENDONLY) != 0)) {
2886 XVA_CLR_REQ(xvap, XAT_APPENDONLY);
2887 XVA_SET_REQ(&tmpxvattr, XAT_APPENDONLY);
2891 if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK)) {
2892 if (xoap->xoa_nounlink !=
2893 ((pzp->zp_flags & ZFS_NOUNLINK) != 0)) {
2896 XVA_CLR_REQ(xvap, XAT_NOUNLINK);
2897 XVA_SET_REQ(&tmpxvattr, XAT_NOUNLINK);
2901 if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE)) {
2902 if (xoap->xoa_immutable !=
2903 ((pzp->zp_flags & ZFS_IMMUTABLE) != 0)) {
2906 XVA_CLR_REQ(xvap, XAT_IMMUTABLE);
2907 XVA_SET_REQ(&tmpxvattr, XAT_IMMUTABLE);
2911 if (XVA_ISSET_REQ(xvap, XAT_NODUMP)) {
2912 if (xoap->xoa_nodump !=
2913 ((pzp->zp_flags & ZFS_NODUMP) != 0)) {
2916 XVA_CLR_REQ(xvap, XAT_NODUMP);
2917 XVA_SET_REQ(&tmpxvattr, XAT_NODUMP);
2921 if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED)) {
2922 if (xoap->xoa_av_modified !=
2923 ((pzp->zp_flags & ZFS_AV_MODIFIED) != 0)) {
2926 XVA_CLR_REQ(xvap, XAT_AV_MODIFIED);
2927 XVA_SET_REQ(&tmpxvattr, XAT_AV_MODIFIED);
2931 if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED)) {
2932 if ((vp->v_type != VREG &&
2933 xoap->xoa_av_quarantined) ||
2934 xoap->xoa_av_quarantined !=
2935 ((pzp->zp_flags & ZFS_AV_QUARANTINED) != 0)) {
2938 XVA_CLR_REQ(xvap, XAT_AV_QUARANTINED);
2939 XVA_SET_REQ(&tmpxvattr, XAT_AV_QUARANTINED);
2943 if (need_policy == FALSE &&
2944 (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP) ||
2945 XVA_ISSET_REQ(xvap, XAT_OPAQUE))) {
2950 mutex_exit(&zp->z_lock);
2952 if (mask & AT_MODE) {
2953 if (zfs_zaccess(zp, ACE_WRITE_ACL, 0, skipaclchk, cr) == 0) {
2954 err = secpolicy_setid_setsticky_clear(vp, vap,
2960 trim_mask |= AT_MODE;
2968 * If trim_mask is set then take ownership
2969 * has been granted or write_acl is present and user
2970 * has the ability to modify mode. In that case remove
2971 * UID|GID and or MODE from mask so that
2972 * secpolicy_vnode_setattr() doesn't revoke it.
2976 saved_mask = vap->va_mask;
2977 vap->va_mask &= ~trim_mask;
2978 if (trim_mask & AT_MODE) {
2980 * Save the mode, as secpolicy_vnode_setattr()
2981 * will overwrite it with ova.va_mode.
2983 saved_mode = vap->va_mode;
2986 err = secpolicy_vnode_setattr(cr, vp, vap, &oldva, flags,
2987 (int (*)(void *, int, cred_t *))zfs_zaccess_unix, zp);
2994 vap->va_mask |= saved_mask;
2995 if (trim_mask & AT_MODE) {
2997 * Recover the mode after
2998 * secpolicy_vnode_setattr().
3000 vap->va_mode = saved_mode;
3006 * secpolicy_vnode_setattr, or take ownership may have
3009 mask = vap->va_mask;
3011 tx = dmu_tx_create(zfsvfs->z_os);
3012 dmu_tx_hold_bonus(tx, zp->z_id);
3014 if (mask & AT_MODE) {
3015 uint64_t pmode = pzp->zp_mode;
3017 new_mode = (pmode & S_IFMT) | (vap->va_mode & ~S_IFMT);
3019 if (err = zfs_acl_chmod_setattr(zp, &aclp, new_mode))
3021 if (pzp->zp_acl.z_acl_extern_obj) {
3022 /* Are we upgrading ACL from old V0 format to new V1 */
3023 if (zfsvfs->z_version <= ZPL_VERSION_FUID &&
3024 pzp->zp_acl.z_acl_version ==
3025 ZFS_ACL_VERSION_INITIAL) {
3026 dmu_tx_hold_free(tx,
3027 pzp->zp_acl.z_acl_extern_obj, 0,
3029 dmu_tx_hold_write(tx, DMU_NEW_OBJECT,
3030 0, aclp->z_acl_bytes);
3032 dmu_tx_hold_write(tx,
3033 pzp->zp_acl.z_acl_extern_obj, 0,
3036 } else if (aclp->z_acl_bytes > ZFS_ACE_SPACE) {
3037 dmu_tx_hold_write(tx, DMU_NEW_OBJECT,
3038 0, aclp->z_acl_bytes);
3042 if (mask & (AT_UID | AT_GID)) {
3043 if (pzp->zp_xattr) {
3044 err = zfs_zget(zp->z_zfsvfs, pzp->zp_xattr, &attrzp);
3047 dmu_tx_hold_bonus(tx, attrzp->z_id);
3049 if (mask & AT_UID) {
3050 new_uid = zfs_fuid_create(zfsvfs,
3051 (uint64_t)vap->va_uid, cr, ZFS_OWNER, &fuidp);
3052 if (new_uid != pzp->zp_uid &&
3053 zfs_usergroup_overquota(zfsvfs, B_FALSE, new_uid)) {
3059 if (mask & AT_GID) {
3060 new_gid = zfs_fuid_create(zfsvfs, (uint64_t)vap->va_gid,
3061 cr, ZFS_GROUP, &fuidp);
3062 if (new_gid != pzp->zp_gid &&
3063 zfs_usergroup_overquota(zfsvfs, B_TRUE, new_gid)) {
3068 fuid_dirtied = zfsvfs->z_fuid_dirty;
3070 if (zfsvfs->z_fuid_obj == 0) {
3071 dmu_tx_hold_bonus(tx, DMU_NEW_OBJECT);
3072 dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0,
3073 FUID_SIZE_ESTIMATE(zfsvfs));
3074 dmu_tx_hold_zap(tx, MASTER_NODE_OBJ,
3077 dmu_tx_hold_bonus(tx, zfsvfs->z_fuid_obj);
3078 dmu_tx_hold_write(tx, zfsvfs->z_fuid_obj, 0,
3079 FUID_SIZE_ESTIMATE(zfsvfs));
3084 err = dmu_tx_assign(tx, TXG_NOWAIT);
3086 if (err == ERESTART)
3091 dmu_buf_will_dirty(zp->z_dbuf, tx);
3094 * Set each attribute requested.
3095 * We group settings according to the locks they need to acquire.
3097 * Note: you cannot set ctime directly, although it will be
3098 * updated as a side-effect of calling this function.
3101 mutex_enter(&zp->z_lock);
3103 if (mask & AT_MODE) {
3104 mutex_enter(&zp->z_acl_lock);
3105 zp->z_phys->zp_mode = new_mode;
3106 err = zfs_aclset_common(zp, aclp, cr, tx);
3107 ASSERT3U(err, ==, 0);
3108 zp->z_acl_cached = aclp;
3110 mutex_exit(&zp->z_acl_lock);
3114 mutex_enter(&attrzp->z_lock);
3116 if (mask & AT_UID) {
3117 pzp->zp_uid = new_uid;
3119 attrzp->z_phys->zp_uid = new_uid;
3122 if (mask & AT_GID) {
3123 pzp->zp_gid = new_gid;
3125 attrzp->z_phys->zp_gid = new_gid;
3129 mutex_exit(&attrzp->z_lock);
3131 if (mask & AT_ATIME)
3132 ZFS_TIME_ENCODE(&vap->va_atime, pzp->zp_atime);
3134 if (mask & AT_MTIME)
3135 ZFS_TIME_ENCODE(&vap->va_mtime, pzp->zp_mtime);
3137 /* XXX - shouldn't this be done *before* the ATIME/MTIME checks? */
3139 zfs_time_stamper_locked(zp, CONTENT_MODIFIED, tx);
3141 zfs_time_stamper_locked(zp, STATE_CHANGED, tx);
3143 * Do this after setting timestamps to prevent timestamp
3144 * update from toggling bit
3147 if (xoap && (mask & AT_XVATTR)) {
3150 * restore trimmed off masks
3151 * so that return masks can be set for caller.
3154 if (XVA_ISSET_REQ(&tmpxvattr, XAT_APPENDONLY)) {
3155 XVA_SET_REQ(xvap, XAT_APPENDONLY);
3157 if (XVA_ISSET_REQ(&tmpxvattr, XAT_NOUNLINK)) {
3158 XVA_SET_REQ(xvap, XAT_NOUNLINK);
3160 if (XVA_ISSET_REQ(&tmpxvattr, XAT_IMMUTABLE)) {
3161 XVA_SET_REQ(xvap, XAT_IMMUTABLE);
3163 if (XVA_ISSET_REQ(&tmpxvattr, XAT_NODUMP)) {
3164 XVA_SET_REQ(xvap, XAT_NODUMP);
3166 if (XVA_ISSET_REQ(&tmpxvattr, XAT_AV_MODIFIED)) {
3167 XVA_SET_REQ(xvap, XAT_AV_MODIFIED);
3169 if (XVA_ISSET_REQ(&tmpxvattr, XAT_AV_QUARANTINED)) {
3170 XVA_SET_REQ(xvap, XAT_AV_QUARANTINED);
3173 if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP)) {
3175 dmu_object_info_t doi;
3177 ASSERT(vp->v_type == VREG);
3179 /* Grow the bonus buffer if necessary. */
3180 dmu_object_info_from_db(zp->z_dbuf, &doi);
3181 len = sizeof (xoap->xoa_av_scanstamp) +
3182 sizeof (znode_phys_t);
3183 if (len > doi.doi_bonus_size)
3184 VERIFY(dmu_set_bonus(zp->z_dbuf, len, tx) == 0);
3186 zfs_xvattr_set(zp, xvap);
3190 zfs_fuid_sync(zfsvfs, tx);
3193 zfs_log_setattr(zilog, tx, TX_SETATTR, zp, vap, mask, fuidp);
3195 mutex_exit(&zp->z_lock);
3199 VN_RELE(ZTOV(attrzp));
3205 zfs_fuid_info_free(fuidp);
3214 if (err == ERESTART)
3221 typedef struct zfs_zlock {
3222 krwlock_t *zl_rwlock; /* lock we acquired */
3223 znode_t *zl_znode; /* znode we held */
3224 struct zfs_zlock *zl_next; /* next in list */
3228 * Drop locks and release vnodes that were held by zfs_rename_lock().
3231 zfs_rename_unlock(zfs_zlock_t **zlpp)
3235 while ((zl = *zlpp) != NULL) {
3236 if (zl->zl_znode != NULL)
3237 VN_RELE(ZTOV(zl->zl_znode));
3238 rw_exit(zl->zl_rwlock);
3239 *zlpp = zl->zl_next;
3240 kmem_free(zl, sizeof (*zl));
3245 * Search back through the directory tree, using the ".." entries.
3246 * Lock each directory in the chain to prevent concurrent renames.
3247 * Fail any attempt to move a directory into one of its own descendants.
3248 * XXX - z_parent_lock can overlap with map or grow locks
3251 zfs_rename_lock(znode_t *szp, znode_t *tdzp, znode_t *sdzp, zfs_zlock_t **zlpp)
3255 uint64_t rootid = zp->z_zfsvfs->z_root;
3256 uint64_t *oidp = &zp->z_id;
3257 krwlock_t *rwlp = &szp->z_parent_lock;
3258 krw_t rw = RW_WRITER;
3261 * First pass write-locks szp and compares to zp->z_id.
3262 * Later passes read-lock zp and compare to zp->z_parent.
3265 if (!rw_tryenter(rwlp, rw)) {
3267 * Another thread is renaming in this path.
3268 * Note that if we are a WRITER, we don't have any
3269 * parent_locks held yet.
3271 if (rw == RW_READER && zp->z_id > szp->z_id) {
3273 * Drop our locks and restart
3275 zfs_rename_unlock(&zl);
3279 rwlp = &szp->z_parent_lock;
3284 * Wait for other thread to drop its locks
3290 zl = kmem_alloc(sizeof (*zl), KM_SLEEP);
3291 zl->zl_rwlock = rwlp;
3292 zl->zl_znode = NULL;
3293 zl->zl_next = *zlpp;
3296 if (*oidp == szp->z_id) /* We're a descendant of szp */
3299 if (*oidp == rootid) /* We've hit the top */
3302 if (rw == RW_READER) { /* i.e. not the first pass */
3303 int error = zfs_zget(zp->z_zfsvfs, *oidp, &zp);
3308 oidp = &zp->z_phys->zp_parent;
3309 rwlp = &zp->z_parent_lock;
3312 } while (zp->z_id != sdzp->z_id);
3318 * Move an entry from the provided source directory to the target
3319 * directory. Change the entry name as indicated.
3321 * IN: sdvp - Source directory containing the "old entry".
3322 * snm - Old entry name.
3323 * tdvp - Target directory to contain the "new entry".
3324 * tnm - New entry name.
3325 * cr - credentials of caller.
3326 * ct - caller context
3327 * flags - case flags
3329 * RETURN: 0 if success
3330 * error code if failure
3333 * sdvp,tdvp - ctime|mtime updated
3337 zfs_rename(vnode_t *sdvp, char *snm, vnode_t *tdvp, char *tnm, cred_t *cr,
3338 caller_context_t *ct, int flags)
3340 znode_t *tdzp, *szp, *tzp;
3341 znode_t *sdzp = VTOZ(sdvp);
3342 zfsvfs_t *zfsvfs = sdzp->z_zfsvfs;
3345 zfs_dirlock_t *sdl, *tdl;
3348 int cmp, serr, terr;
3353 ZFS_VERIFY_ZP(sdzp);
3354 zilog = zfsvfs->z_log;
3357 * Make sure we have the real vp for the target directory.
3359 if (VOP_REALVP(tdvp, &realvp, ct) == 0)
3362 if (tdvp->v_vfsp != sdvp->v_vfsp || zfsctl_is_node(tdvp)) {
3368 ZFS_VERIFY_ZP(tdzp);
3369 if (zfsvfs->z_utf8 && u8_validate(tnm,
3370 strlen(tnm), NULL, U8_VALIDATE_ENTIRE, &error) < 0) {
3375 if (flags & FIGNORECASE)
3384 * This is to prevent the creation of links into attribute space
3385 * by renaming a linked file into/outof an attribute directory.
3386 * See the comment in zfs_link() for why this is considered bad.
3388 if ((tdzp->z_phys->zp_flags & ZFS_XATTR) !=
3389 (sdzp->z_phys->zp_flags & ZFS_XATTR)) {
3395 * Lock source and target directory entries. To prevent deadlock,
3396 * a lock ordering must be defined. We lock the directory with
3397 * the smallest object id first, or if it's a tie, the one with
3398 * the lexically first name.
3400 if (sdzp->z_id < tdzp->z_id) {
3402 } else if (sdzp->z_id > tdzp->z_id) {
3406 * First compare the two name arguments without
3407 * considering any case folding.
3409 int nofold = (zfsvfs->z_norm & ~U8_TEXTPREP_TOUPPER);
3411 cmp = u8_strcmp(snm, tnm, 0, nofold, U8_UNICODE_LATEST, &error);
3412 ASSERT(error == 0 || !zfsvfs->z_utf8);
3415 * POSIX: "If the old argument and the new argument
3416 * both refer to links to the same existing file,
3417 * the rename() function shall return successfully
3418 * and perform no other action."
3424 * If the file system is case-folding, then we may
3425 * have some more checking to do. A case-folding file
3426 * system is either supporting mixed case sensitivity
3427 * access or is completely case-insensitive. Note
3428 * that the file system is always case preserving.
3430 * In mixed sensitivity mode case sensitive behavior
3431 * is the default. FIGNORECASE must be used to
3432 * explicitly request case insensitive behavior.
3434 * If the source and target names provided differ only
3435 * by case (e.g., a request to rename 'tim' to 'Tim'),
3436 * we will treat this as a special case in the
3437 * case-insensitive mode: as long as the source name
3438 * is an exact match, we will allow this to proceed as
3439 * a name-change request.
3441 if ((zfsvfs->z_case == ZFS_CASE_INSENSITIVE ||
3442 (zfsvfs->z_case == ZFS_CASE_MIXED &&
3443 flags & FIGNORECASE)) &&
3444 u8_strcmp(snm, tnm, 0, zfsvfs->z_norm, U8_UNICODE_LATEST,
3447 * case preserving rename request, require exact
3456 * If the source and destination directories are the same, we should
3457 * grab the z_name_lock of that directory only once.
3461 rw_enter(&sdzp->z_name_lock, RW_READER);
3465 serr = zfs_dirent_lock(&sdl, sdzp, snm, &szp,
3466 ZEXISTS | zflg, NULL, NULL);
3467 terr = zfs_dirent_lock(&tdl,
3468 tdzp, tnm, &tzp, ZRENAMING | zflg, NULL, NULL);
3470 terr = zfs_dirent_lock(&tdl,
3471 tdzp, tnm, &tzp, zflg, NULL, NULL);
3472 serr = zfs_dirent_lock(&sdl,
3473 sdzp, snm, &szp, ZEXISTS | ZRENAMING | zflg,
3479 * Source entry invalid or not there.
3482 zfs_dirent_unlock(tdl);
3488 rw_exit(&sdzp->z_name_lock);
3490 if (strcmp(snm, ".") == 0 || strcmp(snm, "..") == 0)
3496 zfs_dirent_unlock(sdl);
3500 rw_exit(&sdzp->z_name_lock);
3502 if (strcmp(tnm, "..") == 0)
3509 * Must have write access at the source to remove the old entry
3510 * and write access at the target to create the new entry.
3511 * Note that if target and source are the same, this can be
3512 * done in a single check.
3515 if (error = zfs_zaccess_rename(sdzp, szp, tdzp, tzp, cr))
3518 if (ZTOV(szp)->v_type == VDIR) {
3520 * Check to make sure rename is valid.
3521 * Can't do a move like this: /usr/a/b to /usr/a/b/c/d
3523 if (error = zfs_rename_lock(szp, tdzp, sdzp, &zl))
3528 * Does target exist?
3532 * Source and target must be the same type.
3534 if (ZTOV(szp)->v_type == VDIR) {
3535 if (ZTOV(tzp)->v_type != VDIR) {
3540 if (ZTOV(tzp)->v_type == VDIR) {
3546 * POSIX dictates that when the source and target
3547 * entries refer to the same file object, rename
3548 * must do nothing and exit without error.
3550 if (szp->z_id == tzp->z_id) {
3556 vnevent_rename_src(ZTOV(szp), sdvp, snm, ct);
3558 vnevent_rename_dest(ZTOV(tzp), tdvp, tnm, ct);
3561 * notify the target directory if it is not the same
3562 * as source directory.
3565 vnevent_rename_dest_dir(tdvp, ct);
3568 tx = dmu_tx_create(zfsvfs->z_os);
3569 dmu_tx_hold_bonus(tx, szp->z_id); /* nlink changes */
3570 dmu_tx_hold_bonus(tx, sdzp->z_id); /* nlink changes */
3571 dmu_tx_hold_zap(tx, sdzp->z_id, FALSE, snm);
3572 dmu_tx_hold_zap(tx, tdzp->z_id, TRUE, tnm);
3574 dmu_tx_hold_bonus(tx, tdzp->z_id); /* nlink changes */
3576 dmu_tx_hold_bonus(tx, tzp->z_id); /* parent changes */
3577 dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL);
3578 error = dmu_tx_assign(tx, TXG_NOWAIT);
3581 zfs_rename_unlock(&zl);
3582 zfs_dirent_unlock(sdl);
3583 zfs_dirent_unlock(tdl);
3586 rw_exit(&sdzp->z_name_lock);
3591 if (error == ERESTART) {
3601 if (tzp) /* Attempt to remove the existing target */
3602 error = zfs_link_destroy(tdl, tzp, tx, zflg, NULL);
3605 error = zfs_link_create(tdl, szp, tx, ZRENAMING);
3607 szp->z_phys->zp_flags |= ZFS_AV_MODIFIED;
3609 error = zfs_link_destroy(sdl, szp, tx, ZRENAMING, NULL);
3612 zfs_log_rename(zilog, tx,
3613 TX_RENAME | (flags & FIGNORECASE ? TX_CI : 0),
3614 sdzp, sdl->dl_name, tdzp, tdl->dl_name, szp);
3616 /* Update path information for the target vnode */
3617 vn_renamepath(tdvp, ZTOV(szp), tnm, strlen(tnm));
3619 #ifdef FREEBSD_NAMECACHE
3630 zfs_rename_unlock(&zl);
3632 zfs_dirent_unlock(sdl);
3633 zfs_dirent_unlock(tdl);
3636 rw_exit(&sdzp->z_name_lock);
3648 * Insert the indicated symbolic reference entry into the directory.
3650 * IN: dvp - Directory to contain new symbolic link.
3651 * link - Name for new symlink entry.
3652 * vap - Attributes of new entry.
3653 * target - Target path of new symlink.
3654 * cr - credentials of caller.
3655 * ct - caller context
3656 * flags - case flags
3658 * RETURN: 0 if success
3659 * error code if failure
3662 * dvp - ctime|mtime updated
3666 zfs_symlink(vnode_t *dvp, vnode_t **vpp, char *name, vattr_t *vap, char *link,
3667 cred_t *cr, kthread_t *td)
3669 znode_t *zp, *dzp = VTOZ(dvp);
3672 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
3674 int len = strlen(link);
3677 zfs_acl_ids_t acl_ids;
3678 boolean_t fuid_dirtied;
3681 ASSERT(vap->va_type == VLNK);
3685 zilog = zfsvfs->z_log;
3687 if (zfsvfs->z_utf8 && u8_validate(name, strlen(name),
3688 NULL, U8_VALIDATE_ENTIRE, &error) < 0) {
3692 if (flags & FIGNORECASE)
3695 if (error = zfs_zaccess(dzp, ACE_ADD_FILE, 0, B_FALSE, cr)) {
3700 if (len > MAXPATHLEN) {
3702 return (ENAMETOOLONG);
3706 * Attempt to lock directory; fail if entry already exists.
3708 error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg, NULL, NULL);
3714 VERIFY(0 == zfs_acl_ids_create(dzp, 0, vap, cr, NULL, &acl_ids));
3715 if (zfs_acl_ids_overquota(zfsvfs, &acl_ids)) {
3716 zfs_acl_ids_free(&acl_ids);
3717 zfs_dirent_unlock(dl);
3721 tx = dmu_tx_create(zfsvfs->z_os);
3722 fuid_dirtied = zfsvfs->z_fuid_dirty;
3723 dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, MAX(1, len));
3724 dmu_tx_hold_bonus(tx, dzp->z_id);
3725 dmu_tx_hold_zap(tx, dzp->z_id, TRUE, name);
3726 if (acl_ids.z_aclp->z_acl_bytes > ZFS_ACE_SPACE)
3727 dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, SPA_MAXBLOCKSIZE);
3729 zfs_fuid_txhold(zfsvfs, tx);
3730 error = dmu_tx_assign(tx, TXG_NOWAIT);
3732 zfs_acl_ids_free(&acl_ids);
3733 zfs_dirent_unlock(dl);
3734 if (error == ERESTART) {
3744 dmu_buf_will_dirty(dzp->z_dbuf, tx);
3747 * Create a new object for the symlink.
3748 * Put the link content into bonus buffer if it will fit;
3749 * otherwise, store it just like any other file data.
3751 if (sizeof (znode_phys_t) + len <= dmu_bonus_max()) {
3752 zfs_mknode(dzp, vap, tx, cr, 0, &zp, len, &acl_ids);
3754 bcopy(link, zp->z_phys + 1, len);
3758 zfs_mknode(dzp, vap, tx, cr, 0, &zp, 0, &acl_ids);
3761 zfs_fuid_sync(zfsvfs, tx);
3763 * Nothing can access the znode yet so no locking needed
3764 * for growing the znode's blocksize.
3766 zfs_grow_blocksize(zp, len, tx);
3768 VERIFY(0 == dmu_buf_hold(zfsvfs->z_os,
3769 zp->z_id, 0, FTAG, &dbp));
3770 dmu_buf_will_dirty(dbp, tx);
3772 ASSERT3U(len, <=, dbp->db_size);
3773 bcopy(link, dbp->db_data, len);
3774 dmu_buf_rele(dbp, FTAG);
3776 zp->z_phys->zp_size = len;
3779 * Insert the new object into the directory.
3781 (void) zfs_link_create(dl, zp, tx, ZNEW);
3783 uint64_t txtype = TX_SYMLINK;
3784 if (flags & FIGNORECASE)
3786 zfs_log_symlink(zilog, tx, txtype, dzp, zp, name, link);
3790 zfs_acl_ids_free(&acl_ids);
3794 zfs_dirent_unlock(dl);
3801 * Return, in the buffer contained in the provided uio structure,
3802 * the symbolic path referred to by vp.
3804 * IN: vp - vnode of symbolic link.
3805 * uoip - structure to contain the link path.
3806 * cr - credentials of caller.
3807 * ct - caller context
3809 * OUT: uio - structure to contain the link path.
3811 * RETURN: 0 if success
3812 * error code if failure
3815 * vp - atime updated
3819 zfs_readlink(vnode_t *vp, uio_t *uio, cred_t *cr, caller_context_t *ct)
3821 znode_t *zp = VTOZ(vp);
3822 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
3829 bufsz = (size_t)zp->z_phys->zp_size;
3830 if (bufsz + sizeof (znode_phys_t) <= zp->z_dbuf->db_size) {
3831 error = uiomove(zp->z_phys + 1,
3832 MIN((size_t)bufsz, uio->uio_resid), UIO_READ, uio);
3835 error = dmu_buf_hold(zfsvfs->z_os, zp->z_id, 0, FTAG, &dbp);
3840 error = uiomove(dbp->db_data,
3841 MIN((size_t)bufsz, uio->uio_resid), UIO_READ, uio);
3842 dmu_buf_rele(dbp, FTAG);
3845 ZFS_ACCESSTIME_STAMP(zfsvfs, zp);
3851 * Insert a new entry into directory tdvp referencing svp.
3853 * IN: tdvp - Directory to contain new entry.
3854 * svp - vnode of new entry.
3855 * name - name of new entry.
3856 * cr - credentials of caller.
3857 * ct - caller context
3859 * RETURN: 0 if success
3860 * error code if failure
3863 * tdvp - ctime|mtime updated
3864 * svp - ctime updated
3868 zfs_link(vnode_t *tdvp, vnode_t *svp, char *name, cred_t *cr,
3869 caller_context_t *ct, int flags)
3871 znode_t *dzp = VTOZ(tdvp);
3873 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
3883 ASSERT(tdvp->v_type == VDIR);
3887 zilog = zfsvfs->z_log;
3889 if (VOP_REALVP(svp, &realvp, ct) == 0)
3893 * POSIX dictates that we return EPERM here.
3894 * Better choices include ENOTSUP or EISDIR.
3896 if (svp->v_type == VDIR) {
3901 if (svp->v_vfsp != tdvp->v_vfsp || zfsctl_is_node(svp)) {
3909 /* Prevent links to .zfs/shares files */
3911 if (szp->z_phys->zp_parent == zfsvfs->z_shares_dir) {
3916 if (zfsvfs->z_utf8 && u8_validate(name,
3917 strlen(name), NULL, U8_VALIDATE_ENTIRE, &error) < 0) {
3921 if (flags & FIGNORECASE)
3925 * We do not support links between attributes and non-attributes
3926 * because of the potential security risk of creating links
3927 * into "normal" file space in order to circumvent restrictions
3928 * imposed in attribute space.
3930 if ((szp->z_phys->zp_flags & ZFS_XATTR) !=
3931 (dzp->z_phys->zp_flags & ZFS_XATTR)) {
3937 owner = zfs_fuid_map_id(zfsvfs, szp->z_phys->zp_uid, cr, ZFS_OWNER);
3938 if (owner != crgetuid(cr) &&
3939 secpolicy_basic_link(svp, cr) != 0) {
3944 if (error = zfs_zaccess(dzp, ACE_ADD_FILE, 0, B_FALSE, cr)) {
3951 * Attempt to lock directory; fail if entry already exists.
3953 error = zfs_dirent_lock(&dl, dzp, name, &tzp, zf, NULL, NULL);
3959 tx = dmu_tx_create(zfsvfs->z_os);
3960 dmu_tx_hold_bonus(tx, szp->z_id);
3961 dmu_tx_hold_zap(tx, dzp->z_id, TRUE, name);
3962 error = dmu_tx_assign(tx, TXG_NOWAIT);
3964 zfs_dirent_unlock(dl);
3965 if (error == ERESTART) {
3975 error = zfs_link_create(dl, szp, tx, 0);
3978 uint64_t txtype = TX_LINK;
3979 if (flags & FIGNORECASE)
3981 zfs_log_link(zilog, tx, txtype, dzp, szp, name);
3986 zfs_dirent_unlock(dl);
3989 vnevent_link(svp, ct);
3998 zfs_inactive(vnode_t *vp, cred_t *cr, caller_context_t *ct)
4000 znode_t *zp = VTOZ(vp);
4001 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
4004 rw_enter(&zfsvfs->z_teardown_inactive_lock, RW_READER);
4005 if (zp->z_dbuf == NULL) {
4007 * The fs has been unmounted, or we did a
4008 * suspend/resume and this file no longer exists.
4011 vp->v_count = 0; /* count arrives as 1 */
4013 vrecycle(vp, curthread);
4014 rw_exit(&zfsvfs->z_teardown_inactive_lock);
4018 if (zp->z_atime_dirty && zp->z_unlinked == 0) {
4019 dmu_tx_t *tx = dmu_tx_create(zfsvfs->z_os);
4021 dmu_tx_hold_bonus(tx, zp->z_id);
4022 error = dmu_tx_assign(tx, TXG_WAIT);
4026 dmu_buf_will_dirty(zp->z_dbuf, tx);
4027 mutex_enter(&zp->z_lock);
4028 zp->z_atime_dirty = 0;
4029 mutex_exit(&zp->z_lock);
4035 rw_exit(&zfsvfs->z_teardown_inactive_lock);
4038 CTASSERT(sizeof(struct zfid_short) <= sizeof(struct fid));
4039 CTASSERT(sizeof(struct zfid_long) <= sizeof(struct fid));
4043 zfs_fid(vnode_t *vp, fid_t *fidp, caller_context_t *ct)
4045 znode_t *zp = VTOZ(vp);
4046 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
4048 uint64_t object = zp->z_id;
4054 gen = (uint32_t)zp->z_gen;
4056 size = (zfsvfs->z_parent != zfsvfs) ? LONG_FID_LEN : SHORT_FID_LEN;
4057 fidp->fid_len = size;
4059 zfid = (zfid_short_t *)fidp;
4061 zfid->zf_len = size;
4063 for (i = 0; i < sizeof (zfid->zf_object); i++)
4064 zfid->zf_object[i] = (uint8_t)(object >> (8 * i));
4066 /* Must have a non-zero generation number to distinguish from .zfs */
4069 for (i = 0; i < sizeof (zfid->zf_gen); i++)
4070 zfid->zf_gen[i] = (uint8_t)(gen >> (8 * i));
4072 if (size == LONG_FID_LEN) {
4073 uint64_t objsetid = dmu_objset_id(zfsvfs->z_os);
4076 zlfid = (zfid_long_t *)fidp;
4078 for (i = 0; i < sizeof (zlfid->zf_setid); i++)
4079 zlfid->zf_setid[i] = (uint8_t)(objsetid >> (8 * i));
4081 /* XXX - this should be the generation number for the objset */
4082 for (i = 0; i < sizeof (zlfid->zf_setgen); i++)
4083 zlfid->zf_setgen[i] = 0;
4091 zfs_pathconf(vnode_t *vp, int cmd, ulong_t *valp, cred_t *cr,
4092 caller_context_t *ct)
4104 case _PC_FILESIZEBITS:
4109 case _PC_XATTR_EXISTS:
4111 zfsvfs = zp->z_zfsvfs;
4115 error = zfs_dirent_lock(&dl, zp, "", &xzp,
4116 ZXATTR | ZEXISTS | ZSHARED, NULL, NULL);
4118 zfs_dirent_unlock(dl);
4119 if (!zfs_dirempty(xzp))
4122 } else if (error == ENOENT) {
4124 * If there aren't extended attributes, it's the
4125 * same as having zero of them.
4133 case _PC_ACL_EXTENDED:
4141 case _PC_ACL_PATH_MAX:
4142 *valp = ACL_MAX_ENTRIES;
4145 case _PC_MIN_HOLE_SIZE:
4146 *valp = (int)SPA_MINBLOCKSIZE;
4150 return (EOPNOTSUPP);
4156 zfs_getsecattr(vnode_t *vp, vsecattr_t *vsecp, int flag, cred_t *cr,
4157 caller_context_t *ct)
4159 znode_t *zp = VTOZ(vp);
4160 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
4162 boolean_t skipaclchk = (flag & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE;
4166 error = zfs_getacl(zp, vsecp, skipaclchk, cr);
4174 zfs_setsecattr(vnode_t *vp, vsecattr_t *vsecp, int flag, cred_t *cr,
4175 caller_context_t *ct)
4177 znode_t *zp = VTOZ(vp);
4178 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
4180 boolean_t skipaclchk = (flag & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE;
4184 error = zfs_setacl(zp, vsecp, skipaclchk, cr);
4190 ioflags(int ioflags)
4194 if (ioflags & IO_APPEND)
4196 if (ioflags & IO_NDELAY)
4198 if (ioflags & IO_SYNC)
4199 flags |= (FSYNC | FDSYNC | FRSYNC);
4205 zfs_getpages(struct vnode *vp, vm_page_t *m, int count, int reqpage)
4207 znode_t *zp = VTOZ(vp);
4208 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
4209 objset_t *os = zp->z_zfsvfs->z_os;
4220 pcount = round_page(count) / PAGE_SIZE;
4222 object = mreq->object;
4225 KASSERT(vp->v_object == object, ("mismatching object"));
4227 VM_OBJECT_LOCK(object);
4228 vm_page_lock_queues();
4229 for (i = 0; i < pcount; i++) {
4234 vm_page_unlock_queues();
4237 if (mreq->valid != VM_PAGE_BITS_ALL)
4238 vm_page_zero_invalid(mreq, TRUE);
4239 VM_OBJECT_UNLOCK(object);
4241 return (VM_PAGER_OK);
4244 PCPU_INC(cnt.v_vnodein);
4245 PCPU_INC(cnt.v_vnodepgsin);
4247 if (IDX_TO_OFF(mreq->pindex) >= object->un_pager.vnp.vnp_size) {
4248 VM_OBJECT_UNLOCK(object);
4250 return (VM_PAGER_BAD);
4254 if (IDX_TO_OFF(mreq->pindex) + size > object->un_pager.vnp.vnp_size)
4255 size = object->un_pager.vnp.vnp_size - IDX_TO_OFF(mreq->pindex);
4257 VM_OBJECT_UNLOCK(object);
4259 va = zfs_map_page(mreq, &sf);
4260 error = dmu_read(os, zp->z_id, IDX_TO_OFF(mreq->pindex),
4261 size, va, DMU_READ_PREFETCH);
4262 if (size != PAGE_SIZE)
4263 bzero(va + size, PAGE_SIZE - size);
4266 VM_OBJECT_LOCK(object);
4269 mreq->valid = VM_PAGE_BITS_ALL;
4270 KASSERT(mreq->dirty == 0, ("zfs_getpages: page %p is dirty", mreq));
4272 VM_OBJECT_UNLOCK(object);
4274 ZFS_ACCESSTIME_STAMP(zfsvfs, zp);
4276 return (error ? VM_PAGER_ERROR : VM_PAGER_OK);
4280 zfs_freebsd_getpages(ap)
4281 struct vop_getpages_args /* {
4286 vm_ooffset_t a_offset;
4290 return (zfs_getpages(ap->a_vp, ap->a_m, ap->a_count, ap->a_reqpage));
4294 zfs_freebsd_open(ap)
4295 struct vop_open_args /* {
4298 struct ucred *a_cred;
4299 struct thread *a_td;
4302 vnode_t *vp = ap->a_vp;
4303 znode_t *zp = VTOZ(vp);
4306 error = zfs_open(&vp, ap->a_mode, ap->a_cred, NULL);
4308 vnode_create_vobject(vp, zp->z_phys->zp_size, ap->a_td);
4313 zfs_freebsd_close(ap)
4314 struct vop_close_args /* {
4317 struct ucred *a_cred;
4318 struct thread *a_td;
4322 return (zfs_close(ap->a_vp, ap->a_fflag, 0, 0, ap->a_cred, NULL));
4326 zfs_freebsd_ioctl(ap)
4327 struct vop_ioctl_args /* {
4337 return (zfs_ioctl(ap->a_vp, ap->a_command, (intptr_t)ap->a_data,
4338 ap->a_fflag, ap->a_cred, NULL, NULL));
4342 zfs_freebsd_read(ap)
4343 struct vop_read_args /* {
4347 struct ucred *a_cred;
4351 return (zfs_read(ap->a_vp, ap->a_uio, ioflags(ap->a_ioflag),
4356 zfs_freebsd_write(ap)
4357 struct vop_write_args /* {
4361 struct ucred *a_cred;
4365 return (zfs_write(ap->a_vp, ap->a_uio, ioflags(ap->a_ioflag),
4370 zfs_freebsd_access(ap)
4371 struct vop_access_args /* {
4373 accmode_t a_accmode;
4374 struct ucred *a_cred;
4375 struct thread *a_td;
4382 * ZFS itself only knowns about VREAD, VWRITE, VEXEC and VAPPEND,
4384 accmode = ap->a_accmode & (VREAD|VWRITE|VEXEC|VAPPEND);
4386 error = zfs_access(ap->a_vp, accmode, 0, ap->a_cred, NULL);
4389 * VADMIN has to be handled by vaccess().
4392 accmode = ap->a_accmode & ~(VREAD|VWRITE|VEXEC|VAPPEND);
4394 vnode_t *vp = ap->a_vp;
4395 znode_t *zp = VTOZ(vp);
4396 znode_phys_t *zphys = zp->z_phys;
4398 error = vaccess(vp->v_type, zphys->zp_mode,
4399 zphys->zp_uid, zphys->zp_gid, accmode, ap->a_cred,
4408 zfs_freebsd_lookup(ap)
4409 struct vop_lookup_args /* {
4410 struct vnode *a_dvp;
4411 struct vnode **a_vpp;
4412 struct componentname *a_cnp;
4415 struct componentname *cnp = ap->a_cnp;
4416 char nm[NAME_MAX + 1];
4418 ASSERT(cnp->cn_namelen < sizeof(nm));
4419 strlcpy(nm, cnp->cn_nameptr, MIN(cnp->cn_namelen + 1, sizeof(nm)));
4421 return (zfs_lookup(ap->a_dvp, nm, ap->a_vpp, cnp, cnp->cn_nameiop,
4422 cnp->cn_cred, cnp->cn_thread, 0));
4426 zfs_freebsd_create(ap)
4427 struct vop_create_args /* {
4428 struct vnode *a_dvp;
4429 struct vnode **a_vpp;
4430 struct componentname *a_cnp;
4431 struct vattr *a_vap;
4434 struct componentname *cnp = ap->a_cnp;
4435 vattr_t *vap = ap->a_vap;
4438 ASSERT(cnp->cn_flags & SAVENAME);
4440 vattr_init_mask(vap);
4441 mode = vap->va_mode & ALLPERMS;
4443 return (zfs_create(ap->a_dvp, cnp->cn_nameptr, vap, !EXCL, mode,
4444 ap->a_vpp, cnp->cn_cred, cnp->cn_thread));
4448 zfs_freebsd_remove(ap)
4449 struct vop_remove_args /* {
4450 struct vnode *a_dvp;
4452 struct componentname *a_cnp;
4456 ASSERT(ap->a_cnp->cn_flags & SAVENAME);
4458 return (zfs_remove(ap->a_dvp, ap->a_cnp->cn_nameptr,
4459 ap->a_cnp->cn_cred, NULL, 0));
4463 zfs_freebsd_mkdir(ap)
4464 struct vop_mkdir_args /* {
4465 struct vnode *a_dvp;
4466 struct vnode **a_vpp;
4467 struct componentname *a_cnp;
4468 struct vattr *a_vap;
4471 vattr_t *vap = ap->a_vap;
4473 ASSERT(ap->a_cnp->cn_flags & SAVENAME);
4475 vattr_init_mask(vap);
4477 return (zfs_mkdir(ap->a_dvp, ap->a_cnp->cn_nameptr, vap, ap->a_vpp,
4478 ap->a_cnp->cn_cred, NULL, 0, NULL));
4482 zfs_freebsd_rmdir(ap)
4483 struct vop_rmdir_args /* {
4484 struct vnode *a_dvp;
4486 struct componentname *a_cnp;
4489 struct componentname *cnp = ap->a_cnp;
4491 ASSERT(cnp->cn_flags & SAVENAME);
4493 return (zfs_rmdir(ap->a_dvp, cnp->cn_nameptr, NULL, cnp->cn_cred, NULL, 0));
4497 zfs_freebsd_readdir(ap)
4498 struct vop_readdir_args /* {
4501 struct ucred *a_cred;
4508 return (zfs_readdir(ap->a_vp, ap->a_uio, ap->a_cred, ap->a_eofflag,
4509 ap->a_ncookies, ap->a_cookies));
4513 zfs_freebsd_fsync(ap)
4514 struct vop_fsync_args /* {
4517 struct thread *a_td;
4522 return (zfs_fsync(ap->a_vp, 0, ap->a_td->td_ucred, NULL));
4526 zfs_freebsd_getattr(ap)
4527 struct vop_getattr_args /* {
4529 struct vattr *a_vap;
4530 struct ucred *a_cred;
4531 struct thread *a_td;
4534 vattr_t *vap = ap->a_vap;
4540 xvap.xva_vattr = *vap;
4541 xvap.xva_vattr.va_mask |= AT_XVATTR;
4543 /* Convert chflags into ZFS-type flags. */
4544 /* XXX: what about SF_SETTABLE?. */
4545 XVA_SET_REQ(&xvap, XAT_IMMUTABLE);
4546 XVA_SET_REQ(&xvap, XAT_APPENDONLY);
4547 XVA_SET_REQ(&xvap, XAT_NOUNLINK);
4548 XVA_SET_REQ(&xvap, XAT_NODUMP);
4549 error = zfs_getattr(ap->a_vp, (vattr_t *)&xvap, 0, ap->a_cred, NULL);
4553 /* Convert ZFS xattr into chflags. */
4554 #define FLAG_CHECK(fflag, xflag, xfield) do { \
4555 if (XVA_ISSET_RTN(&xvap, (xflag)) && (xfield) != 0) \
4556 fflags |= (fflag); \
4558 FLAG_CHECK(SF_IMMUTABLE, XAT_IMMUTABLE,
4559 xvap.xva_xoptattrs.xoa_immutable);
4560 FLAG_CHECK(SF_APPEND, XAT_APPENDONLY,
4561 xvap.xva_xoptattrs.xoa_appendonly);
4562 FLAG_CHECK(SF_NOUNLINK, XAT_NOUNLINK,
4563 xvap.xva_xoptattrs.xoa_nounlink);
4564 FLAG_CHECK(UF_NODUMP, XAT_NODUMP,
4565 xvap.xva_xoptattrs.xoa_nodump);
4567 *vap = xvap.xva_vattr;
4568 vap->va_flags = fflags;
4573 zfs_freebsd_setattr(ap)
4574 struct vop_setattr_args /* {
4576 struct vattr *a_vap;
4577 struct ucred *a_cred;
4578 struct thread *a_td;
4581 vnode_t *vp = ap->a_vp;
4582 vattr_t *vap = ap->a_vap;
4583 cred_t *cred = ap->a_cred;
4588 vattr_init_mask(vap);
4589 vap->va_mask &= ~AT_NOSET;
4592 xvap.xva_vattr = *vap;
4594 zflags = VTOZ(vp)->z_phys->zp_flags;
4596 if (vap->va_flags != VNOVAL) {
4597 zfsvfs_t *zfsvfs = VTOZ(vp)->z_zfsvfs;
4600 if (zfsvfs->z_use_fuids == B_FALSE)
4601 return (EOPNOTSUPP);
4603 fflags = vap->va_flags;
4604 if ((fflags & ~(SF_IMMUTABLE|SF_APPEND|SF_NOUNLINK|UF_NODUMP)) != 0)
4605 return (EOPNOTSUPP);
4607 * Unprivileged processes are not permitted to unset system
4608 * flags, or modify flags if any system flags are set.
4609 * Privileged non-jail processes may not modify system flags
4610 * if securelevel > 0 and any existing system flags are set.
4611 * Privileged jail processes behave like privileged non-jail
4612 * processes if the security.jail.chflags_allowed sysctl is
4613 * is non-zero; otherwise, they behave like unprivileged
4616 if (secpolicy_fs_owner(vp->v_mount, cred) == 0 ||
4617 priv_check_cred(cred, PRIV_VFS_SYSFLAGS, 0) == 0) {
4619 (ZFS_IMMUTABLE | ZFS_APPENDONLY | ZFS_NOUNLINK)) {
4620 error = securelevel_gt(cred, 0);
4626 * Callers may only modify the file flags on objects they
4627 * have VADMIN rights for.
4629 if ((error = VOP_ACCESS(vp, VADMIN, cred, curthread)) != 0)
4632 (ZFS_IMMUTABLE | ZFS_APPENDONLY | ZFS_NOUNLINK)) {
4636 (SF_IMMUTABLE | SF_APPEND | SF_NOUNLINK)) {
4641 #define FLAG_CHANGE(fflag, zflag, xflag, xfield) do { \
4642 if (((fflags & (fflag)) && !(zflags & (zflag))) || \
4643 ((zflags & (zflag)) && !(fflags & (fflag)))) { \
4644 XVA_SET_REQ(&xvap, (xflag)); \
4645 (xfield) = ((fflags & (fflag)) != 0); \
4648 /* Convert chflags into ZFS-type flags. */
4649 /* XXX: what about SF_SETTABLE?. */
4650 FLAG_CHANGE(SF_IMMUTABLE, ZFS_IMMUTABLE, XAT_IMMUTABLE,
4651 xvap.xva_xoptattrs.xoa_immutable);
4652 FLAG_CHANGE(SF_APPEND, ZFS_APPENDONLY, XAT_APPENDONLY,
4653 xvap.xva_xoptattrs.xoa_appendonly);
4654 FLAG_CHANGE(SF_NOUNLINK, ZFS_NOUNLINK, XAT_NOUNLINK,
4655 xvap.xva_xoptattrs.xoa_nounlink);
4656 FLAG_CHANGE(UF_NODUMP, ZFS_NODUMP, XAT_NODUMP,
4657 xvap.xva_xoptattrs.xoa_nodump);
4660 return (zfs_setattr(vp, (vattr_t *)&xvap, 0, cred, NULL));
4664 zfs_freebsd_rename(ap)
4665 struct vop_rename_args /* {
4666 struct vnode *a_fdvp;
4667 struct vnode *a_fvp;
4668 struct componentname *a_fcnp;
4669 struct vnode *a_tdvp;
4670 struct vnode *a_tvp;
4671 struct componentname *a_tcnp;
4674 vnode_t *fdvp = ap->a_fdvp;
4675 vnode_t *fvp = ap->a_fvp;
4676 vnode_t *tdvp = ap->a_tdvp;
4677 vnode_t *tvp = ap->a_tvp;
4680 ASSERT(ap->a_fcnp->cn_flags & (SAVENAME|SAVESTART));
4681 ASSERT(ap->a_tcnp->cn_flags & (SAVENAME|SAVESTART));
4683 error = zfs_rename(fdvp, ap->a_fcnp->cn_nameptr, tdvp,
4684 ap->a_tcnp->cn_nameptr, ap->a_fcnp->cn_cred, NULL, 0);
4699 zfs_freebsd_symlink(ap)
4700 struct vop_symlink_args /* {
4701 struct vnode *a_dvp;
4702 struct vnode **a_vpp;
4703 struct componentname *a_cnp;
4704 struct vattr *a_vap;
4708 struct componentname *cnp = ap->a_cnp;
4709 vattr_t *vap = ap->a_vap;
4711 ASSERT(cnp->cn_flags & SAVENAME);
4713 vap->va_type = VLNK; /* FreeBSD: Syscall only sets va_mode. */
4714 vattr_init_mask(vap);
4716 return (zfs_symlink(ap->a_dvp, ap->a_vpp, cnp->cn_nameptr, vap,
4717 ap->a_target, cnp->cn_cred, cnp->cn_thread));
4721 zfs_freebsd_readlink(ap)
4722 struct vop_readlink_args /* {
4725 struct ucred *a_cred;
4729 return (zfs_readlink(ap->a_vp, ap->a_uio, ap->a_cred, NULL));
4733 zfs_freebsd_link(ap)
4734 struct vop_link_args /* {
4735 struct vnode *a_tdvp;
4737 struct componentname *a_cnp;
4740 struct componentname *cnp = ap->a_cnp;
4742 ASSERT(cnp->cn_flags & SAVENAME);
4744 return (zfs_link(ap->a_tdvp, ap->a_vp, cnp->cn_nameptr, cnp->cn_cred, NULL, 0));
4748 zfs_freebsd_inactive(ap)
4749 struct vop_inactive_args /* {
4751 struct thread *a_td;
4754 vnode_t *vp = ap->a_vp;
4756 zfs_inactive(vp, ap->a_td->td_ucred, NULL);
4761 zfs_reclaim_complete(void *arg, int pending)
4764 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
4766 rw_enter(&zfsvfs->z_teardown_inactive_lock, RW_READER);
4767 if (zp->z_dbuf != NULL) {
4768 ZFS_OBJ_HOLD_ENTER(zfsvfs, zp->z_id);
4769 zfs_znode_dmu_fini(zp);
4770 ZFS_OBJ_HOLD_EXIT(zfsvfs, zp->z_id);
4773 rw_exit(&zfsvfs->z_teardown_inactive_lock);
4775 * If the file system is being unmounted, there is a process waiting
4776 * for us, wake it up.
4778 if (zfsvfs->z_unmounted)
4783 zfs_freebsd_reclaim(ap)
4784 struct vop_reclaim_args /* {
4786 struct thread *a_td;
4789 vnode_t *vp = ap->a_vp;
4790 znode_t *zp = VTOZ(vp);
4791 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
4793 rw_enter(&zfsvfs->z_teardown_inactive_lock, RW_READER);
4798 * Destroy the vm object and flush associated pages.
4800 vnode_destroy_vobject(vp);
4802 mutex_enter(&zp->z_lock);
4803 ASSERT(zp->z_phys != NULL);
4805 mutex_exit(&zp->z_lock);
4809 else if (zp->z_dbuf == NULL)
4811 else /* if (!zp->z_unlinked && zp->z_dbuf != NULL) */ {
4814 locked = MUTEX_HELD(ZFS_OBJ_MUTEX(zfsvfs, zp->z_id)) ? 2 :
4815 ZFS_OBJ_HOLD_TRYENTER(zfsvfs, zp->z_id);
4818 * Lock can't be obtained due to deadlock possibility,
4819 * so defer znode destruction.
4821 TASK_INIT(&zp->z_task, 0, zfs_reclaim_complete, zp);
4822 taskqueue_enqueue(taskqueue_thread, &zp->z_task);
4824 zfs_znode_dmu_fini(zp);
4826 ZFS_OBJ_HOLD_EXIT(zfsvfs, zp->z_id);
4832 ASSERT(vp->v_holdcnt >= 1);
4834 rw_exit(&zfsvfs->z_teardown_inactive_lock);
4840 struct vop_fid_args /* {
4846 return (zfs_fid(ap->a_vp, (void *)ap->a_fid, NULL));
4850 zfs_freebsd_pathconf(ap)
4851 struct vop_pathconf_args /* {
4854 register_t *a_retval;
4860 error = zfs_pathconf(ap->a_vp, ap->a_name, &val, curthread->td_ucred, NULL);
4862 *ap->a_retval = val;
4863 else if (error == EOPNOTSUPP)
4864 error = vop_stdpathconf(ap);
4869 zfs_freebsd_fifo_pathconf(ap)
4870 struct vop_pathconf_args /* {
4873 register_t *a_retval;
4877 switch (ap->a_name) {
4878 case _PC_ACL_EXTENDED:
4880 case _PC_ACL_PATH_MAX:
4881 case _PC_MAC_PRESENT:
4882 return (zfs_freebsd_pathconf(ap));
4884 return (fifo_specops.vop_pathconf(ap));
4889 * FreeBSD's extended attributes namespace defines file name prefix for ZFS'
4890 * extended attribute name:
4893 * system freebsd:system:
4894 * user (none, can be used to access ZFS fsattr(5) attributes
4895 * created on Solaris)
4898 zfs_create_attrname(int attrnamespace, const char *name, char *attrname,
4901 const char *namespace, *prefix, *suffix;
4903 /* We don't allow '/' character in attribute name. */
4904 if (strchr(name, '/') != NULL)
4906 /* We don't allow attribute names that start with "freebsd:" string. */
4907 if (strncmp(name, "freebsd:", 8) == 0)
4910 bzero(attrname, size);
4912 switch (attrnamespace) {
4913 case EXTATTR_NAMESPACE_USER:
4915 prefix = "freebsd:";
4916 namespace = EXTATTR_NAMESPACE_USER_STRING;
4920 * This is the default namespace by which we can access all
4921 * attributes created on Solaris.
4923 prefix = namespace = suffix = "";
4926 case EXTATTR_NAMESPACE_SYSTEM:
4927 prefix = "freebsd:";
4928 namespace = EXTATTR_NAMESPACE_SYSTEM_STRING;
4931 case EXTATTR_NAMESPACE_EMPTY:
4935 if (snprintf(attrname, size, "%s%s%s%s", prefix, namespace, suffix,
4937 return (ENAMETOOLONG);
4943 * Vnode operating to retrieve a named extended attribute.
4946 zfs_getextattr(struct vop_getextattr_args *ap)
4949 IN struct vnode *a_vp;
4950 IN int a_attrnamespace;
4951 IN const char *a_name;
4952 INOUT struct uio *a_uio;
4954 IN struct ucred *a_cred;
4955 IN struct thread *a_td;
4959 zfsvfs_t *zfsvfs = VTOZ(ap->a_vp)->z_zfsvfs;
4960 struct thread *td = ap->a_td;
4961 struct nameidata nd;
4964 vnode_t *xvp = NULL, *vp;
4967 error = extattr_check_cred(ap->a_vp, ap->a_attrnamespace,
4968 ap->a_cred, ap->a_td, VREAD);
4972 error = zfs_create_attrname(ap->a_attrnamespace, ap->a_name, attrname,
4979 error = zfs_lookup(ap->a_vp, NULL, &xvp, NULL, 0, ap->a_cred, td,
4987 NDINIT_ATVP(&nd, LOOKUP, NOFOLLOW | MPSAFE, UIO_SYSSPACE, attrname,
4989 error = vn_open_cred(&nd, &flags, 0, 0, ap->a_cred, NULL);
4991 NDFREE(&nd, NDF_ONLY_PNBUF);
4994 if (error == ENOENT)
4999 if (ap->a_size != NULL) {
5000 error = VOP_GETATTR(vp, &va, ap->a_cred);
5002 *ap->a_size = (size_t)va.va_size;
5003 } else if (ap->a_uio != NULL)
5004 error = VOP_READ(vp, ap->a_uio, IO_UNIT | IO_SYNC, ap->a_cred);
5007 vn_close(vp, flags, ap->a_cred, td);
5014 * Vnode operation to remove a named attribute.
5017 zfs_deleteextattr(struct vop_deleteextattr_args *ap)
5020 IN struct vnode *a_vp;
5021 IN int a_attrnamespace;
5022 IN const char *a_name;
5023 IN struct ucred *a_cred;
5024 IN struct thread *a_td;
5028 zfsvfs_t *zfsvfs = VTOZ(ap->a_vp)->z_zfsvfs;
5029 struct thread *td = ap->a_td;
5030 struct nameidata nd;
5033 vnode_t *xvp = NULL, *vp;
5036 error = extattr_check_cred(ap->a_vp, ap->a_attrnamespace,
5037 ap->a_cred, ap->a_td, VWRITE);
5041 error = zfs_create_attrname(ap->a_attrnamespace, ap->a_name, attrname,
5048 error = zfs_lookup(ap->a_vp, NULL, &xvp, NULL, 0, ap->a_cred, td,
5055 NDINIT_ATVP(&nd, DELETE, NOFOLLOW | LOCKPARENT | LOCKLEAF | MPSAFE,
5056 UIO_SYSSPACE, attrname, xvp, td);
5059 NDFREE(&nd, NDF_ONLY_PNBUF);
5062 if (error == ENOENT)
5066 error = VOP_REMOVE(nd.ni_dvp, vp, &nd.ni_cnd);
5069 if (vp == nd.ni_dvp)
5079 * Vnode operation to set a named attribute.
5082 zfs_setextattr(struct vop_setextattr_args *ap)
5085 IN struct vnode *a_vp;
5086 IN int a_attrnamespace;
5087 IN const char *a_name;
5088 INOUT struct uio *a_uio;
5089 IN struct ucred *a_cred;
5090 IN struct thread *a_td;
5094 zfsvfs_t *zfsvfs = VTOZ(ap->a_vp)->z_zfsvfs;
5095 struct thread *td = ap->a_td;
5096 struct nameidata nd;
5099 vnode_t *xvp = NULL, *vp;
5102 error = extattr_check_cred(ap->a_vp, ap->a_attrnamespace,
5103 ap->a_cred, ap->a_td, VWRITE);
5107 error = zfs_create_attrname(ap->a_attrnamespace, ap->a_name, attrname,
5114 error = zfs_lookup(ap->a_vp, NULL, &xvp, NULL, 0, ap->a_cred, td,
5115 LOOKUP_XATTR | CREATE_XATTR_DIR);
5121 flags = FFLAGS(O_WRONLY | O_CREAT);
5122 NDINIT_ATVP(&nd, LOOKUP, NOFOLLOW | MPSAFE, UIO_SYSSPACE, attrname,
5124 error = vn_open_cred(&nd, &flags, 0600, 0, ap->a_cred, NULL);
5126 NDFREE(&nd, NDF_ONLY_PNBUF);
5134 error = VOP_SETATTR(vp, &va, ap->a_cred);
5136 VOP_WRITE(vp, ap->a_uio, IO_UNIT | IO_SYNC, ap->a_cred);
5139 vn_close(vp, flags, ap->a_cred, td);
5146 * Vnode operation to retrieve extended attributes on a vnode.
5149 zfs_listextattr(struct vop_listextattr_args *ap)
5152 IN struct vnode *a_vp;
5153 IN int a_attrnamespace;
5154 INOUT struct uio *a_uio;
5156 IN struct ucred *a_cred;
5157 IN struct thread *a_td;
5161 zfsvfs_t *zfsvfs = VTOZ(ap->a_vp)->z_zfsvfs;
5162 struct thread *td = ap->a_td;
5163 struct nameidata nd;
5164 char attrprefix[16];
5165 u_char dirbuf[sizeof(struct dirent)];
5168 struct uio auio, *uio = ap->a_uio;
5169 size_t *sizep = ap->a_size;
5171 vnode_t *xvp = NULL, *vp;
5172 int done, error, eof, pos;
5174 error = extattr_check_cred(ap->a_vp, ap->a_attrnamespace,
5175 ap->a_cred, ap->a_td, VREAD);
5179 error = zfs_create_attrname(ap->a_attrnamespace, "", attrprefix,
5180 sizeof(attrprefix));
5183 plen = strlen(attrprefix);
5190 error = zfs_lookup(ap->a_vp, NULL, &xvp, NULL, 0, ap->a_cred, td,
5195 * ENOATTR means that the EA directory does not yet exist,
5196 * i.e. there are no extended attributes there.
5198 if (error == ENOATTR)
5203 NDINIT_ATVP(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | LOCKSHARED | MPSAFE,
5204 UIO_SYSSPACE, ".", xvp, td);
5207 NDFREE(&nd, NDF_ONLY_PNBUF);
5213 auio.uio_iov = &aiov;
5214 auio.uio_iovcnt = 1;
5215 auio.uio_segflg = UIO_SYSSPACE;
5217 auio.uio_rw = UIO_READ;
5218 auio.uio_offset = 0;
5223 aiov.iov_base = (void *)dirbuf;
5224 aiov.iov_len = sizeof(dirbuf);
5225 auio.uio_resid = sizeof(dirbuf);
5226 error = VOP_READDIR(vp, &auio, ap->a_cred, &eof, NULL, NULL);
5227 done = sizeof(dirbuf) - auio.uio_resid;
5230 for (pos = 0; pos < done;) {
5231 dp = (struct dirent *)(dirbuf + pos);
5232 pos += dp->d_reclen;
5234 * XXX: Temporarily we also accept DT_UNKNOWN, as this
5235 * is what we get when attribute was created on Solaris.
5237 if (dp->d_type != DT_REG && dp->d_type != DT_UNKNOWN)
5239 if (plen == 0 && strncmp(dp->d_name, "freebsd:", 8) == 0)
5241 else if (strncmp(dp->d_name, attrprefix, plen) != 0)
5243 nlen = dp->d_namlen - plen;
5246 else if (uio != NULL) {
5248 * Format of extattr name entry is one byte for
5249 * length and the rest for name.
5251 error = uiomove(&nlen, 1, uio->uio_rw, uio);
5253 error = uiomove(dp->d_name + plen, nlen,
5260 } while (!eof && error == 0);
5269 zfs_freebsd_getacl(ap)
5270 struct vop_getacl_args /* {
5279 vsecattr_t vsecattr;
5281 if (ap->a_type != ACL_TYPE_NFS4)
5284 vsecattr.vsa_mask = VSA_ACE | VSA_ACECNT;
5285 if (error = zfs_getsecattr(ap->a_vp, &vsecattr, 0, ap->a_cred, NULL))
5288 error = acl_from_aces(ap->a_aclp, vsecattr.vsa_aclentp, vsecattr.vsa_aclcnt);
5289 if (vsecattr.vsa_aclentp != NULL)
5290 kmem_free(vsecattr.vsa_aclentp, vsecattr.vsa_aclentsz);
5296 zfs_freebsd_setacl(ap)
5297 struct vop_setacl_args /* {
5306 vsecattr_t vsecattr;
5307 int aclbsize; /* size of acl list in bytes */
5310 if (ap->a_type != ACL_TYPE_NFS4)
5313 if (ap->a_aclp->acl_cnt < 1 || ap->a_aclp->acl_cnt > MAX_ACL_ENTRIES)
5317 * With NFSv4 ACLs, chmod(2) may need to add additional entries,
5318 * splitting every entry into two and appending "canonical six"
5319 * entries at the end. Don't allow for setting an ACL that would
5320 * cause chmod(2) to run out of ACL entries.
5322 if (ap->a_aclp->acl_cnt * 2 + 6 > ACL_MAX_ENTRIES)
5325 error = acl_nfs4_check(ap->a_aclp, ap->a_vp->v_type == VDIR);
5329 vsecattr.vsa_mask = VSA_ACE;
5330 aclbsize = ap->a_aclp->acl_cnt * sizeof(ace_t);
5331 vsecattr.vsa_aclentp = kmem_alloc(aclbsize, KM_SLEEP);
5332 aaclp = vsecattr.vsa_aclentp;
5333 vsecattr.vsa_aclentsz = aclbsize;
5335 aces_from_acl(vsecattr.vsa_aclentp, &vsecattr.vsa_aclcnt, ap->a_aclp);
5336 error = zfs_setsecattr(ap->a_vp, &vsecattr, 0, ap->a_cred, NULL);
5337 kmem_free(aaclp, aclbsize);
5343 zfs_freebsd_aclcheck(ap)
5344 struct vop_aclcheck_args /* {
5353 return (EOPNOTSUPP);
5356 struct vop_vector zfs_vnodeops;
5357 struct vop_vector zfs_fifoops;
5358 struct vop_vector zfs_shareops;
5360 struct vop_vector zfs_vnodeops = {
5361 .vop_default = &default_vnodeops,
5362 .vop_inactive = zfs_freebsd_inactive,
5363 .vop_reclaim = zfs_freebsd_reclaim,
5364 .vop_access = zfs_freebsd_access,
5365 #ifdef FREEBSD_NAMECACHE
5366 .vop_lookup = vfs_cache_lookup,
5367 .vop_cachedlookup = zfs_freebsd_lookup,
5369 .vop_lookup = zfs_freebsd_lookup,
5371 .vop_getattr = zfs_freebsd_getattr,
5372 .vop_setattr = zfs_freebsd_setattr,
5373 .vop_create = zfs_freebsd_create,
5374 .vop_mknod = zfs_freebsd_create,
5375 .vop_mkdir = zfs_freebsd_mkdir,
5376 .vop_readdir = zfs_freebsd_readdir,
5377 .vop_fsync = zfs_freebsd_fsync,
5378 .vop_open = zfs_freebsd_open,
5379 .vop_close = zfs_freebsd_close,
5380 .vop_rmdir = zfs_freebsd_rmdir,
5381 .vop_ioctl = zfs_freebsd_ioctl,
5382 .vop_link = zfs_freebsd_link,
5383 .vop_symlink = zfs_freebsd_symlink,
5384 .vop_readlink = zfs_freebsd_readlink,
5385 .vop_read = zfs_freebsd_read,
5386 .vop_write = zfs_freebsd_write,
5387 .vop_remove = zfs_freebsd_remove,
5388 .vop_rename = zfs_freebsd_rename,
5389 .vop_pathconf = zfs_freebsd_pathconf,
5390 .vop_bmap = VOP_EOPNOTSUPP,
5391 .vop_fid = zfs_freebsd_fid,
5392 .vop_getextattr = zfs_getextattr,
5393 .vop_deleteextattr = zfs_deleteextattr,
5394 .vop_setextattr = zfs_setextattr,
5395 .vop_listextattr = zfs_listextattr,
5396 .vop_getacl = zfs_freebsd_getacl,
5397 .vop_setacl = zfs_freebsd_setacl,
5398 .vop_aclcheck = zfs_freebsd_aclcheck,
5399 .vop_getpages = zfs_freebsd_getpages,
5402 struct vop_vector zfs_fifoops = {
5403 .vop_default = &fifo_specops,
5404 .vop_fsync = zfs_freebsd_fsync,
5405 .vop_access = zfs_freebsd_access,
5406 .vop_getattr = zfs_freebsd_getattr,
5407 .vop_inactive = zfs_freebsd_inactive,
5408 .vop_read = VOP_PANIC,
5409 .vop_reclaim = zfs_freebsd_reclaim,
5410 .vop_setattr = zfs_freebsd_setattr,
5411 .vop_write = VOP_PANIC,
5412 .vop_pathconf = zfs_freebsd_fifo_pathconf,
5413 .vop_fid = zfs_freebsd_fid,
5414 .vop_getacl = zfs_freebsd_getacl,
5415 .vop_setacl = zfs_freebsd_setacl,
5416 .vop_aclcheck = zfs_freebsd_aclcheck,
5420 * special share hidden files vnode operations template
5422 struct vop_vector zfs_shareops = {
5423 .vop_default = &default_vnodeops,
5424 .vop_access = zfs_freebsd_access,
5425 .vop_inactive = zfs_freebsd_inactive,
5426 .vop_reclaim = zfs_freebsd_reclaim,
5427 .vop_fid = zfs_freebsd_fid,
5428 .vop_pathconf = zfs_freebsd_pathconf,