4 * The contents of this file are subject to the terms of the
5 * Common Development and Distribution License (the "License").
6 * You may not use this file except in compliance with the License.
8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9 * or http://www.opensolaris.org/os/licensing.
10 * See the License for the specific language governing permissions
11 * and limitations under the License.
13 * When distributing Covered Code, include this CDDL HEADER in each
14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15 * If applicable, add the following below this CDDL HEADER, with the
16 * fields enclosed by brackets "[]" replaced with your own identifying
17 * information: Portions Copyright [yyyy] [name of copyright owner]
22 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
25 /* Portions Copyright 2007 Jeremy Teo */
27 #include <sys/types.h>
28 #include <sys/param.h>
30 #include <sys/systm.h>
31 #include <sys/sysmacros.h>
32 #include <sys/resource.h>
34 #include <sys/vnode.h>
38 #include <sys/taskq.h>
40 #include <sys/atomic.h>
41 #include <sys/namei.h>
43 #include <sys/cmn_err.h>
44 #include <sys/errno.h>
45 #include <sys/unistd.h>
46 #include <sys/zfs_dir.h>
47 #include <sys/zfs_ioctl.h>
48 #include <sys/fs/zfs.h>
54 #include <sys/dirent.h>
55 #include <sys/policy.h>
56 #include <sys/sunddi.h>
57 #include <sys/filio.h>
59 #include <sys/zfs_ctldir.h>
60 #include <sys/zfs_fuid.h>
62 #include <sys/zfs_rlock.h>
63 #include <sys/extdirent.h>
64 #include <sys/kidmap.h>
67 #include <sys/sf_buf.h>
68 #include <sys/sched.h>
74 * Each vnode op performs some logical unit of work. To do this, the ZPL must
75 * properly lock its in-core state, create a DMU transaction, do the work,
76 * record this work in the intent log (ZIL), commit the DMU transaction,
77 * and wait for the intent log to commit if it is a synchronous operation.
78 * Moreover, the vnode ops must work in both normal and log replay context.
79 * The ordering of events is important to avoid deadlocks and references
80 * to freed memory. The example below illustrates the following Big Rules:
82 * (1) A check must be made in each zfs thread for a mounted file system.
83 * This is done avoiding races using ZFS_ENTER(zfsvfs).
84 * A ZFS_EXIT(zfsvfs) is needed before all returns. Any znodes
85 * must be checked with ZFS_VERIFY_ZP(zp). Both of these macros
86 * can return EIO from the calling function.
88 * (2) VN_RELE() should always be the last thing except for zil_commit()
89 * (if necessary) and ZFS_EXIT(). This is for 3 reasons:
90 * First, if it's the last reference, the vnode/znode
91 * can be freed, so the zp may point to freed memory. Second, the last
92 * reference will call zfs_zinactive(), which may induce a lot of work --
93 * pushing cached pages (which acquires range locks) and syncing out
94 * cached atime changes. Third, zfs_zinactive() may require a new tx,
95 * which could deadlock the system if you were already holding one.
96 * If you must call VN_RELE() within a tx then use VN_RELE_ASYNC().
98 * (3) All range locks must be grabbed before calling dmu_tx_assign(),
99 * as they can span dmu_tx_assign() calls.
101 * (4) Always pass TXG_NOWAIT as the second argument to dmu_tx_assign().
102 * This is critical because we don't want to block while holding locks.
103 * Note, in particular, that if a lock is sometimes acquired before
104 * the tx assigns, and sometimes after (e.g. z_lock), then failing to
105 * use a non-blocking assign can deadlock the system. The scenario:
107 * Thread A has grabbed a lock before calling dmu_tx_assign().
108 * Thread B is in an already-assigned tx, and blocks for this lock.
109 * Thread A calls dmu_tx_assign(TXG_WAIT) and blocks in txg_wait_open()
110 * forever, because the previous txg can't quiesce until B's tx commits.
112 * If dmu_tx_assign() returns ERESTART and zfsvfs->z_assign is TXG_NOWAIT,
113 * then drop all locks, call dmu_tx_wait(), and try again.
115 * (5) If the operation succeeded, generate the intent log entry for it
116 * before dropping locks. This ensures that the ordering of events
117 * in the intent log matches the order in which they actually occurred.
118 * During ZIL replay the zfs_log_* functions will update the sequence
119 * number to indicate the zil transaction has replayed.
121 * (6) At the end of each vnode op, the DMU tx must always commit,
122 * regardless of whether there were any errors.
124 * (7) After dropping all locks, invoke zil_commit(zilog, seq, foid)
125 * to ensure that synchronous semantics are provided when necessary.
127 * In general, this is how things should be ordered in each vnode op:
129 * ZFS_ENTER(zfsvfs); // exit if unmounted
131 * zfs_dirent_lock(&dl, ...) // lock directory entry (may VN_HOLD())
132 * rw_enter(...); // grab any other locks you need
133 * tx = dmu_tx_create(...); // get DMU tx
134 * dmu_tx_hold_*(); // hold each object you might modify
135 * error = dmu_tx_assign(tx, TXG_NOWAIT); // try to assign
137 * rw_exit(...); // drop locks
138 * zfs_dirent_unlock(dl); // unlock directory entry
139 * VN_RELE(...); // release held vnodes
140 * if (error == ERESTART) {
145 * dmu_tx_abort(tx); // abort DMU tx
146 * ZFS_EXIT(zfsvfs); // finished in zfs
147 * return (error); // really out of space
149 * error = do_real_work(); // do whatever this VOP does
151 * zfs_log_*(...); // on success, make ZIL entry
152 * dmu_tx_commit(tx); // commit DMU tx -- error or not
153 * rw_exit(...); // drop locks
154 * zfs_dirent_unlock(dl); // unlock directory entry
155 * VN_RELE(...); // release held vnodes
156 * zil_commit(zilog, seq, foid); // synchronous when necessary
157 * ZFS_EXIT(zfsvfs); // finished in zfs
158 * return (error); // done, report error
163 zfs_open(vnode_t **vpp, int flag, cred_t *cr, caller_context_t *ct)
165 znode_t *zp = VTOZ(*vpp);
166 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
171 if ((flag & FWRITE) && (zp->z_phys->zp_flags & ZFS_APPENDONLY) &&
172 ((flag & FAPPEND) == 0)) {
177 if (!zfs_has_ctldir(zp) && zp->z_zfsvfs->z_vscan &&
178 ZTOV(zp)->v_type == VREG &&
179 !(zp->z_phys->zp_flags & ZFS_AV_QUARANTINED) &&
180 zp->z_phys->zp_size > 0) {
181 if (fs_vscan(*vpp, cr, 0) != 0) {
187 /* Keep a count of the synchronous opens in the znode */
188 if (flag & (FSYNC | FDSYNC))
189 atomic_inc_32(&zp->z_sync_cnt);
197 zfs_close(vnode_t *vp, int flag, int count, offset_t offset, cred_t *cr,
198 caller_context_t *ct)
200 znode_t *zp = VTOZ(vp);
201 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
206 /* Decrement the synchronous opens in the znode */
207 if ((flag & (FSYNC | FDSYNC)) && (count == 1))
208 atomic_dec_32(&zp->z_sync_cnt);
211 * Clean up any locks held by this process on the vp.
213 cleanlocks(vp, ddi_get_pid(), 0);
214 cleanshares(vp, ddi_get_pid());
216 if (!zfs_has_ctldir(zp) && zp->z_zfsvfs->z_vscan &&
217 ZTOV(zp)->v_type == VREG &&
218 !(zp->z_phys->zp_flags & ZFS_AV_QUARANTINED) &&
219 zp->z_phys->zp_size > 0)
220 VERIFY(fs_vscan(vp, cr, 1) == 0);
227 * Lseek support for finding holes (cmd == _FIO_SEEK_HOLE) and
228 * data (cmd == _FIO_SEEK_DATA). "off" is an in/out parameter.
231 zfs_holey(vnode_t *vp, u_long cmd, offset_t *off)
233 znode_t *zp = VTOZ(vp);
234 uint64_t noff = (uint64_t)*off; /* new offset */
239 file_sz = zp->z_phys->zp_size;
240 if (noff >= file_sz) {
244 if (cmd == _FIO_SEEK_HOLE)
249 error = dmu_offset_next(zp->z_zfsvfs->z_os, zp->z_id, hole, &noff);
252 if ((error == ESRCH) || (noff > file_sz)) {
254 * Handle the virtual hole at the end of file.
271 zfs_ioctl(vnode_t *vp, u_long com, intptr_t data, int flag, cred_t *cred,
272 int *rvalp, caller_context_t *ct)
284 * The following two ioctls are used by bfu. Faking out,
285 * necessary to avoid bfu errors.
293 if (ddi_copyin((void *)data, &off, sizeof (off), flag))
297 zfsvfs = zp->z_zfsvfs;
301 /* offset parameter is in/out */
302 error = zfs_holey(vp, com, &off);
306 if (ddi_copyout(&off, (void *)data, sizeof (off), flag))
314 page_lookup(vnode_t *vp, int64_t start, int64_t off, int64_t nbytes)
320 VM_OBJECT_LOCK_ASSERT(obj, MA_OWNED);
323 if ((pp = vm_page_lookup(obj, OFF_TO_IDX(start))) != NULL &&
324 vm_page_is_valid(pp, (vm_offset_t)off, nbytes)) {
325 if (vm_page_sleep_if_busy(pp, FALSE, "zfsmwb"))
328 vm_page_lock_queues();
330 vm_page_unlock_queues();
332 if (__predict_false(obj->cache != NULL)) {
333 vm_page_cache_free(obj, OFF_TO_IDX(start),
334 OFF_TO_IDX(start) + 1);
344 page_unlock(vm_page_t pp)
351 zfs_map_page(vm_page_t pp, struct sf_buf **sfp)
355 *sfp = sf_buf_alloc(pp, SFB_CPUPRIVATE);
356 return ((caddr_t)sf_buf_kva(*sfp));
360 zfs_unmap_page(struct sf_buf *sf)
369 * When a file is memory mapped, we must keep the IO data synchronized
370 * between the DMU cache and the memory mapped pages. What this means:
372 * On Write: If we find a memory mapped page, we write to *both*
373 * the page and the dmu buffer.
377 update_pages(vnode_t *vp, int64_t start, int len, objset_t *os, uint64_t oid,
378 int segflg, dmu_tx_t *tx)
384 ASSERT(vp->v_mount != NULL);
388 off = start & PAGEOFFSET;
390 for (start &= PAGEMASK; len > 0; start += PAGESIZE) {
392 uint64_t nbytes = MIN(PAGESIZE - off, len);
394 if ((pp = page_lookup(vp, start, off, nbytes)) != NULL) {
397 VM_OBJECT_UNLOCK(obj);
398 va = zfs_map_page(pp, &sf);
399 if (segflg == UIO_NOCOPY) {
400 (void) dmu_write(os, oid, start+off, nbytes,
403 (void) dmu_read(os, oid, start+off, nbytes,
404 va+off, DMU_READ_PREFETCH);;
414 VM_OBJECT_UNLOCK(obj);
418 * When a file is memory mapped, we must keep the IO data synchronized
419 * between the DMU cache and the memory mapped pages. What this means:
421 * On Read: We "read" preferentially from memory mapped pages,
422 * else we default from the dmu buffer.
424 * NOTE: We will always "break up" the IO into PAGESIZE uiomoves when
425 * the file is memory mapped.
428 mappedread(vnode_t *vp, int nbytes, uio_t *uio)
430 znode_t *zp = VTOZ(vp);
431 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 uint64_t bytes = MIN(PAGESIZE - off, len);
453 if ((m = vm_page_lookup(obj, OFF_TO_IDX(start))) != NULL &&
454 vm_page_is_valid(m, (vm_offset_t)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,
466 sf = sf_buf_alloc(m, SFB_CPUPRIVATE);
467 va = (caddr_t)sf_buf_kva(sf);
468 error = uiomove(va + off, bytes, UIO_READ, uio);
474 } else if (m != NULL && uio->uio_segflg == UIO_NOCOPY) {
476 * The code below is here to make sendfile(2) work
477 * correctly with ZFS. As pointed out by ups@
478 * sendfile(2) should be changed to use VOP_GETPAGES(),
479 * but it pessimize performance of sendfile/UFS, that's
480 * why I handle this special case in ZFS code.
482 if (vm_page_sleep_if_busy(m, FALSE, "zfsmrb"))
485 VM_OBJECT_UNLOCK(obj);
487 error = dmu_read_uio(os, zp->z_id, uio,
493 sf = sf_buf_alloc(m, SFB_CPUPRIVATE);
494 va = (caddr_t)sf_buf_kva(sf);
495 error = dmu_read(os, zp->z_id, start + off,
496 bytes, (void *)(va + off),
503 vm_page_set_valid(m, off, bytes);
506 uio->uio_resid -= bytes;
507 uio->uio_offset += bytes;
517 VM_OBJECT_UNLOCK(obj);
518 if (error == 0 && dirbytes > 0)
519 error = dmu_read_uio(os, zp->z_id, uio, dirbytes);
523 offset_t zfs_read_chunk_size = 1024 * 1024; /* Tunable */
526 * Read bytes from specified file into supplied buffer.
528 * IN: vp - vnode of file to be read from.
529 * uio - structure supplying read location, range info,
531 * ioflag - SYNC flags; used to provide FRSYNC semantics.
532 * cr - credentials of caller.
533 * ct - caller context
535 * OUT: uio - updated offset and range, buffer filled.
537 * RETURN: 0 if success
538 * error code if failure
541 * vp - atime updated if byte count > 0
545 zfs_read(vnode_t *vp, uio_t *uio, int ioflag, cred_t *cr, caller_context_t *ct)
547 znode_t *zp = VTOZ(vp);
548 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
558 if (zp->z_phys->zp_flags & ZFS_AV_QUARANTINED) {
564 * Validate file offset
566 if (uio->uio_loffset < (offset_t)0) {
572 * Fasttrack empty reads
574 if (uio->uio_resid == 0) {
580 * Check for mandatory locks
582 if (MANDMODE((mode_t)zp->z_phys->zp_mode)) {
583 if (error = chklock(vp, FREAD,
584 uio->uio_loffset, uio->uio_resid, uio->uio_fmode, ct)) {
591 * If we're in FRSYNC mode, sync out this znode before reading it.
594 zil_commit(zfsvfs->z_log, zp->z_last_itx, zp->z_id);
597 * Lock the range against changes.
599 rl = zfs_range_lock(zp, uio->uio_loffset, uio->uio_resid, RL_READER);
602 * If we are reading past end-of-file we can skip
603 * to the end; but we might still need to set atime.
605 if (uio->uio_loffset >= zp->z_phys->zp_size) {
610 ASSERT(uio->uio_loffset < zp->z_phys->zp_size);
611 n = MIN(uio->uio_resid, zp->z_phys->zp_size - uio->uio_loffset);
614 nbytes = MIN(n, zfs_read_chunk_size -
615 P2PHASE(uio->uio_loffset, zfs_read_chunk_size));
617 if (vn_has_cached_data(vp))
618 error = mappedread(vp, nbytes, uio);
620 error = dmu_read_uio(os, zp->z_id, uio, nbytes);
622 /* convert checksum errors into IO errors */
632 zfs_range_unlock(rl);
634 ZFS_ACCESSTIME_STAMP(zfsvfs, zp);
640 * Fault in the pages of the first n bytes specified by the uio structure.
641 * 1 byte in each page is touched and the uio struct is unmodified.
642 * Any error will exit this routine as this is only a best
643 * attempt to get the pages resident. This is a copy of ufs_trans_touch().
646 zfs_prefault_write(ssize_t n, struct uio *uio)
652 if (uio->uio_segflg != UIO_USERSPACE)
658 cnt = MIN(iov->iov_len, n);
660 /* empty iov entry */
666 * touch each page in this segment.
672 incr = MIN(cnt, PAGESIZE);
677 * touch the last byte in case it straddles a page.
687 * Write the bytes to a file.
689 * IN: vp - vnode of file to be written to.
690 * uio - structure supplying write location, range info,
692 * ioflag - IO_APPEND flag set if in append mode.
693 * cr - credentials of caller.
694 * ct - caller context (NFS/CIFS fem monitor only)
696 * OUT: uio - updated offset and range.
698 * RETURN: 0 if success
699 * error code if failure
702 * vp - ctime|mtime updated if byte count > 0
706 zfs_write(vnode_t *vp, uio_t *uio, int ioflag, cred_t *cr, caller_context_t *ct)
708 znode_t *zp = VTOZ(vp);
709 rlim64_t limit = MAXOFFSET_T;
710 ssize_t start_resid = uio->uio_resid;
714 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
719 int max_blksz = zfsvfs->z_max_blksz;
725 * Fasttrack empty write
731 if (limit == RLIM64_INFINITY || limit > MAXOFFSET_T)
738 * If immutable or not appending then return EPERM
740 pflags = zp->z_phys->zp_flags;
741 if ((pflags & (ZFS_IMMUTABLE | ZFS_READONLY)) ||
742 ((pflags & ZFS_APPENDONLY) && !(ioflag & FAPPEND) &&
743 (uio->uio_loffset < zp->z_phys->zp_size))) {
748 zilog = zfsvfs->z_log;
751 * Pre-fault the pages to ensure slow (eg NFS) pages
754 zfs_prefault_write(n, uio);
757 * If in append mode, set the io offset pointer to eof.
759 if (ioflag & IO_APPEND) {
761 * Range lock for a file append:
762 * The value for the start of range will be determined by
763 * zfs_range_lock() (to guarantee append semantics).
764 * If this write will cause the block size to increase,
765 * zfs_range_lock() will lock the entire file, so we must
766 * later reduce the range after we grow the block size.
768 rl = zfs_range_lock(zp, 0, n, RL_APPEND);
769 if (rl->r_len == UINT64_MAX) {
770 /* overlocked, zp_size can't change */
771 woff = uio->uio_loffset = zp->z_phys->zp_size;
773 woff = uio->uio_loffset = rl->r_off;
776 woff = uio->uio_loffset;
778 * Validate file offset
786 * If we need to grow the block size then zfs_range_lock()
787 * will lock a wider range than we request here.
788 * Later after growing the block size we reduce the range.
790 rl = zfs_range_lock(zp, woff, n, RL_WRITER);
794 zfs_range_unlock(rl);
799 if ((woff + n) > limit || woff > (limit - n))
803 * Check for mandatory locks
805 if (MANDMODE((mode_t)zp->z_phys->zp_mode) &&
806 (error = chklock(vp, FWRITE, woff, n, uio->uio_fmode, ct)) != 0) {
807 zfs_range_unlock(rl);
811 end_size = MAX(zp->z_phys->zp_size, woff + n);
814 * Write the file in reasonable size chunks. Each chunk is written
815 * in a separate transaction; this keeps the intent log records small
816 * and allows us to do more fine-grained space accounting.
820 woff = uio->uio_loffset;
823 if (zfs_usergroup_overquota(zfsvfs,
824 B_FALSE, zp->z_phys->zp_uid) ||
825 zfs_usergroup_overquota(zfsvfs,
826 B_TRUE, zp->z_phys->zp_gid)) {
828 dmu_return_arcbuf(abuf);
834 * If dmu_assign_arcbuf() is expected to execute with minimum
835 * overhead loan an arc buffer and copy user data to it before
836 * we enter a txg. This avoids holding a txg forever while we
837 * pagefault on a hanging NFS server mapping.
839 if (abuf == NULL && n >= max_blksz &&
840 woff >= zp->z_phys->zp_size &&
841 P2PHASE(woff, max_blksz) == 0 &&
842 zp->z_blksz == max_blksz) {
845 abuf = dmu_request_arcbuf(zp->z_dbuf, max_blksz);
846 ASSERT(abuf != NULL);
847 ASSERT(arc_buf_size(abuf) == max_blksz);
848 if (error = uiocopy(abuf->b_data, max_blksz,
849 UIO_WRITE, uio, &cbytes)) {
850 dmu_return_arcbuf(abuf);
853 ASSERT(cbytes == max_blksz);
857 * Start a transaction.
859 tx = dmu_tx_create(zfsvfs->z_os);
860 dmu_tx_hold_bonus(tx, zp->z_id);
861 dmu_tx_hold_write(tx, zp->z_id, woff, MIN(n, max_blksz));
862 error = dmu_tx_assign(tx, TXG_NOWAIT);
864 if (error == ERESTART) {
871 dmu_return_arcbuf(abuf);
876 * If zfs_range_lock() over-locked we grow the blocksize
877 * and then reduce the lock range. This will only happen
878 * on the first iteration since zfs_range_reduce() will
879 * shrink down r_len to the appropriate size.
881 if (rl->r_len == UINT64_MAX) {
884 if (zp->z_blksz > max_blksz) {
885 ASSERT(!ISP2(zp->z_blksz));
886 new_blksz = MIN(end_size, SPA_MAXBLOCKSIZE);
888 new_blksz = MIN(end_size, max_blksz);
890 zfs_grow_blocksize(zp, new_blksz, tx);
891 zfs_range_reduce(rl, woff, n);
895 * XXX - should we really limit each write to z_max_blksz?
896 * Perhaps we should use SPA_MAXBLOCKSIZE chunks?
898 nbytes = MIN(n, max_blksz - P2PHASE(woff, max_blksz));
900 if (woff + nbytes > zp->z_phys->zp_size)
901 vnode_pager_setsize(vp, woff + nbytes);
904 tx_bytes = uio->uio_resid;
905 error = dmu_write_uio(zfsvfs->z_os, zp->z_id, uio,
907 tx_bytes -= uio->uio_resid;
910 ASSERT(tx_bytes == max_blksz);
911 dmu_assign_arcbuf(zp->z_dbuf, woff, abuf, tx);
912 ASSERT(tx_bytes <= uio->uio_resid);
913 uioskip(uio, tx_bytes);
917 * XXXPJD: There are some cases (triggered by fsx) where
918 * vn_has_cached_data(vp) returns false when it should
919 * return true. This should be investigated.
922 if (tx_bytes && vn_has_cached_data(vp))
924 if (tx_bytes && vp->v_object != NULL)
927 update_pages(vp, woff, tx_bytes, zfsvfs->z_os,
928 zp->z_id, uio->uio_segflg, tx);
932 * If we made no progress, we're done. If we made even
933 * partial progress, update the znode and ZIL accordingly.
942 * Clear Set-UID/Set-GID bits on successful write if not
943 * privileged and at least one of the excute bits is set.
945 * It would be nice to to this after all writes have
946 * been done, but that would still expose the ISUID/ISGID
947 * to another app after the partial write is committed.
949 * Note: we don't call zfs_fuid_map_id() here because
950 * user 0 is not an ephemeral uid.
952 mutex_enter(&zp->z_acl_lock);
953 if ((zp->z_phys->zp_mode & (S_IXUSR | (S_IXUSR >> 3) |
954 (S_IXUSR >> 6))) != 0 &&
955 (zp->z_phys->zp_mode & (S_ISUID | S_ISGID)) != 0 &&
956 secpolicy_vnode_setid_retain(vp, cr,
957 (zp->z_phys->zp_mode & S_ISUID) != 0 &&
958 zp->z_phys->zp_uid == 0) != 0) {
959 zp->z_phys->zp_mode &= ~(S_ISUID | S_ISGID);
961 mutex_exit(&zp->z_acl_lock);
964 * Update time stamp. NOTE: This marks the bonus buffer as
965 * dirty, so we don't have to do it again for zp_size.
967 zfs_time_stamper(zp, CONTENT_MODIFIED, tx);
970 * Update the file size (zp_size) if it has changed;
971 * account for possible concurrent updates.
973 while ((end_size = zp->z_phys->zp_size) < uio->uio_loffset)
974 (void) atomic_cas_64(&zp->z_phys->zp_size, end_size,
976 zfs_log_write(zilog, tx, TX_WRITE, zp, woff, tx_bytes, ioflag);
981 ASSERT(tx_bytes == nbytes);
985 zfs_range_unlock(rl);
988 * If we're in replay mode, or we made no progress, return error.
989 * Otherwise, it's at least a partial write, so it's successful.
991 if (zfsvfs->z_replay || uio->uio_resid == start_resid) {
996 if (ioflag & (FSYNC | FDSYNC))
997 zil_commit(zilog, zp->z_last_itx, zp->z_id);
1004 zfs_get_done(dmu_buf_t *db, void *vzgd)
1006 zgd_t *zgd = (zgd_t *)vzgd;
1007 rl_t *rl = zgd->zgd_rl;
1008 vnode_t *vp = ZTOV(rl->r_zp);
1009 objset_t *os = rl->r_zp->z_zfsvfs->z_os;
1012 vfslocked = VFS_LOCK_GIANT(vp->v_vfsp);
1013 dmu_buf_rele(db, vzgd);
1014 zfs_range_unlock(rl);
1016 * Release the vnode asynchronously as we currently have the
1017 * txg stopped from syncing.
1019 VN_RELE_ASYNC(vp, dsl_pool_vnrele_taskq(dmu_objset_pool(os)));
1020 zil_add_block(zgd->zgd_zilog, zgd->zgd_bp);
1021 kmem_free(zgd, sizeof (zgd_t));
1022 VFS_UNLOCK_GIANT(vfslocked);
1026 * Get data to generate a TX_WRITE intent log record.
1029 zfs_get_data(void *arg, lr_write_t *lr, char *buf, zio_t *zio)
1031 zfsvfs_t *zfsvfs = arg;
1032 objset_t *os = zfsvfs->z_os;
1034 uint64_t off = lr->lr_offset;
1038 int dlen = lr->lr_length; /* length of user data */
1045 * Nothing to do if the file has been removed
1047 if (zfs_zget(zfsvfs, lr->lr_foid, &zp) != 0)
1049 if (zp->z_unlinked) {
1051 * Release the vnode asynchronously as we currently have the
1052 * txg stopped from syncing.
1054 VN_RELE_ASYNC(ZTOV(zp),
1055 dsl_pool_vnrele_taskq(dmu_objset_pool(os)));
1060 * Write records come in two flavors: immediate and indirect.
1061 * For small writes it's cheaper to store the data with the
1062 * log record (immediate); for large writes it's cheaper to
1063 * sync the data and get a pointer to it (indirect) so that
1064 * we don't have to write the data twice.
1066 if (buf != NULL) { /* immediate write */
1067 rl = zfs_range_lock(zp, off, dlen, RL_READER);
1068 /* test for truncation needs to be done while range locked */
1069 if (off >= zp->z_phys->zp_size) {
1073 VERIFY(0 == dmu_read(os, lr->lr_foid, off, dlen, buf,
1074 DMU_READ_NO_PREFETCH));
1075 } else { /* indirect write */
1076 uint64_t boff; /* block starting offset */
1079 * Have to lock the whole block to ensure when it's
1080 * written out and it's checksum is being calculated
1081 * that no one can change the data. We need to re-check
1082 * blocksize after we get the lock in case it's changed!
1085 if (ISP2(zp->z_blksz)) {
1086 boff = P2ALIGN_TYPED(off, zp->z_blksz,
1092 rl = zfs_range_lock(zp, boff, dlen, RL_READER);
1093 if (zp->z_blksz == dlen)
1095 zfs_range_unlock(rl);
1097 /* test for truncation needs to be done while range locked */
1098 if (off >= zp->z_phys->zp_size) {
1102 zgd = (zgd_t *)kmem_alloc(sizeof (zgd_t), KM_SLEEP);
1104 zgd->zgd_zilog = zfsvfs->z_log;
1105 zgd->zgd_bp = &lr->lr_blkptr;
1106 VERIFY(0 == dmu_buf_hold(os, lr->lr_foid, boff, zgd, &db));
1107 ASSERT(boff == db->db_offset);
1108 lr->lr_blkoff = off - boff;
1109 error = dmu_sync(zio, db, &lr->lr_blkptr,
1110 lr->lr_common.lrc_txg, zfs_get_done, zgd);
1111 ASSERT((error && error != EINPROGRESS) ||
1112 lr->lr_length <= zp->z_blksz);
1115 * dmu_sync() can compress a block of zeros to a null
1116 * blkptr but the block size still needs to be passed
1117 * through to replay.
1119 BP_SET_LSIZE(&lr->lr_blkptr, db->db_size);
1120 zil_add_block(zfsvfs->z_log, &lr->lr_blkptr);
1124 * If we get EINPROGRESS, then we need to wait for a
1125 * write IO initiated by dmu_sync() to complete before
1126 * we can release this dbuf. We will finish everything
1127 * up in the zfs_get_done() callback.
1129 if (error == EINPROGRESS) {
1131 } else if (error == EALREADY) {
1132 lr->lr_common.lrc_txtype = TX_WRITE2;
1135 dmu_buf_rele(db, zgd);
1136 kmem_free(zgd, sizeof (zgd_t));
1139 zfs_range_unlock(rl);
1141 * Release the vnode asynchronously as we currently have the
1142 * txg stopped from syncing.
1144 VN_RELE_ASYNC(ZTOV(zp), dsl_pool_vnrele_taskq(dmu_objset_pool(os)));
1150 zfs_access(vnode_t *vp, int mode, int flag, cred_t *cr,
1151 caller_context_t *ct)
1153 znode_t *zp = VTOZ(vp);
1154 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1160 if (flag & V_ACE_MASK)
1161 error = zfs_zaccess(zp, mode, flag, B_FALSE, cr);
1163 error = zfs_zaccess_rwx(zp, mode, flag, cr);
1170 * If vnode is for a device return a specfs vnode instead.
1173 specvp_check(vnode_t **vpp, cred_t *cr)
1177 if (IS_DEVVP(*vpp)) {
1180 svp = specvp(*vpp, (*vpp)->v_rdev, (*vpp)->v_type, cr);
1191 * Lookup an entry in a directory, or an extended attribute directory.
1192 * If it exists, return a held vnode reference for it.
1194 * IN: dvp - vnode of directory to search.
1195 * nm - name of entry to lookup.
1196 * pnp - full pathname to lookup [UNUSED].
1197 * flags - LOOKUP_XATTR set if looking for an attribute.
1198 * rdir - root directory vnode [UNUSED].
1199 * cr - credentials of caller.
1200 * ct - caller context
1201 * direntflags - directory lookup flags
1202 * realpnp - returned pathname.
1204 * OUT: vpp - vnode of located entry, NULL if not found.
1206 * RETURN: 0 if success
1207 * error code if failure
1214 zfs_lookup(vnode_t *dvp, char *nm, vnode_t **vpp, struct componentname *cnp,
1215 int nameiop, cred_t *cr, kthread_t *td, int flags)
1217 znode_t *zdp = VTOZ(dvp);
1218 zfsvfs_t *zfsvfs = zdp->z_zfsvfs;
1220 int *direntflags = NULL;
1221 void *realpnp = NULL;
1224 if (!(flags & (LOOKUP_XATTR | FIGNORECASE))) {
1226 if (dvp->v_type != VDIR) {
1228 } else if (zdp->z_dbuf == NULL) {
1232 if (nm[0] == 0 || (nm[0] == '.' && nm[1] == '\0')) {
1233 error = zfs_fastaccesschk_execute(zdp, cr);
1241 vnode_t *tvp = dnlc_lookup(dvp, nm);
1244 error = zfs_fastaccesschk_execute(zdp, cr);
1249 if (tvp == DNLC_NO_VNODE) {
1254 return (specvp_check(vpp, cr));
1260 DTRACE_PROBE2(zfs__fastpath__lookup__miss, vnode_t *, dvp, char *, nm);
1267 if (flags & LOOKUP_XATTR) {
1270 * If the xattr property is off, refuse the lookup request.
1272 if (!(zfsvfs->z_vfs->vfs_flag & VFS_XATTR)) {
1279 * We don't allow recursive attributes..
1280 * Maybe someday we will.
1282 if (zdp->z_phys->zp_flags & ZFS_XATTR) {
1287 if (error = zfs_get_xattrdir(VTOZ(dvp), vpp, cr, flags)) {
1293 * Do we have permission to get into attribute directory?
1296 if (error = zfs_zaccess(VTOZ(*vpp), ACE_EXECUTE, 0,
1306 if (dvp->v_type != VDIR) {
1312 * Check accessibility of directory.
1315 if (error = zfs_zaccess(zdp, ACE_EXECUTE, 0, B_FALSE, cr)) {
1320 if (zfsvfs->z_utf8 && u8_validate(nm, strlen(nm),
1321 NULL, U8_VALIDATE_ENTIRE, &error) < 0) {
1326 error = zfs_dirlook(zdp, nm, vpp, flags, direntflags, realpnp);
1328 error = specvp_check(vpp, cr);
1330 /* Translate errors and add SAVENAME when needed. */
1331 if (cnp->cn_flags & ISLASTCN) {
1335 if (error == ENOENT) {
1336 error = EJUSTRETURN;
1337 cnp->cn_flags |= SAVENAME;
1343 cnp->cn_flags |= SAVENAME;
1347 if (error == 0 && (nm[0] != '.' || nm[1] != '\0')) {
1350 if (cnp->cn_flags & ISDOTDOT) {
1351 ltype = VOP_ISLOCKED(dvp);
1355 error = vn_lock(*vpp, cnp->cn_lkflags);
1356 if (cnp->cn_flags & ISDOTDOT)
1357 vn_lock(dvp, ltype | LK_RETRY);
1367 #ifdef FREEBSD_NAMECACHE
1369 * Insert name into cache (as non-existent) if appropriate.
1371 if (error == ENOENT && (cnp->cn_flags & MAKEENTRY) && nameiop != CREATE)
1372 cache_enter(dvp, *vpp, cnp);
1374 * Insert name into cache if appropriate.
1376 if (error == 0 && (cnp->cn_flags & MAKEENTRY)) {
1377 if (!(cnp->cn_flags & ISLASTCN) ||
1378 (nameiop != DELETE && nameiop != RENAME)) {
1379 cache_enter(dvp, *vpp, cnp);
1388 * Attempt to create a new entry in a directory. If the entry
1389 * already exists, truncate the file if permissible, else return
1390 * an error. Return the vp of the created or trunc'd file.
1392 * IN: dvp - vnode of directory to put new file entry in.
1393 * name - name of new file entry.
1394 * vap - attributes of new file.
1395 * excl - flag indicating exclusive or non-exclusive mode.
1396 * mode - mode to open file with.
1397 * cr - credentials of caller.
1398 * flag - large file flag [UNUSED].
1399 * ct - caller context
1400 * vsecp - ACL to be set
1402 * OUT: vpp - vnode of created or trunc'd entry.
1404 * RETURN: 0 if success
1405 * error code if failure
1408 * dvp - ctime|mtime updated if new entry created
1409 * vp - ctime|mtime always, atime if new
1414 zfs_create(vnode_t *dvp, char *name, vattr_t *vap, int excl, int mode,
1415 vnode_t **vpp, cred_t *cr, kthread_t *td)
1417 znode_t *zp, *dzp = VTOZ(dvp);
1418 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
1426 gid_t gid = crgetgid(cr);
1427 zfs_acl_ids_t acl_ids;
1428 boolean_t fuid_dirtied;
1433 * If we have an ephemeral id, ACL, or XVATTR then
1434 * make sure file system is at proper version
1437 ksid = crgetsid(cr, KSID_OWNER);
1439 uid = ksid_getid(ksid);
1442 if (zfsvfs->z_use_fuids == B_FALSE &&
1443 (vsecp || (vap->va_mask & AT_XVATTR) ||
1444 IS_EPHEMERAL(crgetuid(cr)) || IS_EPHEMERAL(crgetgid(cr))))
1450 zilog = zfsvfs->z_log;
1452 if (zfsvfs->z_utf8 && u8_validate(name, strlen(name),
1453 NULL, U8_VALIDATE_ENTIRE, &error) < 0) {
1458 if (vap->va_mask & AT_XVATTR) {
1459 if ((error = secpolicy_xvattr(dvp, (xvattr_t *)vap,
1460 crgetuid(cr), cr, vap->va_type)) != 0) {
1468 if ((vap->va_mode & S_ISVTX) && secpolicy_vnode_stky_modify(cr))
1469 vap->va_mode &= ~S_ISVTX;
1471 if (*name == '\0') {
1473 * Null component name refers to the directory itself.
1480 /* possible VN_HOLD(zp) */
1483 if (flag & FIGNORECASE)
1486 error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg,
1489 if (strcmp(name, "..") == 0)
1499 * Create a new file object and update the directory
1502 if (error = zfs_zaccess(dzp, ACE_ADD_FILE, 0, B_FALSE, cr)) {
1507 * We only support the creation of regular files in
1508 * extended attribute directories.
1510 if ((dzp->z_phys->zp_flags & ZFS_XATTR) &&
1511 (vap->va_type != VREG)) {
1517 if ((error = zfs_acl_ids_create(dzp, 0, vap, cr, vsecp,
1520 if (zfs_acl_ids_overquota(zfsvfs, &acl_ids)) {
1521 zfs_acl_ids_free(&acl_ids);
1526 tx = dmu_tx_create(os);
1527 dmu_tx_hold_bonus(tx, DMU_NEW_OBJECT);
1528 fuid_dirtied = zfsvfs->z_fuid_dirty;
1530 zfs_fuid_txhold(zfsvfs, tx);
1531 dmu_tx_hold_bonus(tx, dzp->z_id);
1532 dmu_tx_hold_zap(tx, dzp->z_id, TRUE, name);
1533 if (acl_ids.z_aclp->z_acl_bytes > ZFS_ACE_SPACE) {
1534 dmu_tx_hold_write(tx, DMU_NEW_OBJECT,
1535 0, SPA_MAXBLOCKSIZE);
1537 error = dmu_tx_assign(tx, TXG_NOWAIT);
1539 zfs_acl_ids_free(&acl_ids);
1540 zfs_dirent_unlock(dl);
1541 if (error == ERESTART) {
1550 zfs_mknode(dzp, vap, tx, cr, 0, &zp, 0, &acl_ids);
1553 zfs_fuid_sync(zfsvfs, tx);
1555 (void) zfs_link_create(dl, zp, tx, ZNEW);
1557 txtype = zfs_log_create_txtype(Z_FILE, vsecp, vap);
1558 if (flag & FIGNORECASE)
1560 zfs_log_create(zilog, tx, txtype, dzp, zp, name,
1561 vsecp, acl_ids.z_fuidp, vap);
1562 zfs_acl_ids_free(&acl_ids);
1565 int aflags = (flag & FAPPEND) ? V_APPEND : 0;
1568 * A directory entry already exists for this name.
1571 * Can't truncate an existing file if in exclusive mode.
1578 * Can't open a directory for writing.
1580 if ((ZTOV(zp)->v_type == VDIR) && (mode & S_IWRITE)) {
1585 * Verify requested access to file.
1587 if (mode && (error = zfs_zaccess_rwx(zp, mode, aflags, cr))) {
1591 mutex_enter(&dzp->z_lock);
1593 mutex_exit(&dzp->z_lock);
1596 * Truncate regular files if requested.
1598 if ((ZTOV(zp)->v_type == VREG) &&
1599 (vap->va_mask & AT_SIZE) && (vap->va_size == 0)) {
1600 /* we can't hold any locks when calling zfs_freesp() */
1601 zfs_dirent_unlock(dl);
1603 error = zfs_freesp(zp, 0, 0, mode, TRUE);
1605 vnevent_create(ZTOV(zp), ct);
1611 zfs_dirent_unlock(dl);
1618 error = specvp_check(vpp, cr);
1626 * Remove an entry from a directory.
1628 * IN: dvp - vnode of directory to remove entry from.
1629 * name - name of entry to remove.
1630 * cr - credentials of caller.
1631 * ct - caller context
1632 * flags - case flags
1634 * RETURN: 0 if success
1635 * error code if failure
1639 * vp - ctime (if nlink > 0)
1643 zfs_remove(vnode_t *dvp, char *name, cred_t *cr, caller_context_t *ct,
1646 znode_t *zp, *dzp = VTOZ(dvp);
1647 znode_t *xzp = NULL;
1649 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
1651 uint64_t acl_obj, xattr_obj;
1654 boolean_t may_delete_now, delete_now = FALSE;
1655 boolean_t unlinked, toobig = FALSE;
1657 pathname_t *realnmp = NULL;
1664 zilog = zfsvfs->z_log;
1666 if (flags & FIGNORECASE) {
1674 * Attempt to lock directory; fail if entry doesn't exist.
1676 if (error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg,
1686 if (error = zfs_zaccess_delete(dzp, zp, cr)) {
1691 * Need to use rmdir for removing directories.
1693 if (vp->v_type == VDIR) {
1698 vnevent_remove(vp, dvp, name, ct);
1701 dnlc_remove(dvp, realnmp->pn_buf);
1703 dnlc_remove(dvp, name);
1705 may_delete_now = FALSE;
1708 * We may delete the znode now, or we may put it in the unlinked set;
1709 * it depends on whether we're the last link, and on whether there are
1710 * other holds on the vnode. So we dmu_tx_hold() the right things to
1711 * allow for either case.
1713 tx = dmu_tx_create(zfsvfs->z_os);
1714 dmu_tx_hold_zap(tx, dzp->z_id, FALSE, name);
1715 dmu_tx_hold_bonus(tx, zp->z_id);
1716 if (may_delete_now) {
1718 zp->z_phys->zp_size > zp->z_blksz * DMU_MAX_DELETEBLKCNT;
1719 /* if the file is too big, only hold_free a token amount */
1720 dmu_tx_hold_free(tx, zp->z_id, 0,
1721 (toobig ? DMU_MAX_ACCESS : DMU_OBJECT_END));
1724 /* are there any extended attributes? */
1725 if ((xattr_obj = zp->z_phys->zp_xattr) != 0) {
1726 /* XXX - do we need this if we are deleting? */
1727 dmu_tx_hold_bonus(tx, xattr_obj);
1730 /* are there any additional acls */
1731 if ((acl_obj = zp->z_phys->zp_acl.z_acl_extern_obj) != 0 &&
1733 dmu_tx_hold_free(tx, acl_obj, 0, DMU_OBJECT_END);
1735 /* charge as an update -- would be nice not to charge at all */
1736 dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL);
1738 error = dmu_tx_assign(tx, TXG_NOWAIT);
1740 zfs_dirent_unlock(dl);
1742 if (error == ERESTART) {
1755 * Remove the directory entry.
1757 error = zfs_link_destroy(dl, zp, tx, zflg, &unlinked);
1764 if (0 && unlinked) {
1766 delete_now = may_delete_now && !toobig &&
1767 vp->v_count == 1 && !vn_has_cached_data(vp) &&
1768 zp->z_phys->zp_xattr == xattr_obj &&
1769 zp->z_phys->zp_acl.z_acl_extern_obj == acl_obj;
1774 if (zp->z_phys->zp_xattr) {
1775 error = zfs_zget(zfsvfs, zp->z_phys->zp_xattr, &xzp);
1776 ASSERT3U(error, ==, 0);
1777 ASSERT3U(xzp->z_phys->zp_links, ==, 2);
1778 dmu_buf_will_dirty(xzp->z_dbuf, tx);
1779 mutex_enter(&xzp->z_lock);
1780 xzp->z_unlinked = 1;
1781 xzp->z_phys->zp_links = 0;
1782 mutex_exit(&xzp->z_lock);
1783 zfs_unlinked_add(xzp, tx);
1784 zp->z_phys->zp_xattr = 0; /* probably unnecessary */
1786 mutex_enter(&zp->z_lock);
1789 ASSERT3U(vp->v_count, ==, 0);
1791 mutex_exit(&zp->z_lock);
1792 zfs_znode_delete(zp, tx);
1793 } else if (unlinked) {
1794 zfs_unlinked_add(zp, tx);
1798 if (flags & FIGNORECASE)
1800 zfs_log_remove(zilog, tx, txtype, dzp, name);
1807 zfs_dirent_unlock(dl);
1812 /* this rele is delayed to prevent nesting transactions */
1821 * Create a new directory and insert it into dvp using the name
1822 * provided. Return a pointer to the inserted directory.
1824 * IN: dvp - vnode of directory to add subdir to.
1825 * dirname - name of new directory.
1826 * vap - attributes of new directory.
1827 * cr - credentials of caller.
1828 * ct - caller context
1829 * vsecp - ACL to be set
1831 * OUT: vpp - vnode of created directory.
1833 * RETURN: 0 if success
1834 * error code if failure
1837 * dvp - ctime|mtime updated
1838 * vp - ctime|mtime|atime updated
1842 zfs_mkdir(vnode_t *dvp, char *dirname, vattr_t *vap, vnode_t **vpp, cred_t *cr,
1843 caller_context_t *ct, int flags, vsecattr_t *vsecp)
1845 znode_t *zp, *dzp = VTOZ(dvp);
1846 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
1855 gid_t gid = crgetgid(cr);
1856 zfs_acl_ids_t acl_ids;
1857 boolean_t fuid_dirtied;
1859 ASSERT(vap->va_type == VDIR);
1862 * If we have an ephemeral id, ACL, or XVATTR then
1863 * make sure file system is at proper version
1866 ksid = crgetsid(cr, KSID_OWNER);
1868 uid = ksid_getid(ksid);
1871 if (zfsvfs->z_use_fuids == B_FALSE &&
1872 (vsecp || (vap->va_mask & AT_XVATTR) || IS_EPHEMERAL(crgetuid(cr))||
1873 IS_EPHEMERAL(crgetgid(cr))))
1878 zilog = zfsvfs->z_log;
1880 if (dzp->z_phys->zp_flags & ZFS_XATTR) {
1885 if (zfsvfs->z_utf8 && u8_validate(dirname,
1886 strlen(dirname), NULL, U8_VALIDATE_ENTIRE, &error) < 0) {
1890 if (flags & FIGNORECASE)
1893 if (vap->va_mask & AT_XVATTR)
1894 if ((error = secpolicy_xvattr(dvp, (xvattr_t *)vap,
1895 crgetuid(cr), cr, vap->va_type)) != 0) {
1901 * First make sure the new directory doesn't exist.
1906 if (error = zfs_dirent_lock(&dl, dzp, dirname, &zp, zf,
1912 if (error = zfs_zaccess(dzp, ACE_ADD_SUBDIRECTORY, 0, B_FALSE, cr)) {
1913 zfs_dirent_unlock(dl);
1918 if ((error = zfs_acl_ids_create(dzp, 0, vap, cr, vsecp,
1920 zfs_dirent_unlock(dl);
1924 if (zfs_acl_ids_overquota(zfsvfs, &acl_ids)) {
1925 zfs_acl_ids_free(&acl_ids);
1926 zfs_dirent_unlock(dl);
1932 * Add a new entry to the directory.
1934 tx = dmu_tx_create(zfsvfs->z_os);
1935 dmu_tx_hold_zap(tx, dzp->z_id, TRUE, dirname);
1936 dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL);
1937 fuid_dirtied = zfsvfs->z_fuid_dirty;
1939 zfs_fuid_txhold(zfsvfs, tx);
1940 if (acl_ids.z_aclp->z_acl_bytes > ZFS_ACE_SPACE)
1941 dmu_tx_hold_write(tx, DMU_NEW_OBJECT,
1942 0, SPA_MAXBLOCKSIZE);
1943 error = dmu_tx_assign(tx, TXG_NOWAIT);
1945 zfs_acl_ids_free(&acl_ids);
1946 zfs_dirent_unlock(dl);
1947 if (error == ERESTART) {
1960 zfs_mknode(dzp, vap, tx, cr, 0, &zp, 0, &acl_ids);
1963 zfs_fuid_sync(zfsvfs, tx);
1965 * Now put new name in parent dir.
1967 (void) zfs_link_create(dl, zp, tx, ZNEW);
1971 txtype = zfs_log_create_txtype(Z_DIR, vsecp, vap);
1972 if (flags & FIGNORECASE)
1974 zfs_log_create(zilog, tx, txtype, dzp, zp, dirname, vsecp,
1975 acl_ids.z_fuidp, vap);
1977 zfs_acl_ids_free(&acl_ids);
1980 zfs_dirent_unlock(dl);
1987 * Remove a directory subdir entry. If the current working
1988 * directory is the same as the subdir to be removed, the
1991 * IN: dvp - vnode of directory to remove from.
1992 * name - name of directory to be removed.
1993 * cwd - vnode of current working directory.
1994 * cr - credentials of caller.
1995 * ct - caller context
1996 * flags - case flags
1998 * RETURN: 0 if success
1999 * error code if failure
2002 * dvp - ctime|mtime updated
2006 zfs_rmdir(vnode_t *dvp, char *name, vnode_t *cwd, cred_t *cr,
2007 caller_context_t *ct, int flags)
2009 znode_t *dzp = VTOZ(dvp);
2012 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
2021 zilog = zfsvfs->z_log;
2023 if (flags & FIGNORECASE)
2029 * Attempt to lock directory; fail if entry doesn't exist.
2031 if (error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg,
2039 if (error = zfs_zaccess_delete(dzp, zp, cr)) {
2043 if (vp->v_type != VDIR) {
2053 vnevent_rmdir(vp, dvp, name, ct);
2056 * Grab a lock on the directory to make sure that noone is
2057 * trying to add (or lookup) entries while we are removing it.
2059 rw_enter(&zp->z_name_lock, RW_WRITER);
2062 * Grab a lock on the parent pointer to make sure we play well
2063 * with the treewalk and directory rename code.
2065 rw_enter(&zp->z_parent_lock, RW_WRITER);
2067 tx = dmu_tx_create(zfsvfs->z_os);
2068 dmu_tx_hold_zap(tx, dzp->z_id, FALSE, name);
2069 dmu_tx_hold_bonus(tx, zp->z_id);
2070 dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL);
2071 error = dmu_tx_assign(tx, TXG_NOWAIT);
2073 rw_exit(&zp->z_parent_lock);
2074 rw_exit(&zp->z_name_lock);
2075 zfs_dirent_unlock(dl);
2077 if (error == ERESTART) {
2087 #ifdef FREEBSD_NAMECACHE
2091 error = zfs_link_destroy(dl, zp, tx, zflg, NULL);
2094 uint64_t txtype = TX_RMDIR;
2095 if (flags & FIGNORECASE)
2097 zfs_log_remove(zilog, tx, txtype, dzp, name);
2102 rw_exit(&zp->z_parent_lock);
2103 rw_exit(&zp->z_name_lock);
2104 #ifdef FREEBSD_NAMECACHE
2108 zfs_dirent_unlock(dl);
2117 * Read as many directory entries as will fit into the provided
2118 * buffer from the given directory cursor position (specified in
2119 * the uio structure.
2121 * IN: vp - vnode of directory to read.
2122 * uio - structure supplying read location, range info,
2123 * and return buffer.
2124 * cr - credentials of caller.
2125 * ct - caller context
2126 * flags - case flags
2128 * OUT: uio - updated offset and range, buffer filled.
2129 * eofp - set to true if end-of-file detected.
2131 * RETURN: 0 if success
2132 * error code if failure
2135 * vp - atime updated
2137 * Note that the low 4 bits of the cookie returned by zap is always zero.
2138 * This allows us to use the low range for "special" directory entries:
2139 * We use 0 for '.', and 1 for '..'. If this is the root of the filesystem,
2140 * we use the offset 2 for the '.zfs' directory.
2144 zfs_readdir(vnode_t *vp, uio_t *uio, cred_t *cr, int *eofp, int *ncookies, u_long **cookies)
2146 znode_t *zp = VTOZ(vp);
2150 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
2155 zap_attribute_t zap;
2156 uint_t bytes_wanted;
2157 uint64_t offset; /* must be unsigned; checks for < 1 */
2162 boolean_t check_sysattrs;
2165 u_long *cooks = NULL;
2172 * If we are not given an eof variable,
2179 * Check for valid iov_len.
2181 if (uio->uio_iov->iov_len <= 0) {
2187 * Quit if directory has been removed (posix)
2189 if ((*eofp = zp->z_unlinked) != 0) {
2196 offset = uio->uio_loffset;
2197 prefetch = zp->z_zn_prefetch;
2200 * Initialize the iterator cursor.
2204 * Start iteration from the beginning of the directory.
2206 zap_cursor_init(&zc, os, zp->z_id);
2209 * The offset is a serialized cursor.
2211 zap_cursor_init_serialized(&zc, os, zp->z_id, offset);
2215 * Get space to change directory entries into fs independent format.
2217 iovp = uio->uio_iov;
2218 bytes_wanted = iovp->iov_len;
2219 if (uio->uio_segflg != UIO_SYSSPACE || uio->uio_iovcnt != 1) {
2220 bufsize = bytes_wanted;
2221 outbuf = kmem_alloc(bufsize, KM_SLEEP);
2222 odp = (struct dirent64 *)outbuf;
2224 bufsize = bytes_wanted;
2225 odp = (struct dirent64 *)iovp->iov_base;
2227 eodp = (struct edirent *)odp;
2229 if (ncookies != NULL) {
2231 * Minimum entry size is dirent size and 1 byte for a file name.
2233 ncooks = uio->uio_resid / (sizeof(struct dirent) - sizeof(((struct dirent *)NULL)->d_name) + 1);
2234 cooks = malloc(ncooks * sizeof(u_long), M_TEMP, M_WAITOK);
2239 * If this VFS supports the system attribute view interface; and
2240 * we're looking at an extended attribute directory; and we care
2241 * about normalization conflicts on this vfs; then we must check
2242 * for normalization conflicts with the sysattr name space.
2245 check_sysattrs = vfs_has_feature(vp->v_vfsp, VFSFT_SYSATTR_VIEWS) &&
2246 (vp->v_flag & V_XATTRDIR) && zfsvfs->z_norm &&
2247 (flags & V_RDDIR_ENTFLAGS);
2253 * Transform to file-system independent format
2256 while (outcount < bytes_wanted) {
2262 * Special case `.', `..', and `.zfs'.
2265 (void) strcpy(zap.za_name, ".");
2266 zap.za_normalization_conflict = 0;
2269 } else if (offset == 1) {
2270 (void) strcpy(zap.za_name, "..");
2271 zap.za_normalization_conflict = 0;
2272 objnum = zp->z_phys->zp_parent;
2274 } else if (offset == 2 && zfs_show_ctldir(zp)) {
2275 (void) strcpy(zap.za_name, ZFS_CTLDIR_NAME);
2276 zap.za_normalization_conflict = 0;
2277 objnum = ZFSCTL_INO_ROOT;
2283 if (error = zap_cursor_retrieve(&zc, &zap)) {
2284 if ((*eofp = (error == ENOENT)) != 0)
2290 if (zap.za_integer_length != 8 ||
2291 zap.za_num_integers != 1) {
2292 cmn_err(CE_WARN, "zap_readdir: bad directory "
2293 "entry, obj = %lld, offset = %lld\n",
2294 (u_longlong_t)zp->z_id,
2295 (u_longlong_t)offset);
2300 objnum = ZFS_DIRENT_OBJ(zap.za_first_integer);
2302 * MacOS X can extract the object type here such as:
2303 * uint8_t type = ZFS_DIRENT_TYPE(zap.za_first_integer);
2305 type = ZFS_DIRENT_TYPE(zap.za_first_integer);
2307 if (check_sysattrs && !zap.za_normalization_conflict) {
2309 zap.za_normalization_conflict =
2310 xattr_sysattr_casechk(zap.za_name);
2312 panic("%s:%u: TODO", __func__, __LINE__);
2317 if (flags & V_RDDIR_ACCFILTER) {
2319 * If we have no access at all, don't include
2320 * this entry in the returned information
2323 if (zfs_zget(zp->z_zfsvfs, objnum, &ezp) != 0)
2325 if (!zfs_has_access(ezp, cr)) {
2332 if (flags & V_RDDIR_ENTFLAGS)
2333 reclen = EDIRENT_RECLEN(strlen(zap.za_name));
2335 reclen = DIRENT64_RECLEN(strlen(zap.za_name));
2338 * Will this entry fit in the buffer?
2340 if (outcount + reclen > bufsize) {
2342 * Did we manage to fit anything in the buffer?
2350 if (flags & V_RDDIR_ENTFLAGS) {
2352 * Add extended flag entry:
2354 eodp->ed_ino = objnum;
2355 eodp->ed_reclen = reclen;
2356 /* NOTE: ed_off is the offset for the *next* entry */
2357 next = &(eodp->ed_off);
2358 eodp->ed_eflags = zap.za_normalization_conflict ?
2359 ED_CASE_CONFLICT : 0;
2360 (void) strncpy(eodp->ed_name, zap.za_name,
2361 EDIRENT_NAMELEN(reclen));
2362 eodp = (edirent_t *)((intptr_t)eodp + reclen);
2367 odp->d_ino = objnum;
2368 odp->d_reclen = reclen;
2369 odp->d_namlen = strlen(zap.za_name);
2370 (void) strlcpy(odp->d_name, zap.za_name, odp->d_namlen + 1);
2372 odp = (dirent64_t *)((intptr_t)odp + reclen);
2376 ASSERT(outcount <= bufsize);
2378 /* Prefetch znode */
2380 dmu_prefetch(os, objnum, 0, 0);
2384 * Move to the next entry, fill in the previous offset.
2386 if (offset > 2 || (offset == 2 && !zfs_show_ctldir(zp))) {
2387 zap_cursor_advance(&zc);
2388 offset = zap_cursor_serialize(&zc);
2393 if (cooks != NULL) {
2396 KASSERT(ncooks >= 0, ("ncookies=%d", ncooks));
2399 zp->z_zn_prefetch = B_FALSE; /* a lookup will re-enable pre-fetching */
2401 /* Subtract unused cookies */
2402 if (ncookies != NULL)
2403 *ncookies -= ncooks;
2405 if (uio->uio_segflg == UIO_SYSSPACE && uio->uio_iovcnt == 1) {
2406 iovp->iov_base += outcount;
2407 iovp->iov_len -= outcount;
2408 uio->uio_resid -= outcount;
2409 } else if (error = uiomove(outbuf, (long)outcount, UIO_READ, uio)) {
2411 * Reset the pointer.
2413 offset = uio->uio_loffset;
2417 zap_cursor_fini(&zc);
2418 if (uio->uio_segflg != UIO_SYSSPACE || uio->uio_iovcnt != 1)
2419 kmem_free(outbuf, bufsize);
2421 if (error == ENOENT)
2424 ZFS_ACCESSTIME_STAMP(zfsvfs, zp);
2426 uio->uio_loffset = offset;
2428 if (error != 0 && cookies != NULL) {
2429 free(*cookies, M_TEMP);
2436 ulong_t zfs_fsync_sync_cnt = 4;
2439 zfs_fsync(vnode_t *vp, int syncflag, cred_t *cr, caller_context_t *ct)
2441 znode_t *zp = VTOZ(vp);
2442 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
2444 (void) tsd_set(zfs_fsyncer_key, (void *)zfs_fsync_sync_cnt);
2448 zil_commit(zfsvfs->z_log, zp->z_last_itx, zp->z_id);
2455 * Get the requested file attributes and place them in the provided
2458 * IN: vp - vnode of file.
2459 * vap - va_mask identifies requested attributes.
2460 * If AT_XVATTR set, then optional attrs are requested
2461 * flags - ATTR_NOACLCHECK (CIFS server context)
2462 * cr - credentials of caller.
2463 * ct - caller context
2465 * OUT: vap - attribute values.
2467 * RETURN: 0 (always succeeds)
2471 zfs_getattr(vnode_t *vp, vattr_t *vap, int flags, cred_t *cr,
2472 caller_context_t *ct)
2474 znode_t *zp = VTOZ(vp);
2475 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
2479 u_longlong_t nblocks;
2481 xvattr_t *xvap = (xvattr_t *)vap; /* vap may be an xvattr_t * */
2482 xoptattr_t *xoap = NULL;
2483 boolean_t skipaclchk = (flags & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE;
2490 * If ACL is trivial don't bother looking for ACE_READ_ATTRIBUTES.
2491 * Also, if we are the owner don't bother, since owner should
2492 * always be allowed to read basic attributes of file.
2494 if (!(pzp->zp_flags & ZFS_ACL_TRIVIAL) &&
2495 (pzp->zp_uid != crgetuid(cr))) {
2496 if (error = zfs_zaccess(zp, ACE_READ_ATTRIBUTES, 0,
2504 * Return all attributes. It's cheaper to provide the answer
2505 * than to determine whether we were asked the question.
2508 mutex_enter(&zp->z_lock);
2509 vap->va_type = IFTOVT(pzp->zp_mode);
2510 vap->va_mode = pzp->zp_mode & ~S_IFMT;
2511 zfs_fuid_map_ids(zp, cr, &vap->va_uid, &vap->va_gid);
2512 // vap->va_fsid = zp->z_zfsvfs->z_vfs->vfs_dev;
2513 vap->va_nodeid = zp->z_id;
2514 if ((vp->v_flag & VROOT) && zfs_show_ctldir(zp))
2515 links = pzp->zp_links + 1;
2517 links = pzp->zp_links;
2518 vap->va_nlink = MIN(links, UINT32_MAX); /* nlink_t limit! */
2519 vap->va_size = pzp->zp_size;
2520 vap->va_fsid = vp->v_mount->mnt_stat.f_fsid.val[0];
2521 vap->va_rdev = zfs_cmpldev(pzp->zp_rdev);
2522 vap->va_seq = zp->z_seq;
2523 vap->va_flags = 0; /* FreeBSD: Reset chflags(2) flags. */
2526 * Add in any requested optional attributes and the create time.
2527 * Also set the corresponding bits in the returned attribute bitmap.
2529 if ((xoap = xva_getxoptattr(xvap)) != NULL && zfsvfs->z_use_fuids) {
2530 if (XVA_ISSET_REQ(xvap, XAT_ARCHIVE)) {
2532 ((pzp->zp_flags & ZFS_ARCHIVE) != 0);
2533 XVA_SET_RTN(xvap, XAT_ARCHIVE);
2536 if (XVA_ISSET_REQ(xvap, XAT_READONLY)) {
2537 xoap->xoa_readonly =
2538 ((pzp->zp_flags & ZFS_READONLY) != 0);
2539 XVA_SET_RTN(xvap, XAT_READONLY);
2542 if (XVA_ISSET_REQ(xvap, XAT_SYSTEM)) {
2544 ((pzp->zp_flags & ZFS_SYSTEM) != 0);
2545 XVA_SET_RTN(xvap, XAT_SYSTEM);
2548 if (XVA_ISSET_REQ(xvap, XAT_HIDDEN)) {
2550 ((pzp->zp_flags & ZFS_HIDDEN) != 0);
2551 XVA_SET_RTN(xvap, XAT_HIDDEN);
2554 if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK)) {
2555 xoap->xoa_nounlink =
2556 ((pzp->zp_flags & ZFS_NOUNLINK) != 0);
2557 XVA_SET_RTN(xvap, XAT_NOUNLINK);
2560 if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE)) {
2561 xoap->xoa_immutable =
2562 ((pzp->zp_flags & ZFS_IMMUTABLE) != 0);
2563 XVA_SET_RTN(xvap, XAT_IMMUTABLE);
2566 if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY)) {
2567 xoap->xoa_appendonly =
2568 ((pzp->zp_flags & ZFS_APPENDONLY) != 0);
2569 XVA_SET_RTN(xvap, XAT_APPENDONLY);
2572 if (XVA_ISSET_REQ(xvap, XAT_NODUMP)) {
2574 ((pzp->zp_flags & ZFS_NODUMP) != 0);
2575 XVA_SET_RTN(xvap, XAT_NODUMP);
2578 if (XVA_ISSET_REQ(xvap, XAT_OPAQUE)) {
2580 ((pzp->zp_flags & ZFS_OPAQUE) != 0);
2581 XVA_SET_RTN(xvap, XAT_OPAQUE);
2584 if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED)) {
2585 xoap->xoa_av_quarantined =
2586 ((pzp->zp_flags & ZFS_AV_QUARANTINED) != 0);
2587 XVA_SET_RTN(xvap, XAT_AV_QUARANTINED);
2590 if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED)) {
2591 xoap->xoa_av_modified =
2592 ((pzp->zp_flags & ZFS_AV_MODIFIED) != 0);
2593 XVA_SET_RTN(xvap, XAT_AV_MODIFIED);
2596 if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP) &&
2597 vp->v_type == VREG &&
2598 (pzp->zp_flags & ZFS_BONUS_SCANSTAMP)) {
2600 dmu_object_info_t doi;
2603 * Only VREG files have anti-virus scanstamps, so we
2604 * won't conflict with symlinks in the bonus buffer.
2606 dmu_object_info_from_db(zp->z_dbuf, &doi);
2607 len = sizeof (xoap->xoa_av_scanstamp) +
2608 sizeof (znode_phys_t);
2609 if (len <= doi.doi_bonus_size) {
2611 * pzp points to the start of the
2612 * znode_phys_t. pzp + 1 points to the
2613 * first byte after the znode_phys_t.
2615 (void) memcpy(xoap->xoa_av_scanstamp,
2617 sizeof (xoap->xoa_av_scanstamp));
2618 XVA_SET_RTN(xvap, XAT_AV_SCANSTAMP);
2622 if (XVA_ISSET_REQ(xvap, XAT_CREATETIME)) {
2623 ZFS_TIME_DECODE(&xoap->xoa_createtime, pzp->zp_crtime);
2624 XVA_SET_RTN(xvap, XAT_CREATETIME);
2628 ZFS_TIME_DECODE(&vap->va_atime, pzp->zp_atime);
2629 ZFS_TIME_DECODE(&vap->va_mtime, pzp->zp_mtime);
2630 ZFS_TIME_DECODE(&vap->va_ctime, pzp->zp_ctime);
2631 ZFS_TIME_DECODE(&vap->va_birthtime, pzp->zp_crtime);
2633 mutex_exit(&zp->z_lock);
2635 dmu_object_size_from_db(zp->z_dbuf, &blksize, &nblocks);
2636 vap->va_blksize = blksize;
2637 vap->va_bytes = nblocks << 9; /* nblocks * 512 */
2639 if (zp->z_blksz == 0) {
2641 * Block size hasn't been set; suggest maximal I/O transfers.
2643 vap->va_blksize = zfsvfs->z_max_blksz;
2651 * Set the file attributes to the values contained in the
2654 * IN: vp - vnode of file to be modified.
2655 * vap - new attribute values.
2656 * If AT_XVATTR set, then optional attrs are being set
2657 * flags - ATTR_UTIME set if non-default time values provided.
2658 * - ATTR_NOACLCHECK (CIFS context only).
2659 * cr - credentials of caller.
2660 * ct - caller context
2662 * RETURN: 0 if success
2663 * error code if failure
2666 * vp - ctime updated, mtime updated if size changed.
2670 zfs_setattr(vnode_t *vp, vattr_t *vap, int flags, cred_t *cr,
2671 caller_context_t *ct)
2673 znode_t *zp = VTOZ(vp);
2675 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
2680 uint_t mask = vap->va_mask;
2682 uint64_t saved_mode;
2685 uint64_t new_uid, new_gid;
2687 int need_policy = FALSE;
2689 zfs_fuid_info_t *fuidp = NULL;
2690 xvattr_t *xvap = (xvattr_t *)vap; /* vap may be an xvattr_t * */
2692 zfs_acl_t *aclp = NULL;
2693 boolean_t skipaclchk = (flags & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE;
2694 boolean_t fuid_dirtied = B_FALSE;
2699 if (mask & AT_NOSET)
2706 zilog = zfsvfs->z_log;
2709 * Make sure that if we have ephemeral uid/gid or xvattr specified
2710 * that file system is at proper version level
2713 if (zfsvfs->z_use_fuids == B_FALSE &&
2714 (((mask & AT_UID) && IS_EPHEMERAL(vap->va_uid)) ||
2715 ((mask & AT_GID) && IS_EPHEMERAL(vap->va_gid)) ||
2716 (mask & AT_XVATTR))) {
2721 if (mask & AT_SIZE && vp->v_type == VDIR) {
2726 if (mask & AT_SIZE && vp->v_type != VREG && vp->v_type != VFIFO) {
2732 * If this is an xvattr_t, then get a pointer to the structure of
2733 * optional attributes. If this is NULL, then we have a vattr_t.
2735 xoap = xva_getxoptattr(xvap);
2737 xva_init(&tmpxvattr);
2740 * Immutable files can only alter immutable bit and atime
2742 if ((pzp->zp_flags & ZFS_IMMUTABLE) &&
2743 ((mask & (AT_SIZE|AT_UID|AT_GID|AT_MTIME|AT_MODE)) ||
2744 ((mask & AT_XVATTR) && XVA_ISSET_REQ(xvap, XAT_CREATETIME)))) {
2749 if ((mask & AT_SIZE) && (pzp->zp_flags & ZFS_READONLY)) {
2755 * Verify timestamps doesn't overflow 32 bits.
2756 * ZFS can handle large timestamps, but 32bit syscalls can't
2757 * handle times greater than 2039. This check should be removed
2758 * once large timestamps are fully supported.
2760 if (mask & (AT_ATIME | AT_MTIME)) {
2761 if (((mask & AT_ATIME) && TIMESPEC_OVERFLOW(&vap->va_atime)) ||
2762 ((mask & AT_MTIME) && TIMESPEC_OVERFLOW(&vap->va_mtime))) {
2771 /* Can this be moved to before the top label? */
2772 if (zfsvfs->z_vfs->vfs_flag & VFS_RDONLY) {
2778 * First validate permissions
2781 if (mask & AT_SIZE) {
2782 err = zfs_zaccess(zp, ACE_WRITE_DATA, 0, skipaclchk, cr);
2788 * XXX - Note, we are not providing any open
2789 * mode flags here (like FNDELAY), so we may
2790 * block if there are locks present... this
2791 * should be addressed in openat().
2793 /* XXX - would it be OK to generate a log record here? */
2794 err = zfs_freesp(zp, vap->va_size, 0, 0, FALSE);
2801 if (mask & (AT_ATIME|AT_MTIME) ||
2802 ((mask & AT_XVATTR) && (XVA_ISSET_REQ(xvap, XAT_HIDDEN) ||
2803 XVA_ISSET_REQ(xvap, XAT_READONLY) ||
2804 XVA_ISSET_REQ(xvap, XAT_ARCHIVE) ||
2805 XVA_ISSET_REQ(xvap, XAT_CREATETIME) ||
2806 XVA_ISSET_REQ(xvap, XAT_SYSTEM))))
2807 need_policy = zfs_zaccess(zp, ACE_WRITE_ATTRIBUTES, 0,
2810 if (mask & (AT_UID|AT_GID)) {
2811 int idmask = (mask & (AT_UID|AT_GID));
2816 * NOTE: even if a new mode is being set,
2817 * we may clear S_ISUID/S_ISGID bits.
2820 if (!(mask & AT_MODE))
2821 vap->va_mode = pzp->zp_mode;
2824 * Take ownership or chgrp to group we are a member of
2827 take_owner = (mask & AT_UID) && (vap->va_uid == crgetuid(cr));
2828 take_group = (mask & AT_GID) &&
2829 zfs_groupmember(zfsvfs, vap->va_gid, cr);
2832 * If both AT_UID and AT_GID are set then take_owner and
2833 * take_group must both be set in order to allow taking
2836 * Otherwise, send the check through secpolicy_vnode_setattr()
2840 if (((idmask == (AT_UID|AT_GID)) && take_owner && take_group) ||
2841 ((idmask == AT_UID) && take_owner) ||
2842 ((idmask == AT_GID) && take_group)) {
2843 if (zfs_zaccess(zp, ACE_WRITE_OWNER, 0,
2844 skipaclchk, cr) == 0) {
2846 * Remove setuid/setgid for non-privileged users
2848 secpolicy_setid_clear(vap, vp, cr);
2849 trim_mask = (mask & (AT_UID|AT_GID));
2858 mutex_enter(&zp->z_lock);
2859 oldva.va_mode = pzp->zp_mode;
2860 zfs_fuid_map_ids(zp, cr, &oldva.va_uid, &oldva.va_gid);
2861 if (mask & AT_XVATTR) {
2863 * Update xvattr mask to include only those attributes
2864 * that are actually changing.
2866 * the bits will be restored prior to actually setting
2867 * the attributes so the caller thinks they were set.
2869 if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY)) {
2870 if (xoap->xoa_appendonly !=
2871 ((pzp->zp_flags & ZFS_APPENDONLY) != 0)) {
2874 XVA_CLR_REQ(xvap, XAT_APPENDONLY);
2875 XVA_SET_REQ(&tmpxvattr, XAT_APPENDONLY);
2879 if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK)) {
2880 if (xoap->xoa_nounlink !=
2881 ((pzp->zp_flags & ZFS_NOUNLINK) != 0)) {
2884 XVA_CLR_REQ(xvap, XAT_NOUNLINK);
2885 XVA_SET_REQ(&tmpxvattr, XAT_NOUNLINK);
2889 if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE)) {
2890 if (xoap->xoa_immutable !=
2891 ((pzp->zp_flags & ZFS_IMMUTABLE) != 0)) {
2894 XVA_CLR_REQ(xvap, XAT_IMMUTABLE);
2895 XVA_SET_REQ(&tmpxvattr, XAT_IMMUTABLE);
2899 if (XVA_ISSET_REQ(xvap, XAT_NODUMP)) {
2900 if (xoap->xoa_nodump !=
2901 ((pzp->zp_flags & ZFS_NODUMP) != 0)) {
2904 XVA_CLR_REQ(xvap, XAT_NODUMP);
2905 XVA_SET_REQ(&tmpxvattr, XAT_NODUMP);
2909 if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED)) {
2910 if (xoap->xoa_av_modified !=
2911 ((pzp->zp_flags & ZFS_AV_MODIFIED) != 0)) {
2914 XVA_CLR_REQ(xvap, XAT_AV_MODIFIED);
2915 XVA_SET_REQ(&tmpxvattr, XAT_AV_MODIFIED);
2919 if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED)) {
2920 if ((vp->v_type != VREG &&
2921 xoap->xoa_av_quarantined) ||
2922 xoap->xoa_av_quarantined !=
2923 ((pzp->zp_flags & ZFS_AV_QUARANTINED) != 0)) {
2926 XVA_CLR_REQ(xvap, XAT_AV_QUARANTINED);
2927 XVA_SET_REQ(&tmpxvattr, XAT_AV_QUARANTINED);
2931 if (need_policy == FALSE &&
2932 (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP) ||
2933 XVA_ISSET_REQ(xvap, XAT_OPAQUE))) {
2938 mutex_exit(&zp->z_lock);
2940 if (mask & AT_MODE) {
2941 if (zfs_zaccess(zp, ACE_WRITE_ACL, 0, skipaclchk, cr) == 0) {
2942 err = secpolicy_setid_setsticky_clear(vp, vap,
2948 trim_mask |= AT_MODE;
2956 * If trim_mask is set then take ownership
2957 * has been granted or write_acl is present and user
2958 * has the ability to modify mode. In that case remove
2959 * UID|GID and or MODE from mask so that
2960 * secpolicy_vnode_setattr() doesn't revoke it.
2964 saved_mask = vap->va_mask;
2965 vap->va_mask &= ~trim_mask;
2966 if (trim_mask & AT_MODE) {
2968 * Save the mode, as secpolicy_vnode_setattr()
2969 * will overwrite it with ova.va_mode.
2971 saved_mode = vap->va_mode;
2974 err = secpolicy_vnode_setattr(cr, vp, vap, &oldva, flags,
2975 (int (*)(void *, int, cred_t *))zfs_zaccess_unix, zp);
2982 vap->va_mask |= saved_mask;
2983 if (trim_mask & AT_MODE) {
2985 * Recover the mode after
2986 * secpolicy_vnode_setattr().
2988 vap->va_mode = saved_mode;
2994 * secpolicy_vnode_setattr, or take ownership may have
2997 mask = vap->va_mask;
2999 tx = dmu_tx_create(zfsvfs->z_os);
3000 dmu_tx_hold_bonus(tx, zp->z_id);
3002 if (mask & AT_MODE) {
3003 uint64_t pmode = pzp->zp_mode;
3005 new_mode = (pmode & S_IFMT) | (vap->va_mode & ~S_IFMT);
3007 if (err = zfs_acl_chmod_setattr(zp, &aclp, new_mode))
3009 if (pzp->zp_acl.z_acl_extern_obj) {
3010 /* Are we upgrading ACL from old V0 format to new V1 */
3011 if (zfsvfs->z_version <= ZPL_VERSION_FUID &&
3012 pzp->zp_acl.z_acl_version ==
3013 ZFS_ACL_VERSION_INITIAL) {
3014 dmu_tx_hold_free(tx,
3015 pzp->zp_acl.z_acl_extern_obj, 0,
3017 dmu_tx_hold_write(tx, DMU_NEW_OBJECT,
3018 0, aclp->z_acl_bytes);
3020 dmu_tx_hold_write(tx,
3021 pzp->zp_acl.z_acl_extern_obj, 0,
3024 } else if (aclp->z_acl_bytes > ZFS_ACE_SPACE) {
3025 dmu_tx_hold_write(tx, DMU_NEW_OBJECT,
3026 0, aclp->z_acl_bytes);
3030 if (mask & (AT_UID | AT_GID)) {
3031 if (pzp->zp_xattr) {
3032 err = zfs_zget(zp->z_zfsvfs, pzp->zp_xattr, &attrzp);
3035 dmu_tx_hold_bonus(tx, attrzp->z_id);
3037 if (mask & AT_UID) {
3038 new_uid = zfs_fuid_create(zfsvfs,
3039 (uint64_t)vap->va_uid, cr, ZFS_OWNER, &fuidp);
3040 if (new_uid != pzp->zp_uid &&
3041 zfs_usergroup_overquota(zfsvfs, B_FALSE, new_uid)) {
3047 if (mask & AT_GID) {
3048 new_gid = zfs_fuid_create(zfsvfs, (uint64_t)vap->va_gid,
3049 cr, ZFS_GROUP, &fuidp);
3050 if (new_gid != pzp->zp_gid &&
3051 zfs_usergroup_overquota(zfsvfs, B_TRUE, new_gid)) {
3056 fuid_dirtied = zfsvfs->z_fuid_dirty;
3058 if (zfsvfs->z_fuid_obj == 0) {
3059 dmu_tx_hold_bonus(tx, DMU_NEW_OBJECT);
3060 dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0,
3061 FUID_SIZE_ESTIMATE(zfsvfs));
3062 dmu_tx_hold_zap(tx, MASTER_NODE_OBJ,
3065 dmu_tx_hold_bonus(tx, zfsvfs->z_fuid_obj);
3066 dmu_tx_hold_write(tx, zfsvfs->z_fuid_obj, 0,
3067 FUID_SIZE_ESTIMATE(zfsvfs));
3072 err = dmu_tx_assign(tx, TXG_NOWAIT);
3074 if (err == ERESTART)
3079 dmu_buf_will_dirty(zp->z_dbuf, tx);
3082 * Set each attribute requested.
3083 * We group settings according to the locks they need to acquire.
3085 * Note: you cannot set ctime directly, although it will be
3086 * updated as a side-effect of calling this function.
3089 mutex_enter(&zp->z_lock);
3091 if (mask & AT_MODE) {
3092 mutex_enter(&zp->z_acl_lock);
3093 zp->z_phys->zp_mode = new_mode;
3094 err = zfs_aclset_common(zp, aclp, cr, tx);
3095 ASSERT3U(err, ==, 0);
3096 zp->z_acl_cached = aclp;
3098 mutex_exit(&zp->z_acl_lock);
3102 mutex_enter(&attrzp->z_lock);
3104 if (mask & AT_UID) {
3105 pzp->zp_uid = new_uid;
3107 attrzp->z_phys->zp_uid = new_uid;
3110 if (mask & AT_GID) {
3111 pzp->zp_gid = new_gid;
3113 attrzp->z_phys->zp_gid = new_gid;
3117 mutex_exit(&attrzp->z_lock);
3119 if (mask & AT_ATIME)
3120 ZFS_TIME_ENCODE(&vap->va_atime, pzp->zp_atime);
3122 if (mask & AT_MTIME)
3123 ZFS_TIME_ENCODE(&vap->va_mtime, pzp->zp_mtime);
3125 /* XXX - shouldn't this be done *before* the ATIME/MTIME checks? */
3127 zfs_time_stamper_locked(zp, CONTENT_MODIFIED, tx);
3129 zfs_time_stamper_locked(zp, STATE_CHANGED, tx);
3131 * Do this after setting timestamps to prevent timestamp
3132 * update from toggling bit
3135 if (xoap && (mask & AT_XVATTR)) {
3138 * restore trimmed off masks
3139 * so that return masks can be set for caller.
3142 if (XVA_ISSET_REQ(&tmpxvattr, XAT_APPENDONLY)) {
3143 XVA_SET_REQ(xvap, XAT_APPENDONLY);
3145 if (XVA_ISSET_REQ(&tmpxvattr, XAT_NOUNLINK)) {
3146 XVA_SET_REQ(xvap, XAT_NOUNLINK);
3148 if (XVA_ISSET_REQ(&tmpxvattr, XAT_IMMUTABLE)) {
3149 XVA_SET_REQ(xvap, XAT_IMMUTABLE);
3151 if (XVA_ISSET_REQ(&tmpxvattr, XAT_NODUMP)) {
3152 XVA_SET_REQ(xvap, XAT_NODUMP);
3154 if (XVA_ISSET_REQ(&tmpxvattr, XAT_AV_MODIFIED)) {
3155 XVA_SET_REQ(xvap, XAT_AV_MODIFIED);
3157 if (XVA_ISSET_REQ(&tmpxvattr, XAT_AV_QUARANTINED)) {
3158 XVA_SET_REQ(xvap, XAT_AV_QUARANTINED);
3161 if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP)) {
3163 dmu_object_info_t doi;
3165 ASSERT(vp->v_type == VREG);
3167 /* Grow the bonus buffer if necessary. */
3168 dmu_object_info_from_db(zp->z_dbuf, &doi);
3169 len = sizeof (xoap->xoa_av_scanstamp) +
3170 sizeof (znode_phys_t);
3171 if (len > doi.doi_bonus_size)
3172 VERIFY(dmu_set_bonus(zp->z_dbuf, len, tx) == 0);
3174 zfs_xvattr_set(zp, xvap);
3178 zfs_fuid_sync(zfsvfs, tx);
3181 zfs_log_setattr(zilog, tx, TX_SETATTR, zp, vap, mask, fuidp);
3183 mutex_exit(&zp->z_lock);
3187 VN_RELE(ZTOV(attrzp));
3193 zfs_fuid_info_free(fuidp);
3202 if (err == ERESTART)
3209 typedef struct zfs_zlock {
3210 krwlock_t *zl_rwlock; /* lock we acquired */
3211 znode_t *zl_znode; /* znode we held */
3212 struct zfs_zlock *zl_next; /* next in list */
3216 * Drop locks and release vnodes that were held by zfs_rename_lock().
3219 zfs_rename_unlock(zfs_zlock_t **zlpp)
3223 while ((zl = *zlpp) != NULL) {
3224 if (zl->zl_znode != NULL)
3225 VN_RELE(ZTOV(zl->zl_znode));
3226 rw_exit(zl->zl_rwlock);
3227 *zlpp = zl->zl_next;
3228 kmem_free(zl, sizeof (*zl));
3233 * Search back through the directory tree, using the ".." entries.
3234 * Lock each directory in the chain to prevent concurrent renames.
3235 * Fail any attempt to move a directory into one of its own descendants.
3236 * XXX - z_parent_lock can overlap with map or grow locks
3239 zfs_rename_lock(znode_t *szp, znode_t *tdzp, znode_t *sdzp, zfs_zlock_t **zlpp)
3243 uint64_t rootid = zp->z_zfsvfs->z_root;
3244 uint64_t *oidp = &zp->z_id;
3245 krwlock_t *rwlp = &szp->z_parent_lock;
3246 krw_t rw = RW_WRITER;
3249 * First pass write-locks szp and compares to zp->z_id.
3250 * Later passes read-lock zp and compare to zp->z_parent.
3253 if (!rw_tryenter(rwlp, rw)) {
3255 * Another thread is renaming in this path.
3256 * Note that if we are a WRITER, we don't have any
3257 * parent_locks held yet.
3259 if (rw == RW_READER && zp->z_id > szp->z_id) {
3261 * Drop our locks and restart
3263 zfs_rename_unlock(&zl);
3267 rwlp = &szp->z_parent_lock;
3272 * Wait for other thread to drop its locks
3278 zl = kmem_alloc(sizeof (*zl), KM_SLEEP);
3279 zl->zl_rwlock = rwlp;
3280 zl->zl_znode = NULL;
3281 zl->zl_next = *zlpp;
3284 if (*oidp == szp->z_id) /* We're a descendant of szp */
3287 if (*oidp == rootid) /* We've hit the top */
3290 if (rw == RW_READER) { /* i.e. not the first pass */
3291 int error = zfs_zget(zp->z_zfsvfs, *oidp, &zp);
3296 oidp = &zp->z_phys->zp_parent;
3297 rwlp = &zp->z_parent_lock;
3300 } while (zp->z_id != sdzp->z_id);
3306 * Move an entry from the provided source directory to the target
3307 * directory. Change the entry name as indicated.
3309 * IN: sdvp - Source directory containing the "old entry".
3310 * snm - Old entry name.
3311 * tdvp - Target directory to contain the "new entry".
3312 * tnm - New entry name.
3313 * cr - credentials of caller.
3314 * ct - caller context
3315 * flags - case flags
3317 * RETURN: 0 if success
3318 * error code if failure
3321 * sdvp,tdvp - ctime|mtime updated
3325 zfs_rename(vnode_t *sdvp, char *snm, vnode_t *tdvp, char *tnm, cred_t *cr,
3326 caller_context_t *ct, int flags)
3328 znode_t *tdzp, *szp, *tzp;
3329 znode_t *sdzp = VTOZ(sdvp);
3330 zfsvfs_t *zfsvfs = sdzp->z_zfsvfs;
3333 zfs_dirlock_t *sdl, *tdl;
3336 int cmp, serr, terr;
3341 ZFS_VERIFY_ZP(sdzp);
3342 zilog = zfsvfs->z_log;
3345 * Make sure we have the real vp for the target directory.
3347 if (VOP_REALVP(tdvp, &realvp, ct) == 0)
3350 if (tdvp->v_vfsp != sdvp->v_vfsp || zfsctl_is_node(tdvp)) {
3356 ZFS_VERIFY_ZP(tdzp);
3357 if (zfsvfs->z_utf8 && u8_validate(tnm,
3358 strlen(tnm), NULL, U8_VALIDATE_ENTIRE, &error) < 0) {
3363 if (flags & FIGNORECASE)
3372 * This is to prevent the creation of links into attribute space
3373 * by renaming a linked file into/outof an attribute directory.
3374 * See the comment in zfs_link() for why this is considered bad.
3376 if ((tdzp->z_phys->zp_flags & ZFS_XATTR) !=
3377 (sdzp->z_phys->zp_flags & ZFS_XATTR)) {
3383 * Lock source and target directory entries. To prevent deadlock,
3384 * a lock ordering must be defined. We lock the directory with
3385 * the smallest object id first, or if it's a tie, the one with
3386 * the lexically first name.
3388 if (sdzp->z_id < tdzp->z_id) {
3390 } else if (sdzp->z_id > tdzp->z_id) {
3394 * First compare the two name arguments without
3395 * considering any case folding.
3397 int nofold = (zfsvfs->z_norm & ~U8_TEXTPREP_TOUPPER);
3399 cmp = u8_strcmp(snm, tnm, 0, nofold, U8_UNICODE_LATEST, &error);
3400 ASSERT(error == 0 || !zfsvfs->z_utf8);
3403 * POSIX: "If the old argument and the new argument
3404 * both refer to links to the same existing file,
3405 * the rename() function shall return successfully
3406 * and perform no other action."
3412 * If the file system is case-folding, then we may
3413 * have some more checking to do. A case-folding file
3414 * system is either supporting mixed case sensitivity
3415 * access or is completely case-insensitive. Note
3416 * that the file system is always case preserving.
3418 * In mixed sensitivity mode case sensitive behavior
3419 * is the default. FIGNORECASE must be used to
3420 * explicitly request case insensitive behavior.
3422 * If the source and target names provided differ only
3423 * by case (e.g., a request to rename 'tim' to 'Tim'),
3424 * we will treat this as a special case in the
3425 * case-insensitive mode: as long as the source name
3426 * is an exact match, we will allow this to proceed as
3427 * a name-change request.
3429 if ((zfsvfs->z_case == ZFS_CASE_INSENSITIVE ||
3430 (zfsvfs->z_case == ZFS_CASE_MIXED &&
3431 flags & FIGNORECASE)) &&
3432 u8_strcmp(snm, tnm, 0, zfsvfs->z_norm, U8_UNICODE_LATEST,
3435 * case preserving rename request, require exact
3444 * If the source and destination directories are the same, we should
3445 * grab the z_name_lock of that directory only once.
3449 rw_enter(&sdzp->z_name_lock, RW_READER);
3453 serr = zfs_dirent_lock(&sdl, sdzp, snm, &szp,
3454 ZEXISTS | zflg, NULL, NULL);
3455 terr = zfs_dirent_lock(&tdl,
3456 tdzp, tnm, &tzp, ZRENAMING | zflg, NULL, NULL);
3458 terr = zfs_dirent_lock(&tdl,
3459 tdzp, tnm, &tzp, zflg, NULL, NULL);
3460 serr = zfs_dirent_lock(&sdl,
3461 sdzp, snm, &szp, ZEXISTS | ZRENAMING | zflg,
3467 * Source entry invalid or not there.
3470 zfs_dirent_unlock(tdl);
3476 rw_exit(&sdzp->z_name_lock);
3478 if (strcmp(snm, ".") == 0 || strcmp(snm, "..") == 0)
3484 zfs_dirent_unlock(sdl);
3488 rw_exit(&sdzp->z_name_lock);
3490 if (strcmp(tnm, "..") == 0)
3497 * Must have write access at the source to remove the old entry
3498 * and write access at the target to create the new entry.
3499 * Note that if target and source are the same, this can be
3500 * done in a single check.
3503 if (error = zfs_zaccess_rename(sdzp, szp, tdzp, tzp, cr))
3506 if (ZTOV(szp)->v_type == VDIR) {
3508 * Check to make sure rename is valid.
3509 * Can't do a move like this: /usr/a/b to /usr/a/b/c/d
3511 if (error = zfs_rename_lock(szp, tdzp, sdzp, &zl))
3516 * Does target exist?
3520 * Source and target must be the same type.
3522 if (ZTOV(szp)->v_type == VDIR) {
3523 if (ZTOV(tzp)->v_type != VDIR) {
3528 if (ZTOV(tzp)->v_type == VDIR) {
3534 * POSIX dictates that when the source and target
3535 * entries refer to the same file object, rename
3536 * must do nothing and exit without error.
3538 if (szp->z_id == tzp->z_id) {
3544 vnevent_rename_src(ZTOV(szp), sdvp, snm, ct);
3546 vnevent_rename_dest(ZTOV(tzp), tdvp, tnm, ct);
3549 * notify the target directory if it is not the same
3550 * as source directory.
3553 vnevent_rename_dest_dir(tdvp, ct);
3556 tx = dmu_tx_create(zfsvfs->z_os);
3557 dmu_tx_hold_bonus(tx, szp->z_id); /* nlink changes */
3558 dmu_tx_hold_bonus(tx, sdzp->z_id); /* nlink changes */
3559 dmu_tx_hold_zap(tx, sdzp->z_id, FALSE, snm);
3560 dmu_tx_hold_zap(tx, tdzp->z_id, TRUE, tnm);
3562 dmu_tx_hold_bonus(tx, tdzp->z_id); /* nlink changes */
3564 dmu_tx_hold_bonus(tx, tzp->z_id); /* parent changes */
3565 dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL);
3566 error = dmu_tx_assign(tx, TXG_NOWAIT);
3569 zfs_rename_unlock(&zl);
3570 zfs_dirent_unlock(sdl);
3571 zfs_dirent_unlock(tdl);
3574 rw_exit(&sdzp->z_name_lock);
3579 if (error == ERESTART) {
3589 if (tzp) /* Attempt to remove the existing target */
3590 error = zfs_link_destroy(tdl, tzp, tx, zflg, NULL);
3593 error = zfs_link_create(tdl, szp, tx, ZRENAMING);
3595 szp->z_phys->zp_flags |= ZFS_AV_MODIFIED;
3597 error = zfs_link_destroy(sdl, szp, tx, ZRENAMING, NULL);
3600 zfs_log_rename(zilog, tx,
3601 TX_RENAME | (flags & FIGNORECASE ? TX_CI : 0),
3602 sdzp, sdl->dl_name, tdzp, tdl->dl_name, szp);
3604 /* Update path information for the target vnode */
3605 vn_renamepath(tdvp, ZTOV(szp), tnm, strlen(tnm));
3607 #ifdef FREEBSD_NAMECACHE
3618 zfs_rename_unlock(&zl);
3620 zfs_dirent_unlock(sdl);
3621 zfs_dirent_unlock(tdl);
3624 rw_exit(&sdzp->z_name_lock);
3636 * Insert the indicated symbolic reference entry into the directory.
3638 * IN: dvp - Directory to contain new symbolic link.
3639 * link - Name for new symlink entry.
3640 * vap - Attributes of new entry.
3641 * target - Target path of new symlink.
3642 * cr - credentials of caller.
3643 * ct - caller context
3644 * flags - case flags
3646 * RETURN: 0 if success
3647 * error code if failure
3650 * dvp - ctime|mtime updated
3654 zfs_symlink(vnode_t *dvp, vnode_t **vpp, char *name, vattr_t *vap, char *link,
3655 cred_t *cr, kthread_t *td)
3657 znode_t *zp, *dzp = VTOZ(dvp);
3660 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
3662 int len = strlen(link);
3665 zfs_acl_ids_t acl_ids;
3666 boolean_t fuid_dirtied;
3669 ASSERT(vap->va_type == VLNK);
3673 zilog = zfsvfs->z_log;
3675 if (zfsvfs->z_utf8 && u8_validate(name, strlen(name),
3676 NULL, U8_VALIDATE_ENTIRE, &error) < 0) {
3680 if (flags & FIGNORECASE)
3683 if (error = zfs_zaccess(dzp, ACE_ADD_FILE, 0, B_FALSE, cr)) {
3688 if (len > MAXPATHLEN) {
3690 return (ENAMETOOLONG);
3694 * Attempt to lock directory; fail if entry already exists.
3696 error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg, NULL, NULL);
3702 VERIFY(0 == zfs_acl_ids_create(dzp, 0, vap, cr, NULL, &acl_ids));
3703 if (zfs_acl_ids_overquota(zfsvfs, &acl_ids)) {
3704 zfs_acl_ids_free(&acl_ids);
3705 zfs_dirent_unlock(dl);
3709 tx = dmu_tx_create(zfsvfs->z_os);
3710 fuid_dirtied = zfsvfs->z_fuid_dirty;
3711 dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, MAX(1, len));
3712 dmu_tx_hold_bonus(tx, dzp->z_id);
3713 dmu_tx_hold_zap(tx, dzp->z_id, TRUE, name);
3714 if (acl_ids.z_aclp->z_acl_bytes > ZFS_ACE_SPACE)
3715 dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, SPA_MAXBLOCKSIZE);
3717 zfs_fuid_txhold(zfsvfs, tx);
3718 error = dmu_tx_assign(tx, TXG_NOWAIT);
3720 zfs_acl_ids_free(&acl_ids);
3721 zfs_dirent_unlock(dl);
3722 if (error == ERESTART) {
3732 dmu_buf_will_dirty(dzp->z_dbuf, tx);
3735 * Create a new object for the symlink.
3736 * Put the link content into bonus buffer if it will fit;
3737 * otherwise, store it just like any other file data.
3739 if (sizeof (znode_phys_t) + len <= dmu_bonus_max()) {
3740 zfs_mknode(dzp, vap, tx, cr, 0, &zp, len, &acl_ids);
3742 bcopy(link, zp->z_phys + 1, len);
3746 zfs_mknode(dzp, vap, tx, cr, 0, &zp, 0, &acl_ids);
3749 zfs_fuid_sync(zfsvfs, tx);
3751 * Nothing can access the znode yet so no locking needed
3752 * for growing the znode's blocksize.
3754 zfs_grow_blocksize(zp, len, tx);
3756 VERIFY(0 == dmu_buf_hold(zfsvfs->z_os,
3757 zp->z_id, 0, FTAG, &dbp));
3758 dmu_buf_will_dirty(dbp, tx);
3760 ASSERT3U(len, <=, dbp->db_size);
3761 bcopy(link, dbp->db_data, len);
3762 dmu_buf_rele(dbp, FTAG);
3764 zp->z_phys->zp_size = len;
3767 * Insert the new object into the directory.
3769 (void) zfs_link_create(dl, zp, tx, ZNEW);
3771 uint64_t txtype = TX_SYMLINK;
3772 if (flags & FIGNORECASE)
3774 zfs_log_symlink(zilog, tx, txtype, dzp, zp, name, link);
3778 zfs_acl_ids_free(&acl_ids);
3782 zfs_dirent_unlock(dl);
3789 * Return, in the buffer contained in the provided uio structure,
3790 * the symbolic path referred to by vp.
3792 * IN: vp - vnode of symbolic link.
3793 * uoip - structure to contain the link path.
3794 * cr - credentials of caller.
3795 * ct - caller context
3797 * OUT: uio - structure to contain the link path.
3799 * RETURN: 0 if success
3800 * error code if failure
3803 * vp - atime updated
3807 zfs_readlink(vnode_t *vp, uio_t *uio, cred_t *cr, caller_context_t *ct)
3809 znode_t *zp = VTOZ(vp);
3810 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
3817 bufsz = (size_t)zp->z_phys->zp_size;
3818 if (bufsz + sizeof (znode_phys_t) <= zp->z_dbuf->db_size) {
3819 error = uiomove(zp->z_phys + 1,
3820 MIN((size_t)bufsz, uio->uio_resid), UIO_READ, uio);
3823 error = dmu_buf_hold(zfsvfs->z_os, zp->z_id, 0, FTAG, &dbp);
3828 error = uiomove(dbp->db_data,
3829 MIN((size_t)bufsz, uio->uio_resid), UIO_READ, uio);
3830 dmu_buf_rele(dbp, FTAG);
3833 ZFS_ACCESSTIME_STAMP(zfsvfs, zp);
3839 * Insert a new entry into directory tdvp referencing svp.
3841 * IN: tdvp - Directory to contain new entry.
3842 * svp - vnode of new entry.
3843 * name - name of new entry.
3844 * cr - credentials of caller.
3845 * ct - caller context
3847 * RETURN: 0 if success
3848 * error code if failure
3851 * tdvp - ctime|mtime updated
3852 * svp - ctime updated
3856 zfs_link(vnode_t *tdvp, vnode_t *svp, char *name, cred_t *cr,
3857 caller_context_t *ct, int flags)
3859 znode_t *dzp = VTOZ(tdvp);
3861 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
3871 ASSERT(tdvp->v_type == VDIR);
3875 zilog = zfsvfs->z_log;
3877 if (VOP_REALVP(svp, &realvp, ct) == 0)
3881 * POSIX dictates that we return EPERM here.
3882 * Better choices include ENOTSUP or EISDIR.
3884 if (svp->v_type == VDIR) {
3889 if (svp->v_vfsp != tdvp->v_vfsp || zfsctl_is_node(svp)) {
3897 /* Prevent links to .zfs/shares files */
3899 if (szp->z_phys->zp_parent == zfsvfs->z_shares_dir) {
3904 if (zfsvfs->z_utf8 && u8_validate(name,
3905 strlen(name), NULL, U8_VALIDATE_ENTIRE, &error) < 0) {
3909 if (flags & FIGNORECASE)
3913 * We do not support links between attributes and non-attributes
3914 * because of the potential security risk of creating links
3915 * into "normal" file space in order to circumvent restrictions
3916 * imposed in attribute space.
3918 if ((szp->z_phys->zp_flags & ZFS_XATTR) !=
3919 (dzp->z_phys->zp_flags & ZFS_XATTR)) {
3925 owner = zfs_fuid_map_id(zfsvfs, szp->z_phys->zp_uid, cr, ZFS_OWNER);
3926 if (owner != crgetuid(cr) &&
3927 secpolicy_basic_link(svp, cr) != 0) {
3932 if (error = zfs_zaccess(dzp, ACE_ADD_FILE, 0, B_FALSE, cr)) {
3939 * Attempt to lock directory; fail if entry already exists.
3941 error = zfs_dirent_lock(&dl, dzp, name, &tzp, zf, NULL, NULL);
3947 tx = dmu_tx_create(zfsvfs->z_os);
3948 dmu_tx_hold_bonus(tx, szp->z_id);
3949 dmu_tx_hold_zap(tx, dzp->z_id, TRUE, name);
3950 error = dmu_tx_assign(tx, TXG_NOWAIT);
3952 zfs_dirent_unlock(dl);
3953 if (error == ERESTART) {
3963 error = zfs_link_create(dl, szp, tx, 0);
3966 uint64_t txtype = TX_LINK;
3967 if (flags & FIGNORECASE)
3969 zfs_log_link(zilog, tx, txtype, dzp, szp, name);
3974 zfs_dirent_unlock(dl);
3977 vnevent_link(svp, ct);
3986 zfs_inactive(vnode_t *vp, cred_t *cr, caller_context_t *ct)
3988 znode_t *zp = VTOZ(vp);
3989 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
3992 rw_enter(&zfsvfs->z_teardown_inactive_lock, RW_READER);
3993 if (zp->z_dbuf == NULL) {
3995 * The fs has been unmounted, or we did a
3996 * suspend/resume and this file no longer exists.
3999 vp->v_count = 0; /* count arrives as 1 */
4001 vrecycle(vp, curthread);
4002 rw_exit(&zfsvfs->z_teardown_inactive_lock);
4006 if (zp->z_atime_dirty && zp->z_unlinked == 0) {
4007 dmu_tx_t *tx = dmu_tx_create(zfsvfs->z_os);
4009 dmu_tx_hold_bonus(tx, zp->z_id);
4010 error = dmu_tx_assign(tx, TXG_WAIT);
4014 dmu_buf_will_dirty(zp->z_dbuf, tx);
4015 mutex_enter(&zp->z_lock);
4016 zp->z_atime_dirty = 0;
4017 mutex_exit(&zp->z_lock);
4023 rw_exit(&zfsvfs->z_teardown_inactive_lock);
4026 CTASSERT(sizeof(struct zfid_short) <= sizeof(struct fid));
4027 CTASSERT(sizeof(struct zfid_long) <= sizeof(struct fid));
4031 zfs_fid(vnode_t *vp, fid_t *fidp, caller_context_t *ct)
4033 znode_t *zp = VTOZ(vp);
4034 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
4036 uint64_t object = zp->z_id;
4042 gen = (uint32_t)zp->z_gen;
4044 size = (zfsvfs->z_parent != zfsvfs) ? LONG_FID_LEN : SHORT_FID_LEN;
4045 fidp->fid_len = size;
4047 zfid = (zfid_short_t *)fidp;
4049 zfid->zf_len = size;
4051 for (i = 0; i < sizeof (zfid->zf_object); i++)
4052 zfid->zf_object[i] = (uint8_t)(object >> (8 * i));
4054 /* Must have a non-zero generation number to distinguish from .zfs */
4057 for (i = 0; i < sizeof (zfid->zf_gen); i++)
4058 zfid->zf_gen[i] = (uint8_t)(gen >> (8 * i));
4060 if (size == LONG_FID_LEN) {
4061 uint64_t objsetid = dmu_objset_id(zfsvfs->z_os);
4064 zlfid = (zfid_long_t *)fidp;
4066 for (i = 0; i < sizeof (zlfid->zf_setid); i++)
4067 zlfid->zf_setid[i] = (uint8_t)(objsetid >> (8 * i));
4069 /* XXX - this should be the generation number for the objset */
4070 for (i = 0; i < sizeof (zlfid->zf_setgen); i++)
4071 zlfid->zf_setgen[i] = 0;
4079 zfs_pathconf(vnode_t *vp, int cmd, ulong_t *valp, cred_t *cr,
4080 caller_context_t *ct)
4092 case _PC_FILESIZEBITS:
4097 case _PC_XATTR_EXISTS:
4099 zfsvfs = zp->z_zfsvfs;
4103 error = zfs_dirent_lock(&dl, zp, "", &xzp,
4104 ZXATTR | ZEXISTS | ZSHARED, NULL, NULL);
4106 zfs_dirent_unlock(dl);
4107 if (!zfs_dirempty(xzp))
4110 } else if (error == ENOENT) {
4112 * If there aren't extended attributes, it's the
4113 * same as having zero of them.
4121 case _PC_ACL_EXTENDED:
4129 case _PC_ACL_PATH_MAX:
4130 *valp = ACL_MAX_ENTRIES;
4133 case _PC_MIN_HOLE_SIZE:
4134 *valp = (int)SPA_MINBLOCKSIZE;
4138 return (EOPNOTSUPP);
4144 zfs_getsecattr(vnode_t *vp, vsecattr_t *vsecp, int flag, cred_t *cr,
4145 caller_context_t *ct)
4147 znode_t *zp = VTOZ(vp);
4148 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
4150 boolean_t skipaclchk = (flag & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE;
4154 error = zfs_getacl(zp, vsecp, skipaclchk, cr);
4162 zfs_setsecattr(vnode_t *vp, vsecattr_t *vsecp, int flag, cred_t *cr,
4163 caller_context_t *ct)
4165 znode_t *zp = VTOZ(vp);
4166 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
4168 boolean_t skipaclchk = (flag & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE;
4172 error = zfs_setacl(zp, vsecp, skipaclchk, cr);
4178 zfs_freebsd_open(ap)
4179 struct vop_open_args /* {
4182 struct ucred *a_cred;
4183 struct thread *a_td;
4186 vnode_t *vp = ap->a_vp;
4187 znode_t *zp = VTOZ(vp);
4190 error = zfs_open(&vp, ap->a_mode, ap->a_cred, NULL);
4192 vnode_create_vobject(vp, zp->z_phys->zp_size, ap->a_td);
4197 zfs_freebsd_close(ap)
4198 struct vop_close_args /* {
4201 struct ucred *a_cred;
4202 struct thread *a_td;
4206 return (zfs_close(ap->a_vp, ap->a_fflag, 0, 0, ap->a_cred, NULL));
4210 zfs_freebsd_ioctl(ap)
4211 struct vop_ioctl_args /* {
4221 return (zfs_ioctl(ap->a_vp, ap->a_command, (intptr_t)ap->a_data,
4222 ap->a_fflag, ap->a_cred, NULL, NULL));
4226 zfs_freebsd_read(ap)
4227 struct vop_read_args /* {
4231 struct ucred *a_cred;
4235 return (zfs_read(ap->a_vp, ap->a_uio, ap->a_ioflag, ap->a_cred, NULL));
4239 zfs_freebsd_write(ap)
4240 struct vop_write_args /* {
4244 struct ucred *a_cred;
4248 return (zfs_write(ap->a_vp, ap->a_uio, ap->a_ioflag, ap->a_cred, NULL));
4252 zfs_freebsd_access(ap)
4253 struct vop_access_args /* {
4255 accmode_t a_accmode;
4256 struct ucred *a_cred;
4257 struct thread *a_td;
4264 * ZFS itself only knowns about VREAD, VWRITE, VEXEC and VAPPEND,
4266 accmode = ap->a_accmode & (VREAD|VWRITE|VEXEC|VAPPEND);
4268 error = zfs_access(ap->a_vp, accmode, 0, ap->a_cred, NULL);
4271 * VADMIN has to be handled by vaccess().
4274 accmode = ap->a_accmode & ~(VREAD|VWRITE|VEXEC|VAPPEND);
4276 vnode_t *vp = ap->a_vp;
4277 znode_t *zp = VTOZ(vp);
4278 znode_phys_t *zphys = zp->z_phys;
4280 error = vaccess(vp->v_type, zphys->zp_mode,
4281 zphys->zp_uid, zphys->zp_gid, accmode, ap->a_cred,
4290 zfs_freebsd_lookup(ap)
4291 struct vop_lookup_args /* {
4292 struct vnode *a_dvp;
4293 struct vnode **a_vpp;
4294 struct componentname *a_cnp;
4297 struct componentname *cnp = ap->a_cnp;
4298 char nm[NAME_MAX + 1];
4300 ASSERT(cnp->cn_namelen < sizeof(nm));
4301 strlcpy(nm, cnp->cn_nameptr, MIN(cnp->cn_namelen + 1, sizeof(nm)));
4303 return (zfs_lookup(ap->a_dvp, nm, ap->a_vpp, cnp, cnp->cn_nameiop,
4304 cnp->cn_cred, cnp->cn_thread, 0));
4308 zfs_freebsd_create(ap)
4309 struct vop_create_args /* {
4310 struct vnode *a_dvp;
4311 struct vnode **a_vpp;
4312 struct componentname *a_cnp;
4313 struct vattr *a_vap;
4316 struct componentname *cnp = ap->a_cnp;
4317 vattr_t *vap = ap->a_vap;
4320 ASSERT(cnp->cn_flags & SAVENAME);
4322 vattr_init_mask(vap);
4323 mode = vap->va_mode & ALLPERMS;
4325 return (zfs_create(ap->a_dvp, cnp->cn_nameptr, vap, !EXCL, mode,
4326 ap->a_vpp, cnp->cn_cred, cnp->cn_thread));
4330 zfs_freebsd_remove(ap)
4331 struct vop_remove_args /* {
4332 struct vnode *a_dvp;
4334 struct componentname *a_cnp;
4338 ASSERT(ap->a_cnp->cn_flags & SAVENAME);
4340 return (zfs_remove(ap->a_dvp, ap->a_cnp->cn_nameptr,
4341 ap->a_cnp->cn_cred, NULL, 0));
4345 zfs_freebsd_mkdir(ap)
4346 struct vop_mkdir_args /* {
4347 struct vnode *a_dvp;
4348 struct vnode **a_vpp;
4349 struct componentname *a_cnp;
4350 struct vattr *a_vap;
4353 vattr_t *vap = ap->a_vap;
4355 ASSERT(ap->a_cnp->cn_flags & SAVENAME);
4357 vattr_init_mask(vap);
4359 return (zfs_mkdir(ap->a_dvp, ap->a_cnp->cn_nameptr, vap, ap->a_vpp,
4360 ap->a_cnp->cn_cred, NULL, 0, NULL));
4364 zfs_freebsd_rmdir(ap)
4365 struct vop_rmdir_args /* {
4366 struct vnode *a_dvp;
4368 struct componentname *a_cnp;
4371 struct componentname *cnp = ap->a_cnp;
4373 ASSERT(cnp->cn_flags & SAVENAME);
4375 return (zfs_rmdir(ap->a_dvp, cnp->cn_nameptr, NULL, cnp->cn_cred, NULL, 0));
4379 zfs_freebsd_readdir(ap)
4380 struct vop_readdir_args /* {
4383 struct ucred *a_cred;
4390 return (zfs_readdir(ap->a_vp, ap->a_uio, ap->a_cred, ap->a_eofflag,
4391 ap->a_ncookies, ap->a_cookies));
4395 zfs_freebsd_fsync(ap)
4396 struct vop_fsync_args /* {
4399 struct thread *a_td;
4404 return (zfs_fsync(ap->a_vp, 0, ap->a_td->td_ucred, NULL));
4408 zfs_freebsd_getattr(ap)
4409 struct vop_getattr_args /* {
4411 struct vattr *a_vap;
4412 struct ucred *a_cred;
4413 struct thread *a_td;
4416 vattr_t *vap = ap->a_vap;
4422 xvap.xva_vattr = *vap;
4423 xvap.xva_vattr.va_mask |= AT_XVATTR;
4425 /* Convert chflags into ZFS-type flags. */
4426 /* XXX: what about SF_SETTABLE?. */
4427 XVA_SET_REQ(&xvap, XAT_IMMUTABLE);
4428 XVA_SET_REQ(&xvap, XAT_APPENDONLY);
4429 XVA_SET_REQ(&xvap, XAT_NOUNLINK);
4430 XVA_SET_REQ(&xvap, XAT_NODUMP);
4431 error = zfs_getattr(ap->a_vp, (vattr_t *)&xvap, 0, ap->a_cred, NULL);
4435 /* Convert ZFS xattr into chflags. */
4436 #define FLAG_CHECK(fflag, xflag, xfield) do { \
4437 if (XVA_ISSET_RTN(&xvap, (xflag)) && (xfield) != 0) \
4438 fflags |= (fflag); \
4440 FLAG_CHECK(SF_IMMUTABLE, XAT_IMMUTABLE,
4441 xvap.xva_xoptattrs.xoa_immutable);
4442 FLAG_CHECK(SF_APPEND, XAT_APPENDONLY,
4443 xvap.xva_xoptattrs.xoa_appendonly);
4444 FLAG_CHECK(SF_NOUNLINK, XAT_NOUNLINK,
4445 xvap.xva_xoptattrs.xoa_nounlink);
4446 FLAG_CHECK(UF_NODUMP, XAT_NODUMP,
4447 xvap.xva_xoptattrs.xoa_nodump);
4449 *vap = xvap.xva_vattr;
4450 vap->va_flags = fflags;
4455 zfs_freebsd_setattr(ap)
4456 struct vop_setattr_args /* {
4458 struct vattr *a_vap;
4459 struct ucred *a_cred;
4460 struct thread *a_td;
4463 vnode_t *vp = ap->a_vp;
4464 vattr_t *vap = ap->a_vap;
4465 cred_t *cred = ap->a_cred;
4470 vattr_init_mask(vap);
4471 vap->va_mask &= ~AT_NOSET;
4474 xvap.xva_vattr = *vap;
4476 zflags = VTOZ(vp)->z_phys->zp_flags;
4478 if (vap->va_flags != VNOVAL) {
4479 zfsvfs_t *zfsvfs = VTOZ(vp)->z_zfsvfs;
4482 if (zfsvfs->z_use_fuids == B_FALSE)
4483 return (EOPNOTSUPP);
4485 fflags = vap->va_flags;
4486 if ((fflags & ~(SF_IMMUTABLE|SF_APPEND|SF_NOUNLINK|UF_NODUMP)) != 0)
4487 return (EOPNOTSUPP);
4489 * Unprivileged processes are not permitted to unset system
4490 * flags, or modify flags if any system flags are set.
4491 * Privileged non-jail processes may not modify system flags
4492 * if securelevel > 0 and any existing system flags are set.
4493 * Privileged jail processes behave like privileged non-jail
4494 * processes if the security.jail.chflags_allowed sysctl is
4495 * is non-zero; otherwise, they behave like unprivileged
4498 if (secpolicy_fs_owner(vp->v_mount, cred) == 0 ||
4499 priv_check_cred(cred, PRIV_VFS_SYSFLAGS, 0) == 0) {
4501 (ZFS_IMMUTABLE | ZFS_APPENDONLY | ZFS_NOUNLINK)) {
4502 error = securelevel_gt(cred, 0);
4508 * Callers may only modify the file flags on objects they
4509 * have VADMIN rights for.
4511 if ((error = VOP_ACCESS(vp, VADMIN, cred, curthread)) != 0)
4514 (ZFS_IMMUTABLE | ZFS_APPENDONLY | ZFS_NOUNLINK)) {
4518 (SF_IMMUTABLE | SF_APPEND | SF_NOUNLINK)) {
4523 #define FLAG_CHANGE(fflag, zflag, xflag, xfield) do { \
4524 if (((fflags & (fflag)) && !(zflags & (zflag))) || \
4525 ((zflags & (zflag)) && !(fflags & (fflag)))) { \
4526 XVA_SET_REQ(&xvap, (xflag)); \
4527 (xfield) = ((fflags & (fflag)) != 0); \
4530 /* Convert chflags into ZFS-type flags. */
4531 /* XXX: what about SF_SETTABLE?. */
4532 FLAG_CHANGE(SF_IMMUTABLE, ZFS_IMMUTABLE, XAT_IMMUTABLE,
4533 xvap.xva_xoptattrs.xoa_immutable);
4534 FLAG_CHANGE(SF_APPEND, ZFS_APPENDONLY, XAT_APPENDONLY,
4535 xvap.xva_xoptattrs.xoa_appendonly);
4536 FLAG_CHANGE(SF_NOUNLINK, ZFS_NOUNLINK, XAT_NOUNLINK,
4537 xvap.xva_xoptattrs.xoa_nounlink);
4538 FLAG_CHANGE(UF_NODUMP, ZFS_NODUMP, XAT_NODUMP,
4539 xvap.xva_xoptattrs.xoa_nodump);
4542 return (zfs_setattr(vp, (vattr_t *)&xvap, 0, cred, NULL));
4546 zfs_freebsd_rename(ap)
4547 struct vop_rename_args /* {
4548 struct vnode *a_fdvp;
4549 struct vnode *a_fvp;
4550 struct componentname *a_fcnp;
4551 struct vnode *a_tdvp;
4552 struct vnode *a_tvp;
4553 struct componentname *a_tcnp;
4556 vnode_t *fdvp = ap->a_fdvp;
4557 vnode_t *fvp = ap->a_fvp;
4558 vnode_t *tdvp = ap->a_tdvp;
4559 vnode_t *tvp = ap->a_tvp;
4562 ASSERT(ap->a_fcnp->cn_flags & (SAVENAME|SAVESTART));
4563 ASSERT(ap->a_tcnp->cn_flags & (SAVENAME|SAVESTART));
4565 error = zfs_rename(fdvp, ap->a_fcnp->cn_nameptr, tdvp,
4566 ap->a_tcnp->cn_nameptr, ap->a_fcnp->cn_cred, NULL, 0);
4581 zfs_freebsd_symlink(ap)
4582 struct vop_symlink_args /* {
4583 struct vnode *a_dvp;
4584 struct vnode **a_vpp;
4585 struct componentname *a_cnp;
4586 struct vattr *a_vap;
4590 struct componentname *cnp = ap->a_cnp;
4591 vattr_t *vap = ap->a_vap;
4593 ASSERT(cnp->cn_flags & SAVENAME);
4595 vap->va_type = VLNK; /* FreeBSD: Syscall only sets va_mode. */
4596 vattr_init_mask(vap);
4598 return (zfs_symlink(ap->a_dvp, ap->a_vpp, cnp->cn_nameptr, vap,
4599 ap->a_target, cnp->cn_cred, cnp->cn_thread));
4603 zfs_freebsd_readlink(ap)
4604 struct vop_readlink_args /* {
4607 struct ucred *a_cred;
4611 return (zfs_readlink(ap->a_vp, ap->a_uio, ap->a_cred, NULL));
4615 zfs_freebsd_link(ap)
4616 struct vop_link_args /* {
4617 struct vnode *a_tdvp;
4619 struct componentname *a_cnp;
4622 struct componentname *cnp = ap->a_cnp;
4624 ASSERT(cnp->cn_flags & SAVENAME);
4626 return (zfs_link(ap->a_tdvp, ap->a_vp, cnp->cn_nameptr, cnp->cn_cred, NULL, 0));
4630 zfs_freebsd_inactive(ap)
4631 struct vop_inactive_args /* {
4633 struct thread *a_td;
4636 vnode_t *vp = ap->a_vp;
4638 zfs_inactive(vp, ap->a_td->td_ucred, NULL);
4643 zfs_reclaim_complete(void *arg, int pending)
4646 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
4648 rw_enter(&zfsvfs->z_teardown_inactive_lock, RW_READER);
4649 if (zp->z_dbuf != NULL) {
4650 ZFS_OBJ_HOLD_ENTER(zfsvfs, zp->z_id);
4651 zfs_znode_dmu_fini(zp);
4652 ZFS_OBJ_HOLD_EXIT(zfsvfs, zp->z_id);
4655 rw_exit(&zfsvfs->z_teardown_inactive_lock);
4657 * If the file system is being unmounted, there is a process waiting
4658 * for us, wake it up.
4660 if (zfsvfs->z_unmounted)
4665 zfs_freebsd_reclaim(ap)
4666 struct vop_reclaim_args /* {
4668 struct thread *a_td;
4671 vnode_t *vp = ap->a_vp;
4672 znode_t *zp = VTOZ(vp);
4673 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
4675 rw_enter(&zfsvfs->z_teardown_inactive_lock, RW_READER);
4680 * Destroy the vm object and flush associated pages.
4682 vnode_destroy_vobject(vp);
4684 mutex_enter(&zp->z_lock);
4685 ASSERT(zp->z_phys != NULL);
4687 mutex_exit(&zp->z_lock);
4691 else if (zp->z_dbuf == NULL)
4693 else /* if (!zp->z_unlinked && zp->z_dbuf != NULL) */ {
4696 locked = MUTEX_HELD(ZFS_OBJ_MUTEX(zfsvfs, zp->z_id)) ? 2 :
4697 ZFS_OBJ_HOLD_TRYENTER(zfsvfs, zp->z_id);
4700 * Lock can't be obtained due to deadlock possibility,
4701 * so defer znode destruction.
4703 TASK_INIT(&zp->z_task, 0, zfs_reclaim_complete, zp);
4704 taskqueue_enqueue(taskqueue_thread, &zp->z_task);
4706 zfs_znode_dmu_fini(zp);
4708 ZFS_OBJ_HOLD_EXIT(zfsvfs, zp->z_id);
4714 ASSERT(vp->v_holdcnt >= 1);
4716 rw_exit(&zfsvfs->z_teardown_inactive_lock);
4722 struct vop_fid_args /* {
4728 return (zfs_fid(ap->a_vp, (void *)ap->a_fid, NULL));
4732 zfs_freebsd_pathconf(ap)
4733 struct vop_pathconf_args /* {
4736 register_t *a_retval;
4742 error = zfs_pathconf(ap->a_vp, ap->a_name, &val, curthread->td_ucred, NULL);
4744 *ap->a_retval = val;
4745 else if (error == EOPNOTSUPP)
4746 error = vop_stdpathconf(ap);
4751 zfs_freebsd_fifo_pathconf(ap)
4752 struct vop_pathconf_args /* {
4755 register_t *a_retval;
4759 switch (ap->a_name) {
4760 case _PC_ACL_EXTENDED:
4762 case _PC_ACL_PATH_MAX:
4763 case _PC_MAC_PRESENT:
4764 return (zfs_freebsd_pathconf(ap));
4766 return (fifo_specops.vop_pathconf(ap));
4771 * FreeBSD's extended attributes namespace defines file name prefix for ZFS'
4772 * extended attribute name:
4775 * system freebsd:system:
4776 * user (none, can be used to access ZFS fsattr(5) attributes
4777 * created on Solaris)
4780 zfs_create_attrname(int attrnamespace, const char *name, char *attrname,
4783 const char *namespace, *prefix, *suffix;
4785 /* We don't allow '/' character in attribute name. */
4786 if (strchr(name, '/') != NULL)
4788 /* We don't allow attribute names that start with "freebsd:" string. */
4789 if (strncmp(name, "freebsd:", 8) == 0)
4792 bzero(attrname, size);
4794 switch (attrnamespace) {
4795 case EXTATTR_NAMESPACE_USER:
4797 prefix = "freebsd:";
4798 namespace = EXTATTR_NAMESPACE_USER_STRING;
4802 * This is the default namespace by which we can access all
4803 * attributes created on Solaris.
4805 prefix = namespace = suffix = "";
4808 case EXTATTR_NAMESPACE_SYSTEM:
4809 prefix = "freebsd:";
4810 namespace = EXTATTR_NAMESPACE_SYSTEM_STRING;
4813 case EXTATTR_NAMESPACE_EMPTY:
4817 if (snprintf(attrname, size, "%s%s%s%s", prefix, namespace, suffix,
4819 return (ENAMETOOLONG);
4825 * Vnode operating to retrieve a named extended attribute.
4828 zfs_getextattr(struct vop_getextattr_args *ap)
4831 IN struct vnode *a_vp;
4832 IN int a_attrnamespace;
4833 IN const char *a_name;
4834 INOUT struct uio *a_uio;
4836 IN struct ucred *a_cred;
4837 IN struct thread *a_td;
4841 zfsvfs_t *zfsvfs = VTOZ(ap->a_vp)->z_zfsvfs;
4842 struct thread *td = ap->a_td;
4843 struct nameidata nd;
4846 vnode_t *xvp = NULL, *vp;
4849 error = extattr_check_cred(ap->a_vp, ap->a_attrnamespace,
4850 ap->a_cred, ap->a_td, VREAD);
4854 error = zfs_create_attrname(ap->a_attrnamespace, ap->a_name, attrname,
4861 error = zfs_lookup(ap->a_vp, NULL, &xvp, NULL, 0, ap->a_cred, td,
4869 NDINIT_ATVP(&nd, LOOKUP, NOFOLLOW | MPSAFE, UIO_SYSSPACE, attrname,
4871 error = vn_open_cred(&nd, &flags, 0, 0, ap->a_cred, NULL);
4873 NDFREE(&nd, NDF_ONLY_PNBUF);
4876 if (error == ENOENT)
4881 if (ap->a_size != NULL) {
4882 error = VOP_GETATTR(vp, &va, ap->a_cred);
4884 *ap->a_size = (size_t)va.va_size;
4885 } else if (ap->a_uio != NULL)
4886 error = VOP_READ(vp, ap->a_uio, IO_UNIT | IO_SYNC, ap->a_cred);
4889 vn_close(vp, flags, ap->a_cred, td);
4896 * Vnode operation to remove a named attribute.
4899 zfs_deleteextattr(struct vop_deleteextattr_args *ap)
4902 IN struct vnode *a_vp;
4903 IN int a_attrnamespace;
4904 IN const char *a_name;
4905 IN struct ucred *a_cred;
4906 IN struct thread *a_td;
4910 zfsvfs_t *zfsvfs = VTOZ(ap->a_vp)->z_zfsvfs;
4911 struct thread *td = ap->a_td;
4912 struct nameidata nd;
4915 vnode_t *xvp = NULL, *vp;
4918 error = extattr_check_cred(ap->a_vp, ap->a_attrnamespace,
4919 ap->a_cred, ap->a_td, VWRITE);
4923 error = zfs_create_attrname(ap->a_attrnamespace, ap->a_name, attrname,
4930 error = zfs_lookup(ap->a_vp, NULL, &xvp, NULL, 0, ap->a_cred, td,
4937 NDINIT_ATVP(&nd, DELETE, NOFOLLOW | LOCKPARENT | LOCKLEAF | MPSAFE,
4938 UIO_SYSSPACE, attrname, xvp, td);
4941 NDFREE(&nd, NDF_ONLY_PNBUF);
4944 if (error == ENOENT)
4948 error = VOP_REMOVE(nd.ni_dvp, vp, &nd.ni_cnd);
4951 if (vp == nd.ni_dvp)
4961 * Vnode operation to set a named attribute.
4964 zfs_setextattr(struct vop_setextattr_args *ap)
4967 IN struct vnode *a_vp;
4968 IN int a_attrnamespace;
4969 IN const char *a_name;
4970 INOUT struct uio *a_uio;
4971 IN struct ucred *a_cred;
4972 IN struct thread *a_td;
4976 zfsvfs_t *zfsvfs = VTOZ(ap->a_vp)->z_zfsvfs;
4977 struct thread *td = ap->a_td;
4978 struct nameidata nd;
4981 vnode_t *xvp = NULL, *vp;
4984 error = extattr_check_cred(ap->a_vp, ap->a_attrnamespace,
4985 ap->a_cred, ap->a_td, VWRITE);
4989 error = zfs_create_attrname(ap->a_attrnamespace, ap->a_name, attrname,
4996 error = zfs_lookup(ap->a_vp, NULL, &xvp, NULL, 0, ap->a_cred, td,
4997 LOOKUP_XATTR | CREATE_XATTR_DIR);
5003 flags = FFLAGS(O_WRONLY | O_CREAT);
5004 NDINIT_ATVP(&nd, LOOKUP, NOFOLLOW | MPSAFE, UIO_SYSSPACE, attrname,
5006 error = vn_open_cred(&nd, &flags, 0600, 0, ap->a_cred, NULL);
5008 NDFREE(&nd, NDF_ONLY_PNBUF);
5016 error = VOP_SETATTR(vp, &va, ap->a_cred);
5018 VOP_WRITE(vp, ap->a_uio, IO_UNIT | IO_SYNC, ap->a_cred);
5021 vn_close(vp, flags, ap->a_cred, td);
5028 * Vnode operation to retrieve extended attributes on a vnode.
5031 zfs_listextattr(struct vop_listextattr_args *ap)
5034 IN struct vnode *a_vp;
5035 IN int a_attrnamespace;
5036 INOUT struct uio *a_uio;
5038 IN struct ucred *a_cred;
5039 IN struct thread *a_td;
5043 zfsvfs_t *zfsvfs = VTOZ(ap->a_vp)->z_zfsvfs;
5044 struct thread *td = ap->a_td;
5045 struct nameidata nd;
5046 char attrprefix[16];
5047 u_char dirbuf[sizeof(struct dirent)];
5050 struct uio auio, *uio = ap->a_uio;
5051 size_t *sizep = ap->a_size;
5053 vnode_t *xvp = NULL, *vp;
5054 int done, error, eof, pos;
5056 error = extattr_check_cred(ap->a_vp, ap->a_attrnamespace,
5057 ap->a_cred, ap->a_td, VREAD);
5061 error = zfs_create_attrname(ap->a_attrnamespace, "", attrprefix,
5062 sizeof(attrprefix));
5065 plen = strlen(attrprefix);
5072 error = zfs_lookup(ap->a_vp, NULL, &xvp, NULL, 0, ap->a_cred, td,
5077 * ENOATTR means that the EA directory does not yet exist,
5078 * i.e. there are no extended attributes there.
5080 if (error == ENOATTR)
5085 NDINIT_ATVP(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | LOCKSHARED | MPSAFE,
5086 UIO_SYSSPACE, ".", xvp, td);
5089 NDFREE(&nd, NDF_ONLY_PNBUF);
5095 auio.uio_iov = &aiov;
5096 auio.uio_iovcnt = 1;
5097 auio.uio_segflg = UIO_SYSSPACE;
5099 auio.uio_rw = UIO_READ;
5100 auio.uio_offset = 0;
5105 aiov.iov_base = (void *)dirbuf;
5106 aiov.iov_len = sizeof(dirbuf);
5107 auio.uio_resid = sizeof(dirbuf);
5108 error = VOP_READDIR(vp, &auio, ap->a_cred, &eof, NULL, NULL);
5109 done = sizeof(dirbuf) - auio.uio_resid;
5112 for (pos = 0; pos < done;) {
5113 dp = (struct dirent *)(dirbuf + pos);
5114 pos += dp->d_reclen;
5116 * XXX: Temporarily we also accept DT_UNKNOWN, as this
5117 * is what we get when attribute was created on Solaris.
5119 if (dp->d_type != DT_REG && dp->d_type != DT_UNKNOWN)
5121 if (plen == 0 && strncmp(dp->d_name, "freebsd:", 8) == 0)
5123 else if (strncmp(dp->d_name, attrprefix, plen) != 0)
5125 nlen = dp->d_namlen - plen;
5128 else if (uio != NULL) {
5130 * Format of extattr name entry is one byte for
5131 * length and the rest for name.
5133 error = uiomove(&nlen, 1, uio->uio_rw, uio);
5135 error = uiomove(dp->d_name + plen, nlen,
5142 } while (!eof && error == 0);
5151 zfs_freebsd_getacl(ap)
5152 struct vop_getacl_args /* {
5161 vsecattr_t vsecattr;
5163 if (ap->a_type != ACL_TYPE_NFS4)
5166 vsecattr.vsa_mask = VSA_ACE | VSA_ACECNT;
5167 if (error = zfs_getsecattr(ap->a_vp, &vsecattr, 0, ap->a_cred, NULL))
5170 error = acl_from_aces(ap->a_aclp, vsecattr.vsa_aclentp, vsecattr.vsa_aclcnt);
5171 if (vsecattr.vsa_aclentp != NULL)
5172 kmem_free(vsecattr.vsa_aclentp, vsecattr.vsa_aclentsz);
5178 zfs_freebsd_setacl(ap)
5179 struct vop_setacl_args /* {
5188 vsecattr_t vsecattr;
5189 int aclbsize; /* size of acl list in bytes */
5192 if (ap->a_type != ACL_TYPE_NFS4)
5195 if (ap->a_aclp->acl_cnt < 1 || ap->a_aclp->acl_cnt > MAX_ACL_ENTRIES)
5199 * With NFSv4 ACLs, chmod(2) may need to add additional entries,
5200 * splitting every entry into two and appending "canonical six"
5201 * entries at the end. Don't allow for setting an ACL that would
5202 * cause chmod(2) to run out of ACL entries.
5204 if (ap->a_aclp->acl_cnt * 2 + 6 > ACL_MAX_ENTRIES)
5207 error = acl_nfs4_check(ap->a_aclp, ap->a_vp->v_type == VDIR);
5211 vsecattr.vsa_mask = VSA_ACE;
5212 aclbsize = ap->a_aclp->acl_cnt * sizeof(ace_t);
5213 vsecattr.vsa_aclentp = kmem_alloc(aclbsize, KM_SLEEP);
5214 aaclp = vsecattr.vsa_aclentp;
5215 vsecattr.vsa_aclentsz = aclbsize;
5217 aces_from_acl(vsecattr.vsa_aclentp, &vsecattr.vsa_aclcnt, ap->a_aclp);
5218 error = zfs_setsecattr(ap->a_vp, &vsecattr, 0, ap->a_cred, NULL);
5219 kmem_free(aaclp, aclbsize);
5225 zfs_freebsd_aclcheck(ap)
5226 struct vop_aclcheck_args /* {
5235 return (EOPNOTSUPP);
5238 struct vop_vector zfs_vnodeops;
5239 struct vop_vector zfs_fifoops;
5240 struct vop_vector zfs_shareops;
5242 struct vop_vector zfs_vnodeops = {
5243 .vop_default = &default_vnodeops,
5244 .vop_inactive = zfs_freebsd_inactive,
5245 .vop_reclaim = zfs_freebsd_reclaim,
5246 .vop_access = zfs_freebsd_access,
5247 #ifdef FREEBSD_NAMECACHE
5248 .vop_lookup = vfs_cache_lookup,
5249 .vop_cachedlookup = zfs_freebsd_lookup,
5251 .vop_lookup = zfs_freebsd_lookup,
5253 .vop_getattr = zfs_freebsd_getattr,
5254 .vop_setattr = zfs_freebsd_setattr,
5255 .vop_create = zfs_freebsd_create,
5256 .vop_mknod = zfs_freebsd_create,
5257 .vop_mkdir = zfs_freebsd_mkdir,
5258 .vop_readdir = zfs_freebsd_readdir,
5259 .vop_fsync = zfs_freebsd_fsync,
5260 .vop_open = zfs_freebsd_open,
5261 .vop_close = zfs_freebsd_close,
5262 .vop_rmdir = zfs_freebsd_rmdir,
5263 .vop_ioctl = zfs_freebsd_ioctl,
5264 .vop_link = zfs_freebsd_link,
5265 .vop_symlink = zfs_freebsd_symlink,
5266 .vop_readlink = zfs_freebsd_readlink,
5267 .vop_read = zfs_freebsd_read,
5268 .vop_write = zfs_freebsd_write,
5269 .vop_remove = zfs_freebsd_remove,
5270 .vop_rename = zfs_freebsd_rename,
5271 .vop_pathconf = zfs_freebsd_pathconf,
5272 .vop_bmap = VOP_EOPNOTSUPP,
5273 .vop_fid = zfs_freebsd_fid,
5274 .vop_getextattr = zfs_getextattr,
5275 .vop_deleteextattr = zfs_deleteextattr,
5276 .vop_setextattr = zfs_setextattr,
5277 .vop_listextattr = zfs_listextattr,
5278 .vop_getacl = zfs_freebsd_getacl,
5279 .vop_setacl = zfs_freebsd_setacl,
5280 .vop_aclcheck = zfs_freebsd_aclcheck,
5283 struct vop_vector zfs_fifoops = {
5284 .vop_default = &fifo_specops,
5285 .vop_fsync = zfs_freebsd_fsync,
5286 .vop_access = zfs_freebsd_access,
5287 .vop_getattr = zfs_freebsd_getattr,
5288 .vop_inactive = zfs_freebsd_inactive,
5289 .vop_read = VOP_PANIC,
5290 .vop_reclaim = zfs_freebsd_reclaim,
5291 .vop_setattr = zfs_freebsd_setattr,
5292 .vop_write = VOP_PANIC,
5293 .vop_pathconf = zfs_freebsd_fifo_pathconf,
5294 .vop_fid = zfs_freebsd_fid,
5295 .vop_getacl = zfs_freebsd_getacl,
5296 .vop_setacl = zfs_freebsd_setacl,
5297 .vop_aclcheck = zfs_freebsd_aclcheck,
5301 * special share hidden files vnode operations template
5303 struct vop_vector zfs_shareops = {
5304 .vop_default = &default_vnodeops,
5305 .vop_access = zfs_freebsd_access,
5306 .vop_inactive = zfs_freebsd_inactive,
5307 .vop_reclaim = zfs_freebsd_reclaim,
5308 .vop_fid = zfs_freebsd_fid,
5309 .vop_pathconf = zfs_freebsd_pathconf,