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 ((pp->oflags & VPO_BUSY) != 0) {
327 * Reference the page before unlocking and
328 * sleeping so that the page daemon is less
329 * likely to reclaim it.
331 vm_page_lock_queues();
332 vm_page_flag_set(pp, PG_REFERENCED);
333 vm_page_sleep(pp, "zfsmwb");
337 vm_page_lock_queues();
339 vm_page_unlock_queues();
341 if (__predict_false(obj->cache != NULL)) {
342 vm_page_cache_free(obj, OFF_TO_IDX(start),
343 OFF_TO_IDX(start) + 1);
353 page_unlock(vm_page_t pp)
360 zfs_map_page(vm_page_t pp, struct sf_buf **sfp)
364 *sfp = sf_buf_alloc(pp, SFB_CPUPRIVATE);
365 return ((caddr_t)sf_buf_kva(*sfp));
369 zfs_unmap_page(struct sf_buf *sf)
378 * When a file is memory mapped, we must keep the IO data synchronized
379 * between the DMU cache and the memory mapped pages. What this means:
381 * On Write: If we find a memory mapped page, we write to *both*
382 * the page and the dmu buffer.
386 update_pages(vnode_t *vp, int64_t start, int len, objset_t *os, uint64_t oid,
387 int segflg, dmu_tx_t *tx)
393 ASSERT(vp->v_mount != NULL);
397 off = start & PAGEOFFSET;
399 for (start &= PAGEMASK; len > 0; start += PAGESIZE) {
401 uint64_t nbytes = MIN(PAGESIZE - off, len);
403 if ((pp = page_lookup(vp, start, off, nbytes)) != NULL) {
406 VM_OBJECT_UNLOCK(obj);
407 va = zfs_map_page(pp, &sf);
408 if (segflg == UIO_NOCOPY) {
409 (void) dmu_write(os, oid, start+off, nbytes,
412 (void) dmu_read(os, oid, start+off, nbytes,
413 va+off, DMU_READ_PREFETCH);;
423 VM_OBJECT_UNLOCK(obj);
427 * When a file is memory mapped, we must keep the IO data synchronized
428 * between the DMU cache and the memory mapped pages. What this means:
430 * On Read: We "read" preferentially from memory mapped pages,
431 * else we default from the dmu buffer.
433 * NOTE: We will always "break up" the IO into PAGESIZE uiomoves when
434 * the file is memory mapped.
437 mappedread(vnode_t *vp, int nbytes, uio_t *uio)
439 znode_t *zp = VTOZ(vp);
440 objset_t *os = zp->z_zfsvfs->z_os;
450 ASSERT(vp->v_mount != NULL);
454 start = uio->uio_loffset;
455 off = start & PAGEOFFSET;
458 for (start &= PAGEMASK; len > 0; start += PAGESIZE) {
459 uint64_t bytes = MIN(PAGESIZE - off, len);
462 if ((m = vm_page_lookup(obj, OFF_TO_IDX(start))) != NULL &&
463 vm_page_is_valid(m, (vm_offset_t)off, bytes)) {
464 if ((m->oflags & VPO_BUSY) != 0) {
466 * Reference the page before unlocking and
467 * sleeping so that the page daemon is less
468 * likely to reclaim it.
470 vm_page_lock_queues();
471 vm_page_flag_set(m, PG_REFERENCED);
472 vm_page_sleep(m, "zfsmrb");
477 VM_OBJECT_UNLOCK(obj);
479 error = dmu_read_uio(os, zp->z_id, uio,
485 sf = sf_buf_alloc(m, SFB_CPUPRIVATE);
486 va = (caddr_t)sf_buf_kva(sf);
487 error = uiomove(va + off, bytes, UIO_READ, uio);
493 } else if (m != NULL && uio->uio_segflg == UIO_NOCOPY) {
495 * The code below is here to make sendfile(2) work
496 * correctly with ZFS. As pointed out by ups@
497 * sendfile(2) should be changed to use VOP_GETPAGES(),
498 * but it pessimize performance of sendfile/UFS, that's
499 * why I handle this special case in ZFS code.
501 if ((m->oflags & VPO_BUSY) != 0) {
503 * Reference the page before unlocking and
504 * sleeping so that the page daemon is less
505 * likely to reclaim it.
507 vm_page_lock_queues();
508 vm_page_flag_set(m, PG_REFERENCED);
509 vm_page_sleep(m, "zfsmrb");
513 VM_OBJECT_UNLOCK(obj);
515 error = dmu_read_uio(os, zp->z_id, uio,
521 sf = sf_buf_alloc(m, SFB_CPUPRIVATE);
522 va = (caddr_t)sf_buf_kva(sf);
523 error = dmu_read(os, zp->z_id, start + off,
524 bytes, (void *)(va + off),
531 vm_page_set_valid(m, off, bytes);
534 uio->uio_resid -= bytes;
535 uio->uio_offset += bytes;
545 VM_OBJECT_UNLOCK(obj);
546 if (error == 0 && dirbytes > 0)
547 error = dmu_read_uio(os, zp->z_id, uio, dirbytes);
551 offset_t zfs_read_chunk_size = 1024 * 1024; /* Tunable */
554 * Read bytes from specified file into supplied buffer.
556 * IN: vp - vnode of file to be read from.
557 * uio - structure supplying read location, range info,
559 * ioflag - SYNC flags; used to provide FRSYNC semantics.
560 * cr - credentials of caller.
561 * ct - caller context
563 * OUT: uio - updated offset and range, buffer filled.
565 * RETURN: 0 if success
566 * error code if failure
569 * vp - atime updated if byte count > 0
573 zfs_read(vnode_t *vp, uio_t *uio, int ioflag, cred_t *cr, caller_context_t *ct)
575 znode_t *zp = VTOZ(vp);
576 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
586 if (zp->z_phys->zp_flags & ZFS_AV_QUARANTINED) {
592 * Validate file offset
594 if (uio->uio_loffset < (offset_t)0) {
600 * Fasttrack empty reads
602 if (uio->uio_resid == 0) {
608 * Check for mandatory locks
610 if (MANDMODE((mode_t)zp->z_phys->zp_mode)) {
611 if (error = chklock(vp, FREAD,
612 uio->uio_loffset, uio->uio_resid, uio->uio_fmode, ct)) {
619 * If we're in FRSYNC mode, sync out this znode before reading it.
622 zil_commit(zfsvfs->z_log, zp->z_last_itx, zp->z_id);
625 * Lock the range against changes.
627 rl = zfs_range_lock(zp, uio->uio_loffset, uio->uio_resid, RL_READER);
630 * If we are reading past end-of-file we can skip
631 * to the end; but we might still need to set atime.
633 if (uio->uio_loffset >= zp->z_phys->zp_size) {
638 ASSERT(uio->uio_loffset < zp->z_phys->zp_size);
639 n = MIN(uio->uio_resid, zp->z_phys->zp_size - uio->uio_loffset);
642 nbytes = MIN(n, zfs_read_chunk_size -
643 P2PHASE(uio->uio_loffset, zfs_read_chunk_size));
645 if (vn_has_cached_data(vp))
646 error = mappedread(vp, nbytes, uio);
648 error = dmu_read_uio(os, zp->z_id, uio, nbytes);
650 /* convert checksum errors into IO errors */
660 zfs_range_unlock(rl);
662 ZFS_ACCESSTIME_STAMP(zfsvfs, zp);
668 * Fault in the pages of the first n bytes specified by the uio structure.
669 * 1 byte in each page is touched and the uio struct is unmodified.
670 * Any error will exit this routine as this is only a best
671 * attempt to get the pages resident. This is a copy of ufs_trans_touch().
674 zfs_prefault_write(ssize_t n, struct uio *uio)
680 if (uio->uio_segflg != UIO_USERSPACE)
686 cnt = MIN(iov->iov_len, n);
688 /* empty iov entry */
694 * touch each page in this segment.
700 incr = MIN(cnt, PAGESIZE);
705 * touch the last byte in case it straddles a page.
715 * Write the bytes to a file.
717 * IN: vp - vnode of file to be written to.
718 * uio - structure supplying write location, range info,
720 * ioflag - IO_APPEND flag set if in append mode.
721 * cr - credentials of caller.
722 * ct - caller context (NFS/CIFS fem monitor only)
724 * OUT: uio - updated offset and range.
726 * RETURN: 0 if success
727 * error code if failure
730 * vp - ctime|mtime updated if byte count > 0
734 zfs_write(vnode_t *vp, uio_t *uio, int ioflag, cred_t *cr, caller_context_t *ct)
736 znode_t *zp = VTOZ(vp);
737 rlim64_t limit = MAXOFFSET_T;
738 ssize_t start_resid = uio->uio_resid;
742 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
747 int max_blksz = zfsvfs->z_max_blksz;
753 * Fasttrack empty write
759 if (limit == RLIM64_INFINITY || limit > MAXOFFSET_T)
766 * If immutable or not appending then return EPERM
768 pflags = zp->z_phys->zp_flags;
769 if ((pflags & (ZFS_IMMUTABLE | ZFS_READONLY)) ||
770 ((pflags & ZFS_APPENDONLY) && !(ioflag & FAPPEND) &&
771 (uio->uio_loffset < zp->z_phys->zp_size))) {
776 zilog = zfsvfs->z_log;
779 * Pre-fault the pages to ensure slow (eg NFS) pages
782 zfs_prefault_write(n, uio);
785 * If in append mode, set the io offset pointer to eof.
787 if (ioflag & IO_APPEND) {
789 * Range lock for a file append:
790 * The value for the start of range will be determined by
791 * zfs_range_lock() (to guarantee append semantics).
792 * If this write will cause the block size to increase,
793 * zfs_range_lock() will lock the entire file, so we must
794 * later reduce the range after we grow the block size.
796 rl = zfs_range_lock(zp, 0, n, RL_APPEND);
797 if (rl->r_len == UINT64_MAX) {
798 /* overlocked, zp_size can't change */
799 woff = uio->uio_loffset = zp->z_phys->zp_size;
801 woff = uio->uio_loffset = rl->r_off;
804 woff = uio->uio_loffset;
806 * Validate file offset
814 * If we need to grow the block size then zfs_range_lock()
815 * will lock a wider range than we request here.
816 * Later after growing the block size we reduce the range.
818 rl = zfs_range_lock(zp, woff, n, RL_WRITER);
822 zfs_range_unlock(rl);
827 if ((woff + n) > limit || woff > (limit - n))
831 * Check for mandatory locks
833 if (MANDMODE((mode_t)zp->z_phys->zp_mode) &&
834 (error = chklock(vp, FWRITE, woff, n, uio->uio_fmode, ct)) != 0) {
835 zfs_range_unlock(rl);
839 end_size = MAX(zp->z_phys->zp_size, woff + n);
842 * Write the file in reasonable size chunks. Each chunk is written
843 * in a separate transaction; this keeps the intent log records small
844 * and allows us to do more fine-grained space accounting.
848 woff = uio->uio_loffset;
851 if (zfs_usergroup_overquota(zfsvfs,
852 B_FALSE, zp->z_phys->zp_uid) ||
853 zfs_usergroup_overquota(zfsvfs,
854 B_TRUE, zp->z_phys->zp_gid)) {
856 dmu_return_arcbuf(abuf);
862 * If dmu_assign_arcbuf() is expected to execute with minimum
863 * overhead loan an arc buffer and copy user data to it before
864 * we enter a txg. This avoids holding a txg forever while we
865 * pagefault on a hanging NFS server mapping.
867 if (abuf == NULL && n >= max_blksz &&
868 woff >= zp->z_phys->zp_size &&
869 P2PHASE(woff, max_blksz) == 0 &&
870 zp->z_blksz == max_blksz) {
873 abuf = dmu_request_arcbuf(zp->z_dbuf, max_blksz);
874 ASSERT(abuf != NULL);
875 ASSERT(arc_buf_size(abuf) == max_blksz);
876 if (error = uiocopy(abuf->b_data, max_blksz,
877 UIO_WRITE, uio, &cbytes)) {
878 dmu_return_arcbuf(abuf);
881 ASSERT(cbytes == max_blksz);
885 * Start a transaction.
887 tx = dmu_tx_create(zfsvfs->z_os);
888 dmu_tx_hold_bonus(tx, zp->z_id);
889 dmu_tx_hold_write(tx, zp->z_id, woff, MIN(n, max_blksz));
890 error = dmu_tx_assign(tx, TXG_NOWAIT);
892 if (error == ERESTART) {
899 dmu_return_arcbuf(abuf);
904 * If zfs_range_lock() over-locked we grow the blocksize
905 * and then reduce the lock range. This will only happen
906 * on the first iteration since zfs_range_reduce() will
907 * shrink down r_len to the appropriate size.
909 if (rl->r_len == UINT64_MAX) {
912 if (zp->z_blksz > max_blksz) {
913 ASSERT(!ISP2(zp->z_blksz));
914 new_blksz = MIN(end_size, SPA_MAXBLOCKSIZE);
916 new_blksz = MIN(end_size, max_blksz);
918 zfs_grow_blocksize(zp, new_blksz, tx);
919 zfs_range_reduce(rl, woff, n);
923 * XXX - should we really limit each write to z_max_blksz?
924 * Perhaps we should use SPA_MAXBLOCKSIZE chunks?
926 nbytes = MIN(n, max_blksz - P2PHASE(woff, max_blksz));
928 if (woff + nbytes > zp->z_phys->zp_size)
929 vnode_pager_setsize(vp, woff + nbytes);
932 tx_bytes = uio->uio_resid;
933 error = dmu_write_uio(zfsvfs->z_os, zp->z_id, uio,
935 tx_bytes -= uio->uio_resid;
938 ASSERT(tx_bytes == max_blksz);
939 dmu_assign_arcbuf(zp->z_dbuf, woff, abuf, tx);
940 ASSERT(tx_bytes <= uio->uio_resid);
941 uioskip(uio, tx_bytes);
945 * XXXPJD: There are some cases (triggered by fsx) where
946 * vn_has_cached_data(vp) returns false when it should
947 * return true. This should be investigated.
950 if (tx_bytes && vn_has_cached_data(vp))
952 if (tx_bytes && vp->v_object != NULL)
955 update_pages(vp, woff, tx_bytes, zfsvfs->z_os,
956 zp->z_id, uio->uio_segflg, tx);
960 * If we made no progress, we're done. If we made even
961 * partial progress, update the znode and ZIL accordingly.
970 * Clear Set-UID/Set-GID bits on successful write if not
971 * privileged and at least one of the excute bits is set.
973 * It would be nice to to this after all writes have
974 * been done, but that would still expose the ISUID/ISGID
975 * to another app after the partial write is committed.
977 * Note: we don't call zfs_fuid_map_id() here because
978 * user 0 is not an ephemeral uid.
980 mutex_enter(&zp->z_acl_lock);
981 if ((zp->z_phys->zp_mode & (S_IXUSR | (S_IXUSR >> 3) |
982 (S_IXUSR >> 6))) != 0 &&
983 (zp->z_phys->zp_mode & (S_ISUID | S_ISGID)) != 0 &&
984 secpolicy_vnode_setid_retain(vp, cr,
985 (zp->z_phys->zp_mode & S_ISUID) != 0 &&
986 zp->z_phys->zp_uid == 0) != 0) {
987 zp->z_phys->zp_mode &= ~(S_ISUID | S_ISGID);
989 mutex_exit(&zp->z_acl_lock);
992 * Update time stamp. NOTE: This marks the bonus buffer as
993 * dirty, so we don't have to do it again for zp_size.
995 zfs_time_stamper(zp, CONTENT_MODIFIED, tx);
998 * Update the file size (zp_size) if it has changed;
999 * account for possible concurrent updates.
1001 while ((end_size = zp->z_phys->zp_size) < uio->uio_loffset)
1002 (void) atomic_cas_64(&zp->z_phys->zp_size, end_size,
1004 zfs_log_write(zilog, tx, TX_WRITE, zp, woff, tx_bytes, ioflag);
1009 ASSERT(tx_bytes == nbytes);
1013 zfs_range_unlock(rl);
1016 * If we're in replay mode, or we made no progress, return error.
1017 * Otherwise, it's at least a partial write, so it's successful.
1019 if (zfsvfs->z_replay || uio->uio_resid == start_resid) {
1024 if (ioflag & (FSYNC | FDSYNC))
1025 zil_commit(zilog, zp->z_last_itx, zp->z_id);
1032 zfs_get_done(dmu_buf_t *db, void *vzgd)
1034 zgd_t *zgd = (zgd_t *)vzgd;
1035 rl_t *rl = zgd->zgd_rl;
1036 vnode_t *vp = ZTOV(rl->r_zp);
1037 objset_t *os = rl->r_zp->z_zfsvfs->z_os;
1040 vfslocked = VFS_LOCK_GIANT(vp->v_vfsp);
1041 dmu_buf_rele(db, vzgd);
1042 zfs_range_unlock(rl);
1044 * Release the vnode asynchronously as we currently have the
1045 * txg stopped from syncing.
1047 VN_RELE_ASYNC(vp, dsl_pool_vnrele_taskq(dmu_objset_pool(os)));
1048 zil_add_block(zgd->zgd_zilog, zgd->zgd_bp);
1049 kmem_free(zgd, sizeof (zgd_t));
1050 VFS_UNLOCK_GIANT(vfslocked);
1054 * Get data to generate a TX_WRITE intent log record.
1057 zfs_get_data(void *arg, lr_write_t *lr, char *buf, zio_t *zio)
1059 zfsvfs_t *zfsvfs = arg;
1060 objset_t *os = zfsvfs->z_os;
1062 uint64_t off = lr->lr_offset;
1066 int dlen = lr->lr_length; /* length of user data */
1073 * Nothing to do if the file has been removed
1075 if (zfs_zget(zfsvfs, lr->lr_foid, &zp) != 0)
1077 if (zp->z_unlinked) {
1079 * Release the vnode asynchronously as we currently have the
1080 * txg stopped from syncing.
1082 VN_RELE_ASYNC(ZTOV(zp),
1083 dsl_pool_vnrele_taskq(dmu_objset_pool(os)));
1088 * Write records come in two flavors: immediate and indirect.
1089 * For small writes it's cheaper to store the data with the
1090 * log record (immediate); for large writes it's cheaper to
1091 * sync the data and get a pointer to it (indirect) so that
1092 * we don't have to write the data twice.
1094 if (buf != NULL) { /* immediate write */
1095 rl = zfs_range_lock(zp, off, dlen, RL_READER);
1096 /* test for truncation needs to be done while range locked */
1097 if (off >= zp->z_phys->zp_size) {
1101 VERIFY(0 == dmu_read(os, lr->lr_foid, off, dlen, buf,
1102 DMU_READ_NO_PREFETCH));
1103 } else { /* indirect write */
1104 uint64_t boff; /* block starting offset */
1107 * Have to lock the whole block to ensure when it's
1108 * written out and it's checksum is being calculated
1109 * that no one can change the data. We need to re-check
1110 * blocksize after we get the lock in case it's changed!
1113 if (ISP2(zp->z_blksz)) {
1114 boff = P2ALIGN_TYPED(off, zp->z_blksz,
1120 rl = zfs_range_lock(zp, boff, dlen, RL_READER);
1121 if (zp->z_blksz == dlen)
1123 zfs_range_unlock(rl);
1125 /* test for truncation needs to be done while range locked */
1126 if (off >= zp->z_phys->zp_size) {
1130 zgd = (zgd_t *)kmem_alloc(sizeof (zgd_t), KM_SLEEP);
1132 zgd->zgd_zilog = zfsvfs->z_log;
1133 zgd->zgd_bp = &lr->lr_blkptr;
1134 VERIFY(0 == dmu_buf_hold(os, lr->lr_foid, boff, zgd, &db));
1135 ASSERT(boff == db->db_offset);
1136 lr->lr_blkoff = off - boff;
1137 error = dmu_sync(zio, db, &lr->lr_blkptr,
1138 lr->lr_common.lrc_txg, zfs_get_done, zgd);
1139 ASSERT((error && error != EINPROGRESS) ||
1140 lr->lr_length <= zp->z_blksz);
1143 * dmu_sync() can compress a block of zeros to a null
1144 * blkptr but the block size still needs to be passed
1145 * through to replay.
1147 BP_SET_LSIZE(&lr->lr_blkptr, db->db_size);
1148 zil_add_block(zfsvfs->z_log, &lr->lr_blkptr);
1152 * If we get EINPROGRESS, then we need to wait for a
1153 * write IO initiated by dmu_sync() to complete before
1154 * we can release this dbuf. We will finish everything
1155 * up in the zfs_get_done() callback.
1157 if (error == EINPROGRESS) {
1159 } else if (error == EALREADY) {
1160 lr->lr_common.lrc_txtype = TX_WRITE2;
1163 dmu_buf_rele(db, zgd);
1164 kmem_free(zgd, sizeof (zgd_t));
1167 zfs_range_unlock(rl);
1169 * Release the vnode asynchronously as we currently have the
1170 * txg stopped from syncing.
1172 VN_RELE_ASYNC(ZTOV(zp), dsl_pool_vnrele_taskq(dmu_objset_pool(os)));
1178 zfs_access(vnode_t *vp, int mode, int flag, cred_t *cr,
1179 caller_context_t *ct)
1181 znode_t *zp = VTOZ(vp);
1182 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1188 if (flag & V_ACE_MASK)
1189 error = zfs_zaccess(zp, mode, flag, B_FALSE, cr);
1191 error = zfs_zaccess_rwx(zp, mode, flag, cr);
1198 * If vnode is for a device return a specfs vnode instead.
1201 specvp_check(vnode_t **vpp, cred_t *cr)
1205 if (IS_DEVVP(*vpp)) {
1208 svp = specvp(*vpp, (*vpp)->v_rdev, (*vpp)->v_type, cr);
1219 * Lookup an entry in a directory, or an extended attribute directory.
1220 * If it exists, return a held vnode reference for it.
1222 * IN: dvp - vnode of directory to search.
1223 * nm - name of entry to lookup.
1224 * pnp - full pathname to lookup [UNUSED].
1225 * flags - LOOKUP_XATTR set if looking for an attribute.
1226 * rdir - root directory vnode [UNUSED].
1227 * cr - credentials of caller.
1228 * ct - caller context
1229 * direntflags - directory lookup flags
1230 * realpnp - returned pathname.
1232 * OUT: vpp - vnode of located entry, NULL if not found.
1234 * RETURN: 0 if success
1235 * error code if failure
1242 zfs_lookup(vnode_t *dvp, char *nm, vnode_t **vpp, struct componentname *cnp,
1243 int nameiop, cred_t *cr, kthread_t *td, int flags)
1245 znode_t *zdp = VTOZ(dvp);
1246 zfsvfs_t *zfsvfs = zdp->z_zfsvfs;
1248 int *direntflags = NULL;
1249 void *realpnp = NULL;
1252 if (!(flags & (LOOKUP_XATTR | FIGNORECASE))) {
1254 if (dvp->v_type != VDIR) {
1256 } else if (zdp->z_dbuf == NULL) {
1260 if (nm[0] == 0 || (nm[0] == '.' && nm[1] == '\0')) {
1261 error = zfs_fastaccesschk_execute(zdp, cr);
1269 vnode_t *tvp = dnlc_lookup(dvp, nm);
1272 error = zfs_fastaccesschk_execute(zdp, cr);
1277 if (tvp == DNLC_NO_VNODE) {
1282 return (specvp_check(vpp, cr));
1288 DTRACE_PROBE2(zfs__fastpath__lookup__miss, vnode_t *, dvp, char *, nm);
1295 if (flags & LOOKUP_XATTR) {
1298 * If the xattr property is off, refuse the lookup request.
1300 if (!(zfsvfs->z_vfs->vfs_flag & VFS_XATTR)) {
1307 * We don't allow recursive attributes..
1308 * Maybe someday we will.
1310 if (zdp->z_phys->zp_flags & ZFS_XATTR) {
1315 if (error = zfs_get_xattrdir(VTOZ(dvp), vpp, cr, flags)) {
1321 * Do we have permission to get into attribute directory?
1324 if (error = zfs_zaccess(VTOZ(*vpp), ACE_EXECUTE, 0,
1334 if (dvp->v_type != VDIR) {
1340 * Check accessibility of directory.
1343 if (error = zfs_zaccess(zdp, ACE_EXECUTE, 0, B_FALSE, cr)) {
1348 if (zfsvfs->z_utf8 && u8_validate(nm, strlen(nm),
1349 NULL, U8_VALIDATE_ENTIRE, &error) < 0) {
1354 error = zfs_dirlook(zdp, nm, vpp, flags, direntflags, realpnp);
1356 error = specvp_check(vpp, cr);
1358 /* Translate errors and add SAVENAME when needed. */
1359 if (cnp->cn_flags & ISLASTCN) {
1363 if (error == ENOENT) {
1364 error = EJUSTRETURN;
1365 cnp->cn_flags |= SAVENAME;
1371 cnp->cn_flags |= SAVENAME;
1375 if (error == 0 && (nm[0] != '.' || nm[1] != '\0')) {
1378 if (cnp->cn_flags & ISDOTDOT) {
1379 ltype = VOP_ISLOCKED(dvp);
1383 error = vn_lock(*vpp, cnp->cn_lkflags);
1384 if (cnp->cn_flags & ISDOTDOT)
1385 vn_lock(dvp, ltype | LK_RETRY);
1395 #ifdef FREEBSD_NAMECACHE
1397 * Insert name into cache (as non-existent) if appropriate.
1399 if (error == ENOENT && (cnp->cn_flags & MAKEENTRY) && nameiop != CREATE)
1400 cache_enter(dvp, *vpp, cnp);
1402 * Insert name into cache if appropriate.
1404 if (error == 0 && (cnp->cn_flags & MAKEENTRY)) {
1405 if (!(cnp->cn_flags & ISLASTCN) ||
1406 (nameiop != DELETE && nameiop != RENAME)) {
1407 cache_enter(dvp, *vpp, cnp);
1416 * Attempt to create a new entry in a directory. If the entry
1417 * already exists, truncate the file if permissible, else return
1418 * an error. Return the vp of the created or trunc'd file.
1420 * IN: dvp - vnode of directory to put new file entry in.
1421 * name - name of new file entry.
1422 * vap - attributes of new file.
1423 * excl - flag indicating exclusive or non-exclusive mode.
1424 * mode - mode to open file with.
1425 * cr - credentials of caller.
1426 * flag - large file flag [UNUSED].
1427 * ct - caller context
1428 * vsecp - ACL to be set
1430 * OUT: vpp - vnode of created or trunc'd entry.
1432 * RETURN: 0 if success
1433 * error code if failure
1436 * dvp - ctime|mtime updated if new entry created
1437 * vp - ctime|mtime always, atime if new
1442 zfs_create(vnode_t *dvp, char *name, vattr_t *vap, int excl, int mode,
1443 vnode_t **vpp, cred_t *cr, kthread_t *td)
1445 znode_t *zp, *dzp = VTOZ(dvp);
1446 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
1454 gid_t gid = crgetgid(cr);
1455 zfs_acl_ids_t acl_ids;
1456 boolean_t fuid_dirtied;
1461 * If we have an ephemeral id, ACL, or XVATTR then
1462 * make sure file system is at proper version
1465 ksid = crgetsid(cr, KSID_OWNER);
1467 uid = ksid_getid(ksid);
1470 if (zfsvfs->z_use_fuids == B_FALSE &&
1471 (vsecp || (vap->va_mask & AT_XVATTR) ||
1472 IS_EPHEMERAL(crgetuid(cr)) || IS_EPHEMERAL(crgetgid(cr))))
1478 zilog = zfsvfs->z_log;
1480 if (zfsvfs->z_utf8 && u8_validate(name, strlen(name),
1481 NULL, U8_VALIDATE_ENTIRE, &error) < 0) {
1486 if (vap->va_mask & AT_XVATTR) {
1487 if ((error = secpolicy_xvattr(dvp, (xvattr_t *)vap,
1488 crgetuid(cr), cr, vap->va_type)) != 0) {
1496 if ((vap->va_mode & S_ISVTX) && secpolicy_vnode_stky_modify(cr))
1497 vap->va_mode &= ~S_ISVTX;
1499 if (*name == '\0') {
1501 * Null component name refers to the directory itself.
1508 /* possible VN_HOLD(zp) */
1511 if (flag & FIGNORECASE)
1514 error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg,
1517 if (strcmp(name, "..") == 0)
1527 * Create a new file object and update the directory
1530 if (error = zfs_zaccess(dzp, ACE_ADD_FILE, 0, B_FALSE, cr)) {
1535 * We only support the creation of regular files in
1536 * extended attribute directories.
1538 if ((dzp->z_phys->zp_flags & ZFS_XATTR) &&
1539 (vap->va_type != VREG)) {
1545 if ((error = zfs_acl_ids_create(dzp, 0, vap, cr, vsecp,
1548 if (zfs_acl_ids_overquota(zfsvfs, &acl_ids)) {
1549 zfs_acl_ids_free(&acl_ids);
1554 tx = dmu_tx_create(os);
1555 dmu_tx_hold_bonus(tx, DMU_NEW_OBJECT);
1556 fuid_dirtied = zfsvfs->z_fuid_dirty;
1558 zfs_fuid_txhold(zfsvfs, tx);
1559 dmu_tx_hold_bonus(tx, dzp->z_id);
1560 dmu_tx_hold_zap(tx, dzp->z_id, TRUE, name);
1561 if (acl_ids.z_aclp->z_acl_bytes > ZFS_ACE_SPACE) {
1562 dmu_tx_hold_write(tx, DMU_NEW_OBJECT,
1563 0, SPA_MAXBLOCKSIZE);
1565 error = dmu_tx_assign(tx, TXG_NOWAIT);
1567 zfs_acl_ids_free(&acl_ids);
1568 zfs_dirent_unlock(dl);
1569 if (error == ERESTART) {
1578 zfs_mknode(dzp, vap, tx, cr, 0, &zp, 0, &acl_ids);
1581 zfs_fuid_sync(zfsvfs, tx);
1583 (void) zfs_link_create(dl, zp, tx, ZNEW);
1585 txtype = zfs_log_create_txtype(Z_FILE, vsecp, vap);
1586 if (flag & FIGNORECASE)
1588 zfs_log_create(zilog, tx, txtype, dzp, zp, name,
1589 vsecp, acl_ids.z_fuidp, vap);
1590 zfs_acl_ids_free(&acl_ids);
1593 int aflags = (flag & FAPPEND) ? V_APPEND : 0;
1596 * A directory entry already exists for this name.
1599 * Can't truncate an existing file if in exclusive mode.
1606 * Can't open a directory for writing.
1608 if ((ZTOV(zp)->v_type == VDIR) && (mode & S_IWRITE)) {
1613 * Verify requested access to file.
1615 if (mode && (error = zfs_zaccess_rwx(zp, mode, aflags, cr))) {
1619 mutex_enter(&dzp->z_lock);
1621 mutex_exit(&dzp->z_lock);
1624 * Truncate regular files if requested.
1626 if ((ZTOV(zp)->v_type == VREG) &&
1627 (vap->va_mask & AT_SIZE) && (vap->va_size == 0)) {
1628 /* we can't hold any locks when calling zfs_freesp() */
1629 zfs_dirent_unlock(dl);
1631 error = zfs_freesp(zp, 0, 0, mode, TRUE);
1633 vnevent_create(ZTOV(zp), ct);
1639 zfs_dirent_unlock(dl);
1646 error = specvp_check(vpp, cr);
1654 * Remove an entry from a directory.
1656 * IN: dvp - vnode of directory to remove entry from.
1657 * name - name of entry to remove.
1658 * cr - credentials of caller.
1659 * ct - caller context
1660 * flags - case flags
1662 * RETURN: 0 if success
1663 * error code if failure
1667 * vp - ctime (if nlink > 0)
1671 zfs_remove(vnode_t *dvp, char *name, cred_t *cr, caller_context_t *ct,
1674 znode_t *zp, *dzp = VTOZ(dvp);
1675 znode_t *xzp = NULL;
1677 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
1679 uint64_t acl_obj, xattr_obj;
1682 boolean_t may_delete_now, delete_now = FALSE;
1683 boolean_t unlinked, toobig = FALSE;
1685 pathname_t *realnmp = NULL;
1692 zilog = zfsvfs->z_log;
1694 if (flags & FIGNORECASE) {
1702 * Attempt to lock directory; fail if entry doesn't exist.
1704 if (error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg,
1714 if (error = zfs_zaccess_delete(dzp, zp, cr)) {
1719 * Need to use rmdir for removing directories.
1721 if (vp->v_type == VDIR) {
1726 vnevent_remove(vp, dvp, name, ct);
1729 dnlc_remove(dvp, realnmp->pn_buf);
1731 dnlc_remove(dvp, name);
1733 may_delete_now = FALSE;
1736 * We may delete the znode now, or we may put it in the unlinked set;
1737 * it depends on whether we're the last link, and on whether there are
1738 * other holds on the vnode. So we dmu_tx_hold() the right things to
1739 * allow for either case.
1741 tx = dmu_tx_create(zfsvfs->z_os);
1742 dmu_tx_hold_zap(tx, dzp->z_id, FALSE, name);
1743 dmu_tx_hold_bonus(tx, zp->z_id);
1744 if (may_delete_now) {
1746 zp->z_phys->zp_size > zp->z_blksz * DMU_MAX_DELETEBLKCNT;
1747 /* if the file is too big, only hold_free a token amount */
1748 dmu_tx_hold_free(tx, zp->z_id, 0,
1749 (toobig ? DMU_MAX_ACCESS : DMU_OBJECT_END));
1752 /* are there any extended attributes? */
1753 if ((xattr_obj = zp->z_phys->zp_xattr) != 0) {
1754 /* XXX - do we need this if we are deleting? */
1755 dmu_tx_hold_bonus(tx, xattr_obj);
1758 /* are there any additional acls */
1759 if ((acl_obj = zp->z_phys->zp_acl.z_acl_extern_obj) != 0 &&
1761 dmu_tx_hold_free(tx, acl_obj, 0, DMU_OBJECT_END);
1763 /* charge as an update -- would be nice not to charge at all */
1764 dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL);
1766 error = dmu_tx_assign(tx, TXG_NOWAIT);
1768 zfs_dirent_unlock(dl);
1770 if (error == ERESTART) {
1783 * Remove the directory entry.
1785 error = zfs_link_destroy(dl, zp, tx, zflg, &unlinked);
1792 if (0 && unlinked) {
1794 delete_now = may_delete_now && !toobig &&
1795 vp->v_count == 1 && !vn_has_cached_data(vp) &&
1796 zp->z_phys->zp_xattr == xattr_obj &&
1797 zp->z_phys->zp_acl.z_acl_extern_obj == acl_obj;
1802 if (zp->z_phys->zp_xattr) {
1803 error = zfs_zget(zfsvfs, zp->z_phys->zp_xattr, &xzp);
1804 ASSERT3U(error, ==, 0);
1805 ASSERT3U(xzp->z_phys->zp_links, ==, 2);
1806 dmu_buf_will_dirty(xzp->z_dbuf, tx);
1807 mutex_enter(&xzp->z_lock);
1808 xzp->z_unlinked = 1;
1809 xzp->z_phys->zp_links = 0;
1810 mutex_exit(&xzp->z_lock);
1811 zfs_unlinked_add(xzp, tx);
1812 zp->z_phys->zp_xattr = 0; /* probably unnecessary */
1814 mutex_enter(&zp->z_lock);
1817 ASSERT3U(vp->v_count, ==, 0);
1819 mutex_exit(&zp->z_lock);
1820 zfs_znode_delete(zp, tx);
1821 } else if (unlinked) {
1822 zfs_unlinked_add(zp, tx);
1826 if (flags & FIGNORECASE)
1828 zfs_log_remove(zilog, tx, txtype, dzp, name);
1835 zfs_dirent_unlock(dl);
1840 /* this rele is delayed to prevent nesting transactions */
1849 * Create a new directory and insert it into dvp using the name
1850 * provided. Return a pointer to the inserted directory.
1852 * IN: dvp - vnode of directory to add subdir to.
1853 * dirname - name of new directory.
1854 * vap - attributes of new directory.
1855 * cr - credentials of caller.
1856 * ct - caller context
1857 * vsecp - ACL to be set
1859 * OUT: vpp - vnode of created directory.
1861 * RETURN: 0 if success
1862 * error code if failure
1865 * dvp - ctime|mtime updated
1866 * vp - ctime|mtime|atime updated
1870 zfs_mkdir(vnode_t *dvp, char *dirname, vattr_t *vap, vnode_t **vpp, cred_t *cr,
1871 caller_context_t *ct, int flags, vsecattr_t *vsecp)
1873 znode_t *zp, *dzp = VTOZ(dvp);
1874 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
1883 gid_t gid = crgetgid(cr);
1884 zfs_acl_ids_t acl_ids;
1885 boolean_t fuid_dirtied;
1887 ASSERT(vap->va_type == VDIR);
1890 * If we have an ephemeral id, ACL, or XVATTR then
1891 * make sure file system is at proper version
1894 ksid = crgetsid(cr, KSID_OWNER);
1896 uid = ksid_getid(ksid);
1899 if (zfsvfs->z_use_fuids == B_FALSE &&
1900 (vsecp || (vap->va_mask & AT_XVATTR) || IS_EPHEMERAL(crgetuid(cr))||
1901 IS_EPHEMERAL(crgetgid(cr))))
1906 zilog = zfsvfs->z_log;
1908 if (dzp->z_phys->zp_flags & ZFS_XATTR) {
1913 if (zfsvfs->z_utf8 && u8_validate(dirname,
1914 strlen(dirname), NULL, U8_VALIDATE_ENTIRE, &error) < 0) {
1918 if (flags & FIGNORECASE)
1921 if (vap->va_mask & AT_XVATTR)
1922 if ((error = secpolicy_xvattr(dvp, (xvattr_t *)vap,
1923 crgetuid(cr), cr, vap->va_type)) != 0) {
1929 * First make sure the new directory doesn't exist.
1934 if (error = zfs_dirent_lock(&dl, dzp, dirname, &zp, zf,
1940 if (error = zfs_zaccess(dzp, ACE_ADD_SUBDIRECTORY, 0, B_FALSE, cr)) {
1941 zfs_dirent_unlock(dl);
1946 if ((error = zfs_acl_ids_create(dzp, 0, vap, cr, vsecp,
1948 zfs_dirent_unlock(dl);
1952 if (zfs_acl_ids_overquota(zfsvfs, &acl_ids)) {
1953 zfs_acl_ids_free(&acl_ids);
1954 zfs_dirent_unlock(dl);
1960 * Add a new entry to the directory.
1962 tx = dmu_tx_create(zfsvfs->z_os);
1963 dmu_tx_hold_zap(tx, dzp->z_id, TRUE, dirname);
1964 dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL);
1965 fuid_dirtied = zfsvfs->z_fuid_dirty;
1967 zfs_fuid_txhold(zfsvfs, tx);
1968 if (acl_ids.z_aclp->z_acl_bytes > ZFS_ACE_SPACE)
1969 dmu_tx_hold_write(tx, DMU_NEW_OBJECT,
1970 0, SPA_MAXBLOCKSIZE);
1971 error = dmu_tx_assign(tx, TXG_NOWAIT);
1973 zfs_acl_ids_free(&acl_ids);
1974 zfs_dirent_unlock(dl);
1975 if (error == ERESTART) {
1988 zfs_mknode(dzp, vap, tx, cr, 0, &zp, 0, &acl_ids);
1991 zfs_fuid_sync(zfsvfs, tx);
1993 * Now put new name in parent dir.
1995 (void) zfs_link_create(dl, zp, tx, ZNEW);
1999 txtype = zfs_log_create_txtype(Z_DIR, vsecp, vap);
2000 if (flags & FIGNORECASE)
2002 zfs_log_create(zilog, tx, txtype, dzp, zp, dirname, vsecp,
2003 acl_ids.z_fuidp, vap);
2005 zfs_acl_ids_free(&acl_ids);
2008 zfs_dirent_unlock(dl);
2015 * Remove a directory subdir entry. If the current working
2016 * directory is the same as the subdir to be removed, the
2019 * IN: dvp - vnode of directory to remove from.
2020 * name - name of directory to be removed.
2021 * cwd - vnode of current working directory.
2022 * cr - credentials of caller.
2023 * ct - caller context
2024 * flags - case flags
2026 * RETURN: 0 if success
2027 * error code if failure
2030 * dvp - ctime|mtime updated
2034 zfs_rmdir(vnode_t *dvp, char *name, vnode_t *cwd, cred_t *cr,
2035 caller_context_t *ct, int flags)
2037 znode_t *dzp = VTOZ(dvp);
2040 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
2049 zilog = zfsvfs->z_log;
2051 if (flags & FIGNORECASE)
2057 * Attempt to lock directory; fail if entry doesn't exist.
2059 if (error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg,
2067 if (error = zfs_zaccess_delete(dzp, zp, cr)) {
2071 if (vp->v_type != VDIR) {
2081 vnevent_rmdir(vp, dvp, name, ct);
2084 * Grab a lock on the directory to make sure that noone is
2085 * trying to add (or lookup) entries while we are removing it.
2087 rw_enter(&zp->z_name_lock, RW_WRITER);
2090 * Grab a lock on the parent pointer to make sure we play well
2091 * with the treewalk and directory rename code.
2093 rw_enter(&zp->z_parent_lock, RW_WRITER);
2095 tx = dmu_tx_create(zfsvfs->z_os);
2096 dmu_tx_hold_zap(tx, dzp->z_id, FALSE, name);
2097 dmu_tx_hold_bonus(tx, zp->z_id);
2098 dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL);
2099 error = dmu_tx_assign(tx, TXG_NOWAIT);
2101 rw_exit(&zp->z_parent_lock);
2102 rw_exit(&zp->z_name_lock);
2103 zfs_dirent_unlock(dl);
2105 if (error == ERESTART) {
2115 #ifdef FREEBSD_NAMECACHE
2119 error = zfs_link_destroy(dl, zp, tx, zflg, NULL);
2122 uint64_t txtype = TX_RMDIR;
2123 if (flags & FIGNORECASE)
2125 zfs_log_remove(zilog, tx, txtype, dzp, name);
2130 rw_exit(&zp->z_parent_lock);
2131 rw_exit(&zp->z_name_lock);
2132 #ifdef FREEBSD_NAMECACHE
2136 zfs_dirent_unlock(dl);
2145 * Read as many directory entries as will fit into the provided
2146 * buffer from the given directory cursor position (specified in
2147 * the uio structure.
2149 * IN: vp - vnode of directory to read.
2150 * uio - structure supplying read location, range info,
2151 * and return buffer.
2152 * cr - credentials of caller.
2153 * ct - caller context
2154 * flags - case flags
2156 * OUT: uio - updated offset and range, buffer filled.
2157 * eofp - set to true if end-of-file detected.
2159 * RETURN: 0 if success
2160 * error code if failure
2163 * vp - atime updated
2165 * Note that the low 4 bits of the cookie returned by zap is always zero.
2166 * This allows us to use the low range for "special" directory entries:
2167 * We use 0 for '.', and 1 for '..'. If this is the root of the filesystem,
2168 * we use the offset 2 for the '.zfs' directory.
2172 zfs_readdir(vnode_t *vp, uio_t *uio, cred_t *cr, int *eofp, int *ncookies, u_long **cookies)
2174 znode_t *zp = VTOZ(vp);
2178 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
2183 zap_attribute_t zap;
2184 uint_t bytes_wanted;
2185 uint64_t offset; /* must be unsigned; checks for < 1 */
2190 boolean_t check_sysattrs;
2193 u_long *cooks = NULL;
2200 * If we are not given an eof variable,
2207 * Check for valid iov_len.
2209 if (uio->uio_iov->iov_len <= 0) {
2215 * Quit if directory has been removed (posix)
2217 if ((*eofp = zp->z_unlinked) != 0) {
2224 offset = uio->uio_loffset;
2225 prefetch = zp->z_zn_prefetch;
2228 * Initialize the iterator cursor.
2232 * Start iteration from the beginning of the directory.
2234 zap_cursor_init(&zc, os, zp->z_id);
2237 * The offset is a serialized cursor.
2239 zap_cursor_init_serialized(&zc, os, zp->z_id, offset);
2243 * Get space to change directory entries into fs independent format.
2245 iovp = uio->uio_iov;
2246 bytes_wanted = iovp->iov_len;
2247 if (uio->uio_segflg != UIO_SYSSPACE || uio->uio_iovcnt != 1) {
2248 bufsize = bytes_wanted;
2249 outbuf = kmem_alloc(bufsize, KM_SLEEP);
2250 odp = (struct dirent64 *)outbuf;
2252 bufsize = bytes_wanted;
2253 odp = (struct dirent64 *)iovp->iov_base;
2255 eodp = (struct edirent *)odp;
2257 if (ncookies != NULL) {
2259 * Minimum entry size is dirent size and 1 byte for a file name.
2261 ncooks = uio->uio_resid / (sizeof(struct dirent) - sizeof(((struct dirent *)NULL)->d_name) + 1);
2262 cooks = malloc(ncooks * sizeof(u_long), M_TEMP, M_WAITOK);
2267 * If this VFS supports the system attribute view interface; and
2268 * we're looking at an extended attribute directory; and we care
2269 * about normalization conflicts on this vfs; then we must check
2270 * for normalization conflicts with the sysattr name space.
2273 check_sysattrs = vfs_has_feature(vp->v_vfsp, VFSFT_SYSATTR_VIEWS) &&
2274 (vp->v_flag & V_XATTRDIR) && zfsvfs->z_norm &&
2275 (flags & V_RDDIR_ENTFLAGS);
2281 * Transform to file-system independent format
2284 while (outcount < bytes_wanted) {
2290 * Special case `.', `..', and `.zfs'.
2293 (void) strcpy(zap.za_name, ".");
2294 zap.za_normalization_conflict = 0;
2297 } else if (offset == 1) {
2298 (void) strcpy(zap.za_name, "..");
2299 zap.za_normalization_conflict = 0;
2300 objnum = zp->z_phys->zp_parent;
2302 } else if (offset == 2 && zfs_show_ctldir(zp)) {
2303 (void) strcpy(zap.za_name, ZFS_CTLDIR_NAME);
2304 zap.za_normalization_conflict = 0;
2305 objnum = ZFSCTL_INO_ROOT;
2311 if (error = zap_cursor_retrieve(&zc, &zap)) {
2312 if ((*eofp = (error == ENOENT)) != 0)
2318 if (zap.za_integer_length != 8 ||
2319 zap.za_num_integers != 1) {
2320 cmn_err(CE_WARN, "zap_readdir: bad directory "
2321 "entry, obj = %lld, offset = %lld\n",
2322 (u_longlong_t)zp->z_id,
2323 (u_longlong_t)offset);
2328 objnum = ZFS_DIRENT_OBJ(zap.za_first_integer);
2330 * MacOS X can extract the object type here such as:
2331 * uint8_t type = ZFS_DIRENT_TYPE(zap.za_first_integer);
2333 type = ZFS_DIRENT_TYPE(zap.za_first_integer);
2335 if (check_sysattrs && !zap.za_normalization_conflict) {
2337 zap.za_normalization_conflict =
2338 xattr_sysattr_casechk(zap.za_name);
2340 panic("%s:%u: TODO", __func__, __LINE__);
2345 if (flags & V_RDDIR_ACCFILTER) {
2347 * If we have no access at all, don't include
2348 * this entry in the returned information
2351 if (zfs_zget(zp->z_zfsvfs, objnum, &ezp) != 0)
2353 if (!zfs_has_access(ezp, cr)) {
2360 if (flags & V_RDDIR_ENTFLAGS)
2361 reclen = EDIRENT_RECLEN(strlen(zap.za_name));
2363 reclen = DIRENT64_RECLEN(strlen(zap.za_name));
2366 * Will this entry fit in the buffer?
2368 if (outcount + reclen > bufsize) {
2370 * Did we manage to fit anything in the buffer?
2378 if (flags & V_RDDIR_ENTFLAGS) {
2380 * Add extended flag entry:
2382 eodp->ed_ino = objnum;
2383 eodp->ed_reclen = reclen;
2384 /* NOTE: ed_off is the offset for the *next* entry */
2385 next = &(eodp->ed_off);
2386 eodp->ed_eflags = zap.za_normalization_conflict ?
2387 ED_CASE_CONFLICT : 0;
2388 (void) strncpy(eodp->ed_name, zap.za_name,
2389 EDIRENT_NAMELEN(reclen));
2390 eodp = (edirent_t *)((intptr_t)eodp + reclen);
2395 odp->d_ino = objnum;
2396 odp->d_reclen = reclen;
2397 odp->d_namlen = strlen(zap.za_name);
2398 (void) strlcpy(odp->d_name, zap.za_name, odp->d_namlen + 1);
2400 odp = (dirent64_t *)((intptr_t)odp + reclen);
2404 ASSERT(outcount <= bufsize);
2406 /* Prefetch znode */
2408 dmu_prefetch(os, objnum, 0, 0);
2412 * Move to the next entry, fill in the previous offset.
2414 if (offset > 2 || (offset == 2 && !zfs_show_ctldir(zp))) {
2415 zap_cursor_advance(&zc);
2416 offset = zap_cursor_serialize(&zc);
2421 if (cooks != NULL) {
2424 KASSERT(ncooks >= 0, ("ncookies=%d", ncooks));
2427 zp->z_zn_prefetch = B_FALSE; /* a lookup will re-enable pre-fetching */
2429 /* Subtract unused cookies */
2430 if (ncookies != NULL)
2431 *ncookies -= ncooks;
2433 if (uio->uio_segflg == UIO_SYSSPACE && uio->uio_iovcnt == 1) {
2434 iovp->iov_base += outcount;
2435 iovp->iov_len -= outcount;
2436 uio->uio_resid -= outcount;
2437 } else if (error = uiomove(outbuf, (long)outcount, UIO_READ, uio)) {
2439 * Reset the pointer.
2441 offset = uio->uio_loffset;
2445 zap_cursor_fini(&zc);
2446 if (uio->uio_segflg != UIO_SYSSPACE || uio->uio_iovcnt != 1)
2447 kmem_free(outbuf, bufsize);
2449 if (error == ENOENT)
2452 ZFS_ACCESSTIME_STAMP(zfsvfs, zp);
2454 uio->uio_loffset = offset;
2456 if (error != 0 && cookies != NULL) {
2457 free(*cookies, M_TEMP);
2464 ulong_t zfs_fsync_sync_cnt = 4;
2467 zfs_fsync(vnode_t *vp, int syncflag, cred_t *cr, caller_context_t *ct)
2469 znode_t *zp = VTOZ(vp);
2470 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
2472 (void) tsd_set(zfs_fsyncer_key, (void *)zfs_fsync_sync_cnt);
2476 zil_commit(zfsvfs->z_log, zp->z_last_itx, zp->z_id);
2483 * Get the requested file attributes and place them in the provided
2486 * IN: vp - vnode of file.
2487 * vap - va_mask identifies requested attributes.
2488 * If AT_XVATTR set, then optional attrs are requested
2489 * flags - ATTR_NOACLCHECK (CIFS server context)
2490 * cr - credentials of caller.
2491 * ct - caller context
2493 * OUT: vap - attribute values.
2495 * RETURN: 0 (always succeeds)
2499 zfs_getattr(vnode_t *vp, vattr_t *vap, int flags, cred_t *cr,
2500 caller_context_t *ct)
2502 znode_t *zp = VTOZ(vp);
2503 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
2507 u_longlong_t nblocks;
2509 xvattr_t *xvap = (xvattr_t *)vap; /* vap may be an xvattr_t * */
2510 xoptattr_t *xoap = NULL;
2511 boolean_t skipaclchk = (flags & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE;
2518 * If ACL is trivial don't bother looking for ACE_READ_ATTRIBUTES.
2519 * Also, if we are the owner don't bother, since owner should
2520 * always be allowed to read basic attributes of file.
2522 if (!(pzp->zp_flags & ZFS_ACL_TRIVIAL) &&
2523 (pzp->zp_uid != crgetuid(cr))) {
2524 if (error = zfs_zaccess(zp, ACE_READ_ATTRIBUTES, 0,
2532 * Return all attributes. It's cheaper to provide the answer
2533 * than to determine whether we were asked the question.
2536 mutex_enter(&zp->z_lock);
2537 vap->va_type = IFTOVT(pzp->zp_mode);
2538 vap->va_mode = pzp->zp_mode & ~S_IFMT;
2539 zfs_fuid_map_ids(zp, cr, &vap->va_uid, &vap->va_gid);
2540 // vap->va_fsid = zp->z_zfsvfs->z_vfs->vfs_dev;
2541 vap->va_nodeid = zp->z_id;
2542 if ((vp->v_flag & VROOT) && zfs_show_ctldir(zp))
2543 links = pzp->zp_links + 1;
2545 links = pzp->zp_links;
2546 vap->va_nlink = MIN(links, UINT32_MAX); /* nlink_t limit! */
2547 vap->va_size = pzp->zp_size;
2548 vap->va_fsid = vp->v_mount->mnt_stat.f_fsid.val[0];
2549 vap->va_rdev = zfs_cmpldev(pzp->zp_rdev);
2550 vap->va_seq = zp->z_seq;
2551 vap->va_flags = 0; /* FreeBSD: Reset chflags(2) flags. */
2554 * Add in any requested optional attributes and the create time.
2555 * Also set the corresponding bits in the returned attribute bitmap.
2557 if ((xoap = xva_getxoptattr(xvap)) != NULL && zfsvfs->z_use_fuids) {
2558 if (XVA_ISSET_REQ(xvap, XAT_ARCHIVE)) {
2560 ((pzp->zp_flags & ZFS_ARCHIVE) != 0);
2561 XVA_SET_RTN(xvap, XAT_ARCHIVE);
2564 if (XVA_ISSET_REQ(xvap, XAT_READONLY)) {
2565 xoap->xoa_readonly =
2566 ((pzp->zp_flags & ZFS_READONLY) != 0);
2567 XVA_SET_RTN(xvap, XAT_READONLY);
2570 if (XVA_ISSET_REQ(xvap, XAT_SYSTEM)) {
2572 ((pzp->zp_flags & ZFS_SYSTEM) != 0);
2573 XVA_SET_RTN(xvap, XAT_SYSTEM);
2576 if (XVA_ISSET_REQ(xvap, XAT_HIDDEN)) {
2578 ((pzp->zp_flags & ZFS_HIDDEN) != 0);
2579 XVA_SET_RTN(xvap, XAT_HIDDEN);
2582 if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK)) {
2583 xoap->xoa_nounlink =
2584 ((pzp->zp_flags & ZFS_NOUNLINK) != 0);
2585 XVA_SET_RTN(xvap, XAT_NOUNLINK);
2588 if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE)) {
2589 xoap->xoa_immutable =
2590 ((pzp->zp_flags & ZFS_IMMUTABLE) != 0);
2591 XVA_SET_RTN(xvap, XAT_IMMUTABLE);
2594 if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY)) {
2595 xoap->xoa_appendonly =
2596 ((pzp->zp_flags & ZFS_APPENDONLY) != 0);
2597 XVA_SET_RTN(xvap, XAT_APPENDONLY);
2600 if (XVA_ISSET_REQ(xvap, XAT_NODUMP)) {
2602 ((pzp->zp_flags & ZFS_NODUMP) != 0);
2603 XVA_SET_RTN(xvap, XAT_NODUMP);
2606 if (XVA_ISSET_REQ(xvap, XAT_OPAQUE)) {
2608 ((pzp->zp_flags & ZFS_OPAQUE) != 0);
2609 XVA_SET_RTN(xvap, XAT_OPAQUE);
2612 if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED)) {
2613 xoap->xoa_av_quarantined =
2614 ((pzp->zp_flags & ZFS_AV_QUARANTINED) != 0);
2615 XVA_SET_RTN(xvap, XAT_AV_QUARANTINED);
2618 if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED)) {
2619 xoap->xoa_av_modified =
2620 ((pzp->zp_flags & ZFS_AV_MODIFIED) != 0);
2621 XVA_SET_RTN(xvap, XAT_AV_MODIFIED);
2624 if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP) &&
2625 vp->v_type == VREG &&
2626 (pzp->zp_flags & ZFS_BONUS_SCANSTAMP)) {
2628 dmu_object_info_t doi;
2631 * Only VREG files have anti-virus scanstamps, so we
2632 * won't conflict with symlinks in the bonus buffer.
2634 dmu_object_info_from_db(zp->z_dbuf, &doi);
2635 len = sizeof (xoap->xoa_av_scanstamp) +
2636 sizeof (znode_phys_t);
2637 if (len <= doi.doi_bonus_size) {
2639 * pzp points to the start of the
2640 * znode_phys_t. pzp + 1 points to the
2641 * first byte after the znode_phys_t.
2643 (void) memcpy(xoap->xoa_av_scanstamp,
2645 sizeof (xoap->xoa_av_scanstamp));
2646 XVA_SET_RTN(xvap, XAT_AV_SCANSTAMP);
2650 if (XVA_ISSET_REQ(xvap, XAT_CREATETIME)) {
2651 ZFS_TIME_DECODE(&xoap->xoa_createtime, pzp->zp_crtime);
2652 XVA_SET_RTN(xvap, XAT_CREATETIME);
2656 ZFS_TIME_DECODE(&vap->va_atime, pzp->zp_atime);
2657 ZFS_TIME_DECODE(&vap->va_mtime, pzp->zp_mtime);
2658 ZFS_TIME_DECODE(&vap->va_ctime, pzp->zp_ctime);
2659 ZFS_TIME_DECODE(&vap->va_birthtime, pzp->zp_crtime);
2661 mutex_exit(&zp->z_lock);
2663 dmu_object_size_from_db(zp->z_dbuf, &blksize, &nblocks);
2664 vap->va_blksize = blksize;
2665 vap->va_bytes = nblocks << 9; /* nblocks * 512 */
2667 if (zp->z_blksz == 0) {
2669 * Block size hasn't been set; suggest maximal I/O transfers.
2671 vap->va_blksize = zfsvfs->z_max_blksz;
2679 * Set the file attributes to the values contained in the
2682 * IN: vp - vnode of file to be modified.
2683 * vap - new attribute values.
2684 * If AT_XVATTR set, then optional attrs are being set
2685 * flags - ATTR_UTIME set if non-default time values provided.
2686 * - ATTR_NOACLCHECK (CIFS context only).
2687 * cr - credentials of caller.
2688 * ct - caller context
2690 * RETURN: 0 if success
2691 * error code if failure
2694 * vp - ctime updated, mtime updated if size changed.
2698 zfs_setattr(vnode_t *vp, vattr_t *vap, int flags, cred_t *cr,
2699 caller_context_t *ct)
2701 znode_t *zp = VTOZ(vp);
2703 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
2708 uint_t mask = vap->va_mask;
2710 uint64_t saved_mode;
2713 uint64_t new_uid, new_gid;
2715 int need_policy = FALSE;
2717 zfs_fuid_info_t *fuidp = NULL;
2718 xvattr_t *xvap = (xvattr_t *)vap; /* vap may be an xvattr_t * */
2720 zfs_acl_t *aclp = NULL;
2721 boolean_t skipaclchk = (flags & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE;
2722 boolean_t fuid_dirtied = B_FALSE;
2727 if (mask & AT_NOSET)
2734 zilog = zfsvfs->z_log;
2737 * Make sure that if we have ephemeral uid/gid or xvattr specified
2738 * that file system is at proper version level
2741 if (zfsvfs->z_use_fuids == B_FALSE &&
2742 (((mask & AT_UID) && IS_EPHEMERAL(vap->va_uid)) ||
2743 ((mask & AT_GID) && IS_EPHEMERAL(vap->va_gid)) ||
2744 (mask & AT_XVATTR))) {
2749 if (mask & AT_SIZE && vp->v_type == VDIR) {
2754 if (mask & AT_SIZE && vp->v_type != VREG && vp->v_type != VFIFO) {
2760 * If this is an xvattr_t, then get a pointer to the structure of
2761 * optional attributes. If this is NULL, then we have a vattr_t.
2763 xoap = xva_getxoptattr(xvap);
2765 xva_init(&tmpxvattr);
2768 * Immutable files can only alter immutable bit and atime
2770 if ((pzp->zp_flags & ZFS_IMMUTABLE) &&
2771 ((mask & (AT_SIZE|AT_UID|AT_GID|AT_MTIME|AT_MODE)) ||
2772 ((mask & AT_XVATTR) && XVA_ISSET_REQ(xvap, XAT_CREATETIME)))) {
2777 if ((mask & AT_SIZE) && (pzp->zp_flags & ZFS_READONLY)) {
2783 * Verify timestamps doesn't overflow 32 bits.
2784 * ZFS can handle large timestamps, but 32bit syscalls can't
2785 * handle times greater than 2039. This check should be removed
2786 * once large timestamps are fully supported.
2788 if (mask & (AT_ATIME | AT_MTIME)) {
2789 if (((mask & AT_ATIME) && TIMESPEC_OVERFLOW(&vap->va_atime)) ||
2790 ((mask & AT_MTIME) && TIMESPEC_OVERFLOW(&vap->va_mtime))) {
2799 /* Can this be moved to before the top label? */
2800 if (zfsvfs->z_vfs->vfs_flag & VFS_RDONLY) {
2806 * First validate permissions
2809 if (mask & AT_SIZE) {
2810 err = zfs_zaccess(zp, ACE_WRITE_DATA, 0, skipaclchk, cr);
2816 * XXX - Note, we are not providing any open
2817 * mode flags here (like FNDELAY), so we may
2818 * block if there are locks present... this
2819 * should be addressed in openat().
2821 /* XXX - would it be OK to generate a log record here? */
2822 err = zfs_freesp(zp, vap->va_size, 0, 0, FALSE);
2829 if (mask & (AT_ATIME|AT_MTIME) ||
2830 ((mask & AT_XVATTR) && (XVA_ISSET_REQ(xvap, XAT_HIDDEN) ||
2831 XVA_ISSET_REQ(xvap, XAT_READONLY) ||
2832 XVA_ISSET_REQ(xvap, XAT_ARCHIVE) ||
2833 XVA_ISSET_REQ(xvap, XAT_CREATETIME) ||
2834 XVA_ISSET_REQ(xvap, XAT_SYSTEM))))
2835 need_policy = zfs_zaccess(zp, ACE_WRITE_ATTRIBUTES, 0,
2838 if (mask & (AT_UID|AT_GID)) {
2839 int idmask = (mask & (AT_UID|AT_GID));
2844 * NOTE: even if a new mode is being set,
2845 * we may clear S_ISUID/S_ISGID bits.
2848 if (!(mask & AT_MODE))
2849 vap->va_mode = pzp->zp_mode;
2852 * Take ownership or chgrp to group we are a member of
2855 take_owner = (mask & AT_UID) && (vap->va_uid == crgetuid(cr));
2856 take_group = (mask & AT_GID) &&
2857 zfs_groupmember(zfsvfs, vap->va_gid, cr);
2860 * If both AT_UID and AT_GID are set then take_owner and
2861 * take_group must both be set in order to allow taking
2864 * Otherwise, send the check through secpolicy_vnode_setattr()
2868 if (((idmask == (AT_UID|AT_GID)) && take_owner && take_group) ||
2869 ((idmask == AT_UID) && take_owner) ||
2870 ((idmask == AT_GID) && take_group)) {
2871 if (zfs_zaccess(zp, ACE_WRITE_OWNER, 0,
2872 skipaclchk, cr) == 0) {
2874 * Remove setuid/setgid for non-privileged users
2876 secpolicy_setid_clear(vap, vp, cr);
2877 trim_mask = (mask & (AT_UID|AT_GID));
2886 mutex_enter(&zp->z_lock);
2887 oldva.va_mode = pzp->zp_mode;
2888 zfs_fuid_map_ids(zp, cr, &oldva.va_uid, &oldva.va_gid);
2889 if (mask & AT_XVATTR) {
2891 * Update xvattr mask to include only those attributes
2892 * that are actually changing.
2894 * the bits will be restored prior to actually setting
2895 * the attributes so the caller thinks they were set.
2897 if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY)) {
2898 if (xoap->xoa_appendonly !=
2899 ((pzp->zp_flags & ZFS_APPENDONLY) != 0)) {
2902 XVA_CLR_REQ(xvap, XAT_APPENDONLY);
2903 XVA_SET_REQ(&tmpxvattr, XAT_APPENDONLY);
2907 if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK)) {
2908 if (xoap->xoa_nounlink !=
2909 ((pzp->zp_flags & ZFS_NOUNLINK) != 0)) {
2912 XVA_CLR_REQ(xvap, XAT_NOUNLINK);
2913 XVA_SET_REQ(&tmpxvattr, XAT_NOUNLINK);
2917 if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE)) {
2918 if (xoap->xoa_immutable !=
2919 ((pzp->zp_flags & ZFS_IMMUTABLE) != 0)) {
2922 XVA_CLR_REQ(xvap, XAT_IMMUTABLE);
2923 XVA_SET_REQ(&tmpxvattr, XAT_IMMUTABLE);
2927 if (XVA_ISSET_REQ(xvap, XAT_NODUMP)) {
2928 if (xoap->xoa_nodump !=
2929 ((pzp->zp_flags & ZFS_NODUMP) != 0)) {
2932 XVA_CLR_REQ(xvap, XAT_NODUMP);
2933 XVA_SET_REQ(&tmpxvattr, XAT_NODUMP);
2937 if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED)) {
2938 if (xoap->xoa_av_modified !=
2939 ((pzp->zp_flags & ZFS_AV_MODIFIED) != 0)) {
2942 XVA_CLR_REQ(xvap, XAT_AV_MODIFIED);
2943 XVA_SET_REQ(&tmpxvattr, XAT_AV_MODIFIED);
2947 if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED)) {
2948 if ((vp->v_type != VREG &&
2949 xoap->xoa_av_quarantined) ||
2950 xoap->xoa_av_quarantined !=
2951 ((pzp->zp_flags & ZFS_AV_QUARANTINED) != 0)) {
2954 XVA_CLR_REQ(xvap, XAT_AV_QUARANTINED);
2955 XVA_SET_REQ(&tmpxvattr, XAT_AV_QUARANTINED);
2959 if (need_policy == FALSE &&
2960 (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP) ||
2961 XVA_ISSET_REQ(xvap, XAT_OPAQUE))) {
2966 mutex_exit(&zp->z_lock);
2968 if (mask & AT_MODE) {
2969 if (zfs_zaccess(zp, ACE_WRITE_ACL, 0, skipaclchk, cr) == 0) {
2970 err = secpolicy_setid_setsticky_clear(vp, vap,
2976 trim_mask |= AT_MODE;
2984 * If trim_mask is set then take ownership
2985 * has been granted or write_acl is present and user
2986 * has the ability to modify mode. In that case remove
2987 * UID|GID and or MODE from mask so that
2988 * secpolicy_vnode_setattr() doesn't revoke it.
2992 saved_mask = vap->va_mask;
2993 vap->va_mask &= ~trim_mask;
2994 if (trim_mask & AT_MODE) {
2996 * Save the mode, as secpolicy_vnode_setattr()
2997 * will overwrite it with ova.va_mode.
2999 saved_mode = vap->va_mode;
3002 err = secpolicy_vnode_setattr(cr, vp, vap, &oldva, flags,
3003 (int (*)(void *, int, cred_t *))zfs_zaccess_unix, zp);
3010 vap->va_mask |= saved_mask;
3011 if (trim_mask & AT_MODE) {
3013 * Recover the mode after
3014 * secpolicy_vnode_setattr().
3016 vap->va_mode = saved_mode;
3022 * secpolicy_vnode_setattr, or take ownership may have
3025 mask = vap->va_mask;
3027 tx = dmu_tx_create(zfsvfs->z_os);
3028 dmu_tx_hold_bonus(tx, zp->z_id);
3030 if (mask & AT_MODE) {
3031 uint64_t pmode = pzp->zp_mode;
3033 new_mode = (pmode & S_IFMT) | (vap->va_mode & ~S_IFMT);
3035 if (err = zfs_acl_chmod_setattr(zp, &aclp, new_mode))
3037 if (pzp->zp_acl.z_acl_extern_obj) {
3038 /* Are we upgrading ACL from old V0 format to new V1 */
3039 if (zfsvfs->z_version <= ZPL_VERSION_FUID &&
3040 pzp->zp_acl.z_acl_version ==
3041 ZFS_ACL_VERSION_INITIAL) {
3042 dmu_tx_hold_free(tx,
3043 pzp->zp_acl.z_acl_extern_obj, 0,
3045 dmu_tx_hold_write(tx, DMU_NEW_OBJECT,
3046 0, aclp->z_acl_bytes);
3048 dmu_tx_hold_write(tx,
3049 pzp->zp_acl.z_acl_extern_obj, 0,
3052 } else if (aclp->z_acl_bytes > ZFS_ACE_SPACE) {
3053 dmu_tx_hold_write(tx, DMU_NEW_OBJECT,
3054 0, aclp->z_acl_bytes);
3058 if (mask & (AT_UID | AT_GID)) {
3059 if (pzp->zp_xattr) {
3060 err = zfs_zget(zp->z_zfsvfs, pzp->zp_xattr, &attrzp);
3063 dmu_tx_hold_bonus(tx, attrzp->z_id);
3065 if (mask & AT_UID) {
3066 new_uid = zfs_fuid_create(zfsvfs,
3067 (uint64_t)vap->va_uid, cr, ZFS_OWNER, &fuidp);
3068 if (new_uid != pzp->zp_uid &&
3069 zfs_usergroup_overquota(zfsvfs, B_FALSE, new_uid)) {
3075 if (mask & AT_GID) {
3076 new_gid = zfs_fuid_create(zfsvfs, (uint64_t)vap->va_gid,
3077 cr, ZFS_GROUP, &fuidp);
3078 if (new_gid != pzp->zp_gid &&
3079 zfs_usergroup_overquota(zfsvfs, B_TRUE, new_gid)) {
3084 fuid_dirtied = zfsvfs->z_fuid_dirty;
3086 if (zfsvfs->z_fuid_obj == 0) {
3087 dmu_tx_hold_bonus(tx, DMU_NEW_OBJECT);
3088 dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0,
3089 FUID_SIZE_ESTIMATE(zfsvfs));
3090 dmu_tx_hold_zap(tx, MASTER_NODE_OBJ,
3093 dmu_tx_hold_bonus(tx, zfsvfs->z_fuid_obj);
3094 dmu_tx_hold_write(tx, zfsvfs->z_fuid_obj, 0,
3095 FUID_SIZE_ESTIMATE(zfsvfs));
3100 err = dmu_tx_assign(tx, TXG_NOWAIT);
3102 if (err == ERESTART)
3107 dmu_buf_will_dirty(zp->z_dbuf, tx);
3110 * Set each attribute requested.
3111 * We group settings according to the locks they need to acquire.
3113 * Note: you cannot set ctime directly, although it will be
3114 * updated as a side-effect of calling this function.
3117 mutex_enter(&zp->z_lock);
3119 if (mask & AT_MODE) {
3120 mutex_enter(&zp->z_acl_lock);
3121 zp->z_phys->zp_mode = new_mode;
3122 err = zfs_aclset_common(zp, aclp, cr, tx);
3123 ASSERT3U(err, ==, 0);
3124 zp->z_acl_cached = aclp;
3126 mutex_exit(&zp->z_acl_lock);
3130 mutex_enter(&attrzp->z_lock);
3132 if (mask & AT_UID) {
3133 pzp->zp_uid = new_uid;
3135 attrzp->z_phys->zp_uid = new_uid;
3138 if (mask & AT_GID) {
3139 pzp->zp_gid = new_gid;
3141 attrzp->z_phys->zp_gid = new_gid;
3145 mutex_exit(&attrzp->z_lock);
3147 if (mask & AT_ATIME)
3148 ZFS_TIME_ENCODE(&vap->va_atime, pzp->zp_atime);
3150 if (mask & AT_MTIME)
3151 ZFS_TIME_ENCODE(&vap->va_mtime, pzp->zp_mtime);
3153 /* XXX - shouldn't this be done *before* the ATIME/MTIME checks? */
3155 zfs_time_stamper_locked(zp, CONTENT_MODIFIED, tx);
3157 zfs_time_stamper_locked(zp, STATE_CHANGED, tx);
3159 * Do this after setting timestamps to prevent timestamp
3160 * update from toggling bit
3163 if (xoap && (mask & AT_XVATTR)) {
3166 * restore trimmed off masks
3167 * so that return masks can be set for caller.
3170 if (XVA_ISSET_REQ(&tmpxvattr, XAT_APPENDONLY)) {
3171 XVA_SET_REQ(xvap, XAT_APPENDONLY);
3173 if (XVA_ISSET_REQ(&tmpxvattr, XAT_NOUNLINK)) {
3174 XVA_SET_REQ(xvap, XAT_NOUNLINK);
3176 if (XVA_ISSET_REQ(&tmpxvattr, XAT_IMMUTABLE)) {
3177 XVA_SET_REQ(xvap, XAT_IMMUTABLE);
3179 if (XVA_ISSET_REQ(&tmpxvattr, XAT_NODUMP)) {
3180 XVA_SET_REQ(xvap, XAT_NODUMP);
3182 if (XVA_ISSET_REQ(&tmpxvattr, XAT_AV_MODIFIED)) {
3183 XVA_SET_REQ(xvap, XAT_AV_MODIFIED);
3185 if (XVA_ISSET_REQ(&tmpxvattr, XAT_AV_QUARANTINED)) {
3186 XVA_SET_REQ(xvap, XAT_AV_QUARANTINED);
3189 if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP)) {
3191 dmu_object_info_t doi;
3193 ASSERT(vp->v_type == VREG);
3195 /* Grow the bonus buffer if necessary. */
3196 dmu_object_info_from_db(zp->z_dbuf, &doi);
3197 len = sizeof (xoap->xoa_av_scanstamp) +
3198 sizeof (znode_phys_t);
3199 if (len > doi.doi_bonus_size)
3200 VERIFY(dmu_set_bonus(zp->z_dbuf, len, tx) == 0);
3202 zfs_xvattr_set(zp, xvap);
3206 zfs_fuid_sync(zfsvfs, tx);
3209 zfs_log_setattr(zilog, tx, TX_SETATTR, zp, vap, mask, fuidp);
3211 mutex_exit(&zp->z_lock);
3215 VN_RELE(ZTOV(attrzp));
3221 zfs_fuid_info_free(fuidp);
3230 if (err == ERESTART)
3237 typedef struct zfs_zlock {
3238 krwlock_t *zl_rwlock; /* lock we acquired */
3239 znode_t *zl_znode; /* znode we held */
3240 struct zfs_zlock *zl_next; /* next in list */
3244 * Drop locks and release vnodes that were held by zfs_rename_lock().
3247 zfs_rename_unlock(zfs_zlock_t **zlpp)
3251 while ((zl = *zlpp) != NULL) {
3252 if (zl->zl_znode != NULL)
3253 VN_RELE(ZTOV(zl->zl_znode));
3254 rw_exit(zl->zl_rwlock);
3255 *zlpp = zl->zl_next;
3256 kmem_free(zl, sizeof (*zl));
3261 * Search back through the directory tree, using the ".." entries.
3262 * Lock each directory in the chain to prevent concurrent renames.
3263 * Fail any attempt to move a directory into one of its own descendants.
3264 * XXX - z_parent_lock can overlap with map or grow locks
3267 zfs_rename_lock(znode_t *szp, znode_t *tdzp, znode_t *sdzp, zfs_zlock_t **zlpp)
3271 uint64_t rootid = zp->z_zfsvfs->z_root;
3272 uint64_t *oidp = &zp->z_id;
3273 krwlock_t *rwlp = &szp->z_parent_lock;
3274 krw_t rw = RW_WRITER;
3277 * First pass write-locks szp and compares to zp->z_id.
3278 * Later passes read-lock zp and compare to zp->z_parent.
3281 if (!rw_tryenter(rwlp, rw)) {
3283 * Another thread is renaming in this path.
3284 * Note that if we are a WRITER, we don't have any
3285 * parent_locks held yet.
3287 if (rw == RW_READER && zp->z_id > szp->z_id) {
3289 * Drop our locks and restart
3291 zfs_rename_unlock(&zl);
3295 rwlp = &szp->z_parent_lock;
3300 * Wait for other thread to drop its locks
3306 zl = kmem_alloc(sizeof (*zl), KM_SLEEP);
3307 zl->zl_rwlock = rwlp;
3308 zl->zl_znode = NULL;
3309 zl->zl_next = *zlpp;
3312 if (*oidp == szp->z_id) /* We're a descendant of szp */
3315 if (*oidp == rootid) /* We've hit the top */
3318 if (rw == RW_READER) { /* i.e. not the first pass */
3319 int error = zfs_zget(zp->z_zfsvfs, *oidp, &zp);
3324 oidp = &zp->z_phys->zp_parent;
3325 rwlp = &zp->z_parent_lock;
3328 } while (zp->z_id != sdzp->z_id);
3334 * Move an entry from the provided source directory to the target
3335 * directory. Change the entry name as indicated.
3337 * IN: sdvp - Source directory containing the "old entry".
3338 * snm - Old entry name.
3339 * tdvp - Target directory to contain the "new entry".
3340 * tnm - New entry name.
3341 * cr - credentials of caller.
3342 * ct - caller context
3343 * flags - case flags
3345 * RETURN: 0 if success
3346 * error code if failure
3349 * sdvp,tdvp - ctime|mtime updated
3353 zfs_rename(vnode_t *sdvp, char *snm, vnode_t *tdvp, char *tnm, cred_t *cr,
3354 caller_context_t *ct, int flags)
3356 znode_t *tdzp, *szp, *tzp;
3357 znode_t *sdzp = VTOZ(sdvp);
3358 zfsvfs_t *zfsvfs = sdzp->z_zfsvfs;
3361 zfs_dirlock_t *sdl, *tdl;
3364 int cmp, serr, terr;
3369 ZFS_VERIFY_ZP(sdzp);
3370 zilog = zfsvfs->z_log;
3373 * Make sure we have the real vp for the target directory.
3375 if (VOP_REALVP(tdvp, &realvp, ct) == 0)
3378 if (tdvp->v_vfsp != sdvp->v_vfsp || zfsctl_is_node(tdvp)) {
3384 ZFS_VERIFY_ZP(tdzp);
3385 if (zfsvfs->z_utf8 && u8_validate(tnm,
3386 strlen(tnm), NULL, U8_VALIDATE_ENTIRE, &error) < 0) {
3391 if (flags & FIGNORECASE)
3400 * This is to prevent the creation of links into attribute space
3401 * by renaming a linked file into/outof an attribute directory.
3402 * See the comment in zfs_link() for why this is considered bad.
3404 if ((tdzp->z_phys->zp_flags & ZFS_XATTR) !=
3405 (sdzp->z_phys->zp_flags & ZFS_XATTR)) {
3411 * Lock source and target directory entries. To prevent deadlock,
3412 * a lock ordering must be defined. We lock the directory with
3413 * the smallest object id first, or if it's a tie, the one with
3414 * the lexically first name.
3416 if (sdzp->z_id < tdzp->z_id) {
3418 } else if (sdzp->z_id > tdzp->z_id) {
3422 * First compare the two name arguments without
3423 * considering any case folding.
3425 int nofold = (zfsvfs->z_norm & ~U8_TEXTPREP_TOUPPER);
3427 cmp = u8_strcmp(snm, tnm, 0, nofold, U8_UNICODE_LATEST, &error);
3428 ASSERT(error == 0 || !zfsvfs->z_utf8);
3431 * POSIX: "If the old argument and the new argument
3432 * both refer to links to the same existing file,
3433 * the rename() function shall return successfully
3434 * and perform no other action."
3440 * If the file system is case-folding, then we may
3441 * have some more checking to do. A case-folding file
3442 * system is either supporting mixed case sensitivity
3443 * access or is completely case-insensitive. Note
3444 * that the file system is always case preserving.
3446 * In mixed sensitivity mode case sensitive behavior
3447 * is the default. FIGNORECASE must be used to
3448 * explicitly request case insensitive behavior.
3450 * If the source and target names provided differ only
3451 * by case (e.g., a request to rename 'tim' to 'Tim'),
3452 * we will treat this as a special case in the
3453 * case-insensitive mode: as long as the source name
3454 * is an exact match, we will allow this to proceed as
3455 * a name-change request.
3457 if ((zfsvfs->z_case == ZFS_CASE_INSENSITIVE ||
3458 (zfsvfs->z_case == ZFS_CASE_MIXED &&
3459 flags & FIGNORECASE)) &&
3460 u8_strcmp(snm, tnm, 0, zfsvfs->z_norm, U8_UNICODE_LATEST,
3463 * case preserving rename request, require exact
3472 * If the source and destination directories are the same, we should
3473 * grab the z_name_lock of that directory only once.
3477 rw_enter(&sdzp->z_name_lock, RW_READER);
3481 serr = zfs_dirent_lock(&sdl, sdzp, snm, &szp,
3482 ZEXISTS | zflg, NULL, NULL);
3483 terr = zfs_dirent_lock(&tdl,
3484 tdzp, tnm, &tzp, ZRENAMING | zflg, NULL, NULL);
3486 terr = zfs_dirent_lock(&tdl,
3487 tdzp, tnm, &tzp, zflg, NULL, NULL);
3488 serr = zfs_dirent_lock(&sdl,
3489 sdzp, snm, &szp, ZEXISTS | ZRENAMING | zflg,
3495 * Source entry invalid or not there.
3498 zfs_dirent_unlock(tdl);
3504 rw_exit(&sdzp->z_name_lock);
3506 if (strcmp(snm, ".") == 0 || strcmp(snm, "..") == 0)
3512 zfs_dirent_unlock(sdl);
3516 rw_exit(&sdzp->z_name_lock);
3518 if (strcmp(tnm, "..") == 0)
3525 * Must have write access at the source to remove the old entry
3526 * and write access at the target to create the new entry.
3527 * Note that if target and source are the same, this can be
3528 * done in a single check.
3531 if (error = zfs_zaccess_rename(sdzp, szp, tdzp, tzp, cr))
3534 if (ZTOV(szp)->v_type == VDIR) {
3536 * Check to make sure rename is valid.
3537 * Can't do a move like this: /usr/a/b to /usr/a/b/c/d
3539 if (error = zfs_rename_lock(szp, tdzp, sdzp, &zl))
3544 * Does target exist?
3548 * Source and target must be the same type.
3550 if (ZTOV(szp)->v_type == VDIR) {
3551 if (ZTOV(tzp)->v_type != VDIR) {
3556 if (ZTOV(tzp)->v_type == VDIR) {
3562 * POSIX dictates that when the source and target
3563 * entries refer to the same file object, rename
3564 * must do nothing and exit without error.
3566 if (szp->z_id == tzp->z_id) {
3572 vnevent_rename_src(ZTOV(szp), sdvp, snm, ct);
3574 vnevent_rename_dest(ZTOV(tzp), tdvp, tnm, ct);
3577 * notify the target directory if it is not the same
3578 * as source directory.
3581 vnevent_rename_dest_dir(tdvp, ct);
3584 tx = dmu_tx_create(zfsvfs->z_os);
3585 dmu_tx_hold_bonus(tx, szp->z_id); /* nlink changes */
3586 dmu_tx_hold_bonus(tx, sdzp->z_id); /* nlink changes */
3587 dmu_tx_hold_zap(tx, sdzp->z_id, FALSE, snm);
3588 dmu_tx_hold_zap(tx, tdzp->z_id, TRUE, tnm);
3590 dmu_tx_hold_bonus(tx, tdzp->z_id); /* nlink changes */
3592 dmu_tx_hold_bonus(tx, tzp->z_id); /* parent changes */
3593 dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL);
3594 error = dmu_tx_assign(tx, TXG_NOWAIT);
3597 zfs_rename_unlock(&zl);
3598 zfs_dirent_unlock(sdl);
3599 zfs_dirent_unlock(tdl);
3602 rw_exit(&sdzp->z_name_lock);
3607 if (error == ERESTART) {
3617 if (tzp) /* Attempt to remove the existing target */
3618 error = zfs_link_destroy(tdl, tzp, tx, zflg, NULL);
3621 error = zfs_link_create(tdl, szp, tx, ZRENAMING);
3623 szp->z_phys->zp_flags |= ZFS_AV_MODIFIED;
3625 error = zfs_link_destroy(sdl, szp, tx, ZRENAMING, NULL);
3628 zfs_log_rename(zilog, tx,
3629 TX_RENAME | (flags & FIGNORECASE ? TX_CI : 0),
3630 sdzp, sdl->dl_name, tdzp, tdl->dl_name, szp);
3632 /* Update path information for the target vnode */
3633 vn_renamepath(tdvp, ZTOV(szp), tnm, strlen(tnm));
3635 #ifdef FREEBSD_NAMECACHE
3646 zfs_rename_unlock(&zl);
3648 zfs_dirent_unlock(sdl);
3649 zfs_dirent_unlock(tdl);
3652 rw_exit(&sdzp->z_name_lock);
3664 * Insert the indicated symbolic reference entry into the directory.
3666 * IN: dvp - Directory to contain new symbolic link.
3667 * link - Name for new symlink entry.
3668 * vap - Attributes of new entry.
3669 * target - Target path of new symlink.
3670 * cr - credentials of caller.
3671 * ct - caller context
3672 * flags - case flags
3674 * RETURN: 0 if success
3675 * error code if failure
3678 * dvp - ctime|mtime updated
3682 zfs_symlink(vnode_t *dvp, vnode_t **vpp, char *name, vattr_t *vap, char *link,
3683 cred_t *cr, kthread_t *td)
3685 znode_t *zp, *dzp = VTOZ(dvp);
3688 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
3690 int len = strlen(link);
3693 zfs_acl_ids_t acl_ids;
3694 boolean_t fuid_dirtied;
3697 ASSERT(vap->va_type == VLNK);
3701 zilog = zfsvfs->z_log;
3703 if (zfsvfs->z_utf8 && u8_validate(name, strlen(name),
3704 NULL, U8_VALIDATE_ENTIRE, &error) < 0) {
3708 if (flags & FIGNORECASE)
3711 if (error = zfs_zaccess(dzp, ACE_ADD_FILE, 0, B_FALSE, cr)) {
3716 if (len > MAXPATHLEN) {
3718 return (ENAMETOOLONG);
3722 * Attempt to lock directory; fail if entry already exists.
3724 error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg, NULL, NULL);
3730 VERIFY(0 == zfs_acl_ids_create(dzp, 0, vap, cr, NULL, &acl_ids));
3731 if (zfs_acl_ids_overquota(zfsvfs, &acl_ids)) {
3732 zfs_acl_ids_free(&acl_ids);
3733 zfs_dirent_unlock(dl);
3737 tx = dmu_tx_create(zfsvfs->z_os);
3738 fuid_dirtied = zfsvfs->z_fuid_dirty;
3739 dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, MAX(1, len));
3740 dmu_tx_hold_bonus(tx, dzp->z_id);
3741 dmu_tx_hold_zap(tx, dzp->z_id, TRUE, name);
3742 if (acl_ids.z_aclp->z_acl_bytes > ZFS_ACE_SPACE)
3743 dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, SPA_MAXBLOCKSIZE);
3745 zfs_fuid_txhold(zfsvfs, tx);
3746 error = dmu_tx_assign(tx, TXG_NOWAIT);
3748 zfs_acl_ids_free(&acl_ids);
3749 zfs_dirent_unlock(dl);
3750 if (error == ERESTART) {
3760 dmu_buf_will_dirty(dzp->z_dbuf, tx);
3763 * Create a new object for the symlink.
3764 * Put the link content into bonus buffer if it will fit;
3765 * otherwise, store it just like any other file data.
3767 if (sizeof (znode_phys_t) + len <= dmu_bonus_max()) {
3768 zfs_mknode(dzp, vap, tx, cr, 0, &zp, len, &acl_ids);
3770 bcopy(link, zp->z_phys + 1, len);
3774 zfs_mknode(dzp, vap, tx, cr, 0, &zp, 0, &acl_ids);
3777 zfs_fuid_sync(zfsvfs, tx);
3779 * Nothing can access the znode yet so no locking needed
3780 * for growing the znode's blocksize.
3782 zfs_grow_blocksize(zp, len, tx);
3784 VERIFY(0 == dmu_buf_hold(zfsvfs->z_os,
3785 zp->z_id, 0, FTAG, &dbp));
3786 dmu_buf_will_dirty(dbp, tx);
3788 ASSERT3U(len, <=, dbp->db_size);
3789 bcopy(link, dbp->db_data, len);
3790 dmu_buf_rele(dbp, FTAG);
3792 zp->z_phys->zp_size = len;
3795 * Insert the new object into the directory.
3797 (void) zfs_link_create(dl, zp, tx, ZNEW);
3799 uint64_t txtype = TX_SYMLINK;
3800 if (flags & FIGNORECASE)
3802 zfs_log_symlink(zilog, tx, txtype, dzp, zp, name, link);
3806 zfs_acl_ids_free(&acl_ids);
3810 zfs_dirent_unlock(dl);
3817 * Return, in the buffer contained in the provided uio structure,
3818 * the symbolic path referred to by vp.
3820 * IN: vp - vnode of symbolic link.
3821 * uoip - structure to contain the link path.
3822 * cr - credentials of caller.
3823 * ct - caller context
3825 * OUT: uio - structure to contain the link path.
3827 * RETURN: 0 if success
3828 * error code if failure
3831 * vp - atime updated
3835 zfs_readlink(vnode_t *vp, uio_t *uio, cred_t *cr, caller_context_t *ct)
3837 znode_t *zp = VTOZ(vp);
3838 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
3845 bufsz = (size_t)zp->z_phys->zp_size;
3846 if (bufsz + sizeof (znode_phys_t) <= zp->z_dbuf->db_size) {
3847 error = uiomove(zp->z_phys + 1,
3848 MIN((size_t)bufsz, uio->uio_resid), UIO_READ, uio);
3851 error = dmu_buf_hold(zfsvfs->z_os, zp->z_id, 0, FTAG, &dbp);
3856 error = uiomove(dbp->db_data,
3857 MIN((size_t)bufsz, uio->uio_resid), UIO_READ, uio);
3858 dmu_buf_rele(dbp, FTAG);
3861 ZFS_ACCESSTIME_STAMP(zfsvfs, zp);
3867 * Insert a new entry into directory tdvp referencing svp.
3869 * IN: tdvp - Directory to contain new entry.
3870 * svp - vnode of new entry.
3871 * name - name of new entry.
3872 * cr - credentials of caller.
3873 * ct - caller context
3875 * RETURN: 0 if success
3876 * error code if failure
3879 * tdvp - ctime|mtime updated
3880 * svp - ctime updated
3884 zfs_link(vnode_t *tdvp, vnode_t *svp, char *name, cred_t *cr,
3885 caller_context_t *ct, int flags)
3887 znode_t *dzp = VTOZ(tdvp);
3889 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
3899 ASSERT(tdvp->v_type == VDIR);
3903 zilog = zfsvfs->z_log;
3905 if (VOP_REALVP(svp, &realvp, ct) == 0)
3909 * POSIX dictates that we return EPERM here.
3910 * Better choices include ENOTSUP or EISDIR.
3912 if (svp->v_type == VDIR) {
3917 if (svp->v_vfsp != tdvp->v_vfsp || zfsctl_is_node(svp)) {
3925 /* Prevent links to .zfs/shares files */
3927 if (szp->z_phys->zp_parent == zfsvfs->z_shares_dir) {
3932 if (zfsvfs->z_utf8 && u8_validate(name,
3933 strlen(name), NULL, U8_VALIDATE_ENTIRE, &error) < 0) {
3937 if (flags & FIGNORECASE)
3941 * We do not support links between attributes and non-attributes
3942 * because of the potential security risk of creating links
3943 * into "normal" file space in order to circumvent restrictions
3944 * imposed in attribute space.
3946 if ((szp->z_phys->zp_flags & ZFS_XATTR) !=
3947 (dzp->z_phys->zp_flags & ZFS_XATTR)) {
3953 owner = zfs_fuid_map_id(zfsvfs, szp->z_phys->zp_uid, cr, ZFS_OWNER);
3954 if (owner != crgetuid(cr) &&
3955 secpolicy_basic_link(svp, cr) != 0) {
3960 if (error = zfs_zaccess(dzp, ACE_ADD_FILE, 0, B_FALSE, cr)) {
3967 * Attempt to lock directory; fail if entry already exists.
3969 error = zfs_dirent_lock(&dl, dzp, name, &tzp, zf, NULL, NULL);
3975 tx = dmu_tx_create(zfsvfs->z_os);
3976 dmu_tx_hold_bonus(tx, szp->z_id);
3977 dmu_tx_hold_zap(tx, dzp->z_id, TRUE, name);
3978 error = dmu_tx_assign(tx, TXG_NOWAIT);
3980 zfs_dirent_unlock(dl);
3981 if (error == ERESTART) {
3991 error = zfs_link_create(dl, szp, tx, 0);
3994 uint64_t txtype = TX_LINK;
3995 if (flags & FIGNORECASE)
3997 zfs_log_link(zilog, tx, txtype, dzp, szp, name);
4002 zfs_dirent_unlock(dl);
4005 vnevent_link(svp, ct);
4014 zfs_inactive(vnode_t *vp, cred_t *cr, caller_context_t *ct)
4016 znode_t *zp = VTOZ(vp);
4017 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
4020 rw_enter(&zfsvfs->z_teardown_inactive_lock, RW_READER);
4021 if (zp->z_dbuf == NULL) {
4023 * The fs has been unmounted, or we did a
4024 * suspend/resume and this file no longer exists.
4027 vp->v_count = 0; /* count arrives as 1 */
4029 vrecycle(vp, curthread);
4030 rw_exit(&zfsvfs->z_teardown_inactive_lock);
4034 if (zp->z_atime_dirty && zp->z_unlinked == 0) {
4035 dmu_tx_t *tx = dmu_tx_create(zfsvfs->z_os);
4037 dmu_tx_hold_bonus(tx, zp->z_id);
4038 error = dmu_tx_assign(tx, TXG_WAIT);
4042 dmu_buf_will_dirty(zp->z_dbuf, tx);
4043 mutex_enter(&zp->z_lock);
4044 zp->z_atime_dirty = 0;
4045 mutex_exit(&zp->z_lock);
4051 rw_exit(&zfsvfs->z_teardown_inactive_lock);
4054 CTASSERT(sizeof(struct zfid_short) <= sizeof(struct fid));
4055 CTASSERT(sizeof(struct zfid_long) <= sizeof(struct fid));
4059 zfs_fid(vnode_t *vp, fid_t *fidp, caller_context_t *ct)
4061 znode_t *zp = VTOZ(vp);
4062 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
4064 uint64_t object = zp->z_id;
4070 gen = (uint32_t)zp->z_gen;
4072 size = (zfsvfs->z_parent != zfsvfs) ? LONG_FID_LEN : SHORT_FID_LEN;
4073 fidp->fid_len = size;
4075 zfid = (zfid_short_t *)fidp;
4077 zfid->zf_len = size;
4079 for (i = 0; i < sizeof (zfid->zf_object); i++)
4080 zfid->zf_object[i] = (uint8_t)(object >> (8 * i));
4082 /* Must have a non-zero generation number to distinguish from .zfs */
4085 for (i = 0; i < sizeof (zfid->zf_gen); i++)
4086 zfid->zf_gen[i] = (uint8_t)(gen >> (8 * i));
4088 if (size == LONG_FID_LEN) {
4089 uint64_t objsetid = dmu_objset_id(zfsvfs->z_os);
4092 zlfid = (zfid_long_t *)fidp;
4094 for (i = 0; i < sizeof (zlfid->zf_setid); i++)
4095 zlfid->zf_setid[i] = (uint8_t)(objsetid >> (8 * i));
4097 /* XXX - this should be the generation number for the objset */
4098 for (i = 0; i < sizeof (zlfid->zf_setgen); i++)
4099 zlfid->zf_setgen[i] = 0;
4107 zfs_pathconf(vnode_t *vp, int cmd, ulong_t *valp, cred_t *cr,
4108 caller_context_t *ct)
4120 case _PC_FILESIZEBITS:
4125 case _PC_XATTR_EXISTS:
4127 zfsvfs = zp->z_zfsvfs;
4131 error = zfs_dirent_lock(&dl, zp, "", &xzp,
4132 ZXATTR | ZEXISTS | ZSHARED, NULL, NULL);
4134 zfs_dirent_unlock(dl);
4135 if (!zfs_dirempty(xzp))
4138 } else if (error == ENOENT) {
4140 * If there aren't extended attributes, it's the
4141 * same as having zero of them.
4149 case _PC_ACL_EXTENDED:
4157 case _PC_ACL_PATH_MAX:
4158 *valp = ACL_MAX_ENTRIES;
4161 case _PC_MIN_HOLE_SIZE:
4162 *valp = (int)SPA_MINBLOCKSIZE;
4166 return (EOPNOTSUPP);
4172 zfs_getsecattr(vnode_t *vp, vsecattr_t *vsecp, int flag, cred_t *cr,
4173 caller_context_t *ct)
4175 znode_t *zp = VTOZ(vp);
4176 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
4178 boolean_t skipaclchk = (flag & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE;
4182 error = zfs_getacl(zp, vsecp, skipaclchk, cr);
4190 zfs_setsecattr(vnode_t *vp, vsecattr_t *vsecp, int flag, cred_t *cr,
4191 caller_context_t *ct)
4193 znode_t *zp = VTOZ(vp);
4194 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
4196 boolean_t skipaclchk = (flag & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE;
4200 error = zfs_setacl(zp, vsecp, skipaclchk, cr);
4206 zfs_freebsd_open(ap)
4207 struct vop_open_args /* {
4210 struct ucred *a_cred;
4211 struct thread *a_td;
4214 vnode_t *vp = ap->a_vp;
4215 znode_t *zp = VTOZ(vp);
4218 error = zfs_open(&vp, ap->a_mode, ap->a_cred, NULL);
4220 vnode_create_vobject(vp, zp->z_phys->zp_size, ap->a_td);
4225 zfs_freebsd_close(ap)
4226 struct vop_close_args /* {
4229 struct ucred *a_cred;
4230 struct thread *a_td;
4234 return (zfs_close(ap->a_vp, ap->a_fflag, 0, 0, ap->a_cred, NULL));
4238 zfs_freebsd_ioctl(ap)
4239 struct vop_ioctl_args /* {
4249 return (zfs_ioctl(ap->a_vp, ap->a_command, (intptr_t)ap->a_data,
4250 ap->a_fflag, ap->a_cred, NULL, NULL));
4254 zfs_freebsd_read(ap)
4255 struct vop_read_args /* {
4259 struct ucred *a_cred;
4263 return (zfs_read(ap->a_vp, ap->a_uio, ap->a_ioflag, ap->a_cred, NULL));
4267 zfs_freebsd_write(ap)
4268 struct vop_write_args /* {
4272 struct ucred *a_cred;
4276 return (zfs_write(ap->a_vp, ap->a_uio, ap->a_ioflag, ap->a_cred, NULL));
4280 zfs_freebsd_access(ap)
4281 struct vop_access_args /* {
4283 accmode_t a_accmode;
4284 struct ucred *a_cred;
4285 struct thread *a_td;
4292 * ZFS itself only knowns about VREAD, VWRITE, VEXEC and VAPPEND,
4294 accmode = ap->a_accmode & (VREAD|VWRITE|VEXEC|VAPPEND);
4296 error = zfs_access(ap->a_vp, accmode, 0, ap->a_cred, NULL);
4299 * VADMIN has to be handled by vaccess().
4302 accmode = ap->a_accmode & ~(VREAD|VWRITE|VEXEC|VAPPEND);
4304 vnode_t *vp = ap->a_vp;
4305 znode_t *zp = VTOZ(vp);
4306 znode_phys_t *zphys = zp->z_phys;
4308 error = vaccess(vp->v_type, zphys->zp_mode,
4309 zphys->zp_uid, zphys->zp_gid, accmode, ap->a_cred,
4318 zfs_freebsd_lookup(ap)
4319 struct vop_lookup_args /* {
4320 struct vnode *a_dvp;
4321 struct vnode **a_vpp;
4322 struct componentname *a_cnp;
4325 struct componentname *cnp = ap->a_cnp;
4326 char nm[NAME_MAX + 1];
4328 ASSERT(cnp->cn_namelen < sizeof(nm));
4329 strlcpy(nm, cnp->cn_nameptr, MIN(cnp->cn_namelen + 1, sizeof(nm)));
4331 return (zfs_lookup(ap->a_dvp, nm, ap->a_vpp, cnp, cnp->cn_nameiop,
4332 cnp->cn_cred, cnp->cn_thread, 0));
4336 zfs_freebsd_create(ap)
4337 struct vop_create_args /* {
4338 struct vnode *a_dvp;
4339 struct vnode **a_vpp;
4340 struct componentname *a_cnp;
4341 struct vattr *a_vap;
4344 struct componentname *cnp = ap->a_cnp;
4345 vattr_t *vap = ap->a_vap;
4348 ASSERT(cnp->cn_flags & SAVENAME);
4350 vattr_init_mask(vap);
4351 mode = vap->va_mode & ALLPERMS;
4353 return (zfs_create(ap->a_dvp, cnp->cn_nameptr, vap, !EXCL, mode,
4354 ap->a_vpp, cnp->cn_cred, cnp->cn_thread));
4358 zfs_freebsd_remove(ap)
4359 struct vop_remove_args /* {
4360 struct vnode *a_dvp;
4362 struct componentname *a_cnp;
4366 ASSERT(ap->a_cnp->cn_flags & SAVENAME);
4368 return (zfs_remove(ap->a_dvp, ap->a_cnp->cn_nameptr,
4369 ap->a_cnp->cn_cred, NULL, 0));
4373 zfs_freebsd_mkdir(ap)
4374 struct vop_mkdir_args /* {
4375 struct vnode *a_dvp;
4376 struct vnode **a_vpp;
4377 struct componentname *a_cnp;
4378 struct vattr *a_vap;
4381 vattr_t *vap = ap->a_vap;
4383 ASSERT(ap->a_cnp->cn_flags & SAVENAME);
4385 vattr_init_mask(vap);
4387 return (zfs_mkdir(ap->a_dvp, ap->a_cnp->cn_nameptr, vap, ap->a_vpp,
4388 ap->a_cnp->cn_cred, NULL, 0, NULL));
4392 zfs_freebsd_rmdir(ap)
4393 struct vop_rmdir_args /* {
4394 struct vnode *a_dvp;
4396 struct componentname *a_cnp;
4399 struct componentname *cnp = ap->a_cnp;
4401 ASSERT(cnp->cn_flags & SAVENAME);
4403 return (zfs_rmdir(ap->a_dvp, cnp->cn_nameptr, NULL, cnp->cn_cred, NULL, 0));
4407 zfs_freebsd_readdir(ap)
4408 struct vop_readdir_args /* {
4411 struct ucred *a_cred;
4418 return (zfs_readdir(ap->a_vp, ap->a_uio, ap->a_cred, ap->a_eofflag,
4419 ap->a_ncookies, ap->a_cookies));
4423 zfs_freebsd_fsync(ap)
4424 struct vop_fsync_args /* {
4427 struct thread *a_td;
4432 return (zfs_fsync(ap->a_vp, 0, ap->a_td->td_ucred, NULL));
4436 zfs_freebsd_getattr(ap)
4437 struct vop_getattr_args /* {
4439 struct vattr *a_vap;
4440 struct ucred *a_cred;
4441 struct thread *a_td;
4444 vattr_t *vap = ap->a_vap;
4450 xvap.xva_vattr = *vap;
4451 xvap.xva_vattr.va_mask |= AT_XVATTR;
4453 /* Convert chflags into ZFS-type flags. */
4454 /* XXX: what about SF_SETTABLE?. */
4455 XVA_SET_REQ(&xvap, XAT_IMMUTABLE);
4456 XVA_SET_REQ(&xvap, XAT_APPENDONLY);
4457 XVA_SET_REQ(&xvap, XAT_NOUNLINK);
4458 XVA_SET_REQ(&xvap, XAT_NODUMP);
4459 error = zfs_getattr(ap->a_vp, (vattr_t *)&xvap, 0, ap->a_cred, NULL);
4463 /* Convert ZFS xattr into chflags. */
4464 #define FLAG_CHECK(fflag, xflag, xfield) do { \
4465 if (XVA_ISSET_RTN(&xvap, (xflag)) && (xfield) != 0) \
4466 fflags |= (fflag); \
4468 FLAG_CHECK(SF_IMMUTABLE, XAT_IMMUTABLE,
4469 xvap.xva_xoptattrs.xoa_immutable);
4470 FLAG_CHECK(SF_APPEND, XAT_APPENDONLY,
4471 xvap.xva_xoptattrs.xoa_appendonly);
4472 FLAG_CHECK(SF_NOUNLINK, XAT_NOUNLINK,
4473 xvap.xva_xoptattrs.xoa_nounlink);
4474 FLAG_CHECK(UF_NODUMP, XAT_NODUMP,
4475 xvap.xva_xoptattrs.xoa_nodump);
4477 *vap = xvap.xva_vattr;
4478 vap->va_flags = fflags;
4483 zfs_freebsd_setattr(ap)
4484 struct vop_setattr_args /* {
4486 struct vattr *a_vap;
4487 struct ucred *a_cred;
4488 struct thread *a_td;
4491 vnode_t *vp = ap->a_vp;
4492 vattr_t *vap = ap->a_vap;
4493 cred_t *cred = ap->a_cred;
4498 vattr_init_mask(vap);
4499 vap->va_mask &= ~AT_NOSET;
4502 xvap.xva_vattr = *vap;
4504 zflags = VTOZ(vp)->z_phys->zp_flags;
4506 if (vap->va_flags != VNOVAL) {
4507 zfsvfs_t *zfsvfs = VTOZ(vp)->z_zfsvfs;
4510 if (zfsvfs->z_use_fuids == B_FALSE)
4511 return (EOPNOTSUPP);
4513 fflags = vap->va_flags;
4514 if ((fflags & ~(SF_IMMUTABLE|SF_APPEND|SF_NOUNLINK|UF_NODUMP)) != 0)
4515 return (EOPNOTSUPP);
4517 * Unprivileged processes are not permitted to unset system
4518 * flags, or modify flags if any system flags are set.
4519 * Privileged non-jail processes may not modify system flags
4520 * if securelevel > 0 and any existing system flags are set.
4521 * Privileged jail processes behave like privileged non-jail
4522 * processes if the security.jail.chflags_allowed sysctl is
4523 * is non-zero; otherwise, they behave like unprivileged
4526 if (secpolicy_fs_owner(vp->v_mount, cred) == 0 ||
4527 priv_check_cred(cred, PRIV_VFS_SYSFLAGS, 0) == 0) {
4529 (ZFS_IMMUTABLE | ZFS_APPENDONLY | ZFS_NOUNLINK)) {
4530 error = securelevel_gt(cred, 0);
4536 * Callers may only modify the file flags on objects they
4537 * have VADMIN rights for.
4539 if ((error = VOP_ACCESS(vp, VADMIN, cred, curthread)) != 0)
4542 (ZFS_IMMUTABLE | ZFS_APPENDONLY | ZFS_NOUNLINK)) {
4546 (SF_IMMUTABLE | SF_APPEND | SF_NOUNLINK)) {
4551 #define FLAG_CHANGE(fflag, zflag, xflag, xfield) do { \
4552 if (((fflags & (fflag)) && !(zflags & (zflag))) || \
4553 ((zflags & (zflag)) && !(fflags & (fflag)))) { \
4554 XVA_SET_REQ(&xvap, (xflag)); \
4555 (xfield) = ((fflags & (fflag)) != 0); \
4558 /* Convert chflags into ZFS-type flags. */
4559 /* XXX: what about SF_SETTABLE?. */
4560 FLAG_CHANGE(SF_IMMUTABLE, ZFS_IMMUTABLE, XAT_IMMUTABLE,
4561 xvap.xva_xoptattrs.xoa_immutable);
4562 FLAG_CHANGE(SF_APPEND, ZFS_APPENDONLY, XAT_APPENDONLY,
4563 xvap.xva_xoptattrs.xoa_appendonly);
4564 FLAG_CHANGE(SF_NOUNLINK, ZFS_NOUNLINK, XAT_NOUNLINK,
4565 xvap.xva_xoptattrs.xoa_nounlink);
4566 FLAG_CHANGE(UF_NODUMP, ZFS_NODUMP, XAT_NODUMP,
4567 xvap.xva_xoptattrs.xoa_nodump);
4570 return (zfs_setattr(vp, (vattr_t *)&xvap, 0, cred, NULL));
4574 zfs_freebsd_rename(ap)
4575 struct vop_rename_args /* {
4576 struct vnode *a_fdvp;
4577 struct vnode *a_fvp;
4578 struct componentname *a_fcnp;
4579 struct vnode *a_tdvp;
4580 struct vnode *a_tvp;
4581 struct componentname *a_tcnp;
4584 vnode_t *fdvp = ap->a_fdvp;
4585 vnode_t *fvp = ap->a_fvp;
4586 vnode_t *tdvp = ap->a_tdvp;
4587 vnode_t *tvp = ap->a_tvp;
4590 ASSERT(ap->a_fcnp->cn_flags & (SAVENAME|SAVESTART));
4591 ASSERT(ap->a_tcnp->cn_flags & (SAVENAME|SAVESTART));
4593 error = zfs_rename(fdvp, ap->a_fcnp->cn_nameptr, tdvp,
4594 ap->a_tcnp->cn_nameptr, ap->a_fcnp->cn_cred, NULL, 0);
4609 zfs_freebsd_symlink(ap)
4610 struct vop_symlink_args /* {
4611 struct vnode *a_dvp;
4612 struct vnode **a_vpp;
4613 struct componentname *a_cnp;
4614 struct vattr *a_vap;
4618 struct componentname *cnp = ap->a_cnp;
4619 vattr_t *vap = ap->a_vap;
4621 ASSERT(cnp->cn_flags & SAVENAME);
4623 vap->va_type = VLNK; /* FreeBSD: Syscall only sets va_mode. */
4624 vattr_init_mask(vap);
4626 return (zfs_symlink(ap->a_dvp, ap->a_vpp, cnp->cn_nameptr, vap,
4627 ap->a_target, cnp->cn_cred, cnp->cn_thread));
4631 zfs_freebsd_readlink(ap)
4632 struct vop_readlink_args /* {
4635 struct ucred *a_cred;
4639 return (zfs_readlink(ap->a_vp, ap->a_uio, ap->a_cred, NULL));
4643 zfs_freebsd_link(ap)
4644 struct vop_link_args /* {
4645 struct vnode *a_tdvp;
4647 struct componentname *a_cnp;
4650 struct componentname *cnp = ap->a_cnp;
4652 ASSERT(cnp->cn_flags & SAVENAME);
4654 return (zfs_link(ap->a_tdvp, ap->a_vp, cnp->cn_nameptr, cnp->cn_cred, NULL, 0));
4658 zfs_freebsd_inactive(ap)
4659 struct vop_inactive_args /* {
4661 struct thread *a_td;
4664 vnode_t *vp = ap->a_vp;
4666 zfs_inactive(vp, ap->a_td->td_ucred, NULL);
4671 zfs_reclaim_complete(void *arg, int pending)
4674 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
4676 rw_enter(&zfsvfs->z_teardown_inactive_lock, RW_READER);
4677 if (zp->z_dbuf != NULL) {
4678 ZFS_OBJ_HOLD_ENTER(zfsvfs, zp->z_id);
4679 zfs_znode_dmu_fini(zp);
4680 ZFS_OBJ_HOLD_EXIT(zfsvfs, zp->z_id);
4683 rw_exit(&zfsvfs->z_teardown_inactive_lock);
4685 * If the file system is being unmounted, there is a process waiting
4686 * for us, wake it up.
4688 if (zfsvfs->z_unmounted)
4693 zfs_freebsd_reclaim(ap)
4694 struct vop_reclaim_args /* {
4696 struct thread *a_td;
4699 vnode_t *vp = ap->a_vp;
4700 znode_t *zp = VTOZ(vp);
4701 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
4703 rw_enter(&zfsvfs->z_teardown_inactive_lock, RW_READER);
4708 * Destroy the vm object and flush associated pages.
4710 vnode_destroy_vobject(vp);
4712 mutex_enter(&zp->z_lock);
4713 ASSERT(zp->z_phys != NULL);
4715 mutex_exit(&zp->z_lock);
4719 else if (zp->z_dbuf == NULL)
4721 else /* if (!zp->z_unlinked && zp->z_dbuf != NULL) */ {
4724 locked = MUTEX_HELD(ZFS_OBJ_MUTEX(zfsvfs, zp->z_id)) ? 2 :
4725 ZFS_OBJ_HOLD_TRYENTER(zfsvfs, zp->z_id);
4728 * Lock can't be obtained due to deadlock possibility,
4729 * so defer znode destruction.
4731 TASK_INIT(&zp->z_task, 0, zfs_reclaim_complete, zp);
4732 taskqueue_enqueue(taskqueue_thread, &zp->z_task);
4734 zfs_znode_dmu_fini(zp);
4736 ZFS_OBJ_HOLD_EXIT(zfsvfs, zp->z_id);
4742 ASSERT(vp->v_holdcnt >= 1);
4744 rw_exit(&zfsvfs->z_teardown_inactive_lock);
4750 struct vop_fid_args /* {
4756 return (zfs_fid(ap->a_vp, (void *)ap->a_fid, NULL));
4760 zfs_freebsd_pathconf(ap)
4761 struct vop_pathconf_args /* {
4764 register_t *a_retval;
4770 error = zfs_pathconf(ap->a_vp, ap->a_name, &val, curthread->td_ucred, NULL);
4772 *ap->a_retval = val;
4773 else if (error == EOPNOTSUPP)
4774 error = vop_stdpathconf(ap);
4779 zfs_freebsd_fifo_pathconf(ap)
4780 struct vop_pathconf_args /* {
4783 register_t *a_retval;
4787 switch (ap->a_name) {
4788 case _PC_ACL_EXTENDED:
4790 case _PC_ACL_PATH_MAX:
4791 case _PC_MAC_PRESENT:
4792 return (zfs_freebsd_pathconf(ap));
4794 return (fifo_specops.vop_pathconf(ap));
4799 * FreeBSD's extended attributes namespace defines file name prefix for ZFS'
4800 * extended attribute name:
4803 * system freebsd:system:
4804 * user (none, can be used to access ZFS fsattr(5) attributes
4805 * created on Solaris)
4808 zfs_create_attrname(int attrnamespace, const char *name, char *attrname,
4811 const char *namespace, *prefix, *suffix;
4813 /* We don't allow '/' character in attribute name. */
4814 if (strchr(name, '/') != NULL)
4816 /* We don't allow attribute names that start with "freebsd:" string. */
4817 if (strncmp(name, "freebsd:", 8) == 0)
4820 bzero(attrname, size);
4822 switch (attrnamespace) {
4823 case EXTATTR_NAMESPACE_USER:
4825 prefix = "freebsd:";
4826 namespace = EXTATTR_NAMESPACE_USER_STRING;
4830 * This is the default namespace by which we can access all
4831 * attributes created on Solaris.
4833 prefix = namespace = suffix = "";
4836 case EXTATTR_NAMESPACE_SYSTEM:
4837 prefix = "freebsd:";
4838 namespace = EXTATTR_NAMESPACE_SYSTEM_STRING;
4841 case EXTATTR_NAMESPACE_EMPTY:
4845 if (snprintf(attrname, size, "%s%s%s%s", prefix, namespace, suffix,
4847 return (ENAMETOOLONG);
4853 * Vnode operating to retrieve a named extended attribute.
4856 zfs_getextattr(struct vop_getextattr_args *ap)
4859 IN struct vnode *a_vp;
4860 IN int a_attrnamespace;
4861 IN const char *a_name;
4862 INOUT struct uio *a_uio;
4864 IN struct ucred *a_cred;
4865 IN struct thread *a_td;
4869 zfsvfs_t *zfsvfs = VTOZ(ap->a_vp)->z_zfsvfs;
4870 struct thread *td = ap->a_td;
4871 struct nameidata nd;
4874 vnode_t *xvp = NULL, *vp;
4877 error = extattr_check_cred(ap->a_vp, ap->a_attrnamespace,
4878 ap->a_cred, ap->a_td, VREAD);
4882 error = zfs_create_attrname(ap->a_attrnamespace, ap->a_name, attrname,
4889 error = zfs_lookup(ap->a_vp, NULL, &xvp, NULL, 0, ap->a_cred, td,
4897 NDINIT_ATVP(&nd, LOOKUP, NOFOLLOW | MPSAFE, UIO_SYSSPACE, attrname,
4899 error = vn_open_cred(&nd, &flags, 0, 0, ap->a_cred, NULL);
4901 NDFREE(&nd, NDF_ONLY_PNBUF);
4904 if (error == ENOENT)
4909 if (ap->a_size != NULL) {
4910 error = VOP_GETATTR(vp, &va, ap->a_cred);
4912 *ap->a_size = (size_t)va.va_size;
4913 } else if (ap->a_uio != NULL)
4914 error = VOP_READ(vp, ap->a_uio, IO_UNIT | IO_SYNC, ap->a_cred);
4917 vn_close(vp, flags, ap->a_cred, td);
4924 * Vnode operation to remove a named attribute.
4927 zfs_deleteextattr(struct vop_deleteextattr_args *ap)
4930 IN struct vnode *a_vp;
4931 IN int a_attrnamespace;
4932 IN const char *a_name;
4933 IN struct ucred *a_cred;
4934 IN struct thread *a_td;
4938 zfsvfs_t *zfsvfs = VTOZ(ap->a_vp)->z_zfsvfs;
4939 struct thread *td = ap->a_td;
4940 struct nameidata nd;
4943 vnode_t *xvp = NULL, *vp;
4946 error = extattr_check_cred(ap->a_vp, ap->a_attrnamespace,
4947 ap->a_cred, ap->a_td, VWRITE);
4951 error = zfs_create_attrname(ap->a_attrnamespace, ap->a_name, attrname,
4958 error = zfs_lookup(ap->a_vp, NULL, &xvp, NULL, 0, ap->a_cred, td,
4965 NDINIT_ATVP(&nd, DELETE, NOFOLLOW | LOCKPARENT | LOCKLEAF | MPSAFE,
4966 UIO_SYSSPACE, attrname, xvp, td);
4969 NDFREE(&nd, NDF_ONLY_PNBUF);
4972 if (error == ENOENT)
4976 error = VOP_REMOVE(nd.ni_dvp, vp, &nd.ni_cnd);
4979 if (vp == nd.ni_dvp)
4989 * Vnode operation to set a named attribute.
4992 zfs_setextattr(struct vop_setextattr_args *ap)
4995 IN struct vnode *a_vp;
4996 IN int a_attrnamespace;
4997 IN const char *a_name;
4998 INOUT struct uio *a_uio;
4999 IN struct ucred *a_cred;
5000 IN struct thread *a_td;
5004 zfsvfs_t *zfsvfs = VTOZ(ap->a_vp)->z_zfsvfs;
5005 struct thread *td = ap->a_td;
5006 struct nameidata nd;
5009 vnode_t *xvp = NULL, *vp;
5012 error = extattr_check_cred(ap->a_vp, ap->a_attrnamespace,
5013 ap->a_cred, ap->a_td, VWRITE);
5017 error = zfs_create_attrname(ap->a_attrnamespace, ap->a_name, attrname,
5024 error = zfs_lookup(ap->a_vp, NULL, &xvp, NULL, 0, ap->a_cred, td,
5025 LOOKUP_XATTR | CREATE_XATTR_DIR);
5031 flags = FFLAGS(O_WRONLY | O_CREAT);
5032 NDINIT_ATVP(&nd, LOOKUP, NOFOLLOW | MPSAFE, UIO_SYSSPACE, attrname,
5034 error = vn_open_cred(&nd, &flags, 0600, 0, ap->a_cred, NULL);
5036 NDFREE(&nd, NDF_ONLY_PNBUF);
5044 error = VOP_SETATTR(vp, &va, ap->a_cred);
5046 VOP_WRITE(vp, ap->a_uio, IO_UNIT | IO_SYNC, ap->a_cred);
5049 vn_close(vp, flags, ap->a_cred, td);
5056 * Vnode operation to retrieve extended attributes on a vnode.
5059 zfs_listextattr(struct vop_listextattr_args *ap)
5062 IN struct vnode *a_vp;
5063 IN int a_attrnamespace;
5064 INOUT struct uio *a_uio;
5066 IN struct ucred *a_cred;
5067 IN struct thread *a_td;
5071 zfsvfs_t *zfsvfs = VTOZ(ap->a_vp)->z_zfsvfs;
5072 struct thread *td = ap->a_td;
5073 struct nameidata nd;
5074 char attrprefix[16];
5075 u_char dirbuf[sizeof(struct dirent)];
5078 struct uio auio, *uio = ap->a_uio;
5079 size_t *sizep = ap->a_size;
5081 vnode_t *xvp = NULL, *vp;
5082 int done, error, eof, pos;
5084 error = extattr_check_cred(ap->a_vp, ap->a_attrnamespace,
5085 ap->a_cred, ap->a_td, VREAD);
5089 error = zfs_create_attrname(ap->a_attrnamespace, "", attrprefix,
5090 sizeof(attrprefix));
5093 plen = strlen(attrprefix);
5100 error = zfs_lookup(ap->a_vp, NULL, &xvp, NULL, 0, ap->a_cred, td,
5105 * ENOATTR means that the EA directory does not yet exist,
5106 * i.e. there are no extended attributes there.
5108 if (error == ENOATTR)
5113 NDINIT_ATVP(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | LOCKSHARED | MPSAFE,
5114 UIO_SYSSPACE, ".", xvp, td);
5117 NDFREE(&nd, NDF_ONLY_PNBUF);
5123 auio.uio_iov = &aiov;
5124 auio.uio_iovcnt = 1;
5125 auio.uio_segflg = UIO_SYSSPACE;
5127 auio.uio_rw = UIO_READ;
5128 auio.uio_offset = 0;
5133 aiov.iov_base = (void *)dirbuf;
5134 aiov.iov_len = sizeof(dirbuf);
5135 auio.uio_resid = sizeof(dirbuf);
5136 error = VOP_READDIR(vp, &auio, ap->a_cred, &eof, NULL, NULL);
5137 done = sizeof(dirbuf) - auio.uio_resid;
5140 for (pos = 0; pos < done;) {
5141 dp = (struct dirent *)(dirbuf + pos);
5142 pos += dp->d_reclen;
5144 * XXX: Temporarily we also accept DT_UNKNOWN, as this
5145 * is what we get when attribute was created on Solaris.
5147 if (dp->d_type != DT_REG && dp->d_type != DT_UNKNOWN)
5149 if (plen == 0 && strncmp(dp->d_name, "freebsd:", 8) == 0)
5151 else if (strncmp(dp->d_name, attrprefix, plen) != 0)
5153 nlen = dp->d_namlen - plen;
5156 else if (uio != NULL) {
5158 * Format of extattr name entry is one byte for
5159 * length and the rest for name.
5161 error = uiomove(&nlen, 1, uio->uio_rw, uio);
5163 error = uiomove(dp->d_name + plen, nlen,
5170 } while (!eof && error == 0);
5179 zfs_freebsd_getacl(ap)
5180 struct vop_getacl_args /* {
5189 vsecattr_t vsecattr;
5191 if (ap->a_type != ACL_TYPE_NFS4)
5194 vsecattr.vsa_mask = VSA_ACE | VSA_ACECNT;
5195 if (error = zfs_getsecattr(ap->a_vp, &vsecattr, 0, ap->a_cred, NULL))
5198 error = acl_from_aces(ap->a_aclp, vsecattr.vsa_aclentp, vsecattr.vsa_aclcnt);
5199 if (vsecattr.vsa_aclentp != NULL)
5200 kmem_free(vsecattr.vsa_aclentp, vsecattr.vsa_aclentsz);
5206 zfs_freebsd_setacl(ap)
5207 struct vop_setacl_args /* {
5216 vsecattr_t vsecattr;
5217 int aclbsize; /* size of acl list in bytes */
5220 if (ap->a_type != ACL_TYPE_NFS4)
5223 if (ap->a_aclp->acl_cnt < 1 || ap->a_aclp->acl_cnt > MAX_ACL_ENTRIES)
5227 * With NFSv4 ACLs, chmod(2) may need to add additional entries,
5228 * splitting every entry into two and appending "canonical six"
5229 * entries at the end. Don't allow for setting an ACL that would
5230 * cause chmod(2) to run out of ACL entries.
5232 if (ap->a_aclp->acl_cnt * 2 + 6 > ACL_MAX_ENTRIES)
5235 error = acl_nfs4_check(ap->a_aclp, ap->a_vp->v_type == VDIR);
5239 vsecattr.vsa_mask = VSA_ACE;
5240 aclbsize = ap->a_aclp->acl_cnt * sizeof(ace_t);
5241 vsecattr.vsa_aclentp = kmem_alloc(aclbsize, KM_SLEEP);
5242 aaclp = vsecattr.vsa_aclentp;
5243 vsecattr.vsa_aclentsz = aclbsize;
5245 aces_from_acl(vsecattr.vsa_aclentp, &vsecattr.vsa_aclcnt, ap->a_aclp);
5246 error = zfs_setsecattr(ap->a_vp, &vsecattr, 0, ap->a_cred, NULL);
5247 kmem_free(aaclp, aclbsize);
5253 zfs_freebsd_aclcheck(ap)
5254 struct vop_aclcheck_args /* {
5263 return (EOPNOTSUPP);
5266 struct vop_vector zfs_vnodeops;
5267 struct vop_vector zfs_fifoops;
5268 struct vop_vector zfs_shareops;
5270 struct vop_vector zfs_vnodeops = {
5271 .vop_default = &default_vnodeops,
5272 .vop_inactive = zfs_freebsd_inactive,
5273 .vop_reclaim = zfs_freebsd_reclaim,
5274 .vop_access = zfs_freebsd_access,
5275 #ifdef FREEBSD_NAMECACHE
5276 .vop_lookup = vfs_cache_lookup,
5277 .vop_cachedlookup = zfs_freebsd_lookup,
5279 .vop_lookup = zfs_freebsd_lookup,
5281 .vop_getattr = zfs_freebsd_getattr,
5282 .vop_setattr = zfs_freebsd_setattr,
5283 .vop_create = zfs_freebsd_create,
5284 .vop_mknod = zfs_freebsd_create,
5285 .vop_mkdir = zfs_freebsd_mkdir,
5286 .vop_readdir = zfs_freebsd_readdir,
5287 .vop_fsync = zfs_freebsd_fsync,
5288 .vop_open = zfs_freebsd_open,
5289 .vop_close = zfs_freebsd_close,
5290 .vop_rmdir = zfs_freebsd_rmdir,
5291 .vop_ioctl = zfs_freebsd_ioctl,
5292 .vop_link = zfs_freebsd_link,
5293 .vop_symlink = zfs_freebsd_symlink,
5294 .vop_readlink = zfs_freebsd_readlink,
5295 .vop_read = zfs_freebsd_read,
5296 .vop_write = zfs_freebsd_write,
5297 .vop_remove = zfs_freebsd_remove,
5298 .vop_rename = zfs_freebsd_rename,
5299 .vop_pathconf = zfs_freebsd_pathconf,
5300 .vop_bmap = VOP_EOPNOTSUPP,
5301 .vop_fid = zfs_freebsd_fid,
5302 .vop_getextattr = zfs_getextattr,
5303 .vop_deleteextattr = zfs_deleteextattr,
5304 .vop_setextattr = zfs_setextattr,
5305 .vop_listextattr = zfs_listextattr,
5306 .vop_getacl = zfs_freebsd_getacl,
5307 .vop_setacl = zfs_freebsd_setacl,
5308 .vop_aclcheck = zfs_freebsd_aclcheck,
5311 struct vop_vector zfs_fifoops = {
5312 .vop_default = &fifo_specops,
5313 .vop_fsync = zfs_freebsd_fsync,
5314 .vop_access = zfs_freebsd_access,
5315 .vop_getattr = zfs_freebsd_getattr,
5316 .vop_inactive = zfs_freebsd_inactive,
5317 .vop_read = VOP_PANIC,
5318 .vop_reclaim = zfs_freebsd_reclaim,
5319 .vop_setattr = zfs_freebsd_setattr,
5320 .vop_write = VOP_PANIC,
5321 .vop_pathconf = zfs_freebsd_fifo_pathconf,
5322 .vop_fid = zfs_freebsd_fid,
5323 .vop_getacl = zfs_freebsd_getacl,
5324 .vop_setacl = zfs_freebsd_setacl,
5325 .vop_aclcheck = zfs_freebsd_aclcheck,
5329 * special share hidden files vnode operations template
5331 struct vop_vector zfs_shareops = {
5332 .vop_default = &default_vnodeops,
5333 .vop_access = zfs_freebsd_access,
5334 .vop_inactive = zfs_freebsd_inactive,
5335 .vop_reclaim = zfs_freebsd_reclaim,
5336 .vop_fid = zfs_freebsd_fid,
5337 .vop_pathconf = zfs_freebsd_pathconf,