4 * The contents of this file are subject to the terms of the
5 * Common Development and Distribution License (the "License").
6 * You may not use this file except in compliance with the License.
8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9 * or http://www.opensolaris.org/os/licensing.
10 * See the License for the specific language governing permissions
11 * and limitations under the License.
13 * When distributing Covered Code, include this CDDL HEADER in each
14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15 * If applicable, add the following below this CDDL HEADER, with the
16 * fields enclosed by brackets "[]" replaced with your own identifying
17 * information: Portions Copyright [yyyy] [name of copyright owner]
22 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
25 /* Portions Copyright 2007 Jeremy Teo */
27 #include <sys/types.h>
28 #include <sys/param.h>
30 #include <sys/systm.h>
31 #include <sys/sysmacros.h>
32 #include <sys/resource.h>
34 #include <sys/vnode.h>
38 #include <sys/taskq.h>
40 #include <sys/atomic.h>
41 #include <sys/namei.h>
43 #include <sys/cmn_err.h>
44 #include <sys/errno.h>
45 #include <sys/unistd.h>
46 #include <sys/zfs_dir.h>
47 #include <sys/zfs_ioctl.h>
48 #include <sys/fs/zfs.h>
54 #include <sys/dirent.h>
55 #include <sys/policy.h>
56 #include <sys/sunddi.h>
57 #include <sys/filio.h>
59 #include <sys/zfs_ctldir.h>
60 #include <sys/zfs_fuid.h>
62 #include <sys/zfs_rlock.h>
63 #include <sys/extdirent.h>
64 #include <sys/kidmap.h>
67 #include <sys/sf_buf.h>
68 #include <sys/sched.h>
74 * Each vnode op performs some logical unit of work. To do this, the ZPL must
75 * properly lock its in-core state, create a DMU transaction, do the work,
76 * record this work in the intent log (ZIL), commit the DMU transaction,
77 * and wait for the intent log to commit if it is a synchronous operation.
78 * Moreover, the vnode ops must work in both normal and log replay context.
79 * The ordering of events is important to avoid deadlocks and references
80 * to freed memory. The example below illustrates the following Big Rules:
82 * (1) A check must be made in each zfs thread for a mounted file system.
83 * This is done avoiding races using ZFS_ENTER(zfsvfs).
84 * A ZFS_EXIT(zfsvfs) is needed before all returns. Any znodes
85 * must be checked with ZFS_VERIFY_ZP(zp). Both of these macros
86 * can return EIO from the calling function.
88 * (2) VN_RELE() should always be the last thing except for zil_commit()
89 * (if necessary) and ZFS_EXIT(). This is for 3 reasons:
90 * First, if it's the last reference, the vnode/znode
91 * can be freed, so the zp may point to freed memory. Second, the last
92 * reference will call zfs_zinactive(), which may induce a lot of work --
93 * pushing cached pages (which acquires range locks) and syncing out
94 * cached atime changes. Third, zfs_zinactive() may require a new tx,
95 * which could deadlock the system if you were already holding one.
96 * If you must call VN_RELE() within a tx then use VN_RELE_ASYNC().
98 * (3) All range locks must be grabbed before calling dmu_tx_assign(),
99 * as they can span dmu_tx_assign() calls.
101 * (4) Always pass TXG_NOWAIT as the second argument to dmu_tx_assign().
102 * This is critical because we don't want to block while holding locks.
103 * Note, in particular, that if a lock is sometimes acquired before
104 * the tx assigns, and sometimes after (e.g. z_lock), then failing to
105 * use a non-blocking assign can deadlock the system. The scenario:
107 * Thread A has grabbed a lock before calling dmu_tx_assign().
108 * Thread B is in an already-assigned tx, and blocks for this lock.
109 * Thread A calls dmu_tx_assign(TXG_WAIT) and blocks in txg_wait_open()
110 * forever, because the previous txg can't quiesce until B's tx commits.
112 * If dmu_tx_assign() returns ERESTART and zfsvfs->z_assign is TXG_NOWAIT,
113 * then drop all locks, call dmu_tx_wait(), and try again.
115 * (5) If the operation succeeded, generate the intent log entry for it
116 * before dropping locks. This ensures that the ordering of events
117 * in the intent log matches the order in which they actually occurred.
118 * During ZIL replay the zfs_log_* functions will update the sequence
119 * number to indicate the zil transaction has replayed.
121 * (6) At the end of each vnode op, the DMU tx must always commit,
122 * regardless of whether there were any errors.
124 * (7) After dropping all locks, invoke zil_commit(zilog, seq, foid)
125 * to ensure that synchronous semantics are provided when necessary.
127 * In general, this is how things should be ordered in each vnode op:
129 * ZFS_ENTER(zfsvfs); // exit if unmounted
131 * zfs_dirent_lock(&dl, ...) // lock directory entry (may VN_HOLD())
132 * rw_enter(...); // grab any other locks you need
133 * tx = dmu_tx_create(...); // get DMU tx
134 * dmu_tx_hold_*(); // hold each object you might modify
135 * error = dmu_tx_assign(tx, TXG_NOWAIT); // try to assign
137 * rw_exit(...); // drop locks
138 * zfs_dirent_unlock(dl); // unlock directory entry
139 * VN_RELE(...); // release held vnodes
140 * if (error == ERESTART) {
145 * dmu_tx_abort(tx); // abort DMU tx
146 * ZFS_EXIT(zfsvfs); // finished in zfs
147 * return (error); // really out of space
149 * error = do_real_work(); // do whatever this VOP does
151 * zfs_log_*(...); // on success, make ZIL entry
152 * dmu_tx_commit(tx); // commit DMU tx -- error or not
153 * rw_exit(...); // drop locks
154 * zfs_dirent_unlock(dl); // unlock directory entry
155 * VN_RELE(...); // release held vnodes
156 * zil_commit(zilog, seq, foid); // synchronous when necessary
157 * ZFS_EXIT(zfsvfs); // finished in zfs
158 * return (error); // done, report error
163 zfs_open(vnode_t **vpp, int flag, cred_t *cr, caller_context_t *ct)
165 znode_t *zp = VTOZ(*vpp);
166 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
171 if ((flag & FWRITE) && (zp->z_phys->zp_flags & ZFS_APPENDONLY) &&
172 ((flag & FAPPEND) == 0)) {
177 if (!zfs_has_ctldir(zp) && zp->z_zfsvfs->z_vscan &&
178 ZTOV(zp)->v_type == VREG &&
179 !(zp->z_phys->zp_flags & ZFS_AV_QUARANTINED) &&
180 zp->z_phys->zp_size > 0) {
181 if (fs_vscan(*vpp, cr, 0) != 0) {
187 /* Keep a count of the synchronous opens in the znode */
188 if (flag & (FSYNC | FDSYNC))
189 atomic_inc_32(&zp->z_sync_cnt);
197 zfs_close(vnode_t *vp, int flag, int count, offset_t offset, cred_t *cr,
198 caller_context_t *ct)
200 znode_t *zp = VTOZ(vp);
201 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
206 /* Decrement the synchronous opens in the znode */
207 if ((flag & (FSYNC | FDSYNC)) && (count == 1))
208 atomic_dec_32(&zp->z_sync_cnt);
211 * Clean up any locks held by this process on the vp.
213 cleanlocks(vp, ddi_get_pid(), 0);
214 cleanshares(vp, ddi_get_pid());
216 if (!zfs_has_ctldir(zp) && zp->z_zfsvfs->z_vscan &&
217 ZTOV(zp)->v_type == VREG &&
218 !(zp->z_phys->zp_flags & ZFS_AV_QUARANTINED) &&
219 zp->z_phys->zp_size > 0)
220 VERIFY(fs_vscan(vp, cr, 1) == 0);
227 * Lseek support for finding holes (cmd == _FIO_SEEK_HOLE) and
228 * data (cmd == _FIO_SEEK_DATA). "off" is an in/out parameter.
231 zfs_holey(vnode_t *vp, u_long cmd, offset_t *off)
233 znode_t *zp = VTOZ(vp);
234 uint64_t noff = (uint64_t)*off; /* new offset */
239 file_sz = zp->z_phys->zp_size;
240 if (noff >= file_sz) {
244 if (cmd == _FIO_SEEK_HOLE)
249 error = dmu_offset_next(zp->z_zfsvfs->z_os, zp->z_id, hole, &noff);
252 if ((error == ESRCH) || (noff > file_sz)) {
254 * Handle the virtual hole at the end of file.
271 zfs_ioctl(vnode_t *vp, u_long com, intptr_t data, int flag, cred_t *cred,
272 int *rvalp, caller_context_t *ct)
284 * The following two ioctls are used by bfu. Faking out,
285 * necessary to avoid bfu errors.
293 if (ddi_copyin((void *)data, &off, sizeof (off), flag))
297 zfsvfs = zp->z_zfsvfs;
301 /* offset parameter is in/out */
302 error = zfs_holey(vp, com, &off);
306 if (ddi_copyout(&off, (void *)data, sizeof (off), flag))
314 page_lookup(vnode_t *vp, int64_t start, int64_t off, int64_t nbytes)
320 VM_OBJECT_LOCK_ASSERT(obj, MA_OWNED);
323 if ((pp = vm_page_lookup(obj, OFF_TO_IDX(start))) != NULL &&
324 vm_page_is_valid(pp, (vm_offset_t)off, nbytes)) {
325 if (vm_page_sleep_if_busy(pp, FALSE, "zfsmwb"))
328 vm_page_lock_queues();
330 vm_page_unlock_queues();
332 if (__predict_false(obj->cache != NULL)) {
333 vm_page_cache_free(obj, OFF_TO_IDX(start),
334 OFF_TO_IDX(start) + 1);
344 page_unlock(vm_page_t pp)
351 zfs_map_page(vm_page_t pp, struct sf_buf **sfp)
355 *sfp = sf_buf_alloc(pp, SFB_CPUPRIVATE);
356 return ((caddr_t)sf_buf_kva(*sfp));
360 zfs_unmap_page(struct sf_buf *sf)
369 * When a file is memory mapped, we must keep the IO data synchronized
370 * between the DMU cache and the memory mapped pages. What this means:
372 * On Write: If we find a memory mapped page, we write to *both*
373 * the page and the dmu buffer.
377 update_pages(vnode_t *vp, int64_t start, int len, objset_t *os, uint64_t oid,
378 int segflg, dmu_tx_t *tx)
384 ASSERT(vp->v_mount != NULL);
388 off = start & PAGEOFFSET;
390 for (start &= PAGEMASK; len > 0; start += PAGESIZE) {
392 uint64_t nbytes = MIN(PAGESIZE - off, len);
394 if ((pp = page_lookup(vp, start, off, nbytes)) != NULL) {
397 VM_OBJECT_UNLOCK(obj);
398 va = zfs_map_page(pp, &sf);
399 if (segflg == UIO_NOCOPY) {
400 (void) dmu_write(os, oid, start+off, nbytes,
403 (void) dmu_read(os, oid, start+off, nbytes,
404 va+off, DMU_READ_PREFETCH);;
414 VM_OBJECT_UNLOCK(obj);
418 * When a file is memory mapped, we must keep the IO data synchronized
419 * between the DMU cache and the memory mapped pages. What this means:
421 * On Read: We "read" preferentially from memory mapped pages,
422 * else we default from the dmu buffer.
424 * NOTE: We will always "break up" the IO into PAGESIZE uiomoves when
425 * the file is memory mapped.
428 mappedread(vnode_t *vp, int nbytes, uio_t *uio)
430 znode_t *zp = VTOZ(vp);
431 objset_t *os = zp->z_zfsvfs->z_os;
441 ASSERT(vp->v_mount != NULL);
445 start = uio->uio_loffset;
446 off = start & PAGEOFFSET;
449 for (start &= PAGEMASK; len > 0; start += PAGESIZE) {
450 uint64_t bytes = MIN(PAGESIZE - off, len);
453 if ((m = vm_page_lookup(obj, OFF_TO_IDX(start))) != NULL &&
454 vm_page_is_valid(m, (vm_offset_t)off, bytes)) {
455 if (vm_page_sleep_if_busy(m, FALSE, "zfsmrb"))
458 VM_OBJECT_UNLOCK(obj);
460 error = dmu_read_uio(os, zp->z_id, uio,
466 sf = sf_buf_alloc(m, SFB_CPUPRIVATE);
467 va = (caddr_t)sf_buf_kva(sf);
468 error = uiomove(va + off, bytes, UIO_READ, uio);
474 } else if (m != NULL && uio->uio_segflg == UIO_NOCOPY) {
476 * The code below is here to make sendfile(2) work
477 * correctly with ZFS. As pointed out by ups@
478 * sendfile(2) should be changed to use VOP_GETPAGES(),
479 * but it pessimize performance of sendfile/UFS, that's
480 * why I handle this special case in ZFS code.
482 if (vm_page_sleep_if_busy(m, FALSE, "zfsmrb"))
485 VM_OBJECT_UNLOCK(obj);
487 error = dmu_read_uio(os, zp->z_id, uio,
493 sf = sf_buf_alloc(m, SFB_CPUPRIVATE);
494 va = (caddr_t)sf_buf_kva(sf);
495 error = dmu_read(os, zp->z_id, start + off,
496 bytes, (void *)(va + off),
504 uio->uio_resid -= bytes;
513 VM_OBJECT_UNLOCK(obj);
514 if (error == 0 && dirbytes > 0)
515 error = dmu_read_uio(os, zp->z_id, uio, dirbytes);
519 offset_t zfs_read_chunk_size = 1024 * 1024; /* Tunable */
522 * Read bytes from specified file into supplied buffer.
524 * IN: vp - vnode of file to be read from.
525 * uio - structure supplying read location, range info,
527 * ioflag - SYNC flags; used to provide FRSYNC semantics.
528 * cr - credentials of caller.
529 * ct - caller context
531 * OUT: uio - updated offset and range, buffer filled.
533 * RETURN: 0 if success
534 * error code if failure
537 * vp - atime updated if byte count > 0
541 zfs_read(vnode_t *vp, uio_t *uio, int ioflag, cred_t *cr, caller_context_t *ct)
543 znode_t *zp = VTOZ(vp);
544 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
554 if (zp->z_phys->zp_flags & ZFS_AV_QUARANTINED) {
560 * Validate file offset
562 if (uio->uio_loffset < (offset_t)0) {
568 * Fasttrack empty reads
570 if (uio->uio_resid == 0) {
576 * Check for mandatory locks
578 if (MANDMODE((mode_t)zp->z_phys->zp_mode)) {
579 if (error = chklock(vp, FREAD,
580 uio->uio_loffset, uio->uio_resid, uio->uio_fmode, ct)) {
587 * If we're in FRSYNC mode, sync out this znode before reading it.
590 zil_commit(zfsvfs->z_log, zp->z_last_itx, zp->z_id);
593 * Lock the range against changes.
595 rl = zfs_range_lock(zp, uio->uio_loffset, uio->uio_resid, RL_READER);
598 * If we are reading past end-of-file we can skip
599 * to the end; but we might still need to set atime.
601 if (uio->uio_loffset >= zp->z_phys->zp_size) {
606 ASSERT(uio->uio_loffset < zp->z_phys->zp_size);
607 n = MIN(uio->uio_resid, zp->z_phys->zp_size - uio->uio_loffset);
610 nbytes = MIN(n, zfs_read_chunk_size -
611 P2PHASE(uio->uio_loffset, zfs_read_chunk_size));
613 if (vn_has_cached_data(vp))
614 error = mappedread(vp, nbytes, uio);
616 error = dmu_read_uio(os, zp->z_id, uio, nbytes);
618 /* convert checksum errors into IO errors */
628 zfs_range_unlock(rl);
630 ZFS_ACCESSTIME_STAMP(zfsvfs, zp);
636 * Fault in the pages of the first n bytes specified by the uio structure.
637 * 1 byte in each page is touched and the uio struct is unmodified.
638 * Any error will exit this routine as this is only a best
639 * attempt to get the pages resident. This is a copy of ufs_trans_touch().
642 zfs_prefault_write(ssize_t n, struct uio *uio)
648 if (uio->uio_segflg != UIO_USERSPACE)
654 cnt = MIN(iov->iov_len, n);
656 /* empty iov entry */
662 * touch each page in this segment.
668 incr = MIN(cnt, PAGESIZE);
673 * touch the last byte in case it straddles a page.
683 * Write the bytes to a file.
685 * IN: vp - vnode of file to be written to.
686 * uio - structure supplying write location, range info,
688 * ioflag - IO_APPEND flag set if in append mode.
689 * cr - credentials of caller.
690 * ct - caller context (NFS/CIFS fem monitor only)
692 * OUT: uio - updated offset and range.
694 * RETURN: 0 if success
695 * error code if failure
698 * vp - ctime|mtime updated if byte count > 0
702 zfs_write(vnode_t *vp, uio_t *uio, int ioflag, cred_t *cr, caller_context_t *ct)
704 znode_t *zp = VTOZ(vp);
705 rlim64_t limit = MAXOFFSET_T;
706 ssize_t start_resid = uio->uio_resid;
710 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
715 int max_blksz = zfsvfs->z_max_blksz;
721 * Fasttrack empty write
727 if (limit == RLIM64_INFINITY || limit > MAXOFFSET_T)
734 * If immutable or not appending then return EPERM
736 pflags = zp->z_phys->zp_flags;
737 if ((pflags & (ZFS_IMMUTABLE | ZFS_READONLY)) ||
738 ((pflags & ZFS_APPENDONLY) && !(ioflag & FAPPEND) &&
739 (uio->uio_loffset < zp->z_phys->zp_size))) {
744 zilog = zfsvfs->z_log;
747 * Pre-fault the pages to ensure slow (eg NFS) pages
750 zfs_prefault_write(n, uio);
753 * If in append mode, set the io offset pointer to eof.
755 if (ioflag & IO_APPEND) {
757 * Range lock for a file append:
758 * The value for the start of range will be determined by
759 * zfs_range_lock() (to guarantee append semantics).
760 * If this write will cause the block size to increase,
761 * zfs_range_lock() will lock the entire file, so we must
762 * later reduce the range after we grow the block size.
764 rl = zfs_range_lock(zp, 0, n, RL_APPEND);
765 if (rl->r_len == UINT64_MAX) {
766 /* overlocked, zp_size can't change */
767 woff = uio->uio_loffset = zp->z_phys->zp_size;
769 woff = uio->uio_loffset = rl->r_off;
772 woff = uio->uio_loffset;
774 * Validate file offset
782 * If we need to grow the block size then zfs_range_lock()
783 * will lock a wider range than we request here.
784 * Later after growing the block size we reduce the range.
786 rl = zfs_range_lock(zp, woff, n, RL_WRITER);
790 zfs_range_unlock(rl);
795 if ((woff + n) > limit || woff > (limit - n))
799 * Check for mandatory locks
801 if (MANDMODE((mode_t)zp->z_phys->zp_mode) &&
802 (error = chklock(vp, FWRITE, woff, n, uio->uio_fmode, ct)) != 0) {
803 zfs_range_unlock(rl);
807 end_size = MAX(zp->z_phys->zp_size, woff + n);
810 * Write the file in reasonable size chunks. Each chunk is written
811 * in a separate transaction; this keeps the intent log records small
812 * and allows us to do more fine-grained space accounting.
816 woff = uio->uio_loffset;
819 if (zfs_usergroup_overquota(zfsvfs,
820 B_FALSE, zp->z_phys->zp_uid) ||
821 zfs_usergroup_overquota(zfsvfs,
822 B_TRUE, zp->z_phys->zp_gid)) {
824 dmu_return_arcbuf(abuf);
830 * If dmu_assign_arcbuf() is expected to execute with minimum
831 * overhead loan an arc buffer and copy user data to it before
832 * we enter a txg. This avoids holding a txg forever while we
833 * pagefault on a hanging NFS server mapping.
835 if (abuf == NULL && n >= max_blksz &&
836 woff >= zp->z_phys->zp_size &&
837 P2PHASE(woff, max_blksz) == 0 &&
838 zp->z_blksz == max_blksz) {
841 abuf = dmu_request_arcbuf(zp->z_dbuf, max_blksz);
842 ASSERT(abuf != NULL);
843 ASSERT(arc_buf_size(abuf) == max_blksz);
844 if (error = uiocopy(abuf->b_data, max_blksz,
845 UIO_WRITE, uio, &cbytes)) {
846 dmu_return_arcbuf(abuf);
849 ASSERT(cbytes == max_blksz);
853 * Start a transaction.
855 tx = dmu_tx_create(zfsvfs->z_os);
856 dmu_tx_hold_bonus(tx, zp->z_id);
857 dmu_tx_hold_write(tx, zp->z_id, woff, MIN(n, max_blksz));
858 error = dmu_tx_assign(tx, TXG_NOWAIT);
860 if (error == ERESTART) {
867 dmu_return_arcbuf(abuf);
872 * If zfs_range_lock() over-locked we grow the blocksize
873 * and then reduce the lock range. This will only happen
874 * on the first iteration since zfs_range_reduce() will
875 * shrink down r_len to the appropriate size.
877 if (rl->r_len == UINT64_MAX) {
880 if (zp->z_blksz > max_blksz) {
881 ASSERT(!ISP2(zp->z_blksz));
882 new_blksz = MIN(end_size, SPA_MAXBLOCKSIZE);
884 new_blksz = MIN(end_size, max_blksz);
886 zfs_grow_blocksize(zp, new_blksz, tx);
887 zfs_range_reduce(rl, woff, n);
891 * XXX - should we really limit each write to z_max_blksz?
892 * Perhaps we should use SPA_MAXBLOCKSIZE chunks?
894 nbytes = MIN(n, max_blksz - P2PHASE(woff, max_blksz));
896 if (woff + nbytes > zp->z_phys->zp_size)
897 vnode_pager_setsize(vp, woff + nbytes);
900 tx_bytes = uio->uio_resid;
901 error = dmu_write_uio(zfsvfs->z_os, zp->z_id, uio,
903 tx_bytes -= uio->uio_resid;
906 ASSERT(tx_bytes == max_blksz);
907 dmu_assign_arcbuf(zp->z_dbuf, woff, abuf, tx);
908 ASSERT(tx_bytes <= uio->uio_resid);
909 uioskip(uio, tx_bytes);
913 * XXXPJD: There are some cases (triggered by fsx) where
914 * vn_has_cached_data(vp) returns false when it should
915 * return true. This should be investigated.
918 if (tx_bytes && vn_has_cached_data(vp))
920 if (tx_bytes && vp->v_object != NULL)
923 update_pages(vp, woff, tx_bytes, zfsvfs->z_os,
924 zp->z_id, uio->uio_segflg, tx);
928 * If we made no progress, we're done. If we made even
929 * partial progress, update the znode and ZIL accordingly.
938 * Clear Set-UID/Set-GID bits on successful write if not
939 * privileged and at least one of the excute bits is set.
941 * It would be nice to to this after all writes have
942 * been done, but that would still expose the ISUID/ISGID
943 * to another app after the partial write is committed.
945 * Note: we don't call zfs_fuid_map_id() here because
946 * user 0 is not an ephemeral uid.
948 mutex_enter(&zp->z_acl_lock);
949 if ((zp->z_phys->zp_mode & (S_IXUSR | (S_IXUSR >> 3) |
950 (S_IXUSR >> 6))) != 0 &&
951 (zp->z_phys->zp_mode & (S_ISUID | S_ISGID)) != 0 &&
952 secpolicy_vnode_setid_retain(vp, cr,
953 (zp->z_phys->zp_mode & S_ISUID) != 0 &&
954 zp->z_phys->zp_uid == 0) != 0) {
955 zp->z_phys->zp_mode &= ~(S_ISUID | S_ISGID);
957 mutex_exit(&zp->z_acl_lock);
960 * Update time stamp. NOTE: This marks the bonus buffer as
961 * dirty, so we don't have to do it again for zp_size.
963 zfs_time_stamper(zp, CONTENT_MODIFIED, tx);
966 * Update the file size (zp_size) if it has changed;
967 * account for possible concurrent updates.
969 while ((end_size = zp->z_phys->zp_size) < uio->uio_loffset)
970 (void) atomic_cas_64(&zp->z_phys->zp_size, end_size,
972 zfs_log_write(zilog, tx, TX_WRITE, zp, woff, tx_bytes, ioflag);
977 ASSERT(tx_bytes == nbytes);
981 zfs_range_unlock(rl);
984 * If we're in replay mode, or we made no progress, return error.
985 * Otherwise, it's at least a partial write, so it's successful.
987 if (zfsvfs->z_replay || uio->uio_resid == start_resid) {
992 if (ioflag & (FSYNC | FDSYNC))
993 zil_commit(zilog, zp->z_last_itx, zp->z_id);
1000 zfs_get_done(dmu_buf_t *db, void *vzgd)
1002 zgd_t *zgd = (zgd_t *)vzgd;
1003 rl_t *rl = zgd->zgd_rl;
1004 vnode_t *vp = ZTOV(rl->r_zp);
1005 objset_t *os = rl->r_zp->z_zfsvfs->z_os;
1008 vfslocked = VFS_LOCK_GIANT(vp->v_vfsp);
1009 dmu_buf_rele(db, vzgd);
1010 zfs_range_unlock(rl);
1012 * Release the vnode asynchronously as we currently have the
1013 * txg stopped from syncing.
1015 VN_RELE_ASYNC(vp, dsl_pool_vnrele_taskq(dmu_objset_pool(os)));
1016 zil_add_block(zgd->zgd_zilog, zgd->zgd_bp);
1017 kmem_free(zgd, sizeof (zgd_t));
1018 VFS_UNLOCK_GIANT(vfslocked);
1022 * Get data to generate a TX_WRITE intent log record.
1025 zfs_get_data(void *arg, lr_write_t *lr, char *buf, zio_t *zio)
1027 zfsvfs_t *zfsvfs = arg;
1028 objset_t *os = zfsvfs->z_os;
1030 uint64_t off = lr->lr_offset;
1034 int dlen = lr->lr_length; /* length of user data */
1041 * Nothing to do if the file has been removed
1043 if (zfs_zget(zfsvfs, lr->lr_foid, &zp) != 0)
1045 if (zp->z_unlinked) {
1047 * Release the vnode asynchronously as we currently have the
1048 * txg stopped from syncing.
1050 VN_RELE_ASYNC(ZTOV(zp),
1051 dsl_pool_vnrele_taskq(dmu_objset_pool(os)));
1056 * Write records come in two flavors: immediate and indirect.
1057 * For small writes it's cheaper to store the data with the
1058 * log record (immediate); for large writes it's cheaper to
1059 * sync the data and get a pointer to it (indirect) so that
1060 * we don't have to write the data twice.
1062 if (buf != NULL) { /* immediate write */
1063 rl = zfs_range_lock(zp, off, dlen, RL_READER);
1064 /* test for truncation needs to be done while range locked */
1065 if (off >= zp->z_phys->zp_size) {
1069 VERIFY(0 == dmu_read(os, lr->lr_foid, off, dlen, buf,
1070 DMU_READ_NO_PREFETCH));
1071 } else { /* indirect write */
1072 uint64_t boff; /* block starting offset */
1075 * Have to lock the whole block to ensure when it's
1076 * written out and it's checksum is being calculated
1077 * that no one can change the data. We need to re-check
1078 * blocksize after we get the lock in case it's changed!
1081 if (ISP2(zp->z_blksz)) {
1082 boff = P2ALIGN_TYPED(off, zp->z_blksz,
1088 rl = zfs_range_lock(zp, boff, dlen, RL_READER);
1089 if (zp->z_blksz == dlen)
1091 zfs_range_unlock(rl);
1093 /* test for truncation needs to be done while range locked */
1094 if (off >= zp->z_phys->zp_size) {
1098 zgd = (zgd_t *)kmem_alloc(sizeof (zgd_t), KM_SLEEP);
1100 zgd->zgd_zilog = zfsvfs->z_log;
1101 zgd->zgd_bp = &lr->lr_blkptr;
1102 VERIFY(0 == dmu_buf_hold(os, lr->lr_foid, boff, zgd, &db));
1103 ASSERT(boff == db->db_offset);
1104 lr->lr_blkoff = off - boff;
1105 error = dmu_sync(zio, db, &lr->lr_blkptr,
1106 lr->lr_common.lrc_txg, zfs_get_done, zgd);
1107 ASSERT((error && error != EINPROGRESS) ||
1108 lr->lr_length <= zp->z_blksz);
1111 * dmu_sync() can compress a block of zeros to a null
1112 * blkptr but the block size still needs to be passed
1113 * through to replay.
1115 BP_SET_LSIZE(&lr->lr_blkptr, db->db_size);
1116 zil_add_block(zfsvfs->z_log, &lr->lr_blkptr);
1120 * If we get EINPROGRESS, then we need to wait for a
1121 * write IO initiated by dmu_sync() to complete before
1122 * we can release this dbuf. We will finish everything
1123 * up in the zfs_get_done() callback.
1125 if (error == EINPROGRESS) {
1127 } else if (error == EALREADY) {
1128 lr->lr_common.lrc_txtype = TX_WRITE2;
1131 dmu_buf_rele(db, zgd);
1132 kmem_free(zgd, sizeof (zgd_t));
1135 zfs_range_unlock(rl);
1137 * Release the vnode asynchronously as we currently have the
1138 * txg stopped from syncing.
1140 VN_RELE_ASYNC(ZTOV(zp), dsl_pool_vnrele_taskq(dmu_objset_pool(os)));
1146 zfs_access(vnode_t *vp, int mode, int flag, cred_t *cr,
1147 caller_context_t *ct)
1149 znode_t *zp = VTOZ(vp);
1150 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1156 if (flag & V_ACE_MASK)
1157 error = zfs_zaccess(zp, mode, flag, B_FALSE, cr);
1159 error = zfs_zaccess_rwx(zp, mode, flag, cr);
1166 * If vnode is for a device return a specfs vnode instead.
1169 specvp_check(vnode_t **vpp, cred_t *cr)
1173 if (IS_DEVVP(*vpp)) {
1176 svp = specvp(*vpp, (*vpp)->v_rdev, (*vpp)->v_type, cr);
1187 * Lookup an entry in a directory, or an extended attribute directory.
1188 * If it exists, return a held vnode reference for it.
1190 * IN: dvp - vnode of directory to search.
1191 * nm - name of entry to lookup.
1192 * pnp - full pathname to lookup [UNUSED].
1193 * flags - LOOKUP_XATTR set if looking for an attribute.
1194 * rdir - root directory vnode [UNUSED].
1195 * cr - credentials of caller.
1196 * ct - caller context
1197 * direntflags - directory lookup flags
1198 * realpnp - returned pathname.
1200 * OUT: vpp - vnode of located entry, NULL if not found.
1202 * RETURN: 0 if success
1203 * error code if failure
1210 zfs_lookup(vnode_t *dvp, char *nm, vnode_t **vpp, struct componentname *cnp,
1211 int nameiop, cred_t *cr, kthread_t *td, int flags)
1213 znode_t *zdp = VTOZ(dvp);
1214 zfsvfs_t *zfsvfs = zdp->z_zfsvfs;
1216 int *direntflags = NULL;
1217 void *realpnp = NULL;
1220 if (!(flags & (LOOKUP_XATTR | FIGNORECASE))) {
1222 if (dvp->v_type != VDIR) {
1224 } else if (zdp->z_dbuf == NULL) {
1228 if (nm[0] == 0 || (nm[0] == '.' && nm[1] == '\0')) {
1229 error = zfs_fastaccesschk_execute(zdp, cr);
1237 vnode_t *tvp = dnlc_lookup(dvp, nm);
1240 error = zfs_fastaccesschk_execute(zdp, cr);
1245 if (tvp == DNLC_NO_VNODE) {
1250 return (specvp_check(vpp, cr));
1256 DTRACE_PROBE2(zfs__fastpath__lookup__miss, vnode_t *, dvp, char *, nm);
1263 if (flags & LOOKUP_XATTR) {
1266 * If the xattr property is off, refuse the lookup request.
1268 if (!(zfsvfs->z_vfs->vfs_flag & VFS_XATTR)) {
1275 * We don't allow recursive attributes..
1276 * Maybe someday we will.
1278 if (zdp->z_phys->zp_flags & ZFS_XATTR) {
1283 if (error = zfs_get_xattrdir(VTOZ(dvp), vpp, cr, flags)) {
1289 * Do we have permission to get into attribute directory?
1292 if (error = zfs_zaccess(VTOZ(*vpp), ACE_EXECUTE, 0,
1302 if (dvp->v_type != VDIR) {
1308 * Check accessibility of directory.
1311 if (error = zfs_zaccess(zdp, ACE_EXECUTE, 0, B_FALSE, cr)) {
1316 if (zfsvfs->z_utf8 && u8_validate(nm, strlen(nm),
1317 NULL, U8_VALIDATE_ENTIRE, &error) < 0) {
1322 error = zfs_dirlook(zdp, nm, vpp, flags, direntflags, realpnp);
1324 error = specvp_check(vpp, cr);
1326 /* Translate errors and add SAVENAME when needed. */
1327 if (cnp->cn_flags & ISLASTCN) {
1331 if (error == ENOENT) {
1332 error = EJUSTRETURN;
1333 cnp->cn_flags |= SAVENAME;
1339 cnp->cn_flags |= SAVENAME;
1343 if (error == 0 && (nm[0] != '.' || nm[1] != '\0')) {
1346 if (cnp->cn_flags & ISDOTDOT) {
1347 ltype = VOP_ISLOCKED(dvp);
1351 error = vn_lock(*vpp, cnp->cn_lkflags);
1352 if (cnp->cn_flags & ISDOTDOT)
1353 vn_lock(dvp, ltype | LK_RETRY);
1363 #ifdef FREEBSD_NAMECACHE
1365 * Insert name into cache (as non-existent) if appropriate.
1367 if (error == ENOENT && (cnp->cn_flags & MAKEENTRY) && nameiop != CREATE)
1368 cache_enter(dvp, *vpp, cnp);
1370 * Insert name into cache if appropriate.
1372 if (error == 0 && (cnp->cn_flags & MAKEENTRY)) {
1373 if (!(cnp->cn_flags & ISLASTCN) ||
1374 (nameiop != DELETE && nameiop != RENAME)) {
1375 cache_enter(dvp, *vpp, cnp);
1384 * Attempt to create a new entry in a directory. If the entry
1385 * already exists, truncate the file if permissible, else return
1386 * an error. Return the vp of the created or trunc'd file.
1388 * IN: dvp - vnode of directory to put new file entry in.
1389 * name - name of new file entry.
1390 * vap - attributes of new file.
1391 * excl - flag indicating exclusive or non-exclusive mode.
1392 * mode - mode to open file with.
1393 * cr - credentials of caller.
1394 * flag - large file flag [UNUSED].
1395 * ct - caller context
1396 * vsecp - ACL to be set
1398 * OUT: vpp - vnode of created or trunc'd entry.
1400 * RETURN: 0 if success
1401 * error code if failure
1404 * dvp - ctime|mtime updated if new entry created
1405 * vp - ctime|mtime always, atime if new
1410 zfs_create(vnode_t *dvp, char *name, vattr_t *vap, int excl, int mode,
1411 vnode_t **vpp, cred_t *cr, kthread_t *td)
1413 znode_t *zp, *dzp = VTOZ(dvp);
1414 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
1422 gid_t gid = crgetgid(cr);
1423 zfs_acl_ids_t acl_ids;
1424 boolean_t fuid_dirtied;
1429 * If we have an ephemeral id, ACL, or XVATTR then
1430 * make sure file system is at proper version
1433 ksid = crgetsid(cr, KSID_OWNER);
1435 uid = ksid_getid(ksid);
1438 if (zfsvfs->z_use_fuids == B_FALSE &&
1439 (vsecp || (vap->va_mask & AT_XVATTR) ||
1440 IS_EPHEMERAL(crgetuid(cr)) || IS_EPHEMERAL(crgetgid(cr))))
1446 zilog = zfsvfs->z_log;
1448 if (zfsvfs->z_utf8 && u8_validate(name, strlen(name),
1449 NULL, U8_VALIDATE_ENTIRE, &error) < 0) {
1454 if (vap->va_mask & AT_XVATTR) {
1455 if ((error = secpolicy_xvattr(dvp, (xvattr_t *)vap,
1456 crgetuid(cr), cr, vap->va_type)) != 0) {
1464 if ((vap->va_mode & S_ISVTX) && secpolicy_vnode_stky_modify(cr))
1465 vap->va_mode &= ~S_ISVTX;
1467 if (*name == '\0') {
1469 * Null component name refers to the directory itself.
1476 /* possible VN_HOLD(zp) */
1479 if (flag & FIGNORECASE)
1482 error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg,
1485 if (strcmp(name, "..") == 0)
1495 * Create a new file object and update the directory
1498 if (error = zfs_zaccess(dzp, ACE_ADD_FILE, 0, B_FALSE, cr)) {
1503 * We only support the creation of regular files in
1504 * extended attribute directories.
1506 if ((dzp->z_phys->zp_flags & ZFS_XATTR) &&
1507 (vap->va_type != VREG)) {
1513 if ((error = zfs_acl_ids_create(dzp, 0, vap, cr, vsecp,
1516 if (zfs_acl_ids_overquota(zfsvfs, &acl_ids)) {
1517 zfs_acl_ids_free(&acl_ids);
1522 tx = dmu_tx_create(os);
1523 dmu_tx_hold_bonus(tx, DMU_NEW_OBJECT);
1524 fuid_dirtied = zfsvfs->z_fuid_dirty;
1526 zfs_fuid_txhold(zfsvfs, tx);
1527 dmu_tx_hold_bonus(tx, dzp->z_id);
1528 dmu_tx_hold_zap(tx, dzp->z_id, TRUE, name);
1529 if (acl_ids.z_aclp->z_acl_bytes > ZFS_ACE_SPACE) {
1530 dmu_tx_hold_write(tx, DMU_NEW_OBJECT,
1531 0, SPA_MAXBLOCKSIZE);
1533 error = dmu_tx_assign(tx, TXG_NOWAIT);
1535 zfs_acl_ids_free(&acl_ids);
1536 zfs_dirent_unlock(dl);
1537 if (error == ERESTART) {
1546 zfs_mknode(dzp, vap, tx, cr, 0, &zp, 0, &acl_ids);
1549 zfs_fuid_sync(zfsvfs, tx);
1551 (void) zfs_link_create(dl, zp, tx, ZNEW);
1553 txtype = zfs_log_create_txtype(Z_FILE, vsecp, vap);
1554 if (flag & FIGNORECASE)
1556 zfs_log_create(zilog, tx, txtype, dzp, zp, name,
1557 vsecp, acl_ids.z_fuidp, vap);
1558 zfs_acl_ids_free(&acl_ids);
1561 int aflags = (flag & FAPPEND) ? V_APPEND : 0;
1564 * A directory entry already exists for this name.
1567 * Can't truncate an existing file if in exclusive mode.
1574 * Can't open a directory for writing.
1576 if ((ZTOV(zp)->v_type == VDIR) && (mode & S_IWRITE)) {
1581 * Verify requested access to file.
1583 if (mode && (error = zfs_zaccess_rwx(zp, mode, aflags, cr))) {
1587 mutex_enter(&dzp->z_lock);
1589 mutex_exit(&dzp->z_lock);
1592 * Truncate regular files if requested.
1594 if ((ZTOV(zp)->v_type == VREG) &&
1595 (vap->va_mask & AT_SIZE) && (vap->va_size == 0)) {
1596 /* we can't hold any locks when calling zfs_freesp() */
1597 zfs_dirent_unlock(dl);
1599 error = zfs_freesp(zp, 0, 0, mode, TRUE);
1601 vnevent_create(ZTOV(zp), ct);
1607 zfs_dirent_unlock(dl);
1614 error = specvp_check(vpp, cr);
1622 * Remove an entry from a directory.
1624 * IN: dvp - vnode of directory to remove entry from.
1625 * name - name of entry to remove.
1626 * cr - credentials of caller.
1627 * ct - caller context
1628 * flags - case flags
1630 * RETURN: 0 if success
1631 * error code if failure
1635 * vp - ctime (if nlink > 0)
1639 zfs_remove(vnode_t *dvp, char *name, cred_t *cr, caller_context_t *ct,
1642 znode_t *zp, *dzp = VTOZ(dvp);
1643 znode_t *xzp = NULL;
1645 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
1647 uint64_t acl_obj, xattr_obj;
1650 boolean_t may_delete_now, delete_now = FALSE;
1651 boolean_t unlinked, toobig = FALSE;
1653 pathname_t *realnmp = NULL;
1660 zilog = zfsvfs->z_log;
1662 if (flags & FIGNORECASE) {
1670 * Attempt to lock directory; fail if entry doesn't exist.
1672 if (error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg,
1682 if (error = zfs_zaccess_delete(dzp, zp, cr)) {
1687 * Need to use rmdir for removing directories.
1689 if (vp->v_type == VDIR) {
1694 vnevent_remove(vp, dvp, name, ct);
1697 dnlc_remove(dvp, realnmp->pn_buf);
1699 dnlc_remove(dvp, name);
1701 may_delete_now = FALSE;
1704 * We may delete the znode now, or we may put it in the unlinked set;
1705 * it depends on whether we're the last link, and on whether there are
1706 * other holds on the vnode. So we dmu_tx_hold() the right things to
1707 * allow for either case.
1709 tx = dmu_tx_create(zfsvfs->z_os);
1710 dmu_tx_hold_zap(tx, dzp->z_id, FALSE, name);
1711 dmu_tx_hold_bonus(tx, zp->z_id);
1712 if (may_delete_now) {
1714 zp->z_phys->zp_size > zp->z_blksz * DMU_MAX_DELETEBLKCNT;
1715 /* if the file is too big, only hold_free a token amount */
1716 dmu_tx_hold_free(tx, zp->z_id, 0,
1717 (toobig ? DMU_MAX_ACCESS : DMU_OBJECT_END));
1720 /* are there any extended attributes? */
1721 if ((xattr_obj = zp->z_phys->zp_xattr) != 0) {
1722 /* XXX - do we need this if we are deleting? */
1723 dmu_tx_hold_bonus(tx, xattr_obj);
1726 /* are there any additional acls */
1727 if ((acl_obj = zp->z_phys->zp_acl.z_acl_extern_obj) != 0 &&
1729 dmu_tx_hold_free(tx, acl_obj, 0, DMU_OBJECT_END);
1731 /* charge as an update -- would be nice not to charge at all */
1732 dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL);
1734 error = dmu_tx_assign(tx, TXG_NOWAIT);
1736 zfs_dirent_unlock(dl);
1738 if (error == ERESTART) {
1751 * Remove the directory entry.
1753 error = zfs_link_destroy(dl, zp, tx, zflg, &unlinked);
1760 if (0 && unlinked) {
1762 delete_now = may_delete_now && !toobig &&
1763 vp->v_count == 1 && !vn_has_cached_data(vp) &&
1764 zp->z_phys->zp_xattr == xattr_obj &&
1765 zp->z_phys->zp_acl.z_acl_extern_obj == acl_obj;
1770 if (zp->z_phys->zp_xattr) {
1771 error = zfs_zget(zfsvfs, zp->z_phys->zp_xattr, &xzp);
1772 ASSERT3U(error, ==, 0);
1773 ASSERT3U(xzp->z_phys->zp_links, ==, 2);
1774 dmu_buf_will_dirty(xzp->z_dbuf, tx);
1775 mutex_enter(&xzp->z_lock);
1776 xzp->z_unlinked = 1;
1777 xzp->z_phys->zp_links = 0;
1778 mutex_exit(&xzp->z_lock);
1779 zfs_unlinked_add(xzp, tx);
1780 zp->z_phys->zp_xattr = 0; /* probably unnecessary */
1782 mutex_enter(&zp->z_lock);
1785 ASSERT3U(vp->v_count, ==, 0);
1787 mutex_exit(&zp->z_lock);
1788 zfs_znode_delete(zp, tx);
1789 } else if (unlinked) {
1790 zfs_unlinked_add(zp, tx);
1794 if (flags & FIGNORECASE)
1796 zfs_log_remove(zilog, tx, txtype, dzp, name);
1803 zfs_dirent_unlock(dl);
1808 /* this rele is delayed to prevent nesting transactions */
1817 * Create a new directory and insert it into dvp using the name
1818 * provided. Return a pointer to the inserted directory.
1820 * IN: dvp - vnode of directory to add subdir to.
1821 * dirname - name of new directory.
1822 * vap - attributes of new directory.
1823 * cr - credentials of caller.
1824 * ct - caller context
1825 * vsecp - ACL to be set
1827 * OUT: vpp - vnode of created directory.
1829 * RETURN: 0 if success
1830 * error code if failure
1833 * dvp - ctime|mtime updated
1834 * vp - ctime|mtime|atime updated
1838 zfs_mkdir(vnode_t *dvp, char *dirname, vattr_t *vap, vnode_t **vpp, cred_t *cr,
1839 caller_context_t *ct, int flags, vsecattr_t *vsecp)
1841 znode_t *zp, *dzp = VTOZ(dvp);
1842 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
1851 gid_t gid = crgetgid(cr);
1852 zfs_acl_ids_t acl_ids;
1853 boolean_t fuid_dirtied;
1855 ASSERT(vap->va_type == VDIR);
1858 * If we have an ephemeral id, ACL, or XVATTR then
1859 * make sure file system is at proper version
1862 ksid = crgetsid(cr, KSID_OWNER);
1864 uid = ksid_getid(ksid);
1867 if (zfsvfs->z_use_fuids == B_FALSE &&
1868 (vsecp || (vap->va_mask & AT_XVATTR) || IS_EPHEMERAL(crgetuid(cr))||
1869 IS_EPHEMERAL(crgetgid(cr))))
1874 zilog = zfsvfs->z_log;
1876 if (dzp->z_phys->zp_flags & ZFS_XATTR) {
1881 if (zfsvfs->z_utf8 && u8_validate(dirname,
1882 strlen(dirname), NULL, U8_VALIDATE_ENTIRE, &error) < 0) {
1886 if (flags & FIGNORECASE)
1889 if (vap->va_mask & AT_XVATTR)
1890 if ((error = secpolicy_xvattr(dvp, (xvattr_t *)vap,
1891 crgetuid(cr), cr, vap->va_type)) != 0) {
1897 * First make sure the new directory doesn't exist.
1902 if (error = zfs_dirent_lock(&dl, dzp, dirname, &zp, zf,
1908 if (error = zfs_zaccess(dzp, ACE_ADD_SUBDIRECTORY, 0, B_FALSE, cr)) {
1909 zfs_dirent_unlock(dl);
1914 if ((error = zfs_acl_ids_create(dzp, 0, vap, cr, vsecp,
1916 zfs_dirent_unlock(dl);
1920 if (zfs_acl_ids_overquota(zfsvfs, &acl_ids)) {
1921 zfs_acl_ids_free(&acl_ids);
1922 zfs_dirent_unlock(dl);
1928 * Add a new entry to the directory.
1930 tx = dmu_tx_create(zfsvfs->z_os);
1931 dmu_tx_hold_zap(tx, dzp->z_id, TRUE, dirname);
1932 dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL);
1933 fuid_dirtied = zfsvfs->z_fuid_dirty;
1935 zfs_fuid_txhold(zfsvfs, tx);
1936 if (acl_ids.z_aclp->z_acl_bytes > ZFS_ACE_SPACE)
1937 dmu_tx_hold_write(tx, DMU_NEW_OBJECT,
1938 0, SPA_MAXBLOCKSIZE);
1939 error = dmu_tx_assign(tx, TXG_NOWAIT);
1941 zfs_acl_ids_free(&acl_ids);
1942 zfs_dirent_unlock(dl);
1943 if (error == ERESTART) {
1956 zfs_mknode(dzp, vap, tx, cr, 0, &zp, 0, &acl_ids);
1959 zfs_fuid_sync(zfsvfs, tx);
1961 * Now put new name in parent dir.
1963 (void) zfs_link_create(dl, zp, tx, ZNEW);
1967 txtype = zfs_log_create_txtype(Z_DIR, vsecp, vap);
1968 if (flags & FIGNORECASE)
1970 zfs_log_create(zilog, tx, txtype, dzp, zp, dirname, vsecp,
1971 acl_ids.z_fuidp, vap);
1973 zfs_acl_ids_free(&acl_ids);
1976 zfs_dirent_unlock(dl);
1983 * Remove a directory subdir entry. If the current working
1984 * directory is the same as the subdir to be removed, the
1987 * IN: dvp - vnode of directory to remove from.
1988 * name - name of directory to be removed.
1989 * cwd - vnode of current working directory.
1990 * cr - credentials of caller.
1991 * ct - caller context
1992 * flags - case flags
1994 * RETURN: 0 if success
1995 * error code if failure
1998 * dvp - ctime|mtime updated
2002 zfs_rmdir(vnode_t *dvp, char *name, vnode_t *cwd, cred_t *cr,
2003 caller_context_t *ct, int flags)
2005 znode_t *dzp = VTOZ(dvp);
2008 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
2017 zilog = zfsvfs->z_log;
2019 if (flags & FIGNORECASE)
2025 * Attempt to lock directory; fail if entry doesn't exist.
2027 if (error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg,
2035 if (error = zfs_zaccess_delete(dzp, zp, cr)) {
2039 if (vp->v_type != VDIR) {
2049 vnevent_rmdir(vp, dvp, name, ct);
2052 * Grab a lock on the directory to make sure that noone is
2053 * trying to add (or lookup) entries while we are removing it.
2055 rw_enter(&zp->z_name_lock, RW_WRITER);
2058 * Grab a lock on the parent pointer to make sure we play well
2059 * with the treewalk and directory rename code.
2061 rw_enter(&zp->z_parent_lock, RW_WRITER);
2063 tx = dmu_tx_create(zfsvfs->z_os);
2064 dmu_tx_hold_zap(tx, dzp->z_id, FALSE, name);
2065 dmu_tx_hold_bonus(tx, zp->z_id);
2066 dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL);
2067 error = dmu_tx_assign(tx, TXG_NOWAIT);
2069 rw_exit(&zp->z_parent_lock);
2070 rw_exit(&zp->z_name_lock);
2071 zfs_dirent_unlock(dl);
2073 if (error == ERESTART) {
2083 #ifdef FREEBSD_NAMECACHE
2087 error = zfs_link_destroy(dl, zp, tx, zflg, NULL);
2090 uint64_t txtype = TX_RMDIR;
2091 if (flags & FIGNORECASE)
2093 zfs_log_remove(zilog, tx, txtype, dzp, name);
2098 rw_exit(&zp->z_parent_lock);
2099 rw_exit(&zp->z_name_lock);
2100 #ifdef FREEBSD_NAMECACHE
2104 zfs_dirent_unlock(dl);
2113 * Read as many directory entries as will fit into the provided
2114 * buffer from the given directory cursor position (specified in
2115 * the uio structure.
2117 * IN: vp - vnode of directory to read.
2118 * uio - structure supplying read location, range info,
2119 * and return buffer.
2120 * cr - credentials of caller.
2121 * ct - caller context
2122 * flags - case flags
2124 * OUT: uio - updated offset and range, buffer filled.
2125 * eofp - set to true if end-of-file detected.
2127 * RETURN: 0 if success
2128 * error code if failure
2131 * vp - atime updated
2133 * Note that the low 4 bits of the cookie returned by zap is always zero.
2134 * This allows us to use the low range for "special" directory entries:
2135 * We use 0 for '.', and 1 for '..'. If this is the root of the filesystem,
2136 * we use the offset 2 for the '.zfs' directory.
2140 zfs_readdir(vnode_t *vp, uio_t *uio, cred_t *cr, int *eofp, int *ncookies, u_long **cookies)
2142 znode_t *zp = VTOZ(vp);
2146 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
2151 zap_attribute_t zap;
2152 uint_t bytes_wanted;
2153 uint64_t offset; /* must be unsigned; checks for < 1 */
2158 boolean_t check_sysattrs;
2161 u_long *cooks = NULL;
2168 * If we are not given an eof variable,
2175 * Check for valid iov_len.
2177 if (uio->uio_iov->iov_len <= 0) {
2183 * Quit if directory has been removed (posix)
2185 if ((*eofp = zp->z_unlinked) != 0) {
2192 offset = uio->uio_loffset;
2193 prefetch = zp->z_zn_prefetch;
2196 * Initialize the iterator cursor.
2200 * Start iteration from the beginning of the directory.
2202 zap_cursor_init(&zc, os, zp->z_id);
2205 * The offset is a serialized cursor.
2207 zap_cursor_init_serialized(&zc, os, zp->z_id, offset);
2211 * Get space to change directory entries into fs independent format.
2213 iovp = uio->uio_iov;
2214 bytes_wanted = iovp->iov_len;
2215 if (uio->uio_segflg != UIO_SYSSPACE || uio->uio_iovcnt != 1) {
2216 bufsize = bytes_wanted;
2217 outbuf = kmem_alloc(bufsize, KM_SLEEP);
2218 odp = (struct dirent64 *)outbuf;
2220 bufsize = bytes_wanted;
2221 odp = (struct dirent64 *)iovp->iov_base;
2223 eodp = (struct edirent *)odp;
2225 if (ncookies != NULL) {
2227 * Minimum entry size is dirent size and 1 byte for a file name.
2229 ncooks = uio->uio_resid / (sizeof(struct dirent) - sizeof(((struct dirent *)NULL)->d_name) + 1);
2230 cooks = malloc(ncooks * sizeof(u_long), M_TEMP, M_WAITOK);
2235 * If this VFS supports the system attribute view interface; and
2236 * we're looking at an extended attribute directory; and we care
2237 * about normalization conflicts on this vfs; then we must check
2238 * for normalization conflicts with the sysattr name space.
2241 check_sysattrs = vfs_has_feature(vp->v_vfsp, VFSFT_SYSATTR_VIEWS) &&
2242 (vp->v_flag & V_XATTRDIR) && zfsvfs->z_norm &&
2243 (flags & V_RDDIR_ENTFLAGS);
2249 * Transform to file-system independent format
2252 while (outcount < bytes_wanted) {
2258 * Special case `.', `..', and `.zfs'.
2261 (void) strcpy(zap.za_name, ".");
2262 zap.za_normalization_conflict = 0;
2265 } else if (offset == 1) {
2266 (void) strcpy(zap.za_name, "..");
2267 zap.za_normalization_conflict = 0;
2268 objnum = zp->z_phys->zp_parent;
2270 } else if (offset == 2 && zfs_show_ctldir(zp)) {
2271 (void) strcpy(zap.za_name, ZFS_CTLDIR_NAME);
2272 zap.za_normalization_conflict = 0;
2273 objnum = ZFSCTL_INO_ROOT;
2279 if (error = zap_cursor_retrieve(&zc, &zap)) {
2280 if ((*eofp = (error == ENOENT)) != 0)
2286 if (zap.za_integer_length != 8 ||
2287 zap.za_num_integers != 1) {
2288 cmn_err(CE_WARN, "zap_readdir: bad directory "
2289 "entry, obj = %lld, offset = %lld\n",
2290 (u_longlong_t)zp->z_id,
2291 (u_longlong_t)offset);
2296 objnum = ZFS_DIRENT_OBJ(zap.za_first_integer);
2298 * MacOS X can extract the object type here such as:
2299 * uint8_t type = ZFS_DIRENT_TYPE(zap.za_first_integer);
2301 type = ZFS_DIRENT_TYPE(zap.za_first_integer);
2303 if (check_sysattrs && !zap.za_normalization_conflict) {
2305 zap.za_normalization_conflict =
2306 xattr_sysattr_casechk(zap.za_name);
2308 panic("%s:%u: TODO", __func__, __LINE__);
2313 if (flags & V_RDDIR_ACCFILTER) {
2315 * If we have no access at all, don't include
2316 * this entry in the returned information
2319 if (zfs_zget(zp->z_zfsvfs, objnum, &ezp) != 0)
2321 if (!zfs_has_access(ezp, cr)) {
2328 if (flags & V_RDDIR_ENTFLAGS)
2329 reclen = EDIRENT_RECLEN(strlen(zap.za_name));
2331 reclen = DIRENT64_RECLEN(strlen(zap.za_name));
2334 * Will this entry fit in the buffer?
2336 if (outcount + reclen > bufsize) {
2338 * Did we manage to fit anything in the buffer?
2346 if (flags & V_RDDIR_ENTFLAGS) {
2348 * Add extended flag entry:
2350 eodp->ed_ino = objnum;
2351 eodp->ed_reclen = reclen;
2352 /* NOTE: ed_off is the offset for the *next* entry */
2353 next = &(eodp->ed_off);
2354 eodp->ed_eflags = zap.za_normalization_conflict ?
2355 ED_CASE_CONFLICT : 0;
2356 (void) strncpy(eodp->ed_name, zap.za_name,
2357 EDIRENT_NAMELEN(reclen));
2358 eodp = (edirent_t *)((intptr_t)eodp + reclen);
2363 odp->d_ino = objnum;
2364 odp->d_reclen = reclen;
2365 odp->d_namlen = strlen(zap.za_name);
2366 (void) strlcpy(odp->d_name, zap.za_name, odp->d_namlen + 1);
2368 odp = (dirent64_t *)((intptr_t)odp + reclen);
2372 ASSERT(outcount <= bufsize);
2374 /* Prefetch znode */
2376 dmu_prefetch(os, objnum, 0, 0);
2380 * Move to the next entry, fill in the previous offset.
2382 if (offset > 2 || (offset == 2 && !zfs_show_ctldir(zp))) {
2383 zap_cursor_advance(&zc);
2384 offset = zap_cursor_serialize(&zc);
2389 if (cooks != NULL) {
2392 KASSERT(ncooks >= 0, ("ncookies=%d", ncooks));
2395 zp->z_zn_prefetch = B_FALSE; /* a lookup will re-enable pre-fetching */
2397 /* Subtract unused cookies */
2398 if (ncookies != NULL)
2399 *ncookies -= ncooks;
2401 if (uio->uio_segflg == UIO_SYSSPACE && uio->uio_iovcnt == 1) {
2402 iovp->iov_base += outcount;
2403 iovp->iov_len -= outcount;
2404 uio->uio_resid -= outcount;
2405 } else if (error = uiomove(outbuf, (long)outcount, UIO_READ, uio)) {
2407 * Reset the pointer.
2409 offset = uio->uio_loffset;
2413 zap_cursor_fini(&zc);
2414 if (uio->uio_segflg != UIO_SYSSPACE || uio->uio_iovcnt != 1)
2415 kmem_free(outbuf, bufsize);
2417 if (error == ENOENT)
2420 ZFS_ACCESSTIME_STAMP(zfsvfs, zp);
2422 uio->uio_loffset = offset;
2424 if (error != 0 && cookies != NULL) {
2425 free(*cookies, M_TEMP);
2432 ulong_t zfs_fsync_sync_cnt = 4;
2435 zfs_fsync(vnode_t *vp, int syncflag, cred_t *cr, caller_context_t *ct)
2437 znode_t *zp = VTOZ(vp);
2438 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
2440 (void) tsd_set(zfs_fsyncer_key, (void *)zfs_fsync_sync_cnt);
2444 zil_commit(zfsvfs->z_log, zp->z_last_itx, zp->z_id);
2451 * Get the requested file attributes and place them in the provided
2454 * IN: vp - vnode of file.
2455 * vap - va_mask identifies requested attributes.
2456 * If AT_XVATTR set, then optional attrs are requested
2457 * flags - ATTR_NOACLCHECK (CIFS server context)
2458 * cr - credentials of caller.
2459 * ct - caller context
2461 * OUT: vap - attribute values.
2463 * RETURN: 0 (always succeeds)
2467 zfs_getattr(vnode_t *vp, vattr_t *vap, int flags, cred_t *cr,
2468 caller_context_t *ct)
2470 znode_t *zp = VTOZ(vp);
2471 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
2475 u_longlong_t nblocks;
2477 xvattr_t *xvap = (xvattr_t *)vap; /* vap may be an xvattr_t * */
2478 xoptattr_t *xoap = NULL;
2479 boolean_t skipaclchk = (flags & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE;
2486 * If ACL is trivial don't bother looking for ACE_READ_ATTRIBUTES.
2487 * Also, if we are the owner don't bother, since owner should
2488 * always be allowed to read basic attributes of file.
2490 if (!(pzp->zp_flags & ZFS_ACL_TRIVIAL) &&
2491 (pzp->zp_uid != crgetuid(cr))) {
2492 if (error = zfs_zaccess(zp, ACE_READ_ATTRIBUTES, 0,
2500 * Return all attributes. It's cheaper to provide the answer
2501 * than to determine whether we were asked the question.
2504 mutex_enter(&zp->z_lock);
2505 vap->va_type = IFTOVT(pzp->zp_mode);
2506 vap->va_mode = pzp->zp_mode & ~S_IFMT;
2507 zfs_fuid_map_ids(zp, cr, &vap->va_uid, &vap->va_gid);
2508 // vap->va_fsid = zp->z_zfsvfs->z_vfs->vfs_dev;
2509 vap->va_nodeid = zp->z_id;
2510 if ((vp->v_flag & VROOT) && zfs_show_ctldir(zp))
2511 links = pzp->zp_links + 1;
2513 links = pzp->zp_links;
2514 vap->va_nlink = MIN(links, UINT32_MAX); /* nlink_t limit! */
2515 vap->va_size = pzp->zp_size;
2516 vap->va_fsid = vp->v_mount->mnt_stat.f_fsid.val[0];
2517 vap->va_rdev = zfs_cmpldev(pzp->zp_rdev);
2518 vap->va_seq = zp->z_seq;
2519 vap->va_flags = 0; /* FreeBSD: Reset chflags(2) flags. */
2522 * Add in any requested optional attributes and the create time.
2523 * Also set the corresponding bits in the returned attribute bitmap.
2525 if ((xoap = xva_getxoptattr(xvap)) != NULL && zfsvfs->z_use_fuids) {
2526 if (XVA_ISSET_REQ(xvap, XAT_ARCHIVE)) {
2528 ((pzp->zp_flags & ZFS_ARCHIVE) != 0);
2529 XVA_SET_RTN(xvap, XAT_ARCHIVE);
2532 if (XVA_ISSET_REQ(xvap, XAT_READONLY)) {
2533 xoap->xoa_readonly =
2534 ((pzp->zp_flags & ZFS_READONLY) != 0);
2535 XVA_SET_RTN(xvap, XAT_READONLY);
2538 if (XVA_ISSET_REQ(xvap, XAT_SYSTEM)) {
2540 ((pzp->zp_flags & ZFS_SYSTEM) != 0);
2541 XVA_SET_RTN(xvap, XAT_SYSTEM);
2544 if (XVA_ISSET_REQ(xvap, XAT_HIDDEN)) {
2546 ((pzp->zp_flags & ZFS_HIDDEN) != 0);
2547 XVA_SET_RTN(xvap, XAT_HIDDEN);
2550 if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK)) {
2551 xoap->xoa_nounlink =
2552 ((pzp->zp_flags & ZFS_NOUNLINK) != 0);
2553 XVA_SET_RTN(xvap, XAT_NOUNLINK);
2556 if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE)) {
2557 xoap->xoa_immutable =
2558 ((pzp->zp_flags & ZFS_IMMUTABLE) != 0);
2559 XVA_SET_RTN(xvap, XAT_IMMUTABLE);
2562 if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY)) {
2563 xoap->xoa_appendonly =
2564 ((pzp->zp_flags & ZFS_APPENDONLY) != 0);
2565 XVA_SET_RTN(xvap, XAT_APPENDONLY);
2568 if (XVA_ISSET_REQ(xvap, XAT_NODUMP)) {
2570 ((pzp->zp_flags & ZFS_NODUMP) != 0);
2571 XVA_SET_RTN(xvap, XAT_NODUMP);
2574 if (XVA_ISSET_REQ(xvap, XAT_OPAQUE)) {
2576 ((pzp->zp_flags & ZFS_OPAQUE) != 0);
2577 XVA_SET_RTN(xvap, XAT_OPAQUE);
2580 if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED)) {
2581 xoap->xoa_av_quarantined =
2582 ((pzp->zp_flags & ZFS_AV_QUARANTINED) != 0);
2583 XVA_SET_RTN(xvap, XAT_AV_QUARANTINED);
2586 if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED)) {
2587 xoap->xoa_av_modified =
2588 ((pzp->zp_flags & ZFS_AV_MODIFIED) != 0);
2589 XVA_SET_RTN(xvap, XAT_AV_MODIFIED);
2592 if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP) &&
2593 vp->v_type == VREG &&
2594 (pzp->zp_flags & ZFS_BONUS_SCANSTAMP)) {
2596 dmu_object_info_t doi;
2599 * Only VREG files have anti-virus scanstamps, so we
2600 * won't conflict with symlinks in the bonus buffer.
2602 dmu_object_info_from_db(zp->z_dbuf, &doi);
2603 len = sizeof (xoap->xoa_av_scanstamp) +
2604 sizeof (znode_phys_t);
2605 if (len <= doi.doi_bonus_size) {
2607 * pzp points to the start of the
2608 * znode_phys_t. pzp + 1 points to the
2609 * first byte after the znode_phys_t.
2611 (void) memcpy(xoap->xoa_av_scanstamp,
2613 sizeof (xoap->xoa_av_scanstamp));
2614 XVA_SET_RTN(xvap, XAT_AV_SCANSTAMP);
2618 if (XVA_ISSET_REQ(xvap, XAT_CREATETIME)) {
2619 ZFS_TIME_DECODE(&xoap->xoa_createtime, pzp->zp_crtime);
2620 XVA_SET_RTN(xvap, XAT_CREATETIME);
2624 ZFS_TIME_DECODE(&vap->va_atime, pzp->zp_atime);
2625 ZFS_TIME_DECODE(&vap->va_mtime, pzp->zp_mtime);
2626 ZFS_TIME_DECODE(&vap->va_ctime, pzp->zp_ctime);
2627 ZFS_TIME_DECODE(&vap->va_birthtime, pzp->zp_crtime);
2629 mutex_exit(&zp->z_lock);
2631 dmu_object_size_from_db(zp->z_dbuf, &blksize, &nblocks);
2632 vap->va_blksize = blksize;
2633 vap->va_bytes = nblocks << 9; /* nblocks * 512 */
2635 if (zp->z_blksz == 0) {
2637 * Block size hasn't been set; suggest maximal I/O transfers.
2639 vap->va_blksize = zfsvfs->z_max_blksz;
2647 * Set the file attributes to the values contained in the
2650 * IN: vp - vnode of file to be modified.
2651 * vap - new attribute values.
2652 * If AT_XVATTR set, then optional attrs are being set
2653 * flags - ATTR_UTIME set if non-default time values provided.
2654 * - ATTR_NOACLCHECK (CIFS context only).
2655 * cr - credentials of caller.
2656 * ct - caller context
2658 * RETURN: 0 if success
2659 * error code if failure
2662 * vp - ctime updated, mtime updated if size changed.
2666 zfs_setattr(vnode_t *vp, vattr_t *vap, int flags, cred_t *cr,
2667 caller_context_t *ct)
2669 znode_t *zp = VTOZ(vp);
2671 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
2676 uint_t mask = vap->va_mask;
2678 uint64_t saved_mode;
2681 uint64_t new_uid, new_gid;
2683 int need_policy = FALSE;
2685 zfs_fuid_info_t *fuidp = NULL;
2686 xvattr_t *xvap = (xvattr_t *)vap; /* vap may be an xvattr_t * */
2688 zfs_acl_t *aclp = NULL;
2689 boolean_t skipaclchk = (flags & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE;
2690 boolean_t fuid_dirtied = B_FALSE;
2695 if (mask & AT_NOSET)
2702 zilog = zfsvfs->z_log;
2705 * Make sure that if we have ephemeral uid/gid or xvattr specified
2706 * that file system is at proper version level
2709 if (zfsvfs->z_use_fuids == B_FALSE &&
2710 (((mask & AT_UID) && IS_EPHEMERAL(vap->va_uid)) ||
2711 ((mask & AT_GID) && IS_EPHEMERAL(vap->va_gid)) ||
2712 (mask & AT_XVATTR))) {
2717 if (mask & AT_SIZE && vp->v_type == VDIR) {
2722 if (mask & AT_SIZE && vp->v_type != VREG && vp->v_type != VFIFO) {
2728 * If this is an xvattr_t, then get a pointer to the structure of
2729 * optional attributes. If this is NULL, then we have a vattr_t.
2731 xoap = xva_getxoptattr(xvap);
2733 xva_init(&tmpxvattr);
2736 * Immutable files can only alter immutable bit and atime
2738 if ((pzp->zp_flags & ZFS_IMMUTABLE) &&
2739 ((mask & (AT_SIZE|AT_UID|AT_GID|AT_MTIME|AT_MODE)) ||
2740 ((mask & AT_XVATTR) && XVA_ISSET_REQ(xvap, XAT_CREATETIME)))) {
2745 if ((mask & AT_SIZE) && (pzp->zp_flags & ZFS_READONLY)) {
2751 * Verify timestamps doesn't overflow 32 bits.
2752 * ZFS can handle large timestamps, but 32bit syscalls can't
2753 * handle times greater than 2039. This check should be removed
2754 * once large timestamps are fully supported.
2756 if (mask & (AT_ATIME | AT_MTIME)) {
2757 if (((mask & AT_ATIME) && TIMESPEC_OVERFLOW(&vap->va_atime)) ||
2758 ((mask & AT_MTIME) && TIMESPEC_OVERFLOW(&vap->va_mtime))) {
2767 /* Can this be moved to before the top label? */
2768 if (zfsvfs->z_vfs->vfs_flag & VFS_RDONLY) {
2774 * First validate permissions
2777 if (mask & AT_SIZE) {
2778 err = zfs_zaccess(zp, ACE_WRITE_DATA, 0, skipaclchk, cr);
2784 * XXX - Note, we are not providing any open
2785 * mode flags here (like FNDELAY), so we may
2786 * block if there are locks present... this
2787 * should be addressed in openat().
2789 /* XXX - would it be OK to generate a log record here? */
2790 err = zfs_freesp(zp, vap->va_size, 0, 0, FALSE);
2797 if (mask & (AT_ATIME|AT_MTIME) ||
2798 ((mask & AT_XVATTR) && (XVA_ISSET_REQ(xvap, XAT_HIDDEN) ||
2799 XVA_ISSET_REQ(xvap, XAT_READONLY) ||
2800 XVA_ISSET_REQ(xvap, XAT_ARCHIVE) ||
2801 XVA_ISSET_REQ(xvap, XAT_CREATETIME) ||
2802 XVA_ISSET_REQ(xvap, XAT_SYSTEM))))
2803 need_policy = zfs_zaccess(zp, ACE_WRITE_ATTRIBUTES, 0,
2806 if (mask & (AT_UID|AT_GID)) {
2807 int idmask = (mask & (AT_UID|AT_GID));
2812 * NOTE: even if a new mode is being set,
2813 * we may clear S_ISUID/S_ISGID bits.
2816 if (!(mask & AT_MODE))
2817 vap->va_mode = pzp->zp_mode;
2820 * Take ownership or chgrp to group we are a member of
2823 take_owner = (mask & AT_UID) && (vap->va_uid == crgetuid(cr));
2824 take_group = (mask & AT_GID) &&
2825 zfs_groupmember(zfsvfs, vap->va_gid, cr);
2828 * If both AT_UID and AT_GID are set then take_owner and
2829 * take_group must both be set in order to allow taking
2832 * Otherwise, send the check through secpolicy_vnode_setattr()
2836 if (((idmask == (AT_UID|AT_GID)) && take_owner && take_group) ||
2837 ((idmask == AT_UID) && take_owner) ||
2838 ((idmask == AT_GID) && take_group)) {
2839 if (zfs_zaccess(zp, ACE_WRITE_OWNER, 0,
2840 skipaclchk, cr) == 0) {
2842 * Remove setuid/setgid for non-privileged users
2844 secpolicy_setid_clear(vap, vp, cr);
2845 trim_mask = (mask & (AT_UID|AT_GID));
2854 mutex_enter(&zp->z_lock);
2855 oldva.va_mode = pzp->zp_mode;
2856 zfs_fuid_map_ids(zp, cr, &oldva.va_uid, &oldva.va_gid);
2857 if (mask & AT_XVATTR) {
2859 * Update xvattr mask to include only those attributes
2860 * that are actually changing.
2862 * the bits will be restored prior to actually setting
2863 * the attributes so the caller thinks they were set.
2865 if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY)) {
2866 if (xoap->xoa_appendonly !=
2867 ((pzp->zp_flags & ZFS_APPENDONLY) != 0)) {
2870 XVA_CLR_REQ(xvap, XAT_APPENDONLY);
2871 XVA_SET_REQ(&tmpxvattr, XAT_APPENDONLY);
2875 if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK)) {
2876 if (xoap->xoa_nounlink !=
2877 ((pzp->zp_flags & ZFS_NOUNLINK) != 0)) {
2880 XVA_CLR_REQ(xvap, XAT_NOUNLINK);
2881 XVA_SET_REQ(&tmpxvattr, XAT_NOUNLINK);
2885 if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE)) {
2886 if (xoap->xoa_immutable !=
2887 ((pzp->zp_flags & ZFS_IMMUTABLE) != 0)) {
2890 XVA_CLR_REQ(xvap, XAT_IMMUTABLE);
2891 XVA_SET_REQ(&tmpxvattr, XAT_IMMUTABLE);
2895 if (XVA_ISSET_REQ(xvap, XAT_NODUMP)) {
2896 if (xoap->xoa_nodump !=
2897 ((pzp->zp_flags & ZFS_NODUMP) != 0)) {
2900 XVA_CLR_REQ(xvap, XAT_NODUMP);
2901 XVA_SET_REQ(&tmpxvattr, XAT_NODUMP);
2905 if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED)) {
2906 if (xoap->xoa_av_modified !=
2907 ((pzp->zp_flags & ZFS_AV_MODIFIED) != 0)) {
2910 XVA_CLR_REQ(xvap, XAT_AV_MODIFIED);
2911 XVA_SET_REQ(&tmpxvattr, XAT_AV_MODIFIED);
2915 if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED)) {
2916 if ((vp->v_type != VREG &&
2917 xoap->xoa_av_quarantined) ||
2918 xoap->xoa_av_quarantined !=
2919 ((pzp->zp_flags & ZFS_AV_QUARANTINED) != 0)) {
2922 XVA_CLR_REQ(xvap, XAT_AV_QUARANTINED);
2923 XVA_SET_REQ(&tmpxvattr, XAT_AV_QUARANTINED);
2927 if (need_policy == FALSE &&
2928 (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP) ||
2929 XVA_ISSET_REQ(xvap, XAT_OPAQUE))) {
2934 mutex_exit(&zp->z_lock);
2936 if (mask & AT_MODE) {
2937 if (zfs_zaccess(zp, ACE_WRITE_ACL, 0, skipaclchk, cr) == 0) {
2938 err = secpolicy_setid_setsticky_clear(vp, vap,
2944 trim_mask |= AT_MODE;
2952 * If trim_mask is set then take ownership
2953 * has been granted or write_acl is present and user
2954 * has the ability to modify mode. In that case remove
2955 * UID|GID and or MODE from mask so that
2956 * secpolicy_vnode_setattr() doesn't revoke it.
2960 saved_mask = vap->va_mask;
2961 vap->va_mask &= ~trim_mask;
2962 if (trim_mask & AT_MODE) {
2964 * Save the mode, as secpolicy_vnode_setattr()
2965 * will overwrite it with ova.va_mode.
2967 saved_mode = vap->va_mode;
2970 err = secpolicy_vnode_setattr(cr, vp, vap, &oldva, flags,
2971 (int (*)(void *, int, cred_t *))zfs_zaccess_unix, zp);
2978 vap->va_mask |= saved_mask;
2979 if (trim_mask & AT_MODE) {
2981 * Recover the mode after
2982 * secpolicy_vnode_setattr().
2984 vap->va_mode = saved_mode;
2990 * secpolicy_vnode_setattr, or take ownership may have
2993 mask = vap->va_mask;
2995 tx = dmu_tx_create(zfsvfs->z_os);
2996 dmu_tx_hold_bonus(tx, zp->z_id);
2998 if (mask & AT_MODE) {
2999 uint64_t pmode = pzp->zp_mode;
3001 new_mode = (pmode & S_IFMT) | (vap->va_mode & ~S_IFMT);
3003 if (err = zfs_acl_chmod_setattr(zp, &aclp, new_mode))
3005 if (pzp->zp_acl.z_acl_extern_obj) {
3006 /* Are we upgrading ACL from old V0 format to new V1 */
3007 if (zfsvfs->z_version <= ZPL_VERSION_FUID &&
3008 pzp->zp_acl.z_acl_version ==
3009 ZFS_ACL_VERSION_INITIAL) {
3010 dmu_tx_hold_free(tx,
3011 pzp->zp_acl.z_acl_extern_obj, 0,
3013 dmu_tx_hold_write(tx, DMU_NEW_OBJECT,
3014 0, aclp->z_acl_bytes);
3016 dmu_tx_hold_write(tx,
3017 pzp->zp_acl.z_acl_extern_obj, 0,
3020 } else if (aclp->z_acl_bytes > ZFS_ACE_SPACE) {
3021 dmu_tx_hold_write(tx, DMU_NEW_OBJECT,
3022 0, aclp->z_acl_bytes);
3026 if (mask & (AT_UID | AT_GID)) {
3027 if (pzp->zp_xattr) {
3028 err = zfs_zget(zp->z_zfsvfs, pzp->zp_xattr, &attrzp);
3031 dmu_tx_hold_bonus(tx, attrzp->z_id);
3033 if (mask & AT_UID) {
3034 new_uid = zfs_fuid_create(zfsvfs,
3035 (uint64_t)vap->va_uid, cr, ZFS_OWNER, &fuidp);
3036 if (new_uid != pzp->zp_uid &&
3037 zfs_usergroup_overquota(zfsvfs, B_FALSE, new_uid)) {
3043 if (mask & AT_GID) {
3044 new_gid = zfs_fuid_create(zfsvfs, (uint64_t)vap->va_gid,
3045 cr, ZFS_GROUP, &fuidp);
3046 if (new_gid != pzp->zp_gid &&
3047 zfs_usergroup_overquota(zfsvfs, B_TRUE, new_gid)) {
3052 fuid_dirtied = zfsvfs->z_fuid_dirty;
3054 if (zfsvfs->z_fuid_obj == 0) {
3055 dmu_tx_hold_bonus(tx, DMU_NEW_OBJECT);
3056 dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0,
3057 FUID_SIZE_ESTIMATE(zfsvfs));
3058 dmu_tx_hold_zap(tx, MASTER_NODE_OBJ,
3061 dmu_tx_hold_bonus(tx, zfsvfs->z_fuid_obj);
3062 dmu_tx_hold_write(tx, zfsvfs->z_fuid_obj, 0,
3063 FUID_SIZE_ESTIMATE(zfsvfs));
3068 err = dmu_tx_assign(tx, TXG_NOWAIT);
3070 if (err == ERESTART)
3075 dmu_buf_will_dirty(zp->z_dbuf, tx);
3078 * Set each attribute requested.
3079 * We group settings according to the locks they need to acquire.
3081 * Note: you cannot set ctime directly, although it will be
3082 * updated as a side-effect of calling this function.
3085 mutex_enter(&zp->z_lock);
3087 if (mask & AT_MODE) {
3088 mutex_enter(&zp->z_acl_lock);
3089 zp->z_phys->zp_mode = new_mode;
3090 err = zfs_aclset_common(zp, aclp, cr, tx);
3091 ASSERT3U(err, ==, 0);
3092 zp->z_acl_cached = aclp;
3094 mutex_exit(&zp->z_acl_lock);
3098 mutex_enter(&attrzp->z_lock);
3100 if (mask & AT_UID) {
3101 pzp->zp_uid = new_uid;
3103 attrzp->z_phys->zp_uid = new_uid;
3106 if (mask & AT_GID) {
3107 pzp->zp_gid = new_gid;
3109 attrzp->z_phys->zp_gid = new_gid;
3113 mutex_exit(&attrzp->z_lock);
3115 if (mask & AT_ATIME)
3116 ZFS_TIME_ENCODE(&vap->va_atime, pzp->zp_atime);
3118 if (mask & AT_MTIME)
3119 ZFS_TIME_ENCODE(&vap->va_mtime, pzp->zp_mtime);
3121 /* XXX - shouldn't this be done *before* the ATIME/MTIME checks? */
3123 zfs_time_stamper_locked(zp, CONTENT_MODIFIED, tx);
3125 zfs_time_stamper_locked(zp, STATE_CHANGED, tx);
3127 * Do this after setting timestamps to prevent timestamp
3128 * update from toggling bit
3131 if (xoap && (mask & AT_XVATTR)) {
3134 * restore trimmed off masks
3135 * so that return masks can be set for caller.
3138 if (XVA_ISSET_REQ(&tmpxvattr, XAT_APPENDONLY)) {
3139 XVA_SET_REQ(xvap, XAT_APPENDONLY);
3141 if (XVA_ISSET_REQ(&tmpxvattr, XAT_NOUNLINK)) {
3142 XVA_SET_REQ(xvap, XAT_NOUNLINK);
3144 if (XVA_ISSET_REQ(&tmpxvattr, XAT_IMMUTABLE)) {
3145 XVA_SET_REQ(xvap, XAT_IMMUTABLE);
3147 if (XVA_ISSET_REQ(&tmpxvattr, XAT_NODUMP)) {
3148 XVA_SET_REQ(xvap, XAT_NODUMP);
3150 if (XVA_ISSET_REQ(&tmpxvattr, XAT_AV_MODIFIED)) {
3151 XVA_SET_REQ(xvap, XAT_AV_MODIFIED);
3153 if (XVA_ISSET_REQ(&tmpxvattr, XAT_AV_QUARANTINED)) {
3154 XVA_SET_REQ(xvap, XAT_AV_QUARANTINED);
3157 if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP)) {
3159 dmu_object_info_t doi;
3161 ASSERT(vp->v_type == VREG);
3163 /* Grow the bonus buffer if necessary. */
3164 dmu_object_info_from_db(zp->z_dbuf, &doi);
3165 len = sizeof (xoap->xoa_av_scanstamp) +
3166 sizeof (znode_phys_t);
3167 if (len > doi.doi_bonus_size)
3168 VERIFY(dmu_set_bonus(zp->z_dbuf, len, tx) == 0);
3170 zfs_xvattr_set(zp, xvap);
3174 zfs_fuid_sync(zfsvfs, tx);
3177 zfs_log_setattr(zilog, tx, TX_SETATTR, zp, vap, mask, fuidp);
3179 mutex_exit(&zp->z_lock);
3183 VN_RELE(ZTOV(attrzp));
3189 zfs_fuid_info_free(fuidp);
3198 if (err == ERESTART)
3205 typedef struct zfs_zlock {
3206 krwlock_t *zl_rwlock; /* lock we acquired */
3207 znode_t *zl_znode; /* znode we held */
3208 struct zfs_zlock *zl_next; /* next in list */
3212 * Drop locks and release vnodes that were held by zfs_rename_lock().
3215 zfs_rename_unlock(zfs_zlock_t **zlpp)
3219 while ((zl = *zlpp) != NULL) {
3220 if (zl->zl_znode != NULL)
3221 VN_RELE(ZTOV(zl->zl_znode));
3222 rw_exit(zl->zl_rwlock);
3223 *zlpp = zl->zl_next;
3224 kmem_free(zl, sizeof (*zl));
3229 * Search back through the directory tree, using the ".." entries.
3230 * Lock each directory in the chain to prevent concurrent renames.
3231 * Fail any attempt to move a directory into one of its own descendants.
3232 * XXX - z_parent_lock can overlap with map or grow locks
3235 zfs_rename_lock(znode_t *szp, znode_t *tdzp, znode_t *sdzp, zfs_zlock_t **zlpp)
3239 uint64_t rootid = zp->z_zfsvfs->z_root;
3240 uint64_t *oidp = &zp->z_id;
3241 krwlock_t *rwlp = &szp->z_parent_lock;
3242 krw_t rw = RW_WRITER;
3245 * First pass write-locks szp and compares to zp->z_id.
3246 * Later passes read-lock zp and compare to zp->z_parent.
3249 if (!rw_tryenter(rwlp, rw)) {
3251 * Another thread is renaming in this path.
3252 * Note that if we are a WRITER, we don't have any
3253 * parent_locks held yet.
3255 if (rw == RW_READER && zp->z_id > szp->z_id) {
3257 * Drop our locks and restart
3259 zfs_rename_unlock(&zl);
3263 rwlp = &szp->z_parent_lock;
3268 * Wait for other thread to drop its locks
3274 zl = kmem_alloc(sizeof (*zl), KM_SLEEP);
3275 zl->zl_rwlock = rwlp;
3276 zl->zl_znode = NULL;
3277 zl->zl_next = *zlpp;
3280 if (*oidp == szp->z_id) /* We're a descendant of szp */
3283 if (*oidp == rootid) /* We've hit the top */
3286 if (rw == RW_READER) { /* i.e. not the first pass */
3287 int error = zfs_zget(zp->z_zfsvfs, *oidp, &zp);
3292 oidp = &zp->z_phys->zp_parent;
3293 rwlp = &zp->z_parent_lock;
3296 } while (zp->z_id != sdzp->z_id);
3302 * Move an entry from the provided source directory to the target
3303 * directory. Change the entry name as indicated.
3305 * IN: sdvp - Source directory containing the "old entry".
3306 * snm - Old entry name.
3307 * tdvp - Target directory to contain the "new entry".
3308 * tnm - New entry name.
3309 * cr - credentials of caller.
3310 * ct - caller context
3311 * flags - case flags
3313 * RETURN: 0 if success
3314 * error code if failure
3317 * sdvp,tdvp - ctime|mtime updated
3321 zfs_rename(vnode_t *sdvp, char *snm, vnode_t *tdvp, char *tnm, cred_t *cr,
3322 caller_context_t *ct, int flags)
3324 znode_t *tdzp, *szp, *tzp;
3325 znode_t *sdzp = VTOZ(sdvp);
3326 zfsvfs_t *zfsvfs = sdzp->z_zfsvfs;
3329 zfs_dirlock_t *sdl, *tdl;
3332 int cmp, serr, terr;
3337 ZFS_VERIFY_ZP(sdzp);
3338 zilog = zfsvfs->z_log;
3341 * Make sure we have the real vp for the target directory.
3343 if (VOP_REALVP(tdvp, &realvp, ct) == 0)
3346 if (tdvp->v_vfsp != sdvp->v_vfsp || zfsctl_is_node(tdvp)) {
3352 ZFS_VERIFY_ZP(tdzp);
3353 if (zfsvfs->z_utf8 && u8_validate(tnm,
3354 strlen(tnm), NULL, U8_VALIDATE_ENTIRE, &error) < 0) {
3359 if (flags & FIGNORECASE)
3368 * This is to prevent the creation of links into attribute space
3369 * by renaming a linked file into/outof an attribute directory.
3370 * See the comment in zfs_link() for why this is considered bad.
3372 if ((tdzp->z_phys->zp_flags & ZFS_XATTR) !=
3373 (sdzp->z_phys->zp_flags & ZFS_XATTR)) {
3379 * Lock source and target directory entries. To prevent deadlock,
3380 * a lock ordering must be defined. We lock the directory with
3381 * the smallest object id first, or if it's a tie, the one with
3382 * the lexically first name.
3384 if (sdzp->z_id < tdzp->z_id) {
3386 } else if (sdzp->z_id > tdzp->z_id) {
3390 * First compare the two name arguments without
3391 * considering any case folding.
3393 int nofold = (zfsvfs->z_norm & ~U8_TEXTPREP_TOUPPER);
3395 cmp = u8_strcmp(snm, tnm, 0, nofold, U8_UNICODE_LATEST, &error);
3396 ASSERT(error == 0 || !zfsvfs->z_utf8);
3399 * POSIX: "If the old argument and the new argument
3400 * both refer to links to the same existing file,
3401 * the rename() function shall return successfully
3402 * and perform no other action."
3408 * If the file system is case-folding, then we may
3409 * have some more checking to do. A case-folding file
3410 * system is either supporting mixed case sensitivity
3411 * access or is completely case-insensitive. Note
3412 * that the file system is always case preserving.
3414 * In mixed sensitivity mode case sensitive behavior
3415 * is the default. FIGNORECASE must be used to
3416 * explicitly request case insensitive behavior.
3418 * If the source and target names provided differ only
3419 * by case (e.g., a request to rename 'tim' to 'Tim'),
3420 * we will treat this as a special case in the
3421 * case-insensitive mode: as long as the source name
3422 * is an exact match, we will allow this to proceed as
3423 * a name-change request.
3425 if ((zfsvfs->z_case == ZFS_CASE_INSENSITIVE ||
3426 (zfsvfs->z_case == ZFS_CASE_MIXED &&
3427 flags & FIGNORECASE)) &&
3428 u8_strcmp(snm, tnm, 0, zfsvfs->z_norm, U8_UNICODE_LATEST,
3431 * case preserving rename request, require exact
3440 * If the source and destination directories are the same, we should
3441 * grab the z_name_lock of that directory only once.
3445 rw_enter(&sdzp->z_name_lock, RW_READER);
3449 serr = zfs_dirent_lock(&sdl, sdzp, snm, &szp,
3450 ZEXISTS | zflg, NULL, NULL);
3451 terr = zfs_dirent_lock(&tdl,
3452 tdzp, tnm, &tzp, ZRENAMING | zflg, NULL, NULL);
3454 terr = zfs_dirent_lock(&tdl,
3455 tdzp, tnm, &tzp, zflg, NULL, NULL);
3456 serr = zfs_dirent_lock(&sdl,
3457 sdzp, snm, &szp, ZEXISTS | ZRENAMING | zflg,
3463 * Source entry invalid or not there.
3466 zfs_dirent_unlock(tdl);
3472 rw_exit(&sdzp->z_name_lock);
3474 if (strcmp(snm, ".") == 0 || strcmp(snm, "..") == 0)
3480 zfs_dirent_unlock(sdl);
3484 rw_exit(&sdzp->z_name_lock);
3486 if (strcmp(tnm, "..") == 0)
3493 * Must have write access at the source to remove the old entry
3494 * and write access at the target to create the new entry.
3495 * Note that if target and source are the same, this can be
3496 * done in a single check.
3499 if (error = zfs_zaccess_rename(sdzp, szp, tdzp, tzp, cr))
3502 if (ZTOV(szp)->v_type == VDIR) {
3504 * Check to make sure rename is valid.
3505 * Can't do a move like this: /usr/a/b to /usr/a/b/c/d
3507 if (error = zfs_rename_lock(szp, tdzp, sdzp, &zl))
3512 * Does target exist?
3516 * Source and target must be the same type.
3518 if (ZTOV(szp)->v_type == VDIR) {
3519 if (ZTOV(tzp)->v_type != VDIR) {
3524 if (ZTOV(tzp)->v_type == VDIR) {
3530 * POSIX dictates that when the source and target
3531 * entries refer to the same file object, rename
3532 * must do nothing and exit without error.
3534 if (szp->z_id == tzp->z_id) {
3540 vnevent_rename_src(ZTOV(szp), sdvp, snm, ct);
3542 vnevent_rename_dest(ZTOV(tzp), tdvp, tnm, ct);
3545 * notify the target directory if it is not the same
3546 * as source directory.
3549 vnevent_rename_dest_dir(tdvp, ct);
3552 tx = dmu_tx_create(zfsvfs->z_os);
3553 dmu_tx_hold_bonus(tx, szp->z_id); /* nlink changes */
3554 dmu_tx_hold_bonus(tx, sdzp->z_id); /* nlink changes */
3555 dmu_tx_hold_zap(tx, sdzp->z_id, FALSE, snm);
3556 dmu_tx_hold_zap(tx, tdzp->z_id, TRUE, tnm);
3558 dmu_tx_hold_bonus(tx, tdzp->z_id); /* nlink changes */
3560 dmu_tx_hold_bonus(tx, tzp->z_id); /* parent changes */
3561 dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL);
3562 error = dmu_tx_assign(tx, TXG_NOWAIT);
3565 zfs_rename_unlock(&zl);
3566 zfs_dirent_unlock(sdl);
3567 zfs_dirent_unlock(tdl);
3570 rw_exit(&sdzp->z_name_lock);
3575 if (error == ERESTART) {
3585 if (tzp) /* Attempt to remove the existing target */
3586 error = zfs_link_destroy(tdl, tzp, tx, zflg, NULL);
3589 error = zfs_link_create(tdl, szp, tx, ZRENAMING);
3591 szp->z_phys->zp_flags |= ZFS_AV_MODIFIED;
3593 error = zfs_link_destroy(sdl, szp, tx, ZRENAMING, NULL);
3596 zfs_log_rename(zilog, tx,
3597 TX_RENAME | (flags & FIGNORECASE ? TX_CI : 0),
3598 sdzp, sdl->dl_name, tdzp, tdl->dl_name, szp);
3600 /* Update path information for the target vnode */
3601 vn_renamepath(tdvp, ZTOV(szp), tnm, strlen(tnm));
3603 #ifdef FREEBSD_NAMECACHE
3614 zfs_rename_unlock(&zl);
3616 zfs_dirent_unlock(sdl);
3617 zfs_dirent_unlock(tdl);
3620 rw_exit(&sdzp->z_name_lock);
3632 * Insert the indicated symbolic reference entry into the directory.
3634 * IN: dvp - Directory to contain new symbolic link.
3635 * link - Name for new symlink entry.
3636 * vap - Attributes of new entry.
3637 * target - Target path of new symlink.
3638 * cr - credentials of caller.
3639 * ct - caller context
3640 * flags - case flags
3642 * RETURN: 0 if success
3643 * error code if failure
3646 * dvp - ctime|mtime updated
3650 zfs_symlink(vnode_t *dvp, vnode_t **vpp, char *name, vattr_t *vap, char *link,
3651 cred_t *cr, kthread_t *td)
3653 znode_t *zp, *dzp = VTOZ(dvp);
3656 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
3658 int len = strlen(link);
3661 zfs_acl_ids_t acl_ids;
3662 boolean_t fuid_dirtied;
3665 ASSERT(vap->va_type == VLNK);
3669 zilog = zfsvfs->z_log;
3671 if (zfsvfs->z_utf8 && u8_validate(name, strlen(name),
3672 NULL, U8_VALIDATE_ENTIRE, &error) < 0) {
3676 if (flags & FIGNORECASE)
3679 if (error = zfs_zaccess(dzp, ACE_ADD_FILE, 0, B_FALSE, cr)) {
3684 if (len > MAXPATHLEN) {
3686 return (ENAMETOOLONG);
3690 * Attempt to lock directory; fail if entry already exists.
3692 error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg, NULL, NULL);
3698 VERIFY(0 == zfs_acl_ids_create(dzp, 0, vap, cr, NULL, &acl_ids));
3699 if (zfs_acl_ids_overquota(zfsvfs, &acl_ids)) {
3700 zfs_acl_ids_free(&acl_ids);
3701 zfs_dirent_unlock(dl);
3705 tx = dmu_tx_create(zfsvfs->z_os);
3706 fuid_dirtied = zfsvfs->z_fuid_dirty;
3707 dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, MAX(1, len));
3708 dmu_tx_hold_bonus(tx, dzp->z_id);
3709 dmu_tx_hold_zap(tx, dzp->z_id, TRUE, name);
3710 if (acl_ids.z_aclp->z_acl_bytes > ZFS_ACE_SPACE)
3711 dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, SPA_MAXBLOCKSIZE);
3713 zfs_fuid_txhold(zfsvfs, tx);
3714 error = dmu_tx_assign(tx, TXG_NOWAIT);
3716 zfs_acl_ids_free(&acl_ids);
3717 zfs_dirent_unlock(dl);
3718 if (error == ERESTART) {
3728 dmu_buf_will_dirty(dzp->z_dbuf, tx);
3731 * Create a new object for the symlink.
3732 * Put the link content into bonus buffer if it will fit;
3733 * otherwise, store it just like any other file data.
3735 if (sizeof (znode_phys_t) + len <= dmu_bonus_max()) {
3736 zfs_mknode(dzp, vap, tx, cr, 0, &zp, len, &acl_ids);
3738 bcopy(link, zp->z_phys + 1, len);
3742 zfs_mknode(dzp, vap, tx, cr, 0, &zp, 0, &acl_ids);
3745 zfs_fuid_sync(zfsvfs, tx);
3747 * Nothing can access the znode yet so no locking needed
3748 * for growing the znode's blocksize.
3750 zfs_grow_blocksize(zp, len, tx);
3752 VERIFY(0 == dmu_buf_hold(zfsvfs->z_os,
3753 zp->z_id, 0, FTAG, &dbp));
3754 dmu_buf_will_dirty(dbp, tx);
3756 ASSERT3U(len, <=, dbp->db_size);
3757 bcopy(link, dbp->db_data, len);
3758 dmu_buf_rele(dbp, FTAG);
3760 zp->z_phys->zp_size = len;
3763 * Insert the new object into the directory.
3765 (void) zfs_link_create(dl, zp, tx, ZNEW);
3767 uint64_t txtype = TX_SYMLINK;
3768 if (flags & FIGNORECASE)
3770 zfs_log_symlink(zilog, tx, txtype, dzp, zp, name, link);
3774 zfs_acl_ids_free(&acl_ids);
3778 zfs_dirent_unlock(dl);
3785 * Return, in the buffer contained in the provided uio structure,
3786 * the symbolic path referred to by vp.
3788 * IN: vp - vnode of symbolic link.
3789 * uoip - structure to contain the link path.
3790 * cr - credentials of caller.
3791 * ct - caller context
3793 * OUT: uio - structure to contain the link path.
3795 * RETURN: 0 if success
3796 * error code if failure
3799 * vp - atime updated
3803 zfs_readlink(vnode_t *vp, uio_t *uio, cred_t *cr, caller_context_t *ct)
3805 znode_t *zp = VTOZ(vp);
3806 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
3813 bufsz = (size_t)zp->z_phys->zp_size;
3814 if (bufsz + sizeof (znode_phys_t) <= zp->z_dbuf->db_size) {
3815 error = uiomove(zp->z_phys + 1,
3816 MIN((size_t)bufsz, uio->uio_resid), UIO_READ, uio);
3819 error = dmu_buf_hold(zfsvfs->z_os, zp->z_id, 0, FTAG, &dbp);
3824 error = uiomove(dbp->db_data,
3825 MIN((size_t)bufsz, uio->uio_resid), UIO_READ, uio);
3826 dmu_buf_rele(dbp, FTAG);
3829 ZFS_ACCESSTIME_STAMP(zfsvfs, zp);
3835 * Insert a new entry into directory tdvp referencing svp.
3837 * IN: tdvp - Directory to contain new entry.
3838 * svp - vnode of new entry.
3839 * name - name of new entry.
3840 * cr - credentials of caller.
3841 * ct - caller context
3843 * RETURN: 0 if success
3844 * error code if failure
3847 * tdvp - ctime|mtime updated
3848 * svp - ctime updated
3852 zfs_link(vnode_t *tdvp, vnode_t *svp, char *name, cred_t *cr,
3853 caller_context_t *ct, int flags)
3855 znode_t *dzp = VTOZ(tdvp);
3857 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
3867 ASSERT(tdvp->v_type == VDIR);
3871 zilog = zfsvfs->z_log;
3873 if (VOP_REALVP(svp, &realvp, ct) == 0)
3877 * POSIX dictates that we return EPERM here.
3878 * Better choices include ENOTSUP or EISDIR.
3880 if (svp->v_type == VDIR) {
3885 if (svp->v_vfsp != tdvp->v_vfsp || zfsctl_is_node(svp)) {
3893 /* Prevent links to .zfs/shares files */
3895 if (szp->z_phys->zp_parent == zfsvfs->z_shares_dir) {
3900 if (zfsvfs->z_utf8 && u8_validate(name,
3901 strlen(name), NULL, U8_VALIDATE_ENTIRE, &error) < 0) {
3905 if (flags & FIGNORECASE)
3909 * We do not support links between attributes and non-attributes
3910 * because of the potential security risk of creating links
3911 * into "normal" file space in order to circumvent restrictions
3912 * imposed in attribute space.
3914 if ((szp->z_phys->zp_flags & ZFS_XATTR) !=
3915 (dzp->z_phys->zp_flags & ZFS_XATTR)) {
3921 owner = zfs_fuid_map_id(zfsvfs, szp->z_phys->zp_uid, cr, ZFS_OWNER);
3922 if (owner != crgetuid(cr) &&
3923 secpolicy_basic_link(svp, cr) != 0) {
3928 if (error = zfs_zaccess(dzp, ACE_ADD_FILE, 0, B_FALSE, cr)) {
3935 * Attempt to lock directory; fail if entry already exists.
3937 error = zfs_dirent_lock(&dl, dzp, name, &tzp, zf, NULL, NULL);
3943 tx = dmu_tx_create(zfsvfs->z_os);
3944 dmu_tx_hold_bonus(tx, szp->z_id);
3945 dmu_tx_hold_zap(tx, dzp->z_id, TRUE, name);
3946 error = dmu_tx_assign(tx, TXG_NOWAIT);
3948 zfs_dirent_unlock(dl);
3949 if (error == ERESTART) {
3959 error = zfs_link_create(dl, szp, tx, 0);
3962 uint64_t txtype = TX_LINK;
3963 if (flags & FIGNORECASE)
3965 zfs_log_link(zilog, tx, txtype, dzp, szp, name);
3970 zfs_dirent_unlock(dl);
3973 vnevent_link(svp, ct);
3982 zfs_inactive(vnode_t *vp, cred_t *cr, caller_context_t *ct)
3984 znode_t *zp = VTOZ(vp);
3985 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
3988 rw_enter(&zfsvfs->z_teardown_inactive_lock, RW_READER);
3989 if (zp->z_dbuf == NULL) {
3991 * The fs has been unmounted, or we did a
3992 * suspend/resume and this file no longer exists.
3995 vp->v_count = 0; /* count arrives as 1 */
3997 vrecycle(vp, curthread);
3998 rw_exit(&zfsvfs->z_teardown_inactive_lock);
4002 if (zp->z_atime_dirty && zp->z_unlinked == 0) {
4003 dmu_tx_t *tx = dmu_tx_create(zfsvfs->z_os);
4005 dmu_tx_hold_bonus(tx, zp->z_id);
4006 error = dmu_tx_assign(tx, TXG_WAIT);
4010 dmu_buf_will_dirty(zp->z_dbuf, tx);
4011 mutex_enter(&zp->z_lock);
4012 zp->z_atime_dirty = 0;
4013 mutex_exit(&zp->z_lock);
4019 rw_exit(&zfsvfs->z_teardown_inactive_lock);
4022 CTASSERT(sizeof(struct zfid_short) <= sizeof(struct fid));
4023 CTASSERT(sizeof(struct zfid_long) <= sizeof(struct fid));
4027 zfs_fid(vnode_t *vp, fid_t *fidp, caller_context_t *ct)
4029 znode_t *zp = VTOZ(vp);
4030 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
4032 uint64_t object = zp->z_id;
4038 gen = (uint32_t)zp->z_gen;
4040 size = (zfsvfs->z_parent != zfsvfs) ? LONG_FID_LEN : SHORT_FID_LEN;
4041 fidp->fid_len = size;
4043 zfid = (zfid_short_t *)fidp;
4045 zfid->zf_len = size;
4047 for (i = 0; i < sizeof (zfid->zf_object); i++)
4048 zfid->zf_object[i] = (uint8_t)(object >> (8 * i));
4050 /* Must have a non-zero generation number to distinguish from .zfs */
4053 for (i = 0; i < sizeof (zfid->zf_gen); i++)
4054 zfid->zf_gen[i] = (uint8_t)(gen >> (8 * i));
4056 if (size == LONG_FID_LEN) {
4057 uint64_t objsetid = dmu_objset_id(zfsvfs->z_os);
4060 zlfid = (zfid_long_t *)fidp;
4062 for (i = 0; i < sizeof (zlfid->zf_setid); i++)
4063 zlfid->zf_setid[i] = (uint8_t)(objsetid >> (8 * i));
4065 /* XXX - this should be the generation number for the objset */
4066 for (i = 0; i < sizeof (zlfid->zf_setgen); i++)
4067 zlfid->zf_setgen[i] = 0;
4075 zfs_pathconf(vnode_t *vp, int cmd, ulong_t *valp, cred_t *cr,
4076 caller_context_t *ct)
4088 case _PC_FILESIZEBITS:
4093 case _PC_XATTR_EXISTS:
4095 zfsvfs = zp->z_zfsvfs;
4099 error = zfs_dirent_lock(&dl, zp, "", &xzp,
4100 ZXATTR | ZEXISTS | ZSHARED, NULL, NULL);
4102 zfs_dirent_unlock(dl);
4103 if (!zfs_dirempty(xzp))
4106 } else if (error == ENOENT) {
4108 * If there aren't extended attributes, it's the
4109 * same as having zero of them.
4117 case _PC_ACL_EXTENDED:
4125 case _PC_ACL_PATH_MAX:
4126 *valp = ACL_MAX_ENTRIES;
4129 case _PC_MIN_HOLE_SIZE:
4130 *valp = (int)SPA_MINBLOCKSIZE;
4134 return (EOPNOTSUPP);
4140 zfs_getsecattr(vnode_t *vp, vsecattr_t *vsecp, int flag, cred_t *cr,
4141 caller_context_t *ct)
4143 znode_t *zp = VTOZ(vp);
4144 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
4146 boolean_t skipaclchk = (flag & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE;
4150 error = zfs_getacl(zp, vsecp, skipaclchk, cr);
4158 zfs_setsecattr(vnode_t *vp, vsecattr_t *vsecp, int flag, cred_t *cr,
4159 caller_context_t *ct)
4161 znode_t *zp = VTOZ(vp);
4162 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
4164 boolean_t skipaclchk = (flag & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE;
4168 error = zfs_setacl(zp, vsecp, skipaclchk, cr);
4174 zfs_freebsd_open(ap)
4175 struct vop_open_args /* {
4178 struct ucred *a_cred;
4179 struct thread *a_td;
4182 vnode_t *vp = ap->a_vp;
4183 znode_t *zp = VTOZ(vp);
4186 error = zfs_open(&vp, ap->a_mode, ap->a_cred, NULL);
4188 vnode_create_vobject(vp, zp->z_phys->zp_size, ap->a_td);
4193 zfs_freebsd_close(ap)
4194 struct vop_close_args /* {
4197 struct ucred *a_cred;
4198 struct thread *a_td;
4202 return (zfs_close(ap->a_vp, ap->a_fflag, 0, 0, ap->a_cred, NULL));
4206 zfs_freebsd_ioctl(ap)
4207 struct vop_ioctl_args /* {
4217 return (zfs_ioctl(ap->a_vp, ap->a_command, (intptr_t)ap->a_data,
4218 ap->a_fflag, ap->a_cred, NULL, NULL));
4222 zfs_freebsd_read(ap)
4223 struct vop_read_args /* {
4227 struct ucred *a_cred;
4231 return (zfs_read(ap->a_vp, ap->a_uio, ap->a_ioflag, ap->a_cred, NULL));
4235 zfs_freebsd_write(ap)
4236 struct vop_write_args /* {
4240 struct ucred *a_cred;
4244 return (zfs_write(ap->a_vp, ap->a_uio, ap->a_ioflag, ap->a_cred, NULL));
4248 zfs_freebsd_access(ap)
4249 struct vop_access_args /* {
4251 accmode_t a_accmode;
4252 struct ucred *a_cred;
4253 struct thread *a_td;
4260 * ZFS itself only knowns about VREAD, VWRITE, VEXEC and VAPPEND,
4262 accmode = ap->a_accmode & (VREAD|VWRITE|VEXEC|VAPPEND);
4264 error = zfs_access(ap->a_vp, accmode, 0, ap->a_cred, NULL);
4267 * VADMIN has to be handled by vaccess().
4270 accmode = ap->a_accmode & ~(VREAD|VWRITE|VEXEC|VAPPEND);
4272 vnode_t *vp = ap->a_vp;
4273 znode_t *zp = VTOZ(vp);
4274 znode_phys_t *zphys = zp->z_phys;
4276 error = vaccess(vp->v_type, zphys->zp_mode,
4277 zphys->zp_uid, zphys->zp_gid, accmode, ap->a_cred,
4286 zfs_freebsd_lookup(ap)
4287 struct vop_lookup_args /* {
4288 struct vnode *a_dvp;
4289 struct vnode **a_vpp;
4290 struct componentname *a_cnp;
4293 struct componentname *cnp = ap->a_cnp;
4294 char nm[NAME_MAX + 1];
4296 ASSERT(cnp->cn_namelen < sizeof(nm));
4297 strlcpy(nm, cnp->cn_nameptr, MIN(cnp->cn_namelen + 1, sizeof(nm)));
4299 return (zfs_lookup(ap->a_dvp, nm, ap->a_vpp, cnp, cnp->cn_nameiop,
4300 cnp->cn_cred, cnp->cn_thread, 0));
4304 zfs_freebsd_create(ap)
4305 struct vop_create_args /* {
4306 struct vnode *a_dvp;
4307 struct vnode **a_vpp;
4308 struct componentname *a_cnp;
4309 struct vattr *a_vap;
4312 struct componentname *cnp = ap->a_cnp;
4313 vattr_t *vap = ap->a_vap;
4316 ASSERT(cnp->cn_flags & SAVENAME);
4318 vattr_init_mask(vap);
4319 mode = vap->va_mode & ALLPERMS;
4321 return (zfs_create(ap->a_dvp, cnp->cn_nameptr, vap, !EXCL, mode,
4322 ap->a_vpp, cnp->cn_cred, cnp->cn_thread));
4326 zfs_freebsd_remove(ap)
4327 struct vop_remove_args /* {
4328 struct vnode *a_dvp;
4330 struct componentname *a_cnp;
4334 ASSERT(ap->a_cnp->cn_flags & SAVENAME);
4336 return (zfs_remove(ap->a_dvp, ap->a_cnp->cn_nameptr,
4337 ap->a_cnp->cn_cred, NULL, 0));
4341 zfs_freebsd_mkdir(ap)
4342 struct vop_mkdir_args /* {
4343 struct vnode *a_dvp;
4344 struct vnode **a_vpp;
4345 struct componentname *a_cnp;
4346 struct vattr *a_vap;
4349 vattr_t *vap = ap->a_vap;
4351 ASSERT(ap->a_cnp->cn_flags & SAVENAME);
4353 vattr_init_mask(vap);
4355 return (zfs_mkdir(ap->a_dvp, ap->a_cnp->cn_nameptr, vap, ap->a_vpp,
4356 ap->a_cnp->cn_cred, NULL, 0, NULL));
4360 zfs_freebsd_rmdir(ap)
4361 struct vop_rmdir_args /* {
4362 struct vnode *a_dvp;
4364 struct componentname *a_cnp;
4367 struct componentname *cnp = ap->a_cnp;
4369 ASSERT(cnp->cn_flags & SAVENAME);
4371 return (zfs_rmdir(ap->a_dvp, cnp->cn_nameptr, NULL, cnp->cn_cred, NULL, 0));
4375 zfs_freebsd_readdir(ap)
4376 struct vop_readdir_args /* {
4379 struct ucred *a_cred;
4386 return (zfs_readdir(ap->a_vp, ap->a_uio, ap->a_cred, ap->a_eofflag,
4387 ap->a_ncookies, ap->a_cookies));
4391 zfs_freebsd_fsync(ap)
4392 struct vop_fsync_args /* {
4395 struct thread *a_td;
4400 return (zfs_fsync(ap->a_vp, 0, ap->a_td->td_ucred, NULL));
4404 zfs_freebsd_getattr(ap)
4405 struct vop_getattr_args /* {
4407 struct vattr *a_vap;
4408 struct ucred *a_cred;
4409 struct thread *a_td;
4412 vattr_t *vap = ap->a_vap;
4418 xvap.xva_vattr = *vap;
4419 xvap.xva_vattr.va_mask |= AT_XVATTR;
4421 /* Convert chflags into ZFS-type flags. */
4422 /* XXX: what about SF_SETTABLE?. */
4423 XVA_SET_REQ(&xvap, XAT_IMMUTABLE);
4424 XVA_SET_REQ(&xvap, XAT_APPENDONLY);
4425 XVA_SET_REQ(&xvap, XAT_NOUNLINK);
4426 XVA_SET_REQ(&xvap, XAT_NODUMP);
4427 error = zfs_getattr(ap->a_vp, (vattr_t *)&xvap, 0, ap->a_cred, NULL);
4431 /* Convert ZFS xattr into chflags. */
4432 #define FLAG_CHECK(fflag, xflag, xfield) do { \
4433 if (XVA_ISSET_RTN(&xvap, (xflag)) && (xfield) != 0) \
4434 fflags |= (fflag); \
4436 FLAG_CHECK(SF_IMMUTABLE, XAT_IMMUTABLE,
4437 xvap.xva_xoptattrs.xoa_immutable);
4438 FLAG_CHECK(SF_APPEND, XAT_APPENDONLY,
4439 xvap.xva_xoptattrs.xoa_appendonly);
4440 FLAG_CHECK(SF_NOUNLINK, XAT_NOUNLINK,
4441 xvap.xva_xoptattrs.xoa_nounlink);
4442 FLAG_CHECK(UF_NODUMP, XAT_NODUMP,
4443 xvap.xva_xoptattrs.xoa_nodump);
4445 *vap = xvap.xva_vattr;
4446 vap->va_flags = fflags;
4451 zfs_freebsd_setattr(ap)
4452 struct vop_setattr_args /* {
4454 struct vattr *a_vap;
4455 struct ucred *a_cred;
4456 struct thread *a_td;
4459 vnode_t *vp = ap->a_vp;
4460 vattr_t *vap = ap->a_vap;
4461 cred_t *cred = ap->a_cred;
4466 vattr_init_mask(vap);
4467 vap->va_mask &= ~AT_NOSET;
4470 xvap.xva_vattr = *vap;
4472 zflags = VTOZ(vp)->z_phys->zp_flags;
4474 if (vap->va_flags != VNOVAL) {
4475 zfsvfs_t *zfsvfs = VTOZ(vp)->z_zfsvfs;
4478 if (zfsvfs->z_use_fuids == B_FALSE)
4479 return (EOPNOTSUPP);
4481 fflags = vap->va_flags;
4482 if ((fflags & ~(SF_IMMUTABLE|SF_APPEND|SF_NOUNLINK|UF_NODUMP)) != 0)
4483 return (EOPNOTSUPP);
4485 * Unprivileged processes are not permitted to unset system
4486 * flags, or modify flags if any system flags are set.
4487 * Privileged non-jail processes may not modify system flags
4488 * if securelevel > 0 and any existing system flags are set.
4489 * Privileged jail processes behave like privileged non-jail
4490 * processes if the security.jail.chflags_allowed sysctl is
4491 * is non-zero; otherwise, they behave like unprivileged
4494 if (secpolicy_fs_owner(vp->v_mount, cred) == 0 ||
4495 priv_check_cred(cred, PRIV_VFS_SYSFLAGS, 0) == 0) {
4497 (ZFS_IMMUTABLE | ZFS_APPENDONLY | ZFS_NOUNLINK)) {
4498 error = securelevel_gt(cred, 0);
4504 * Callers may only modify the file flags on objects they
4505 * have VADMIN rights for.
4507 if ((error = VOP_ACCESS(vp, VADMIN, cred, curthread)) != 0)
4510 (ZFS_IMMUTABLE | ZFS_APPENDONLY | ZFS_NOUNLINK)) {
4514 (SF_IMMUTABLE | SF_APPEND | SF_NOUNLINK)) {
4519 #define FLAG_CHANGE(fflag, zflag, xflag, xfield) do { \
4520 if (((fflags & (fflag)) && !(zflags & (zflag))) || \
4521 ((zflags & (zflag)) && !(fflags & (fflag)))) { \
4522 XVA_SET_REQ(&xvap, (xflag)); \
4523 (xfield) = ((fflags & (fflag)) != 0); \
4526 /* Convert chflags into ZFS-type flags. */
4527 /* XXX: what about SF_SETTABLE?. */
4528 FLAG_CHANGE(SF_IMMUTABLE, ZFS_IMMUTABLE, XAT_IMMUTABLE,
4529 xvap.xva_xoptattrs.xoa_immutable);
4530 FLAG_CHANGE(SF_APPEND, ZFS_APPENDONLY, XAT_APPENDONLY,
4531 xvap.xva_xoptattrs.xoa_appendonly);
4532 FLAG_CHANGE(SF_NOUNLINK, ZFS_NOUNLINK, XAT_NOUNLINK,
4533 xvap.xva_xoptattrs.xoa_nounlink);
4534 FLAG_CHANGE(UF_NODUMP, ZFS_NODUMP, XAT_NODUMP,
4535 xvap.xva_xoptattrs.xoa_nodump);
4538 return (zfs_setattr(vp, (vattr_t *)&xvap, 0, cred, NULL));
4542 zfs_freebsd_rename(ap)
4543 struct vop_rename_args /* {
4544 struct vnode *a_fdvp;
4545 struct vnode *a_fvp;
4546 struct componentname *a_fcnp;
4547 struct vnode *a_tdvp;
4548 struct vnode *a_tvp;
4549 struct componentname *a_tcnp;
4552 vnode_t *fdvp = ap->a_fdvp;
4553 vnode_t *fvp = ap->a_fvp;
4554 vnode_t *tdvp = ap->a_tdvp;
4555 vnode_t *tvp = ap->a_tvp;
4558 ASSERT(ap->a_fcnp->cn_flags & (SAVENAME|SAVESTART));
4559 ASSERT(ap->a_tcnp->cn_flags & (SAVENAME|SAVESTART));
4561 error = zfs_rename(fdvp, ap->a_fcnp->cn_nameptr, tdvp,
4562 ap->a_tcnp->cn_nameptr, ap->a_fcnp->cn_cred, NULL, 0);
4577 zfs_freebsd_symlink(ap)
4578 struct vop_symlink_args /* {
4579 struct vnode *a_dvp;
4580 struct vnode **a_vpp;
4581 struct componentname *a_cnp;
4582 struct vattr *a_vap;
4586 struct componentname *cnp = ap->a_cnp;
4587 vattr_t *vap = ap->a_vap;
4589 ASSERT(cnp->cn_flags & SAVENAME);
4591 vap->va_type = VLNK; /* FreeBSD: Syscall only sets va_mode. */
4592 vattr_init_mask(vap);
4594 return (zfs_symlink(ap->a_dvp, ap->a_vpp, cnp->cn_nameptr, vap,
4595 ap->a_target, cnp->cn_cred, cnp->cn_thread));
4599 zfs_freebsd_readlink(ap)
4600 struct vop_readlink_args /* {
4603 struct ucred *a_cred;
4607 return (zfs_readlink(ap->a_vp, ap->a_uio, ap->a_cred, NULL));
4611 zfs_freebsd_link(ap)
4612 struct vop_link_args /* {
4613 struct vnode *a_tdvp;
4615 struct componentname *a_cnp;
4618 struct componentname *cnp = ap->a_cnp;
4620 ASSERT(cnp->cn_flags & SAVENAME);
4622 return (zfs_link(ap->a_tdvp, ap->a_vp, cnp->cn_nameptr, cnp->cn_cred, NULL, 0));
4626 zfs_freebsd_inactive(ap)
4627 struct vop_inactive_args /* {
4629 struct thread *a_td;
4632 vnode_t *vp = ap->a_vp;
4634 zfs_inactive(vp, ap->a_td->td_ucred, NULL);
4639 zfs_reclaim_complete(void *arg, int pending)
4642 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
4644 rw_enter(&zfsvfs->z_teardown_inactive_lock, RW_READER);
4645 if (zp->z_dbuf != NULL) {
4646 ZFS_OBJ_HOLD_ENTER(zfsvfs, zp->z_id);
4647 zfs_znode_dmu_fini(zp);
4648 ZFS_OBJ_HOLD_EXIT(zfsvfs, zp->z_id);
4651 rw_exit(&zfsvfs->z_teardown_inactive_lock);
4653 * If the file system is being unmounted, there is a process waiting
4654 * for us, wake it up.
4656 if (zfsvfs->z_unmounted)
4661 zfs_freebsd_reclaim(ap)
4662 struct vop_reclaim_args /* {
4664 struct thread *a_td;
4667 vnode_t *vp = ap->a_vp;
4668 znode_t *zp = VTOZ(vp);
4669 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
4671 rw_enter(&zfsvfs->z_teardown_inactive_lock, RW_READER);
4676 * Destroy the vm object and flush associated pages.
4678 vnode_destroy_vobject(vp);
4680 mutex_enter(&zp->z_lock);
4681 ASSERT(zp->z_phys != NULL);
4683 mutex_exit(&zp->z_lock);
4687 else if (zp->z_dbuf == NULL)
4689 else /* if (!zp->z_unlinked && zp->z_dbuf != NULL) */ {
4692 locked = MUTEX_HELD(ZFS_OBJ_MUTEX(zfsvfs, zp->z_id)) ? 2 :
4693 ZFS_OBJ_HOLD_TRYENTER(zfsvfs, zp->z_id);
4696 * Lock can't be obtained due to deadlock possibility,
4697 * so defer znode destruction.
4699 TASK_INIT(&zp->z_task, 0, zfs_reclaim_complete, zp);
4700 taskqueue_enqueue(taskqueue_thread, &zp->z_task);
4702 zfs_znode_dmu_fini(zp);
4704 ZFS_OBJ_HOLD_EXIT(zfsvfs, zp->z_id);
4710 ASSERT(vp->v_holdcnt >= 1);
4712 rw_exit(&zfsvfs->z_teardown_inactive_lock);
4718 struct vop_fid_args /* {
4724 return (zfs_fid(ap->a_vp, (void *)ap->a_fid, NULL));
4728 zfs_freebsd_pathconf(ap)
4729 struct vop_pathconf_args /* {
4732 register_t *a_retval;
4738 error = zfs_pathconf(ap->a_vp, ap->a_name, &val, curthread->td_ucred, NULL);
4740 *ap->a_retval = val;
4741 else if (error == EOPNOTSUPP)
4742 error = vop_stdpathconf(ap);
4747 zfs_freebsd_fifo_pathconf(ap)
4748 struct vop_pathconf_args /* {
4751 register_t *a_retval;
4755 switch (ap->a_name) {
4756 case _PC_ACL_EXTENDED:
4758 case _PC_ACL_PATH_MAX:
4759 case _PC_MAC_PRESENT:
4760 return (zfs_freebsd_pathconf(ap));
4762 return (fifo_specops.vop_pathconf(ap));
4767 * FreeBSD's extended attributes namespace defines file name prefix for ZFS'
4768 * extended attribute name:
4771 * system freebsd:system:
4772 * user (none, can be used to access ZFS fsattr(5) attributes
4773 * created on Solaris)
4776 zfs_create_attrname(int attrnamespace, const char *name, char *attrname,
4779 const char *namespace, *prefix, *suffix;
4781 /* We don't allow '/' character in attribute name. */
4782 if (strchr(name, '/') != NULL)
4784 /* We don't allow attribute names that start with "freebsd:" string. */
4785 if (strncmp(name, "freebsd:", 8) == 0)
4788 bzero(attrname, size);
4790 switch (attrnamespace) {
4791 case EXTATTR_NAMESPACE_USER:
4793 prefix = "freebsd:";
4794 namespace = EXTATTR_NAMESPACE_USER_STRING;
4798 * This is the default namespace by which we can access all
4799 * attributes created on Solaris.
4801 prefix = namespace = suffix = "";
4804 case EXTATTR_NAMESPACE_SYSTEM:
4805 prefix = "freebsd:";
4806 namespace = EXTATTR_NAMESPACE_SYSTEM_STRING;
4809 case EXTATTR_NAMESPACE_EMPTY:
4813 if (snprintf(attrname, size, "%s%s%s%s", prefix, namespace, suffix,
4815 return (ENAMETOOLONG);
4821 * Vnode operating to retrieve a named extended attribute.
4824 zfs_getextattr(struct vop_getextattr_args *ap)
4827 IN struct vnode *a_vp;
4828 IN int a_attrnamespace;
4829 IN const char *a_name;
4830 INOUT struct uio *a_uio;
4832 IN struct ucred *a_cred;
4833 IN struct thread *a_td;
4837 zfsvfs_t *zfsvfs = VTOZ(ap->a_vp)->z_zfsvfs;
4838 struct thread *td = ap->a_td;
4839 struct nameidata nd;
4842 vnode_t *xvp = NULL, *vp;
4845 error = extattr_check_cred(ap->a_vp, ap->a_attrnamespace,
4846 ap->a_cred, ap->a_td, VREAD);
4850 error = zfs_create_attrname(ap->a_attrnamespace, ap->a_name, attrname,
4857 error = zfs_lookup(ap->a_vp, NULL, &xvp, NULL, 0, ap->a_cred, td,
4865 NDINIT_ATVP(&nd, LOOKUP, NOFOLLOW | MPSAFE, UIO_SYSSPACE, attrname,
4867 error = vn_open_cred(&nd, &flags, 0, 0, ap->a_cred, NULL);
4869 NDFREE(&nd, NDF_ONLY_PNBUF);
4872 if (error == ENOENT)
4877 if (ap->a_size != NULL) {
4878 error = VOP_GETATTR(vp, &va, ap->a_cred);
4880 *ap->a_size = (size_t)va.va_size;
4881 } else if (ap->a_uio != NULL)
4882 error = VOP_READ(vp, ap->a_uio, IO_UNIT | IO_SYNC, ap->a_cred);
4885 vn_close(vp, flags, ap->a_cred, td);
4892 * Vnode operation to remove a named attribute.
4895 zfs_deleteextattr(struct vop_deleteextattr_args *ap)
4898 IN struct vnode *a_vp;
4899 IN int a_attrnamespace;
4900 IN const char *a_name;
4901 IN struct ucred *a_cred;
4902 IN struct thread *a_td;
4906 zfsvfs_t *zfsvfs = VTOZ(ap->a_vp)->z_zfsvfs;
4907 struct thread *td = ap->a_td;
4908 struct nameidata nd;
4911 vnode_t *xvp = NULL, *vp;
4914 error = extattr_check_cred(ap->a_vp, ap->a_attrnamespace,
4915 ap->a_cred, ap->a_td, VWRITE);
4919 error = zfs_create_attrname(ap->a_attrnamespace, ap->a_name, attrname,
4926 error = zfs_lookup(ap->a_vp, NULL, &xvp, NULL, 0, ap->a_cred, td,
4933 NDINIT_ATVP(&nd, DELETE, NOFOLLOW | LOCKPARENT | LOCKLEAF | MPSAFE,
4934 UIO_SYSSPACE, attrname, xvp, td);
4937 NDFREE(&nd, NDF_ONLY_PNBUF);
4940 if (error == ENOENT)
4944 error = VOP_REMOVE(nd.ni_dvp, vp, &nd.ni_cnd);
4947 if (vp == nd.ni_dvp)
4957 * Vnode operation to set a named attribute.
4960 zfs_setextattr(struct vop_setextattr_args *ap)
4963 IN struct vnode *a_vp;
4964 IN int a_attrnamespace;
4965 IN const char *a_name;
4966 INOUT struct uio *a_uio;
4967 IN struct ucred *a_cred;
4968 IN struct thread *a_td;
4972 zfsvfs_t *zfsvfs = VTOZ(ap->a_vp)->z_zfsvfs;
4973 struct thread *td = ap->a_td;
4974 struct nameidata nd;
4977 vnode_t *xvp = NULL, *vp;
4980 error = extattr_check_cred(ap->a_vp, ap->a_attrnamespace,
4981 ap->a_cred, ap->a_td, VWRITE);
4985 error = zfs_create_attrname(ap->a_attrnamespace, ap->a_name, attrname,
4992 error = zfs_lookup(ap->a_vp, NULL, &xvp, NULL, 0, ap->a_cred, td,
4993 LOOKUP_XATTR | CREATE_XATTR_DIR);
4999 flags = FFLAGS(O_WRONLY | O_CREAT);
5000 NDINIT_ATVP(&nd, LOOKUP, NOFOLLOW | MPSAFE, UIO_SYSSPACE, attrname,
5002 error = vn_open_cred(&nd, &flags, 0600, 0, ap->a_cred, NULL);
5004 NDFREE(&nd, NDF_ONLY_PNBUF);
5012 error = VOP_SETATTR(vp, &va, ap->a_cred);
5014 VOP_WRITE(vp, ap->a_uio, IO_UNIT | IO_SYNC, ap->a_cred);
5017 vn_close(vp, flags, ap->a_cred, td);
5024 * Vnode operation to retrieve extended attributes on a vnode.
5027 zfs_listextattr(struct vop_listextattr_args *ap)
5030 IN struct vnode *a_vp;
5031 IN int a_attrnamespace;
5032 INOUT struct uio *a_uio;
5034 IN struct ucred *a_cred;
5035 IN struct thread *a_td;
5039 zfsvfs_t *zfsvfs = VTOZ(ap->a_vp)->z_zfsvfs;
5040 struct thread *td = ap->a_td;
5041 struct nameidata nd;
5042 char attrprefix[16];
5043 u_char dirbuf[sizeof(struct dirent)];
5046 struct uio auio, *uio = ap->a_uio;
5047 size_t *sizep = ap->a_size;
5049 vnode_t *xvp = NULL, *vp;
5050 int done, error, eof, pos;
5052 error = extattr_check_cred(ap->a_vp, ap->a_attrnamespace,
5053 ap->a_cred, ap->a_td, VREAD);
5057 error = zfs_create_attrname(ap->a_attrnamespace, "", attrprefix,
5058 sizeof(attrprefix));
5061 plen = strlen(attrprefix);
5068 error = zfs_lookup(ap->a_vp, NULL, &xvp, NULL, 0, ap->a_cred, td,
5073 * ENOATTR means that the EA directory does not yet exist,
5074 * i.e. there are no extended attributes there.
5076 if (error == ENOATTR)
5081 NDINIT_ATVP(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | LOCKSHARED | MPSAFE,
5082 UIO_SYSSPACE, ".", xvp, td);
5085 NDFREE(&nd, NDF_ONLY_PNBUF);
5091 auio.uio_iov = &aiov;
5092 auio.uio_iovcnt = 1;
5093 auio.uio_segflg = UIO_SYSSPACE;
5095 auio.uio_rw = UIO_READ;
5096 auio.uio_offset = 0;
5101 aiov.iov_base = (void *)dirbuf;
5102 aiov.iov_len = sizeof(dirbuf);
5103 auio.uio_resid = sizeof(dirbuf);
5104 error = VOP_READDIR(vp, &auio, ap->a_cred, &eof, NULL, NULL);
5105 done = sizeof(dirbuf) - auio.uio_resid;
5108 for (pos = 0; pos < done;) {
5109 dp = (struct dirent *)(dirbuf + pos);
5110 pos += dp->d_reclen;
5112 * XXX: Temporarily we also accept DT_UNKNOWN, as this
5113 * is what we get when attribute was created on Solaris.
5115 if (dp->d_type != DT_REG && dp->d_type != DT_UNKNOWN)
5117 if (plen == 0 && strncmp(dp->d_name, "freebsd:", 8) == 0)
5119 else if (strncmp(dp->d_name, attrprefix, plen) != 0)
5121 nlen = dp->d_namlen - plen;
5124 else if (uio != NULL) {
5126 * Format of extattr name entry is one byte for
5127 * length and the rest for name.
5129 error = uiomove(&nlen, 1, uio->uio_rw, uio);
5131 error = uiomove(dp->d_name + plen, nlen,
5138 } while (!eof && error == 0);
5147 zfs_freebsd_getacl(ap)
5148 struct vop_getacl_args /* {
5157 vsecattr_t vsecattr;
5159 if (ap->a_type != ACL_TYPE_NFS4)
5162 vsecattr.vsa_mask = VSA_ACE | VSA_ACECNT;
5163 if (error = zfs_getsecattr(ap->a_vp, &vsecattr, 0, ap->a_cred, NULL))
5166 error = acl_from_aces(ap->a_aclp, vsecattr.vsa_aclentp, vsecattr.vsa_aclcnt);
5167 if (vsecattr.vsa_aclentp != NULL)
5168 kmem_free(vsecattr.vsa_aclentp, vsecattr.vsa_aclentsz);
5174 zfs_freebsd_setacl(ap)
5175 struct vop_setacl_args /* {
5184 vsecattr_t vsecattr;
5185 int aclbsize; /* size of acl list in bytes */
5188 if (ap->a_type != ACL_TYPE_NFS4)
5191 if (ap->a_aclp->acl_cnt < 1 || ap->a_aclp->acl_cnt > MAX_ACL_ENTRIES)
5195 * With NFSv4 ACLs, chmod(2) may need to add additional entries,
5196 * splitting every entry into two and appending "canonical six"
5197 * entries at the end. Don't allow for setting an ACL that would
5198 * cause chmod(2) to run out of ACL entries.
5200 if (ap->a_aclp->acl_cnt * 2 + 6 > ACL_MAX_ENTRIES)
5203 error = acl_nfs4_check(ap->a_aclp, ap->a_vp->v_type == VDIR);
5207 vsecattr.vsa_mask = VSA_ACE;
5208 aclbsize = ap->a_aclp->acl_cnt * sizeof(ace_t);
5209 vsecattr.vsa_aclentp = kmem_alloc(aclbsize, KM_SLEEP);
5210 aaclp = vsecattr.vsa_aclentp;
5211 vsecattr.vsa_aclentsz = aclbsize;
5213 aces_from_acl(vsecattr.vsa_aclentp, &vsecattr.vsa_aclcnt, ap->a_aclp);
5214 error = zfs_setsecattr(ap->a_vp, &vsecattr, 0, ap->a_cred, NULL);
5215 kmem_free(aaclp, aclbsize);
5221 zfs_freebsd_aclcheck(ap)
5222 struct vop_aclcheck_args /* {
5231 return (EOPNOTSUPP);
5234 struct vop_vector zfs_vnodeops;
5235 struct vop_vector zfs_fifoops;
5236 struct vop_vector zfs_shareops;
5238 struct vop_vector zfs_vnodeops = {
5239 .vop_default = &default_vnodeops,
5240 .vop_inactive = zfs_freebsd_inactive,
5241 .vop_reclaim = zfs_freebsd_reclaim,
5242 .vop_access = zfs_freebsd_access,
5243 #ifdef FREEBSD_NAMECACHE
5244 .vop_lookup = vfs_cache_lookup,
5245 .vop_cachedlookup = zfs_freebsd_lookup,
5247 .vop_lookup = zfs_freebsd_lookup,
5249 .vop_getattr = zfs_freebsd_getattr,
5250 .vop_setattr = zfs_freebsd_setattr,
5251 .vop_create = zfs_freebsd_create,
5252 .vop_mknod = zfs_freebsd_create,
5253 .vop_mkdir = zfs_freebsd_mkdir,
5254 .vop_readdir = zfs_freebsd_readdir,
5255 .vop_fsync = zfs_freebsd_fsync,
5256 .vop_open = zfs_freebsd_open,
5257 .vop_close = zfs_freebsd_close,
5258 .vop_rmdir = zfs_freebsd_rmdir,
5259 .vop_ioctl = zfs_freebsd_ioctl,
5260 .vop_link = zfs_freebsd_link,
5261 .vop_symlink = zfs_freebsd_symlink,
5262 .vop_readlink = zfs_freebsd_readlink,
5263 .vop_read = zfs_freebsd_read,
5264 .vop_write = zfs_freebsd_write,
5265 .vop_remove = zfs_freebsd_remove,
5266 .vop_rename = zfs_freebsd_rename,
5267 .vop_pathconf = zfs_freebsd_pathconf,
5268 .vop_bmap = VOP_EOPNOTSUPP,
5269 .vop_fid = zfs_freebsd_fid,
5270 .vop_getextattr = zfs_getextattr,
5271 .vop_deleteextattr = zfs_deleteextattr,
5272 .vop_setextattr = zfs_setextattr,
5273 .vop_listextattr = zfs_listextattr,
5274 .vop_getacl = zfs_freebsd_getacl,
5275 .vop_setacl = zfs_freebsd_setacl,
5276 .vop_aclcheck = zfs_freebsd_aclcheck,
5279 struct vop_vector zfs_fifoops = {
5280 .vop_default = &fifo_specops,
5281 .vop_fsync = zfs_freebsd_fsync,
5282 .vop_access = zfs_freebsd_access,
5283 .vop_getattr = zfs_freebsd_getattr,
5284 .vop_inactive = zfs_freebsd_inactive,
5285 .vop_read = VOP_PANIC,
5286 .vop_reclaim = zfs_freebsd_reclaim,
5287 .vop_setattr = zfs_freebsd_setattr,
5288 .vop_write = VOP_PANIC,
5289 .vop_pathconf = zfs_freebsd_fifo_pathconf,
5290 .vop_fid = zfs_freebsd_fid,
5291 .vop_getacl = zfs_freebsd_getacl,
5292 .vop_setacl = zfs_freebsd_setacl,
5293 .vop_aclcheck = zfs_freebsd_aclcheck,
5297 * special share hidden files vnode operations template
5299 struct vop_vector zfs_shareops = {
5300 .vop_default = &default_vnodeops,
5301 .vop_access = zfs_freebsd_access,
5302 .vop_inactive = zfs_freebsd_inactive,
5303 .vop_reclaim = zfs_freebsd_reclaim,
5304 .vop_fid = zfs_freebsd_fid,
5305 .vop_pathconf = zfs_freebsd_pathconf,