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 */
26 /* Portions Copyright 2010 Robert Milkowski */
30 #include <sys/types.h>
31 #include <sys/param.h>
33 #include <sys/systm.h>
34 #include <sys/sysmacros.h>
35 #include <sys/resource.h>
37 #include <sys/vfs_opreg.h>
38 #include <sys/vnode.h>
42 #include <sys/taskq.h>
44 #include <sys/vmsystm.h>
45 #include <sys/atomic.h>
47 #include <vm/seg_vn.h>
51 #include <vm/seg_kpm.h>
53 #include <sys/pathname.h>
54 #include <sys/cmn_err.h>
55 #include <sys/errno.h>
56 #include <sys/unistd.h>
57 #include <sys/zfs_dir.h>
58 #include <sys/zfs_acl.h>
59 #include <sys/zfs_ioctl.h>
60 #include <sys/fs/zfs.h>
62 #include <sys/dmu_objset.h>
68 #include <sys/dirent.h>
69 #include <sys/policy.h>
70 #include <sys/sunddi.h>
73 #include "fs/fs_subr.h"
74 #include <sys/zfs_fuid.h>
75 #include <sys/zfs_sa.h>
76 #include <sys/zfs_vnops.h>
78 #include <sys/zfs_rlock.h>
79 #include <sys/extdirent.h>
80 #include <sys/kidmap.h>
87 * Each vnode op performs some logical unit of work. To do this, the ZPL must
88 * properly lock its in-core state, create a DMU transaction, do the work,
89 * record this work in the intent log (ZIL), commit the DMU transaction,
90 * and wait for the intent log to commit if it is a synchronous operation.
91 * Moreover, the vnode ops must work in both normal and log replay context.
92 * The ordering of events is important to avoid deadlocks and references
93 * to freed memory. The example below illustrates the following Big Rules:
95 * (1) A check must be made in each zfs thread for a mounted file system.
96 * This is done avoiding races using ZFS_ENTER(zfsvfs).
97 * A ZFS_EXIT(zfsvfs) is needed before all returns. Any znodes
98 * must be checked with ZFS_VERIFY_ZP(zp). Both of these macros
99 * can return EIO from the calling function.
101 * (2) VN_RELE() should always be the last thing except for zil_commit()
102 * (if necessary) and ZFS_EXIT(). This is for 3 reasons:
103 * First, if it's the last reference, the vnode/znode
104 * can be freed, so the zp may point to freed memory. Second, the last
105 * reference will call zfs_zinactive(), which may induce a lot of work --
106 * pushing cached pages (which acquires range locks) and syncing out
107 * cached atime changes. Third, zfs_zinactive() may require a new tx,
108 * which could deadlock the system if you were already holding one.
109 * If you must call VN_RELE() within a tx then use VN_RELE_ASYNC().
111 * (3) All range locks must be grabbed before calling dmu_tx_assign(),
112 * as they can span dmu_tx_assign() calls.
114 * (4) Always pass TXG_NOWAIT as the second argument to dmu_tx_assign().
115 * This is critical because we don't want to block while holding locks.
116 * Note, in particular, that if a lock is sometimes acquired before
117 * the tx assigns, and sometimes after (e.g. z_lock), then failing to
118 * use a non-blocking assign can deadlock the system. The scenario:
120 * Thread A has grabbed a lock before calling dmu_tx_assign().
121 * Thread B is in an already-assigned tx, and blocks for this lock.
122 * Thread A calls dmu_tx_assign(TXG_WAIT) and blocks in txg_wait_open()
123 * forever, because the previous txg can't quiesce until B's tx commits.
125 * If dmu_tx_assign() returns ERESTART and zfsvfs->z_assign is TXG_NOWAIT,
126 * then drop all locks, call dmu_tx_wait(), and try again.
128 * (5) If the operation succeeded, generate the intent log entry for it
129 * before dropping locks. This ensures that the ordering of events
130 * in the intent log matches the order in which they actually occurred.
131 * During ZIL replay the zfs_log_* functions will update the sequence
132 * number to indicate the zil transaction has replayed.
134 * (6) At the end of each vnode op, the DMU tx must always commit,
135 * regardless of whether there were any errors.
137 * (7) After dropping all locks, invoke zil_commit(zilog, foid)
138 * to ensure that synchronous semantics are provided when necessary.
140 * In general, this is how things should be ordered in each vnode op:
142 * ZFS_ENTER(zfsvfs); // exit if unmounted
144 * zfs_dirent_lock(&dl, ...) // lock directory entry (may VN_HOLD())
145 * rw_enter(...); // grab any other locks you need
146 * tx = dmu_tx_create(...); // get DMU tx
147 * dmu_tx_hold_*(); // hold each object you might modify
148 * error = dmu_tx_assign(tx, TXG_NOWAIT); // try to assign
150 * rw_exit(...); // drop locks
151 * zfs_dirent_unlock(dl); // unlock directory entry
152 * VN_RELE(...); // release held vnodes
153 * if (error == ERESTART) {
158 * dmu_tx_abort(tx); // abort DMU tx
159 * ZFS_EXIT(zfsvfs); // finished in zfs
160 * return (error); // really out of space
162 * error = do_real_work(); // do whatever this VOP does
164 * zfs_log_*(...); // on success, make ZIL entry
165 * dmu_tx_commit(tx); // commit DMU tx -- error or not
166 * rw_exit(...); // drop locks
167 * zfs_dirent_unlock(dl); // unlock directory entry
168 * VN_RELE(...); // release held vnodes
169 * zil_commit(zilog, foid); // synchronous when necessary
170 * ZFS_EXIT(zfsvfs); // finished in zfs
171 * return (error); // done, report error
176 zfs_open(vnode_t **vpp, int flag, cred_t *cr, caller_context_t *ct)
178 znode_t *zp = VTOZ(*vpp);
179 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
184 if ((flag & FWRITE) && (zp->z_pflags & ZFS_APPENDONLY) &&
185 ((flag & FAPPEND) == 0)) {
190 if (!zfs_has_ctldir(zp) && zp->z_zfsvfs->z_vscan &&
191 ZTOV(zp)->v_type == VREG &&
192 !(zp->z_pflags & ZFS_AV_QUARANTINED) && zp->z_size > 0) {
193 if (fs_vscan(*vpp, cr, 0) != 0) {
199 /* Keep a count of the synchronous opens in the znode */
200 if (flag & (FSYNC | FDSYNC))
201 atomic_inc_32(&zp->z_sync_cnt);
209 zfs_close(vnode_t *vp, int flag, int count, offset_t offset, cred_t *cr,
210 caller_context_t *ct)
212 znode_t *zp = VTOZ(vp);
213 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
216 * Clean up any locks held by this process on the vp.
218 cleanlocks(vp, ddi_get_pid(), 0);
219 cleanshares(vp, ddi_get_pid());
224 /* Decrement the synchronous opens in the znode */
225 if ((flag & (FSYNC | FDSYNC)) && (count == 1))
226 atomic_dec_32(&zp->z_sync_cnt);
228 if (!zfs_has_ctldir(zp) && zp->z_zfsvfs->z_vscan &&
229 ZTOV(zp)->v_type == VREG &&
230 !(zp->z_pflags & ZFS_AV_QUARANTINED) && zp->z_size > 0)
231 VERIFY(fs_vscan(vp, cr, 1) == 0);
238 * Lseek support for finding holes (cmd == _FIO_SEEK_HOLE) and
239 * data (cmd == _FIO_SEEK_DATA). "off" is an in/out parameter.
242 zfs_holey(vnode_t *vp, int cmd, offset_t *off)
244 znode_t *zp = VTOZ(vp);
245 uint64_t noff = (uint64_t)*off; /* new offset */
250 file_sz = zp->z_size;
251 if (noff >= file_sz) {
255 if (cmd == _FIO_SEEK_HOLE)
260 error = dmu_offset_next(zp->z_zfsvfs->z_os, zp->z_id, hole, &noff);
263 if ((error == ESRCH) || (noff > file_sz)) {
265 * Handle the virtual hole at the end of file.
282 zfs_ioctl(vnode_t *vp, int com, intptr_t data, int flag, cred_t *cred,
283 int *rvalp, caller_context_t *ct)
292 return (zfs_sync(vp->v_vfsp, 0, cred));
295 * The following two ioctls are used by bfu. Faking out,
296 * necessary to avoid bfu errors.
304 if (ddi_copyin((void *)data, &off, sizeof (off), flag))
308 zfsvfs = zp->z_zfsvfs;
312 /* offset parameter is in/out */
313 error = zfs_holey(vp, com, &off);
317 if (ddi_copyout(&off, (void *)data, sizeof (off), flag))
324 #if defined(_KERNEL) && defined(HAVE_UIO_RW)
326 * Utility functions to map and unmap a single physical page. These
327 * are used to manage the mappable copies of ZFS file data, and therefore
328 * do not update ref/mod bits.
331 zfs_map_page(page_t *pp, enum seg_rw rw)
334 return (hat_kpm_mapin(pp, 0));
335 ASSERT(rw == S_READ || rw == S_WRITE);
336 return (ppmapin(pp, PROT_READ | ((rw == S_WRITE) ? PROT_WRITE : 0),
341 zfs_unmap_page(page_t *pp, caddr_t addr)
344 hat_kpm_mapout(pp, 0, addr);
349 #endif /* _KERNEL && HAVE_UIO_RW */
352 * When a file is memory mapped, we must keep the IO data synchronized
353 * between the DMU cache and the memory mapped pages. What this means:
355 * On Write: If we find a memory mapped page, we write to *both*
356 * the page and the dmu buffer.
359 update_pages(vnode_t *vp, int64_t start, int len, objset_t *os, uint64_t oid)
363 off = start & PAGEOFFSET;
364 for (start &= PAGEMASK; len > 0; start += PAGESIZE) {
366 uint64_t nbytes = MIN(PAGESIZE - off, len);
368 if (pp = page_lookup(vp, start, SE_SHARED)) {
371 va = zfs_map_page(pp, S_WRITE);
372 (void) dmu_read(os, oid, start+off, nbytes, va+off,
374 zfs_unmap_page(pp, va);
383 * When a file is memory mapped, we must keep the IO data synchronized
384 * between the DMU cache and the memory mapped pages. What this means:
386 * On Read: We "read" preferentially from memory mapped pages,
387 * else we default from the dmu buffer.
389 * NOTE: We will always "break up" the IO into PAGESIZE uiomoves when
390 * the file is memory mapped.
393 mappedread(vnode_t *vp, int nbytes, uio_t *uio)
395 znode_t *zp = VTOZ(vp);
396 objset_t *os = zp->z_zfsvfs->z_os;
401 start = uio->uio_loffset;
402 off = start & PAGEOFFSET;
403 for (start &= PAGEMASK; len > 0; start += PAGESIZE) {
405 uint64_t bytes = MIN(PAGESIZE - off, len);
407 if (pp = page_lookup(vp, start, SE_SHARED)) {
410 va = zfs_map_page(pp, S_READ);
411 error = uiomove(va + off, bytes, UIO_READ, uio);
412 zfs_unmap_page(pp, va);
415 error = dmu_read_uio(os, zp->z_id, uio, bytes);
425 offset_t zfs_read_chunk_size = 1024 * 1024; /* Tunable */
428 * Read bytes from specified file into supplied buffer.
430 * IN: vp - vnode of file to be read from.
431 * uio - structure supplying read location, range info,
433 * ioflag - SYNC flags; used to provide FRSYNC semantics.
434 * cr - credentials of caller.
435 * ct - caller context
437 * OUT: uio - updated offset and range, buffer filled.
439 * RETURN: 0 if success
440 * error code if failure
443 * vp - atime updated if byte count > 0
447 zfs_read(vnode_t *vp, uio_t *uio, int ioflag, cred_t *cr, caller_context_t *ct)
449 znode_t *zp = VTOZ(vp);
450 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
461 if (zp->z_pflags & ZFS_AV_QUARANTINED) {
467 * Validate file offset
469 if (uio->uio_loffset < (offset_t)0) {
475 * Fasttrack empty reads
477 if (uio->uio_resid == 0) {
483 * Check for mandatory locks
485 if (MANDMODE(zp->z_mode)) {
486 if (error = chklock(vp, FREAD,
487 uio->uio_loffset, uio->uio_resid, uio->uio_fmode, ct)) {
494 * If we're in FRSYNC mode, sync out this znode before reading it.
496 if (ioflag & FRSYNC || zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS)
497 zil_commit(zfsvfs->z_log, zp->z_id);
500 * Lock the range against changes.
502 rl = zfs_range_lock(zp, uio->uio_loffset, uio->uio_resid, RL_READER);
505 * If we are reading past end-of-file we can skip
506 * to the end; but we might still need to set atime.
508 if (uio->uio_loffset >= zp->z_size) {
513 ASSERT(uio->uio_loffset < zp->z_size);
514 n = MIN(uio->uio_resid, zp->z_size - uio->uio_loffset);
516 if ((uio->uio_extflg == UIO_XUIO) &&
517 (((xuio_t *)uio)->xu_type == UIOTYPE_ZEROCOPY)) {
519 int blksz = zp->z_blksz;
520 uint64_t offset = uio->uio_loffset;
522 xuio = (xuio_t *)uio;
524 nblk = (P2ROUNDUP(offset + n, blksz) - P2ALIGN(offset,
527 ASSERT(offset + n <= blksz);
530 (void) dmu_xuio_init(xuio, nblk);
532 if (vn_has_cached_data(vp)) {
534 * For simplicity, we always allocate a full buffer
535 * even if we only expect to read a portion of a block.
537 while (--nblk >= 0) {
538 (void) dmu_xuio_add(xuio,
539 dmu_request_arcbuf(sa_get_db(zp->z_sa_hdl),
546 nbytes = MIN(n, zfs_read_chunk_size -
547 P2PHASE(uio->uio_loffset, zfs_read_chunk_size));
549 if (vn_has_cached_data(vp))
550 error = mappedread(vp, nbytes, uio);
552 error = dmu_read_uio(os, zp->z_id, uio, nbytes);
554 /* convert checksum errors into IO errors */
563 zfs_range_unlock(rl);
565 ZFS_ACCESSTIME_STAMP(zfsvfs, zp);
566 zfs_inode_update(zp);
570 EXPORT_SYMBOL(zfs_read);
573 * Write the bytes to a file.
575 * IN: vp - vnode of file to be written to.
576 * uio - structure supplying write location, range info,
578 * ioflag - FAPPEND flag set if in append mode.
579 * cr - credentials of caller.
580 * ct - caller context (NFS/CIFS fem monitor only)
582 * OUT: uio - updated offset and range.
584 * RETURN: 0 if success
585 * error code if failure
588 * vp - ctime|mtime updated if byte count > 0
593 zfs_write(vnode_t *vp, uio_t *uio, int ioflag, cred_t *cr, caller_context_t *ct)
595 znode_t *zp = VTOZ(vp);
596 rlim64_t limit = uio->uio_llimit;
597 ssize_t start_resid = uio->uio_resid;
601 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
606 int max_blksz = zfsvfs->z_max_blksz;
612 int iovcnt = uio->uio_iovcnt;
613 iovec_t *iovp = uio->uio_iov;
616 sa_bulk_attr_t bulk[4];
617 uint64_t mtime[2], ctime[2];
620 * Fasttrack empty write
626 if (limit == RLIM64_INFINITY || limit > MAXOFFSET_T)
632 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL, &mtime, 16);
633 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL, &ctime, 16);
634 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs), NULL,
636 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL,
640 * If immutable or not appending then return EPERM
642 if ((zp->z_pflags & (ZFS_IMMUTABLE | ZFS_READONLY)) ||
643 ((zp->z_pflags & ZFS_APPENDONLY) && !(ioflag & FAPPEND) &&
644 (uio->uio_loffset < zp->z_size))) {
649 zilog = zfsvfs->z_log;
652 * Validate file offset
654 woff = ioflag & FAPPEND ? zp->z_size : uio->uio_loffset;
661 * Check for mandatory locks before calling zfs_range_lock()
662 * in order to prevent a deadlock with locks set via fcntl().
664 if (MANDMODE((mode_t)zp->z_mode) &&
665 (error = chklock(vp, FWRITE, woff, n, uio->uio_fmode, ct)) != 0) {
671 * Pre-fault the pages to ensure slow (eg NFS) pages
673 * Skip this if uio contains loaned arc_buf.
675 if ((uio->uio_extflg == UIO_XUIO) &&
676 (((xuio_t *)uio)->xu_type == UIOTYPE_ZEROCOPY))
677 xuio = (xuio_t *)uio;
679 uio_prefaultpages(MIN(n, max_blksz), uio);
682 * If in append mode, set the io offset pointer to eof.
684 if (ioflag & FAPPEND) {
686 * Obtain an appending range lock to guarantee file append
687 * semantics. We reset the write offset once we have the lock.
689 rl = zfs_range_lock(zp, 0, n, RL_APPEND);
691 if (rl->r_len == UINT64_MAX) {
693 * We overlocked the file because this write will cause
694 * the file block size to increase.
695 * Note that zp_size cannot change with this lock held.
699 uio->uio_loffset = woff;
702 * Note that if the file block size will change as a result of
703 * this write, then this range lock will lock the entire file
704 * so that we can re-write the block safely.
706 rl = zfs_range_lock(zp, woff, n, RL_WRITER);
710 zfs_range_unlock(rl);
715 if ((woff + n) > limit || woff > (limit - n))
718 /* Will this write extend the file length? */
719 write_eof = (woff + n > zp->z_size);
721 end_size = MAX(zp->z_size, woff + n);
724 * Write the file in reasonable size chunks. Each chunk is written
725 * in a separate transaction; this keeps the intent log records small
726 * and allows us to do more fine-grained space accounting.
730 woff = uio->uio_loffset;
732 if (zfs_owner_overquota(zfsvfs, zp, B_FALSE) ||
733 zfs_owner_overquota(zfsvfs, zp, B_TRUE)) {
735 dmu_return_arcbuf(abuf);
740 if (xuio && abuf == NULL) {
741 ASSERT(i_iov < iovcnt);
743 abuf = dmu_xuio_arcbuf(xuio, i_iov);
744 dmu_xuio_clear(xuio, i_iov);
745 DTRACE_PROBE3(zfs_cp_write, int, i_iov,
746 iovec_t *, aiov, arc_buf_t *, abuf);
747 ASSERT((aiov->iov_base == abuf->b_data) ||
748 ((char *)aiov->iov_base - (char *)abuf->b_data +
749 aiov->iov_len == arc_buf_size(abuf)));
751 } else if (abuf == NULL && n >= max_blksz &&
752 woff >= zp->z_size &&
753 P2PHASE(woff, max_blksz) == 0 &&
754 zp->z_blksz == max_blksz) {
756 * This write covers a full block. "Borrow" a buffer
757 * from the dmu so that we can fill it before we enter
758 * a transaction. This avoids the possibility of
759 * holding up the transaction if the data copy hangs
760 * up on a pagefault (e.g., from an NFS server mapping).
764 abuf = dmu_request_arcbuf(sa_get_db(zp->z_sa_hdl),
766 ASSERT(abuf != NULL);
767 ASSERT(arc_buf_size(abuf) == max_blksz);
768 if ((error = uiocopy(abuf->b_data, max_blksz,
769 UIO_WRITE, uio, &cbytes))) {
770 dmu_return_arcbuf(abuf);
773 ASSERT(cbytes == max_blksz);
777 * Start a transaction.
779 tx = dmu_tx_create(zfsvfs->z_os);
780 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
781 dmu_tx_hold_write(tx, zp->z_id, woff, MIN(n, max_blksz));
782 zfs_sa_upgrade_txholds(tx, zp);
783 error = dmu_tx_assign(tx, TXG_NOWAIT);
785 if (error == ERESTART) {
792 dmu_return_arcbuf(abuf);
797 * If zfs_range_lock() over-locked we grow the blocksize
798 * and then reduce the lock range. This will only happen
799 * on the first iteration since zfs_range_reduce() will
800 * shrink down r_len to the appropriate size.
802 if (rl->r_len == UINT64_MAX) {
805 if (zp->z_blksz > max_blksz) {
806 ASSERT(!ISP2(zp->z_blksz));
807 new_blksz = MIN(end_size, SPA_MAXBLOCKSIZE);
809 new_blksz = MIN(end_size, max_blksz);
811 zfs_grow_blocksize(zp, new_blksz, tx);
812 zfs_range_reduce(rl, woff, n);
816 * XXX - should we really limit each write to z_max_blksz?
817 * Perhaps we should use SPA_MAXBLOCKSIZE chunks?
819 nbytes = MIN(n, max_blksz - P2PHASE(woff, max_blksz));
822 tx_bytes = uio->uio_resid;
823 error = dmu_write_uio_dbuf(sa_get_db(zp->z_sa_hdl),
825 tx_bytes -= uio->uio_resid;
828 ASSERT(xuio == NULL || tx_bytes == aiov->iov_len);
830 * If this is not a full block write, but we are
831 * extending the file past EOF and this data starts
832 * block-aligned, use assign_arcbuf(). Otherwise,
833 * write via dmu_write().
835 if (tx_bytes < max_blksz && (!write_eof ||
836 aiov->iov_base != abuf->b_data)) {
838 dmu_write(zfsvfs->z_os, zp->z_id, woff,
839 aiov->iov_len, aiov->iov_base, tx);
840 dmu_return_arcbuf(abuf);
841 xuio_stat_wbuf_copied();
843 ASSERT(xuio || tx_bytes == max_blksz);
844 dmu_assign_arcbuf(sa_get_db(zp->z_sa_hdl),
847 ASSERT(tx_bytes <= uio->uio_resid);
848 uioskip(uio, tx_bytes);
850 if (tx_bytes && vn_has_cached_data(vp)) {
851 update_pages(vp, woff,
852 tx_bytes, zfsvfs->z_os, zp->z_id);
856 * If we made no progress, we're done. If we made even
857 * partial progress, update the znode and ZIL accordingly.
860 (void) sa_update(zp->z_sa_hdl, SA_ZPL_SIZE(zfsvfs),
861 (void *)&zp->z_size, sizeof (uint64_t), tx);
868 * Clear Set-UID/Set-GID bits on successful write if not
869 * privileged and at least one of the excute bits is set.
871 * It would be nice to to this after all writes have
872 * been done, but that would still expose the ISUID/ISGID
873 * to another app after the partial write is committed.
875 * Note: we don't call zfs_fuid_map_id() here because
876 * user 0 is not an ephemeral uid.
878 mutex_enter(&zp->z_acl_lock);
879 if ((zp->z_mode & (S_IXUSR | (S_IXUSR >> 3) |
880 (S_IXUSR >> 6))) != 0 &&
881 (zp->z_mode & (S_ISUID | S_ISGID)) != 0 &&
882 secpolicy_vnode_setid_retain(cr,
883 (zp->z_mode & S_ISUID) != 0 && zp->z_uid == 0) != 0) {
885 zp->z_mode &= ~(S_ISUID | S_ISGID);
886 newmode = zp->z_mode;
887 (void) sa_update(zp->z_sa_hdl, SA_ZPL_MODE(zfsvfs),
888 (void *)&newmode, sizeof (uint64_t), tx);
890 mutex_exit(&zp->z_acl_lock);
892 zfs_tstamp_update_setup(zp, CONTENT_MODIFIED, mtime, ctime,
896 * Update the file size (zp_size) if it has changed;
897 * account for possible concurrent updates.
899 while ((end_size = zp->z_size) < uio->uio_loffset) {
900 (void) atomic_cas_64(&zp->z_size, end_size,
905 * If we are replaying and eof is non zero then force
906 * the file size to the specified eof. Note, there's no
907 * concurrency during replay.
909 if (zfsvfs->z_replay && zfsvfs->z_replay_eof != 0)
910 zp->z_size = zfsvfs->z_replay_eof;
912 error = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx);
914 zfs_log_write(zilog, tx, TX_WRITE, zp, woff, tx_bytes, ioflag);
919 ASSERT(tx_bytes == nbytes);
923 uio_prefaultpages(MIN(n, max_blksz), uio);
926 zfs_range_unlock(rl);
929 * If we're in replay mode, or we made no progress, return error.
930 * Otherwise, it's at least a partial write, so it's successful.
932 if (zfsvfs->z_replay || uio->uio_resid == start_resid) {
937 if (ioflag & (FSYNC | FDSYNC) ||
938 zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS)
939 zil_commit(zilog, zp->z_id);
941 zfs_inode_update(zp);
945 EXPORT_SYMBOL(zfs_write);
948 zfs_get_done(zgd_t *zgd, int error)
950 znode_t *zp = zgd->zgd_private;
951 objset_t *os = zp->z_zfsvfs->z_os;
954 dmu_buf_rele(zgd->zgd_db, zgd);
956 zfs_range_unlock(zgd->zgd_rl);
959 * Release the vnode asynchronously as we currently have the
960 * txg stopped from syncing.
962 VN_RELE_ASYNC(ZTOV(zp), dsl_pool_vnrele_taskq(dmu_objset_pool(os)));
964 if (error == 0 && zgd->zgd_bp)
965 zil_add_block(zgd->zgd_zilog, zgd->zgd_bp);
967 kmem_free(zgd, sizeof (zgd_t));
971 static int zil_fault_io = 0;
975 * Get data to generate a TX_WRITE intent log record.
978 zfs_get_data(void *arg, lr_write_t *lr, char *buf, zio_t *zio)
980 zfsvfs_t *zfsvfs = arg;
981 objset_t *os = zfsvfs->z_os;
983 uint64_t object = lr->lr_foid;
984 uint64_t offset = lr->lr_offset;
985 uint64_t size = lr->lr_length;
986 blkptr_t *bp = &lr->lr_blkptr;
995 * Nothing to do if the file has been removed
997 if (zfs_zget(zfsvfs, object, &zp) != 0)
999 if (zp->z_unlinked) {
1001 * Release the vnode asynchronously as we currently have the
1002 * txg stopped from syncing.
1004 VN_RELE_ASYNC(ZTOV(zp),
1005 dsl_pool_vnrele_taskq(dmu_objset_pool(os)));
1009 zgd = (zgd_t *)kmem_zalloc(sizeof (zgd_t), KM_SLEEP);
1010 zgd->zgd_zilog = zfsvfs->z_log;
1011 zgd->zgd_private = zp;
1014 * Write records come in two flavors: immediate and indirect.
1015 * For small writes it's cheaper to store the data with the
1016 * log record (immediate); for large writes it's cheaper to
1017 * sync the data and get a pointer to it (indirect) so that
1018 * we don't have to write the data twice.
1020 if (buf != NULL) { /* immediate write */
1021 zgd->zgd_rl = zfs_range_lock(zp, offset, size, RL_READER);
1022 /* test for truncation needs to be done while range locked */
1023 if (offset >= zp->z_size) {
1026 error = dmu_read(os, object, offset, size, buf,
1027 DMU_READ_NO_PREFETCH);
1029 ASSERT(error == 0 || error == ENOENT);
1030 } else { /* indirect write */
1032 * Have to lock the whole block to ensure when it's
1033 * written out and it's checksum is being calculated
1034 * that no one can change the data. We need to re-check
1035 * blocksize after we get the lock in case it's changed!
1040 blkoff = ISP2(size) ? P2PHASE(offset, size) : offset;
1042 zgd->zgd_rl = zfs_range_lock(zp, offset, size,
1044 if (zp->z_blksz == size)
1047 zfs_range_unlock(zgd->zgd_rl);
1049 /* test for truncation needs to be done while range locked */
1050 if (lr->lr_offset >= zp->z_size)
1059 error = dmu_buf_hold(os, object, offset, zgd, &db,
1060 DMU_READ_NO_PREFETCH);
1066 ASSERT(db->db_offset == offset);
1067 ASSERT(db->db_size == size);
1069 error = dmu_sync(zio, lr->lr_common.lrc_txg,
1071 ASSERT(error || lr->lr_length <= zp->z_blksz);
1074 * On success, we need to wait for the write I/O
1075 * initiated by dmu_sync() to complete before we can
1076 * release this dbuf. We will finish everything up
1077 * in the zfs_get_done() callback.
1082 if (error == EALREADY) {
1083 lr->lr_common.lrc_txtype = TX_WRITE2;
1089 zfs_get_done(zgd, error);
1096 zfs_access(vnode_t *vp, int mode, int flag, cred_t *cr,
1097 caller_context_t *ct)
1099 znode_t *zp = VTOZ(vp);
1100 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1106 if (flag & V_ACE_MASK)
1107 error = zfs_zaccess(zp, mode, flag, B_FALSE, cr);
1109 error = zfs_zaccess_rwx(zp, mode, flag, cr);
1116 * If vnode is for a device return a specfs vnode instead.
1119 specvp_check(vnode_t **vpp, cred_t *cr)
1123 if (IS_DEVVP(*vpp)) {
1126 svp = specvp(*vpp, (*vpp)->v_rdev, (*vpp)->v_type, cr);
1137 * Lookup an entry in a directory, or an extended attribute directory.
1138 * If it exists, return a held vnode reference for it.
1140 * IN: dvp - vnode of directory to search.
1141 * nm - name of entry to lookup.
1142 * pnp - full pathname to lookup [UNUSED].
1143 * flags - LOOKUP_XATTR set if looking for an attribute.
1144 * rdir - root directory vnode [UNUSED].
1145 * cr - credentials of caller.
1146 * ct - caller context
1147 * direntflags - directory lookup flags
1148 * realpnp - returned pathname.
1150 * OUT: vpp - vnode of located entry, NULL if not found.
1152 * RETURN: 0 if success
1153 * error code if failure
1160 zfs_lookup(vnode_t *dvp, char *nm, vnode_t **vpp, struct pathname *pnp,
1161 int flags, vnode_t *rdir, cred_t *cr, caller_context_t *ct,
1162 int *direntflags, pathname_t *realpnp)
1164 znode_t *zdp = VTOZ(dvp);
1165 zfsvfs_t *zfsvfs = zdp->z_zfsvfs;
1169 if (!(flags & (LOOKUP_XATTR | FIGNORECASE))) {
1171 if (dvp->v_type != VDIR) {
1173 } else if (zdp->z_sa_hdl == NULL) {
1177 if (nm[0] == 0 || (nm[0] == '.' && nm[1] == '\0')) {
1178 error = zfs_fastaccesschk_execute(zdp, cr);
1186 vnode_t *tvp = dnlc_lookup(dvp, nm);
1189 error = zfs_fastaccesschk_execute(zdp, cr);
1194 if (tvp == DNLC_NO_VNODE) {
1199 return (specvp_check(vpp, cr));
1205 DTRACE_PROBE2(zfs__fastpath__lookup__miss, vnode_t *, dvp, char *, nm);
1212 if (flags & LOOKUP_XATTR) {
1214 * If the xattr property is off, refuse the lookup request.
1216 if (!(zfsvfs->z_vfs->vfs_flag & VFS_XATTR)) {
1222 * We don't allow recursive attributes..
1223 * Maybe someday we will.
1225 if (zdp->z_pflags & ZFS_XATTR) {
1230 if ((error = zfs_get_xattrdir(VTOZ(dvp), vpp, cr, flags))) {
1236 * Do we have permission to get into attribute directory?
1239 if ((error = zfs_zaccess(VTOZ(*vpp), ACE_EXECUTE, 0,
1249 if (dvp->v_type != VDIR) {
1255 * Check accessibility of directory.
1258 if ((error = zfs_zaccess(zdp, ACE_EXECUTE, 0, B_FALSE, cr))) {
1263 if (zfsvfs->z_utf8 && u8_validate(nm, strlen(nm),
1264 NULL, U8_VALIDATE_ENTIRE, &error) < 0) {
1269 error = zfs_dirlook(zdp, nm, vpp, flags, direntflags, realpnp);
1272 zfs_inode_update(VTOZ(*vpp));
1274 error = specvp_check(vpp, cr);
1280 EXPORT_SYMBOL(zfs_lookup);
1283 * Attempt to create a new entry in a directory. If the entry
1284 * already exists, truncate the file if permissible, else return
1285 * an error. Return the vp of the created or trunc'd file.
1287 * IN: dvp - vnode of directory to put new file entry in.
1288 * name - name of new file entry.
1289 * vap - attributes of new file.
1290 * excl - flag indicating exclusive or non-exclusive mode.
1291 * mode - mode to open file with.
1292 * cr - credentials of caller.
1293 * flag - large file flag [UNUSED].
1294 * ct - caller context
1295 * vsecp - ACL to be set
1297 * OUT: vpp - vnode of created or trunc'd entry.
1299 * RETURN: 0 if success
1300 * error code if failure
1303 * dvp - ctime|mtime updated if new entry created
1304 * vp - ctime|mtime always, atime if new
1309 zfs_create(vnode_t *dvp, char *name, vattr_t *vap, int excl,
1310 int mode, vnode_t **vpp, cred_t *cr, int flag, caller_context_t *ct,
1313 znode_t *zp, *dzp = VTOZ(dvp);
1314 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
1323 zfs_acl_ids_t acl_ids;
1324 boolean_t fuid_dirtied;
1325 boolean_t have_acl = B_FALSE;
1328 * If we have an ephemeral id, ACL, or XVATTR then
1329 * make sure file system is at proper version
1333 ksid = crgetsid(cr, KSID_OWNER);
1335 uid = ksid_getid(ksid);
1339 if (zfsvfs->z_use_fuids == B_FALSE &&
1340 (vsecp || (vap->va_mask & AT_XVATTR) ||
1341 IS_EPHEMERAL(uid) || IS_EPHEMERAL(gid)))
1347 zilog = zfsvfs->z_log;
1349 if (zfsvfs->z_utf8 && u8_validate(name, strlen(name),
1350 NULL, U8_VALIDATE_ENTIRE, &error) < 0) {
1355 if (vap->va_mask & AT_XVATTR) {
1356 if ((error = secpolicy_xvattr((xvattr_t *)vap,
1357 crgetuid(cr), cr, vap->va_type)) != 0) {
1365 if ((vap->va_mode & VSVTX) && secpolicy_vnode_stky_modify(cr))
1366 vap->va_mode &= ~VSVTX;
1368 if (*name == '\0') {
1370 * Null component name refers to the directory itself.
1377 /* possible VN_HOLD(zp) */
1380 if (flag & FIGNORECASE)
1383 error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg,
1387 zfs_acl_ids_free(&acl_ids);
1388 if (strcmp(name, "..") == 0)
1399 * Create a new file object and update the directory
1402 if ((error = zfs_zaccess(dzp, ACE_ADD_FILE, 0, B_FALSE, cr))) {
1404 zfs_acl_ids_free(&acl_ids);
1409 * We only support the creation of regular files in
1410 * extended attribute directories.
1413 if ((dzp->z_pflags & ZFS_XATTR) &&
1414 (vap->va_type != VREG)) {
1416 zfs_acl_ids_free(&acl_ids);
1421 if (!have_acl && (error = zfs_acl_ids_create(dzp, 0, vap,
1422 cr, vsecp, &acl_ids)) != 0)
1426 if (zfs_acl_ids_overquota(zfsvfs, &acl_ids)) {
1427 zfs_acl_ids_free(&acl_ids);
1432 tx = dmu_tx_create(os);
1434 dmu_tx_hold_sa_create(tx, acl_ids.z_aclp->z_acl_bytes +
1435 ZFS_SA_BASE_ATTR_SIZE);
1437 fuid_dirtied = zfsvfs->z_fuid_dirty;
1439 zfs_fuid_txhold(zfsvfs, tx);
1440 dmu_tx_hold_zap(tx, dzp->z_id, TRUE, name);
1441 dmu_tx_hold_sa(tx, dzp->z_sa_hdl, B_FALSE);
1442 if (!zfsvfs->z_use_sa &&
1443 acl_ids.z_aclp->z_acl_bytes > ZFS_ACE_SPACE) {
1444 dmu_tx_hold_write(tx, DMU_NEW_OBJECT,
1445 0, acl_ids.z_aclp->z_acl_bytes);
1447 error = dmu_tx_assign(tx, TXG_NOWAIT);
1449 zfs_dirent_unlock(dl);
1450 if (error == ERESTART) {
1455 zfs_acl_ids_free(&acl_ids);
1460 zfs_mknode(dzp, vap, tx, cr, 0, &zp, &acl_ids);
1463 zfs_fuid_sync(zfsvfs, tx);
1465 (void) zfs_link_create(dl, zp, tx, ZNEW);
1466 txtype = zfs_log_create_txtype(Z_FILE, vsecp, vap);
1467 if (flag & FIGNORECASE)
1469 zfs_log_create(zilog, tx, txtype, dzp, zp, name,
1470 vsecp, acl_ids.z_fuidp, vap);
1471 zfs_acl_ids_free(&acl_ids);
1474 int aflags = (flag & FAPPEND) ? V_APPEND : 0;
1477 zfs_acl_ids_free(&acl_ids);
1481 * A directory entry already exists for this name.
1484 * Can't truncate an existing file if in exclusive mode.
1491 * Can't open a directory for writing.
1493 if ((ZTOV(zp)->v_type == VDIR) && (mode & S_IWRITE)) {
1498 * Verify requested access to file.
1500 if (mode && (error = zfs_zaccess_rwx(zp, mode, aflags, cr))) {
1504 mutex_enter(&dzp->z_lock);
1506 mutex_exit(&dzp->z_lock);
1509 * Truncate regular files if requested.
1511 if ((ZTOV(zp)->v_type == VREG) &&
1512 (vap->va_mask & AT_SIZE) && (vap->va_size == 0)) {
1513 /* we can't hold any locks when calling zfs_freesp() */
1514 zfs_dirent_unlock(dl);
1516 error = zfs_freesp(zp, 0, 0, mode, TRUE);
1518 vnevent_create(ZTOV(zp), ct);
1525 zfs_dirent_unlock(dl);
1531 zfs_inode_update(dzp);
1532 zfs_inode_update(zp);
1534 error = specvp_check(vpp, cr);
1537 if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS)
1538 zil_commit(zilog, 0);
1543 EXPORT_SYMBOL(zfs_create);
1546 * Remove an entry from a directory.
1548 * IN: dvp - vnode of directory to remove entry from.
1549 * name - name of entry to remove.
1550 * cr - credentials of caller.
1551 * ct - caller context
1552 * flags - case flags
1554 * RETURN: 0 if success
1555 * error code if failure
1559 * vp - ctime (if nlink > 0)
1562 uint64_t null_xattr = 0;
1566 zfs_remove(vnode_t *dvp, char *name, cred_t *cr, caller_context_t *ct,
1569 znode_t *zp, *dzp = VTOZ(dvp);
1572 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
1574 uint64_t acl_obj, xattr_obj;
1575 uint64_t xattr_obj_unlinked = 0;
1579 boolean_t may_delete_now, delete_now = FALSE;
1580 boolean_t unlinked, toobig = FALSE;
1582 pathname_t *realnmp = NULL;
1589 zilog = zfsvfs->z_log;
1591 if (flags & FIGNORECASE) {
1601 * Attempt to lock directory; fail if entry doesn't exist.
1603 if ((error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg,
1613 if ((error = zfs_zaccess_delete(dzp, zp, cr))) {
1618 * Need to use rmdir for removing directories.
1620 if (vp->v_type == VDIR) {
1625 vnevent_remove(vp, dvp, name, ct);
1628 dnlc_remove(dvp, realnmp->pn_buf);
1630 dnlc_remove(dvp, name);
1632 mutex_enter(&vp->v_lock);
1633 may_delete_now = ((vp->v_count == 1) && (!vn_has_cached_data(vp)));
1634 mutex_exit(&vp->v_lock);
1637 * We may delete the znode now, or we may put it in the unlinked set;
1638 * it depends on whether we're the last link, and on whether there are
1639 * other holds on the vnode. So we dmu_tx_hold() the right things to
1640 * allow for either case.
1643 tx = dmu_tx_create(zfsvfs->z_os);
1644 dmu_tx_hold_zap(tx, dzp->z_id, FALSE, name);
1645 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
1646 zfs_sa_upgrade_txholds(tx, zp);
1647 zfs_sa_upgrade_txholds(tx, dzp);
1648 if (may_delete_now) {
1650 zp->z_size > zp->z_blksz * DMU_MAX_DELETEBLKCNT;
1651 /* if the file is too big, only hold_free a token amount */
1652 dmu_tx_hold_free(tx, zp->z_id, 0,
1653 (toobig ? DMU_MAX_ACCESS : DMU_OBJECT_END));
1656 /* are there any extended attributes? */
1657 error = sa_lookup(zp->z_sa_hdl, SA_ZPL_XATTR(zfsvfs),
1658 &xattr_obj, sizeof (xattr_obj));
1659 if (error == 0 && xattr_obj) {
1660 error = zfs_zget(zfsvfs, xattr_obj, &xzp);
1661 ASSERT3U(error, ==, 0);
1662 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE);
1663 dmu_tx_hold_sa(tx, xzp->z_sa_hdl, B_FALSE);
1666 mutex_enter(&zp->z_lock);
1667 if ((acl_obj = zfs_external_acl(zp)) != 0 && may_delete_now)
1668 dmu_tx_hold_free(tx, acl_obj, 0, DMU_OBJECT_END);
1669 mutex_exit(&zp->z_lock);
1671 /* charge as an update -- would be nice not to charge at all */
1672 dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL);
1674 error = dmu_tx_assign(tx, TXG_NOWAIT);
1676 zfs_dirent_unlock(dl);
1680 if (error == ERESTART) {
1693 * Remove the directory entry.
1695 error = zfs_link_destroy(dl, zp, tx, zflg, &unlinked);
1705 * Hold z_lock so that we can make sure that the ACL obj
1706 * hasn't changed. Could have been deleted due to
1709 mutex_enter(&zp->z_lock);
1710 mutex_enter(&vp->v_lock);
1711 (void) sa_lookup(zp->z_sa_hdl, SA_ZPL_XATTR(zfsvfs),
1712 &xattr_obj_unlinked, sizeof (xattr_obj_unlinked));
1713 delete_now = may_delete_now && !toobig &&
1714 vp->v_count == 1 && !vn_has_cached_data(vp) &&
1715 xattr_obj == xattr_obj_unlinked && zfs_external_acl(zp) ==
1717 mutex_exit(&vp->v_lock);
1721 if (xattr_obj_unlinked) {
1722 ASSERT3U(xzp->z_links, ==, 2);
1723 mutex_enter(&xzp->z_lock);
1724 xzp->z_unlinked = 1;
1726 error = sa_update(xzp->z_sa_hdl, SA_ZPL_LINKS(zfsvfs),
1727 &xzp->z_links, sizeof (xzp->z_links), tx);
1728 ASSERT3U(error, ==, 0);
1729 mutex_exit(&xzp->z_lock);
1730 zfs_unlinked_add(xzp, tx);
1733 error = sa_remove(zp->z_sa_hdl,
1734 SA_ZPL_XATTR(zfsvfs), tx);
1736 error = sa_update(zp->z_sa_hdl,
1737 SA_ZPL_XATTR(zfsvfs), &null_xattr,
1738 sizeof (uint64_t), tx);
1739 ASSERT3U(error, ==, 0);
1741 mutex_enter(&vp->v_lock);
1743 ASSERT3U(vp->v_count, ==, 0);
1744 mutex_exit(&vp->v_lock);
1745 mutex_exit(&zp->z_lock);
1746 zfs_znode_delete(zp, tx);
1747 } else if (unlinked) {
1748 mutex_exit(&zp->z_lock);
1749 zfs_unlinked_add(zp, tx);
1753 if (flags & FIGNORECASE)
1755 zfs_log_remove(zilog, tx, txtype, dzp, name, obj);
1762 zfs_dirent_unlock(dl);
1763 zfs_inode_update(dzp);
1764 zfs_inode_update(zp);
1766 zfs_inode_update(xzp);
1773 if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS)
1774 zil_commit(zilog, 0);
1779 EXPORT_SYMBOL(zfs_remove);
1782 * Create a new directory and insert it into dvp using the name
1783 * provided. Return a pointer to the inserted directory.
1785 * IN: dvp - vnode of directory to add subdir to.
1786 * dirname - name of new directory.
1787 * vap - attributes of new directory.
1788 * cr - credentials of caller.
1789 * ct - caller context
1790 * vsecp - ACL to be set
1792 * OUT: vpp - vnode of created directory.
1794 * RETURN: 0 if success
1795 * error code if failure
1798 * dvp - ctime|mtime updated
1799 * vp - ctime|mtime|atime updated
1803 zfs_mkdir(vnode_t *dvp, char *dirname, vattr_t *vap, vnode_t **vpp, cred_t *cr,
1804 caller_context_t *ct, int flags, vsecattr_t *vsecp)
1806 znode_t *zp, *dzp = VTOZ(dvp);
1807 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
1816 gid_t gid = crgetgid(cr);
1817 zfs_acl_ids_t acl_ids;
1818 boolean_t fuid_dirtied;
1820 ASSERT(vap->va_type == VDIR);
1823 * If we have an ephemeral id, ACL, or XVATTR then
1824 * make sure file system is at proper version
1827 ksid = crgetsid(cr, KSID_OWNER);
1829 uid = ksid_getid(ksid);
1832 if (zfsvfs->z_use_fuids == B_FALSE &&
1833 (vsecp || (vap->va_mask & AT_XVATTR) ||
1834 IS_EPHEMERAL(uid) || IS_EPHEMERAL(gid)))
1839 zilog = zfsvfs->z_log;
1841 if (dzp->z_pflags & ZFS_XATTR) {
1846 if (zfsvfs->z_utf8 && u8_validate(dirname,
1847 strlen(dirname), NULL, U8_VALIDATE_ENTIRE, &error) < 0) {
1851 if (flags & FIGNORECASE)
1854 if (vap->va_mask & AT_XVATTR) {
1855 if ((error = secpolicy_xvattr((xvattr_t *)vap,
1856 crgetuid(cr), cr, vap->va_type)) != 0) {
1862 if ((error = zfs_acl_ids_create(dzp, 0, vap, cr,
1863 vsecp, &acl_ids)) != 0) {
1868 * First make sure the new directory doesn't exist.
1870 * Existence is checked first to make sure we don't return
1871 * EACCES instead of EEXIST which can cause some applications
1877 if ((error = zfs_dirent_lock(&dl, dzp, dirname, &zp, zf,
1879 zfs_acl_ids_free(&acl_ids);
1884 if ((error = zfs_zaccess(dzp, ACE_ADD_SUBDIRECTORY, 0, B_FALSE, cr))) {
1885 zfs_acl_ids_free(&acl_ids);
1886 zfs_dirent_unlock(dl);
1891 if (zfs_acl_ids_overquota(zfsvfs, &acl_ids)) {
1892 zfs_acl_ids_free(&acl_ids);
1893 zfs_dirent_unlock(dl);
1899 * Add a new entry to the directory.
1901 tx = dmu_tx_create(zfsvfs->z_os);
1902 dmu_tx_hold_zap(tx, dzp->z_id, TRUE, dirname);
1903 dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL);
1904 fuid_dirtied = zfsvfs->z_fuid_dirty;
1906 zfs_fuid_txhold(zfsvfs, tx);
1907 if (!zfsvfs->z_use_sa && acl_ids.z_aclp->z_acl_bytes > ZFS_ACE_SPACE) {
1908 dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0,
1909 acl_ids.z_aclp->z_acl_bytes);
1912 dmu_tx_hold_sa_create(tx, acl_ids.z_aclp->z_acl_bytes +
1913 ZFS_SA_BASE_ATTR_SIZE);
1915 error = dmu_tx_assign(tx, TXG_NOWAIT);
1917 zfs_dirent_unlock(dl);
1918 if (error == ERESTART) {
1923 zfs_acl_ids_free(&acl_ids);
1932 zfs_mknode(dzp, vap, tx, cr, 0, &zp, &acl_ids);
1935 zfs_fuid_sync(zfsvfs, tx);
1938 * Now put new name in parent dir.
1940 (void) zfs_link_create(dl, zp, tx, ZNEW);
1944 txtype = zfs_log_create_txtype(Z_DIR, vsecp, vap);
1945 if (flags & FIGNORECASE)
1947 zfs_log_create(zilog, tx, txtype, dzp, zp, dirname, vsecp,
1948 acl_ids.z_fuidp, vap);
1950 zfs_acl_ids_free(&acl_ids);
1954 zfs_dirent_unlock(dl);
1956 if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS)
1957 zil_commit(zilog, 0);
1959 zfs_inode_update(dzp);
1960 zfs_inode_update(zp);
1964 EXPORT_SYMBOL(zfs_mkdir);
1967 * Remove a directory subdir entry. If the current working
1968 * directory is the same as the subdir to be removed, the
1971 * IN: dvp - vnode of directory to remove from.
1972 * name - name of directory to be removed.
1973 * cwd - vnode of current working directory.
1974 * cr - credentials of caller.
1975 * ct - caller context
1976 * flags - case flags
1978 * RETURN: 0 if success
1979 * error code if failure
1982 * dvp - ctime|mtime updated
1986 zfs_rmdir(vnode_t *dvp, char *name, vnode_t *cwd, cred_t *cr,
1987 caller_context_t *ct, int flags)
1989 znode_t *dzp = VTOZ(dvp);
1992 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
2001 zilog = zfsvfs->z_log;
2003 if (flags & FIGNORECASE)
2009 * Attempt to lock directory; fail if entry doesn't exist.
2011 if ((error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg,
2019 if ((error = zfs_zaccess_delete(dzp, zp, cr))) {
2023 if (vp->v_type != VDIR) {
2033 vnevent_rmdir(vp, dvp, name, ct);
2036 * Grab a lock on the directory to make sure that noone is
2037 * trying to add (or lookup) entries while we are removing it.
2039 rw_enter(&zp->z_name_lock, RW_WRITER);
2042 * Grab a lock on the parent pointer to make sure we play well
2043 * with the treewalk and directory rename code.
2045 rw_enter(&zp->z_parent_lock, RW_WRITER);
2047 tx = dmu_tx_create(zfsvfs->z_os);
2048 dmu_tx_hold_zap(tx, dzp->z_id, FALSE, name);
2049 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
2050 dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL);
2051 zfs_sa_upgrade_txholds(tx, zp);
2052 zfs_sa_upgrade_txholds(tx, dzp);
2053 error = dmu_tx_assign(tx, TXG_NOWAIT);
2055 rw_exit(&zp->z_parent_lock);
2056 rw_exit(&zp->z_name_lock);
2057 zfs_dirent_unlock(dl);
2059 if (error == ERESTART) {
2069 error = zfs_link_destroy(dl, zp, tx, zflg, NULL);
2072 uint64_t txtype = TX_RMDIR;
2073 if (flags & FIGNORECASE)
2075 zfs_log_remove(zilog, tx, txtype, dzp, name, ZFS_NO_OBJECT);
2080 rw_exit(&zp->z_parent_lock);
2081 rw_exit(&zp->z_name_lock);
2083 zfs_dirent_unlock(dl);
2087 if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS)
2088 zil_commit(zilog, 0);
2090 zfs_inode_update(dzp);
2091 zfs_inode_update(zp);
2095 EXPORT_SYMBOL(zfs_rmdir);
2098 * Read as many directory entries as will fit into the provided
2099 * buffer from the given directory cursor position (specified in
2100 * the uio structure.
2102 * IN: vp - vnode of directory to read.
2103 * uio - structure supplying read location, range info,
2104 * and return buffer.
2105 * cr - credentials of caller.
2106 * ct - caller context
2107 * flags - case flags
2109 * OUT: uio - updated offset and range, buffer filled.
2110 * eofp - set to true if end-of-file detected.
2112 * RETURN: 0 if success
2113 * error code if failure
2116 * vp - atime updated
2118 * Note that the low 4 bits of the cookie returned by zap is always zero.
2119 * This allows us to use the low range for "special" directory entries:
2120 * We use 0 for '.', and 1 for '..'. If this is the root of the filesystem,
2121 * we use the offset 2 for the '.zfs' directory.
2125 zfs_readdir(vnode_t *vp, uio_t *uio, cred_t *cr, int *eofp,
2126 caller_context_t *ct, int flags)
2128 znode_t *zp = VTOZ(vp);
2132 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
2137 zap_attribute_t zap;
2138 uint_t bytes_wanted;
2139 uint64_t offset; /* must be unsigned; checks for < 1 */
2145 boolean_t check_sysattrs;
2150 if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_PARENT(zfsvfs),
2151 &parent, sizeof (parent))) != 0) {
2157 * If we are not given an eof variable,
2164 * Check for valid iov_len.
2166 if (uio->uio_iov->iov_len <= 0) {
2172 * Quit if directory has been removed (posix)
2174 if ((*eofp = zp->z_unlinked) != 0) {
2181 offset = uio->uio_loffset;
2182 prefetch = zp->z_zn_prefetch;
2185 * Initialize the iterator cursor.
2189 * Start iteration from the beginning of the directory.
2191 zap_cursor_init(&zc, os, zp->z_id);
2194 * The offset is a serialized cursor.
2196 zap_cursor_init_serialized(&zc, os, zp->z_id, offset);
2200 * Get space to change directory entries into fs independent format.
2202 iovp = uio->uio_iov;
2203 bytes_wanted = iovp->iov_len;
2204 if (uio->uio_segflg != UIO_SYSSPACE || uio->uio_iovcnt != 1) {
2205 bufsize = bytes_wanted;
2206 outbuf = kmem_alloc(bufsize, KM_SLEEP);
2207 odp = (struct dirent64 *)outbuf;
2209 bufsize = bytes_wanted;
2210 odp = (struct dirent64 *)iovp->iov_base;
2212 eodp = (struct edirent *)odp;
2215 * If this VFS supports the system attribute view interface; and
2216 * we're looking at an extended attribute directory; and we care
2217 * about normalization conflicts on this vfs; then we must check
2218 * for normalization conflicts with the sysattr name space.
2220 check_sysattrs = vfs_has_feature(vp->v_vfsp, VFSFT_SYSATTR_VIEWS) &&
2221 (vp->v_flag & V_XATTRDIR) && zfsvfs->z_norm &&
2222 (flags & V_RDDIR_ENTFLAGS);
2225 * Transform to file-system independent format
2228 while (outcount < bytes_wanted) {
2231 off64_t *next = NULL;
2234 * Special case `.', `..', and `.zfs'.
2237 (void) strcpy(zap.za_name, ".");
2238 zap.za_normalization_conflict = 0;
2240 } else if (offset == 1) {
2241 (void) strcpy(zap.za_name, "..");
2242 zap.za_normalization_conflict = 0;
2244 } else if (offset == 2 && zfs_show_ctldir(zp)) {
2245 (void) strcpy(zap.za_name, ZFS_CTLDIR_NAME);
2246 zap.za_normalization_conflict = 0;
2247 objnum = ZFSCTL_INO_ROOT;
2252 if (error = zap_cursor_retrieve(&zc, &zap)) {
2253 if ((*eofp = (error == ENOENT)) != 0)
2259 if (zap.za_integer_length != 8 ||
2260 zap.za_num_integers != 1) {
2261 cmn_err(CE_WARN, "zap_readdir: bad directory "
2262 "entry, obj = %lld, offset = %lld\n",
2263 (u_longlong_t)zp->z_id,
2264 (u_longlong_t)offset);
2269 objnum = ZFS_DIRENT_OBJ(zap.za_first_integer);
2271 * MacOS X can extract the object type here such as:
2272 * uint8_t type = ZFS_DIRENT_TYPE(zap.za_first_integer);
2275 if (check_sysattrs && !zap.za_normalization_conflict) {
2276 zap.za_normalization_conflict =
2277 xattr_sysattr_casechk(zap.za_name);
2281 if (flags & V_RDDIR_ACCFILTER) {
2283 * If we have no access at all, don't include
2284 * this entry in the returned information
2287 if (zfs_zget(zp->z_zfsvfs, objnum, &ezp) != 0)
2289 if (!zfs_has_access(ezp, cr)) {
2296 if (flags & V_RDDIR_ENTFLAGS)
2297 reclen = EDIRENT_RECLEN(strlen(zap.za_name));
2299 reclen = DIRENT64_RECLEN(strlen(zap.za_name));
2302 * Will this entry fit in the buffer?
2304 if (outcount + reclen > bufsize) {
2306 * Did we manage to fit anything in the buffer?
2314 if (flags & V_RDDIR_ENTFLAGS) {
2316 * Add extended flag entry:
2318 eodp->ed_ino = objnum;
2319 eodp->ed_reclen = reclen;
2320 /* NOTE: ed_off is the offset for the *next* entry */
2321 next = &(eodp->ed_off);
2322 eodp->ed_eflags = zap.za_normalization_conflict ?
2323 ED_CASE_CONFLICT : 0;
2324 (void) strncpy(eodp->ed_name, zap.za_name,
2325 EDIRENT_NAMELEN(reclen));
2326 eodp = (edirent_t *)((intptr_t)eodp + reclen);
2331 odp->d_ino = objnum;
2332 odp->d_reclen = reclen;
2333 /* NOTE: d_off is the offset for the *next* entry */
2334 next = &(odp->d_off);
2335 (void) strncpy(odp->d_name, zap.za_name,
2336 DIRENT64_NAMELEN(reclen));
2337 odp = (dirent64_t *)((intptr_t)odp + reclen);
2341 ASSERT(outcount <= bufsize);
2343 /* Prefetch znode */
2345 dmu_prefetch(os, objnum, 0, 0);
2349 * Move to the next entry, fill in the previous offset.
2351 if (offset > 2 || (offset == 2 && !zfs_show_ctldir(zp))) {
2352 zap_cursor_advance(&zc);
2353 offset = zap_cursor_serialize(&zc);
2360 zp->z_zn_prefetch = B_FALSE; /* a lookup will re-enable pre-fetching */
2362 if (uio->uio_segflg == UIO_SYSSPACE && uio->uio_iovcnt == 1) {
2363 iovp->iov_base += outcount;
2364 iovp->iov_len -= outcount;
2365 uio->uio_resid -= outcount;
2366 } else if (error = uiomove(outbuf, (long)outcount, UIO_READ, uio)) {
2368 * Reset the pointer.
2370 offset = uio->uio_loffset;
2374 zap_cursor_fini(&zc);
2375 if (uio->uio_segflg != UIO_SYSSPACE || uio->uio_iovcnt != 1)
2376 kmem_free(outbuf, bufsize);
2378 if (error == ENOENT)
2381 ZFS_ACCESSTIME_STAMP(zfsvfs, zp);
2383 uio->uio_loffset = offset;
2388 ulong_t zfs_fsync_sync_cnt = 4;
2391 zfs_fsync(vnode_t *vp, int syncflag, cred_t *cr, caller_context_t *ct)
2393 znode_t *zp = VTOZ(vp);
2394 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
2397 * Regardless of whether this is required for standards conformance,
2398 * this is the logical behavior when fsync() is called on a file with
2399 * dirty pages. We use B_ASYNC since the ZIL transactions are already
2400 * going to be pushed out as part of the zil_commit().
2402 if (vn_has_cached_data(vp) && !(syncflag & FNODSYNC) &&
2403 (vp->v_type == VREG) && !(IS_SWAPVP(vp)))
2404 (void) VOP_PUTPAGE(vp, (offset_t)0, (size_t)0, B_ASYNC, cr, ct);
2406 (void) tsd_set(zfs_fsyncer_key, (void *)zfs_fsync_sync_cnt);
2408 if (zfsvfs->z_os->os_sync != ZFS_SYNC_DISABLED) {
2411 zil_commit(zfsvfs->z_log, zp->z_id);
2416 EXPORT_SYMBOL(zfs_fsync);
2420 * Get the requested file attributes and place them in the provided
2423 * IN: vp - vnode of file.
2424 * vap - va_mask identifies requested attributes.
2425 * If AT_XVATTR set, then optional attrs are requested
2426 * flags - ATTR_NOACLCHECK (CIFS server context)
2427 * cr - credentials of caller.
2428 * ct - caller context
2430 * OUT: vap - attribute values.
2432 * RETURN: 0 (always succeeds)
2436 zfs_getattr(vnode_t *vp, vattr_t *vap, int flags, cred_t *cr,
2437 caller_context_t *ct)
2439 znode_t *zp = VTOZ(vp);
2440 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
2443 uint64_t mtime[2], ctime[2];
2444 xvattr_t *xvap = (xvattr_t *)vap; /* vap may be an xvattr_t * */
2445 xoptattr_t *xoap = NULL;
2446 boolean_t skipaclchk = (flags & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE;
2447 sa_bulk_attr_t bulk[2];
2453 zfs_fuid_map_ids(zp, cr, &vap->va_uid, &vap->va_gid);
2455 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL, &mtime, 16);
2456 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL, &ctime, 16);
2458 if ((error = sa_bulk_lookup(zp->z_sa_hdl, bulk, count)) != 0) {
2464 * If ACL is trivial don't bother looking for ACE_READ_ATTRIBUTES.
2465 * Also, if we are the owner don't bother, since owner should
2466 * always be allowed to read basic attributes of file.
2468 if (!(zp->z_pflags & ZFS_ACL_TRIVIAL) &&
2469 (vap->va_uid != crgetuid(cr))) {
2470 if ((error = zfs_zaccess(zp, ACE_READ_ATTRIBUTES, 0,
2478 * Return all attributes. It's cheaper to provide the answer
2479 * than to determine whether we were asked the question.
2482 mutex_enter(&zp->z_lock);
2483 vap->va_type = vp->v_type;
2484 vap->va_mode = zp->z_mode & MODEMASK;
2485 vap->va_fsid = zp->z_zfsvfs->z_vfs->vfs_dev;
2486 vap->va_nodeid = zp->z_id;
2487 if ((vp->v_flag & VROOT) && zfs_show_ctldir(zp))
2488 links = zp->z_links + 1;
2490 links = zp->z_links;
2491 vap->va_nlink = MIN(links, UINT32_MAX); /* nlink_t limit! */
2492 vap->va_size = zp->z_size;
2493 vap->va_rdev = vp->v_rdev;
2494 vap->va_seq = zp->z_seq;
2497 * Add in any requested optional attributes and the create time.
2498 * Also set the corresponding bits in the returned attribute bitmap.
2500 if ((xoap = xva_getxoptattr(xvap)) != NULL && zfsvfs->z_use_fuids) {
2501 if (XVA_ISSET_REQ(xvap, XAT_ARCHIVE)) {
2503 ((zp->z_pflags & ZFS_ARCHIVE) != 0);
2504 XVA_SET_RTN(xvap, XAT_ARCHIVE);
2507 if (XVA_ISSET_REQ(xvap, XAT_READONLY)) {
2508 xoap->xoa_readonly =
2509 ((zp->z_pflags & ZFS_READONLY) != 0);
2510 XVA_SET_RTN(xvap, XAT_READONLY);
2513 if (XVA_ISSET_REQ(xvap, XAT_SYSTEM)) {
2515 ((zp->z_pflags & ZFS_SYSTEM) != 0);
2516 XVA_SET_RTN(xvap, XAT_SYSTEM);
2519 if (XVA_ISSET_REQ(xvap, XAT_HIDDEN)) {
2521 ((zp->z_pflags & ZFS_HIDDEN) != 0);
2522 XVA_SET_RTN(xvap, XAT_HIDDEN);
2525 if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK)) {
2526 xoap->xoa_nounlink =
2527 ((zp->z_pflags & ZFS_NOUNLINK) != 0);
2528 XVA_SET_RTN(xvap, XAT_NOUNLINK);
2531 if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE)) {
2532 xoap->xoa_immutable =
2533 ((zp->z_pflags & ZFS_IMMUTABLE) != 0);
2534 XVA_SET_RTN(xvap, XAT_IMMUTABLE);
2537 if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY)) {
2538 xoap->xoa_appendonly =
2539 ((zp->z_pflags & ZFS_APPENDONLY) != 0);
2540 XVA_SET_RTN(xvap, XAT_APPENDONLY);
2543 if (XVA_ISSET_REQ(xvap, XAT_NODUMP)) {
2545 ((zp->z_pflags & ZFS_NODUMP) != 0);
2546 XVA_SET_RTN(xvap, XAT_NODUMP);
2549 if (XVA_ISSET_REQ(xvap, XAT_OPAQUE)) {
2551 ((zp->z_pflags & ZFS_OPAQUE) != 0);
2552 XVA_SET_RTN(xvap, XAT_OPAQUE);
2555 if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED)) {
2556 xoap->xoa_av_quarantined =
2557 ((zp->z_pflags & ZFS_AV_QUARANTINED) != 0);
2558 XVA_SET_RTN(xvap, XAT_AV_QUARANTINED);
2561 if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED)) {
2562 xoap->xoa_av_modified =
2563 ((zp->z_pflags & ZFS_AV_MODIFIED) != 0);
2564 XVA_SET_RTN(xvap, XAT_AV_MODIFIED);
2567 if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP) &&
2568 vp->v_type == VREG) {
2569 zfs_sa_get_scanstamp(zp, xvap);
2572 if (XVA_ISSET_REQ(xvap, XAT_CREATETIME)) {
2575 (void) sa_lookup(zp->z_sa_hdl, SA_ZPL_CRTIME(zfsvfs),
2576 times, sizeof (times));
2577 ZFS_TIME_DECODE(&xoap->xoa_createtime, times);
2578 XVA_SET_RTN(xvap, XAT_CREATETIME);
2581 if (XVA_ISSET_REQ(xvap, XAT_REPARSE)) {
2582 xoap->xoa_reparse = ((zp->z_pflags & ZFS_REPARSE) != 0);
2583 XVA_SET_RTN(xvap, XAT_REPARSE);
2585 if (XVA_ISSET_REQ(xvap, XAT_GEN)) {
2586 xoap->xoa_generation = zp->z_gen;
2587 XVA_SET_RTN(xvap, XAT_GEN);
2590 if (XVA_ISSET_REQ(xvap, XAT_OFFLINE)) {
2592 ((zp->z_pflags & ZFS_OFFLINE) != 0);
2593 XVA_SET_RTN(xvap, XAT_OFFLINE);
2596 if (XVA_ISSET_REQ(xvap, XAT_SPARSE)) {
2598 ((zp->z_pflags & ZFS_SPARSE) != 0);
2599 XVA_SET_RTN(xvap, XAT_SPARSE);
2603 ZFS_TIME_DECODE(&vap->va_atime, zp->z_atime);
2604 ZFS_TIME_DECODE(&vap->va_mtime, mtime);
2605 ZFS_TIME_DECODE(&vap->va_ctime, ctime);
2607 mutex_exit(&zp->z_lock);
2609 sa_object_size(zp->z_sa_hdl, &vap->va_blksize, &vap->va_nblocks);
2611 if (zp->z_blksz == 0) {
2613 * Block size hasn't been set; suggest maximal I/O transfers.
2615 vap->va_blksize = zfsvfs->z_max_blksz;
2621 EXPORT_SYMBOL(zfs_getattr);
2624 * Set the file attributes to the values contained in the
2627 * IN: vp - vnode of file to be modified.
2628 * vap - new attribute values.
2629 * If AT_XVATTR set, then optional attrs are being set
2630 * flags - ATTR_UTIME set if non-default time values provided.
2631 * - ATTR_NOACLCHECK (CIFS context only).
2632 * cr - credentials of caller.
2633 * ct - caller context
2635 * RETURN: 0 if success
2636 * error code if failure
2639 * vp - ctime updated, mtime updated if size changed.
2643 zfs_setattr(vnode_t *vp, vattr_t *vap, int flags, cred_t *cr,
2644 caller_context_t *ct)
2646 znode_t *zp = VTOZ(vp);
2647 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
2652 uint_t mask = vap->va_mask;
2656 uint64_t new_uid, new_gid;
2658 uint64_t mtime[2], ctime[2];
2660 int need_policy = FALSE;
2662 zfs_fuid_info_t *fuidp = NULL;
2663 xvattr_t *xvap = (xvattr_t *)vap; /* vap may be an xvattr_t * */
2666 boolean_t skipaclchk = (flags & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE;
2667 boolean_t fuid_dirtied = B_FALSE;
2668 sa_bulk_attr_t bulk[7], xattr_bulk[7];
2669 int count = 0, xattr_count = 0;
2674 if (mask & AT_NOSET)
2680 zilog = zfsvfs->z_log;
2683 * Make sure that if we have ephemeral uid/gid or xvattr specified
2684 * that file system is at proper version level
2687 if (zfsvfs->z_use_fuids == B_FALSE &&
2688 (((mask & AT_UID) && IS_EPHEMERAL(vap->va_uid)) ||
2689 ((mask & AT_GID) && IS_EPHEMERAL(vap->va_gid)) ||
2690 (mask & AT_XVATTR))) {
2695 if (mask & AT_SIZE && vp->v_type == VDIR) {
2700 if (mask & AT_SIZE && vp->v_type != VREG && vp->v_type != VFIFO) {
2706 * If this is an xvattr_t, then get a pointer to the structure of
2707 * optional attributes. If this is NULL, then we have a vattr_t.
2709 xoap = xva_getxoptattr(xvap);
2711 xva_init(&tmpxvattr);
2714 * Immutable files can only alter immutable bit and atime
2716 if ((zp->z_pflags & ZFS_IMMUTABLE) &&
2717 ((mask & (AT_SIZE|AT_UID|AT_GID|AT_MTIME|AT_MODE)) ||
2718 ((mask & AT_XVATTR) && XVA_ISSET_REQ(xvap, XAT_CREATETIME)))) {
2723 if ((mask & AT_SIZE) && (zp->z_pflags & ZFS_READONLY)) {
2729 * Verify timestamps doesn't overflow 32 bits.
2730 * ZFS can handle large timestamps, but 32bit syscalls can't
2731 * handle times greater than 2039. This check should be removed
2732 * once large timestamps are fully supported.
2734 if (mask & (AT_ATIME | AT_MTIME)) {
2735 if (((mask & AT_ATIME) && TIMESPEC_OVERFLOW(&vap->va_atime)) ||
2736 ((mask & AT_MTIME) && TIMESPEC_OVERFLOW(&vap->va_mtime))) {
2746 /* Can this be moved to before the top label? */
2747 if (zfsvfs->z_vfs->vfs_flag & VFS_RDONLY) {
2753 * First validate permissions
2756 if (mask & AT_SIZE) {
2757 err = zfs_zaccess(zp, ACE_WRITE_DATA, 0, skipaclchk, cr);
2763 * XXX - Note, we are not providing any open
2764 * mode flags here (like FNDELAY), so we may
2765 * block if there are locks present... this
2766 * should be addressed in openat().
2768 /* XXX - would it be OK to generate a log record here? */
2769 err = zfs_freesp(zp, vap->va_size, 0, 0, FALSE);
2776 if (mask & (AT_ATIME|AT_MTIME) ||
2777 ((mask & AT_XVATTR) && (XVA_ISSET_REQ(xvap, XAT_HIDDEN) ||
2778 XVA_ISSET_REQ(xvap, XAT_READONLY) ||
2779 XVA_ISSET_REQ(xvap, XAT_ARCHIVE) ||
2780 XVA_ISSET_REQ(xvap, XAT_OFFLINE) ||
2781 XVA_ISSET_REQ(xvap, XAT_SPARSE) ||
2782 XVA_ISSET_REQ(xvap, XAT_CREATETIME) ||
2783 XVA_ISSET_REQ(xvap, XAT_SYSTEM)))) {
2784 need_policy = zfs_zaccess(zp, ACE_WRITE_ATTRIBUTES, 0,
2788 if (mask & (AT_UID|AT_GID)) {
2789 int idmask = (mask & (AT_UID|AT_GID));
2794 * NOTE: even if a new mode is being set,
2795 * we may clear S_ISUID/S_ISGID bits.
2798 if (!(mask & AT_MODE))
2799 vap->va_mode = zp->z_mode;
2802 * Take ownership or chgrp to group we are a member of
2805 take_owner = (mask & AT_UID) && (vap->va_uid == crgetuid(cr));
2806 take_group = (mask & AT_GID) &&
2807 zfs_groupmember(zfsvfs, vap->va_gid, cr);
2810 * If both AT_UID and AT_GID are set then take_owner and
2811 * take_group must both be set in order to allow taking
2814 * Otherwise, send the check through secpolicy_vnode_setattr()
2818 if (((idmask == (AT_UID|AT_GID)) && take_owner && take_group) ||
2819 ((idmask == AT_UID) && take_owner) ||
2820 ((idmask == AT_GID) && take_group)) {
2821 if (zfs_zaccess(zp, ACE_WRITE_OWNER, 0,
2822 skipaclchk, cr) == 0) {
2824 * Remove setuid/setgid for non-privileged users
2826 secpolicy_setid_clear(vap, cr);
2827 trim_mask = (mask & (AT_UID|AT_GID));
2836 mutex_enter(&zp->z_lock);
2837 oldva.va_mode = zp->z_mode;
2838 zfs_fuid_map_ids(zp, cr, &oldva.va_uid, &oldva.va_gid);
2839 if (mask & AT_XVATTR) {
2841 * Update xvattr mask to include only those attributes
2842 * that are actually changing.
2844 * the bits will be restored prior to actually setting
2845 * the attributes so the caller thinks they were set.
2847 if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY)) {
2848 if (xoap->xoa_appendonly !=
2849 ((zp->z_pflags & ZFS_APPENDONLY) != 0)) {
2852 XVA_CLR_REQ(xvap, XAT_APPENDONLY);
2853 XVA_SET_REQ(&tmpxvattr, XAT_APPENDONLY);
2857 if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK)) {
2858 if (xoap->xoa_nounlink !=
2859 ((zp->z_pflags & ZFS_NOUNLINK) != 0)) {
2862 XVA_CLR_REQ(xvap, XAT_NOUNLINK);
2863 XVA_SET_REQ(&tmpxvattr, XAT_NOUNLINK);
2867 if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE)) {
2868 if (xoap->xoa_immutable !=
2869 ((zp->z_pflags & ZFS_IMMUTABLE) != 0)) {
2872 XVA_CLR_REQ(xvap, XAT_IMMUTABLE);
2873 XVA_SET_REQ(&tmpxvattr, XAT_IMMUTABLE);
2877 if (XVA_ISSET_REQ(xvap, XAT_NODUMP)) {
2878 if (xoap->xoa_nodump !=
2879 ((zp->z_pflags & ZFS_NODUMP) != 0)) {
2882 XVA_CLR_REQ(xvap, XAT_NODUMP);
2883 XVA_SET_REQ(&tmpxvattr, XAT_NODUMP);
2887 if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED)) {
2888 if (xoap->xoa_av_modified !=
2889 ((zp->z_pflags & ZFS_AV_MODIFIED) != 0)) {
2892 XVA_CLR_REQ(xvap, XAT_AV_MODIFIED);
2893 XVA_SET_REQ(&tmpxvattr, XAT_AV_MODIFIED);
2897 if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED)) {
2898 if ((vp->v_type != VREG &&
2899 xoap->xoa_av_quarantined) ||
2900 xoap->xoa_av_quarantined !=
2901 ((zp->z_pflags & ZFS_AV_QUARANTINED) != 0)) {
2904 XVA_CLR_REQ(xvap, XAT_AV_QUARANTINED);
2905 XVA_SET_REQ(&tmpxvattr, XAT_AV_QUARANTINED);
2909 if (XVA_ISSET_REQ(xvap, XAT_REPARSE)) {
2910 mutex_exit(&zp->z_lock);
2915 if (need_policy == FALSE &&
2916 (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP) ||
2917 XVA_ISSET_REQ(xvap, XAT_OPAQUE))) {
2922 mutex_exit(&zp->z_lock);
2924 if (mask & AT_MODE) {
2925 if (zfs_zaccess(zp, ACE_WRITE_ACL, 0, skipaclchk, cr) == 0) {
2926 err = secpolicy_setid_setsticky_clear(vp, vap,
2932 trim_mask |= AT_MODE;
2940 * If trim_mask is set then take ownership
2941 * has been granted or write_acl is present and user
2942 * has the ability to modify mode. In that case remove
2943 * UID|GID and or MODE from mask so that
2944 * secpolicy_vnode_setattr() doesn't revoke it.
2948 saved_mask = vap->va_mask;
2949 vap->va_mask &= ~trim_mask;
2951 err = secpolicy_vnode_setattr(cr, vp, vap, &oldva, flags,
2952 (int (*)(void *, int, cred_t *))zfs_zaccess_unix, zp);
2959 vap->va_mask |= saved_mask;
2963 * secpolicy_vnode_setattr, or take ownership may have
2966 mask = vap->va_mask;
2968 if ((mask & (AT_UID | AT_GID))) {
2969 err = sa_lookup(zp->z_sa_hdl, SA_ZPL_XATTR(zfsvfs),
2970 &xattr_obj, sizeof (xattr_obj));
2972 if (err == 0 && xattr_obj) {
2973 err = zfs_zget(zp->z_zfsvfs, xattr_obj, &attrzp);
2977 if (mask & AT_UID) {
2978 new_uid = zfs_fuid_create(zfsvfs,
2979 (uint64_t)vap->va_uid, cr, ZFS_OWNER, &fuidp);
2980 if (new_uid != zp->z_uid &&
2981 zfs_fuid_overquota(zfsvfs, B_FALSE, new_uid)) {
2983 VN_RELE(ZTOV(attrzp));
2989 if (mask & AT_GID) {
2990 new_gid = zfs_fuid_create(zfsvfs, (uint64_t)vap->va_gid,
2991 cr, ZFS_GROUP, &fuidp);
2992 if (new_gid != zp->z_gid &&
2993 zfs_fuid_overquota(zfsvfs, B_TRUE, new_gid)) {
2995 VN_RELE(ZTOV(attrzp));
3001 tx = dmu_tx_create(zfsvfs->z_os);
3003 if (mask & AT_MODE) {
3004 uint64_t pmode = zp->z_mode;
3006 new_mode = (pmode & S_IFMT) | (vap->va_mode & ~S_IFMT);
3008 zfs_acl_chmod_setattr(zp, &aclp, new_mode);
3010 mutex_enter(&zp->z_lock);
3011 if (!zp->z_is_sa && ((acl_obj = zfs_external_acl(zp)) != 0)) {
3013 * Are we upgrading ACL from old V0 format
3016 if (zfsvfs->z_version >= ZPL_VERSION_FUID &&
3017 zfs_znode_acl_version(zp) ==
3018 ZFS_ACL_VERSION_INITIAL) {
3019 dmu_tx_hold_free(tx, acl_obj, 0,
3021 dmu_tx_hold_write(tx, DMU_NEW_OBJECT,
3022 0, aclp->z_acl_bytes);
3024 dmu_tx_hold_write(tx, acl_obj, 0,
3027 } else if (!zp->z_is_sa && aclp->z_acl_bytes > ZFS_ACE_SPACE) {
3028 dmu_tx_hold_write(tx, DMU_NEW_OBJECT,
3029 0, aclp->z_acl_bytes);
3031 mutex_exit(&zp->z_lock);
3032 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE);
3034 if ((mask & AT_XVATTR) &&
3035 XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP))
3036 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE);
3038 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
3042 dmu_tx_hold_sa(tx, attrzp->z_sa_hdl, B_FALSE);
3045 fuid_dirtied = zfsvfs->z_fuid_dirty;
3047 zfs_fuid_txhold(zfsvfs, tx);
3049 zfs_sa_upgrade_txholds(tx, zp);
3051 err = dmu_tx_assign(tx, TXG_NOWAIT);
3053 if (err == ERESTART)
3060 * Set each attribute requested.
3061 * We group settings according to the locks they need to acquire.
3063 * Note: you cannot set ctime directly, although it will be
3064 * updated as a side-effect of calling this function.
3068 if (mask & (AT_UID|AT_GID|AT_MODE))
3069 mutex_enter(&zp->z_acl_lock);
3070 mutex_enter(&zp->z_lock);
3072 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL,
3073 &zp->z_pflags, sizeof (zp->z_pflags));
3076 if (mask & (AT_UID|AT_GID|AT_MODE))
3077 mutex_enter(&attrzp->z_acl_lock);
3078 mutex_enter(&attrzp->z_lock);
3079 SA_ADD_BULK_ATTR(xattr_bulk, xattr_count,
3080 SA_ZPL_FLAGS(zfsvfs), NULL, &attrzp->z_pflags,
3081 sizeof (attrzp->z_pflags));
3084 if (mask & (AT_UID|AT_GID)) {
3086 if (mask & AT_UID) {
3087 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zfsvfs), NULL,
3088 &new_uid, sizeof (new_uid));
3089 zp->z_uid = new_uid;
3091 SA_ADD_BULK_ATTR(xattr_bulk, xattr_count,
3092 SA_ZPL_UID(zfsvfs), NULL, &new_uid,
3094 attrzp->z_uid = new_uid;
3098 if (mask & AT_GID) {
3099 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zfsvfs),
3100 NULL, &new_gid, sizeof (new_gid));
3101 zp->z_gid = new_gid;
3103 SA_ADD_BULK_ATTR(xattr_bulk, xattr_count,
3104 SA_ZPL_GID(zfsvfs), NULL, &new_gid,
3106 attrzp->z_gid = new_gid;
3109 if (!(mask & AT_MODE)) {
3110 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs),
3111 NULL, &new_mode, sizeof (new_mode));
3112 new_mode = zp->z_mode;
3114 err = zfs_acl_chown_setattr(zp);
3117 err = zfs_acl_chown_setattr(attrzp);
3122 if (mask & AT_MODE) {
3123 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs), NULL,
3124 &new_mode, sizeof (new_mode));
3125 zp->z_mode = new_mode;
3126 ASSERT3U((uintptr_t)aclp, !=, NULL);
3127 err = zfs_aclset_common(zp, aclp, cr, tx);
3128 ASSERT3U(err, ==, 0);
3129 if (zp->z_acl_cached)
3130 zfs_acl_free(zp->z_acl_cached);
3131 zp->z_acl_cached = aclp;
3136 if (mask & AT_ATIME) {
3137 ZFS_TIME_ENCODE(&vap->va_atime, zp->z_atime);
3138 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zfsvfs), NULL,
3139 &zp->z_atime, sizeof (zp->z_atime));
3142 if (mask & AT_MTIME) {
3143 ZFS_TIME_ENCODE(&vap->va_mtime, mtime);
3144 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL,
3145 mtime, sizeof (mtime));
3148 /* XXX - shouldn't this be done *before* the ATIME/MTIME checks? */
3149 if (mask & AT_SIZE && !(mask & AT_MTIME)) {
3150 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs),
3151 NULL, mtime, sizeof (mtime));
3152 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL,
3153 &ctime, sizeof (ctime));
3154 zfs_tstamp_update_setup(zp, CONTENT_MODIFIED, mtime, ctime,
3156 } else if (mask != 0) {
3157 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL,
3158 &ctime, sizeof (ctime));
3159 zfs_tstamp_update_setup(zp, STATE_CHANGED, mtime, ctime,
3162 SA_ADD_BULK_ATTR(xattr_bulk, xattr_count,
3163 SA_ZPL_CTIME(zfsvfs), NULL,
3164 &ctime, sizeof (ctime));
3165 zfs_tstamp_update_setup(attrzp, STATE_CHANGED,
3166 mtime, ctime, B_TRUE);
3170 * Do this after setting timestamps to prevent timestamp
3171 * update from toggling bit
3174 if (xoap && (mask & AT_XVATTR)) {
3177 * restore trimmed off masks
3178 * so that return masks can be set for caller.
3181 if (XVA_ISSET_REQ(&tmpxvattr, XAT_APPENDONLY)) {
3182 XVA_SET_REQ(xvap, XAT_APPENDONLY);
3184 if (XVA_ISSET_REQ(&tmpxvattr, XAT_NOUNLINK)) {
3185 XVA_SET_REQ(xvap, XAT_NOUNLINK);
3187 if (XVA_ISSET_REQ(&tmpxvattr, XAT_IMMUTABLE)) {
3188 XVA_SET_REQ(xvap, XAT_IMMUTABLE);
3190 if (XVA_ISSET_REQ(&tmpxvattr, XAT_NODUMP)) {
3191 XVA_SET_REQ(xvap, XAT_NODUMP);
3193 if (XVA_ISSET_REQ(&tmpxvattr, XAT_AV_MODIFIED)) {
3194 XVA_SET_REQ(xvap, XAT_AV_MODIFIED);
3196 if (XVA_ISSET_REQ(&tmpxvattr, XAT_AV_QUARANTINED)) {
3197 XVA_SET_REQ(xvap, XAT_AV_QUARANTINED);
3200 if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP))
3201 ASSERT(vp->v_type == VREG);
3203 zfs_xvattr_set(zp, xvap, tx);
3207 zfs_fuid_sync(zfsvfs, tx);
3210 zfs_log_setattr(zilog, tx, TX_SETATTR, zp, vap, mask, fuidp);
3212 mutex_exit(&zp->z_lock);
3213 if (mask & (AT_UID|AT_GID|AT_MODE))
3214 mutex_exit(&zp->z_acl_lock);
3217 if (mask & (AT_UID|AT_GID|AT_MODE))
3218 mutex_exit(&attrzp->z_acl_lock);
3219 mutex_exit(&attrzp->z_lock);
3222 if (err == 0 && attrzp) {
3223 err2 = sa_bulk_update(attrzp->z_sa_hdl, xattr_bulk,
3229 VN_RELE(ZTOV(attrzp));
3234 zfs_fuid_info_free(fuidp);
3240 if (err == ERESTART)
3243 err2 = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx);
3245 zfs_inode_update(zp);
3249 if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS)
3250 zil_commit(zilog, 0);
3255 EXPORT_SYMBOL(zfs_setattr);
3257 typedef struct zfs_zlock {
3258 krwlock_t *zl_rwlock; /* lock we acquired */
3259 znode_t *zl_znode; /* znode we held */
3260 struct zfs_zlock *zl_next; /* next in list */
3264 * Drop locks and release vnodes that were held by zfs_rename_lock().
3267 zfs_rename_unlock(zfs_zlock_t **zlpp)
3271 while ((zl = *zlpp) != NULL) {
3272 if (zl->zl_znode != NULL)
3273 VN_RELE(ZTOV(zl->zl_znode));
3274 rw_exit(zl->zl_rwlock);
3275 *zlpp = zl->zl_next;
3276 kmem_free(zl, sizeof (*zl));
3281 * Search back through the directory tree, using the ".." entries.
3282 * Lock each directory in the chain to prevent concurrent renames.
3283 * Fail any attempt to move a directory into one of its own descendants.
3284 * XXX - z_parent_lock can overlap with map or grow locks
3287 zfs_rename_lock(znode_t *szp, znode_t *tdzp, znode_t *sdzp, zfs_zlock_t **zlpp)
3291 uint64_t rootid = zp->z_zfsvfs->z_root;
3292 uint64_t oidp = zp->z_id;
3293 krwlock_t *rwlp = &szp->z_parent_lock;
3294 krw_t rw = RW_WRITER;
3297 * First pass write-locks szp and compares to zp->z_id.
3298 * Later passes read-lock zp and compare to zp->z_parent.
3301 if (!rw_tryenter(rwlp, rw)) {
3303 * Another thread is renaming in this path.
3304 * Note that if we are a WRITER, we don't have any
3305 * parent_locks held yet.
3307 if (rw == RW_READER && zp->z_id > szp->z_id) {
3309 * Drop our locks and restart
3311 zfs_rename_unlock(&zl);
3315 rwlp = &szp->z_parent_lock;
3320 * Wait for other thread to drop its locks
3326 zl = kmem_alloc(sizeof (*zl), KM_SLEEP);
3327 zl->zl_rwlock = rwlp;
3328 zl->zl_znode = NULL;
3329 zl->zl_next = *zlpp;
3332 if (oidp == szp->z_id) /* We're a descendant of szp */
3335 if (oidp == rootid) /* We've hit the top */
3338 if (rw == RW_READER) { /* i.e. not the first pass */
3339 int error = zfs_zget(zp->z_zfsvfs, oidp, &zp);
3344 (void) sa_lookup(zp->z_sa_hdl, SA_ZPL_PARENT(zp->z_zfsvfs),
3345 &oidp, sizeof (oidp));
3346 rwlp = &zp->z_parent_lock;
3349 } while (zp->z_id != sdzp->z_id);
3355 * Move an entry from the provided source directory to the target
3356 * directory. Change the entry name as indicated.
3358 * IN: sdvp - Source directory containing the "old entry".
3359 * snm - Old entry name.
3360 * tdvp - Target directory to contain the "new entry".
3361 * tnm - New entry name.
3362 * cr - credentials of caller.
3363 * ct - caller context
3364 * flags - case flags
3366 * RETURN: 0 if success
3367 * error code if failure
3370 * sdvp,tdvp - ctime|mtime updated
3374 zfs_rename(vnode_t *sdvp, char *snm, vnode_t *tdvp, char *tnm, cred_t *cr,
3375 caller_context_t *ct, int flags)
3377 znode_t *tdzp, *szp, *tzp;
3378 znode_t *sdzp = VTOZ(sdvp);
3379 zfsvfs_t *zfsvfs = sdzp->z_zfsvfs;
3382 zfs_dirlock_t *sdl, *tdl;
3385 int cmp, serr, terr;
3390 ZFS_VERIFY_ZP(sdzp);
3391 zilog = zfsvfs->z_log;
3394 * Make sure we have the real vp for the target directory.
3396 if (VOP_REALVP(tdvp, &realvp, ct) == 0)
3399 if (tdvp->v_vfsp != sdvp->v_vfsp || zfsctl_is_node(tdvp)) {
3405 ZFS_VERIFY_ZP(tdzp);
3406 if (zfsvfs->z_utf8 && u8_validate(tnm,
3407 strlen(tnm), NULL, U8_VALIDATE_ENTIRE, &error) < 0) {
3412 if (flags & FIGNORECASE)
3421 * This is to prevent the creation of links into attribute space
3422 * by renaming a linked file into/outof an attribute directory.
3423 * See the comment in zfs_link() for why this is considered bad.
3425 if ((tdzp->z_pflags & ZFS_XATTR) != (sdzp->z_pflags & ZFS_XATTR)) {
3431 * Lock source and target directory entries. To prevent deadlock,
3432 * a lock ordering must be defined. We lock the directory with
3433 * the smallest object id first, or if it's a tie, the one with
3434 * the lexically first name.
3436 if (sdzp->z_id < tdzp->z_id) {
3438 } else if (sdzp->z_id > tdzp->z_id) {
3442 * First compare the two name arguments without
3443 * considering any case folding.
3445 int nofold = (zfsvfs->z_norm & ~U8_TEXTPREP_TOUPPER);
3447 cmp = u8_strcmp(snm, tnm, 0, nofold, U8_UNICODE_LATEST, &error);
3448 ASSERT(error == 0 || !zfsvfs->z_utf8);
3451 * POSIX: "If the old argument and the new argument
3452 * both refer to links to the same existing file,
3453 * the rename() function shall return successfully
3454 * and perform no other action."
3460 * If the file system is case-folding, then we may
3461 * have some more checking to do. A case-folding file
3462 * system is either supporting mixed case sensitivity
3463 * access or is completely case-insensitive. Note
3464 * that the file system is always case preserving.
3466 * In mixed sensitivity mode case sensitive behavior
3467 * is the default. FIGNORECASE must be used to
3468 * explicitly request case insensitive behavior.
3470 * If the source and target names provided differ only
3471 * by case (e.g., a request to rename 'tim' to 'Tim'),
3472 * we will treat this as a special case in the
3473 * case-insensitive mode: as long as the source name
3474 * is an exact match, we will allow this to proceed as
3475 * a name-change request.
3477 if ((zfsvfs->z_case == ZFS_CASE_INSENSITIVE ||
3478 (zfsvfs->z_case == ZFS_CASE_MIXED &&
3479 flags & FIGNORECASE)) &&
3480 u8_strcmp(snm, tnm, 0, zfsvfs->z_norm, U8_UNICODE_LATEST,
3483 * case preserving rename request, require exact
3492 * If the source and destination directories are the same, we should
3493 * grab the z_name_lock of that directory only once.
3497 rw_enter(&sdzp->z_name_lock, RW_READER);
3501 serr = zfs_dirent_lock(&sdl, sdzp, snm, &szp,
3502 ZEXISTS | zflg, NULL, NULL);
3503 terr = zfs_dirent_lock(&tdl,
3504 tdzp, tnm, &tzp, ZRENAMING | zflg, NULL, NULL);
3506 terr = zfs_dirent_lock(&tdl,
3507 tdzp, tnm, &tzp, zflg, NULL, NULL);
3508 serr = zfs_dirent_lock(&sdl,
3509 sdzp, snm, &szp, ZEXISTS | ZRENAMING | zflg,
3515 * Source entry invalid or not there.
3518 zfs_dirent_unlock(tdl);
3524 rw_exit(&sdzp->z_name_lock);
3526 if (strcmp(snm, "..") == 0)
3532 zfs_dirent_unlock(sdl);
3536 rw_exit(&sdzp->z_name_lock);
3538 if (strcmp(tnm, "..") == 0)
3545 * Must have write access at the source to remove the old entry
3546 * and write access at the target to create the new entry.
3547 * Note that if target and source are the same, this can be
3548 * done in a single check.
3551 if ((error = zfs_zaccess_rename(sdzp, szp, tdzp, tzp, cr)))
3554 if (ZTOV(szp)->v_type == VDIR) {
3556 * Check to make sure rename is valid.
3557 * Can't do a move like this: /usr/a/b to /usr/a/b/c/d
3559 if ((error = zfs_rename_lock(szp, tdzp, sdzp, &zl)))
3564 * Does target exist?
3568 * Source and target must be the same type.
3570 if (ZTOV(szp)->v_type == VDIR) {
3571 if (ZTOV(tzp)->v_type != VDIR) {
3576 if (ZTOV(tzp)->v_type == VDIR) {
3582 * POSIX dictates that when the source and target
3583 * entries refer to the same file object, rename
3584 * must do nothing and exit without error.
3586 if (szp->z_id == tzp->z_id) {
3592 vnevent_rename_src(ZTOV(szp), sdvp, snm, ct);
3594 vnevent_rename_dest(ZTOV(tzp), tdvp, tnm, ct);
3597 * notify the target directory if it is not the same
3598 * as source directory.
3601 vnevent_rename_dest_dir(tdvp, ct);
3604 tx = dmu_tx_create(zfsvfs->z_os);
3605 dmu_tx_hold_sa(tx, szp->z_sa_hdl, B_FALSE);
3606 dmu_tx_hold_sa(tx, sdzp->z_sa_hdl, B_FALSE);
3607 dmu_tx_hold_zap(tx, sdzp->z_id, FALSE, snm);
3608 dmu_tx_hold_zap(tx, tdzp->z_id, TRUE, tnm);
3610 dmu_tx_hold_sa(tx, tdzp->z_sa_hdl, B_FALSE);
3611 zfs_sa_upgrade_txholds(tx, tdzp);
3614 dmu_tx_hold_sa(tx, tzp->z_sa_hdl, B_FALSE);
3615 zfs_sa_upgrade_txholds(tx, tzp);
3618 zfs_sa_upgrade_txholds(tx, szp);
3619 dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL);
3620 error = dmu_tx_assign(tx, TXG_NOWAIT);
3623 zfs_rename_unlock(&zl);
3624 zfs_dirent_unlock(sdl);
3625 zfs_dirent_unlock(tdl);
3628 rw_exit(&sdzp->z_name_lock);
3633 if (error == ERESTART) {
3643 if (tzp) /* Attempt to remove the existing target */
3644 error = zfs_link_destroy(tdl, tzp, tx, zflg, NULL);
3647 error = zfs_link_create(tdl, szp, tx, ZRENAMING);
3649 szp->z_pflags |= ZFS_AV_MODIFIED;
3651 error = sa_update(szp->z_sa_hdl, SA_ZPL_FLAGS(zfsvfs),
3652 (void *)&szp->z_pflags, sizeof (uint64_t), tx);
3653 ASSERT3U(error, ==, 0);
3655 error = zfs_link_destroy(sdl, szp, tx, ZRENAMING, NULL);
3657 zfs_log_rename(zilog, tx, TX_RENAME |
3658 (flags & FIGNORECASE ? TX_CI : 0), sdzp,
3659 sdl->dl_name, tdzp, tdl->dl_name, szp);
3662 * Update path information for the target vnode
3664 vn_renamepath(tdvp, ZTOV(szp), tnm,
3668 * At this point, we have successfully created
3669 * the target name, but have failed to remove
3670 * the source name. Since the create was done
3671 * with the ZRENAMING flag, there are
3672 * complications; for one, the link count is
3673 * wrong. The easiest way to deal with this
3674 * is to remove the newly created target, and
3675 * return the original error. This must
3676 * succeed; fortunately, it is very unlikely to
3677 * fail, since we just created it.
3679 VERIFY3U(zfs_link_destroy(tdl, szp, tx,
3680 ZRENAMING, NULL), ==, 0);
3688 zfs_rename_unlock(&zl);
3690 zfs_dirent_unlock(sdl);
3691 zfs_dirent_unlock(tdl);
3693 zfs_inode_update(sdzp);
3695 rw_exit(&sdzp->z_name_lock);
3698 zfs_inode_update(tdzp);
3700 zfs_inode_update(szp);
3703 zfs_inode_update(tzp);
3707 if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS)
3708 zil_commit(zilog, 0);
3713 EXPORT_SYMBOL(zfs_rename);
3716 * Insert the indicated symbolic reference entry into the directory.
3718 * IN: dvp - Directory to contain new symbolic link.
3719 * link - Name for new symlink entry.
3720 * vap - Attributes of new entry.
3721 * target - Target path of new symlink.
3722 * cr - credentials of caller.
3723 * ct - caller context
3724 * flags - case flags
3726 * RETURN: 0 if success
3727 * error code if failure
3730 * dvp - ctime|mtime updated
3734 zfs_symlink(vnode_t *dvp, char *name, vattr_t *vap, char *link, cred_t *cr,
3735 caller_context_t *ct, int flags)
3737 znode_t *zp, *dzp = VTOZ(dvp);
3740 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
3742 uint64_t len = strlen(link);
3745 zfs_acl_ids_t acl_ids;
3746 boolean_t fuid_dirtied;
3747 uint64_t txtype = TX_SYMLINK;
3749 ASSERT(vap->va_type == VLNK);
3753 zilog = zfsvfs->z_log;
3755 if (zfsvfs->z_utf8 && u8_validate(name, strlen(name),
3756 NULL, U8_VALIDATE_ENTIRE, &error) < 0) {
3760 if (flags & FIGNORECASE)
3763 if (len > MAXPATHLEN) {
3765 return (ENAMETOOLONG);
3768 if ((error = zfs_acl_ids_create(dzp, 0,
3769 vap, cr, NULL, &acl_ids)) != 0) {
3775 * Attempt to lock directory; fail if entry already exists.
3777 error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg, NULL, NULL);
3779 zfs_acl_ids_free(&acl_ids);
3784 if ((error = zfs_zaccess(dzp, ACE_ADD_FILE, 0, B_FALSE, cr))) {
3785 zfs_acl_ids_free(&acl_ids);
3786 zfs_dirent_unlock(dl);
3791 if (zfs_acl_ids_overquota(zfsvfs, &acl_ids)) {
3792 zfs_acl_ids_free(&acl_ids);
3793 zfs_dirent_unlock(dl);
3797 tx = dmu_tx_create(zfsvfs->z_os);
3798 fuid_dirtied = zfsvfs->z_fuid_dirty;
3799 dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, MAX(1, len));
3800 dmu_tx_hold_zap(tx, dzp->z_id, TRUE, name);
3801 dmu_tx_hold_sa_create(tx, acl_ids.z_aclp->z_acl_bytes +
3802 ZFS_SA_BASE_ATTR_SIZE + len);
3803 dmu_tx_hold_sa(tx, dzp->z_sa_hdl, B_FALSE);
3804 if (!zfsvfs->z_use_sa && acl_ids.z_aclp->z_acl_bytes > ZFS_ACE_SPACE) {
3805 dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0,
3806 acl_ids.z_aclp->z_acl_bytes);
3809 zfs_fuid_txhold(zfsvfs, tx);
3810 error = dmu_tx_assign(tx, TXG_NOWAIT);
3812 zfs_dirent_unlock(dl);
3813 if (error == ERESTART) {
3818 zfs_acl_ids_free(&acl_ids);
3825 * Create a new object for the symlink.
3826 * for version 4 ZPL datsets the symlink will be an SA attribute
3828 zfs_mknode(dzp, vap, tx, cr, 0, &zp, &acl_ids);
3831 zfs_fuid_sync(zfsvfs, tx);
3833 mutex_enter(&zp->z_lock);
3835 error = sa_update(zp->z_sa_hdl, SA_ZPL_SYMLINK(zfsvfs),
3838 zfs_sa_symlink(zp, link, len, tx);
3839 mutex_exit(&zp->z_lock);
3842 (void) sa_update(zp->z_sa_hdl, SA_ZPL_SIZE(zfsvfs),
3843 &zp->z_size, sizeof (zp->z_size), tx);
3845 * Insert the new object into the directory.
3847 (void) zfs_link_create(dl, zp, tx, ZNEW);
3849 if (flags & FIGNORECASE)
3851 zfs_log_symlink(zilog, tx, txtype, dzp, zp, name, link);
3853 zfs_inode_update(dzp);
3854 zfs_inode_update(zp);
3856 zfs_acl_ids_free(&acl_ids);
3860 zfs_dirent_unlock(dl);
3864 if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS)
3865 zil_commit(zilog, 0);
3870 EXPORT_SYMBOL(zfs_symlink);
3873 * Return, in the buffer contained in the provided uio structure,
3874 * the symbolic path referred to by vp.
3876 * IN: vp - vnode of symbolic link.
3877 * uoip - structure to contain the link path.
3878 * cr - credentials of caller.
3879 * ct - caller context
3881 * OUT: uio - structure to contain the link path.
3883 * RETURN: 0 if success
3884 * error code if failure
3887 * vp - atime updated
3891 zfs_readlink(vnode_t *vp, uio_t *uio, cred_t *cr, caller_context_t *ct)
3893 znode_t *zp = VTOZ(vp);
3894 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
3900 mutex_enter(&zp->z_lock);
3902 error = sa_lookup_uio(zp->z_sa_hdl,
3903 SA_ZPL_SYMLINK(zfsvfs), uio);
3905 error = zfs_sa_readlink(zp, uio);
3906 mutex_exit(&zp->z_lock);
3908 ZFS_ACCESSTIME_STAMP(zfsvfs, zp);
3910 zfs_inode_update(zp);
3914 EXPORT_SYMBOL(zfs_readlink);
3917 * Insert a new entry into directory tdvp referencing svp.
3919 * IN: tdvp - Directory to contain new entry.
3920 * svp - vnode of new entry.
3921 * name - name of new entry.
3922 * cr - credentials of caller.
3923 * ct - caller context
3925 * RETURN: 0 if success
3926 * error code if failure
3929 * tdvp - ctime|mtime updated
3930 * svp - ctime updated
3934 zfs_link(vnode_t *tdvp, vnode_t *svp, char *name, cred_t *cr,
3935 caller_context_t *ct, int flags)
3937 znode_t *dzp = VTOZ(tdvp);
3939 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
3949 ASSERT(tdvp->v_type == VDIR);
3953 zilog = zfsvfs->z_log;
3955 if (VOP_REALVP(svp, &realvp, ct) == 0)
3959 * POSIX dictates that we return EPERM here.
3960 * Better choices include ENOTSUP or EISDIR.
3962 if (svp->v_type == VDIR) {
3967 if (svp->v_vfsp != tdvp->v_vfsp || zfsctl_is_node(svp)) {
3975 /* Prevent links to .zfs/shares files */
3977 if ((error = sa_lookup(szp->z_sa_hdl, SA_ZPL_PARENT(zfsvfs),
3978 &parent, sizeof (uint64_t))) != 0) {
3982 if (parent == zfsvfs->z_shares_dir) {
3987 if (zfsvfs->z_utf8 && u8_validate(name,
3988 strlen(name), NULL, U8_VALIDATE_ENTIRE, &error) < 0) {
3992 if (flags & FIGNORECASE)
3996 * We do not support links between attributes and non-attributes
3997 * because of the potential security risk of creating links
3998 * into "normal" file space in order to circumvent restrictions
3999 * imposed in attribute space.
4001 if ((szp->z_pflags & ZFS_XATTR) != (dzp->z_pflags & ZFS_XATTR)) {
4007 owner = zfs_fuid_map_id(zfsvfs, szp->z_uid, cr, ZFS_OWNER);
4008 if (owner != crgetuid(cr) && secpolicy_basic_link(cr) != 0) {
4013 if ((error = zfs_zaccess(dzp, ACE_ADD_FILE, 0, B_FALSE, cr))) {
4020 * Attempt to lock directory; fail if entry already exists.
4022 error = zfs_dirent_lock(&dl, dzp, name, &tzp, zf, NULL, NULL);
4028 tx = dmu_tx_create(zfsvfs->z_os);
4029 dmu_tx_hold_sa(tx, szp->z_sa_hdl, B_FALSE);
4030 dmu_tx_hold_zap(tx, dzp->z_id, TRUE, name);
4031 zfs_sa_upgrade_txholds(tx, szp);
4032 zfs_sa_upgrade_txholds(tx, dzp);
4033 error = dmu_tx_assign(tx, TXG_NOWAIT);
4035 zfs_dirent_unlock(dl);
4036 if (error == ERESTART) {
4046 error = zfs_link_create(dl, szp, tx, 0);
4049 uint64_t txtype = TX_LINK;
4050 if (flags & FIGNORECASE)
4052 zfs_log_link(zilog, tx, txtype, dzp, szp, name);
4057 zfs_dirent_unlock(dl);
4060 vnevent_link(svp, ct);
4063 if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS)
4064 zil_commit(zilog, 0);
4066 zfs_inode_update(dzp);
4067 zfs_inode_update(szp);
4071 EXPORT_SYMBOL(zfs_link);
4074 * zfs_null_putapage() is used when the file system has been force
4075 * unmounted. It just drops the pages.
4079 zfs_null_putapage(vnode_t *vp, page_t *pp, u_offset_t *offp,
4080 size_t *lenp, int flags, cred_t *cr)
4082 pvn_write_done(pp, B_INVAL|B_FORCE|B_ERROR);
4087 * Push a page out to disk, klustering if possible.
4089 * IN: vp - file to push page to.
4090 * pp - page to push.
4091 * flags - additional flags.
4092 * cr - credentials of caller.
4094 * OUT: offp - start of range pushed.
4095 * lenp - len of range pushed.
4097 * RETURN: 0 if success
4098 * error code if failure
4100 * NOTE: callers must have locked the page to be pushed. On
4101 * exit, the page (and all other pages in the kluster) must be
4106 zfs_putapage(vnode_t *vp, page_t *pp, u_offset_t *offp,
4107 size_t *lenp, int flags, cred_t *cr)
4109 znode_t *zp = VTOZ(vp);
4110 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
4112 u_offset_t off, koff;
4119 * If our blocksize is bigger than the page size, try to kluster
4120 * multiple pages so that we write a full block (thus avoiding
4121 * a read-modify-write).
4123 if (off < zp->z_size && zp->z_blksz > PAGESIZE) {
4124 klen = P2ROUNDUP((ulong_t)zp->z_blksz, PAGESIZE);
4125 koff = ISP2(klen) ? P2ALIGN(off, (u_offset_t)klen) : 0;
4126 ASSERT(koff <= zp->z_size);
4127 if (koff + klen > zp->z_size)
4128 klen = P2ROUNDUP(zp->z_size - koff, (uint64_t)PAGESIZE);
4129 pp = pvn_write_kluster(vp, pp, &off, &len, koff, klen, flags);
4131 ASSERT3U(btop(len), ==, btopr(len));
4134 * Can't push pages past end-of-file.
4136 if (off >= zp->z_size) {
4137 /* ignore all pages */
4140 } else if (off + len > zp->z_size) {
4141 int npages = btopr(zp->z_size - off);
4144 page_list_break(&pp, &trunc, npages);
4145 /* ignore pages past end of file */
4147 pvn_write_done(trunc, flags);
4148 len = zp->z_size - off;
4151 if (zfs_owner_overquota(zfsvfs, zp, B_FALSE) ||
4152 zfs_owner_overquota(zfsvfs, zp, B_TRUE)) {
4157 tx = dmu_tx_create(zfsvfs->z_os);
4158 dmu_tx_hold_write(tx, zp->z_id, off, len);
4160 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
4161 zfs_sa_upgrade_txholds(tx, zp);
4162 err = dmu_tx_assign(tx, TXG_NOWAIT);
4164 if (err == ERESTART) {
4173 if (zp->z_blksz <= PAGESIZE) {
4174 caddr_t va = zfs_map_page(pp, S_READ);
4175 ASSERT3U(len, <=, PAGESIZE);
4176 dmu_write(zfsvfs->z_os, zp->z_id, off, len, va, tx);
4177 zfs_unmap_page(pp, va);
4179 err = dmu_write_pages(zfsvfs->z_os, zp->z_id, off, len, pp, tx);
4183 uint64_t mtime[2], ctime[2];
4184 sa_bulk_attr_t bulk[3];
4187 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL,
4189 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL,
4191 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL,
4193 zfs_tstamp_update_setup(zp, CONTENT_MODIFIED, mtime, ctime,
4195 zfs_log_write(zfsvfs->z_log, tx, TX_WRITE, zp, off, len, 0);
4200 pvn_write_done(pp, (err ? B_ERROR : 0) | flags);
4210 * Copy the portion of the file indicated from pages into the file.
4211 * The pages are stored in a page list attached to the files vnode.
4213 * IN: vp - vnode of file to push page data to.
4214 * off - position in file to put data.
4215 * len - amount of data to write.
4216 * flags - flags to control the operation.
4217 * cr - credentials of caller.
4218 * ct - caller context.
4220 * RETURN: 0 if success
4221 * error code if failure
4224 * vp - ctime|mtime updated
4228 zfs_putpage(vnode_t *vp, offset_t off, size_t len, int flags, cred_t *cr,
4229 caller_context_t *ct)
4231 znode_t *zp = VTOZ(vp);
4232 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
4244 * Align this request to the file block size in case we kluster.
4245 * XXX - this can result in pretty aggresive locking, which can
4246 * impact simultanious read/write access. One option might be
4247 * to break up long requests (len == 0) into block-by-block
4248 * operations to get narrower locking.
4250 blksz = zp->z_blksz;
4252 io_off = P2ALIGN_TYPED(off, blksz, u_offset_t);
4255 if (len > 0 && ISP2(blksz))
4256 io_len = P2ROUNDUP_TYPED(len + (off - io_off), blksz, size_t);
4262 * Search the entire vp list for pages >= io_off.
4264 rl = zfs_range_lock(zp, io_off, UINT64_MAX, RL_WRITER);
4265 error = pvn_vplist_dirty(vp, io_off, zfs_putapage, flags, cr);
4268 rl = zfs_range_lock(zp, io_off, io_len, RL_WRITER);
4270 if (off > zp->z_size) {
4271 /* past end of file */
4272 zfs_range_unlock(rl);
4277 len = MIN(io_len, P2ROUNDUP(zp->z_size, PAGESIZE) - io_off);
4279 for (off = io_off; io_off < off + len; io_off += io_len) {
4280 if ((flags & B_INVAL) || ((flags & B_ASYNC) == 0)) {
4281 pp = page_lookup(vp, io_off,
4282 (flags & (B_INVAL | B_FREE)) ? SE_EXCL : SE_SHARED);
4284 pp = page_lookup_nowait(vp, io_off,
4285 (flags & B_FREE) ? SE_EXCL : SE_SHARED);
4288 if (pp != NULL && pvn_getdirty(pp, flags)) {
4292 * Found a dirty page to push
4294 err = zfs_putapage(vp, pp, &io_off, &io_len, flags, cr);
4302 zfs_range_unlock(rl);
4303 if ((flags & B_ASYNC) == 0 || zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS)
4304 zil_commit(zfsvfs->z_log, zp->z_id);
4311 zfs_inactive(vnode_t *vp, cred_t *cr, caller_context_t *ct)
4313 znode_t *zp = VTOZ(vp);
4314 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
4317 rw_enter(&zfsvfs->z_teardown_inactive_lock, RW_READER);
4318 if (zp->z_sa_hdl == NULL) {
4320 * The fs has been unmounted, or we did a
4321 * suspend/resume and this file no longer exists.
4323 if (vn_has_cached_data(vp)) {
4324 (void) pvn_vplist_dirty(vp, 0, zfs_null_putapage,
4328 mutex_enter(&zp->z_lock);
4329 mutex_enter(&vp->v_lock);
4330 ASSERT(vp->v_count == 1);
4332 mutex_exit(&vp->v_lock);
4333 mutex_exit(&zp->z_lock);
4334 rw_exit(&zfsvfs->z_teardown_inactive_lock);
4340 * Attempt to push any data in the page cache. If this fails
4341 * we will get kicked out later in zfs_zinactive().
4343 if (vn_has_cached_data(vp)) {
4344 (void) pvn_vplist_dirty(vp, 0, zfs_putapage, B_INVAL|B_ASYNC,
4348 if (zp->z_atime_dirty && zp->z_unlinked == 0) {
4349 dmu_tx_t *tx = dmu_tx_create(zfsvfs->z_os);
4351 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
4352 zfs_sa_upgrade_txholds(tx, zp);
4353 error = dmu_tx_assign(tx, TXG_WAIT);
4357 mutex_enter(&zp->z_lock);
4358 (void) sa_update(zp->z_sa_hdl, SA_ZPL_ATIME(zfsvfs),
4359 (void *)&zp->z_atime, sizeof (zp->z_atime), tx);
4360 zp->z_atime_dirty = 0;
4361 mutex_exit(&zp->z_lock);
4367 rw_exit(&zfsvfs->z_teardown_inactive_lock);
4369 EXPORT_SYMBOL(zfs_inactive);
4372 * Bounds-check the seek operation.
4374 * IN: vp - vnode seeking within
4375 * ooff - old file offset
4376 * noffp - pointer to new file offset
4377 * ct - caller context
4379 * RETURN: 0 if success
4380 * EINVAL if new offset invalid
4384 zfs_seek(vnode_t *vp, offset_t ooff, offset_t *noffp,
4385 caller_context_t *ct)
4387 if (vp->v_type == VDIR)
4389 return ((*noffp < 0 || *noffp > MAXOFFSET_T) ? EINVAL : 0);
4393 * Pre-filter the generic locking function to trap attempts to place
4394 * a mandatory lock on a memory mapped file.
4397 zfs_frlock(vnode_t *vp, int cmd, flock64_t *bfp, int flag, offset_t offset,
4398 flk_callback_t *flk_cbp, cred_t *cr, caller_context_t *ct)
4400 znode_t *zp = VTOZ(vp);
4401 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
4407 * We are following the UFS semantics with respect to mapcnt
4408 * here: If we see that the file is mapped already, then we will
4409 * return an error, but we don't worry about races between this
4410 * function and zfs_map().
4412 if (zp->z_mapcnt > 0 && MANDMODE(zp->z_mode)) {
4417 return (fs_frlock(vp, cmd, bfp, flag, offset, flk_cbp, cr, ct));
4421 * If we can't find a page in the cache, we will create a new page
4422 * and fill it with file data. For efficiency, we may try to fill
4423 * multiple pages at once (klustering) to fill up the supplied page
4424 * list. Note that the pages to be filled are held with an exclusive
4425 * lock to prevent access by other threads while they are being filled.
4428 zfs_fillpage(vnode_t *vp, u_offset_t off, struct seg *seg,
4429 caddr_t addr, page_t *pl[], size_t plsz, enum seg_rw rw)
4431 znode_t *zp = VTOZ(vp);
4432 page_t *pp, *cur_pp;
4433 objset_t *os = zp->z_zfsvfs->z_os;
4434 u_offset_t io_off, total;
4438 if (plsz == PAGESIZE || zp->z_blksz <= PAGESIZE) {
4440 * We only have a single page, don't bother klustering
4444 pp = page_create_va(vp, io_off, io_len,
4445 PG_EXCL | PG_WAIT, seg, addr);
4448 * Try to find enough pages to fill the page list
4450 pp = pvn_read_kluster(vp, off, seg, addr, &io_off,
4451 &io_len, off, plsz, 0);
4455 * The page already exists, nothing to do here.
4462 * Fill the pages in the kluster.
4465 for (total = io_off + io_len; io_off < total; io_off += PAGESIZE) {
4468 ASSERT3U(io_off, ==, cur_pp->p_offset);
4469 va = zfs_map_page(cur_pp, S_WRITE);
4470 err = dmu_read(os, zp->z_id, io_off, PAGESIZE, va,
4472 zfs_unmap_page(cur_pp, va);
4474 /* On error, toss the entire kluster */
4475 pvn_read_done(pp, B_ERROR);
4476 /* convert checksum errors into IO errors */
4481 cur_pp = cur_pp->p_next;
4485 * Fill in the page list array from the kluster starting
4486 * from the desired offset `off'.
4487 * NOTE: the page list will always be null terminated.
4489 pvn_plist_init(pp, pl, plsz, off, io_len, rw);
4490 ASSERT(pl == NULL || (*pl)->p_offset == off);
4496 * Return pointers to the pages for the file region [off, off + len]
4497 * in the pl array. If plsz is greater than len, this function may
4498 * also return page pointers from after the specified region
4499 * (i.e. the region [off, off + plsz]). These additional pages are
4500 * only returned if they are already in the cache, or were created as
4501 * part of a klustered read.
4503 * IN: vp - vnode of file to get data from.
4504 * off - position in file to get data from.
4505 * len - amount of data to retrieve.
4506 * plsz - length of provided page list.
4507 * seg - segment to obtain pages for.
4508 * addr - virtual address of fault.
4509 * rw - mode of created pages.
4510 * cr - credentials of caller.
4511 * ct - caller context.
4513 * OUT: protp - protection mode of created pages.
4514 * pl - list of pages created.
4516 * RETURN: 0 if success
4517 * error code if failure
4520 * vp - atime updated
4524 zfs_getpage(vnode_t *vp, offset_t off, size_t len, uint_t *protp,
4525 page_t *pl[], size_t plsz, struct seg *seg, caddr_t addr,
4526 enum seg_rw rw, cred_t *cr, caller_context_t *ct)
4528 znode_t *zp = VTOZ(vp);
4529 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
4533 /* we do our own caching, faultahead is unnecessary */
4536 else if (len > plsz)
4539 len = P2ROUNDUP(len, PAGESIZE);
4540 ASSERT(plsz >= len);
4549 * Loop through the requested range [off, off + len) looking
4550 * for pages. If we don't find a page, we will need to create
4551 * a new page and fill it with data from the file.
4554 if (*pl = page_lookup(vp, off, SE_SHARED))
4556 else if (err = zfs_fillpage(vp, off, seg, addr, pl, plsz, rw))
4559 ASSERT3U((*pl)->p_offset, ==, off);
4563 ASSERT3U(len, >=, PAGESIZE);
4566 ASSERT3U(plsz, >=, PAGESIZE);
4573 * Fill out the page array with any pages already in the cache.
4576 (*pl++ = page_lookup_nowait(vp, off, SE_SHARED))) {
4583 * Release any pages we have previously locked.
4588 ZFS_ACCESSTIME_STAMP(zfsvfs, zp);
4598 * Request a memory map for a section of a file. This code interacts
4599 * with common code and the VM system as follows:
4601 * common code calls mmap(), which ends up in smmap_common()
4603 * this calls VOP_MAP(), which takes you into (say) zfs
4605 * zfs_map() calls as_map(), passing segvn_create() as the callback
4607 * segvn_create() creates the new segment and calls VOP_ADDMAP()
4609 * zfs_addmap() updates z_mapcnt
4613 zfs_map(vnode_t *vp, offset_t off, struct as *as, caddr_t *addrp,
4614 size_t len, uchar_t prot, uchar_t maxprot, uint_t flags, cred_t *cr,
4615 caller_context_t *ct)
4617 znode_t *zp = VTOZ(vp);
4618 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
4619 segvn_crargs_t vn_a;
4625 if ((prot & PROT_WRITE) && (zp->z_pflags &
4626 (ZFS_IMMUTABLE | ZFS_READONLY | ZFS_APPENDONLY))) {
4631 if ((prot & (PROT_READ | PROT_EXEC)) &&
4632 (zp->z_pflags & ZFS_AV_QUARANTINED)) {
4637 if (vp->v_flag & VNOMAP) {
4642 if (off < 0 || len > MAXOFFSET_T - off) {
4647 if (vp->v_type != VREG) {
4653 * If file is locked, disallow mapping.
4655 if (MANDMODE(zp->z_mode) && vn_has_flocks(vp)) {
4661 error = choose_addr(as, addrp, len, off, ADDR_VACALIGN, flags);
4669 vn_a.offset = (u_offset_t)off;
4670 vn_a.type = flags & MAP_TYPE;
4672 vn_a.maxprot = maxprot;
4675 vn_a.flags = flags & ~MAP_TYPE;
4677 vn_a.lgrp_mem_policy_flags = 0;
4679 error = as_map(as, *addrp, len, segvn_create, &vn_a);
4688 zfs_addmap(vnode_t *vp, offset_t off, struct as *as, caddr_t addr,
4689 size_t len, uchar_t prot, uchar_t maxprot, uint_t flags, cred_t *cr,
4690 caller_context_t *ct)
4692 uint64_t pages = btopr(len);
4694 atomic_add_64(&VTOZ(vp)->z_mapcnt, pages);
4699 * The reason we push dirty pages as part of zfs_delmap() is so that we get a
4700 * more accurate mtime for the associated file. Since we don't have a way of
4701 * detecting when the data was actually modified, we have to resort to
4702 * heuristics. If an explicit msync() is done, then we mark the mtime when the
4703 * last page is pushed. The problem occurs when the msync() call is omitted,
4704 * which by far the most common case:
4712 * putpage() via fsflush
4714 * If we wait until fsflush to come along, we can have a modification time that
4715 * is some arbitrary point in the future. In order to prevent this in the
4716 * common case, we flush pages whenever a (MAP_SHARED, PROT_WRITE) mapping is
4721 zfs_delmap(vnode_t *vp, offset_t off, struct as *as, caddr_t addr,
4722 size_t len, uint_t prot, uint_t maxprot, uint_t flags, cred_t *cr,
4723 caller_context_t *ct)
4725 uint64_t pages = btopr(len);
4727 ASSERT3U(VTOZ(vp)->z_mapcnt, >=, pages);
4728 atomic_add_64(&VTOZ(vp)->z_mapcnt, -pages);
4730 if ((flags & MAP_SHARED) && (prot & PROT_WRITE) &&
4731 vn_has_cached_data(vp))
4732 (void) VOP_PUTPAGE(vp, off, len, B_ASYNC, cr, ct);
4738 * Free or allocate space in a file. Currently, this function only
4739 * supports the `F_FREESP' command. However, this command is somewhat
4740 * misnamed, as its functionality includes the ability to allocate as
4741 * well as free space.
4743 * IN: vp - vnode of file to free data in.
4744 * cmd - action to take (only F_FREESP supported).
4745 * bfp - section of file to free/alloc.
4746 * flag - current file open mode flags.
4747 * offset - current file offset.
4748 * cr - credentials of caller [UNUSED].
4749 * ct - caller context.
4751 * RETURN: 0 if success
4752 * error code if failure
4755 * vp - ctime|mtime updated
4759 zfs_space(vnode_t *vp, int cmd, flock64_t *bfp, int flag,
4760 offset_t offset, cred_t *cr, caller_context_t *ct)
4762 znode_t *zp = VTOZ(vp);
4763 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
4770 if (cmd != F_FREESP) {
4775 if ((error = convoff(vp, bfp, 0, offset))) {
4780 if (bfp->l_len < 0) {
4786 len = bfp->l_len; /* 0 means from off to end of file */
4788 error = zfs_freesp(zp, off, len, flag, TRUE);
4793 EXPORT_SYMBOL(zfs_space);
4797 zfs_fid(vnode_t *vp, fid_t *fidp, caller_context_t *ct)
4799 znode_t *zp = VTOZ(vp);
4800 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
4803 uint64_t object = zp->z_id;
4810 if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_GEN(zfsvfs),
4811 &gen64, sizeof (uint64_t))) != 0) {
4816 gen = (uint32_t)gen64;
4818 size = (zfsvfs->z_parent != zfsvfs) ? LONG_FID_LEN : SHORT_FID_LEN;
4819 if (fidp->fid_len < size) {
4820 fidp->fid_len = size;
4825 zfid = (zfid_short_t *)fidp;
4827 zfid->zf_len = size;
4829 for (i = 0; i < sizeof (zfid->zf_object); i++)
4830 zfid->zf_object[i] = (uint8_t)(object >> (8 * i));
4832 /* Must have a non-zero generation number to distinguish from .zfs */
4835 for (i = 0; i < sizeof (zfid->zf_gen); i++)
4836 zfid->zf_gen[i] = (uint8_t)(gen >> (8 * i));
4838 if (size == LONG_FID_LEN) {
4839 uint64_t objsetid = dmu_objset_id(zfsvfs->z_os);
4842 zlfid = (zfid_long_t *)fidp;
4844 for (i = 0; i < sizeof (zlfid->zf_setid); i++)
4845 zlfid->zf_setid[i] = (uint8_t)(objsetid >> (8 * i));
4847 /* XXX - this should be the generation number for the objset */
4848 for (i = 0; i < sizeof (zlfid->zf_setgen); i++)
4849 zlfid->zf_setgen[i] = 0;
4855 EXPORT_SYMBOL(zfs_fid);
4858 zfs_pathconf(vnode_t *vp, int cmd, ulong_t *valp, cred_t *cr,
4859 caller_context_t *ct)
4871 case _PC_FILESIZEBITS:
4875 case _PC_XATTR_EXISTS:
4877 zfsvfs = zp->z_zfsvfs;
4881 error = zfs_dirent_lock(&dl, zp, "", &xzp,
4882 ZXATTR | ZEXISTS | ZSHARED, NULL, NULL);
4884 zfs_dirent_unlock(dl);
4885 if (!zfs_dirempty(xzp))
4888 } else if (error == ENOENT) {
4890 * If there aren't extended attributes, it's the
4891 * same as having zero of them.
4898 case _PC_SATTR_ENABLED:
4899 case _PC_SATTR_EXISTS:
4900 *valp = vfs_has_feature(vp->v_vfsp, VFSFT_SYSATTR_VIEWS) &&
4901 (vp->v_type == VREG || vp->v_type == VDIR);
4904 case _PC_ACCESS_FILTERING:
4905 *valp = vfs_has_feature(vp->v_vfsp, VFSFT_ACCESS_FILTER) &&
4909 case _PC_ACL_ENABLED:
4910 *valp = _ACL_ACE_ENABLED;
4913 case _PC_MIN_HOLE_SIZE:
4914 *valp = (ulong_t)SPA_MINBLOCKSIZE;
4917 case _PC_TIMESTAMP_RESOLUTION:
4918 /* nanosecond timestamp resolution */
4923 return (fs_pathconf(vp, cmd, valp, cr, ct));
4929 zfs_getsecattr(vnode_t *vp, vsecattr_t *vsecp, int flag, cred_t *cr,
4930 caller_context_t *ct)
4932 znode_t *zp = VTOZ(vp);
4933 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
4935 boolean_t skipaclchk = (flag & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE;
4939 error = zfs_getacl(zp, vsecp, skipaclchk, cr);
4944 EXPORT_SYMBOL(zfs_getsecattr);
4948 zfs_setsecattr(vnode_t *vp, vsecattr_t *vsecp, int flag, cred_t *cr,
4949 caller_context_t *ct)
4951 znode_t *zp = VTOZ(vp);
4952 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
4954 boolean_t skipaclchk = (flag & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE;
4955 zilog_t *zilog = zfsvfs->z_log;
4960 error = zfs_setacl(zp, vsecp, skipaclchk, cr);
4962 if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS)
4963 zil_commit(zilog, 0);
4968 EXPORT_SYMBOL(zfs_setsecattr);
4971 * Tunable, both must be a power of 2.
4973 * zcr_blksz_min: the smallest read we may consider to loan out an arcbuf
4974 * zcr_blksz_max: if set to less than the file block size, allow loaning out of
4975 * an arcbuf for a partial block read
4977 int zcr_blksz_min = (1 << 10); /* 1K */
4978 int zcr_blksz_max = (1 << 17); /* 128K */
4982 zfs_reqzcbuf(vnode_t *vp, enum uio_rw ioflag, xuio_t *xuio, cred_t *cr,
4983 caller_context_t *ct)
4985 znode_t *zp = VTOZ(vp);
4986 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
4987 int max_blksz = zfsvfs->z_max_blksz;
4988 uio_t *uio = &xuio->xu_uio;
4989 ssize_t size = uio->uio_resid;
4990 offset_t offset = uio->uio_loffset;
4995 int preamble, postamble;
4997 if (xuio->xu_type != UIOTYPE_ZEROCOPY)
5005 * Loan out an arc_buf for write if write size is bigger than
5006 * max_blksz, and the file's block size is also max_blksz.
5009 if (size < blksz || zp->z_blksz != blksz) {
5014 * Caller requests buffers for write before knowing where the
5015 * write offset might be (e.g. NFS TCP write).
5020 preamble = P2PHASE(offset, blksz);
5022 preamble = blksz - preamble;
5027 postamble = P2PHASE(size, blksz);
5030 fullblk = size / blksz;
5031 (void) dmu_xuio_init(xuio,
5032 (preamble != 0) + fullblk + (postamble != 0));
5033 DTRACE_PROBE3(zfs_reqzcbuf_align, int, preamble,
5034 int, postamble, int,
5035 (preamble != 0) + fullblk + (postamble != 0));
5038 * Have to fix iov base/len for partial buffers. They
5039 * currently represent full arc_buf's.
5042 /* data begins in the middle of the arc_buf */
5043 abuf = dmu_request_arcbuf(sa_get_db(zp->z_sa_hdl),
5046 (void) dmu_xuio_add(xuio, abuf,
5047 blksz - preamble, preamble);
5050 for (i = 0; i < fullblk; i++) {
5051 abuf = dmu_request_arcbuf(sa_get_db(zp->z_sa_hdl),
5054 (void) dmu_xuio_add(xuio, abuf, 0, blksz);
5058 /* data ends in the middle of the arc_buf */
5059 abuf = dmu_request_arcbuf(sa_get_db(zp->z_sa_hdl),
5062 (void) dmu_xuio_add(xuio, abuf, 0, postamble);
5067 * Loan out an arc_buf for read if the read size is larger than
5068 * the current file block size. Block alignment is not
5069 * considered. Partial arc_buf will be loaned out for read.
5071 blksz = zp->z_blksz;
5072 if (blksz < zcr_blksz_min)
5073 blksz = zcr_blksz_min;
5074 if (blksz > zcr_blksz_max)
5075 blksz = zcr_blksz_max;
5076 /* avoid potential complexity of dealing with it */
5077 if (blksz > max_blksz) {
5082 maxsize = zp->z_size - uio->uio_loffset;
5086 if (size < blksz || vn_has_cached_data(vp)) {
5096 uio->uio_extflg = UIO_XUIO;
5097 XUIO_XUZC_RW(xuio) = ioflag;
5104 zfs_retzcbuf(vnode_t *vp, xuio_t *xuio, cred_t *cr, caller_context_t *ct)
5108 int ioflag = XUIO_XUZC_RW(xuio);
5110 ASSERT(xuio->xu_type == UIOTYPE_ZEROCOPY);
5112 i = dmu_xuio_cnt(xuio);
5114 abuf = dmu_xuio_arcbuf(xuio, i);
5116 * if abuf == NULL, it must be a write buffer
5117 * that has been returned in zfs_write().
5120 dmu_return_arcbuf(abuf);
5121 ASSERT(abuf || ioflag == UIO_WRITE);
5124 dmu_xuio_fini(xuio);
5129 * Predeclare these here so that the compiler assumes that
5130 * this is an "old style" function declaration that does
5131 * not include arguments => we won't get type mismatch errors
5132 * in the initializations that follow.
5134 static int zfs_inval();
5135 static int zfs_isdir();
5149 * Directory vnode operations template
5151 vnodeops_t *zfs_dvnodeops;
5152 const fs_operation_def_t zfs_dvnodeops_template[] = {
5153 VOPNAME_OPEN, { .vop_open = zfs_open },
5154 VOPNAME_CLOSE, { .vop_close = zfs_close },
5155 VOPNAME_READ, { .error = zfs_isdir },
5156 VOPNAME_WRITE, { .error = zfs_isdir },
5157 VOPNAME_IOCTL, { .vop_ioctl = zfs_ioctl },
5158 VOPNAME_GETATTR, { .vop_getattr = zfs_getattr },
5159 VOPNAME_SETATTR, { .vop_setattr = zfs_setattr },
5160 VOPNAME_ACCESS, { .vop_access = zfs_access },
5161 VOPNAME_LOOKUP, { .vop_lookup = zfs_lookup },
5162 VOPNAME_CREATE, { .vop_create = zfs_create },
5163 VOPNAME_REMOVE, { .vop_remove = zfs_remove },
5164 VOPNAME_LINK, { .vop_link = zfs_link },
5165 VOPNAME_RENAME, { .vop_rename = zfs_rename },
5166 VOPNAME_MKDIR, { .vop_mkdir = zfs_mkdir },
5167 VOPNAME_RMDIR, { .vop_rmdir = zfs_rmdir },
5168 VOPNAME_READDIR, { .vop_readdir = zfs_readdir },
5169 VOPNAME_SYMLINK, { .vop_symlink = zfs_symlink },
5170 VOPNAME_FSYNC, { .vop_fsync = zfs_fsync },
5171 VOPNAME_INACTIVE, { .vop_inactive = zfs_inactive },
5172 VOPNAME_FID, { .vop_fid = zfs_fid },
5173 VOPNAME_SEEK, { .vop_seek = zfs_seek },
5174 VOPNAME_PATHCONF, { .vop_pathconf = zfs_pathconf },
5175 VOPNAME_GETSECATTR, { .vop_getsecattr = zfs_getsecattr },
5176 VOPNAME_SETSECATTR, { .vop_setsecattr = zfs_setsecattr },
5177 VOPNAME_VNEVENT, { .vop_vnevent = fs_vnevent_support },
5182 * Regular file vnode operations template
5184 vnodeops_t *zfs_fvnodeops;
5185 const fs_operation_def_t zfs_fvnodeops_template[] = {
5186 VOPNAME_OPEN, { .vop_open = zfs_open },
5187 VOPNAME_CLOSE, { .vop_close = zfs_close },
5188 VOPNAME_READ, { .vop_read = zfs_read },
5189 VOPNAME_WRITE, { .vop_write = zfs_write },
5190 VOPNAME_IOCTL, { .vop_ioctl = zfs_ioctl },
5191 VOPNAME_GETATTR, { .vop_getattr = zfs_getattr },
5192 VOPNAME_SETATTR, { .vop_setattr = zfs_setattr },
5193 VOPNAME_ACCESS, { .vop_access = zfs_access },
5194 VOPNAME_LOOKUP, { .vop_lookup = zfs_lookup },
5195 VOPNAME_RENAME, { .vop_rename = zfs_rename },
5196 VOPNAME_FSYNC, { .vop_fsync = zfs_fsync },
5197 VOPNAME_INACTIVE, { .vop_inactive = zfs_inactive },
5198 VOPNAME_FID, { .vop_fid = zfs_fid },
5199 VOPNAME_SEEK, { .vop_seek = zfs_seek },
5200 VOPNAME_FRLOCK, { .vop_frlock = zfs_frlock },
5201 VOPNAME_SPACE, { .vop_space = zfs_space },
5202 VOPNAME_GETPAGE, { .vop_getpage = zfs_getpage },
5203 VOPNAME_PUTPAGE, { .vop_putpage = zfs_putpage },
5204 VOPNAME_MAP, { .vop_map = zfs_map },
5205 VOPNAME_ADDMAP, { .vop_addmap = zfs_addmap },
5206 VOPNAME_DELMAP, { .vop_delmap = zfs_delmap },
5207 VOPNAME_PATHCONF, { .vop_pathconf = zfs_pathconf },
5208 VOPNAME_GETSECATTR, { .vop_getsecattr = zfs_getsecattr },
5209 VOPNAME_SETSECATTR, { .vop_setsecattr = zfs_setsecattr },
5210 VOPNAME_VNEVENT, { .vop_vnevent = fs_vnevent_support },
5211 VOPNAME_REQZCBUF, { .vop_reqzcbuf = zfs_reqzcbuf },
5212 VOPNAME_RETZCBUF, { .vop_retzcbuf = zfs_retzcbuf },
5217 * Symbolic link vnode operations template
5219 vnodeops_t *zfs_symvnodeops;
5220 const fs_operation_def_t zfs_symvnodeops_template[] = {
5221 VOPNAME_GETATTR, { .vop_getattr = zfs_getattr },
5222 VOPNAME_SETATTR, { .vop_setattr = zfs_setattr },
5223 VOPNAME_ACCESS, { .vop_access = zfs_access },
5224 VOPNAME_RENAME, { .vop_rename = zfs_rename },
5225 VOPNAME_READLINK, { .vop_readlink = zfs_readlink },
5226 VOPNAME_INACTIVE, { .vop_inactive = zfs_inactive },
5227 VOPNAME_FID, { .vop_fid = zfs_fid },
5228 VOPNAME_PATHCONF, { .vop_pathconf = zfs_pathconf },
5229 VOPNAME_VNEVENT, { .vop_vnevent = fs_vnevent_support },
5234 * special share hidden files vnode operations template
5236 vnodeops_t *zfs_sharevnodeops;
5237 const fs_operation_def_t zfs_sharevnodeops_template[] = {
5238 VOPNAME_GETATTR, { .vop_getattr = zfs_getattr },
5239 VOPNAME_ACCESS, { .vop_access = zfs_access },
5240 VOPNAME_INACTIVE, { .vop_inactive = zfs_inactive },
5241 VOPNAME_FID, { .vop_fid = zfs_fid },
5242 VOPNAME_PATHCONF, { .vop_pathconf = zfs_pathconf },
5243 VOPNAME_GETSECATTR, { .vop_getsecattr = zfs_getsecattr },
5244 VOPNAME_SETSECATTR, { .vop_setsecattr = zfs_setsecattr },
5245 VOPNAME_VNEVENT, { .vop_vnevent = fs_vnevent_support },
5250 * Extended attribute directory vnode operations template
5251 * This template is identical to the directory vnodes
5252 * operation template except for restricted operations:
5255 * Note that there are other restrictions embedded in:
5256 * zfs_create() - restrict type to VREG
5257 * zfs_link() - no links into/out of attribute space
5258 * zfs_rename() - no moves into/out of attribute space
5260 vnodeops_t *zfs_xdvnodeops;
5261 const fs_operation_def_t zfs_xdvnodeops_template[] = {
5262 VOPNAME_OPEN, { .vop_open = zfs_open },
5263 VOPNAME_CLOSE, { .vop_close = zfs_close },
5264 VOPNAME_IOCTL, { .vop_ioctl = zfs_ioctl },
5265 VOPNAME_GETATTR, { .vop_getattr = zfs_getattr },
5266 VOPNAME_SETATTR, { .vop_setattr = zfs_setattr },
5267 VOPNAME_ACCESS, { .vop_access = zfs_access },
5268 VOPNAME_LOOKUP, { .vop_lookup = zfs_lookup },
5269 VOPNAME_CREATE, { .vop_create = zfs_create },
5270 VOPNAME_REMOVE, { .vop_remove = zfs_remove },
5271 VOPNAME_LINK, { .vop_link = zfs_link },
5272 VOPNAME_RENAME, { .vop_rename = zfs_rename },
5273 VOPNAME_MKDIR, { .error = zfs_inval },
5274 VOPNAME_RMDIR, { .vop_rmdir = zfs_rmdir },
5275 VOPNAME_READDIR, { .vop_readdir = zfs_readdir },
5276 VOPNAME_SYMLINK, { .error = zfs_inval },
5277 VOPNAME_FSYNC, { .vop_fsync = zfs_fsync },
5278 VOPNAME_INACTIVE, { .vop_inactive = zfs_inactive },
5279 VOPNAME_FID, { .vop_fid = zfs_fid },
5280 VOPNAME_SEEK, { .vop_seek = zfs_seek },
5281 VOPNAME_PATHCONF, { .vop_pathconf = zfs_pathconf },
5282 VOPNAME_GETSECATTR, { .vop_getsecattr = zfs_getsecattr },
5283 VOPNAME_SETSECATTR, { .vop_setsecattr = zfs_setsecattr },
5284 VOPNAME_VNEVENT, { .vop_vnevent = fs_vnevent_support },
5289 * Error vnode operations template
5291 vnodeops_t *zfs_evnodeops;
5292 const fs_operation_def_t zfs_evnodeops_template[] = {
5293 VOPNAME_INACTIVE, { .vop_inactive = zfs_inactive },
5294 VOPNAME_PATHCONF, { .vop_pathconf = zfs_pathconf },
5297 #endif /* HAVE_ZPL */