usr/src/uts/common/fs/zfs/zfs_vnops.c

/*
 * CDDL HEADER START
 *
 * The contents of this file are subject to the terms of the
 * Common Development and Distribution License (the "License").
 * You may not use this file except in compliance with the License.
 *
 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
 * or http://www.opensolaris.org/os/licensing.
 * See the License for the specific language governing permissions
 * and limitations under the License.
 *
 * When distributing Covered Code, include this CDDL HEADER in each
 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
 * If applicable, add the following below this CDDL HEADER, with the
 * fields enclosed by brackets "[]" replaced with your own identifying
 * information: Portions Copyright [yyyy] [name of copyright owner]
 *
 * CDDL HEADER END
 */

/*
 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
 * Copyright (c) 2012, 2017 by Delphix. All rights reserved.
 * Copyright (c) 2014 Integros [integros.com]
 * Copyright 2015 Joyent, Inc.
 * Copyright 2017 Nexenta Systems, Inc.
 */

/* Portions Copyright 2007 Jeremy Teo */
/* Portions Copyright 2010 Robert Milkowski */

#include <sys/types.h>
#include <sys/param.h>
#include <sys/time.h>
#include <sys/systm.h>
#include <sys/sysmacros.h>
#include <sys/resource.h>
#include <sys/vfs.h>
#include <sys/vfs_opreg.h>
#include <sys/vnode.h>
#include <sys/file.h>
#include <sys/stat.h>
#include <sys/kmem.h>
#include <sys/taskq.h>
#include <sys/uio.h>
#include <sys/vmsystm.h>
#include <sys/atomic.h>
#include <sys/vm.h>
#include <vm/seg_vn.h>
#include <vm/pvn.h>
#include <vm/as.h>
#include <vm/kpm.h>
#include <vm/seg_kpm.h>
#include <sys/mman.h>
#include <sys/pathname.h>
#include <sys/cmn_err.h>
#include <sys/errno.h>
#include <sys/unistd.h>
#include <sys/zfs_dir.h>
#include <sys/zfs_acl.h>
#include <sys/zfs_ioctl.h>
#include <sys/fs/zfs.h>
#include <sys/dmu.h>
#include <sys/dmu_objset.h>
#include <sys/spa.h>
#include <sys/txg.h>
#include <sys/dbuf.h>
#include <sys/zap.h>
#include <sys/sa.h>
#include <sys/dirent.h>
#include <sys/policy.h>
#include <sys/sunddi.h>
#include <sys/filio.h>
#include <sys/sid.h>
#include "fs/fs_subr.h"
#include <sys/zfs_ctldir.h>
#include <sys/zfs_fuid.h>
#include <sys/zfs_sa.h>
#include <sys/dnlc.h>
#include <sys/zfs_rlock.h>
#include <sys/extdirent.h>
#include <sys/kidmap.h>
#include <sys/cred.h>
#include <sys/attr.h>
#include <sys/zil.h>

/*
 * Programming rules.
 *
 * Each vnode op performs some logical unit of work.  To do this, the ZPL must
 * properly lock its in-core state, create a DMU transaction, do the work,
 * record this work in the intent log (ZIL), commit the DMU transaction,
 * and wait for the intent log to commit if it is a synchronous operation.
 * Moreover, the vnode ops must work in both normal and log replay context.
 * The ordering of events is important to avoid deadlocks and references
 * to freed memory.  The example below illustrates the following Big Rules:
 *
 *  (1)	A check must be made in each zfs thread for a mounted file system.
 *	This is done avoiding races using ZFS_ENTER(zfsvfs).
 *	A ZFS_EXIT(zfsvfs) is needed before all returns.  Any znodes
 *	must be checked with ZFS_VERIFY_ZP(zp).  Both of these macros
 *	can return EIO from the calling function.
 *
 *  (2)	VN_RELE() should always be the last thing except for zil_commit()
 *	(if necessary) and ZFS_EXIT(). This is for 3 reasons:
 *	First, if it's the last reference, the vnode/znode
 *	can be freed, so the zp may point to freed memory.  Second, the last
 *	reference will call zfs_zinactive(), which may induce a lot of work --
 *	pushing cached pages (which acquires range locks) and syncing out
 *	cached atime changes.  Third, zfs_zinactive() may require a new tx,
 *	which could deadlock the system if you were already holding one.
 *	If you must call VN_RELE() within a tx then use VN_RELE_ASYNC().
 *
 *  (3)	All range locks must be grabbed before calling dmu_tx_assign(),
 *	as they can span dmu_tx_assign() calls.
 *
 *  (4) If ZPL locks are held, pass TXG_NOWAIT as the second argument to
 *      dmu_tx_assign().  This is critical because we don't want to block
 *      while holding locks.
 *
 *	If no ZPL locks are held (aside from ZFS_ENTER()), use TXG_WAIT.  This
 *	reduces lock contention and CPU usage when we must wait (note that if
 *	throughput is constrained by the storage, nearly every transaction
 *	must wait).
 *
 *      Note, in particular, that if a lock is sometimes acquired before
 *      the tx assigns, and sometimes after (e.g. z_lock), then failing
 *      to use a non-blocking assign can deadlock the system.  The scenario:
 *
 *	Thread A has grabbed a lock before calling dmu_tx_assign().
 *	Thread B is in an already-assigned tx, and blocks for this lock.
 *	Thread A calls dmu_tx_assign(TXG_WAIT) and blocks in txg_wait_open()
 *	forever, because the previous txg can't quiesce until B's tx commits.
 *
 *	If dmu_tx_assign() returns ERESTART and zfsvfs->z_assign is TXG_NOWAIT,
 *	then drop all locks, call dmu_tx_wait(), and try again.  On subsequent
 *	calls to dmu_tx_assign(), pass TXG_NOTHROTTLE in addition to TXG_NOWAIT,
 *	to indicate that this operation has already called dmu_tx_wait().
 *	This will ensure that we don't retry forever, waiting a short bit
 *	each time.
 *
 *  (5)	If the operation succeeded, generate the intent log entry for it
 *	before dropping locks.  This ensures that the ordering of events
 *	in the intent log matches the order in which they actually occurred.
 *	During ZIL replay the zfs_log_* functions will update the sequence
 *	number to indicate the zil transaction has replayed.
 *
 *  (6)	At the end of each vnode op, the DMU tx must always commit,
 *	regardless of whether there were any errors.
 *
 *  (7)	After dropping all locks, invoke zil_commit(zilog, foid)
 *	to ensure that synchronous semantics are provided when necessary.
 *
 * In general, this is how things should be ordered in each vnode op:
 *
 *	ZFS_ENTER(zfsvfs);		// exit if unmounted
 * top:
 *	zfs_dirent_lock(&dl, ...)	// lock directory entry (may VN_HOLD())
 *	rw_enter(...);			// grab any other locks you need
 *	tx = dmu_tx_create(...);	// get DMU tx
 *	dmu_tx_hold_*();		// hold each object you might modify
 *	error = dmu_tx_assign(tx, (waited ? TXG_NOTHROTTLE : 0) | TXG_NOWAIT);
 *	if (error) {
 *		rw_exit(...);		// drop locks
 *		zfs_dirent_unlock(dl);	// unlock directory entry
 *		VN_RELE(...);		// release held vnodes
 *		if (error == ERESTART) {
 *			waited = B_TRUE;
 *			dmu_tx_wait(tx);
 *			dmu_tx_abort(tx);
 *			goto top;
 *		}
 *		dmu_tx_abort(tx);	// abort DMU tx
 *		ZFS_EXIT(zfsvfs);	// finished in zfs
 *		return (error);		// really out of space
 *	}
 *	error = do_real_work();		// do whatever this VOP does
 *	if (error == 0)
 *		zfs_log_*(...);		// on success, make ZIL entry
 *	dmu_tx_commit(tx);		// commit DMU tx -- error or not
 *	rw_exit(...);			// drop locks
 *	zfs_dirent_unlock(dl);		// unlock directory entry
 *	VN_RELE(...);			// release held vnodes
 *	zil_commit(zilog, foid);	// synchronous when necessary
 *	ZFS_EXIT(zfsvfs);		// finished in zfs
 *	return (error);			// done, report error
 */

/* ARGSUSED */
static int
zfs_open(vnode_t **vpp, int flag, cred_t *cr, caller_context_t *ct)
{
	znode_t	*zp = VTOZ(*vpp);
	zfsvfs_t *zfsvfs = zp->z_zfsvfs;

	ZFS_ENTER(zfsvfs);
	ZFS_VERIFY_ZP(zp);

	if ((flag & FWRITE) && (zp->z_pflags & ZFS_APPENDONLY) &&
	    ((flag & FAPPEND) == 0)) {
		ZFS_EXIT(zfsvfs);
		return (SET_ERROR(EPERM));
	}

	if (!zfs_has_ctldir(zp) && zp->z_zfsvfs->z_vscan &&
	    ZTOV(zp)->v_type == VREG &&
	    !(zp->z_pflags & ZFS_AV_QUARANTINED) && zp->z_size > 0) {
		if (fs_vscan(*vpp, cr, 0) != 0) {
			ZFS_EXIT(zfsvfs);
			return (SET_ERROR(EACCES));
		}
	}

	/* Keep a count of the synchronous opens in the znode */
	if (flag & (FSYNC | FDSYNC))
		atomic_inc_32(&zp->z_sync_cnt);

	ZFS_EXIT(zfsvfs);
	return (0);
}

/* ARGSUSED */
static int
zfs_close(vnode_t *vp, int flag, int count, offset_t offset, cred_t *cr,
    caller_context_t *ct)
{
	znode_t	*zp = VTOZ(vp);
	zfsvfs_t *zfsvfs = zp->z_zfsvfs;

	/*
	 * Clean up any locks held by this process on the vp.
	 */
	cleanlocks(vp, ddi_get_pid(), 0);
	cleanshares(vp, ddi_get_pid());

	ZFS_ENTER(zfsvfs);
	ZFS_VERIFY_ZP(zp);

	/* Decrement the synchronous opens in the znode */
	if ((flag & (FSYNC | FDSYNC)) && (count == 1))
		atomic_dec_32(&zp->z_sync_cnt);

	if (!zfs_has_ctldir(zp) && zp->z_zfsvfs->z_vscan &&
	    ZTOV(zp)->v_type == VREG &&
	    !(zp->z_pflags & ZFS_AV_QUARANTINED) && zp->z_size > 0)
		VERIFY(fs_vscan(vp, cr, 1) == 0);

	ZFS_EXIT(zfsvfs);
	return (0);
}

/*
 * Lseek support for finding holes (cmd == _FIO_SEEK_HOLE) and
 * data (cmd == _FIO_SEEK_DATA). "off" is an in/out parameter.
 */
static int
zfs_holey(vnode_t *vp, int cmd, offset_t *off)
{
	znode_t	*zp = VTOZ(vp);
	uint64_t noff = (uint64_t)*off; /* new offset */
	uint64_t file_sz;
	int error;
	boolean_t hole;

	file_sz = zp->z_size;
	if (noff >= file_sz)  {
		return (SET_ERROR(ENXIO));
	}

	if (cmd == _FIO_SEEK_HOLE)
		hole = B_TRUE;
	else
		hole = B_FALSE;

	error = dmu_offset_next(zp->z_zfsvfs->z_os, zp->z_id, hole, &noff);

	if (error == ESRCH)
		return (SET_ERROR(ENXIO));

	/*
	 * We could find a hole that begins after the logical end-of-file,
	 * because dmu_offset_next() only works on whole blocks.  If the
	 * EOF falls mid-block, then indicate that the "virtual hole"
	 * at the end of the file begins at the logical EOF, rather than
	 * at the end of the last block.
	 */
	if (noff > file_sz) {
		ASSERT(hole);
		noff = file_sz;
	}

	if (noff < *off)
		return (error);
	*off = noff;
	return (error);
}

/* ARGSUSED */
static int
zfs_ioctl(vnode_t *vp, int com, intptr_t data, int flag, cred_t *cred,
    int *rvalp, caller_context_t *ct)
{
	offset_t off;
	offset_t ndata;
	dmu_object_info_t doi;
	int error;
	zfsvfs_t *zfsvfs;
	znode_t *zp;

	switch (com) {
	case _FIOFFS:
	{
		return (zfs_sync(vp->v_vfsp, 0, cred));

		/*
		 * The following two ioctls are used by bfu.  Faking out,
		 * necessary to avoid bfu errors.
		 */
	}
	case _FIOGDIO:
	case _FIOSDIO:
	{
		return (0);
	}

	case _FIO_SEEK_DATA:
	case _FIO_SEEK_HOLE:
	{
		if (ddi_copyin((void *)data, &off, sizeof (off), flag))
			return (SET_ERROR(EFAULT));

		zp = VTOZ(vp);
		zfsvfs = zp->z_zfsvfs;
		ZFS_ENTER(zfsvfs);
		ZFS_VERIFY_ZP(zp);

		/* offset parameter is in/out */
		error = zfs_holey(vp, com, &off);
		ZFS_EXIT(zfsvfs);
		if (error)
			return (error);
		if (ddi_copyout(&off, (void *)data, sizeof (off), flag))
			return (SET_ERROR(EFAULT));
		return (0);
	}
	case _FIO_COUNT_FILLED:
	{
		/*
		 * _FIO_COUNT_FILLED adds a new ioctl command which
		 * exposes the number of filled blocks in a
		 * ZFS object.
		 */
		zp = VTOZ(vp);
		zfsvfs = zp->z_zfsvfs;
		ZFS_ENTER(zfsvfs);
		ZFS_VERIFY_ZP(zp);

		/*
		 * Wait for all dirty blocks for this object
		 * to get synced out to disk, and the DMU info
		 * updated.
		 */
		error = dmu_object_wait_synced(zfsvfs->z_os, zp->z_id);
		if (error) {
			ZFS_EXIT(zfsvfs);
			return (error);
		}

		/*
		 * Retrieve fill count from DMU object.
		 */
		error = dmu_object_info(zfsvfs->z_os, zp->z_id, &doi);
		if (error) {
			ZFS_EXIT(zfsvfs);
			return (error);
		}

		ndata = doi.doi_fill_count;

		ZFS_EXIT(zfsvfs);
		if (ddi_copyout(&ndata, (void *)data, sizeof (ndata), flag))
			return (SET_ERROR(EFAULT));
		return (0);
	}
	}
	return (SET_ERROR(ENOTTY));
}

/*
 * Utility functions to map and unmap a single physical page.  These
 * are used to manage the mappable copies of ZFS file data, and therefore
 * do not update ref/mod bits.
 */
caddr_t
zfs_map_page(page_t *pp, enum seg_rw rw)
{
	if (kpm_enable)
		return (hat_kpm_mapin(pp, 0));
	ASSERT(rw == S_READ || rw == S_WRITE);
	return (ppmapin(pp, PROT_READ | ((rw == S_WRITE) ? PROT_WRITE : 0),
	    (caddr_t)-1));
}

void
zfs_unmap_page(page_t *pp, caddr_t addr)
{
	if (kpm_enable) {
		hat_kpm_mapout(pp, 0, addr);
	} else {
		ppmapout(addr);
	}
}

/*
 * When a file is memory mapped, we must keep the IO data synchronized
 * between the DMU cache and the memory mapped pages.  What this means:
 *
 * On Write:	If we find a memory mapped page, we write to *both*
 *		the page and the dmu buffer.
 */
static void
update_pages(vnode_t *vp, int64_t start, int len, objset_t *os, uint64_t oid)
{
	int64_t	off;

	off = start & PAGEOFFSET;
	for (start &= PAGEMASK; len > 0; start += PAGESIZE) {
		page_t *pp;
		uint64_t nbytes = MIN(PAGESIZE - off, len);

		if (pp = page_lookup(vp, start, SE_SHARED)) {
			caddr_t va;

			va = zfs_map_page(pp, S_WRITE);
			(void) dmu_read(os, oid, start+off, nbytes, va+off,
			    DMU_READ_PREFETCH);
			zfs_unmap_page(pp, va);
			page_unlock(pp);
		}
		len -= nbytes;
		off = 0;
	}
}

/*
 * When a file is memory mapped, we must keep the IO data synchronized
 * between the DMU cache and the memory mapped pages.  What this means:
 *
 * On Read:	We "read" preferentially from memory mapped pages,
 *		else we default from the dmu buffer.
 *
 * NOTE: We will always "break up" the IO into PAGESIZE uiomoves when
 *	 the file is memory mapped.
 */
static int
mappedread(vnode_t *vp, int nbytes, uio_t *uio)
{
	znode_t *zp = VTOZ(vp);
	int64_t	start, off;
	int len = nbytes;
	int error = 0;

	start = uio->uio_loffset;
	off = start & PAGEOFFSET;
	for (start &= PAGEMASK; len > 0; start += PAGESIZE) {
		page_t *pp;
		uint64_t bytes = MIN(PAGESIZE - off, len);

		if (pp = page_lookup(vp, start, SE_SHARED)) {
			caddr_t va;

			va = zfs_map_page(pp, S_READ);
			error = uiomove(va + off, bytes, UIO_READ, uio);
			zfs_unmap_page(pp, va);
			page_unlock(pp);
		} else {
			error = dmu_read_uio_dbuf(sa_get_db(zp->z_sa_hdl),
			    uio, bytes);
		}
		len -= bytes;
		off = 0;
		if (error)
			break;
	}
	return (error);
}

offset_t zfs_read_chunk_size = 1024 * 1024; /* Tunable */

/*
 * Read bytes from specified file into supplied buffer.
 *
 *	IN:	vp	- vnode of file to be read from.
 *		uio	- structure supplying read location, range info,
 *			  and return buffer.
 *		ioflag	- SYNC flags; used to provide FRSYNC semantics.
 *		cr	- credentials of caller.
 *		ct	- caller context
 *
 *	OUT:	uio	- updated offset and range, buffer filled.
 *
 *	RETURN:	0 on success, error code on failure.
 *
 * Side Effects:
 *	vp - atime updated if byte count > 0
 */
/* ARGSUSED */
static int
zfs_read(vnode_t *vp, uio_t *uio, int ioflag, cred_t *cr, caller_context_t *ct)
{
	znode_t		*zp = VTOZ(vp);
	zfsvfs_t	*zfsvfs = zp->z_zfsvfs;
	ssize_t		n, nbytes;
	int		error = 0;
	rl_t		*rl;
	xuio_t		*xuio = NULL;

	ZFS_ENTER(zfsvfs);
	ZFS_VERIFY_ZP(zp);

	if (zp->z_pflags & ZFS_AV_QUARANTINED) {
		ZFS_EXIT(zfsvfs);
		return (SET_ERROR(EACCES));
	}

	/*
	 * Validate file offset
	 */
	if (uio->uio_loffset < (offset_t)0) {
		ZFS_EXIT(zfsvfs);
		return (SET_ERROR(EINVAL));
	}

	/*
	 * Fasttrack empty reads
	 */
	if (uio->uio_resid == 0) {
		ZFS_EXIT(zfsvfs);
		return (0);
	}

	/*
	 * Check for mandatory locks
	 */
	if (MANDMODE(zp->z_mode)) {
		if (error = chklock(vp, FREAD,
		    uio->uio_loffset, uio->uio_resid, uio->uio_fmode, ct)) {
			ZFS_EXIT(zfsvfs);
			return (error);
		}
	}

	/*
	 * If we're in FRSYNC mode, sync out this znode before reading it.
	 */
	if (ioflag & FRSYNC || zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS)
		zil_commit(zfsvfs->z_log, zp->z_id);

	/*
	 * Lock the range against changes.
	 */
	rl = zfs_range_lock(zp, uio->uio_loffset, uio->uio_resid, RL_READER);

	/*
	 * If we are reading past end-of-file we can skip
	 * to the end; but we might still need to set atime.
	 */
	if (uio->uio_loffset >= zp->z_size) {
		error = 0;
		goto out;
	}

	ASSERT(uio->uio_loffset < zp->z_size);
	n = MIN(uio->uio_resid, zp->z_size - uio->uio_loffset);

	if ((uio->uio_extflg == UIO_XUIO) &&
	    (((xuio_t *)uio)->xu_type == UIOTYPE_ZEROCOPY)) {
		int nblk;
		int blksz = zp->z_blksz;
		uint64_t offset = uio->uio_loffset;

		xuio = (xuio_t *)uio;
		if ((ISP2(blksz))) {
			nblk = (P2ROUNDUP(offset + n, blksz) - P2ALIGN(offset,
			    blksz)) / blksz;
		} else {
			ASSERT(offset + n <= blksz);
			nblk = 1;
		}
		(void) dmu_xuio_init(xuio, nblk);

		if (vn_has_cached_data(vp)) {
			/*
			 * For simplicity, we always allocate a full buffer
			 * even if we only expect to read a portion of a block.
			 */
			while (--nblk >= 0) {
				(void) dmu_xuio_add(xuio,
				    dmu_request_arcbuf(sa_get_db(zp->z_sa_hdl),
				    blksz), 0, blksz);
			}
		}
	}

	while (n > 0) {
		nbytes = MIN(n, zfs_read_chunk_size -
		    P2PHASE(uio->uio_loffset, zfs_read_chunk_size));

		if (vn_has_cached_data(vp)) {
			error = mappedread(vp, nbytes, uio);
		} else {
			error = dmu_read_uio_dbuf(sa_get_db(zp->z_sa_hdl),
			    uio, nbytes);
		}
		if (error) {
			/* convert checksum errors into IO errors */
			if (error == ECKSUM)
				error = SET_ERROR(EIO);
			break;
		}

		n -= nbytes;
	}
out:
	zfs_range_unlock(rl);

	ZFS_ACCESSTIME_STAMP(zfsvfs, zp);
	ZFS_EXIT(zfsvfs);
	return (error);
}

/*
 * Write the bytes to a file.
 *
 *	IN:	vp	- vnode of file to be written to.
 *		uio	- structure supplying write location, range info,
 *			  and data buffer.
 *		ioflag	- FAPPEND, FSYNC, and/or FDSYNC.  FAPPEND is
 *			  set if in append mode.
 *		cr	- credentials of caller.
 *		ct	- caller context (NFS/CIFS fem monitor only)
 *
 *	OUT:	uio	- updated offset and range.
 *
 *	RETURN:	0 on success, error code on failure.
 *
 * Timestamps:
 *	vp - ctime|mtime updated if byte count > 0
 */

/* ARGSUSED */
static int
zfs_write(vnode_t *vp, uio_t *uio, int ioflag, cred_t *cr, caller_context_t *ct)
{
	znode_t		*zp = VTOZ(vp);
	rlim64_t	limit = uio->uio_llimit;
	ssize_t		start_resid = uio->uio_resid;
	ssize_t		tx_bytes;
	uint64_t	end_size;
	dmu_tx_t	*tx;
	zfsvfs_t	*zfsvfs = zp->z_zfsvfs;
	zilog_t		*zilog;
	offset_t	woff;
	ssize_t		n, nbytes;
	rl_t		*rl;
	int		max_blksz = zfsvfs->z_max_blksz;
	int		error = 0;
	arc_buf_t	*abuf;
	iovec_t		*aiov = NULL;
	xuio_t		*xuio = NULL;
	int		i_iov = 0;
	int		iovcnt = uio->uio_iovcnt;
	iovec_t		*iovp = uio->uio_iov;
	int		write_eof;
	int		count = 0;
	sa_bulk_attr_t	bulk[4];
	uint64_t	mtime[2], ctime[2];

	/*
	 * Fasttrack empty write
	 */
	n = start_resid;
	if (n == 0)
		return (0);

	if (limit == RLIM64_INFINITY || limit > MAXOFFSET_T)
		limit = MAXOFFSET_T;

	ZFS_ENTER(zfsvfs);
	ZFS_VERIFY_ZP(zp);

	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL, &mtime, 16);
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL, &ctime, 16);
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs), NULL,
	    &zp->z_size, 8);
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL,
	    &zp->z_pflags, 8);

	/*
	 * In a case vp->v_vfsp != zp->z_zfsvfs->z_vfs (e.g. snapshots) our
	 * callers might not be able to detect properly that we are read-only,
	 * so check it explicitly here.
	 */
	if (zfsvfs->z_vfs->vfs_flag & VFS_RDONLY) {
		ZFS_EXIT(zfsvfs);
		return (SET_ERROR(EROFS));
	}

	/*
	 * If immutable or not appending then return EPERM.
	 * Intentionally allow ZFS_READONLY through here.
	 * See zfs_zaccess_common()
	 */
	if ((zp->z_pflags & ZFS_IMMUTABLE) ||
	    ((zp->z_pflags & ZFS_APPENDONLY) && !(ioflag & FAPPEND) &&
	    (uio->uio_loffset < zp->z_size))) {
		ZFS_EXIT(zfsvfs);
		return (SET_ERROR(EPERM));
	}

	zilog = zfsvfs->z_log;

	/*
	 * Validate file offset
	 */
	woff = ioflag & FAPPEND ? zp->z_size : uio->uio_loffset;
	if (woff < 0) {
		ZFS_EXIT(zfsvfs);
		return (SET_ERROR(EINVAL));
	}

	/*
	 * Check for mandatory locks before calling zfs_range_lock()
	 * in order to prevent a deadlock with locks set via fcntl().
	 */
	if (MANDMODE((mode_t)zp->z_mode) &&
	    (error = chklock(vp, FWRITE, woff, n, uio->uio_fmode, ct)) != 0) {
		ZFS_EXIT(zfsvfs);
		return (error);
	}

	/*
	 * Pre-fault the pages to ensure slow (eg NFS) pages
	 * don't hold up txg.
	 * Skip this if uio contains loaned arc_buf.
	 */
	if ((uio->uio_extflg == UIO_XUIO) &&
	    (((xuio_t *)uio)->xu_type == UIOTYPE_ZEROCOPY))
		xuio = (xuio_t *)uio;
	else
		uio_prefaultpages(MIN(n, max_blksz), uio);

	/*
	 * If in append mode, set the io offset pointer to eof.
	 */
	if (ioflag & FAPPEND) {
		/*
		 * Obtain an appending range lock to guarantee file append
		 * semantics.  We reset the write offset once we have the lock.
		 */
		rl = zfs_range_lock(zp, 0, n, RL_APPEND);
		woff = rl->r_off;
		if (rl->r_len == UINT64_MAX) {
			/*
			 * We overlocked the file because this write will cause
			 * the file block size to increase.
			 * Note that zp_size cannot change with this lock held.
			 */
			woff = zp->z_size;
		}
		uio->uio_loffset = woff;
	} else {
		/*
		 * Note that if the file block size will change as a result of
		 * this write, then this range lock will lock the entire file
		 * so that we can re-write the block safely.
		 */
		rl = zfs_range_lock(zp, woff, n, RL_WRITER);
	}

	if (woff >= limit) {
		zfs_range_unlock(rl);
		ZFS_EXIT(zfsvfs);
		return (SET_ERROR(EFBIG));
	}

	if ((woff + n) > limit || woff > (limit - n))
		n = limit - woff;

	/* Will this write extend the file length? */
	write_eof = (woff + n > zp->z_size);

	end_size = MAX(zp->z_size, woff + n);

	/*
	 * Write the file in reasonable size chunks.  Each chunk is written
	 * in a separate transaction; this keeps the intent log records small
	 * and allows us to do more fine-grained space accounting.
	 */
	while (n > 0) {
		abuf = NULL;
		woff = uio->uio_loffset;
		if (zfs_owner_overquota(zfsvfs, zp, B_FALSE) ||
		    zfs_owner_overquota(zfsvfs, zp, B_TRUE)) {
			if (abuf != NULL)
				dmu_return_arcbuf(abuf);
			error = SET_ERROR(EDQUOT);
			break;
		}

		if (xuio && abuf == NULL) {
			ASSERT(i_iov < iovcnt);
			aiov = &iovp[i_iov];
			abuf = dmu_xuio_arcbuf(xuio, i_iov);
			dmu_xuio_clear(xuio, i_iov);
			DTRACE_PROBE3(zfs_cp_write, int, i_iov,
			    iovec_t *, aiov, arc_buf_t *, abuf);
			ASSERT((aiov->iov_base == abuf->b_data) ||
			    ((char *)aiov->iov_base - (char *)abuf->b_data +
			    aiov->iov_len == arc_buf_size(abuf)));
			i_iov++;
		} else if (abuf == NULL && n >= max_blksz &&
		    woff >= zp->z_size &&
		    P2PHASE(woff, max_blksz) == 0 &&
		    zp->z_blksz == max_blksz) {
			/*
			 * This write covers a full block.  "Borrow" a buffer
			 * from the dmu so that we can fill it before we enter
			 * a transaction.  This avoids the possibility of
			 * holding up the transaction if the data copy hangs
			 * up on a pagefault (e.g., from an NFS server mapping).
			 */
			size_t cbytes;

			abuf = dmu_request_arcbuf(sa_get_db(zp->z_sa_hdl),
			    max_blksz);
			ASSERT(abuf != NULL);
			ASSERT(arc_buf_size(abuf) == max_blksz);
			if (error = uiocopy(abuf->b_data, max_blksz,
			    UIO_WRITE, uio, &cbytes)) {
				dmu_return_arcbuf(abuf);
				break;
			}
			ASSERT(cbytes == max_blksz);
		}

		/*
		 * Start a transaction.
		 */
		tx = dmu_tx_create(zfsvfs->z_os);
		dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
		dmu_tx_hold_write(tx, zp->z_id, woff, MIN(n, max_blksz));
		zfs_sa_upgrade_txholds(tx, zp);
		error = dmu_tx_assign(tx, TXG_WAIT);
		if (error) {
			dmu_tx_abort(tx);
			if (abuf != NULL)
				dmu_return_arcbuf(abuf);
			break;
		}

		/*
		 * If zfs_range_lock() over-locked we grow the blocksize
		 * and then reduce the lock range.  This will only happen
		 * on the first iteration since zfs_range_reduce() will
		 * shrink down r_len to the appropriate size.
		 */
		if (rl->r_len == UINT64_MAX) {
			uint64_t new_blksz;

			if (zp->z_blksz > max_blksz) {
				/*
				 * File's blocksize is already larger than the
				 * "recordsize" property.  Only let it grow to
				 * the next power of 2.
				 */
				ASSERT(!ISP2(zp->z_blksz));
				new_blksz = MIN(end_size,
				    1 << highbit64(zp->z_blksz));
			} else {
				new_blksz = MIN(end_size, max_blksz);
			}
			zfs_grow_blocksize(zp, new_blksz, tx);
			zfs_range_reduce(rl, woff, n);
		}

		/*
		 * XXX - should we really limit each write to z_max_blksz?
		 * Perhaps we should use SPA_MAXBLOCKSIZE chunks?
		 */
		nbytes = MIN(n, max_blksz - P2PHASE(woff, max_blksz));

		if (abuf == NULL) {
			tx_bytes = uio->uio_resid;
			error = dmu_write_uio_dbuf(sa_get_db(zp->z_sa_hdl),
			    uio, nbytes, tx);
			tx_bytes -= uio->uio_resid;
		} else {
			tx_bytes = nbytes;
			ASSERT(xuio == NULL || tx_bytes == aiov->iov_len);
			/*
			 * If this is not a full block write, but we are
			 * extending the file past EOF and this data starts
			 * block-aligned, use assign_arcbuf().  Otherwise,
			 * write via dmu_write().
			 */
			if (tx_bytes < max_blksz && (!write_eof ||
			    aiov->iov_base != abuf->b_data)) {
				ASSERT(xuio);
				dmu_write(zfsvfs->z_os, zp->z_id, woff,
				    aiov->iov_len, aiov->iov_base, tx);
				dmu_return_arcbuf(abuf);
				xuio_stat_wbuf_copied();
			} else {
				ASSERT(xuio || tx_bytes == max_blksz);
				dmu_assign_arcbuf(sa_get_db(zp->z_sa_hdl),
				    woff, abuf, tx);
			}
			ASSERT(tx_bytes <= uio->uio_resid);
			uioskip(uio, tx_bytes);
		}
		if (tx_bytes && vn_has_cached_data(vp)) {
			update_pages(vp, woff,
			    tx_bytes, zfsvfs->z_os, zp->z_id);
		}

		/*
		 * If we made no progress, we're done.  If we made even
		 * partial progress, update the znode and ZIL accordingly.
		 */
		if (tx_bytes == 0) {
			(void) sa_update(zp->z_sa_hdl, SA_ZPL_SIZE(zfsvfs),
			    (void *)&zp->z_size, sizeof (uint64_t), tx);
			dmu_tx_commit(tx);
			ASSERT(error != 0);
			break;
		}

		/*
		 * Clear Set-UID/Set-GID bits on successful write if not
		 * privileged and at least one of the excute bits is set.
		 *
		 * It would be nice to to this after all writes have
		 * been done, but that would still expose the ISUID/ISGID
		 * to another app after the partial write is committed.
		 *
		 * Note: we don't call zfs_fuid_map_id() here because
		 * user 0 is not an ephemeral uid.
		 */
		mutex_enter(&zp->z_acl_lock);
		if ((zp->z_mode & (S_IXUSR | (S_IXUSR >> 3) |
		    (S_IXUSR >> 6))) != 0 &&
		    (zp->z_mode & (S_ISUID | S_ISGID)) != 0 &&
		    secpolicy_vnode_setid_retain(cr,
		    (zp->z_mode & S_ISUID) != 0 && zp->z_uid == 0) != 0) {
			uint64_t newmode;
			zp->z_mode &= ~(S_ISUID | S_ISGID);
			newmode = zp->z_mode;
			(void) sa_update(zp->z_sa_hdl, SA_ZPL_MODE(zfsvfs),
			    (void *)&newmode, sizeof (uint64_t), tx);
		}
		mutex_exit(&zp->z_acl_lock);

		zfs_tstamp_update_setup(zp, CONTENT_MODIFIED, mtime, ctime,
		    B_TRUE);

		/*
		 * Update the file size (zp_size) if it has changed;
		 * account for possible concurrent updates.
		 */
		while ((end_size = zp->z_size) < uio->uio_loffset) {
			(void) atomic_cas_64(&zp->z_size, end_size,
			    uio->uio_loffset);
			ASSERT(error == 0);
		}
		/*
		 * If we are replaying and eof is non zero then force
		 * the file size to the specified eof. Note, there's no
		 * concurrency during replay.
		 */
		if (zfsvfs->z_replay && zfsvfs->z_replay_eof != 0)
			zp->z_size = zfsvfs->z_replay_eof;

		error = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx);

		zfs_log_write(zilog, tx, TX_WRITE, zp, woff, tx_bytes, ioflag);
		dmu_tx_commit(tx);

		if (error != 0)
			break;
		ASSERT(tx_bytes == nbytes);
		n -= nbytes;

		if (!xuio && n > 0)
			uio_prefaultpages(MIN(n, max_blksz), uio);
	}

	zfs_range_unlock(rl);

	/*
	 * If we're in replay mode, or we made no progress, return error.
	 * Otherwise, it's at least a partial write, so it's successful.
	 */
	if (zfsvfs->z_replay || uio->uio_resid == start_resid) {
		ZFS_EXIT(zfsvfs);
		return (error);
	}

	if (ioflag & (FSYNC | FDSYNC) ||
	    zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS)
		zil_commit(zilog, zp->z_id);

	ZFS_EXIT(zfsvfs);
	return (0);
}

/* ARGSUSED */
void
zfs_get_done(zgd_t *zgd, int error)
{
	znode_t *zp = zgd->zgd_private;
	objset_t *os = zp->z_zfsvfs->z_os;

	if (zgd->zgd_db)
		dmu_buf_rele(zgd->zgd_db, zgd);

	zfs_range_unlock(zgd->zgd_rl);

	/*
	 * Release the vnode asynchronously as we currently have the
	 * txg stopped from syncing.
	 */
	VN_RELE_ASYNC(ZTOV(zp), dsl_pool_vnrele_taskq(dmu_objset_pool(os)));

	kmem_free(zgd, sizeof (zgd_t));
}

#ifdef DEBUG
static int zil_fault_io = 0;
#endif

/*
 * Get data to generate a TX_WRITE intent log record.
 */
int
zfs_get_data(void *arg, lr_write_t *lr, char *buf, struct lwb *lwb, zio_t *zio)
{
	zfsvfs_t *zfsvfs = arg;
	objset_t *os = zfsvfs->z_os;
	znode_t *zp;
	uint64_t object = lr->lr_foid;
	uint64_t offset = lr->lr_offset;
	uint64_t size = lr->lr_length;
	dmu_buf_t *db;
	zgd_t *zgd;
	int error = 0;

	ASSERT3P(lwb, !=, NULL);
	ASSERT3P(zio, !=, NULL);
	ASSERT3U(size, !=, 0);

	/*
	 * Nothing to do if the file has been removed
	 */
	if (zfs_zget(zfsvfs, object, &zp) != 0)
		return (SET_ERROR(ENOENT));
	if (zp->z_unlinked) {
		/*
		 * Release the vnode asynchronously as we currently have the
		 * txg stopped from syncing.
		 */
		VN_RELE_ASYNC(ZTOV(zp),
		    dsl_pool_vnrele_taskq(dmu_objset_pool(os)));
		return (SET_ERROR(ENOENT));
	}

	zgd = (zgd_t *)kmem_zalloc(sizeof (zgd_t), KM_SLEEP);
	zgd->zgd_lwb = lwb;
	zgd->zgd_private = zp;

	/*
	 * Write records come in two flavors: immediate and indirect.
	 * For small writes it's cheaper to store the data with the
	 * log record (immediate); for large writes it's cheaper to
	 * sync the data and get a pointer to it (indirect) so that
	 * we don't have to write the data twice.
	 */
	if (buf != NULL) { /* immediate write */
		zgd->zgd_rl = zfs_range_lock(zp, offset, size, RL_READER);
		/* test for truncation needs to be done while range locked */
		if (offset >= zp->z_size) {
			error = SET_ERROR(ENOENT);
		} else {
			error = dmu_read(os, object, offset, size, buf,
			    DMU_READ_NO_PREFETCH);
		}
		ASSERT(error == 0 || error == ENOENT);
	} else { /* indirect write */
		/*
		 * Have to lock the whole block to ensure when it's
		 * written out and its checksum is being calculated
		 * that no one can change the data. We need to re-check
		 * blocksize after we get the lock in case it's changed!
		 */
		for (;;) {
			uint64_t blkoff;
			size = zp->z_blksz;
			blkoff = ISP2(size) ? P2PHASE(offset, size) : offset;
			offset -= blkoff;
			zgd->zgd_rl = zfs_range_lock(zp, offset, size,
			    RL_READER);
			if (zp->z_blksz == size)
				break;
			offset += blkoff;
			zfs_range_unlock(zgd->zgd_rl);
		}
		/* test for truncation needs to be done while range locked */
		if (lr->lr_offset >= zp->z_size)
			error = SET_ERROR(ENOENT);
#ifdef DEBUG
		if (zil_fault_io) {
			error = SET_ERROR(EIO);
			zil_fault_io = 0;
		}
#endif
		if (error == 0)
			error = dmu_buf_hold(os, object, offset, zgd, &db,
			    DMU_READ_NO_PREFETCH);

		if (error == 0) {
			blkptr_t *bp = &lr->lr_blkptr;

			zgd->zgd_db = db;
			zgd->zgd_bp = bp;

			ASSERT(db->db_offset == offset);
			ASSERT(db->db_size == size);

			error = dmu_sync(zio, lr->lr_common.lrc_txg,
			    zfs_get_done, zgd);
			ASSERT(error || lr->lr_length <= size);

			/*
			 * On success, we need to wait for the write I/O
			 * initiated by dmu_sync() to complete before we can
			 * release this dbuf.  We will finish everything up
			 * in the zfs_get_done() callback.
			 */
			if (error == 0)
				return (0);

			if (error == EALREADY) {
				lr->lr_common.lrc_txtype = TX_WRITE2;
				/*
				 * TX_WRITE2 relies on the data previously
				 * written by the TX_WRITE that caused
				 * EALREADY.  We zero out the BP because
				 * it is the old, currently-on-disk BP.
				 */
				zgd->zgd_bp = NULL;
				BP_ZERO(bp);
				error = 0;
			}
		}
	}

	zfs_get_done(zgd, error);

	return (error);
}

/*ARGSUSED*/
static int
zfs_access(vnode_t *vp, int mode, int flag, cred_t *cr,
    caller_context_t *ct)
{
	znode_t *zp = VTOZ(vp);
	zfsvfs_t *zfsvfs = zp->z_zfsvfs;
	int error;

	ZFS_ENTER(zfsvfs);
	ZFS_VERIFY_ZP(zp);

	if (flag & V_ACE_MASK)
		error = zfs_zaccess(zp, mode, flag, B_FALSE, cr);
	else
		error = zfs_zaccess_rwx(zp, mode, flag, cr);

	ZFS_EXIT(zfsvfs);
	return (error);
}

/*
 * If vnode is for a device return a specfs vnode instead.
 */
static int
specvp_check(vnode_t **vpp, cred_t *cr)
{
	int error = 0;

	if (IS_DEVVP(*vpp)) {
		struct vnode *svp;

		svp = specvp(*vpp, (*vpp)->v_rdev, (*vpp)->v_type, cr);
		VN_RELE(*vpp);
		if (svp == NULL)
			error = SET_ERROR(ENOSYS);
		*vpp = svp;
	}
	return (error);
}


/*
 * Lookup an entry in a directory, or an extended attribute directory.
 * If it exists, return a held vnode reference for it.
 *
 *	IN:	dvp	- vnode of directory to search.
 *		nm	- name of entry to lookup.
 *		pnp	- full pathname to lookup [UNUSED].
 *		flags	- LOOKUP_XATTR set if looking for an attribute.
 *		rdir	- root directory vnode [UNUSED].
 *		cr	- credentials of caller.
 *		ct	- caller context
 *		direntflags - directory lookup flags
 *		realpnp - returned pathname.
 *
 *	OUT:	vpp	- vnode of located entry, NULL if not found.
 *
 *	RETURN:	0 on success, error code on failure.
 *
 * Timestamps:
 *	NA
 */
/* ARGSUSED */
static int
zfs_lookup(vnode_t *dvp, char *nm, vnode_t **vpp, struct pathname *pnp,
    int flags, vnode_t *rdir, cred_t *cr,  caller_context_t *ct,
    int *direntflags, pathname_t *realpnp)
{
	znode_t *zdp = VTOZ(dvp);
	zfsvfs_t *zfsvfs = zdp->z_zfsvfs;
	int	error = 0;

	/*
	 * Fast path lookup, however we must skip DNLC lookup
	 * for case folding or normalizing lookups because the
	 * DNLC code only stores the passed in name.  This means
	 * creating 'a' and removing 'A' on a case insensitive
	 * file system would work, but DNLC still thinks 'a'
	 * exists and won't let you create it again on the next
	 * pass through fast path.
	 */
	if (!(flags & (LOOKUP_XATTR | FIGNORECASE))) {

		if (dvp->v_type != VDIR) {
			return (SET_ERROR(ENOTDIR));
		} else if (zdp->z_sa_hdl == NULL) {
			return (SET_ERROR(EIO));
		}

		if (nm[0] == 0 || (nm[0] == '.' && nm[1] == '\0')) {
			error = zfs_fastaccesschk_execute(zdp, cr);
			if (!error) {
				*vpp = dvp;
				VN_HOLD(*vpp);
				return (0);
			}
			return (error);
		} else if (!zdp->z_zfsvfs->z_norm &&
		    (zdp->z_zfsvfs->z_case == ZFS_CASE_SENSITIVE)) {

			vnode_t *tvp = dnlc_lookup(dvp, nm);

			if (tvp) {
				error = zfs_fastaccesschk_execute(zdp, cr);
				if (error) {
					VN_RELE(tvp);
					return (error);
				}
				if (tvp == DNLC_NO_VNODE) {
					VN_RELE(tvp);
					return (SET_ERROR(ENOENT));
				} else {
					*vpp = tvp;
					return (specvp_check(vpp, cr));
				}
			}
		}
	}

	DTRACE_PROBE2(zfs__fastpath__lookup__miss, vnode_t *, dvp, char *, nm);

	ZFS_ENTER(zfsvfs);
	ZFS_VERIFY_ZP(zdp);

	*vpp = NULL;

	if (flags & LOOKUP_XATTR) {
		/*
		 * If the xattr property is off, refuse the lookup request.
		 */
		if (!(zfsvfs->z_vfs->vfs_flag & VFS_XATTR)) {
			ZFS_EXIT(zfsvfs);
			return (SET_ERROR(EINVAL));
		}

		/*
		 * We don't allow recursive attributes..
		 * Maybe someday we will.
		 */
		if (zdp->z_pflags & ZFS_XATTR) {
			ZFS_EXIT(zfsvfs);
			return (SET_ERROR(EINVAL));
		}

		if (error = zfs_get_xattrdir(VTOZ(dvp), vpp, cr, flags)) {
			ZFS_EXIT(zfsvfs);
			return (error);
		}

		/*
		 * Do we have permission to get into attribute directory?
		 */

		if (error = zfs_zaccess(VTOZ(*vpp), ACE_EXECUTE, 0,
		    B_FALSE, cr)) {
			VN_RELE(*vpp);
			*vpp = NULL;
		}

		ZFS_EXIT(zfsvfs);
		return (error);
	}

	if (dvp->v_type != VDIR) {
		ZFS_EXIT(zfsvfs);
		return (SET_ERROR(ENOTDIR));
	}

	/*
	 * Check accessibility of directory.
	 */

	if (error = zfs_zaccess(zdp, ACE_EXECUTE, 0, B_FALSE, cr)) {
		ZFS_EXIT(zfsvfs);
		return (error);
	}

	if (zfsvfs->z_utf8 && u8_validate(nm, strlen(nm),
	    NULL, U8_VALIDATE_ENTIRE, &error) < 0) {
		ZFS_EXIT(zfsvfs);
		return (SET_ERROR(EILSEQ));
	}

	error = zfs_dirlook(zdp, nm, vpp, flags, direntflags, realpnp);
	if (error == 0)
		error = specvp_check(vpp, cr);

	ZFS_EXIT(zfsvfs);
	return (error);
}

/*
 * Attempt to create a new entry in a directory.  If the entry
 * already exists, truncate the file if permissible, else return
 * an error.  Return the vp of the created or trunc'd file.
 *
 *	IN:	dvp	- vnode of directory to put new file entry in.
 *		name	- name of new file entry.
 *		vap	- attributes of new file.
 *		excl	- flag indicating exclusive or non-exclusive mode.
 *		mode	- mode to open file with.
 *		cr	- credentials of caller.
 *		flag	- large file flag [UNUSED].
 *		ct	- caller context
 *		vsecp	- ACL to be set
 *
 *	OUT:	vpp	- vnode of created or trunc'd entry.
 *
 *	RETURN:	0 on success, error code on failure.
 *
 * Timestamps:
 *	dvp - ctime|mtime updated if new entry created
 *	 vp - ctime|mtime always, atime if new
 */

/* ARGSUSED */
static int
zfs_create(vnode_t *dvp, char *name, vattr_t *vap, vcexcl_t excl,
    int mode, vnode_t **vpp, cred_t *cr, int flag, caller_context_t *ct,
    vsecattr_t *vsecp)
{
	znode_t		*zp, *dzp = VTOZ(dvp);
	zfsvfs_t	*zfsvfs = dzp->z_zfsvfs;
	zilog_t		*zilog;
	objset_t	*os;
	zfs_dirlock_t	*dl;
	dmu_tx_t	*tx;
	int		error;
	ksid_t		*ksid;
	uid_t		uid;
	gid_t		gid = crgetgid(cr);
	zfs_acl_ids_t   acl_ids;
	boolean_t	fuid_dirtied;
	boolean_t	have_acl = B_FALSE;
	boolean_t	waited = B_FALSE;

	/*
	 * If we have an ephemeral id, ACL, or XVATTR then
	 * make sure file system is at proper version
	 */

	ksid = crgetsid(cr, KSID_OWNER);
	if (ksid)
		uid = ksid_getid(ksid);
	else
		uid = crgetuid(cr);

	if (zfsvfs->z_use_fuids == B_FALSE &&
	    (vsecp || (vap->va_mask & AT_XVATTR) ||
	    IS_EPHEMERAL(uid) || IS_EPHEMERAL(gid)))
		return (SET_ERROR(EINVAL));

	ZFS_ENTER(zfsvfs);
	ZFS_VERIFY_ZP(dzp);
	os = zfsvfs->z_os;
	zilog = zfsvfs->z_log;

	if (zfsvfs->z_utf8 && u8_validate(name, strlen(name),
	    NULL, U8_VALIDATE_ENTIRE, &error) < 0) {
		ZFS_EXIT(zfsvfs);
		return (SET_ERROR(EILSEQ));
	}

	if (vap->va_mask & AT_XVATTR) {
		if ((error = secpolicy_xvattr((xvattr_t *)vap,
		    crgetuid(cr), cr, vap->va_type)) != 0) {
			ZFS_EXIT(zfsvfs);
			return (error);
		}
	}
top:
	*vpp = NULL;

	if ((vap->va_mode & VSVTX) && secpolicy_vnode_stky_modify(cr))
		vap->va_mode &= ~VSVTX;

	if (*name == '\0') {
		/*
		 * Null component name refers to the directory itself.
		 */
		VN_HOLD(dvp);
		zp = dzp;
		dl = NULL;
		error = 0;
	} else {
		/* possible VN_HOLD(zp) */
		int zflg = 0;

		if (flag & FIGNORECASE)
			zflg |= ZCILOOK;

		error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg,
		    NULL, NULL);
		if (error) {
			if (have_acl)
				zfs_acl_ids_free(&acl_ids);
			if (strcmp(name, "..") == 0)
				error = SET_ERROR(EISDIR);
			ZFS_EXIT(zfsvfs);
			return (error);
		}
	}

	if (zp == NULL) {
		uint64_t txtype;

		/*
		 * Create a new file object and update the directory
		 * to reference it.
		 */
		if (error = zfs_zaccess(dzp, ACE_ADD_FILE, 0, B_FALSE, cr)) {
			if (have_acl)
				zfs_acl_ids_free(&acl_ids);
			goto out;
		}

		/*
		 * We only support the creation of regular files in
		 * extended attribute directories.
		 */

		if ((dzp->z_pflags & ZFS_XATTR) &&
		    (vap->va_type != VREG)) {
			if (have_acl)
				zfs_acl_ids_free(&acl_ids);
			error = SET_ERROR(EINVAL);
			goto out;
		}

		if (!have_acl && (error = zfs_acl_ids_create(dzp, 0, vap,
		    cr, vsecp, &acl_ids)) != 0)
			goto out;
		have_acl = B_TRUE;

		if (zfs_acl_ids_overquota(zfsvfs, &acl_ids)) {
			zfs_acl_ids_free(&acl_ids);
			error = SET_ERROR(EDQUOT);
			goto out;
		}

		tx = dmu_tx_create(os);

		dmu_tx_hold_sa_create(tx, acl_ids.z_aclp->z_acl_bytes +
		    ZFS_SA_BASE_ATTR_SIZE);

		fuid_dirtied = zfsvfs->z_fuid_dirty;
		if (fuid_dirtied)
			zfs_fuid_txhold(zfsvfs, tx);
		dmu_tx_hold_zap(tx, dzp->z_id, TRUE, name);
		dmu_tx_hold_sa(tx, dzp->z_sa_hdl, B_FALSE);
		if (!zfsvfs->z_use_sa &&
		    acl_ids.z_aclp->z_acl_bytes > ZFS_ACE_SPACE) {
			dmu_tx_hold_write(tx, DMU_NEW_OBJECT,
			    0, acl_ids.z_aclp->z_acl_bytes);
		}
		error = dmu_tx_assign(tx,
		    (waited ? TXG_NOTHROTTLE : 0) | TXG_NOWAIT);
		if (error) {
			zfs_dirent_unlock(dl);
			if (error == ERESTART) {
				waited = B_TRUE;
				dmu_tx_wait(tx);
				dmu_tx_abort(tx);
				goto top;
			}
			zfs_acl_ids_free(&acl_ids);
			dmu_tx_abort(tx);
			ZFS_EXIT(zfsvfs);
			return (error);
		}
		zfs_mknode(dzp, vap, tx, cr, 0, &zp, &acl_ids);

		if (fuid_dirtied)
			zfs_fuid_sync(zfsvfs, tx);

		(void) zfs_link_create(dl, zp, tx, ZNEW);
		txtype = zfs_log_create_txtype(Z_FILE, vsecp, vap);
		if (flag & FIGNORECASE)
			txtype |= TX_CI;
		zfs_log_create(zilog, tx, txtype, dzp, zp, name,
		    vsecp, acl_ids.z_fuidp, vap);
		zfs_acl_ids_free(&acl_ids);
		dmu_tx_commit(tx);
	} else {
		int aflags = (flag & FAPPEND) ? V_APPEND : 0;

		if (have_acl)
			zfs_acl_ids_free(&acl_ids);
		have_acl = B_FALSE;

		/*
		 * A directory entry already exists for this name.
		 */
		/*
		 * Can't truncate an existing file if in exclusive mode.
		 */
		if (excl == EXCL) {
			error = SET_ERROR(EEXIST);
			goto out;
		}
		/*
		 * Can't open a directory for writing.
		 */
		if ((ZTOV(zp)->v_type == VDIR) && (mode & S_IWRITE)) {
			error = SET_ERROR(EISDIR);
			goto out;
		}
		/*
		 * Verify requested access to file.
		 */
		if (mode && (error = zfs_zaccess_rwx(zp, mode, aflags, cr))) {
			goto out;
		}

		mutex_enter(&dzp->z_lock);
		dzp->z_seq++;
		mutex_exit(&dzp->z_lock);

		/*
		 * Truncate regular files if requested.
		 */
		if ((ZTOV(zp)->v_type == VREG) &&
		    (vap->va_mask & AT_SIZE) && (vap->va_size == 0)) {
			/* we can't hold any locks when calling zfs_freesp() */
			zfs_dirent_unlock(dl);
			dl = NULL;
			error = zfs_freesp(zp, 0, 0, mode, TRUE);
			if (error == 0) {
				vnevent_create(ZTOV(zp), ct);
			}
		}
	}
out:

	if (dl)
		zfs_dirent_unlock(dl);

	if (error) {
		if (zp)
			VN_RELE(ZTOV(zp));
	} else {
		*vpp = ZTOV(zp);
		error = specvp_check(vpp, cr);
	}

	if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS)
		zil_commit(zilog, 0);

	ZFS_EXIT(zfsvfs);
	return (error);
}

/*
 * Remove an entry from a directory.
 *
 *	IN:	dvp	- vnode of directory to remove entry from.
 *		name	- name of entry to remove.
 *		cr	- credentials of caller.
 *		ct	- caller context
 *		flags	- case flags
 *
 *	RETURN:	0 on success, error code on failure.
 *
 * Timestamps:
 *	dvp - ctime|mtime
 *	 vp - ctime (if nlink > 0)
 */

uint64_t null_xattr = 0;

/*ARGSUSED*/
static int
zfs_remove(vnode_t *dvp, char *name, cred_t *cr, caller_context_t *ct,
    int flags)
{
	znode_t		*zp, *dzp = VTOZ(dvp);
	znode_t		*xzp;
	vnode_t		*vp;
	zfsvfs_t	*zfsvfs = dzp->z_zfsvfs;
	zilog_t		*zilog;
	uint64_t	acl_obj, xattr_obj;
	uint64_t	xattr_obj_unlinked = 0;
	uint64_t	obj = 0;
	zfs_dirlock_t	*dl;
	dmu_tx_t	*tx;
	boolean_t	may_delete_now, delete_now = FALSE;
	boolean_t	unlinked, toobig = FALSE;
	uint64_t	txtype;
	pathname_t	*realnmp = NULL;
	pathname_t	realnm;
	int		error;
	int		zflg = ZEXISTS;
	boolean_t	waited = B_FALSE;

	ZFS_ENTER(zfsvfs);
	ZFS_VERIFY_ZP(dzp);
	zilog = zfsvfs->z_log;

	if (flags & FIGNORECASE) {
		zflg |= ZCILOOK;
		pn_alloc(&realnm);
		realnmp = &realnm;
	}

top:
	xattr_obj = 0;
	xzp = NULL;
	/*
	 * Attempt to lock directory; fail if entry doesn't exist.
	 */
	if (error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg,
	    NULL, realnmp)) {
		if (realnmp)
			pn_free(realnmp);
		ZFS_EXIT(zfsvfs);
		return (error);
	}

	vp = ZTOV(zp);

	if (error = zfs_zaccess_delete(dzp, zp, cr)) {
		goto out;
	}

	/*
	 * Need to use rmdir for removing directories.
	 */
	if (vp->v_type == VDIR) {
		error = SET_ERROR(EPERM);
		goto out;
	}

	vnevent_remove(vp, dvp, name, ct);

	if (realnmp)
		dnlc_remove(dvp, realnmp->pn_buf);
	else
		dnlc_remove(dvp, name);

	mutex_enter(&vp->v_lock);
	may_delete_now = vp->v_count == 1 && !vn_has_cached_data(vp);
	mutex_exit(&vp->v_lock);

	/*
	 * We may delete the znode now, or we may put it in the unlinked set;
	 * it depends on whether we're the last link, and on whether there are
	 * other holds on the vnode.  So we dmu_tx_hold() the right things to
	 * allow for either case.
	 */
	obj = zp->z_id;
	tx = dmu_tx_create(zfsvfs->z_os);
	dmu_tx_hold_zap(tx, dzp->z_id, FALSE, name);
	dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
	zfs_sa_upgrade_txholds(tx, zp);
	zfs_sa_upgrade_txholds(tx, dzp);
	if (may_delete_now) {
		toobig =
		    zp->z_size > zp->z_blksz * DMU_MAX_DELETEBLKCNT;
		/* if the file is too big, only hold_free a token amount */
		dmu_tx_hold_free(tx, zp->z_id, 0,
		    (toobig ? DMU_MAX_ACCESS : DMU_OBJECT_END));
	}

	/* are there any extended attributes? */
	error = sa_lookup(zp->z_sa_hdl, SA_ZPL_XATTR(zfsvfs),
	    &xattr_obj, sizeof (xattr_obj));
	if (error == 0 && xattr_obj) {
		error = zfs_zget(zfsvfs, xattr_obj, &xzp);
		ASSERT0(error);
		dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE);
		dmu_tx_hold_sa(tx, xzp->z_sa_hdl, B_FALSE);
	}

	mutex_enter(&zp->z_lock);
	if ((acl_obj = zfs_external_acl(zp)) != 0 && may_delete_now)
		dmu_tx_hold_free(tx, acl_obj, 0, DMU_OBJECT_END);
	mutex_exit(&zp->z_lock);

	/* charge as an update -- would be nice not to charge at all */
	dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL);

	/*
	 * Mark this transaction as typically resulting in a net free of space
	 */
	dmu_tx_mark_netfree(tx);

	error = dmu_tx_assign(tx, (waited ? TXG_NOTHROTTLE : 0) | TXG_NOWAIT);
	if (error) {
		zfs_dirent_unlock(dl);
		VN_RELE(vp);
		if (xzp)
			VN_RELE(ZTOV(xzp));
		if (error == ERESTART) {
			waited = B_TRUE;
			dmu_tx_wait(tx);
			dmu_tx_abort(tx);
			goto top;
		}
		if (realnmp)
			pn_free(realnmp);
		dmu_tx_abort(tx);
		ZFS_EXIT(zfsvfs);
		return (error);
	}

	/*
	 * Remove the directory entry.
	 */
	error = zfs_link_destroy(dl, zp, tx, zflg, &unlinked);

	if (error) {
		dmu_tx_commit(tx);
		goto out;
	}

	if (unlinked) {
		/*
		 * Hold z_lock so that we can make sure that the ACL obj
		 * hasn't changed.  Could have been deleted due to
		 * zfs_sa_upgrade().
		 */
		mutex_enter(&zp->z_lock);
		mutex_enter(&vp->v_lock);
		(void) sa_lookup(zp->z_sa_hdl, SA_ZPL_XATTR(zfsvfs),
		    &xattr_obj_unlinked, sizeof (xattr_obj_unlinked));
		delete_now = may_delete_now && !toobig &&
		    vp->v_count == 1 && !vn_has_cached_data(vp) &&
		    xattr_obj == xattr_obj_unlinked && zfs_external_acl(zp) ==
		    acl_obj;
		mutex_exit(&vp->v_lock);
	}

	if (delete_now) {
		if (xattr_obj_unlinked) {
			ASSERT3U(xzp->z_links, ==, 2);
			mutex_enter(&xzp->z_lock);
			xzp->z_unlinked = 1;
			xzp->z_links = 0;
			error = sa_update(xzp->z_sa_hdl, SA_ZPL_LINKS(zfsvfs),
			    &xzp->z_links, sizeof (xzp->z_links), tx);
			ASSERT3U(error,  ==,  0);
			mutex_exit(&xzp->z_lock);
			zfs_unlinked_add(xzp, tx);

			if (zp->z_is_sa)
				error = sa_remove(zp->z_sa_hdl,
				    SA_ZPL_XATTR(zfsvfs), tx);
			else
				error = sa_update(zp->z_sa_hdl,
				    SA_ZPL_XATTR(zfsvfs), &null_xattr,
				    sizeof (uint64_t), tx);
			ASSERT0(error);
		}
		mutex_enter(&vp->v_lock);
		VN_RELE_LOCKED(vp);
		ASSERT0(vp->v_count);
		mutex_exit(&vp->v_lock);
		mutex_exit(&zp->z_lock);
		zfs_znode_delete(zp, tx);
	} else if (unlinked) {
		mutex_exit(&zp->z_lock);
		zfs_unlinked_add(zp, tx);
	}

	txtype = TX_REMOVE;
	if (flags & FIGNORECASE)
		txtype |= TX_CI;
	zfs_log_remove(zilog, tx, txtype, dzp, name, obj);

	dmu_tx_commit(tx);
out:
	if (realnmp)
		pn_free(realnmp);

	zfs_dirent_unlock(dl);

	if (!delete_now)
		VN_RELE(vp);
	if (xzp)
		VN_RELE(ZTOV(xzp));

	if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS)
		zil_commit(zilog, 0);

	ZFS_EXIT(zfsvfs);
	return (error);
}

/*
 * Create a new directory and insert it into dvp using the name
 * provided.  Return a pointer to the inserted directory.
 *
 *	IN:	dvp	- vnode of directory to add subdir to.
 *		dirname	- name of new directory.
 *		vap	- attributes of new directory.
 *		cr	- credentials of caller.
 *		ct	- caller context
 *		flags	- case flags
 *		vsecp	- ACL to be set
 *
 *	OUT:	vpp	- vnode of created directory.
 *
 *	RETURN:	0 on success, error code on failure.
 *
 * Timestamps:
 *	dvp - ctime|mtime updated
 *	 vp - ctime|mtime|atime updated
 */
/*ARGSUSED*/
static int
zfs_mkdir(vnode_t *dvp, char *dirname, vattr_t *vap, vnode_t **vpp, cred_t *cr,
    caller_context_t *ct, int flags, vsecattr_t *vsecp)
{
	znode_t		*zp, *dzp = VTOZ(dvp);
	zfsvfs_t	*zfsvfs = dzp->z_zfsvfs;
	zilog_t		*zilog;
	zfs_dirlock_t	*dl;
	uint64_t	txtype;
	dmu_tx_t	*tx;
	int		error;
	int		zf = ZNEW;
	ksid_t		*ksid;
	uid_t		uid;
	gid_t		gid = crgetgid(cr);
	zfs_acl_ids_t   acl_ids;
	boolean_t	fuid_dirtied;
	boolean_t	waited = B_FALSE;

	ASSERT(vap->va_type == VDIR);

	/*
	 * If we have an ephemeral id, ACL, or XVATTR then
	 * make sure file system is at proper version
	 */

	ksid = crgetsid(cr, KSID_OWNER);
	if (ksid)
		uid = ksid_getid(ksid);
	else
		uid = crgetuid(cr);
	if (zfsvfs->z_use_fuids == B_FALSE &&
	    (vsecp || (vap->va_mask & AT_XVATTR) ||
	    IS_EPHEMERAL(uid) || IS_EPHEMERAL(gid)))
		return (SET_ERROR(EINVAL));

	ZFS_ENTER(zfsvfs);
	ZFS_VERIFY_ZP(dzp);
	zilog = zfsvfs->z_log;

	if (dzp->z_pflags & ZFS_XATTR) {
		ZFS_EXIT(zfsvfs);
		return (SET_ERROR(EINVAL));
	}

	if (zfsvfs->z_utf8 && u8_validate(dirname,
	    strlen(dirname), NULL, U8_VALIDATE_ENTIRE, &error) < 0) {
		ZFS_EXIT(zfsvfs);
		return (SET_ERROR(EILSEQ));
	}
	if (flags & FIGNORECASE)
		zf |= ZCILOOK;

	if (vap->va_mask & AT_XVATTR) {
		if ((error = secpolicy_xvattr((xvattr_t *)vap,
		    crgetuid(cr), cr, vap->va_type)) != 0) {
			ZFS_EXIT(zfsvfs);
			return (error);
		}
	}

	if ((error = zfs_acl_ids_create(dzp, 0, vap, cr,
	    vsecp, &acl_ids)) != 0) {
		ZFS_EXIT(zfsvfs);
		return (error);
	}
	/*
	 * First make sure the new directory doesn't exist.
	 *
	 * Existence is checked first to make sure we don't return
	 * EACCES instead of EEXIST which can cause some applications
	 * to fail.
	 */
top:
	*vpp = NULL;

	if (error = zfs_dirent_lock(&dl, dzp, dirname, &zp, zf,
	    NULL, NULL)) {
		zfs_acl_ids_free(&acl_ids);
		ZFS_EXIT(zfsvfs);
		return (error);
	}

	if (error = zfs_zaccess(dzp, ACE_ADD_SUBDIRECTORY, 0, B_FALSE, cr)) {
		zfs_acl_ids_free(&acl_ids);
		zfs_dirent_unlock(dl);
		ZFS_EXIT(zfsvfs);
		return (error);
	}

	if (zfs_acl_ids_overquota(zfsvfs, &acl_ids)) {
		zfs_acl_ids_free(&acl_ids);
		zfs_dirent_unlock(dl);
		ZFS_EXIT(zfsvfs);
		return (SET_ERROR(EDQUOT));
	}

	/*
	 * Add a new entry to the directory.
	 */
	tx = dmu_tx_create(zfsvfs->z_os);
	dmu_tx_hold_zap(tx, dzp->z_id, TRUE, dirname);
	dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL);
	fuid_dirtied = zfsvfs->z_fuid_dirty;
	if (fuid_dirtied)
		zfs_fuid_txhold(zfsvfs, tx);
	if (!zfsvfs->z_use_sa && acl_ids.z_aclp->z_acl_bytes > ZFS_ACE_SPACE) {
		dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0,
		    acl_ids.z_aclp->z_acl_bytes);
	}

	dmu_tx_hold_sa_create(tx, acl_ids.z_aclp->z_acl_bytes +
	    ZFS_SA_BASE_ATTR_SIZE);

	error = dmu_tx_assign(tx, (waited ? TXG_NOTHROTTLE : 0) | TXG_NOWAIT);
	if (error) {
		zfs_dirent_unlock(dl);
		if (error == ERESTART) {
			waited = B_TRUE;
			dmu_tx_wait(tx);
			dmu_tx_abort(tx);
			goto top;
		}
		zfs_acl_ids_free(&acl_ids);
		dmu_tx_abort(tx);
		ZFS_EXIT(zfsvfs);
		return (error);
	}

	/*
	 * Create new node.
	 */
	zfs_mknode(dzp, vap, tx, cr, 0, &zp, &acl_ids);

	if (fuid_dirtied)
		zfs_fuid_sync(zfsvfs, tx);

	/*
	 * Now put new name in parent dir.
	 */
	(void) zfs_link_create(dl, zp, tx, ZNEW);

	*vpp = ZTOV(zp);

	txtype = zfs_log_create_txtype(Z_DIR, vsecp, vap);
	if (flags & FIGNORECASE)
		txtype |= TX_CI;
	zfs_log_create(zilog, tx, txtype, dzp, zp, dirname, vsecp,
	    acl_ids.z_fuidp, vap);

	zfs_acl_ids_free(&acl_ids);

	dmu_tx_commit(tx);

	zfs_dirent_unlock(dl);

	if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS)
		zil_commit(zilog, 0);

	ZFS_EXIT(zfsvfs);
	return (0);
}

/*
 * Remove a directory subdir entry.  If the current working
 * directory is the same as the subdir to be removed, the
 * remove will fail.
 *
 *	IN:	dvp	- vnode of directory to remove from.
 *		name	- name of directory to be removed.
 *		cwd	- vnode of current working directory.
 *		cr	- credentials of caller.
 *		ct	- caller context
 *		flags	- case flags
 *
 *	RETURN:	0 on success, error code on failure.
 *
 * Timestamps:
 *	dvp - ctime|mtime updated
 */
/*ARGSUSED*/
static int
zfs_rmdir(vnode_t *dvp, char *name, vnode_t *cwd, cred_t *cr,
    caller_context_t *ct, int flags)
{
	znode_t		*dzp = VTOZ(dvp);
	znode_t		*zp;
	vnode_t		*vp;
	zfsvfs_t	*zfsvfs = dzp->z_zfsvfs;
	zilog_t		*zilog;
	zfs_dirlock_t	*dl;
	dmu_tx_t	*tx;
	int		error;
	int		zflg = ZEXISTS;
	boolean_t	waited = B_FALSE;

	ZFS_ENTER(zfsvfs);
	ZFS_VERIFY_ZP(dzp);
	zilog = zfsvfs->z_log;

	if (flags & FIGNORECASE)
		zflg |= ZCILOOK;
top:
	zp = NULL;

	/*
	 * Attempt to lock directory; fail if entry doesn't exist.
	 */
	if (error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg,
	    NULL, NULL)) {
		ZFS_EXIT(zfsvfs);
		return (error);
	}

	vp = ZTOV(zp);

	if (error = zfs_zaccess_delete(dzp, zp, cr)) {
		goto out;
	}

	if (vp->v_type != VDIR) {
		error = SET_ERROR(ENOTDIR);
		goto out;
	}

	if (vp == cwd) {
		error = SET_ERROR(EINVAL);
		goto out;
	}

	vnevent_rmdir(vp, dvp, name, ct);

	/*
	 * Grab a lock on the directory to make sure that noone is
	 * trying to add (or lookup) entries while we are removing it.
	 */
	rw_enter(&zp->z_name_lock, RW_WRITER);

	/*
	 * Grab a lock on the parent pointer to make sure we play well
	 * with the treewalk and directory rename code.
	 */
	rw_enter(&zp->z_parent_lock, RW_WRITER);

	tx = dmu_tx_create(zfsvfs->z_os);
	dmu_tx_hold_zap(tx, dzp->z_id, FALSE, name);
	dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
	dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL);
	zfs_sa_upgrade_txholds(tx, zp);
	zfs_sa_upgrade_txholds(tx, dzp);
	dmu_tx_mark_netfree(tx);
	error = dmu_tx_assign(tx, (waited ? TXG_NOTHROTTLE : 0) | TXG_NOWAIT);
	if (error) {
		rw_exit(&zp->z_parent_lock);
		rw_exit(&zp->z_name_lock);
		zfs_dirent_unlock(dl);
		VN_RELE(vp);
		if (error == ERESTART) {
			waited = B_TRUE;
			dmu_tx_wait(tx);
			dmu_tx_abort(tx);
			goto top;
		}
		dmu_tx_abort(tx);
		ZFS_EXIT(zfsvfs);
		return (error);
	}

	error = zfs_link_destroy(dl, zp, tx, zflg, NULL);

	if (error == 0) {
		uint64_t txtype = TX_RMDIR;
		if (flags & FIGNORECASE)
			txtype |= TX_CI;
		zfs_log_remove(zilog, tx, txtype, dzp, name, ZFS_NO_OBJECT);
	}

	dmu_tx_commit(tx);

	rw_exit(&zp->z_parent_lock);
	rw_exit(&zp->z_name_lock);
out:
	zfs_dirent_unlock(dl);

	VN_RELE(vp);

	if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS)
		zil_commit(zilog, 0);

	ZFS_EXIT(zfsvfs);
	return (error);
}

/*
 * Read as many directory entries as will fit into the provided
 * buffer from the given directory cursor position (specified in
 * the uio structure).
 *
 *	IN:	vp	- vnode of directory to read.
 *		uio	- structure supplying read location, range info,
 *			  and return buffer.
 *		cr	- credentials of caller.
 *		ct	- caller context
 *		flags	- case flags
 *
 *	OUT:	uio	- updated offset and range, buffer filled.
 *		eofp	- set to true if end-of-file detected.
 *
 *	RETURN:	0 on success, error code on failure.
 *
 * Timestamps:
 *	vp - atime updated
 *
 * Note that the low 4 bits of the cookie returned by zap is always zero.
 * This allows us to use the low range for "special" directory entries:
 * We use 0 for '.', and 1 for '..'.  If this is the root of the filesystem,
 * we use the offset 2 for the '.zfs' directory.
 */
/* ARGSUSED */
static int
zfs_readdir(vnode_t *vp, uio_t *uio, cred_t *cr, int *eofp,
    caller_context_t *ct, int flags)
{
	znode_t		*zp = VTOZ(vp);
	iovec_t		*iovp;
	edirent_t	*eodp;
	dirent64_t	*odp;
	zfsvfs_t	*zfsvfs = zp->z_zfsvfs;
	objset_t	*os;
	caddr_t		outbuf;
	size_t		bufsize;
	zap_cursor_t	zc;
	zap_attribute_t	zap;
	uint_t		bytes_wanted;
	uint64_t	offset; /* must be unsigned; checks for < 1 */
	uint64_t	parent;
	int		local_eof;
	int		outcount;
	int		error;
	uint8_t		prefetch;
	boolean_t	check_sysattrs;

	ZFS_ENTER(zfsvfs);
	ZFS_VERIFY_ZP(zp);

	if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_PARENT(zfsvfs),
	    &parent, sizeof (parent))) != 0) {
		ZFS_EXIT(zfsvfs);
		return (error);
	}

	/*
	 * If we are not given an eof variable,
	 * use a local one.
	 */
	if (eofp == NULL)
		eofp = &local_eof;

	/*
	 * Check for valid iov_len.
	 */
	if (uio->uio_iov->iov_len <= 0) {
		ZFS_EXIT(zfsvfs);
		return (SET_ERROR(EINVAL));
	}

	/*
	 * Quit if directory has been removed (posix)
	 */
	if ((*eofp = zp->z_unlinked) != 0) {
		ZFS_EXIT(zfsvfs);
		return (0);
	}

	error = 0;
	os = zfsvfs->z_os;
	offset = uio->uio_loffset;
	prefetch = zp->z_zn_prefetch;

	/*
	 * Initialize the iterator cursor.
	 */
	if (offset <= 3) {
		/*
		 * Start iteration from the beginning of the directory.
		 */
		zap_cursor_init(&zc, os, zp->z_id);
	} else {
		/*
		 * The offset is a serialized cursor.
		 */
		zap_cursor_init_serialized(&zc, os, zp->z_id, offset);
	}

	/*
	 * Get space to change directory entries into fs independent format.
	 */
	iovp = uio->uio_iov;
	bytes_wanted = iovp->iov_len;
	if (uio->uio_segflg != UIO_SYSSPACE || uio->uio_iovcnt != 1) {
		bufsize = bytes_wanted;
		outbuf = kmem_alloc(bufsize, KM_SLEEP);
		odp = (struct dirent64 *)outbuf;
	} else {
		bufsize = bytes_wanted;
		outbuf = NULL;
		odp = (struct dirent64 *)iovp->iov_base;
	}
	eodp = (struct edirent *)odp;

	/*
	 * If this VFS supports the system attribute view interface; and
	 * we're looking at an extended attribute directory; and we care
	 * about normalization conflicts on this vfs; then we must check
	 * for normalization conflicts with the sysattr name space.
	 */
	check_sysattrs = vfs_has_feature(vp->v_vfsp, VFSFT_SYSATTR_VIEWS) &&
	    (vp->v_flag & V_XATTRDIR) && zfsvfs->z_norm &&
	    (flags & V_RDDIR_ENTFLAGS);

	/*
	 * Transform to file-system independent format
	 */
	outcount = 0;
	while (outcount < bytes_wanted) {
		ino64_t objnum;
		ushort_t reclen;
		off64_t *next = NULL;

		/*
		 * Special case `.', `..', and `.zfs'.
		 */
		if (offset == 0) {
			(void) strcpy(zap.za_name, ".");
			zap.za_normalization_conflict = 0;
			objnum = zp->z_id;
		} else if (offset == 1) {
			(void) strcpy(zap.za_name, "..");
			zap.za_normalization_conflict = 0;
			objnum = parent;
		} else if (offset == 2 && zfs_show_ctldir(zp)) {
			(void) strcpy(zap.za_name, ZFS_CTLDIR_NAME);
			zap.za_normalization_conflict = 0;
			objnum = ZFSCTL_INO_ROOT;
		} else {
			/*
			 * Grab next entry.
			 */
			if (error = zap_cursor_retrieve(&zc, &zap)) {
				if ((*eofp = (error == ENOENT)) != 0)
					break;
				else
					goto update;
			}

			if (zap.za_integer_length != 8 ||
			    zap.za_num_integers != 1) {
				cmn_err(CE_WARN, "zap_readdir: bad directory "
				    "entry, obj = %lld, offset = %lld\n",
				    (u_longlong_t)zp->z_id,
				    (u_longlong_t)offset);
				error = SET_ERROR(ENXIO);
				goto update;
			}

			objnum = ZFS_DIRENT_OBJ(zap.za_first_integer);
			/*
			 * MacOS X can extract the object type here such as:
			 * uint8_t type = ZFS_DIRENT_TYPE(zap.za_first_integer);
			 */

			if (check_sysattrs && !zap.za_normalization_conflict) {
				zap.za_normalization_conflict =
				    xattr_sysattr_casechk(zap.za_name);
			}
		}

		if (flags & V_RDDIR_ACCFILTER) {
			/*
			 * If we have no access at all, don't include
			 * this entry in the returned information
			 */
			znode_t	*ezp;
			if (zfs_zget(zp->z_zfsvfs, objnum, &ezp) != 0)
				goto skip_entry;
			if (!zfs_has_access(ezp, cr)) {
				VN_RELE(ZTOV(ezp));
				goto skip_entry;
			}
			VN_RELE(ZTOV(ezp));
		}

		if (flags & V_RDDIR_ENTFLAGS)
			reclen = EDIRENT_RECLEN(strlen(zap.za_name));
		else
			reclen = DIRENT64_RECLEN(strlen(zap.za_name));

		/*
		 * Will this entry fit in the buffer?
		 */
		if (outcount + reclen > bufsize) {
			/*
			 * Did we manage to fit anything in the buffer?
			 */
			if (!outcount) {
				error = SET_ERROR(EINVAL);
				goto update;
			}
			break;
		}
		if (flags & V_RDDIR_ENTFLAGS) {
			/*
			 * Add extended flag entry:
			 */
			eodp->ed_ino = objnum;
			eodp->ed_reclen = reclen;
			/* NOTE: ed_off is the offset for the *next* entry */
			next = &(eodp->ed_off);
			eodp->ed_eflags = zap.za_normalization_conflict ?
			    ED_CASE_CONFLICT : 0;
			(void) strncpy(eodp->ed_name, zap.za_name,
			    EDIRENT_NAMELEN(reclen));
			eodp = (edirent_t *)((intptr_t)eodp + reclen);
		} else {
			/*
			 * Add normal entry:
			 */
			odp->d_ino = objnum;
			odp->d_reclen = reclen;
			/* NOTE: d_off is the offset for the *next* entry */
			next = &(odp->d_off);
			(void) strncpy(odp->d_name, zap.za_name,
			    DIRENT64_NAMELEN(reclen));
			odp = (dirent64_t *)((intptr_t)odp + reclen);
		}
		outcount += reclen;

		ASSERT(outcount <= bufsize);

		/* Prefetch znode */
		if (prefetch)
			dmu_prefetch(os, objnum, 0, 0, 0,
			    ZIO_PRIORITY_SYNC_READ);

	skip_entry:
		/*
		 * Move to the next entry, fill in the previous offset.
		 */
		if (offset > 2 || (offset == 2 && !zfs_show_ctldir(zp))) {
			zap_cursor_advance(&zc);
			offset = zap_cursor_serialize(&zc);
		} else {
			offset += 1;
		}
		if (next)
			*next = offset;
	}
	zp->z_zn_prefetch = B_FALSE; /* a lookup will re-enable pre-fetching */

	if (uio->uio_segflg == UIO_SYSSPACE && uio->uio_iovcnt == 1) {
		iovp->iov_base += outcount;
		iovp->iov_len -= outcount;
		uio->uio_resid -= outcount;
	} else if (error = uiomove(outbuf, (long)outcount, UIO_READ, uio)) {
		/*
		 * Reset the pointer.
		 */
		offset = uio->uio_loffset;
	}

update:
	zap_cursor_fini(&zc);
	if (uio->uio_segflg != UIO_SYSSPACE || uio->uio_iovcnt != 1)
		kmem_free(outbuf, bufsize);

	if (error == ENOENT)
		error = 0;

	ZFS_ACCESSTIME_STAMP(zfsvfs, zp);

	uio->uio_loffset = offset;
	ZFS_EXIT(zfsvfs);
	return (error);
}

ulong_t zfs_fsync_sync_cnt = 4;

static int
zfs_fsync(vnode_t *vp, int syncflag, cred_t *cr, caller_context_t *ct)
{
	znode_t	*zp = VTOZ(vp);
	zfsvfs_t *zfsvfs = zp->z_zfsvfs;

	/*
	 * Regardless of whether this is required for standards conformance,
	 * this is the logical behavior when fsync() is called on a file with
	 * dirty pages.  We use B_ASYNC since the ZIL transactions are already
	 * going to be pushed out as part of the zil_commit().
	 */
	if (vn_has_cached_data(vp) && !(syncflag & FNODSYNC) &&
	    (vp->v_type == VREG) && !(IS_SWAPVP(vp)))
		(void) VOP_PUTPAGE(vp, (offset_t)0, (size_t)0, B_ASYNC, cr, ct);

	(void) tsd_set(zfs_fsyncer_key, (void *)zfs_fsync_sync_cnt);

	if (zfsvfs->z_os->os_sync != ZFS_SYNC_DISABLED) {
		ZFS_ENTER(zfsvfs);
		ZFS_VERIFY_ZP(zp);
		zil_commit(zfsvfs->z_log, zp->z_id);
		ZFS_EXIT(zfsvfs);
	}
	return (0);
}


/*
 * Get the requested file attributes and place them in the provided
 * vattr structure.
 *
 *	IN:	vp	- vnode of file.
 *		vap	- va_mask identifies requested attributes.
 *			  If AT_XVATTR set, then optional attrs are requested
 *		flags	- ATTR_NOACLCHECK (CIFS server context)
 *		cr	- credentials of caller.
 *		ct	- caller context
 *
 *	OUT:	vap	- attribute values.
 *
 *	RETURN:	0 (always succeeds).
 */
/* ARGSUSED */
static int
zfs_getattr(vnode_t *vp, vattr_t *vap, int flags, cred_t *cr,
    caller_context_t *ct)
{
	znode_t *zp = VTOZ(vp);
	zfsvfs_t *zfsvfs = zp->z_zfsvfs;
	int	error = 0;
	uint64_t links;
	uint64_t mtime[2], ctime[2];
	xvattr_t *xvap = (xvattr_t *)vap;	/* vap may be an xvattr_t * */
	xoptattr_t *xoap = NULL;
	boolean_t skipaclchk = (flags & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE;
	sa_bulk_attr_t bulk[2];
	int count = 0;

	ZFS_ENTER(zfsvfs);
	ZFS_VERIFY_ZP(zp);

	zfs_fuid_map_ids(zp, cr, &vap->va_uid, &vap->va_gid);

	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL, &mtime, 16);
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL, &ctime, 16);

	if ((error = sa_bulk_lookup(zp->z_sa_hdl, bulk, count)) != 0) {
		ZFS_EXIT(zfsvfs);
		return (error);
	}

	/*
	 * If ACL is trivial don't bother looking for ACE_READ_ATTRIBUTES.
	 * Also, if we are the owner don't bother, since owner should
	 * always be allowed to read basic attributes of file.
	 */
	if (!(zp->z_pflags & ZFS_ACL_TRIVIAL) &&
	    (vap->va_uid != crgetuid(cr))) {
		if (error = zfs_zaccess(zp, ACE_READ_ATTRIBUTES, 0,
		    skipaclchk, cr)) {
			ZFS_EXIT(zfsvfs);
			return (error);
		}
	}

	/*
	 * Return all attributes.  It's cheaper to provide the answer
	 * than to determine whether we were asked the question.
	 */

	mutex_enter(&zp->z_lock);
	vap->va_type = vp->v_type;
	vap->va_mode = zp->z_mode & MODEMASK;
	vap->va_fsid = zp->z_zfsvfs->z_vfs->vfs_dev;
	vap->va_nodeid = zp->z_id;
	if ((vp->v_flag & VROOT) && zfs_show_ctldir(zp))
		links = zp->z_links + 1;
	else
		links = zp->z_links;
	vap->va_nlink = MIN(links, UINT32_MAX);	/* nlink_t limit! */
	vap->va_size = zp->z_size;
	vap->va_rdev = vp->v_rdev;
	vap->va_seq = zp->z_seq;

	/*
	 * Add in any requested optional attributes and the create time.
	 * Also set the corresponding bits in the returned attribute bitmap.
	 */
	if ((xoap = xva_getxoptattr(xvap)) != NULL && zfsvfs->z_use_fuids) {
		if (XVA_ISSET_REQ(xvap, XAT_ARCHIVE)) {
			xoap->xoa_archive =
			    ((zp->z_pflags & ZFS_ARCHIVE) != 0);
			XVA_SET_RTN(xvap, XAT_ARCHIVE);
		}

		if (XVA_ISSET_REQ(xvap, XAT_READONLY)) {
			xoap->xoa_readonly =
			    ((zp->z_pflags & ZFS_READONLY) != 0);
			XVA_SET_RTN(xvap, XAT_READONLY);
		}

		if (XVA_ISSET_REQ(xvap, XAT_SYSTEM)) {
			xoap->xoa_system =
			    ((zp->z_pflags & ZFS_SYSTEM) != 0);
			XVA_SET_RTN(xvap, XAT_SYSTEM);
		}

		if (XVA_ISSET_REQ(xvap, XAT_HIDDEN)) {
			xoap->xoa_hidden =
			    ((zp->z_pflags & ZFS_HIDDEN) != 0);
			XVA_SET_RTN(xvap, XAT_HIDDEN);
		}

		if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK)) {
			xoap->xoa_nounlink =
			    ((zp->z_pflags & ZFS_NOUNLINK) != 0);
			XVA_SET_RTN(xvap, XAT_NOUNLINK);
		}

		if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE)) {
			xoap->xoa_immutable =
			    ((zp->z_pflags & ZFS_IMMUTABLE) != 0);
			XVA_SET_RTN(xvap, XAT_IMMUTABLE);
		}

		if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY)) {
			xoap->xoa_appendonly =
			    ((zp->z_pflags & ZFS_APPENDONLY) != 0);
			XVA_SET_RTN(xvap, XAT_APPENDONLY);
		}

		if (XVA_ISSET_REQ(xvap, XAT_NODUMP)) {
			xoap->xoa_nodump =
			    ((zp->z_pflags & ZFS_NODUMP) != 0);
			XVA_SET_RTN(xvap, XAT_NODUMP);
		}

		if (XVA_ISSET_REQ(xvap, XAT_OPAQUE)) {
			xoap->xoa_opaque =
			    ((zp->z_pflags & ZFS_OPAQUE) != 0);
			XVA_SET_RTN(xvap, XAT_OPAQUE);
		}

		if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED)) {
			xoap->xoa_av_quarantined =
			    ((zp->z_pflags & ZFS_AV_QUARANTINED) != 0);
			XVA_SET_RTN(xvap, XAT_AV_QUARANTINED);
		}

		if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED)) {
			xoap->xoa_av_modified =
			    ((zp->z_pflags & ZFS_AV_MODIFIED) != 0);
			XVA_SET_RTN(xvap, XAT_AV_MODIFIED);
		}

		if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP) &&
		    vp->v_type == VREG) {
			zfs_sa_get_scanstamp(zp, xvap);
		}

		if (XVA_ISSET_REQ(xvap, XAT_CREATETIME)) {
			uint64_t times[2];

			(void) sa_lookup(zp->z_sa_hdl, SA_ZPL_CRTIME(zfsvfs),
			    times, sizeof (times));
			ZFS_TIME_DECODE(&xoap->xoa_createtime, times);
			XVA_SET_RTN(xvap, XAT_CREATETIME);
		}

		if (XVA_ISSET_REQ(xvap, XAT_REPARSE)) {
			xoap->xoa_reparse = ((zp->z_pflags & ZFS_REPARSE) != 0);
			XVA_SET_RTN(xvap, XAT_REPARSE);
		}
		if (XVA_ISSET_REQ(xvap, XAT_GEN)) {
			xoap->xoa_generation = zp->z_gen;
			XVA_SET_RTN(xvap, XAT_GEN);
		}

		if (XVA_ISSET_REQ(xvap, XAT_OFFLINE)) {
			xoap->xoa_offline =
			    ((zp->z_pflags & ZFS_OFFLINE) != 0);
			XVA_SET_RTN(xvap, XAT_OFFLINE);
		}

		if (XVA_ISSET_REQ(xvap, XAT_SPARSE)) {
			xoap->xoa_sparse =
			    ((zp->z_pflags & ZFS_SPARSE) != 0);
			XVA_SET_RTN(xvap, XAT_SPARSE);
		}
	}

	ZFS_TIME_DECODE(&vap->va_atime, zp->z_atime);
	ZFS_TIME_DECODE(&vap->va_mtime, mtime);
	ZFS_TIME_DECODE(&vap->va_ctime, ctime);

	mutex_exit(&zp->z_lock);

	sa_object_size(zp->z_sa_hdl, &vap->va_blksize, &vap->va_nblocks);

	if (zp->z_blksz == 0) {
		/*
		 * Block size hasn't been set; suggest maximal I/O transfers.
		 */
		vap->va_blksize = zfsvfs->z_max_blksz;
	}

	ZFS_EXIT(zfsvfs);
	return (0);
}

/*
 * Set the file attributes to the values contained in the
 * vattr structure.
 *
 *	IN:	vp	- vnode of file to be modified.
 *		vap	- new attribute values.
 *			  If AT_XVATTR set, then optional attrs are being set
 *		flags	- ATTR_UTIME set if non-default time values provided.
 *			- ATTR_NOACLCHECK (CIFS context only).
 *		cr	- credentials of caller.
 *		ct	- caller context
 *
 *	RETURN:	0 on success, error code on failure.
 *
 * Timestamps:
 *	vp - ctime updated, mtime updated if size changed.
 */
/* ARGSUSED */
static int
zfs_setattr(vnode_t *vp, vattr_t *vap, int flags, cred_t *cr,
    caller_context_t *ct)
{
	znode_t		*zp = VTOZ(vp);
	zfsvfs_t	*zfsvfs = zp->z_zfsvfs;
	zilog_t		*zilog;
	dmu_tx_t	*tx;
	vattr_t		oldva;
	xvattr_t	tmpxvattr;
	uint_t		mask = vap->va_mask;
	uint_t		saved_mask = 0;
	int		trim_mask = 0;
	uint64_t	new_mode;
	uint64_t	new_uid, new_gid;
	uint64_t	xattr_obj;
	uint64_t	mtime[2], ctime[2];
	znode_t		*attrzp;
	int		need_policy = FALSE;
	int		err, err2;
	zfs_fuid_info_t *fuidp = NULL;
	xvattr_t *xvap = (xvattr_t *)vap;	/* vap may be an xvattr_t * */
	xoptattr_t	*xoap;
	zfs_acl_t	*aclp;
	boolean_t skipaclchk = (flags & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE;
	boolean_t	fuid_dirtied = B_FALSE;
	sa_bulk_attr_t	bulk[7], xattr_bulk[7];
	int		count = 0, xattr_count = 0;

	if (mask == 0)
		return (0);

	if (mask & AT_NOSET)
		return (SET_ERROR(EINVAL));

	ZFS_ENTER(zfsvfs);
	ZFS_VERIFY_ZP(zp);

	zilog = zfsvfs->z_log;

	/*
	 * Make sure that if we have ephemeral uid/gid or xvattr specified
	 * that file system is at proper version level
	 */

	if (zfsvfs->z_use_fuids == B_FALSE &&
	    (((mask & AT_UID) && IS_EPHEMERAL(vap->va_uid)) ||
	    ((mask & AT_GID) && IS_EPHEMERAL(vap->va_gid)) ||
	    (mask & AT_XVATTR))) {
		ZFS_EXIT(zfsvfs);
		return (SET_ERROR(EINVAL));
	}

	if (mask & AT_SIZE && vp->v_type == VDIR) {
		ZFS_EXIT(zfsvfs);
		return (SET_ERROR(EISDIR));
	}

	if (mask & AT_SIZE && vp->v_type != VREG && vp->v_type != VFIFO) {
		ZFS_EXIT(zfsvfs);
		return (SET_ERROR(EINVAL));
	}

	/*
	 * If this is an xvattr_t, then get a pointer to the structure of
	 * optional attributes.  If this is NULL, then we have a vattr_t.
	 */
	xoap = xva_getxoptattr(xvap);

	xva_init(&tmpxvattr);

	/*
	 * Immutable files can only alter immutable bit and atime
	 */
	if ((zp->z_pflags & ZFS_IMMUTABLE) &&
	    ((mask & (AT_SIZE|AT_UID|AT_GID|AT_MTIME|AT_MODE)) ||
	    ((mask & AT_XVATTR) && XVA_ISSET_REQ(xvap, XAT_CREATETIME)))) {
		ZFS_EXIT(zfsvfs);
		return (SET_ERROR(EPERM));
	}

	/*
	 * Note: ZFS_READONLY is handled in zfs_zaccess_common.
	 */

	/*
	 * Verify timestamps doesn't overflow 32 bits.
	 * ZFS can handle large timestamps, but 32bit syscalls can't
	 * handle times greater than 2039.  This check should be removed
	 * once large timestamps are fully supported.
	 */
	if (mask & (AT_ATIME | AT_MTIME)) {
		if (((mask & AT_ATIME) && TIMESPEC_OVERFLOW(&vap->va_atime)) ||
		    ((mask & AT_MTIME) && TIMESPEC_OVERFLOW(&vap->va_mtime))) {
			ZFS_EXIT(zfsvfs);
			return (SET_ERROR(EOVERFLOW));
		}
	}

top:
	attrzp = NULL;
	aclp = NULL;

	/* Can this be moved to before the top label? */
	if (zfsvfs->z_vfs->vfs_flag & VFS_RDONLY) {
		ZFS_EXIT(zfsvfs);
		return (SET_ERROR(EROFS));
	}