Home | History | Annotate | Line # | Download | only in ffs
ffs_vnops.c revision 1.112
      1 /*	$NetBSD: ffs_vnops.c,v 1.112 2009/03/29 10:29:00 ad Exp $	*/
      2 
      3 /*-
      4  * Copyright (c) 2008, 2009 The NetBSD Foundation, Inc.
      5  * All rights reserved.
      6  *
      7  * This code is derived from software contributed to The NetBSD Foundation
      8  * by Wasabi Systems, Inc, and by Andrew Doran.
      9  *
     10  * Redistribution and use in source and binary forms, with or without
     11  * modification, are permitted provided that the following conditions
     12  * are met:
     13  * 1. Redistributions of source code must retain the above copyright
     14  *    notice, this list of conditions and the following disclaimer.
     15  * 2. Redistributions in binary form must reproduce the above copyright
     16  *    notice, this list of conditions and the following disclaimer in the
     17  *    documentation and/or other materials provided with the distribution.
     18  *
     19  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
     20  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
     21  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
     22  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
     23  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
     24  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
     25  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
     26  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
     27  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
     28  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
     29  * POSSIBILITY OF SUCH DAMAGE.
     30  */
     31 
     32 /*
     33  * Copyright (c) 1982, 1986, 1989, 1993
     34  *	The Regents of the University of California.  All rights reserved.
     35  *
     36  * Redistribution and use in source and binary forms, with or without
     37  * modification, are permitted provided that the following conditions
     38  * are met:
     39  * 1. Redistributions of source code must retain the above copyright
     40  *    notice, this list of conditions and the following disclaimer.
     41  * 2. Redistributions in binary form must reproduce the above copyright
     42  *    notice, this list of conditions and the following disclaimer in the
     43  *    documentation and/or other materials provided with the distribution.
     44  * 3. Neither the name of the University nor the names of its contributors
     45  *    may be used to endorse or promote products derived from this software
     46  *    without specific prior written permission.
     47  *
     48  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
     49  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
     50  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
     51  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
     52  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
     53  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
     54  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
     55  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
     56  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
     57  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
     58  * SUCH DAMAGE.
     59  *
     60  *	@(#)ffs_vnops.c	8.15 (Berkeley) 5/14/95
     61  */
     62 
     63 #include <sys/cdefs.h>
     64 __KERNEL_RCSID(0, "$NetBSD: ffs_vnops.c,v 1.112 2009/03/29 10:29:00 ad Exp $");
     65 
     66 #if defined(_KERNEL_OPT)
     67 #include "opt_ffs.h"
     68 #include "opt_wapbl.h"
     69 #endif
     70 
     71 #include <sys/param.h>
     72 #include <sys/systm.h>
     73 #include <sys/resourcevar.h>
     74 #include <sys/kernel.h>
     75 #include <sys/file.h>
     76 #include <sys/stat.h>
     77 #include <sys/buf.h>
     78 #include <sys/event.h>
     79 #include <sys/proc.h>
     80 #include <sys/mount.h>
     81 #include <sys/vnode.h>
     82 #include <sys/pool.h>
     83 #include <sys/signalvar.h>
     84 #include <sys/kauth.h>
     85 #include <sys/wapbl.h>
     86 #include <sys/fstrans.h>
     87 
     88 #include <miscfs/fifofs/fifo.h>
     89 #include <miscfs/genfs/genfs.h>
     90 #include <miscfs/specfs/specdev.h>
     91 
     92 #include <ufs/ufs/inode.h>
     93 #include <ufs/ufs/dir.h>
     94 #include <ufs/ufs/ufs_extern.h>
     95 #include <ufs/ufs/ufsmount.h>
     96 #include <ufs/ufs/ufs_wapbl.h>
     97 
     98 #include <ufs/ffs/fs.h>
     99 #include <ufs/ffs/ffs_extern.h>
    100 
    101 #include <uvm/uvm.h>
    102 
    103 /* Global vfs data structures for ufs. */
    104 int (**ffs_vnodeop_p)(void *);
    105 const struct vnodeopv_entry_desc ffs_vnodeop_entries[] = {
    106 	{ &vop_default_desc, vn_default_error },
    107 	{ &vop_lookup_desc, ufs_lookup },		/* lookup */
    108 	{ &vop_create_desc, ufs_create },		/* create */
    109 	{ &vop_whiteout_desc, ufs_whiteout },		/* whiteout */
    110 	{ &vop_mknod_desc, ufs_mknod },			/* mknod */
    111 	{ &vop_open_desc, ufs_open },			/* open */
    112 	{ &vop_close_desc, ufs_close },			/* close */
    113 	{ &vop_access_desc, ufs_access },		/* access */
    114 	{ &vop_getattr_desc, ufs_getattr },		/* getattr */
    115 	{ &vop_setattr_desc, ufs_setattr },		/* setattr */
    116 	{ &vop_read_desc, ffs_read },			/* read */
    117 	{ &vop_write_desc, ffs_write },			/* write */
    118 	{ &vop_ioctl_desc, ufs_ioctl },			/* ioctl */
    119 	{ &vop_fcntl_desc, ufs_fcntl },			/* fcntl */
    120 	{ &vop_poll_desc, ufs_poll },			/* poll */
    121 	{ &vop_kqfilter_desc, genfs_kqfilter },		/* kqfilter */
    122 	{ &vop_revoke_desc, ufs_revoke },		/* revoke */
    123 	{ &vop_mmap_desc, ufs_mmap },			/* mmap */
    124 	{ &vop_fsync_desc, ffs_fsync },			/* fsync */
    125 	{ &vop_seek_desc, ufs_seek },			/* seek */
    126 	{ &vop_remove_desc, ufs_remove },		/* remove */
    127 	{ &vop_link_desc, ufs_link },			/* link */
    128 	{ &vop_rename_desc, ufs_rename },		/* rename */
    129 	{ &vop_mkdir_desc, ufs_mkdir },			/* mkdir */
    130 	{ &vop_rmdir_desc, ufs_rmdir },			/* rmdir */
    131 	{ &vop_symlink_desc, ufs_symlink },		/* symlink */
    132 	{ &vop_readdir_desc, ufs_readdir },		/* readdir */
    133 	{ &vop_readlink_desc, ufs_readlink },		/* readlink */
    134 	{ &vop_abortop_desc, ufs_abortop },		/* abortop */
    135 	{ &vop_inactive_desc, ufs_inactive },		/* inactive */
    136 	{ &vop_reclaim_desc, ffs_reclaim },		/* reclaim */
    137 	{ &vop_lock_desc, ffs_lock },			/* lock */
    138 	{ &vop_unlock_desc, ffs_unlock },		/* unlock */
    139 	{ &vop_bmap_desc, ufs_bmap },			/* bmap */
    140 	{ &vop_strategy_desc, ufs_strategy },		/* strategy */
    141 	{ &vop_print_desc, ufs_print },			/* print */
    142 	{ &vop_islocked_desc, ffs_islocked },		/* islocked */
    143 	{ &vop_pathconf_desc, ufs_pathconf },		/* pathconf */
    144 	{ &vop_advlock_desc, ufs_advlock },		/* advlock */
    145 	{ &vop_bwrite_desc, vn_bwrite },		/* bwrite */
    146 	{ &vop_getpages_desc, genfs_getpages },		/* getpages */
    147 	{ &vop_putpages_desc, genfs_putpages },		/* putpages */
    148 	{ &vop_openextattr_desc, ffs_openextattr },	/* openextattr */
    149 	{ &vop_closeextattr_desc, ffs_closeextattr },	/* closeextattr */
    150 	{ &vop_getextattr_desc, ffs_getextattr },	/* getextattr */
    151 	{ &vop_setextattr_desc, ffs_setextattr },	/* setextattr */
    152 	{ &vop_listextattr_desc, ffs_listextattr },	/* listextattr */
    153 	{ &vop_deleteextattr_desc, ffs_deleteextattr },	/* deleteextattr */
    154 	{ NULL, NULL }
    155 };
    156 const struct vnodeopv_desc ffs_vnodeop_opv_desc =
    157 	{ &ffs_vnodeop_p, ffs_vnodeop_entries };
    158 
    159 int (**ffs_specop_p)(void *);
    160 const struct vnodeopv_entry_desc ffs_specop_entries[] = {
    161 	{ &vop_default_desc, vn_default_error },
    162 	{ &vop_lookup_desc, spec_lookup },		/* lookup */
    163 	{ &vop_create_desc, spec_create },		/* create */
    164 	{ &vop_mknod_desc, spec_mknod },		/* mknod */
    165 	{ &vop_open_desc, spec_open },			/* open */
    166 	{ &vop_close_desc, ufsspec_close },		/* close */
    167 	{ &vop_access_desc, ufs_access },		/* access */
    168 	{ &vop_getattr_desc, ufs_getattr },		/* getattr */
    169 	{ &vop_setattr_desc, ufs_setattr },		/* setattr */
    170 	{ &vop_read_desc, ufsspec_read },		/* read */
    171 	{ &vop_write_desc, ufsspec_write },		/* write */
    172 	{ &vop_ioctl_desc, spec_ioctl },		/* ioctl */
    173 	{ &vop_fcntl_desc, ufs_fcntl },			/* fcntl */
    174 	{ &vop_poll_desc, spec_poll },			/* poll */
    175 	{ &vop_kqfilter_desc, spec_kqfilter },		/* kqfilter */
    176 	{ &vop_revoke_desc, spec_revoke },		/* revoke */
    177 	{ &vop_mmap_desc, spec_mmap },			/* mmap */
    178 	{ &vop_fsync_desc, ffs_fsync },			/* fsync */
    179 	{ &vop_seek_desc, spec_seek },			/* seek */
    180 	{ &vop_remove_desc, spec_remove },		/* remove */
    181 	{ &vop_link_desc, spec_link },			/* link */
    182 	{ &vop_rename_desc, spec_rename },		/* rename */
    183 	{ &vop_mkdir_desc, spec_mkdir },		/* mkdir */
    184 	{ &vop_rmdir_desc, spec_rmdir },		/* rmdir */
    185 	{ &vop_symlink_desc, spec_symlink },		/* symlink */
    186 	{ &vop_readdir_desc, spec_readdir },		/* readdir */
    187 	{ &vop_readlink_desc, spec_readlink },		/* readlink */
    188 	{ &vop_abortop_desc, spec_abortop },		/* abortop */
    189 	{ &vop_inactive_desc, ufs_inactive },		/* inactive */
    190 	{ &vop_reclaim_desc, ffs_reclaim },		/* reclaim */
    191 	{ &vop_lock_desc, ffs_lock },			/* lock */
    192 	{ &vop_unlock_desc, ffs_unlock },		/* unlock */
    193 	{ &vop_bmap_desc, spec_bmap },			/* bmap */
    194 	{ &vop_strategy_desc, spec_strategy },		/* strategy */
    195 	{ &vop_print_desc, ufs_print },			/* print */
    196 	{ &vop_islocked_desc, ffs_islocked },		/* islocked */
    197 	{ &vop_pathconf_desc, spec_pathconf },		/* pathconf */
    198 	{ &vop_advlock_desc, spec_advlock },		/* advlock */
    199 	{ &vop_bwrite_desc, vn_bwrite },		/* bwrite */
    200 	{ &vop_getpages_desc, spec_getpages },		/* getpages */
    201 	{ &vop_putpages_desc, spec_putpages },		/* putpages */
    202 	{ &vop_openextattr_desc, ffs_openextattr },	/* openextattr */
    203 	{ &vop_closeextattr_desc, ffs_closeextattr },	/* closeextattr */
    204 	{ &vop_getextattr_desc, ffs_getextattr },	/* getextattr */
    205 	{ &vop_setextattr_desc, ffs_setextattr },	/* setextattr */
    206 	{ &vop_listextattr_desc, ffs_listextattr },	/* listextattr */
    207 	{ &vop_deleteextattr_desc, ffs_deleteextattr },	/* deleteextattr */
    208 	{ NULL, NULL }
    209 };
    210 const struct vnodeopv_desc ffs_specop_opv_desc =
    211 	{ &ffs_specop_p, ffs_specop_entries };
    212 
    213 int (**ffs_fifoop_p)(void *);
    214 const struct vnodeopv_entry_desc ffs_fifoop_entries[] = {
    215 	{ &vop_default_desc, vn_default_error },
    216 	{ &vop_lookup_desc, fifo_lookup },		/* lookup */
    217 	{ &vop_create_desc, fifo_create },		/* create */
    218 	{ &vop_mknod_desc, fifo_mknod },		/* mknod */
    219 	{ &vop_open_desc, fifo_open },			/* open */
    220 	{ &vop_close_desc, ufsfifo_close },		/* close */
    221 	{ &vop_access_desc, ufs_access },		/* access */
    222 	{ &vop_getattr_desc, ufs_getattr },		/* getattr */
    223 	{ &vop_setattr_desc, ufs_setattr },		/* setattr */
    224 	{ &vop_read_desc, ufsfifo_read },		/* read */
    225 	{ &vop_write_desc, ufsfifo_write },		/* write */
    226 	{ &vop_ioctl_desc, fifo_ioctl },		/* ioctl */
    227 	{ &vop_fcntl_desc, ufs_fcntl },			/* fcntl */
    228 	{ &vop_poll_desc, fifo_poll },			/* poll */
    229 	{ &vop_kqfilter_desc, fifo_kqfilter },		/* kqfilter */
    230 	{ &vop_revoke_desc, fifo_revoke },		/* revoke */
    231 	{ &vop_mmap_desc, fifo_mmap },			/* mmap */
    232 	{ &vop_fsync_desc, ffs_fsync },			/* fsync */
    233 	{ &vop_seek_desc, fifo_seek },			/* seek */
    234 	{ &vop_remove_desc, fifo_remove },		/* remove */
    235 	{ &vop_link_desc, fifo_link },			/* link */
    236 	{ &vop_rename_desc, fifo_rename },		/* rename */
    237 	{ &vop_mkdir_desc, fifo_mkdir },		/* mkdir */
    238 	{ &vop_rmdir_desc, fifo_rmdir },		/* rmdir */
    239 	{ &vop_symlink_desc, fifo_symlink },		/* symlink */
    240 	{ &vop_readdir_desc, fifo_readdir },		/* readdir */
    241 	{ &vop_readlink_desc, fifo_readlink },		/* readlink */
    242 	{ &vop_abortop_desc, fifo_abortop },		/* abortop */
    243 	{ &vop_inactive_desc, ufs_inactive },		/* inactive */
    244 	{ &vop_reclaim_desc, ffs_reclaim },		/* reclaim */
    245 	{ &vop_lock_desc, ffs_lock },			/* lock */
    246 	{ &vop_unlock_desc, ffs_unlock },		/* unlock */
    247 	{ &vop_bmap_desc, fifo_bmap },			/* bmap */
    248 	{ &vop_strategy_desc, fifo_strategy },		/* strategy */
    249 	{ &vop_print_desc, ufs_print },			/* print */
    250 	{ &vop_islocked_desc, ffs_islocked },		/* islocked */
    251 	{ &vop_pathconf_desc, fifo_pathconf },		/* pathconf */
    252 	{ &vop_advlock_desc, fifo_advlock },		/* advlock */
    253 	{ &vop_bwrite_desc, vn_bwrite },		/* bwrite */
    254 	{ &vop_putpages_desc, fifo_putpages }, 		/* putpages */
    255 	{ &vop_openextattr_desc, ffs_openextattr },	/* openextattr */
    256 	{ &vop_closeextattr_desc, ffs_closeextattr },	/* closeextattr */
    257 	{ &vop_getextattr_desc, ffs_getextattr },	/* getextattr */
    258 	{ &vop_setextattr_desc, ffs_setextattr },	/* setextattr */
    259 	{ &vop_listextattr_desc, ffs_listextattr },	/* listextattr */
    260 	{ &vop_deleteextattr_desc, ffs_deleteextattr },	/* deleteextattr */
    261 	{ NULL, NULL }
    262 };
    263 const struct vnodeopv_desc ffs_fifoop_opv_desc =
    264 	{ &ffs_fifoop_p, ffs_fifoop_entries };
    265 
    266 #include <ufs/ufs/ufs_readwrite.c>
    267 
    268 int
    269 ffs_fsync(void *v)
    270 {
    271 	struct vop_fsync_args /* {
    272 		struct vnode *a_vp;
    273 		kauth_cred_t a_cred;
    274 		int a_flags;
    275 		off_t a_offlo;
    276 		off_t a_offhi;
    277 		struct lwp *a_l;
    278 	} */ *ap = v;
    279 	struct buf *bp;
    280 	int num, error, i;
    281 	struct indir ia[NIADDR + 1];
    282 	int bsize;
    283 	daddr_t blk_high;
    284 	struct vnode *vp;
    285 #ifdef WAPBL
    286 	struct mount *mp;
    287 #endif
    288 
    289 	vp = ap->a_vp;
    290 
    291 	fstrans_start(vp->v_mount, FSTRANS_LAZY);
    292 	if ((ap->a_offlo == 0 && ap->a_offhi == 0) || (vp->v_type != VREG)) {
    293 		error = ffs_full_fsync(vp, ap->a_flags);
    294 		goto out;
    295 	}
    296 
    297 	bsize = vp->v_mount->mnt_stat.f_iosize;
    298 	blk_high = ap->a_offhi / bsize;
    299 	if (ap->a_offhi % bsize != 0)
    300 		blk_high++;
    301 
    302 	/*
    303 	 * First, flush all pages in range.
    304 	 */
    305 
    306 	mutex_enter(&vp->v_interlock);
    307 	error = VOP_PUTPAGES(vp, trunc_page(ap->a_offlo),
    308 	    round_page(ap->a_offhi), PGO_CLEANIT |
    309 	    ((ap->a_flags & FSYNC_WAIT) ? PGO_SYNCIO : 0));
    310 	if (error) {
    311 		goto out;
    312 	}
    313 
    314 #ifdef WAPBL
    315 	mp = wapbl_vptomp(vp);
    316 	if (mp->mnt_wapbl) {
    317 		/*
    318 		 * Don't bother writing out metadata if the syncer is
    319 		 * making the request.  We will let the sync vnode
    320 		 * write it out in a single burst through a call to
    321 		 * VFS_SYNC().
    322 		 */
    323 		if ((ap->a_flags & (FSYNC_DATAONLY | FSYNC_LAZY)) != 0) {
    324 			fstrans_done(vp->v_mount);
    325 			return 0;
    326 		}
    327 		error = 0;
    328 		if (vp->v_tag == VT_UFS && VTOI(vp)->i_flag &
    329 		    (IN_ACCESS | IN_CHANGE | IN_UPDATE | IN_MODIFY |
    330 				 IN_MODIFIED | IN_ACCESSED)) {
    331 			error = UFS_WAPBL_BEGIN(mp);
    332 			if (error) {
    333 				fstrans_done(vp->v_mount);
    334 				return error;
    335 			}
    336 			error = ffs_update(vp, NULL, NULL, UPDATE_CLOSE |
    337 			    ((ap->a_flags & FSYNC_WAIT) ? UPDATE_WAIT : 0));
    338 			UFS_WAPBL_END(mp);
    339 		}
    340 		if (error || (ap->a_flags & FSYNC_NOLOG) != 0) {
    341 			fstrans_done(vp->v_mount);
    342 			return error;
    343 		}
    344 		error = wapbl_flush(mp->mnt_wapbl, 0);
    345 		fstrans_done(vp->v_mount);
    346 		return error;
    347 	}
    348 #endif /* WAPBL */
    349 
    350 	/*
    351 	 * Then, flush indirect blocks.
    352 	 */
    353 
    354 	if (blk_high >= NDADDR) {
    355 		error = ufs_getlbns(vp, blk_high, ia, &num);
    356 		if (error)
    357 			goto out;
    358 
    359 		mutex_enter(&bufcache_lock);
    360 		for (i = 0; i < num; i++) {
    361 			if ((bp = incore(vp, ia[i].in_lbn)) == NULL)
    362 				continue;
    363 			if ((bp->b_cflags & BC_BUSY) != 0 ||
    364 			    (bp->b_oflags & BO_DELWRI) == 0)
    365 				continue;
    366 			bp->b_cflags |= BC_BUSY | BC_VFLUSH;
    367 			mutex_exit(&bufcache_lock);
    368 			bawrite(bp);
    369 			mutex_enter(&bufcache_lock);
    370 		}
    371 		mutex_exit(&bufcache_lock);
    372 	}
    373 
    374 	if (ap->a_flags & FSYNC_WAIT) {
    375 		mutex_enter(&vp->v_interlock);
    376 		while (vp->v_numoutput > 0)
    377 			cv_wait(&vp->v_cv, &vp->v_interlock);
    378 		mutex_exit(&vp->v_interlock);
    379 	}
    380 
    381 	error = ffs_update(vp, NULL, NULL, UPDATE_CLOSE |
    382 	    (((ap->a_flags & (FSYNC_WAIT | FSYNC_DATAONLY)) == FSYNC_WAIT)
    383 	    ? UPDATE_WAIT : 0));
    384 
    385 	if (error == 0 && ap->a_flags & FSYNC_CACHE) {
    386 		int l = 0;
    387 		VOP_IOCTL(VTOI(vp)->i_devvp, DIOCCACHESYNC, &l, FWRITE,
    388 			curlwp->l_cred);
    389 	}
    390 
    391 out:
    392 	fstrans_done(vp->v_mount);
    393 	return error;
    394 }
    395 
    396 /*
    397  * Synch an open file.  Called for VOP_FSYNC().
    398  */
    399 /* ARGSUSED */
    400 int
    401 ffs_full_fsync(struct vnode *vp, int flags)
    402 {
    403 	struct buf *bp, *nbp;
    404 	int error, passes, skipmeta, waitfor, i;
    405 	struct mount *mp;
    406 
    407 	KASSERT(VTOI(vp) != NULL);
    408 	KASSERT(vp->v_tag == VT_UFS);
    409 
    410 	error = 0;
    411 
    412 	mp = vp->v_mount;
    413 	if (vp->v_type == VBLK && vp->v_specmountpoint != NULL) {
    414 		mp = vp->v_specmountpoint;
    415 	} else {
    416 		mp = vp->v_mount;
    417 	}
    418 
    419 	/*
    420 	 * Flush all dirty data associated with the vnode.
    421 	 */
    422 	if (vp->v_type == VREG || vp->v_type == VBLK) {
    423 		int pflags = PGO_ALLPAGES | PGO_CLEANIT;
    424 
    425 		if ((flags & FSYNC_WAIT))
    426 			pflags |= PGO_SYNCIO;
    427 		if (vp->v_type == VREG &&
    428 		    fstrans_getstate(mp) == FSTRANS_SUSPENDING)
    429 			pflags |= PGO_FREE;
    430 		mutex_enter(&vp->v_interlock);
    431 		error = VOP_PUTPAGES(vp, 0, 0, pflags);
    432 		if (error)
    433 			return error;
    434 	}
    435 
    436 #ifdef WAPBL
    437 	mp = wapbl_vptomp(vp);
    438 	if (mp && mp->mnt_wapbl) {
    439 		/*
    440 		 * Don't bother writing out metadata if the syncer is
    441 		 * making the request.  We will let the sync vnode
    442 		 * write it out in a single burst through a call to
    443 		 * VFS_SYNC().
    444 		 */
    445 		if ((flags & (FSYNC_DATAONLY | FSYNC_LAZY)) != 0)
    446 			return 0;
    447 
    448 		if ((VTOI(vp)->i_flag & (IN_ACCESS | IN_CHANGE | IN_UPDATE
    449 		    | IN_MODIFY | IN_MODIFIED | IN_ACCESSED)) != 0) {
    450 			error = UFS_WAPBL_BEGIN(mp);
    451 			if (error)
    452 				return error;
    453 			error = ffs_update(vp, NULL, NULL, UPDATE_CLOSE |
    454 			    ((flags & FSYNC_WAIT) ? UPDATE_WAIT : 0));
    455 			UFS_WAPBL_END(mp);
    456 		}
    457 		if (error || (flags & FSYNC_NOLOG) != 0)
    458 			return error;
    459 
    460 		/*
    461 		 * Don't flush the log if the vnode being flushed
    462 		 * contains no dirty buffers that could be in the log.
    463 		 */
    464 		if (!LIST_EMPTY(&vp->v_dirtyblkhd)) {
    465 			error = wapbl_flush(mp->mnt_wapbl, 0);
    466 			if (error)
    467 				return error;
    468 		}
    469 
    470 		if ((flags & FSYNC_WAIT) != 0) {
    471 			mutex_enter(&vp->v_interlock);
    472 			while (vp->v_numoutput != 0)
    473 				cv_wait(&vp->v_cv, &vp->v_interlock);
    474 			mutex_exit(&vp->v_interlock);
    475 		}
    476 
    477 		return error;
    478 	}
    479 #endif /* WAPBL */
    480 
    481 	/*
    482 	 * Write out metadata for non-logging file systems. XXX This block
    483 	 * should be simplified now that softdep is gone.
    484 	 */
    485 	passes = NIADDR + 1;
    486 	skipmeta = 0;
    487 	if (flags & FSYNC_WAIT)
    488 		skipmeta = 1;
    489 
    490 loop:
    491 	mutex_enter(&bufcache_lock);
    492 	LIST_FOREACH(bp, &vp->v_dirtyblkhd, b_vnbufs) {
    493 		bp->b_cflags &= ~BC_SCANNED;
    494 	}
    495 	for (bp = LIST_FIRST(&vp->v_dirtyblkhd); bp; bp = nbp) {
    496 		nbp = LIST_NEXT(bp, b_vnbufs);
    497 		if (bp->b_cflags & (BC_BUSY | BC_SCANNED))
    498 			continue;
    499 		if ((bp->b_oflags & BO_DELWRI) == 0)
    500 			panic("ffs_fsync: not dirty");
    501 		if (skipmeta && bp->b_lblkno < 0)
    502 			continue;
    503 		bp->b_cflags |= BC_BUSY | BC_VFLUSH | BC_SCANNED;
    504 		mutex_exit(&bufcache_lock);
    505 		/*
    506 		 * On our final pass through, do all I/O synchronously
    507 		 * so that we can find out if our flush is failing
    508 		 * because of write errors.
    509 		 */
    510 		if (passes > 0 || !(flags & FSYNC_WAIT))
    511 			(void) bawrite(bp);
    512 		else if ((error = bwrite(bp)) != 0)
    513 			return (error);
    514 		/*
    515 		 * Since we unlocked during the I/O, we need
    516 		 * to start from a known point.
    517 		 */
    518 		mutex_enter(&bufcache_lock);
    519 		nbp = LIST_FIRST(&vp->v_dirtyblkhd);
    520 	}
    521 	mutex_exit(&bufcache_lock);
    522 	if (skipmeta) {
    523 		skipmeta = 0;
    524 		goto loop;
    525 	}
    526 
    527 	if ((flags & FSYNC_WAIT) != 0) {
    528 		mutex_enter(&vp->v_interlock);
    529 		while (vp->v_numoutput) {
    530 			cv_wait(&vp->v_cv, &vp->v_interlock);
    531 		}
    532 		mutex_exit(&vp->v_interlock);
    533 
    534 		/*
    535 		 * Ensure that any filesystem metadata associated
    536 		 * with the vnode has been written.
    537 		 */
    538 		if (!LIST_EMPTY(&vp->v_dirtyblkhd)) {
    539 			/*
    540 			* Block devices associated with filesystems may
    541 			* have new I/O requests posted for them even if
    542 			* the vnode is locked, so no amount of trying will
    543 			* get them clean. Thus we give block devices a
    544 			* good effort, then just give up. For all other file
    545 			* types, go around and try again until it is clean.
    546 			*/
    547 			if (passes > 0) {
    548 				passes--;
    549 				goto loop;
    550 			}
    551 #ifdef DIAGNOSTIC
    552 			if (vp->v_type != VBLK)
    553 				vprint("ffs_fsync: dirty", vp);
    554 #endif
    555 		}
    556 	}
    557 
    558 	waitfor = (flags & FSYNC_WAIT) ? UPDATE_WAIT : 0;
    559 	error = ffs_update(vp, NULL, NULL, UPDATE_CLOSE | waitfor);
    560 
    561 	if (error == 0 && (flags & FSYNC_CACHE) != 0) {
    562 		(void)VOP_IOCTL(VTOI(vp)->i_devvp, DIOCCACHESYNC, &i, FWRITE,
    563 		    kauth_cred_get());
    564 	}
    565 
    566 	return error;
    567 }
    568 
    569 /*
    570  * Reclaim an inode so that it can be used for other purposes.
    571  */
    572 int
    573 ffs_reclaim(void *v)
    574 {
    575 	struct vop_reclaim_args /* {
    576 		struct vnode *a_vp;
    577 		struct lwp *a_l;
    578 	} */ *ap = v;
    579 	struct vnode *vp = ap->a_vp;
    580 	struct inode *ip = VTOI(vp);
    581 	struct mount *mp = vp->v_mount;
    582 	struct ufsmount *ump = ip->i_ump;
    583 	void *data;
    584 	int error;
    585 
    586 	fstrans_start(mp, FSTRANS_LAZY);
    587 	if ((error = ufs_reclaim(vp)) != 0) {
    588 		fstrans_done(mp);
    589 		return (error);
    590 	}
    591 	if (ip->i_din.ffs1_din != NULL) {
    592 		if (ump->um_fstype == UFS1)
    593 			pool_cache_put(ffs_dinode1_cache, ip->i_din.ffs1_din);
    594 		else
    595 			pool_cache_put(ffs_dinode2_cache, ip->i_din.ffs2_din);
    596 	}
    597 	/*
    598 	 * To interlock with ffs_sync().
    599 	 */
    600 	genfs_node_destroy(vp);
    601 	mutex_enter(&vp->v_interlock);
    602 	data = vp->v_data;
    603 	vp->v_data = NULL;
    604 	mutex_exit(&vp->v_interlock);
    605 
    606 	/*
    607 	 * XXX MFS ends up here, too, to free an inode.  Should we create
    608 	 * XXX a separate pool for MFS inodes?
    609 	 */
    610 	pool_cache_put(ffs_inode_cache, data);
    611 	fstrans_done(mp);
    612 	return (0);
    613 }
    614 
    615 /*
    616  * Return the last logical file offset that should be written for this file
    617  * if we're doing a write that ends at "size".
    618  */
    619 
    620 void
    621 ffs_gop_size(struct vnode *vp, off_t size, off_t *eobp, int flags)
    622 {
    623 	struct inode *ip = VTOI(vp);
    624 	struct fs *fs = ip->i_fs;
    625 	daddr_t olbn, nlbn;
    626 
    627 	olbn = lblkno(fs, ip->i_size);
    628 	nlbn = lblkno(fs, size);
    629 	if (nlbn < NDADDR && olbn <= nlbn) {
    630 		*eobp = fragroundup(fs, size);
    631 	} else {
    632 		*eobp = blkroundup(fs, size);
    633 	}
    634 }
    635 
    636 int
    637 ffs_openextattr(void *v)
    638 {
    639 	struct vop_openextattr_args /* {
    640 		struct vnode *a_vp;
    641 		kauth_cred_t a_cred;
    642 		struct proc *a_p;
    643 	} */ *ap = v;
    644 	struct inode *ip = VTOI(ap->a_vp);
    645 	struct fs *fs = ip->i_fs;
    646 
    647 	/* Not supported for UFS1 file systems. */
    648 	if (fs->fs_magic == FS_UFS1_MAGIC)
    649 		return (EOPNOTSUPP);
    650 
    651 	/* XXX Not implemented for UFS2 file systems. */
    652 	return (EOPNOTSUPP);
    653 }
    654 
    655 int
    656 ffs_closeextattr(void *v)
    657 {
    658 	struct vop_closeextattr_args /* {
    659 		struct vnode *a_vp;
    660 		int a_commit;
    661 		kauth_cred_t a_cred;
    662 		struct proc *a_p;
    663 	} */ *ap = v;
    664 	struct inode *ip = VTOI(ap->a_vp);
    665 	struct fs *fs = ip->i_fs;
    666 
    667 	/* Not supported for UFS1 file systems. */
    668 	if (fs->fs_magic == FS_UFS1_MAGIC)
    669 		return (EOPNOTSUPP);
    670 
    671 	/* XXX Not implemented for UFS2 file systems. */
    672 	return (EOPNOTSUPP);
    673 }
    674 
    675 int
    676 ffs_getextattr(void *v)
    677 {
    678 	struct vop_getextattr_args /* {
    679 		struct vnode *a_vp;
    680 		int a_attrnamespace;
    681 		const char *a_name;
    682 		struct uio *a_uio;
    683 		size_t *a_size;
    684 		kauth_cred_t a_cred;
    685 		struct proc *a_p;
    686 	} */ *ap = v;
    687 	struct vnode *vp = ap->a_vp;
    688 	struct inode *ip = VTOI(vp);
    689 	struct fs *fs = ip->i_fs;
    690 
    691 	if (fs->fs_magic == FS_UFS1_MAGIC) {
    692 #ifdef UFS_EXTATTR
    693 		int error;
    694 
    695 		fstrans_start(vp->v_mount, FSTRANS_SHARED);
    696 		error = ufs_getextattr(ap);
    697 		fstrans_done(vp->v_mount);
    698 		return error;
    699 #else
    700 		return (EOPNOTSUPP);
    701 #endif
    702 	}
    703 
    704 	/* XXX Not implemented for UFS2 file systems. */
    705 	return (EOPNOTSUPP);
    706 }
    707 
    708 int
    709 ffs_setextattr(void *v)
    710 {
    711 	struct vop_setextattr_args /* {
    712 		struct vnode *a_vp;
    713 		int a_attrnamespace;
    714 		const char *a_name;
    715 		struct uio *a_uio;
    716 		kauth_cred_t a_cred;
    717 		struct proc *a_p;
    718 	} */ *ap = v;
    719 	struct vnode *vp = ap->a_vp;
    720 	struct inode *ip = VTOI(vp);
    721 	struct fs *fs = ip->i_fs;
    722 
    723 	if (fs->fs_magic == FS_UFS1_MAGIC) {
    724 #ifdef UFS_EXTATTR
    725 		int error;
    726 
    727 		fstrans_start(vp->v_mount, FSTRANS_SHARED);
    728 		error = ufs_setextattr(ap);
    729 		fstrans_done(vp->v_mount);
    730 		return error;
    731 #else
    732 		return (EOPNOTSUPP);
    733 #endif
    734 	}
    735 
    736 	/* XXX Not implemented for UFS2 file systems. */
    737 	return (EOPNOTSUPP);
    738 }
    739 
    740 int
    741 ffs_listextattr(void *v)
    742 {
    743 	struct vop_listextattr_args /* {
    744 		struct vnode *a_vp;
    745 		int a_attrnamespace;
    746 		struct uio *a_uio;
    747 		size_t *a_size;
    748 		kauth_cred_t a_cred;
    749 		struct proc *a_p;
    750 	} */ *ap = v;
    751 	struct inode *ip = VTOI(ap->a_vp);
    752 	struct fs *fs = ip->i_fs;
    753 
    754 	/* Not supported for UFS1 file systems. */
    755 	if (fs->fs_magic == FS_UFS1_MAGIC)
    756 		return (EOPNOTSUPP);
    757 
    758 	/* XXX Not implemented for UFS2 file systems. */
    759 	return (EOPNOTSUPP);
    760 }
    761 
    762 int
    763 ffs_deleteextattr(void *v)
    764 {
    765 	struct vop_deleteextattr_args /* {
    766 		struct vnode *a_vp;
    767 		int a_attrnamespace;
    768 		kauth_cred_t a_cred;
    769 		struct proc *a_p;
    770 	} */ *ap = v;
    771 	struct vnode *vp = ap->a_vp;
    772 	struct inode *ip = VTOI(vp);
    773 	struct fs *fs = ip->i_fs;
    774 
    775 	if (fs->fs_magic == FS_UFS1_MAGIC) {
    776 #ifdef UFS_EXTATTR
    777 		int error;
    778 
    779 		fstrans_start(vp->v_mount, FSTRANS_SHARED);
    780 		error = ufs_deleteextattr(ap);
    781 		fstrans_done(vp->v_mount);
    782 		return error;
    783 #else
    784 		return (EOPNOTSUPP);
    785 #endif
    786 	}
    787 
    788 	/* XXX Not implemented for UFS2 file systems. */
    789 	return (EOPNOTSUPP);
    790 }
    791 
    792 /*
    793  * Lock the node.
    794  */
    795 int
    796 ffs_lock(void *v)
    797 {
    798 	struct vop_lock_args /* {
    799 		struct vnode *a_vp;
    800 		int a_flags;
    801 	} */ *ap = v;
    802 	struct vnode *vp = ap->a_vp;
    803 	struct mount *mp = vp->v_mount;
    804 	int flags = ap->a_flags;
    805 
    806 	if ((flags & LK_INTERLOCK) != 0) {
    807 		mutex_exit(&vp->v_interlock);
    808 		flags &= ~LK_INTERLOCK;
    809 	}
    810 
    811 	/*
    812 	 * Fake lock during file system suspension.
    813 	 */
    814 	if ((vp->v_type == VREG || vp->v_type == VDIR) &&
    815 	    fstrans_is_owner(mp) &&
    816 	    fstrans_getstate(mp) == FSTRANS_SUSPENDING) {
    817 		return 0;
    818 	}
    819 
    820 	return (vlockmgr(vp->v_vnlock, flags));
    821 }
    822 
    823 /*
    824  * Unlock the node.
    825  */
    826 int
    827 ffs_unlock(void *v)
    828 {
    829 	struct vop_unlock_args /* {
    830 		struct vnode *a_vp;
    831 		int a_flags;
    832 	} */ *ap = v;
    833 	struct vnode *vp = ap->a_vp;
    834 	struct mount *mp = vp->v_mount;
    835 
    836 	KASSERT(ap->a_flags == 0);
    837 
    838 	/*
    839 	 * Fake unlock during file system suspension.
    840 	 */
    841 	if ((vp->v_type == VREG || vp->v_type == VDIR) &&
    842 	    fstrans_is_owner(mp) &&
    843 	    fstrans_getstate(mp) == FSTRANS_SUSPENDING) {
    844 		return 0;
    845 	}
    846 	return (vlockmgr(vp->v_vnlock, LK_RELEASE));
    847 }
    848 
    849 /*
    850  * Return whether or not the node is locked.
    851  */
    852 int
    853 ffs_islocked(void *v)
    854 {
    855 	struct vop_islocked_args /* {
    856 		struct vnode *a_vp;
    857 	} */ *ap = v;
    858 	struct vnode *vp = ap->a_vp;
    859 
    860 	return (vlockstatus(vp->v_vnlock));
    861 }
    862