Home | History | Annotate | Line # | Download | only in lfs
lfs_vfsops.c revision 1.31
      1 /*	$NetBSD: lfs_vfsops.c,v 1.31 1999/04/11 23:58:17 perseant Exp $	*/
      2 
      3 /*-
      4  * Copyright (c) 1999 The NetBSD Foundation, Inc.
      5  * All rights reserved.
      6  *
      7  * This code is derived from software contributed to The NetBSD Foundation
      8  * by Konrad E. Schroder <perseant (at) hhhh.org>.
      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  * 3. All advertising materials mentioning features or use of this software
     19  *    must display the following acknowledgement:
     20  *      This product includes software developed by the NetBSD
     21  *      Foundation, Inc. and its contributors.
     22  * 4. Neither the name of The NetBSD Foundation nor the names of its
     23  *    contributors may be used to endorse or promote products derived
     24  *    from this software without specific prior written permission.
     25  *
     26  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
     27  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
     28  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
     29  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
     30  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
     31  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
     32  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
     33  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
     34  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
     35  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
     36  * POSSIBILITY OF SUCH DAMAGE.
     37  */
     38 /*-
     39  * Copyright (c) 1989, 1991, 1993, 1994
     40  *	The Regents of the University of California.  All rights reserved.
     41  *
     42  * Redistribution and use in source and binary forms, with or without
     43  * modification, are permitted provided that the following conditions
     44  * are met:
     45  * 1. Redistributions of source code must retain the above copyright
     46  *    notice, this list of conditions and the following disclaimer.
     47  * 2. Redistributions in binary form must reproduce the above copyright
     48  *    notice, this list of conditions and the following disclaimer in the
     49  *    documentation and/or other materials provided with the distribution.
     50  * 3. All advertising materials mentioning features or use of this software
     51  *    must display the following acknowledgement:
     52  *	This product includes software developed by the University of
     53  *	California, Berkeley and its contributors.
     54  * 4. Neither the name of the University nor the names of its contributors
     55  *    may be used to endorse or promote products derived from this software
     56  *    without specific prior written permission.
     57  *
     58  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
     59  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
     60  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
     61  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
     62  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
     63  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
     64  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
     65  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
     66  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
     67  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
     68  * SUCH DAMAGE.
     69  *
     70  *	@(#)lfs_vfsops.c	8.20 (Berkeley) 6/10/95
     71  */
     72 
     73 #if defined(_KERNEL) && !defined(_LKM)
     74 #include "opt_quota.h"
     75 #endif
     76 
     77 #include <sys/param.h>
     78 #include <sys/systm.h>
     79 #include <sys/namei.h>
     80 #include <sys/proc.h>
     81 #include <sys/kernel.h>
     82 #include <sys/vnode.h>
     83 #include <sys/mount.h>
     84 #include <sys/buf.h>
     85 #include <sys/mbuf.h>
     86 #include <sys/file.h>
     87 #include <sys/disklabel.h>
     88 #include <sys/ioctl.h>
     89 #include <sys/errno.h>
     90 #include <sys/malloc.h>
     91 #include <sys/pool.h>
     92 #include <sys/socket.h>
     93 #include <vm/vm.h>
     94 #include <sys/sysctl.h>
     95 
     96 #include <miscfs/specfs/specdev.h>
     97 
     98 #include <ufs/ufs/quota.h>
     99 #include <ufs/ufs/inode.h>
    100 #include <ufs/ufs/ufsmount.h>
    101 #include <ufs/ufs/ufs_extern.h>
    102 
    103 #include <ufs/lfs/lfs.h>
    104 #include <ufs/lfs/lfs_extern.h>
    105 
    106 int lfs_mountfs __P((struct vnode *, struct mount *, struct proc *));
    107 
    108 extern struct vnodeopv_desc lfs_vnodeop_opv_desc;
    109 extern struct vnodeopv_desc lfs_specop_opv_desc;
    110 extern struct vnodeopv_desc lfs_fifoop_opv_desc;
    111 
    112 struct vnodeopv_desc *lfs_vnodeopv_descs[] = {
    113 	&lfs_vnodeop_opv_desc,
    114 	&lfs_specop_opv_desc,
    115 	&lfs_fifoop_opv_desc,
    116 	NULL,
    117 };
    118 
    119 struct vfsops lfs_vfsops = {
    120 	MOUNT_LFS,
    121 	lfs_mount,
    122 	ufs_start,
    123 	lfs_unmount,
    124 	ufs_root,
    125 	ufs_quotactl,
    126 	lfs_statfs,
    127 	lfs_sync,
    128 	lfs_vget,
    129 	lfs_fhtovp,
    130 	lfs_vptofh,
    131 	lfs_init,
    132 	lfs_sysctl,
    133 	NULL,
    134 	ufs_check_export,
    135 	lfs_vnodeopv_descs,
    136 };
    137 
    138 struct pool lfs_inode_pool;
    139 
    140 /*
    141  * Initialize the filesystem, most work done by ufs_init.
    142  */
    143 void
    144 lfs_init()
    145 {
    146 	ufs_init();
    147 
    148 	/*
    149 	 * XXX Same structure as FFS inodes?  Should we share a common pool?
    150 	 */
    151 	pool_init(&lfs_inode_pool, sizeof(struct inode), 0, 0, 0,
    152 		  "lfsinopl", 0, pool_page_alloc_nointr, pool_page_free_nointr,
    153 		  M_LFSNODE);
    154 }
    155 
    156 /*
    157  * Called by main() when ufs is going to be mounted as root.
    158  */
    159 int
    160 lfs_mountroot()
    161 {
    162 	extern struct vnode *rootvp;
    163 	struct mount *mp;
    164 	struct proc *p = curproc;	/* XXX */
    165 	int error;
    166 
    167 	/*
    168 	 * Get vnodes for swapdev and rootdev.
    169 	 */
    170 	if ((error = bdevvp(rootdev, &rootvp))) {
    171 		printf("lfs_mountroot: can't setup bdevvp's");
    172 		return (error);
    173 	}
    174 	if ((error = vfs_rootmountalloc(MOUNT_LFS, "root_device", &mp)))
    175 		return (error);
    176 	if ((error = lfs_mountfs(rootvp, mp, p))) {
    177 		mp->mnt_op->vfs_refcount--;
    178 		vfs_unbusy(mp);
    179 		free(mp, M_MOUNT);
    180 		return (error);
    181 	}
    182 	simple_lock(&mountlist_slock);
    183 	CIRCLEQ_INSERT_TAIL(&mountlist, mp, mnt_list);
    184 	simple_unlock(&mountlist_slock);
    185 	(void)lfs_statfs(mp, &mp->mnt_stat, p);
    186 	vfs_unbusy(mp);
    187 	return (0);
    188 }
    189 
    190 /*
    191  * VFS Operations.
    192  *
    193  * mount system call
    194  */
    195 int
    196 lfs_mount(mp, path, data, ndp, p)
    197 	register struct mount *mp;
    198 	const char *path;
    199 	void *data;
    200 	struct nameidata *ndp;
    201 	struct proc *p;
    202 {
    203 	struct vnode *devvp;
    204 	struct ufs_args args;
    205 	struct ufsmount *ump = NULL;
    206 	register struct lfs *fs = NULL;				/* LFS */
    207 	size_t size;
    208 	int error;
    209 	mode_t accessmode;
    210 
    211 	error = copyin(data, (caddr_t)&args, sizeof (struct ufs_args));
    212 	if (error)
    213 		return (error);
    214 
    215 #if 0
    216 	/* Until LFS can do NFS right.		XXX */
    217 	if (args.export.ex_flags & MNT_EXPORTED)
    218 		return (EINVAL);
    219 #endif
    220 
    221 	/*
    222 	 * If updating, check whether changing from read-only to
    223 	 * read/write; if there is no device name, that's all we do.
    224 	 */
    225 	if (mp->mnt_flag & MNT_UPDATE) {
    226 		ump = VFSTOUFS(mp);
    227 		fs = ump->um_lfs;
    228 		if (fs->lfs_ronly && (mp->mnt_flag & MNT_WANTRDWR)) {
    229 			/*
    230 			 * If upgrade to read-write by non-root, then verify
    231 			 * that user has necessary permissions on the device.
    232 			 */
    233 			if (p->p_ucred->cr_uid != 0) {
    234 				vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY);
    235 				error = VOP_ACCESS(ump->um_devvp, VREAD|VWRITE,
    236 						   p->p_ucred, p);
    237 				VOP_UNLOCK(ump->um_devvp, 0);
    238 				if (error)
    239 					return (error);
    240 			}
    241 			fs->lfs_ronly = 0;
    242 		}
    243 		if (args.fspec == 0) {
    244 			/*
    245 			 * Process export requests.
    246 			 */
    247 			return (vfs_export(mp, &ump->um_export, &args.export));
    248 		}
    249 	}
    250 	/*
    251 	 * Not an update, or updating the name: look up the name
    252 	 * and verify that it refers to a sensible block device.
    253 	 */
    254 	NDINIT(ndp, LOOKUP, FOLLOW, UIO_USERSPACE, args.fspec, p);
    255 	if ((error = namei(ndp)) != 0)
    256 		return (error);
    257 	devvp = ndp->ni_vp;
    258 	if (devvp->v_type != VBLK) {
    259 		vrele(devvp);
    260 		return (ENOTBLK);
    261 	}
    262 	if (major(devvp->v_rdev) >= nblkdev) {
    263 		vrele(devvp);
    264 		return (ENXIO);
    265 	}
    266 	/*
    267 	 * If mount by non-root, then verify that user has necessary
    268 	 * permissions on the device.
    269 	 */
    270 	if (p->p_ucred->cr_uid != 0) {
    271 		accessmode = VREAD;
    272 		if ((mp->mnt_flag & MNT_RDONLY) == 0)
    273 			accessmode |= VWRITE;
    274 		vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
    275 		error = VOP_ACCESS(devvp, accessmode, p->p_ucred, p);
    276 		if (error) {
    277 			vput(devvp);
    278 			return (error);
    279 		}
    280 		VOP_UNLOCK(devvp, 0);
    281 	}
    282 	if ((mp->mnt_flag & MNT_UPDATE) == 0)
    283 		error = lfs_mountfs(devvp, mp, p);		/* LFS */
    284 	else {
    285 		if (devvp != ump->um_devvp)
    286 			error = EINVAL;	/* needs translation */
    287 		else
    288 			vrele(devvp);
    289 	}
    290 	if (error) {
    291 		vrele(devvp);
    292 		return (error);
    293 	}
    294 	ump = VFSTOUFS(mp);
    295 	fs = ump->um_lfs;					/* LFS */
    296 	(void)copyinstr(path, fs->lfs_fsmnt, sizeof(fs->lfs_fsmnt) - 1, &size);
    297 	bzero(fs->lfs_fsmnt + size, sizeof(fs->lfs_fsmnt) - size);
    298 	bcopy(fs->lfs_fsmnt, mp->mnt_stat.f_mntonname, MNAMELEN);
    299 	(void) copyinstr(args.fspec, mp->mnt_stat.f_mntfromname, MNAMELEN - 1,
    300 			 &size);
    301 	bzero(mp->mnt_stat.f_mntfromname + size, MNAMELEN - size);
    302 	return (0);
    303 }
    304 
    305 /*
    306  * Common code for mount and mountroot
    307  * LFS specific
    308  */
    309 int
    310 lfs_mountfs(devvp, mp, p)
    311 	register struct vnode *devvp;
    312 	struct mount *mp;
    313 	struct proc *p;
    314 {
    315 	extern struct vnode *rootvp;
    316 	struct dlfs *dfs, *adfs;
    317 	register struct lfs *fs;
    318 	register struct ufsmount *ump;
    319 	struct vnode *vp;
    320 	struct buf *bp, *abp;
    321 	struct partinfo dpart;
    322 	dev_t dev;
    323 	int error, i, ronly, size;
    324 	struct ucred *cred;
    325         SEGUSE *sup;
    326 
    327 	cred = p ? p->p_ucred : NOCRED;
    328 	/*
    329 	 * Disallow multiple mounts of the same device.
    330 	 * Disallow mounting of a device that is currently in use
    331 	 * (except for root, which might share swap device for miniroot).
    332 	 * Flush out any old buffers remaining from a previous use.
    333 	 */
    334 	if ((error = vfs_mountedon(devvp)) != 0)
    335 		return (error);
    336 	if (vcount(devvp) > 1 && devvp != rootvp)
    337 		return (EBUSY);
    338 	if ((error = vinvalbuf(devvp, V_SAVE, cred, p, 0, 0)) != 0)
    339 		return (error);
    340 
    341 	ronly = (mp->mnt_flag & MNT_RDONLY) != 0;
    342 	error = VOP_OPEN(devvp, ronly ? FREAD : FREAD|FWRITE, FSCRED, p);
    343 	if (error)
    344 		return (error);
    345 	if (VOP_IOCTL(devvp, DIOCGPART, (caddr_t)&dpart, FREAD, cred, p) != 0)
    346 		size = DEV_BSIZE;
    347 	else
    348 		size = dpart.disklab->d_secsize;
    349 
    350 	/* Don't free random space on error. */
    351 	bp = NULL;
    352 	ump = NULL;
    353 
    354 	/* Read in the superblock. */
    355 	error = bread(devvp, LFS_LABELPAD / size, LFS_SBPAD, cred, &bp);
    356 	if (error)
    357 		goto out;
    358 	dfs = (struct dlfs *)bp->b_data;
    359 
    360 	/*
    361 	 * Check the second superblock to see which is newer; then mount
    362 	 * using the older of the two.  This is necessary to ensure that
    363 	 * the filesystem is valid if it was not unmounted cleanly.
    364 	 */
    365 	error = bread(devvp, dfs->dlfs_sboffs[1], LFS_SBPAD, cred, &abp);
    366 	if (error)
    367 		goto out;
    368 	adfs = (struct dlfs *)abp->b_data;
    369 
    370 	if (adfs->dlfs_tstamp < dfs->dlfs_tstamp) /* XXX KS - 1s resolution? */
    371 		dfs = adfs;
    372 
    373 	/* Check the basics. */
    374 	if (dfs->dlfs_magic != LFS_MAGIC || dfs->dlfs_bsize > MAXBSIZE ||
    375 	    dfs->dlfs_bsize < sizeof(struct dlfs)) {
    376 		error = EINVAL;		/* XXX needs translation */
    377 		goto out;
    378 	}
    379 
    380 	/* Allocate the mount structure, copy the superblock into it. */
    381 	fs = malloc(sizeof(struct lfs), M_UFSMNT, M_WAITOK);
    382 	memcpy(&fs->lfs_dlfs, dfs, sizeof(struct dlfs));
    383 	ump = malloc(sizeof *ump, M_UFSMNT, M_WAITOK);
    384 	memset((caddr_t)ump, 0, sizeof *ump);
    385 	ump->um_lfs = fs;
    386 	if (sizeof(struct lfs) < LFS_SBPAD)			/* XXX why? */
    387 		bp->b_flags |= B_INVAL;
    388 	brelse(bp);
    389 	bp = NULL;
    390 	brelse(abp);
    391 	abp = NULL;
    392 
    393 	/* Set up the I/O information */
    394 	fs->lfs_iocount = 0;
    395 	fs->lfs_dirvcount = 0;
    396 	fs->lfs_activesb = 0;
    397 #ifdef LFS_CANNOT_ROLLFW
    398 	fs->lfs_sbactive = NULL;
    399 #endif
    400 #ifdef LFS_TRACK_IOS
    401 	for (i=0;i<LFS_THROTTLE;i++)
    402 		fs->lfs_pending[i] = LFS_UNUSED_DADDR;
    403 #endif
    404 
    405 	/* Set up the ifile and lock aflags */
    406 	fs->lfs_doifile = 0;
    407 	fs->lfs_writer = 0;
    408 	fs->lfs_dirops = 0;
    409 	fs->lfs_seglock = 0;
    410 
    411 	/* Set the file system readonly/modify bits. */
    412 	fs->lfs_ronly = ronly;
    413 	if (ronly == 0)
    414 		fs->lfs_fmod = 1;
    415 
    416 	/* Initialize the mount structure. */
    417 	dev = devvp->v_rdev;
    418 	mp->mnt_data = (qaddr_t)ump;
    419 	mp->mnt_stat.f_fsid.val[0] = (long)dev;
    420 	mp->mnt_stat.f_fsid.val[1] = makefstype(MOUNT_LFS);
    421 	mp->mnt_maxsymlinklen = fs->lfs_maxsymlinklen;
    422 	mp->mnt_flag |= MNT_LOCAL;
    423 	ump->um_flags = 0;
    424 	ump->um_mountp = mp;
    425 	ump->um_dev = dev;
    426 	ump->um_devvp = devvp;
    427 	ump->um_bptrtodb = 0;
    428 	ump->um_seqinc = 1 << fs->lfs_fsbtodb;
    429 	ump->um_nindir = fs->lfs_nindir;
    430 	for (i = 0; i < MAXQUOTAS; i++)
    431 		ump->um_quotas[i] = NULLVP;
    432 	devvp->v_specflags |= SI_MOUNTEDON;
    433 
    434 	/*
    435 	 * We use the ifile vnode for almost every operation.  Instead of
    436 	 * retrieving it from the hash table each time we retrieve it here,
    437 	 * artificially increment the reference count and keep a pointer
    438 	 * to it in the incore copy of the superblock.
    439 	 */
    440 	if ((error = VFS_VGET(mp, LFS_IFILE_INUM, &vp)) != 0)
    441 		goto out;
    442 	fs->lfs_ivnode = vp;
    443 	VREF(vp);
    444 	vput(vp);
    445 
    446 	/*
    447 	 * Mark the current segment as ACTIVE, since we're going to
    448 	 * be writing to it.
    449 	 */
    450         LFS_SEGENTRY(sup, fs, datosn(fs, fs->lfs_offset), bp);
    451         sup->su_flags |= SEGUSE_DIRTY | SEGUSE_ACTIVE;
    452         (void) VOP_BWRITE(bp);
    453 
    454 	return (0);
    455 out:
    456 	if (bp)
    457 		brelse(bp);
    458 	if (abp)
    459 		brelse(abp);
    460 	(void)VOP_CLOSE(devvp, ronly ? FREAD : FREAD|FWRITE, cred, p);
    461 	if (ump) {
    462 		free(ump->um_lfs, M_UFSMNT);
    463 		free(ump, M_UFSMNT);
    464 		mp->mnt_data = (qaddr_t)0;
    465 	}
    466 	return (error);
    467 }
    468 
    469 /*
    470  * unmount system call
    471  */
    472 int
    473 lfs_unmount(mp, mntflags, p)
    474 	struct mount *mp;
    475 	int mntflags;
    476 	struct proc *p;
    477 {
    478 	register struct ufsmount *ump;
    479 	register struct lfs *fs;
    480 	int error, flags, ronly;
    481 	extern int lfs_allclean_wakeup;
    482 
    483 	flags = 0;
    484 	if (mntflags & MNT_FORCE)
    485 		flags |= FORCECLOSE;
    486 
    487 	ump = VFSTOUFS(mp);
    488 	fs = ump->um_lfs;
    489 #ifdef QUOTA
    490 	if (mp->mnt_flag & MNT_QUOTA) {
    491 		int i;
    492 		error = vflush(mp, fs->lfs_ivnode, SKIPSYSTEM|flags);
    493 		if (error)
    494 			return (error);
    495 		for (i = 0; i < MAXQUOTAS; i++) {
    496 			if (ump->um_quotas[i] == NULLVP)
    497 				continue;
    498 			quotaoff(p, mp, i);
    499 		}
    500 		/*
    501 		 * Here we fall through to vflush again to ensure
    502 		 * that we have gotten rid of all the system vnodes.
    503 		 */
    504 	}
    505 #endif
    506 	if ((error = vflush(mp, fs->lfs_ivnode, flags)) != 0)
    507 		return (error);
    508 	fs->lfs_clean = 1;
    509 	if ((error = VFS_SYNC(mp, 1, p->p_ucred, p)) != 0)
    510 		return (error);
    511 	if (fs->lfs_ivnode->v_dirtyblkhd.lh_first)
    512 		panic("lfs_unmount: still dirty blocks on ifile vnode\n");
    513 	vrele(fs->lfs_ivnode);
    514 	vgone(fs->lfs_ivnode);
    515 
    516 	ronly = !fs->lfs_ronly;
    517 	ump->um_devvp->v_specflags &= ~SI_MOUNTEDON;
    518 	error = VOP_CLOSE(ump->um_devvp,
    519 	    ronly ? FREAD : FREAD|FWRITE, NOCRED, p);
    520 	vrele(ump->um_devvp);
    521 
    522 	/* XXX KS - wake up the cleaner so it can die */
    523 	wakeup(&fs->lfs_nextseg);
    524 	wakeup(&lfs_allclean_wakeup);
    525 
    526 	free(fs, M_UFSMNT);
    527 	free(ump, M_UFSMNT);
    528 	mp->mnt_data = (qaddr_t)0;
    529 	mp->mnt_flag &= ~MNT_LOCAL;
    530 	return (error);
    531 }
    532 
    533 /*
    534  * Get file system statistics.
    535  */
    536 int
    537 lfs_statfs(mp, sbp, p)
    538 	struct mount *mp;
    539 	register struct statfs *sbp;
    540 	struct proc *p;
    541 {
    542 	register struct lfs *fs;
    543 	register struct ufsmount *ump;
    544 
    545 	ump = VFSTOUFS(mp);
    546 	fs = ump->um_lfs;
    547 	if (fs->lfs_magic != LFS_MAGIC)
    548 		panic("lfs_statfs: magic");
    549 	sbp->f_type = 0;
    550 	sbp->f_bsize = fs->lfs_fsize;
    551 	sbp->f_iosize = fs->lfs_bsize;
    552 	sbp->f_blocks = dbtofrags(fs, fs->lfs_dsize);
    553 	sbp->f_bfree = dbtofrags(fs, fs->lfs_bfree);
    554 	/*
    555 	 * To compute the available space.  Subtract the minimum free
    556 	 * from the total number of blocks in the file system.	Set avail
    557 	 * to the smaller of this number and fs->lfs_bfree.
    558 	 *
    559 	 * XXX KS - is my modification below what is desired?  (This
    560 	 * will, e.g., change the report when the cleaner runs.)
    561 	 */
    562 #if 0
    563         sbp->f_bavail = (long) ((u_int64_t) fs->lfs_dsize * (u_int64_t)
    564 		(100 - fs->lfs_minfree) / (u_int64_t) 100);
    565 	sbp->f_bavail =
    566 	    sbp->f_bavail > fs->lfs_bfree ? fs->lfs_bfree : sbp->f_bavail;
    567 #else
    568 	sbp->f_bavail = (long) ((u_int64_t) fs->lfs_dsize * (u_int64_t)
    569 				(100 - fs->lfs_minfree) / (u_int64_t) 100)
    570 		- (u_int64_t)(fs->lfs_dsize - fs->lfs_bfree);
    571 #endif
    572 	sbp->f_bavail = dbtofrags(fs, sbp->f_bavail);
    573 	sbp->f_files = dbtofsb(fs,fs->lfs_bfree) * INOPB(fs);
    574 	sbp->f_ffree = sbp->f_files - fs->lfs_nfiles;
    575 	if (sbp != &mp->mnt_stat) {
    576 		bcopy(mp->mnt_stat.f_mntonname, sbp->f_mntonname, MNAMELEN);
    577 		bcopy(mp->mnt_stat.f_mntfromname, sbp->f_mntfromname, MNAMELEN);
    578 	}
    579 	strncpy(sbp->f_fstypename, mp->mnt_op->vfs_name, MFSNAMELEN);
    580 	return (0);
    581 }
    582 
    583 /*
    584  * Go through the disk queues to initiate sandbagged IO;
    585  * go through the inodes to write those that have been modified;
    586  * initiate the writing of the super block if it has been modified.
    587  *
    588  * Note: we are always called with the filesystem marked `MPBUSY'.
    589  */
    590 int
    591 lfs_sync(mp, waitfor, cred, p)
    592 	struct mount *mp;
    593 	int waitfor;
    594 	struct ucred *cred;
    595 	struct proc *p;
    596 {
    597 	int error;
    598 	struct lfs *fs;
    599 
    600 	fs = ((struct ufsmount *)mp->mnt_data)->ufsmount_u.lfs;
    601 	while(fs->lfs_dirops)
    602 		error = tsleep(&fs->lfs_dirops, PRIBIO + 1, "lfs_dirops", 0);
    603 	fs->lfs_writer++;
    604 
    605 	/* All syncs must be checkpoints until roll-forward is implemented. */
    606 	error = lfs_segwrite(mp, SEGM_CKP | (waitfor ? SEGM_SYNC : 0));
    607 	if(--fs->lfs_writer==0)
    608 		wakeup(&fs->lfs_dirops);
    609 #ifdef QUOTA
    610 	qsync(mp);
    611 #endif
    612 	return (error);
    613 }
    614 
    615 #ifdef USE_UFS_HASHLOCK
    616 extern struct lock ufs_hashlock;
    617 #endif
    618 
    619 /*
    620  * Look up an LFS dinode number to find its incore vnode.  If not already
    621  * in core, read it in from the specified device.  Return the inode locked.
    622  * Detection and handling of mount points must be done by the calling routine.
    623  */
    624 int
    625 lfs_vget(mp, ino, vpp)
    626 	struct mount *mp;
    627 	ino_t ino;
    628 	struct vnode **vpp;
    629 {
    630 	register struct lfs *fs;
    631 	register struct inode *ip;
    632 	struct buf *bp;
    633 	struct ifile *ifp;
    634 	struct vnode *vp;
    635 	struct ufsmount *ump;
    636 	ufs_daddr_t daddr;
    637 	dev_t dev;
    638 	int error;
    639 #ifdef LFS_ATIME_IFILE
    640 	struct timespec ts;
    641 #endif
    642 
    643 	ump = VFSTOUFS(mp);
    644 	dev = ump->um_dev;
    645 
    646 #ifdef USE_UFS_HASHLOCK
    647 	do {
    648 #endif
    649 		if ((*vpp = ufs_ihashget(dev, ino)) != NULL)
    650 			return (0);
    651 #ifdef USE_UFS_HASHLOCK
    652 	} while (lockmgr(&ufs_hashlock, LK_EXCLUSIVE|LK_SLEEPFAIL, 0));
    653 #endif
    654 
    655 	/* Translate the inode number to a disk address. */
    656 	fs = ump->um_lfs;
    657 	if (ino == LFS_IFILE_INUM)
    658 		daddr = fs->lfs_idaddr;
    659 	else {
    660 		LFS_IENTRY(ifp, fs, ino, bp);
    661 		daddr = ifp->if_daddr;
    662 #ifdef LFS_ATIME_IFILE
    663 		ts = ifp->if_atime; /* structure copy */
    664 #endif
    665 		brelse(bp);
    666 		if (daddr == LFS_UNUSED_DADDR) {
    667 #ifdef USE_UFS_HASHLOCK
    668 			lockmgr(&ufs_hashlock, LK_RELEASE, 0);
    669 #endif
    670 			return (ENOENT);
    671 		}
    672 	}
    673 
    674 	/* Allocate new vnode/inode. */
    675 	if ((error = lfs_vcreate(mp, ino, &vp)) != 0) {
    676 		*vpp = NULL;
    677 #ifdef USE_UFS_HASHLOCK
    678 		lockmgr(&ufs_hashlock, LK_RELEASE, 0);
    679 #endif
    680 		return (error);
    681 	}
    682 
    683 	/*
    684 	 * Put it onto its hash chain and lock it so that other requests for
    685 	 * this inode will block if they arrive while we are sleeping waiting
    686 	 * for old data structures to be purged or for the contents of the
    687 	 * disk portion of this inode to be read.
    688 	 */
    689 	ip = VTOI(vp);
    690 	ufs_ihashins(ip);
    691 #ifdef USE_UFS_HASHLOCK
    692 	lockmgr(&ufs_hashlock, LK_RELEASE, 0);
    693 #endif
    694 
    695 	/*
    696 	 * XXX
    697 	 * This may not need to be here, logically it should go down with
    698 	 * the i_devvp initialization.
    699 	 * Ask Kirk.
    700 	 */
    701 	ip->i_lfs = ump->um_lfs;
    702 
    703 	/* Read in the disk contents for the inode, copy into the inode. */
    704 	error = bread(ump->um_devvp, daddr, (int)fs->lfs_bsize, NOCRED, &bp);
    705 	if (error) {
    706 		/*
    707 		 * The inode does not contain anything useful, so it would
    708 		 * be misleading to leave it on its hash chain. With mode
    709 		 * still zero, it will be unlinked and returned to the free
    710 		 * list by vput().
    711 		 */
    712 		vput(vp);
    713 		brelse(bp);
    714 		*vpp = NULL;
    715 		return (error);
    716 	}
    717 	ip->i_din.ffs_din = *lfs_ifind(fs, ino, (struct dinode *)bp->b_data);
    718 #ifdef LFS_ATIME_IFILE
    719 	ip->i_ffs_atime = ts.tv_sec;
    720 	ip->i_ffs_atimensec = ts.tv_nsec;
    721 #endif
    722 	brelse(bp);
    723 
    724 	/*
    725 	 * Initialize the vnode from the inode, check for aliases.  In all
    726 	 * cases re-init ip, the underlying vnode/inode may have changed.
    727 	 */
    728 	error = ufs_vinit(mp, lfs_specop_p, lfs_fifoop_p, &vp);
    729 	if (error) {
    730 		vput(vp);
    731 		*vpp = NULL;
    732 		return (error);
    733 	}
    734 	if(vp->v_type == VNON) {
    735 		printf("lfs_vget: ino %d is type VNON! (ifmt %o)\n", ip->i_number, (ip->i_ffs_mode&IFMT)>>12);
    736 #ifdef DDB
    737 		Debugger();
    738 #endif
    739 	}
    740 	/*
    741 	 * Finish inode initialization now that aliasing has been resolved.
    742 	 */
    743 	ip->i_devvp = ump->um_devvp;
    744 	VREF(ip->i_devvp);
    745 	*vpp = vp;
    746 
    747 	return (0);
    748 }
    749 
    750 /*
    751  * File handle to vnode
    752  *
    753  * Have to be really careful about stale file handles:
    754  * - check that the inode number is valid
    755  * - call lfs_vget() to get the locked inode
    756  * - check for an unallocated inode (i_mode == 0)
    757  *
    758  * XXX
    759  * use ifile to see if inode is allocated instead of reading off disk
    760  * what is the relationship between my generational number and the NFS
    761  * generational number.
    762  */
    763 int
    764 lfs_fhtovp(mp, fhp, vpp)
    765 	register struct mount *mp;
    766 	struct fid *fhp;
    767 	struct vnode **vpp;
    768 {
    769 	register struct ufid *ufhp;
    770 
    771 	ufhp = (struct ufid *)fhp;
    772 	if (ufhp->ufid_ino < ROOTINO)
    773 		return (ESTALE);
    774 	return (ufs_fhtovp(mp, ufhp, vpp));
    775 }
    776 
    777 /*
    778  * Vnode pointer to File handle
    779  */
    780 /* ARGSUSED */
    781 int
    782 lfs_vptofh(vp, fhp)
    783 	struct vnode *vp;
    784 	struct fid *fhp;
    785 {
    786 	register struct inode *ip;
    787 	register struct ufid *ufhp;
    788 
    789 	ip = VTOI(vp);
    790 	ufhp = (struct ufid *)fhp;
    791 	ufhp->ufid_len = sizeof(struct ufid);
    792 	ufhp->ufid_ino = ip->i_number;
    793 	ufhp->ufid_gen = ip->i_ffs_gen;
    794 	return (0);
    795 }
    796 
    797 int
    798 lfs_sysctl(name, namelen, oldp, oldlenp, newp, newlen, p)
    799 	int *name;
    800 	u_int namelen;
    801 	void *oldp;
    802 	size_t *oldlenp;
    803 	void *newp;
    804 	size_t newlen;
    805 	struct proc *p;
    806 {
    807 	extern int lfs_writeindir, lfs_dostats, lfs_clean_vnhead;
    808 	extern struct lfs_stats lfs_stats;
    809 	int error;
    810 
    811 	/* all sysctl names at this level are terminal */
    812 	if (namelen != 1)
    813 		return (ENOTDIR);
    814 
    815 	switch (name[0]) {
    816 	case LFS_WRITEINDIR:
    817 		return (sysctl_int(oldp, oldlenp, newp, newlen,
    818 				   &lfs_writeindir));
    819 	case LFS_CLEAN_VNHEAD:
    820 		return (sysctl_int(oldp, oldlenp, newp, newlen,
    821 				   &lfs_clean_vnhead));
    822 	case LFS_DOSTATS:
    823 		if((error = sysctl_int(oldp, oldlenp, newp, newlen,
    824 				       &lfs_dostats)))
    825 			return error;
    826 		if(lfs_dostats == 0)
    827 			memset(&lfs_stats,0,sizeof(lfs_stats));
    828 		return 0;
    829 	case LFS_STATS:
    830 		return (sysctl_rdstruct(oldp, oldlenp, newp,
    831 					&lfs_stats, sizeof(lfs_stats)));
    832 	default:
    833 		return (EOPNOTSUPP);
    834 	}
    835 	/* NOTREACHED */
    836 }
    837