Home | History | Annotate | Line # | Download | only in ffs
ffs_vfsops.c revision 1.33
      1 /*	$NetBSD: ffs_vfsops.c,v 1.33 1998/03/01 02:23:15 fvdl Exp $	*/
      2 
      3 /*
      4  * Copyright (c) 1989, 1991, 1993, 1994
      5  *	The Regents of the University of California.  All rights reserved.
      6  *
      7  * Redistribution and use in source and binary forms, with or without
      8  * modification, are permitted provided that the following conditions
      9  * are met:
     10  * 1. Redistributions of source code must retain the above copyright
     11  *    notice, this list of conditions and the following disclaimer.
     12  * 2. Redistributions in binary form must reproduce the above copyright
     13  *    notice, this list of conditions and the following disclaimer in the
     14  *    documentation and/or other materials provided with the distribution.
     15  * 3. All advertising materials mentioning features or use of this software
     16  *    must display the following acknowledgement:
     17  *	This product includes software developed by the University of
     18  *	California, Berkeley and its contributors.
     19  * 4. Neither the name of the University nor the names of its contributors
     20  *    may be used to endorse or promote products derived from this software
     21  *    without specific prior written permission.
     22  *
     23  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
     24  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
     25  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
     26  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
     27  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
     28  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
     29  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
     30  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
     31  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
     32  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
     33  * SUCH DAMAGE.
     34  *
     35  *	@(#)ffs_vfsops.c	8.31 (Berkeley) 5/20/95
     36  */
     37 
     38 #include <sys/param.h>
     39 #include <sys/systm.h>
     40 #include <sys/namei.h>
     41 #include <sys/proc.h>
     42 #include <sys/kernel.h>
     43 #include <sys/vnode.h>
     44 #include <sys/socket.h>
     45 #include <sys/mount.h>
     46 #include <sys/buf.h>
     47 #include <sys/device.h>
     48 #include <sys/mbuf.h>
     49 #include <sys/file.h>
     50 #include <sys/disklabel.h>
     51 #include <sys/ioctl.h>
     52 #include <sys/errno.h>
     53 #include <sys/malloc.h>
     54 #include <sys/lock.h>
     55 #include <vm/vm.h>
     56 #include <sys/sysctl.h>
     57 
     58 #include <miscfs/specfs/specdev.h>
     59 
     60 #include <ufs/ufs/quota.h>
     61 #include <ufs/ufs/ufsmount.h>
     62 #include <ufs/ufs/inode.h>
     63 #include <ufs/ufs/dir.h>
     64 #include <ufs/ufs/ufs_extern.h>
     65 
     66 #include <ufs/ffs/fs.h>
     67 #include <ufs/ffs/ffs_extern.h>
     68 
     69 extern struct lock ufs_hashlock;
     70 
     71 int ffs_sbupdate __P((struct ufsmount *, int));
     72 
     73 extern struct vnodeopv_desc ffs_vnodeop_opv_desc;
     74 extern struct vnodeopv_desc ffs_specop_opv_desc;
     75 #ifdef FIFO
     76 extern struct vnodeopv_desc ffs_fifoop_opv_desc;
     77 #endif
     78 
     79 struct vnodeopv_desc *ffs_vnodeopv_descs[] = {
     80 	&ffs_vnodeop_opv_desc,
     81 	&ffs_specop_opv_desc,
     82 #ifdef FIFO
     83 	&ffs_fifoop_opv_desc,
     84 #endif
     85 	NULL,
     86 };
     87 
     88 struct vfsops ffs_vfsops = {
     89 	MOUNT_FFS,
     90 	ffs_mount,
     91 	ufs_start,
     92 	ffs_unmount,
     93 	ufs_root,
     94 	ufs_quotactl,
     95 	ffs_statfs,
     96 	ffs_sync,
     97 	ffs_vget,
     98 	ffs_fhtovp,
     99 	ffs_vptofh,
    100 	ffs_init,
    101 	ffs_sysctl,
    102 	ffs_mountroot,
    103 	ffs_vnodeopv_descs,
    104 };
    105 
    106 /*
    107  * Called by main() when ffs is going to be mounted as root.
    108  */
    109 
    110 int
    111 ffs_mountroot()
    112 {
    113 	extern struct vnode *rootvp;
    114 	struct fs *fs;
    115 	struct mount *mp;
    116 	struct proc *p = curproc;	/* XXX */
    117 	struct ufsmount *ump;
    118 	int error;
    119 
    120 	if (root_device->dv_class != DV_DISK)
    121 		return (ENODEV);
    122 
    123 	/*
    124 	 * Get vnodes for rootdev.
    125 	 */
    126 	if (bdevvp(rootdev, &rootvp))
    127 		panic("ffs_mountroot: can't setup bdevvp's");
    128 
    129 	if ((error = vfs_rootmountalloc(MOUNT_FFS, "root_device", &mp)))
    130 		return (error);
    131 	if ((error = ffs_mountfs(rootvp, mp, p)) != 0) {
    132 		mp->mnt_op->vfs_refcount--;
    133 		vfs_unbusy(mp);
    134 		free(mp, M_MOUNT);
    135 		return (error);
    136 	}
    137 	simple_lock(&mountlist_slock);
    138 	CIRCLEQ_INSERT_TAIL(&mountlist, mp, mnt_list);
    139 	simple_unlock(&mountlist_slock);
    140 	ump = VFSTOUFS(mp);
    141 	fs = ump->um_fs;
    142 	bzero(fs->fs_fsmnt, sizeof(fs->fs_fsmnt));
    143 	(void)copystr(mp->mnt_stat.f_mntonname, fs->fs_fsmnt, MNAMELEN - 1, 0);
    144 	(void)ffs_statfs(mp, &mp->mnt_stat, p);
    145 	vfs_unbusy(mp);
    146 	inittodr(fs->fs_time);
    147 	return (0);
    148 }
    149 
    150 /*
    151  * VFS Operations.
    152  *
    153  * mount system call
    154  */
    155 int
    156 ffs_mount(mp, path, data, ndp, p)
    157 	register struct mount *mp;
    158 	const char *path;
    159 	void *data;
    160 	struct nameidata *ndp;
    161 	struct proc *p;
    162 {
    163 	struct vnode *devvp;
    164 	struct ufs_args args;
    165 	struct ufsmount *ump = NULL;
    166 	register struct fs *fs;
    167 	size_t size;
    168 	int error, flags;
    169 	mode_t accessmode;
    170 
    171 	error = copyin(data, (caddr_t)&args, sizeof (struct ufs_args));
    172 	if (error)
    173 		return (error);
    174 	/*
    175 	 * If updating, check whether changing from read-only to
    176 	 * read/write; if there is no device name, that's all we do.
    177 	 */
    178 	if (mp->mnt_flag & MNT_UPDATE) {
    179 		ump = VFSTOUFS(mp);
    180 		fs = ump->um_fs;
    181 		if (fs->fs_ronly == 0 && (mp->mnt_flag & MNT_RDONLY)) {
    182 			flags = WRITECLOSE;
    183 			if (mp->mnt_flag & MNT_FORCE)
    184 				flags |= FORCECLOSE;
    185 			error = ffs_flushfiles(mp, flags, p);
    186 			if (error == 0 &&
    187 			    ffs_cgupdate(ump, MNT_WAIT) == 0 &&
    188 			    fs->fs_clean & FS_WASCLEAN) {
    189 				fs->fs_clean = FS_ISCLEAN;
    190 				(void) ffs_sbupdate(ump, MNT_WAIT);
    191 			}
    192 			if (error)
    193 				return (error);
    194 			fs->fs_ronly = 1;
    195 		}
    196 		if (mp->mnt_flag & MNT_RELOAD) {
    197 			error = ffs_reload(mp, ndp->ni_cnd.cn_cred, p);
    198 			if (error)
    199 				return (error);
    200 		}
    201 		if (fs->fs_ronly && (mp->mnt_flag & MNT_WANTRDWR)) {
    202 			/*
    203 			 * If upgrade to read-write by non-root, then verify
    204 			 * that user has necessary permissions on the device.
    205 			 */
    206 			if (p->p_ucred->cr_uid != 0) {
    207 				devvp = ump->um_devvp;
    208 				vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
    209 				error = VOP_ACCESS(devvp, VREAD | VWRITE,
    210 						   p->p_ucred, p);
    211 				VOP_UNLOCK(devvp, 0);
    212 				if (error)
    213 					return (error);
    214 			}
    215 			fs->fs_ronly = 0;
    216 			fs->fs_clean <<= 1;
    217 			fs->fs_fmod = 1;
    218 		}
    219 		if (args.fspec == 0) {
    220 			/*
    221 			 * Process export requests.
    222 			 */
    223 			return (vfs_export(mp, &ump->um_export, &args.export));
    224 		}
    225 	}
    226 	/*
    227 	 * Not an update, or updating the name: look up the name
    228 	 * and verify that it refers to a sensible block device.
    229 	 */
    230 	NDINIT(ndp, LOOKUP, FOLLOW, UIO_USERSPACE, args.fspec, p);
    231 	if ((error = namei(ndp)) != 0)
    232 		return (error);
    233 	devvp = ndp->ni_vp;
    234 
    235 	if (devvp->v_type != VBLK) {
    236 		vrele(devvp);
    237 		return (ENOTBLK);
    238 	}
    239 	if (major(devvp->v_rdev) >= nblkdev) {
    240 		vrele(devvp);
    241 		return (ENXIO);
    242 	}
    243 	/*
    244 	 * If mount by non-root, then verify that user has necessary
    245 	 * permissions on the device.
    246 	 */
    247 	if (p->p_ucred->cr_uid != 0) {
    248 		accessmode = VREAD;
    249 		if ((mp->mnt_flag & MNT_RDONLY) == 0)
    250 			accessmode |= VWRITE;
    251 		vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
    252 		error = VOP_ACCESS(devvp, accessmode, p->p_ucred, p);
    253 		VOP_UNLOCK(devvp, 0);
    254 		if (error) {
    255 			vrele(devvp);
    256 			return (error);
    257 		}
    258 	}
    259 	if ((mp->mnt_flag & MNT_UPDATE) == 0)
    260 		error = ffs_mountfs(devvp, mp, p);
    261 	else {
    262 		if (devvp != ump->um_devvp)
    263 			error = EINVAL;	/* needs translation */
    264 		else
    265 			vrele(devvp);
    266 	}
    267 	if (error) {
    268 		vrele(devvp);
    269 		return (error);
    270 	}
    271 	ump = VFSTOUFS(mp);
    272 	fs = ump->um_fs;
    273 	(void) copyinstr(path, fs->fs_fsmnt, sizeof(fs->fs_fsmnt) - 1, &size);
    274 	bzero(fs->fs_fsmnt + size, sizeof(fs->fs_fsmnt) - size);
    275 	bcopy(fs->fs_fsmnt, mp->mnt_stat.f_mntonname, MNAMELEN);
    276 	(void) copyinstr(args.fspec, mp->mnt_stat.f_mntfromname, MNAMELEN - 1,
    277 	    &size);
    278 	bzero(mp->mnt_stat.f_mntfromname + size, MNAMELEN - size);
    279 	if (fs->fs_fmod != 0) {	/* XXX */
    280 		fs->fs_fmod = 0;
    281 		if (fs->fs_clean & FS_WASCLEAN)
    282 			fs->fs_time = time.tv_sec;
    283 		else
    284 			printf("%s: file system not clean; please fsck(8)\n",
    285 			    mp->mnt_stat.f_mntfromname);
    286 		(void) ffs_cgupdate(ump, MNT_WAIT);
    287 	}
    288 	return (0);
    289 }
    290 
    291 /*
    292  * Reload all incore data for a filesystem (used after running fsck on
    293  * the root filesystem and finding things to fix). The filesystem must
    294  * be mounted read-only.
    295  *
    296  * Things to do to update the mount:
    297  *	1) invalidate all cached meta-data.
    298  *	2) re-read superblock from disk.
    299  *	3) re-read summary information from disk.
    300  *	4) invalidate all inactive vnodes.
    301  *	5) invalidate all cached file data.
    302  *	6) re-read inode data for all active vnodes.
    303  */
    304 int
    305 ffs_reload(mountp, cred, p)
    306 	register struct mount *mountp;
    307 	struct ucred *cred;
    308 	struct proc *p;
    309 {
    310 	register struct vnode *vp, *nvp, *devvp;
    311 	struct inode *ip;
    312 	struct csum *space;
    313 	struct buf *bp;
    314 	struct fs *fs, *newfs;
    315 	struct partinfo dpart;
    316 	int i, blks, size, error;
    317 	int32_t *lp;
    318 
    319 	if ((mountp->mnt_flag & MNT_RDONLY) == 0)
    320 		return (EINVAL);
    321 	/*
    322 	 * Step 1: invalidate all cached meta-data.
    323 	 */
    324 	devvp = VFSTOUFS(mountp)->um_devvp;
    325 	if (vinvalbuf(devvp, 0, cred, p, 0, 0))
    326 		panic("ffs_reload: dirty1");
    327 	/*
    328 	 * Step 2: re-read superblock from disk.
    329 	 */
    330 	if (VOP_IOCTL(devvp, DIOCGPART, (caddr_t)&dpart, FREAD, NOCRED, p) != 0)
    331 		size = DEV_BSIZE;
    332 	else
    333 		size = dpart.disklab->d_secsize;
    334 	error = bread(devvp, (ufs_daddr_t)(SBOFF / size), SBSIZE, NOCRED, &bp);
    335 	if (error)
    336 		return (error);
    337 	newfs = (struct fs *)bp->b_data;
    338 	if (newfs->fs_magic != FS_MAGIC || newfs->fs_bsize > MAXBSIZE ||
    339 	    newfs->fs_bsize < sizeof(struct fs)) {
    340 		brelse(bp);
    341 		return (EIO);		/* XXX needs translation */
    342 	}
    343 	fs = VFSTOUFS(mountp)->um_fs;
    344 	/*
    345 	 * Copy pointer fields back into superblock before copying in	XXX
    346 	 * new superblock. These should really be in the ufsmount.	XXX
    347 	 * Note that important parameters (eg fs_ncg) are unchanged.
    348 	 */
    349 	bcopy(&fs->fs_csp[0], &newfs->fs_csp[0], sizeof(fs->fs_csp));
    350 	newfs->fs_maxcluster = fs->fs_maxcluster;
    351 	bcopy(newfs, fs, (u_int)fs->fs_sbsize);
    352 	if (fs->fs_sbsize < SBSIZE)
    353 		bp->b_flags |= B_INVAL;
    354 	brelse(bp);
    355 	mountp->mnt_maxsymlinklen = fs->fs_maxsymlinklen;
    356 	ffs_oldfscompat(fs);
    357 	/*
    358 	 * Step 3: re-read summary information from disk.
    359 	 */
    360 	blks = howmany(fs->fs_cssize, fs->fs_fsize);
    361 	space = fs->fs_csp[0];
    362 	for (i = 0; i < blks; i += fs->fs_frag) {
    363 		size = fs->fs_bsize;
    364 		if (i + fs->fs_frag > blks)
    365 			size = (blks - i) * fs->fs_fsize;
    366 		error = bread(devvp, fsbtodb(fs, fs->fs_csaddr + i), size,
    367 			      NOCRED, &bp);
    368 		if (error)
    369 			return (error);
    370 		bcopy(bp->b_data, fs->fs_csp[fragstoblks(fs, i)], (u_int)size);
    371 		brelse(bp);
    372 	}
    373 	/*
    374 	 * We no longer know anything about clusters per cylinder group.
    375 	 */
    376 	if (fs->fs_contigsumsize > 0) {
    377 		lp = fs->fs_maxcluster;
    378 		for (i = 0; i < fs->fs_ncg; i++)
    379 			*lp++ = fs->fs_contigsumsize;
    380 	}
    381 
    382 loop:
    383 	simple_lock(&mntvnode_slock);
    384 	for (vp = mountp->mnt_vnodelist.lh_first; vp != NULL; vp = nvp) {
    385 		if (vp->v_mount != mountp) {
    386 			simple_unlock(&mntvnode_slock);
    387 			goto loop;
    388 		}
    389 		nvp = vp->v_mntvnodes.le_next;
    390 		/*
    391 		 * Step 4: invalidate all inactive vnodes.
    392 		 */
    393 		if (vrecycle(vp, &mntvnode_slock, p))
    394 			goto loop;
    395 		/*
    396 		 * Step 5: invalidate all cached file data.
    397 		 */
    398 		simple_lock(&vp->v_interlock);
    399 		simple_unlock(&mntvnode_slock);
    400 		if (vget(vp, LK_EXCLUSIVE | LK_INTERLOCK))
    401 			goto loop;
    402 		if (vinvalbuf(vp, 0, cred, p, 0, 0))
    403 			panic("ffs_reload: dirty2");
    404 		/*
    405 		 * Step 6: re-read inode data for all active vnodes.
    406 		 */
    407 		ip = VTOI(vp);
    408 		error = bread(devvp, fsbtodb(fs, ino_to_fsba(fs, ip->i_number)),
    409 			      (int)fs->fs_bsize, NOCRED, &bp);
    410 		if (error) {
    411 			vput(vp);
    412 			return (error);
    413 		}
    414 		ip->i_din.ffs_din = *((struct dinode *)bp->b_data +
    415 		    ino_to_fsbo(fs, ip->i_number));
    416 		brelse(bp);
    417 		vput(vp);
    418 		simple_lock(&mntvnode_slock);
    419 	}
    420 	simple_unlock(&mntvnode_slock);
    421 	return (0);
    422 }
    423 
    424 /*
    425  * Common code for mount and mountroot
    426  */
    427 int
    428 ffs_mountfs(devvp, mp, p)
    429 	register struct vnode *devvp;
    430 	struct mount *mp;
    431 	struct proc *p;
    432 {
    433 	register struct ufsmount *ump;
    434 	struct buf *bp;
    435 	register struct fs *fs;
    436 	dev_t dev;
    437 	struct partinfo dpart;
    438 	caddr_t base, space;
    439 	int blks;
    440 	int error, i, size, ronly;
    441 	int32_t *lp;
    442 	struct ucred *cred;
    443 	extern struct vnode *rootvp;
    444 	u_int64_t maxfilesize;					/* XXX */
    445 
    446 	dev = devvp->v_rdev;
    447 	cred = p ? p->p_ucred : NOCRED;
    448 	/*
    449 	 * Disallow multiple mounts of the same device.
    450 	 * Disallow mounting of a device that is currently in use
    451 	 * (except for root, which might share swap device for miniroot).
    452 	 * Flush out any old buffers remaining from a previous use.
    453 	 */
    454 	if ((error = vfs_mountedon(devvp)) != 0)
    455 		return (error);
    456 	if (vcount(devvp) > 1 && devvp != rootvp)
    457 		return (EBUSY);
    458 	if ((error = vinvalbuf(devvp, V_SAVE, cred, p, 0, 0)) != 0)
    459 		return (error);
    460 
    461 	ronly = (mp->mnt_flag & MNT_RDONLY) != 0;
    462 	error = VOP_OPEN(devvp, ronly ? FREAD : FREAD|FWRITE, FSCRED, p);
    463 	if (error)
    464 		return (error);
    465 	if (VOP_IOCTL(devvp, DIOCGPART, (caddr_t)&dpart, FREAD, cred, p) != 0)
    466 		size = DEV_BSIZE;
    467 	else
    468 		size = dpart.disklab->d_secsize;
    469 
    470 	bp = NULL;
    471 	ump = NULL;
    472 	error = bread(devvp, (ufs_daddr_t)(SBOFF / size), SBSIZE, cred, &bp);
    473 	if (error)
    474 		goto out;
    475 	fs = (struct fs *)bp->b_data;
    476 	if (fs->fs_magic != FS_MAGIC || fs->fs_bsize > MAXBSIZE ||
    477 	    fs->fs_bsize < sizeof(struct fs)) {
    478 		error = EINVAL;		/* XXX needs translation */
    479 		goto out;
    480 	}
    481 	/* XXX updating 4.2 FFS superblocks trashes rotational layout tables */
    482 	if (fs->fs_postblformat == FS_42POSTBLFMT && !ronly) {
    483 		error = EROFS;		/* XXX what should be returned? */
    484 		goto out;
    485 	}
    486 	ump = malloc(sizeof *ump, M_UFSMNT, M_WAITOK);
    487 	bzero((caddr_t)ump, sizeof *ump);
    488 	ump->um_fs = malloc((u_long)fs->fs_sbsize, M_UFSMNT,
    489 	    M_WAITOK);
    490 	bcopy(bp->b_data, ump->um_fs, (u_int)fs->fs_sbsize);
    491 	if (fs->fs_sbsize < SBSIZE)
    492 		bp->b_flags |= B_INVAL;
    493 	brelse(bp);
    494 	bp = NULL;
    495 	fs = ump->um_fs;
    496 	fs->fs_ronly = ronly;
    497 	if (ronly == 0) {
    498 		fs->fs_clean <<= 1;
    499 		fs->fs_fmod = 1;
    500 	}
    501 	size = fs->fs_cssize;
    502 	blks = howmany(size, fs->fs_fsize);
    503 	if (fs->fs_contigsumsize > 0)
    504 		size += fs->fs_ncg * sizeof(int32_t);
    505 	base = space = malloc((u_long)size, M_UFSMNT, M_WAITOK);
    506 	for (i = 0; i < blks; i += fs->fs_frag) {
    507 		size = fs->fs_bsize;
    508 		if (i + fs->fs_frag > blks)
    509 			size = (blks - i) * fs->fs_fsize;
    510 		error = bread(devvp, fsbtodb(fs, fs->fs_csaddr + i), size,
    511 			      cred, &bp);
    512 		if (error) {
    513 			free(base, M_UFSMNT);
    514 			goto out;
    515 		}
    516 		bcopy(bp->b_data, space, (u_int)size);
    517 		fs->fs_csp[fragstoblks(fs, i)] = (struct csum *)space;
    518 		space += size;
    519 		brelse(bp);
    520 		bp = NULL;
    521 	}
    522 	if (fs->fs_contigsumsize > 0) {
    523 		fs->fs_maxcluster = lp = (int32_t *)space;
    524 		for (i = 0; i < fs->fs_ncg; i++)
    525 			*lp++ = fs->fs_contigsumsize;
    526 	}
    527 	mp->mnt_data = (qaddr_t)ump;
    528 	mp->mnt_stat.f_fsid.val[0] = (long)dev;
    529 	mp->mnt_stat.f_fsid.val[1] = makefstype(MOUNT_FFS);
    530 	mp->mnt_maxsymlinklen = fs->fs_maxsymlinklen;
    531 	mp->mnt_flag |= MNT_LOCAL;
    532 	ump->um_mountp = mp;
    533 	ump->um_dev = dev;
    534 	ump->um_devvp = devvp;
    535 	ump->um_nindir = fs->fs_nindir;
    536 	ump->um_bptrtodb = fs->fs_fsbtodb;
    537 	ump->um_seqinc = fs->fs_frag;
    538 	for (i = 0; i < MAXQUOTAS; i++)
    539 		ump->um_quotas[i] = NULLVP;
    540 	devvp->v_specflags |= SI_MOUNTEDON;
    541 	ffs_oldfscompat(fs);
    542 	ump->um_savedmaxfilesize = fs->fs_maxfilesize;		/* XXX */
    543 	maxfilesize = (u_int64_t)0x80000000 * fs->fs_bsize - 1;	/* XXX */
    544 	if (fs->fs_maxfilesize > maxfilesize)			/* XXX */
    545 		fs->fs_maxfilesize = maxfilesize;		/* XXX */
    546 	return (0);
    547 out:
    548 	if (bp)
    549 		brelse(bp);
    550 	(void)VOP_CLOSE(devvp, ronly ? FREAD : FREAD|FWRITE, cred, p);
    551 	if (ump) {
    552 		free(ump->um_fs, M_UFSMNT);
    553 		free(ump, M_UFSMNT);
    554 		mp->mnt_data = (qaddr_t)0;
    555 	}
    556 	return (error);
    557 }
    558 
    559 /*
    560  * Sanity checks for old file systems.
    561  *
    562  * XXX - goes away some day.
    563  */
    564 int
    565 ffs_oldfscompat(fs)
    566 	struct fs *fs;
    567 {
    568 	int i;
    569 
    570 	fs->fs_npsect = max(fs->fs_npsect, fs->fs_nsect);	/* XXX */
    571 	fs->fs_interleave = max(fs->fs_interleave, 1);		/* XXX */
    572 	if (fs->fs_postblformat == FS_42POSTBLFMT)		/* XXX */
    573 		fs->fs_nrpos = 8;				/* XXX */
    574 	if (fs->fs_inodefmt < FS_44INODEFMT) {			/* XXX */
    575 		u_int64_t sizepb = fs->fs_bsize;		/* XXX */
    576 								/* XXX */
    577 		fs->fs_maxfilesize = fs->fs_bsize * NDADDR - 1;	/* XXX */
    578 		for (i = 0; i < NIADDR; i++) {			/* XXX */
    579 			sizepb *= NINDIR(fs);			/* XXX */
    580 			fs->fs_maxfilesize += sizepb;		/* XXX */
    581 		}						/* XXX */
    582 		fs->fs_qbmask = ~fs->fs_bmask;			/* XXX */
    583 		fs->fs_qfmask = ~fs->fs_fmask;			/* XXX */
    584 	}							/* XXX */
    585 	return (0);
    586 }
    587 
    588 /*
    589  * unmount system call
    590  */
    591 int
    592 ffs_unmount(mp, mntflags, p)
    593 	struct mount *mp;
    594 	int mntflags;
    595 	struct proc *p;
    596 {
    597 	register struct ufsmount *ump;
    598 	register struct fs *fs;
    599 	int error, flags;
    600 
    601 	flags = 0;
    602 	if (mntflags & MNT_FORCE)
    603 		flags |= FORCECLOSE;
    604 	if ((error = ffs_flushfiles(mp, flags, p)) != 0)
    605 		return (error);
    606 	ump = VFSTOUFS(mp);
    607 	fs = ump->um_fs;
    608 	if (fs->fs_ronly == 0 &&
    609 	    ffs_cgupdate(ump, MNT_WAIT) == 0 &&
    610 	    fs->fs_clean & FS_WASCLEAN) {
    611 		fs->fs_clean = FS_ISCLEAN;
    612 		(void) ffs_sbupdate(ump, MNT_WAIT);
    613 	}
    614 	ump->um_devvp->v_specflags &= ~SI_MOUNTEDON;
    615 	error = VOP_CLOSE(ump->um_devvp, fs->fs_ronly ? FREAD : FREAD|FWRITE,
    616 		NOCRED, p);
    617 	vrele(ump->um_devvp);
    618 	free(fs->fs_csp[0], M_UFSMNT);
    619 	free(fs, M_UFSMNT);
    620 	free(ump, M_UFSMNT);
    621 	mp->mnt_data = (qaddr_t)0;
    622 	mp->mnt_flag &= ~MNT_LOCAL;
    623 	return (error);
    624 }
    625 
    626 /*
    627  * Flush out all the files in a filesystem.
    628  */
    629 int
    630 ffs_flushfiles(mp, flags, p)
    631 	register struct mount *mp;
    632 	int flags;
    633 	struct proc *p;
    634 {
    635 	extern int doforce;
    636 	register struct ufsmount *ump;
    637 	int error;
    638 
    639 	if (!doforce)
    640 		flags &= ~FORCECLOSE;
    641 	ump = VFSTOUFS(mp);
    642 #ifdef QUOTA
    643 	if (mp->mnt_flag & MNT_QUOTA) {
    644 		int i;
    645 		if ((error = vflush(mp, NULLVP, SKIPSYSTEM|flags)) != 0)
    646 			return (error);
    647 		for (i = 0; i < MAXQUOTAS; i++) {
    648 			if (ump->um_quotas[i] == NULLVP)
    649 				continue;
    650 			quotaoff(p, mp, i);
    651 		}
    652 		/*
    653 		 * Here we fall through to vflush again to ensure
    654 		 * that we have gotten rid of all the system vnodes.
    655 		 */
    656 	}
    657 #endif
    658 	error = vflush(mp, NULLVP, flags);
    659 	return (error);
    660 }
    661 
    662 /*
    663  * Get file system statistics.
    664  */
    665 int
    666 ffs_statfs(mp, sbp, p)
    667 	struct mount *mp;
    668 	register struct statfs *sbp;
    669 	struct proc *p;
    670 {
    671 	register struct ufsmount *ump;
    672 	register struct fs *fs;
    673 
    674 	ump = VFSTOUFS(mp);
    675 	fs = ump->um_fs;
    676 	if (fs->fs_magic != FS_MAGIC)
    677 		panic("ffs_statfs");
    678 #ifdef COMPAT_09
    679 	sbp->f_type = 1;
    680 #else
    681 	sbp->f_type = 0;
    682 #endif
    683 	sbp->f_bsize = fs->fs_fsize;
    684 	sbp->f_iosize = fs->fs_bsize;
    685 	sbp->f_blocks = fs->fs_dsize;
    686 	sbp->f_bfree = fs->fs_cstotal.cs_nbfree * fs->fs_frag +
    687 		fs->fs_cstotal.cs_nffree;
    688 	sbp->f_bavail = (long) (((u_int64_t) fs->fs_dsize * (u_int64_t)
    689 		(100 - fs->fs_minfree) / (u_int64_t) 100) -
    690 		(u_int64_t) (fs->fs_dsize - sbp->f_bfree));
    691 	sbp->f_files =  fs->fs_ncg * fs->fs_ipg - ROOTINO;
    692 	sbp->f_ffree = fs->fs_cstotal.cs_nifree;
    693 	if (sbp != &mp->mnt_stat) {
    694 		bcopy(mp->mnt_stat.f_mntonname, sbp->f_mntonname, MNAMELEN);
    695 		bcopy(mp->mnt_stat.f_mntfromname, sbp->f_mntfromname, MNAMELEN);
    696 	}
    697 	strncpy(sbp->f_fstypename, mp->mnt_op->vfs_name, MFSNAMELEN);
    698 	return (0);
    699 }
    700 
    701 /*
    702  * Go through the disk queues to initiate sandbagged IO;
    703  * go through the inodes to write those that have been modified;
    704  * initiate the writing of the super block if it has been modified.
    705  *
    706  * Note: we are always called with the filesystem marked `MPBUSY'.
    707  */
    708 int
    709 ffs_sync(mp, waitfor, cred, p)
    710 	struct mount *mp;
    711 	int waitfor;
    712 	struct ucred *cred;
    713 	struct proc *p;
    714 {
    715 	struct vnode *vp, *nvp;
    716 	struct inode *ip;
    717 	struct ufsmount *ump = VFSTOUFS(mp);
    718 	struct fs *fs;
    719 	int error, allerror = 0;
    720 
    721 	fs = ump->um_fs;
    722 	if (fs->fs_fmod != 0 && fs->fs_ronly != 0) {		/* XXX */
    723 		printf("fs = %s\n", fs->fs_fsmnt);
    724 		panic("update: rofs mod");
    725 	}
    726 	/*
    727 	 * Write back each (modified) inode.
    728 	 */
    729 	simple_lock(&mntvnode_slock);
    730 loop:
    731 	for (vp = mp->mnt_vnodelist.lh_first; vp != NULL; vp = nvp) {
    732 		/*
    733 		 * If the vnode that we are about to sync is no longer
    734 		 * associated with this mount point, start over.
    735 		 */
    736 		if (vp->v_mount != mp)
    737 			goto loop;
    738 		simple_lock(&vp->v_interlock);
    739 		nvp = vp->v_mntvnodes.le_next;
    740 		ip = VTOI(vp);
    741 		if ((ip->i_flag &
    742 		    (IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE)) == 0 &&
    743 		    vp->v_dirtyblkhd.lh_first == NULL) {
    744 			simple_unlock(&vp->v_interlock);
    745 			continue;
    746 		}
    747 		simple_unlock(&mntvnode_slock);
    748 		error = vget(vp, LK_EXCLUSIVE | LK_NOWAIT | LK_INTERLOCK);
    749 		if (error) {
    750 			simple_lock(&mntvnode_slock);
    751 			if (error == ENOENT)
    752 				goto loop;
    753 			continue;
    754 		}
    755 		if ((error = VOP_FSYNC(vp, cred, waitfor, p)) != 0)
    756 			allerror = error;
    757 		vput(vp);
    758 		simple_lock(&mntvnode_slock);
    759 	}
    760 	simple_unlock(&mntvnode_slock);
    761 	/*
    762 	 * Force stale file system control information to be flushed.
    763 	 */
    764 	if ((error = VOP_FSYNC(ump->um_devvp, cred, waitfor, p)) != 0)
    765 		allerror = error;
    766 #ifdef QUOTA
    767 	qsync(mp);
    768 #endif
    769 	/*
    770 	 * Write back modified superblock.
    771 	 */
    772 	if (fs->fs_fmod != 0) {
    773 		fs->fs_fmod = 0;
    774 		fs->fs_time = time.tv_sec;
    775 		allerror = ffs_cgupdate(ump, waitfor);
    776 	}
    777 	return (allerror);
    778 }
    779 
    780 /*
    781  * Look up a FFS dinode number to find its incore vnode, otherwise read it
    782  * in from disk.  If it is in core, wait for the lock bit to clear, then
    783  * return the inode locked.  Detection and handling of mount points must be
    784  * done by the calling routine.
    785  */
    786 int
    787 ffs_vget(mp, ino, vpp)
    788 	struct mount *mp;
    789 	ino_t ino;
    790 	struct vnode **vpp;
    791 {
    792 	struct fs *fs;
    793 	struct inode *ip;
    794 	struct ufsmount *ump;
    795 	struct buf *bp;
    796 	struct vnode *vp;
    797 	dev_t dev;
    798 	int type, error;
    799 
    800 	ump = VFSTOUFS(mp);
    801 	dev = ump->um_dev;
    802 	do {
    803 		if ((*vpp = ufs_ihashget(dev, ino)) != NULL)
    804 			return (0);
    805 	} while (lockmgr(&ufs_hashlock, LK_EXCLUSIVE|LK_SLEEPFAIL, 0));
    806 
    807 	/* Allocate a new vnode/inode. */
    808 	if ((error = getnewvnode(VT_UFS, mp, ffs_vnodeop_p, &vp)) != 0) {
    809 		*vpp = NULL;
    810 		lockmgr(&ufs_hashlock, LK_RELEASE, 0);
    811 		return (error);
    812 	}
    813 	type = ump->um_devvp->v_tag == VT_MFS ? M_MFSNODE : M_FFSNODE; /* XXX */
    814 	MALLOC(ip, struct inode *, sizeof(struct inode), type, M_WAITOK);
    815 	bzero((caddr_t)ip, sizeof(struct inode));
    816 	lockinit(&ip->i_lock, PINOD, "inode", 0, 0);
    817 	vp->v_data = ip;
    818 	ip->i_vnode = vp;
    819 	ip->i_fs = fs = ump->um_fs;
    820 	ip->i_dev = dev;
    821 	ip->i_number = ino;
    822 #ifdef QUOTA
    823 	{
    824 		int i;
    825 
    826 		for (i = 0; i < MAXQUOTAS; i++)
    827 			ip->i_dquot[i] = NODQUOT;
    828 	}
    829 #endif
    830 	/*
    831 	 * Put it onto its hash chain and lock it so that other requests for
    832 	 * this inode will block if they arrive while we are sleeping waiting
    833 	 * for old data structures to be purged or for the contents of the
    834 	 * disk portion of this inode to be read.
    835 	 */
    836 	ufs_ihashins(ip);
    837 	lockmgr(&ufs_hashlock, LK_RELEASE, 0);
    838 
    839 	/* Read in the disk contents for the inode, copy into the inode. */
    840 	error = bread(ump->um_devvp, fsbtodb(fs, ino_to_fsba(fs, ino)),
    841 		      (int)fs->fs_bsize, NOCRED, &bp);
    842 	if (error) {
    843 		/*
    844 		 * The inode does not contain anything useful, so it would
    845 		 * be misleading to leave it on its hash chain. With mode
    846 		 * still zero, it will be unlinked and returned to the free
    847 		 * list by vput().
    848 		 */
    849 		vput(vp);
    850 		brelse(bp);
    851 		*vpp = NULL;
    852 		return (error);
    853 	}
    854 	ip->i_din.ffs_din = *((struct dinode *)bp->b_data + ino_to_fsbo(fs, ino));
    855 	brelse(bp);
    856 
    857 	/*
    858 	 * Initialize the vnode from the inode, check for aliases.
    859 	 * Note that the underlying vnode may have changed.
    860 	 */
    861 	error = ufs_vinit(mp, ffs_specop_p, FFS_FIFOOPS, &vp);
    862 	if (error) {
    863 		vput(vp);
    864 		*vpp = NULL;
    865 		return (error);
    866 	}
    867 	/*
    868 	 * Finish inode initialization now that aliasing has been resolved.
    869 	 */
    870 	ip->i_devvp = ump->um_devvp;
    871 	VREF(ip->i_devvp);
    872 	/*
    873 	 * Ensure that uid and gid are correct. This is a temporary
    874 	 * fix until fsck has been changed to do the update.
    875 	 */
    876 	if (fs->fs_inodefmt < FS_44INODEFMT) {		/* XXX */
    877 		ip->i_ffs_uid = ip->i_din.ffs_din.di_ouid;		/* XXX */
    878 		ip->i_ffs_gid = ip->i_din.ffs_din.di_ogid;		/* XXX */
    879 	}						/* XXX */
    880 
    881 	*vpp = vp;
    882 	return (0);
    883 }
    884 
    885 /*
    886  * File handle to vnode
    887  *
    888  * Have to be really careful about stale file handles:
    889  * - check that the inode number is valid
    890  * - call ffs_vget() to get the locked inode
    891  * - check for an unallocated inode (i_mode == 0)
    892  * - check that the given client host has export rights and return
    893  *   those rights via. exflagsp and credanonp
    894  */
    895 int
    896 ffs_fhtovp(mp, fhp, nam, vpp, exflagsp, credanonp)
    897 	register struct mount *mp;
    898 	struct fid *fhp;
    899 	struct mbuf *nam;
    900 	struct vnode **vpp;
    901 	int *exflagsp;
    902 	struct ucred **credanonp;
    903 {
    904 	register struct ufid *ufhp;
    905 	struct fs *fs;
    906 
    907 	ufhp = (struct ufid *)fhp;
    908 	fs = VFSTOUFS(mp)->um_fs;
    909 	if (ufhp->ufid_ino < ROOTINO ||
    910 	    ufhp->ufid_ino >= fs->fs_ncg * fs->fs_ipg)
    911 		return (ESTALE);
    912 	return (ufs_check_export(mp, ufhp, nam, vpp, exflagsp, credanonp));
    913 }
    914 
    915 /*
    916  * Vnode pointer to File handle
    917  */
    918 /* ARGSUSED */
    919 int
    920 ffs_vptofh(vp, fhp)
    921 	struct vnode *vp;
    922 	struct fid *fhp;
    923 {
    924 	register struct inode *ip;
    925 	register struct ufid *ufhp;
    926 
    927 	ip = VTOI(vp);
    928 	ufhp = (struct ufid *)fhp;
    929 	ufhp->ufid_len = sizeof(struct ufid);
    930 	ufhp->ufid_ino = ip->i_number;
    931 	ufhp->ufid_gen = ip->i_ffs_gen;
    932 	return (0);
    933 }
    934 
    935 void
    936 ffs_init()
    937 {
    938 	ufs_init();
    939 }
    940 
    941 int
    942 ffs_sysctl(name, namelen, oldp, oldlenp, newp, newlen, p)
    943 	int *name;
    944 	u_int namelen;
    945 	void *oldp;
    946 	size_t *oldlenp;
    947 	void *newp;
    948 	size_t newlen;
    949 	struct proc *p;
    950 {
    951 	extern int doclusterread, doclusterwrite, doreallocblks, doasyncfree;
    952 
    953 	/* all sysctl names at this level are terminal */
    954 	if (namelen != 1)
    955 		return (ENOTDIR);		/* overloaded */
    956 
    957 	switch (name[0]) {
    958 	case FFS_CLUSTERREAD:
    959 		return (sysctl_int(oldp, oldlenp, newp, newlen,
    960 		    &doclusterread));
    961 	case FFS_CLUSTERWRITE:
    962 		return (sysctl_int(oldp, oldlenp, newp, newlen,
    963 		    &doclusterwrite));
    964 	case FFS_REALLOCBLKS:
    965 		return (sysctl_int(oldp, oldlenp, newp, newlen,
    966 		    &doreallocblks));
    967 	case FFS_ASYNCFREE:
    968 		return (sysctl_int(oldp, oldlenp, newp, newlen, &doasyncfree));
    969 	default:
    970 		return (EOPNOTSUPP);
    971 	}
    972 	/* NOTREACHED */
    973 }
    974 
    975 /*
    976  * Write a superblock and associated information back to disk.
    977  */
    978 int
    979 ffs_sbupdate(mp, waitfor)
    980 	struct ufsmount *mp;
    981 	int waitfor;
    982 {
    983 	register struct fs *dfs, *fs = mp->um_fs;
    984 	register struct buf *bp;
    985 	int i, error = 0;
    986 
    987 	bp = getblk(mp->um_devvp, SBOFF >> (fs->fs_fshift - fs->fs_fsbtodb),
    988 	    (int)fs->fs_sbsize, 0, 0);
    989 	bcopy((caddr_t)fs, bp->b_data, (u_int)fs->fs_sbsize);
    990 	/* Restore compatibility to old file systems.		   XXX */
    991 	dfs = (struct fs *)bp->b_data;				/* XXX */
    992 	if (fs->fs_postblformat == FS_42POSTBLFMT)		/* XXX */
    993 		dfs->fs_nrpos = -1;				/* XXX */
    994 	if (fs->fs_inodefmt < FS_44INODEFMT) {			/* XXX */
    995 		int32_t *lp, tmp;				/* XXX */
    996 								/* XXX */
    997 		lp = (int32_t *)&dfs->fs_qbmask;		/* XXX */
    998 		tmp = lp[4];					/* XXX */
    999 		for (i = 4; i > 0; i--)				/* XXX */
   1000 			lp[i] = lp[i-1];			/* XXX */
   1001 		lp[0] = tmp;					/* XXX */
   1002 	}							/* XXX */
   1003 	dfs->fs_maxfilesize = mp->um_savedmaxfilesize;		/* XXX */
   1004 	if (waitfor == MNT_WAIT)
   1005 		error = bwrite(bp);
   1006 	else
   1007 		bawrite(bp);
   1008 	return (error);
   1009 }
   1010 
   1011 int
   1012 ffs_cgupdate(mp, waitfor)
   1013 	struct ufsmount *mp;
   1014 	int waitfor;
   1015 {
   1016 	register struct fs *fs = mp->um_fs;
   1017 	register struct buf *bp;
   1018 	int blks;
   1019 	caddr_t space;
   1020 	int i, size, error = 0, allerror = 0;
   1021 
   1022 	allerror = ffs_sbupdate(mp, waitfor);
   1023 	blks = howmany(fs->fs_cssize, fs->fs_fsize);
   1024 	space = (caddr_t)fs->fs_csp[0];
   1025 	for (i = 0; i < blks; i += fs->fs_frag) {
   1026 		size = fs->fs_bsize;
   1027 		if (i + fs->fs_frag > blks)
   1028 			size = (blks - i) * fs->fs_fsize;
   1029 		bp = getblk(mp->um_devvp, fsbtodb(fs, fs->fs_csaddr + i),
   1030 		    size, 0, 0);
   1031 		bcopy(space, bp->b_data, (u_int)size);
   1032 		space += size;
   1033 		if (waitfor == MNT_WAIT)
   1034 			error = bwrite(bp);
   1035 		else
   1036 			bawrite(bp);
   1037 	}
   1038 	if (!allerror && error)
   1039 		allerror = error;
   1040 	return (allerror);
   1041 }
   1042