Home | History | Annotate | Line # | Download | only in ffs
ffs_vfsops.c revision 1.18
      1 /*	$NetBSD: ffs_vfsops.c,v 1.18 1995/12/19 23:27:53 cgd 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	8.14 (Berkeley) 11/28/94
     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/mbuf.h>
     48 #include <sys/file.h>
     49 #include <sys/disklabel.h>
     50 #include <sys/ioctl.h>
     51 #include <sys/errno.h>
     52 #include <sys/malloc.h>
     53 
     54 #include <miscfs/specfs/specdev.h>
     55 
     56 #include <ufs/ufs/quota.h>
     57 #include <ufs/ufs/ufsmount.h>
     58 #include <ufs/ufs/inode.h>
     59 #include <ufs/ufs/ufs_extern.h>
     60 
     61 #include <ufs/ffs/fs.h>
     62 #include <ufs/ffs/ffs_extern.h>
     63 
     64 int ffs_sbupdate __P((struct ufsmount *, int));
     65 
     66 struct vfsops ffs_vfsops = {
     67 	MOUNT_FFS,
     68 	ffs_mount,
     69 	ufs_start,
     70 	ffs_unmount,
     71 	ufs_root,
     72 	ufs_quotactl,
     73 	ffs_statfs,
     74 	ffs_sync,
     75 	ffs_vget,
     76 	ffs_fhtovp,
     77 	ffs_vptofh,
     78 	ffs_init,
     79 };
     80 
     81 extern u_long nextgennumber;
     82 
     83 /*
     84  * Called by main() when ufs is going to be mounted as root.
     85  *
     86  * Name is updated by mount(8) after booting.
     87  */
     88 #define ROOTNAME	"root_device"
     89 
     90 ffs_mountroot()
     91 {
     92 	extern struct vnode *rootvp;
     93 	register struct fs *fs;
     94 	register struct mount *mp;
     95 	struct proc *p = curproc;	/* XXX */
     96 	struct ufsmount *ump;
     97 	size_t size;
     98 	int error;
     99 
    100 	/*
    101 	 * Get vnodes for swapdev and rootdev.
    102 	 */
    103 	if (bdevvp(swapdev, &swapdev_vp) || bdevvp(rootdev, &rootvp))
    104 		panic("ffs_mountroot: can't setup bdevvp's");
    105 
    106 	mp = malloc((u_long)sizeof(struct mount), M_MOUNT, M_WAITOK);
    107 	bzero((char *)mp, (u_long)sizeof(struct mount));
    108 	mp->mnt_op = &ffs_vfsops;
    109 	mp->mnt_flag = MNT_RDONLY;
    110 	if (error = ffs_mountfs(rootvp, mp, p)) {
    111 		free(mp, M_MOUNT);
    112 		return (error);
    113 	}
    114 	if (error = vfs_lock(mp)) {
    115 		(void)ffs_unmount(mp, 0, p);
    116 		free(mp, M_MOUNT);
    117 		return (error);
    118 	}
    119 	CIRCLEQ_INSERT_TAIL(&mountlist, mp, mnt_list);
    120 	mp->mnt_vnodecovered = NULLVP;
    121 	ump = VFSTOUFS(mp);
    122 	fs = ump->um_fs;
    123 	bzero(fs->fs_fsmnt, sizeof(fs->fs_fsmnt));
    124 	fs->fs_fsmnt[0] = '/';
    125 	bcopy(fs->fs_fsmnt, mp->mnt_stat.f_mntonname, MNAMELEN);
    126 	(void) copystr(ROOTNAME, mp->mnt_stat.f_mntfromname, MNAMELEN - 1,
    127 	    &size);
    128 	bzero(mp->mnt_stat.f_mntfromname + size, MNAMELEN - size);
    129 	(void)ffs_statfs(mp, &mp->mnt_stat, p);
    130 	vfs_unlock(mp);
    131 	inittodr(fs->fs_time);
    132 	return (0);
    133 }
    134 
    135 /*
    136  * VFS Operations.
    137  *
    138  * mount system call
    139  */
    140 int
    141 ffs_mount(mp, path, data, ndp, p)
    142 	register struct mount *mp;
    143 	char *path;
    144 	caddr_t data;
    145 	struct nameidata *ndp;
    146 	struct proc *p;
    147 {
    148 	struct vnode *devvp;
    149 	struct ufs_args args;
    150 	struct ufsmount *ump;
    151 	register struct fs *fs;
    152 	size_t size;
    153 	int error, flags;
    154 	mode_t accessmode;
    155 
    156 	if (error = copyin(data, (caddr_t)&args, sizeof (struct ufs_args)))
    157 		return (error);
    158 	/*
    159 	 * If updating, check whether changing from read-only to
    160 	 * read/write; if there is no device name, that's all we do.
    161 	 */
    162 	if (mp->mnt_flag & MNT_UPDATE) {
    163 		ump = VFSTOUFS(mp);
    164 		fs = ump->um_fs;
    165 		if (fs->fs_ronly == 0 && (mp->mnt_flag & MNT_RDONLY)) {
    166 			flags = WRITECLOSE;
    167 			if (mp->mnt_flag & MNT_FORCE)
    168 				flags |= FORCECLOSE;
    169 			if (vfs_busy(mp))
    170 				return (EBUSY);
    171 			error = ffs_flushfiles(mp, flags, p);
    172 			if (error == 0 &&
    173 			    ffs_cgupdate(ump, MNT_WAIT) == 0 &&
    174 			    fs->fs_clean & FS_WASCLEAN) {
    175 				fs->fs_clean = FS_ISCLEAN;
    176 				(void) ffs_sbupdate(ump, MNT_WAIT);
    177 			}
    178 			vfs_unbusy(mp);
    179 			if (error)
    180 				return (error);
    181 			fs->fs_ronly = 1;
    182 		}
    183 		if (mp->mnt_flag & MNT_RELOAD) {
    184 			error = ffs_reload(mp, ndp->ni_cnd.cn_cred, p);
    185 			if (error)
    186 				return (error);
    187 		}
    188 		if (fs->fs_ronly && (mp->mnt_flag & MNT_WANTRDWR)) {
    189 			/*
    190 			 * If upgrade to read-write by non-root, then verify
    191 			 * that user has necessary permissions on the device.
    192 			 */
    193 			if (p->p_ucred->cr_uid != 0) {
    194 				devvp = ump->um_devvp;
    195 				VOP_LOCK(devvp);
    196 				if (error = VOP_ACCESS(devvp, VREAD | VWRITE,
    197 				    p->p_ucred, p)) {
    198 					VOP_UNLOCK(devvp);
    199 					return (error);
    200 				}
    201 				VOP_UNLOCK(devvp);
    202 			}
    203 			fs->fs_ronly = 0;
    204 			fs->fs_clean <<= 1;
    205 			fs->fs_fmod = 1;
    206 		}
    207 		if (args.fspec == 0) {
    208 			/*
    209 			 * Process export requests.
    210 			 */
    211 			return (vfs_export(mp, &ump->um_export, &args.export));
    212 		}
    213 	}
    214 	/*
    215 	 * Not an update, or updating the name: look up the name
    216 	 * and verify that it refers to a sensible block device.
    217 	 */
    218 	NDINIT(ndp, LOOKUP, FOLLOW, UIO_USERSPACE, args.fspec, p);
    219 	if (error = namei(ndp))
    220 		return (error);
    221 	devvp = ndp->ni_vp;
    222 
    223 	if (devvp->v_type != VBLK) {
    224 		vrele(devvp);
    225 		return (ENOTBLK);
    226 	}
    227 	if (major(devvp->v_rdev) >= nblkdev) {
    228 		vrele(devvp);
    229 		return (ENXIO);
    230 	}
    231 	/*
    232 	 * If mount by non-root, then verify that user has necessary
    233 	 * permissions on the device.
    234 	 */
    235 	if (p->p_ucred->cr_uid != 0) {
    236 		accessmode = VREAD;
    237 		if ((mp->mnt_flag & MNT_RDONLY) == 0)
    238 			accessmode |= VWRITE;
    239 		VOP_LOCK(devvp);
    240 		if (error = VOP_ACCESS(devvp, accessmode, p->p_ucred, p)) {
    241 			vput(devvp);
    242 			return (error);
    243 		}
    244 		VOP_UNLOCK(devvp);
    245 	}
    246 	if ((mp->mnt_flag & MNT_UPDATE) == 0)
    247 		error = ffs_mountfs(devvp, mp, p);
    248 	else {
    249 		if (devvp != ump->um_devvp)
    250 			error = EINVAL;	/* needs translation */
    251 		else
    252 			vrele(devvp);
    253 	}
    254 	if (error) {
    255 		vrele(devvp);
    256 		return (error);
    257 	}
    258 	ump = VFSTOUFS(mp);
    259 	fs = ump->um_fs;
    260 	(void) copyinstr(path, fs->fs_fsmnt, sizeof(fs->fs_fsmnt) - 1, &size);
    261 	bzero(fs->fs_fsmnt + size, sizeof(fs->fs_fsmnt) - size);
    262 	bcopy(fs->fs_fsmnt, mp->mnt_stat.f_mntonname, MNAMELEN);
    263 	(void) copyinstr(args.fspec, mp->mnt_stat.f_mntfromname, MNAMELEN - 1,
    264 	    &size);
    265 	bzero(mp->mnt_stat.f_mntfromname + size, MNAMELEN - size);
    266 	if (fs->fs_fmod != 0) {	/* XXX */
    267 		fs->fs_fmod = 0;
    268 		if (fs->fs_clean & FS_WASCLEAN)
    269 			fs->fs_time = time.tv_sec;
    270 		else
    271 			printf("%s: file system not clean; please fsck(8)\n",
    272 			    mp->mnt_stat.f_mntfromname);
    273 		(void) ffs_cgupdate(ump, MNT_WAIT);
    274 	}
    275 	return (0);
    276 }
    277 
    278 /*
    279  * Reload all incore data for a filesystem (used after running fsck on
    280  * the root filesystem and finding things to fix). The filesystem must
    281  * be mounted read-only.
    282  *
    283  * Things to do to update the mount:
    284  *	1) invalidate all cached meta-data.
    285  *	2) re-read superblock from disk.
    286  *	3) re-read summary information from disk.
    287  *	4) invalidate all inactive vnodes.
    288  *	5) invalidate all cached file data.
    289  *	6) re-read inode data for all active vnodes.
    290  */
    291 ffs_reload(mountp, cred, p)
    292 	register struct mount *mountp;
    293 	struct ucred *cred;
    294 	struct proc *p;
    295 {
    296 	register struct vnode *vp, *nvp, *devvp;
    297 	struct inode *ip;
    298 	struct csum *space;
    299 	struct buf *bp;
    300 	struct fs *fs, *newfs;
    301 	struct partinfo dpart;
    302 	int i, blks, size, error;
    303 	int32_t *lp;
    304 
    305 	if ((mountp->mnt_flag & MNT_RDONLY) == 0)
    306 		return (EINVAL);
    307 	/*
    308 	 * Step 1: invalidate all cached meta-data.
    309 	 */
    310 	devvp = VFSTOUFS(mountp)->um_devvp;
    311 	if (vinvalbuf(devvp, 0, cred, p, 0, 0))
    312 		panic("ffs_reload: dirty1");
    313 	/*
    314 	 * Step 2: re-read superblock from disk.
    315 	 */
    316 	if (VOP_IOCTL(devvp, DIOCGPART, (caddr_t)&dpart, FREAD, NOCRED, p) != 0)
    317 		size = DEV_BSIZE;
    318 	else
    319 		size = dpart.disklab->d_secsize;
    320 	if (error = bread(devvp, (daddr_t)(SBOFF / size), SBSIZE, NOCRED, &bp))
    321 		return (error);
    322 	newfs = (struct fs *)bp->b_data;
    323 	if (newfs->fs_magic != FS_MAGIC || newfs->fs_bsize > MAXBSIZE ||
    324 	    newfs->fs_bsize < sizeof(struct fs)) {
    325 		brelse(bp);
    326 		return (EIO);		/* XXX needs translation */
    327 	}
    328 	fs = VFSTOUFS(mountp)->um_fs;
    329 	/*
    330 	 * Copy pointer fields back into superblock before copying in	XXX
    331 	 * new superblock. These should really be in the ufsmount.	XXX
    332 	 * Note that important parameters (eg fs_ncg) are unchanged.
    333 	 */
    334 	bcopy(&fs->fs_csp[0], &newfs->fs_csp[0], sizeof(fs->fs_csp));
    335 	newfs->fs_maxcluster = fs->fs_maxcluster;
    336 	bcopy(newfs, fs, (u_int)fs->fs_sbsize);
    337 	if (fs->fs_sbsize < SBSIZE)
    338 		bp->b_flags |= B_INVAL;
    339 	brelse(bp);
    340 	mountp->mnt_maxsymlinklen = fs->fs_maxsymlinklen;
    341 	ffs_oldfscompat(fs);
    342 	/*
    343 	 * Step 3: re-read summary information from disk.
    344 	 */
    345 	blks = howmany(fs->fs_cssize, fs->fs_fsize);
    346 	space = fs->fs_csp[0];
    347 	for (i = 0; i < blks; i += fs->fs_frag) {
    348 		size = fs->fs_bsize;
    349 		if (i + fs->fs_frag > blks)
    350 			size = (blks - i) * fs->fs_fsize;
    351 		if (error = bread(devvp, fsbtodb(fs, fs->fs_csaddr + i), size,
    352 		    NOCRED, &bp))
    353 			return (error);
    354 		bcopy(bp->b_data, fs->fs_csp[fragstoblks(fs, i)], (u_int)size);
    355 		brelse(bp);
    356 	}
    357 	/*
    358 	 * We no longer know anything about clusters per cylinder group.
    359 	 */
    360 	if (fs->fs_contigsumsize > 0) {
    361 		lp = fs->fs_maxcluster;
    362 		for (i = 0; i < fs->fs_ncg; i++)
    363 			*lp++ = fs->fs_contigsumsize;
    364 	}
    365 
    366 loop:
    367 	for (vp = mountp->mnt_vnodelist.lh_first; vp != NULL; vp = nvp) {
    368 		nvp = vp->v_mntvnodes.le_next;
    369 		/*
    370 		 * Step 4: invalidate all inactive vnodes.
    371 		 */
    372 		if (vp->v_usecount == 0) {
    373 			vgone(vp);
    374 			continue;
    375 		}
    376 		/*
    377 		 * Step 5: invalidate all cached file data.
    378 		 */
    379 		if (vget(vp, 1))
    380 			goto loop;
    381 		if (vinvalbuf(vp, 0, cred, p, 0, 0))
    382 			panic("ffs_reload: dirty2");
    383 		/*
    384 		 * Step 6: re-read inode data for all active vnodes.
    385 		 */
    386 		ip = VTOI(vp);
    387 		if (error =
    388 		    bread(devvp, fsbtodb(fs, ino_to_fsba(fs, ip->i_number)),
    389 		    (int)fs->fs_bsize, NOCRED, &bp)) {
    390 			vput(vp);
    391 			return (error);
    392 		}
    393 		ip->i_din = *((struct dinode *)bp->b_data +
    394 		    ino_to_fsbo(fs, ip->i_number));
    395 		brelse(bp);
    396 		vput(vp);
    397 		if (vp->v_mount != mountp)
    398 			goto loop;
    399 	}
    400 	return (0);
    401 }
    402 
    403 /*
    404  * Common code for mount and mountroot
    405  */
    406 int
    407 ffs_mountfs(devvp, mp, p)
    408 	register struct vnode *devvp;
    409 	struct mount *mp;
    410 	struct proc *p;
    411 {
    412 	register struct ufsmount *ump;
    413 	struct buf *bp;
    414 	register struct fs *fs;
    415 	dev_t dev;
    416 	struct partinfo dpart;
    417 	caddr_t base, space;
    418 	int blks;
    419 	int error, i, size, ronly;
    420 	int32_t *lp;
    421 	struct ucred *cred;
    422 	extern struct vnode *rootvp;
    423 	u_int64_t maxfilesize;					/* XXX */
    424 
    425 	dev = devvp->v_rdev;
    426 	cred = p ? p->p_ucred : NOCRED;
    427 	/*
    428 	 * Disallow multiple mounts of the same device.
    429 	 * Disallow mounting of a device that is currently in use
    430 	 * (except for root, which might share swap device for miniroot).
    431 	 * Flush out any old buffers remaining from a previous use.
    432 	 */
    433 	if (error = vfs_mountedon(devvp))
    434 		return (error);
    435 	if (vcount(devvp) > 1 && devvp != rootvp)
    436 		return (EBUSY);
    437 	if (error = vinvalbuf(devvp, V_SAVE, cred, p, 0, 0))
    438 		return (error);
    439 
    440 	ronly = (mp->mnt_flag & MNT_RDONLY) != 0;
    441 	if (error = VOP_OPEN(devvp, ronly ? FREAD : FREAD|FWRITE, FSCRED, p))
    442 		return (error);
    443 	if (VOP_IOCTL(devvp, DIOCGPART, (caddr_t)&dpart, FREAD, cred, p) != 0)
    444 		size = DEV_BSIZE;
    445 	else
    446 		size = dpart.disklab->d_secsize;
    447 
    448 	bp = NULL;
    449 	ump = NULL;
    450 	if (error = bread(devvp, (daddr_t)(SBOFF / size), SBSIZE, cred, &bp))
    451 		goto out;
    452 	fs = (struct fs *)bp->b_data;
    453 	if (fs->fs_magic != FS_MAGIC || fs->fs_bsize > MAXBSIZE ||
    454 	    fs->fs_bsize < sizeof(struct fs)) {
    455 		error = EINVAL;		/* XXX needs translation */
    456 		goto out;
    457 	}
    458 	/* XXX updating 4.2 FFS superblocks trashes rotational layout tables */
    459 	if (fs->fs_postblformat == FS_42POSTBLFMT && !ronly) {
    460 		error = EROFS;		/* XXX what should be returned? */
    461 		goto out;
    462 	}
    463 	ump = malloc(sizeof *ump, M_UFSMNT, M_WAITOK);
    464 	bzero((caddr_t)ump, sizeof *ump);
    465 	ump->um_fs = malloc((u_long)fs->fs_sbsize, M_UFSMNT,
    466 	    M_WAITOK);
    467 	bcopy(bp->b_data, ump->um_fs, (u_int)fs->fs_sbsize);
    468 	if (fs->fs_sbsize < SBSIZE)
    469 		bp->b_flags |= B_INVAL;
    470 	brelse(bp);
    471 	bp = NULL;
    472 	fs = ump->um_fs;
    473 	fs->fs_ronly = ronly;
    474 	if (ronly == 0) {
    475 		fs->fs_clean <<= 1;
    476 		fs->fs_fmod = 1;
    477 	}
    478 	size = fs->fs_cssize;
    479 	blks = howmany(size, fs->fs_fsize);
    480 	if (fs->fs_contigsumsize > 0)
    481 		size += fs->fs_ncg * sizeof(int32_t);
    482 	base = space = malloc((u_long)size, M_UFSMNT, M_WAITOK);
    483 	for (i = 0; i < blks; i += fs->fs_frag) {
    484 		size = fs->fs_bsize;
    485 		if (i + fs->fs_frag > blks)
    486 			size = (blks - i) * fs->fs_fsize;
    487 		if (error = bread(devvp, fsbtodb(fs, fs->fs_csaddr + i), size,
    488 		    cred, &bp)) {
    489 			free(base, M_UFSMNT);
    490 			goto out;
    491 		}
    492 		bcopy(bp->b_data, space, (u_int)size);
    493 		fs->fs_csp[fragstoblks(fs, i)] = (struct csum *)space;
    494 		space += size;
    495 		brelse(bp);
    496 		bp = NULL;
    497 	}
    498 	if (fs->fs_contigsumsize > 0) {
    499 		fs->fs_maxcluster = lp = (int32_t *)space;
    500 		for (i = 0; i < fs->fs_ncg; i++)
    501 			*lp++ = fs->fs_contigsumsize;
    502 	}
    503 	mp->mnt_data = (qaddr_t)ump;
    504 	mp->mnt_stat.f_fsid.val[0] = (long)dev;
    505 	mp->mnt_stat.f_fsid.val[1] = makefstype(MOUNT_FFS);
    506 	mp->mnt_maxsymlinklen = fs->fs_maxsymlinklen;
    507 	mp->mnt_flag |= MNT_LOCAL;
    508 	ump->um_mountp = mp;
    509 	ump->um_dev = dev;
    510 	ump->um_devvp = devvp;
    511 	ump->um_nindir = fs->fs_nindir;
    512 	ump->um_bptrtodb = fs->fs_fsbtodb;
    513 	ump->um_seqinc = fs->fs_frag;
    514 	for (i = 0; i < MAXQUOTAS; i++)
    515 		ump->um_quotas[i] = NULLVP;
    516 	devvp->v_specflags |= SI_MOUNTEDON;
    517 	ffs_oldfscompat(fs);
    518 	ump->um_savedmaxfilesize = fs->fs_maxfilesize;		/* XXX */
    519 	maxfilesize = (u_int64_t)0x80000000 * fs->fs_bsize - 1;	/* XXX */
    520 	if (fs->fs_maxfilesize > maxfilesize)			/* XXX */
    521 		fs->fs_maxfilesize = maxfilesize;		/* XXX */
    522 	return (0);
    523 out:
    524 	if (bp)
    525 		brelse(bp);
    526 	(void)VOP_CLOSE(devvp, ronly ? FREAD : FREAD|FWRITE, cred, p);
    527 	if (ump) {
    528 		free(ump->um_fs, M_UFSMNT);
    529 		free(ump, M_UFSMNT);
    530 		mp->mnt_data = (qaddr_t)0;
    531 	}
    532 	return (error);
    533 }
    534 
    535 /*
    536  * Sanity checks for old file systems.
    537  *
    538  * XXX - goes away some day.
    539  */
    540 ffs_oldfscompat(fs)
    541 	struct fs *fs;
    542 {
    543 	int i;
    544 
    545 	fs->fs_npsect = max(fs->fs_npsect, fs->fs_nsect);	/* XXX */
    546 	fs->fs_interleave = max(fs->fs_interleave, 1);		/* XXX */
    547 	if (fs->fs_postblformat == FS_42POSTBLFMT)		/* XXX */
    548 		fs->fs_nrpos = 8;				/* XXX */
    549 	if (fs->fs_inodefmt < FS_44INODEFMT) {			/* XXX */
    550 		u_int64_t sizepb = fs->fs_bsize;		/* XXX */
    551 								/* XXX */
    552 		fs->fs_maxfilesize = fs->fs_bsize * NDADDR - 1;	/* XXX */
    553 		for (i = 0; i < NIADDR; i++) {			/* XXX */
    554 			sizepb *= NINDIR(fs);			/* XXX */
    555 			fs->fs_maxfilesize += sizepb;		/* XXX */
    556 		}						/* XXX */
    557 		fs->fs_qbmask = ~fs->fs_bmask;			/* XXX */
    558 		fs->fs_qfmask = ~fs->fs_fmask;			/* XXX */
    559 	}							/* XXX */
    560 	return (0);
    561 }
    562 
    563 /*
    564  * unmount system call
    565  */
    566 int
    567 ffs_unmount(mp, mntflags, p)
    568 	struct mount *mp;
    569 	int mntflags;
    570 	struct proc *p;
    571 {
    572 	register struct ufsmount *ump;
    573 	register struct fs *fs;
    574 	int error, flags;
    575 
    576 	flags = 0;
    577 	if (mntflags & MNT_FORCE)
    578 		flags |= FORCECLOSE;
    579 	if (error = ffs_flushfiles(mp, flags, p))
    580 		return (error);
    581 	ump = VFSTOUFS(mp);
    582 	fs = ump->um_fs;
    583 	if (fs->fs_ronly == 0 &&
    584 	    ffs_cgupdate(ump, MNT_WAIT) == 0 &&
    585 	    fs->fs_clean & FS_WASCLEAN) {
    586 		fs->fs_clean = FS_ISCLEAN;
    587 		(void) ffs_sbupdate(ump, MNT_WAIT);
    588 	}
    589 	ump->um_devvp->v_specflags &= ~SI_MOUNTEDON;
    590 	error = VOP_CLOSE(ump->um_devvp, fs->fs_ronly ? FREAD : FREAD|FWRITE,
    591 		NOCRED, p);
    592 	vrele(ump->um_devvp);
    593 	free(fs->fs_csp[0], M_UFSMNT);
    594 	free(fs, M_UFSMNT);
    595 	free(ump, M_UFSMNT);
    596 	mp->mnt_data = (qaddr_t)0;
    597 	mp->mnt_flag &= ~MNT_LOCAL;
    598 	return (error);
    599 }
    600 
    601 /*
    602  * Flush out all the files in a filesystem.
    603  */
    604 ffs_flushfiles(mp, flags, p)
    605 	register struct mount *mp;
    606 	int flags;
    607 	struct proc *p;
    608 {
    609 	extern int doforce;
    610 	register struct ufsmount *ump;
    611 	int i, error;
    612 
    613 	if (!doforce)
    614 		flags &= ~FORCECLOSE;
    615 	ump = VFSTOUFS(mp);
    616 #ifdef QUOTA
    617 	if (mp->mnt_flag & MNT_QUOTA) {
    618 		if (error = vflush(mp, NULLVP, SKIPSYSTEM|flags))
    619 			return (error);
    620 		for (i = 0; i < MAXQUOTAS; i++) {
    621 			if (ump->um_quotas[i] == NULLVP)
    622 				continue;
    623 			quotaoff(p, mp, i);
    624 		}
    625 		/*
    626 		 * Here we fall through to vflush again to ensure
    627 		 * that we have gotten rid of all the system vnodes.
    628 		 */
    629 	}
    630 #endif
    631 	error = vflush(mp, NULLVP, flags);
    632 	return (error);
    633 }
    634 
    635 /*
    636  * Get file system statistics.
    637  */
    638 int
    639 ffs_statfs(mp, sbp, p)
    640 	struct mount *mp;
    641 	register struct statfs *sbp;
    642 	struct proc *p;
    643 {
    644 	register struct ufsmount *ump;
    645 	register struct fs *fs;
    646 
    647 	ump = VFSTOUFS(mp);
    648 	fs = ump->um_fs;
    649 	if (fs->fs_magic != FS_MAGIC)
    650 		panic("ffs_statfs");
    651 #ifdef COMPAT_09
    652 	sbp->f_type = 1;
    653 #else
    654 	sbp->f_type = 0;
    655 #endif
    656 	sbp->f_bsize = fs->fs_fsize;
    657 	sbp->f_iosize = fs->fs_bsize;
    658 	sbp->f_blocks = fs->fs_dsize;
    659 	sbp->f_bfree = fs->fs_cstotal.cs_nbfree * fs->fs_frag +
    660 		fs->fs_cstotal.cs_nffree;
    661 	sbp->f_bavail = (fs->fs_dsize * (100 - fs->fs_minfree) / 100) -
    662 		(fs->fs_dsize - sbp->f_bfree);
    663 	sbp->f_files =  fs->fs_ncg * fs->fs_ipg - ROOTINO;
    664 	sbp->f_ffree = fs->fs_cstotal.cs_nifree;
    665 	if (sbp != &mp->mnt_stat) {
    666 		bcopy(mp->mnt_stat.f_mntonname, sbp->f_mntonname, MNAMELEN);
    667 		bcopy(mp->mnt_stat.f_mntfromname, sbp->f_mntfromname, MNAMELEN);
    668 	}
    669 	strncpy(sbp->f_fstypename, mp->mnt_op->vfs_name, MFSNAMELEN);
    670 	return (0);
    671 }
    672 
    673 /*
    674  * Go through the disk queues to initiate sandbagged IO;
    675  * go through the inodes to write those that have been modified;
    676  * initiate the writing of the super block if it has been modified.
    677  *
    678  * Note: we are always called with the filesystem marked `MPBUSY'.
    679  */
    680 int
    681 ffs_sync(mp, waitfor, cred, p)
    682 	struct mount *mp;
    683 	int waitfor;
    684 	struct ucred *cred;
    685 	struct proc *p;
    686 {
    687 	register struct vnode *vp;
    688 	register struct inode *ip;
    689 	register struct ufsmount *ump = VFSTOUFS(mp);
    690 	register struct fs *fs;
    691 	int error, allerror = 0;
    692 
    693 	fs = ump->um_fs;
    694 	/*
    695 	 * Write back modified superblock.
    696 	 * Consistency check that the superblock
    697 	 * is still in the buffer cache.
    698 	 */
    699 	if (fs->fs_fmod != 0) {
    700 		if (fs->fs_ronly != 0) {		/* XXX */
    701 			printf("fs = %s\n", fs->fs_fsmnt);
    702 			panic("update: rofs mod");
    703 		}
    704 		fs->fs_fmod = 0;
    705 		fs->fs_time = time.tv_sec;
    706 		allerror = ffs_cgupdate(ump, waitfor);
    707 	}
    708 	/*
    709 	 * Write back each (modified) inode.
    710 	 */
    711 loop:
    712 	for (vp = mp->mnt_vnodelist.lh_first;
    713 	     vp != NULL;
    714 	     vp = vp->v_mntvnodes.le_next) {
    715 		/*
    716 		 * If the vnode that we are about to sync is no longer
    717 		 * associated with this mount point, start over.
    718 		 */
    719 		if (vp->v_mount != mp)
    720 			goto loop;
    721 		if (VOP_ISLOCKED(vp))
    722 			continue;
    723 		ip = VTOI(vp);
    724 		if ((ip->i_flag &
    725 		    (IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE)) == 0 &&
    726 		    vp->v_dirtyblkhd.lh_first == NULL)
    727 			continue;
    728 		if (vget(vp, 1))
    729 			goto loop;
    730 		if (error = VOP_FSYNC(vp, cred, waitfor, p))
    731 			allerror = error;
    732 		vput(vp);
    733 	}
    734 	/*
    735 	 * Force stale file system control information to be flushed.
    736 	 */
    737 	if (error = VOP_FSYNC(ump->um_devvp, cred, waitfor, p))
    738 		allerror = error;
    739 #ifdef QUOTA
    740 	qsync(mp);
    741 #endif
    742 	return (allerror);
    743 }
    744 
    745 /*
    746  * Look up a FFS dinode number to find its incore vnode, otherwise read it
    747  * in from disk.  If it is in core, wait for the lock bit to clear, then
    748  * return the inode locked.  Detection and handling of mount points must be
    749  * done by the calling routine.
    750  */
    751 int
    752 ffs_vget(mp, ino, vpp)
    753 	struct mount *mp;
    754 	ino_t ino;
    755 	struct vnode **vpp;
    756 {
    757 	register struct fs *fs;
    758 	register struct inode *ip;
    759 	struct ufsmount *ump;
    760 	struct buf *bp;
    761 	struct vnode *vp;
    762 	dev_t dev;
    763 	int i, type, error;
    764 
    765 	ump = VFSTOUFS(mp);
    766 	dev = ump->um_dev;
    767 	if ((*vpp = ufs_ihashget(dev, ino)) != NULL)
    768 		return (0);
    769 
    770 	/* Allocate a new vnode/inode. */
    771 	if (error = getnewvnode(VT_UFS, mp, ffs_vnodeop_p, &vp)) {
    772 		*vpp = NULL;
    773 		return (error);
    774 	}
    775 	type = ump->um_devvp->v_tag == VT_MFS ? M_MFSNODE : M_FFSNODE; /* XXX */
    776 	MALLOC(ip, struct inode *, sizeof(struct inode), type, M_WAITOK);
    777 	bzero((caddr_t)ip, sizeof(struct inode));
    778 	vp->v_data = ip;
    779 	ip->i_vnode = vp;
    780 	ip->i_fs = fs = ump->um_fs;
    781 	ip->i_dev = dev;
    782 	ip->i_number = ino;
    783 #ifdef QUOTA
    784 	for (i = 0; i < MAXQUOTAS; i++)
    785 		ip->i_dquot[i] = NODQUOT;
    786 #endif
    787 	/*
    788 	 * Put it onto its hash chain and lock it so that other requests for
    789 	 * this inode will block if they arrive while we are sleeping waiting
    790 	 * for old data structures to be purged or for the contents of the
    791 	 * disk portion of this inode to be read.
    792 	 */
    793 	ufs_ihashins(ip);
    794 
    795 	/* Read in the disk contents for the inode, copy into the inode. */
    796 	if (error = bread(ump->um_devvp, fsbtodb(fs, ino_to_fsba(fs, ino)),
    797 	    (int)fs->fs_bsize, NOCRED, &bp)) {
    798 		/*
    799 		 * The inode does not contain anything useful, so it would
    800 		 * be misleading to leave it on its hash chain. With mode
    801 		 * still zero, it will be unlinked and returned to the free
    802 		 * list by vput().
    803 		 */
    804 		vput(vp);
    805 		brelse(bp);
    806 		*vpp = NULL;
    807 		return (error);
    808 	}
    809 	ip->i_din = *((struct dinode *)bp->b_data + ino_to_fsbo(fs, ino));
    810 	brelse(bp);
    811 
    812 	/*
    813 	 * Initialize the vnode from the inode, check for aliases.
    814 	 * Note that the underlying vnode may have changed.
    815 	 */
    816 	if (error = ufs_vinit(mp, ffs_specop_p, FFS_FIFOOPS, &vp)) {
    817 		vput(vp);
    818 		*vpp = NULL;
    819 		return (error);
    820 	}
    821 	/*
    822 	 * Finish inode initialization now that aliasing has been resolved.
    823 	 */
    824 	ip->i_devvp = ump->um_devvp;
    825 	VREF(ip->i_devvp);
    826 	/*
    827 	 * Set up a generation number for this inode if it does not
    828 	 * already have one. This should only happen on old filesystems.
    829 	 */
    830 	if (ip->i_gen == 0) {
    831 		if (++nextgennumber < (u_long)time.tv_sec)
    832 			nextgennumber = time.tv_sec;
    833 		ip->i_gen = nextgennumber;
    834 		if ((vp->v_mount->mnt_flag & MNT_RDONLY) == 0)
    835 			ip->i_flag |= IN_MODIFIED;
    836 	}
    837 	/*
    838 	 * Ensure that uid and gid are correct. This is a temporary
    839 	 * fix until fsck has been changed to do the update.
    840 	 */
    841 	if (fs->fs_inodefmt < FS_44INODEFMT) {		/* XXX */
    842 		ip->i_uid = ip->i_din.di_ouid;		/* XXX */
    843 		ip->i_gid = ip->i_din.di_ogid;		/* XXX */
    844 	}						/* XXX */
    845 
    846 	*vpp = vp;
    847 	return (0);
    848 }
    849 
    850 /*
    851  * File handle to vnode
    852  *
    853  * Have to be really careful about stale file handles:
    854  * - check that the inode number is valid
    855  * - call ffs_vget() to get the locked inode
    856  * - check for an unallocated inode (i_mode == 0)
    857  * - check that the given client host has export rights and return
    858  *   those rights via. exflagsp and credanonp
    859  */
    860 int
    861 ffs_fhtovp(mp, fhp, nam, vpp, exflagsp, credanonp)
    862 	register struct mount *mp;
    863 	struct fid *fhp;
    864 	struct mbuf *nam;
    865 	struct vnode **vpp;
    866 	int *exflagsp;
    867 	struct ucred **credanonp;
    868 {
    869 	register struct ufid *ufhp;
    870 	struct fs *fs;
    871 
    872 	ufhp = (struct ufid *)fhp;
    873 	fs = VFSTOUFS(mp)->um_fs;
    874 	if (ufhp->ufid_ino < ROOTINO ||
    875 	    ufhp->ufid_ino >= fs->fs_ncg * fs->fs_ipg)
    876 		return (ESTALE);
    877 	return (ufs_check_export(mp, ufhp, nam, vpp, exflagsp, credanonp));
    878 }
    879 
    880 /*
    881  * Vnode pointer to File handle
    882  */
    883 /* ARGSUSED */
    884 ffs_vptofh(vp, fhp)
    885 	struct vnode *vp;
    886 	struct fid *fhp;
    887 {
    888 	register struct inode *ip;
    889 	register struct ufid *ufhp;
    890 
    891 	ip = VTOI(vp);
    892 	ufhp = (struct ufid *)fhp;
    893 	ufhp->ufid_len = sizeof(struct ufid);
    894 	ufhp->ufid_ino = ip->i_number;
    895 	ufhp->ufid_gen = ip->i_gen;
    896 	return (0);
    897 }
    898 
    899 /*
    900  * Write a superblock and associated information back to disk.
    901  */
    902 int
    903 ffs_sbupdate(mp, waitfor)
    904 	struct ufsmount *mp;
    905 	int waitfor;
    906 {
    907 	register struct fs *dfs, *fs = mp->um_fs;
    908 	register struct buf *bp;
    909 	int i, error = 0;
    910 
    911 	bp = getblk(mp->um_devvp, SBOFF >> (fs->fs_fshift - fs->fs_fsbtodb),
    912 	    (int)fs->fs_sbsize, 0, 0);
    913 	bcopy((caddr_t)fs, bp->b_data, (u_int)fs->fs_sbsize);
    914 	/* Restore compatibility to old file systems.		   XXX */
    915 	dfs = (struct fs *)bp->b_data;				/* XXX */
    916 	if (fs->fs_postblformat == FS_42POSTBLFMT)		/* XXX */
    917 		dfs->fs_nrpos = -1;				/* XXX */
    918 	if (fs->fs_inodefmt < FS_44INODEFMT) {			/* XXX */
    919 		int32_t *lp, tmp;				/* XXX */
    920 								/* XXX */
    921 		lp = (int32_t *)&dfs->fs_qbmask;		/* XXX */
    922 		tmp = lp[4];					/* XXX */
    923 		for (i = 4; i > 0; i--)				/* XXX */
    924 			lp[i] = lp[i-1];			/* XXX */
    925 		lp[0] = tmp;					/* XXX */
    926 	}							/* XXX */
    927 	dfs->fs_maxfilesize = mp->um_savedmaxfilesize;		/* XXX */
    928 	if (waitfor == MNT_WAIT)
    929 		error = bwrite(bp);
    930 	else
    931 		bawrite(bp);
    932 	return (error);
    933 }
    934 
    935 int
    936 ffs_cgupdate(mp, waitfor)
    937 	struct ufsmount *mp;
    938 	int waitfor;
    939 {
    940 	register struct fs *fs = mp->um_fs;
    941 	register struct buf *bp;
    942 	int blks;
    943 	caddr_t space;
    944 	int i, size, error = 0, allerror = 0;
    945 
    946 	allerror = ffs_sbupdate(mp, waitfor);
    947 	blks = howmany(fs->fs_cssize, fs->fs_fsize);
    948 	space = (caddr_t)fs->fs_csp[0];
    949 	for (i = 0; i < blks; i += fs->fs_frag) {
    950 		size = fs->fs_bsize;
    951 		if (i + fs->fs_frag > blks)
    952 			size = (blks - i) * fs->fs_fsize;
    953 		bp = getblk(mp->um_devvp, fsbtodb(fs, fs->fs_csaddr + i),
    954 		    size, 0, 0);
    955 		bcopy(space, bp->b_data, (u_int)size);
    956 		space += size;
    957 		if (waitfor == MNT_WAIT)
    958 			error = bwrite(bp);
    959 		else
    960 			bawrite(bp);
    961 	}
    962 	if (!allerror && error)
    963 		allerror = error;
    964 	return (allerror);
    965 }
    966