Home | History | Annotate | Line # | Download | only in ffs
ffs_vfsops.c revision 1.1.1.2
      1 /*
      2  * Copyright (c) 1989, 1991, 1993, 1994
      3  *	The Regents of the University of California.  All rights reserved.
      4  *
      5  * Redistribution and use in source and binary forms, with or without
      6  * modification, are permitted provided that the following conditions
      7  * are met:
      8  * 1. Redistributions of source code must retain the above copyright
      9  *    notice, this list of conditions and the following disclaimer.
     10  * 2. Redistributions in binary form must reproduce the above copyright
     11  *    notice, this list of conditions and the following disclaimer in the
     12  *    documentation and/or other materials provided with the distribution.
     13  * 3. All advertising materials mentioning features or use of this software
     14  *    must display the following acknowledgement:
     15  *	This product includes software developed by the University of
     16  *	California, Berkeley and its contributors.
     17  * 4. Neither the name of the University nor the names of its contributors
     18  *    may be used to endorse or promote products derived from this software
     19  *    without specific prior written permission.
     20  *
     21  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
     22  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
     23  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
     24  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
     25  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
     26  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
     27  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
     28  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
     29  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
     30  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
     31  * SUCH DAMAGE.
     32  *
     33  *	@(#)ffs_vfsops.c	8.31 (Berkeley) 5/20/95
     34  */
     35 
     36 #include <sys/param.h>
     37 #include <sys/systm.h>
     38 #include <sys/namei.h>
     39 #include <sys/proc.h>
     40 #include <sys/kernel.h>
     41 #include <sys/vnode.h>
     42 #include <sys/socket.h>
     43 #include <sys/mount.h>
     44 #include <sys/buf.h>
     45 #include <sys/mbuf.h>
     46 #include <sys/file.h>
     47 #include <sys/disklabel.h>
     48 #include <sys/ioctl.h>
     49 #include <sys/errno.h>
     50 #include <sys/malloc.h>
     51 
     52 #include <miscfs/specfs/specdev.h>
     53 
     54 #include <ufs/ufs/quota.h>
     55 #include <ufs/ufs/ufsmount.h>
     56 #include <ufs/ufs/inode.h>
     57 #include <ufs/ufs/ufs_extern.h>
     58 
     59 #include <ufs/ffs/fs.h>
     60 #include <ufs/ffs/ffs_extern.h>
     61 
     62 int ffs_sbupdate __P((struct ufsmount *, int));
     63 
     64 struct vfsops ufs_vfsops = {
     65 	ffs_mount,
     66 	ufs_start,
     67 	ffs_unmount,
     68 	ufs_root,
     69 	ufs_quotactl,
     70 	ffs_statfs,
     71 	ffs_sync,
     72 	ffs_vget,
     73 	ffs_fhtovp,
     74 	ffs_vptofh,
     75 	ffs_init,
     76 	ffs_sysctl,
     77 };
     78 
     79 extern u_long nextgennumber;
     80 
     81 /*
     82  * Called by main() when ufs is going to be mounted as root.
     83  */
     84 ffs_mountroot()
     85 {
     86 	extern struct vnode *rootvp;
     87 	struct fs *fs;
     88 	struct mount *mp;
     89 	struct proc *p = curproc;	/* XXX */
     90 	struct ufsmount *ump;
     91 	u_int size;
     92 	int error;
     93 
     94 	/*
     95 	 * Get vnodes for swapdev and rootdev.
     96 	 */
     97 	if ((error = bdevvp(swapdev, &swapdev_vp)) ||
     98 	    (error = bdevvp(rootdev, &rootvp))) {
     99 		printf("ffs_mountroot: can't setup bdevvp's");
    100 		return (error);
    101 	}
    102 	if (error = vfs_rootmountalloc("ufs", "root_device", &mp))
    103 		return (error);
    104 	if (error = ffs_mountfs(rootvp, mp, p)) {
    105 		mp->mnt_vfc->vfc_refcount--;
    106 		vfs_unbusy(mp, p);
    107 		free(mp, M_MOUNT);
    108 		return (error);
    109 	}
    110 	simple_lock(&mountlist_slock);
    111 	CIRCLEQ_INSERT_TAIL(&mountlist, mp, mnt_list);
    112 	simple_unlock(&mountlist_slock);
    113 	ump = VFSTOUFS(mp);
    114 	fs = ump->um_fs;
    115 	(void) copystr(mp->mnt_stat.f_mntonname, fs->fs_fsmnt, MNAMELEN - 1, 0);
    116 	(void)ffs_statfs(mp, &mp->mnt_stat, p);
    117 	vfs_unbusy(mp, p);
    118 	inittodr(fs->fs_time);
    119 	return (0);
    120 }
    121 
    122 /*
    123  * VFS Operations.
    124  *
    125  * mount system call
    126  */
    127 int
    128 ffs_mount(mp, path, data, ndp, p)
    129 	register struct mount *mp;
    130 	char *path;
    131 	caddr_t data;
    132 	struct nameidata *ndp;
    133 	struct proc *p;
    134 {
    135 	struct vnode *devvp;
    136 	struct ufs_args args;
    137 	struct ufsmount *ump;
    138 	register struct fs *fs;
    139 	u_int size;
    140 	int error, flags;
    141 	mode_t accessmode;
    142 
    143 	if (error = copyin(data, (caddr_t)&args, sizeof (struct ufs_args)))
    144 		return (error);
    145 	/*
    146 	 * If updating, check whether changing from read-only to
    147 	 * read/write; if there is no device name, that's all we do.
    148 	 */
    149 	if (mp->mnt_flag & MNT_UPDATE) {
    150 		ump = VFSTOUFS(mp);
    151 		fs = ump->um_fs;
    152 		if (fs->fs_ronly == 0 && (mp->mnt_flag & MNT_RDONLY)) {
    153 			flags = WRITECLOSE;
    154 			if (mp->mnt_flag & MNT_FORCE)
    155 				flags |= FORCECLOSE;
    156 			if (error = ffs_flushfiles(mp, flags, p))
    157 				return (error);
    158 			fs->fs_clean = 1;
    159 			fs->fs_ronly = 1;
    160 			if (error = ffs_sbupdate(ump, MNT_WAIT)) {
    161 				fs->fs_clean = 0;
    162 				fs->fs_ronly = 0;
    163 				return (error);
    164 			}
    165 		}
    166 		if ((mp->mnt_flag & MNT_RELOAD) &&
    167 		    (error = ffs_reload(mp, ndp->ni_cnd.cn_cred, p)))
    168 			return (error);
    169 		if (fs->fs_ronly && (mp->mnt_flag & MNT_WANTRDWR)) {
    170 			/*
    171 			 * If upgrade to read-write by non-root, then verify
    172 			 * that user has necessary permissions on the device.
    173 			 */
    174 			if (p->p_ucred->cr_uid != 0) {
    175 				devvp = ump->um_devvp;
    176 				vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY, p);
    177 				if (error = VOP_ACCESS(devvp, VREAD | VWRITE,
    178 				    p->p_ucred, p)) {
    179 					VOP_UNLOCK(devvp, 0, p);
    180 					return (error);
    181 				}
    182 				VOP_UNLOCK(devvp, 0, p);
    183 			}
    184 			fs->fs_ronly = 0;
    185 			fs->fs_clean = 0;
    186 			(void) ffs_sbupdate(ump, MNT_WAIT);
    187 		}
    188 		if (args.fspec == 0) {
    189 			/*
    190 			 * Process export requests.
    191 			 */
    192 			return (vfs_export(mp, &ump->um_export, &args.export));
    193 		}
    194 	}
    195 	/*
    196 	 * Not an update, or updating the name: look up the name
    197 	 * and verify that it refers to a sensible block device.
    198 	 */
    199 	NDINIT(ndp, LOOKUP, FOLLOW, UIO_USERSPACE, args.fspec, p);
    200 	if (error = namei(ndp))
    201 		return (error);
    202 	devvp = ndp->ni_vp;
    203 
    204 	if (devvp->v_type != VBLK) {
    205 		vrele(devvp);
    206 		return (ENOTBLK);
    207 	}
    208 	if (major(devvp->v_rdev) >= nblkdev) {
    209 		vrele(devvp);
    210 		return (ENXIO);
    211 	}
    212 	/*
    213 	 * If mount by non-root, then verify that user has necessary
    214 	 * permissions on the device.
    215 	 */
    216 	if (p->p_ucred->cr_uid != 0) {
    217 		accessmode = VREAD;
    218 		if ((mp->mnt_flag & MNT_RDONLY) == 0)
    219 			accessmode |= VWRITE;
    220 		vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY, p);
    221 		if (error = VOP_ACCESS(devvp, accessmode, p->p_ucred, p)) {
    222 			vput(devvp);
    223 			return (error);
    224 		}
    225 		VOP_UNLOCK(devvp, 0, p);
    226 	}
    227 	if ((mp->mnt_flag & MNT_UPDATE) == 0)
    228 		error = ffs_mountfs(devvp, mp, p);
    229 	else {
    230 		if (devvp != ump->um_devvp)
    231 			error = EINVAL;	/* needs translation */
    232 		else
    233 			vrele(devvp);
    234 	}
    235 	if (error) {
    236 		vrele(devvp);
    237 		return (error);
    238 	}
    239 	ump = VFSTOUFS(mp);
    240 	fs = ump->um_fs;
    241 	(void) copyinstr(path, fs->fs_fsmnt, sizeof(fs->fs_fsmnt) - 1, &size);
    242 	bzero(fs->fs_fsmnt + size, sizeof(fs->fs_fsmnt) - size);
    243 	bcopy((caddr_t)fs->fs_fsmnt, (caddr_t)mp->mnt_stat.f_mntonname,
    244 	    MNAMELEN);
    245 	(void) copyinstr(args.fspec, mp->mnt_stat.f_mntfromname, MNAMELEN - 1,
    246 	    &size);
    247 	bzero(mp->mnt_stat.f_mntfromname + size, MNAMELEN - size);
    248 	(void)ffs_statfs(mp, &mp->mnt_stat, p);
    249 	return (0);
    250 }
    251 
    252 /*
    253  * Reload all incore data for a filesystem (used after running fsck on
    254  * the root filesystem and finding things to fix). The filesystem must
    255  * be mounted read-only.
    256  *
    257  * Things to do to update the mount:
    258  *	1) invalidate all cached meta-data.
    259  *	2) re-read superblock from disk.
    260  *	3) re-read summary information from disk.
    261  *	4) invalidate all inactive vnodes.
    262  *	5) invalidate all cached file data.
    263  *	6) re-read inode data for all active vnodes.
    264  */
    265 ffs_reload(mountp, cred, p)
    266 	register struct mount *mountp;
    267 	struct ucred *cred;
    268 	struct proc *p;
    269 {
    270 	register struct vnode *vp, *nvp, *devvp;
    271 	struct inode *ip;
    272 	struct csum *space;
    273 	struct buf *bp;
    274 	struct fs *fs, *newfs;
    275 	struct partinfo dpart;
    276 	int i, blks, size, error;
    277 	int32_t *lp;
    278 
    279 	if ((mountp->mnt_flag & MNT_RDONLY) == 0)
    280 		return (EINVAL);
    281 	/*
    282 	 * Step 1: invalidate all cached meta-data.
    283 	 */
    284 	devvp = VFSTOUFS(mountp)->um_devvp;
    285 	if (vinvalbuf(devvp, 0, cred, p, 0, 0))
    286 		panic("ffs_reload: dirty1");
    287 	/*
    288 	 * Step 2: re-read superblock from disk.
    289 	 */
    290 	if (VOP_IOCTL(devvp, DIOCGPART, (caddr_t)&dpart, FREAD, NOCRED, p) != 0)
    291 		size = DEV_BSIZE;
    292 	else
    293 		size = dpart.disklab->d_secsize;
    294 	if (error = bread(devvp, (ufs_daddr_t)(SBOFF/size), SBSIZE, NOCRED,&bp))
    295 		return (error);
    296 	newfs = (struct fs *)bp->b_data;
    297 	if (newfs->fs_magic != FS_MAGIC || newfs->fs_bsize > MAXBSIZE ||
    298 	    newfs->fs_bsize < sizeof(struct fs)) {
    299 		brelse(bp);
    300 		return (EIO);		/* XXX needs translation */
    301 	}
    302 	fs = VFSTOUFS(mountp)->um_fs;
    303 	/*
    304 	 * Copy pointer fields back into superblock before copying in	XXX
    305 	 * new superblock. These should really be in the ufsmount.	XXX
    306 	 * Note that important parameters (eg fs_ncg) are unchanged.
    307 	 */
    308 	bcopy(&fs->fs_csp[0], &newfs->fs_csp[0], sizeof(fs->fs_csp));
    309 	newfs->fs_maxcluster = fs->fs_maxcluster;
    310 	bcopy(newfs, fs, (u_int)fs->fs_sbsize);
    311 	if (fs->fs_sbsize < SBSIZE)
    312 		bp->b_flags |= B_INVAL;
    313 	brelse(bp);
    314 	mountp->mnt_maxsymlinklen = fs->fs_maxsymlinklen;
    315 	ffs_oldfscompat(fs);
    316 	/*
    317 	 * Step 3: re-read summary information from disk.
    318 	 */
    319 	blks = howmany(fs->fs_cssize, fs->fs_fsize);
    320 	space = fs->fs_csp[0];
    321 	for (i = 0; i < blks; i += fs->fs_frag) {
    322 		size = fs->fs_bsize;
    323 		if (i + fs->fs_frag > blks)
    324 			size = (blks - i) * fs->fs_fsize;
    325 		if (error = bread(devvp, fsbtodb(fs, fs->fs_csaddr + i), size,
    326 		    NOCRED, &bp))
    327 			return (error);
    328 		bcopy(bp->b_data, fs->fs_csp[fragstoblks(fs, i)], (u_int)size);
    329 		brelse(bp);
    330 	}
    331 	/*
    332 	 * We no longer know anything about clusters per cylinder group.
    333 	 */
    334 	if (fs->fs_contigsumsize > 0) {
    335 		lp = fs->fs_maxcluster;
    336 		for (i = 0; i < fs->fs_ncg; i++)
    337 			*lp++ = fs->fs_contigsumsize;
    338 	}
    339 
    340 loop:
    341 	simple_lock(&mntvnode_slock);
    342 	for (vp = mountp->mnt_vnodelist.lh_first; vp != NULL; vp = nvp) {
    343 		if (vp->v_mount != mountp) {
    344 			simple_unlock(&mntvnode_slock);
    345 			goto loop;
    346 		}
    347 		nvp = vp->v_mntvnodes.le_next;
    348 		/*
    349 		 * Step 4: invalidate all inactive vnodes.
    350 		 */
    351 		if (vrecycle(vp, &mntvnode_slock, p))
    352 			goto loop;
    353 		/*
    354 		 * Step 5: invalidate all cached file data.
    355 		 */
    356 		simple_lock(&vp->v_interlock);
    357 		simple_unlock(&mntvnode_slock);
    358 		if (vget(vp, LK_EXCLUSIVE | LK_INTERLOCK, p)) {
    359 			goto loop;
    360 		}
    361 		if (vinvalbuf(vp, 0, cred, p, 0, 0))
    362 			panic("ffs_reload: dirty2");
    363 		/*
    364 		 * Step 6: re-read inode data for all active vnodes.
    365 		 */
    366 		ip = VTOI(vp);
    367 		if (error =
    368 		    bread(devvp, fsbtodb(fs, ino_to_fsba(fs, ip->i_number)),
    369 		    (int)fs->fs_bsize, NOCRED, &bp)) {
    370 			vput(vp);
    371 			return (error);
    372 		}
    373 		ip->i_din = *((struct dinode *)bp->b_data +
    374 		    ino_to_fsbo(fs, ip->i_number));
    375 		brelse(bp);
    376 		vput(vp);
    377 		simple_lock(&mntvnode_slock);
    378 	}
    379 	simple_unlock(&mntvnode_slock);
    380 	return (0);
    381 }
    382 
    383 /*
    384  * Common code for mount and mountroot
    385  */
    386 int
    387 ffs_mountfs(devvp, mp, p)
    388 	register struct vnode *devvp;
    389 	struct mount *mp;
    390 	struct proc *p;
    391 {
    392 	register struct ufsmount *ump;
    393 	struct buf *bp;
    394 	register struct fs *fs;
    395 	dev_t dev;
    396 	struct partinfo dpart;
    397 	caddr_t base, space;
    398 	int error, i, blks, size, ronly;
    399 	int32_t *lp;
    400 	struct ucred *cred;
    401 	extern struct vnode *rootvp;
    402 	u_int64_t maxfilesize;					/* XXX */
    403 
    404 	dev = devvp->v_rdev;
    405 	cred = p ? p->p_ucred : NOCRED;
    406 	/*
    407 	 * Disallow multiple mounts of the same device.
    408 	 * Disallow mounting of a device that is currently in use
    409 	 * (except for root, which might share swap device for miniroot).
    410 	 * Flush out any old buffers remaining from a previous use.
    411 	 */
    412 	if (error = vfs_mountedon(devvp))
    413 		return (error);
    414 	if (vcount(devvp) > 1 && devvp != rootvp)
    415 		return (EBUSY);
    416 	if (error = vinvalbuf(devvp, V_SAVE, cred, p, 0, 0))
    417 		return (error);
    418 
    419 	ronly = (mp->mnt_flag & MNT_RDONLY) != 0;
    420 	if (error = VOP_OPEN(devvp, ronly ? FREAD : FREAD|FWRITE, FSCRED, p))
    421 		return (error);
    422 	if (VOP_IOCTL(devvp, DIOCGPART, (caddr_t)&dpart, FREAD, cred, p) != 0)
    423 		size = DEV_BSIZE;
    424 	else
    425 		size = dpart.disklab->d_secsize;
    426 
    427 	bp = NULL;
    428 	ump = NULL;
    429 	if (error = bread(devvp, (ufs_daddr_t)(SBOFF/size), SBSIZE, cred, &bp))
    430 		goto out;
    431 	fs = (struct fs *)bp->b_data;
    432 	if (fs->fs_magic != FS_MAGIC || fs->fs_bsize > MAXBSIZE ||
    433 	    fs->fs_bsize < sizeof(struct fs)) {
    434 		error = EINVAL;		/* XXX needs translation */
    435 		goto out;
    436 	}
    437 	/* XXX updating 4.2 FFS superblocks trashes rotational layout tables */
    438 	if (fs->fs_postblformat == FS_42POSTBLFMT && !ronly) {
    439 		error = EROFS;          /* needs translation */
    440 		goto out;
    441 	}
    442 	ump = malloc(sizeof *ump, M_UFSMNT, M_WAITOK);
    443 	bzero((caddr_t)ump, sizeof *ump);
    444 	ump->um_fs = malloc((u_long)fs->fs_sbsize, M_UFSMNT,
    445 	    M_WAITOK);
    446 	bcopy(bp->b_data, ump->um_fs, (u_int)fs->fs_sbsize);
    447 	if (fs->fs_sbsize < SBSIZE)
    448 		bp->b_flags |= B_INVAL;
    449 	brelse(bp);
    450 	bp = NULL;
    451 	fs = ump->um_fs;
    452 	fs->fs_ronly = ronly;
    453 	size = fs->fs_cssize;
    454 	blks = howmany(size, fs->fs_fsize);
    455 	if (fs->fs_contigsumsize > 0)
    456 		size += fs->fs_ncg * sizeof(int32_t);
    457 	base = space = malloc((u_long)size, M_UFSMNT, M_WAITOK);
    458 	for (i = 0; i < blks; i += fs->fs_frag) {
    459 		size = fs->fs_bsize;
    460 		if (i + fs->fs_frag > blks)
    461 			size = (blks - i) * fs->fs_fsize;
    462 		if (error = bread(devvp, fsbtodb(fs, fs->fs_csaddr + i), size,
    463 		    cred, &bp)) {
    464 			free(base, M_UFSMNT);
    465 			goto out;
    466 		}
    467 		bcopy(bp->b_data, space, (u_int)size);
    468 		fs->fs_csp[fragstoblks(fs, i)] = (struct csum *)space;
    469 		space += size;
    470 		brelse(bp);
    471 		bp = NULL;
    472 	}
    473 	if (fs->fs_contigsumsize > 0) {
    474 		fs->fs_maxcluster = lp = (int32_t *)space;
    475 		for (i = 0; i < fs->fs_ncg; i++)
    476 			*lp++ = fs->fs_contigsumsize;
    477 	}
    478 	mp->mnt_data = (qaddr_t)ump;
    479 	mp->mnt_stat.f_fsid.val[0] = (long)dev;
    480 	mp->mnt_stat.f_fsid.val[1] = mp->mnt_vfc->vfc_typenum;
    481 	mp->mnt_maxsymlinklen = fs->fs_maxsymlinklen;
    482 	ump->um_mountp = mp;
    483 	ump->um_dev = dev;
    484 	ump->um_devvp = devvp;
    485 	ump->um_nindir = fs->fs_nindir;
    486 	ump->um_bptrtodb = fs->fs_fsbtodb;
    487 	ump->um_seqinc = fs->fs_frag;
    488 	for (i = 0; i < MAXQUOTAS; i++)
    489 		ump->um_quotas[i] = NULLVP;
    490 	devvp->v_specflags |= SI_MOUNTEDON;
    491 	ffs_oldfscompat(fs);
    492 	ump->um_savedmaxfilesize = fs->fs_maxfilesize;		/* XXX */
    493 	maxfilesize = (u_int64_t)0x40000000 * fs->fs_bsize - 1;	/* XXX */
    494 	if (fs->fs_maxfilesize > maxfilesize)			/* XXX */
    495 		fs->fs_maxfilesize = maxfilesize;		/* XXX */
    496 	if (ronly == 0) {
    497 		fs->fs_clean = 0;
    498 		(void) ffs_sbupdate(ump, MNT_WAIT);
    499 	}
    500 	return (0);
    501 out:
    502 	if (bp)
    503 		brelse(bp);
    504 	(void)VOP_CLOSE(devvp, ronly ? FREAD : FREAD|FWRITE, cred, p);
    505 	if (ump) {
    506 		free(ump->um_fs, M_UFSMNT);
    507 		free(ump, M_UFSMNT);
    508 		mp->mnt_data = (qaddr_t)0;
    509 	}
    510 	return (error);
    511 }
    512 
    513 /*
    514  * Sanity checks for old file systems.
    515  *
    516  * XXX - goes away some day.
    517  */
    518 ffs_oldfscompat(fs)
    519 	struct fs *fs;
    520 {
    521 	int i;
    522 
    523 	fs->fs_npsect = max(fs->fs_npsect, fs->fs_nsect);	/* XXX */
    524 	fs->fs_interleave = max(fs->fs_interleave, 1);		/* XXX */
    525 	if (fs->fs_postblformat == FS_42POSTBLFMT)		/* XXX */
    526 		fs->fs_nrpos = 8;				/* XXX */
    527 	if (fs->fs_inodefmt < FS_44INODEFMT) {			/* XXX */
    528 		u_int64_t sizepb = fs->fs_bsize;		/* XXX */
    529 								/* XXX */
    530 		fs->fs_maxfilesize = fs->fs_bsize * NDADDR - 1;	/* XXX */
    531 		for (i = 0; i < NIADDR; i++) {			/* XXX */
    532 			sizepb *= NINDIR(fs);			/* XXX */
    533 			fs->fs_maxfilesize += sizepb;		/* XXX */
    534 		}						/* XXX */
    535 		fs->fs_qbmask = ~fs->fs_bmask;			/* XXX */
    536 		fs->fs_qfmask = ~fs->fs_fmask;			/* XXX */
    537 	}							/* XXX */
    538 	return (0);
    539 }
    540 
    541 /*
    542  * unmount system call
    543  */
    544 int
    545 ffs_unmount(mp, mntflags, p)
    546 	struct mount *mp;
    547 	int mntflags;
    548 	struct proc *p;
    549 {
    550 	register struct ufsmount *ump;
    551 	register struct fs *fs;
    552 	int error, flags;
    553 
    554 	flags = 0;
    555 	if (mntflags & MNT_FORCE)
    556 		flags |= FORCECLOSE;
    557 	if (error = ffs_flushfiles(mp, flags, p))
    558 		return (error);
    559 	ump = VFSTOUFS(mp);
    560 	fs = ump->um_fs;
    561 	if (fs->fs_ronly == 0) {
    562 		fs->fs_clean = 1;
    563 		if (error = ffs_sbupdate(ump, MNT_WAIT)) {
    564 			fs->fs_clean = 0;
    565 			return (error);
    566 		}
    567 	}
    568 	ump->um_devvp->v_specflags &= ~SI_MOUNTEDON;
    569 	error = VOP_CLOSE(ump->um_devvp, fs->fs_ronly ? FREAD : FREAD|FWRITE,
    570 		NOCRED, p);
    571 	vrele(ump->um_devvp);
    572 	free(fs->fs_csp[0], M_UFSMNT);
    573 	free(fs, M_UFSMNT);
    574 	free(ump, M_UFSMNT);
    575 	mp->mnt_data = (qaddr_t)0;
    576 	return (error);
    577 }
    578 
    579 /*
    580  * Flush out all the files in a filesystem.
    581  */
    582 ffs_flushfiles(mp, flags, p)
    583 	register struct mount *mp;
    584 	int flags;
    585 	struct proc *p;
    586 {
    587 	register struct ufsmount *ump;
    588 	int i, error;
    589 
    590 	ump = VFSTOUFS(mp);
    591 #ifdef QUOTA
    592 	if (mp->mnt_flag & MNT_QUOTA) {
    593 		if (error = vflush(mp, NULLVP, SKIPSYSTEM|flags))
    594 			return (error);
    595 		for (i = 0; i < MAXQUOTAS; i++) {
    596 			if (ump->um_quotas[i] == NULLVP)
    597 				continue;
    598 			quotaoff(p, mp, i);
    599 		}
    600 		/*
    601 		 * Here we fall through to vflush again to ensure
    602 		 * that we have gotten rid of all the system vnodes.
    603 		 */
    604 	}
    605 #endif
    606 	error = vflush(mp, NULLVP, flags);
    607 	return (error);
    608 }
    609 
    610 /*
    611  * Get file system statistics.
    612  */
    613 int
    614 ffs_statfs(mp, sbp, p)
    615 	struct mount *mp;
    616 	register struct statfs *sbp;
    617 	struct proc *p;
    618 {
    619 	register struct ufsmount *ump;
    620 	register struct fs *fs;
    621 
    622 	ump = VFSTOUFS(mp);
    623 	fs = ump->um_fs;
    624 	if (fs->fs_magic != FS_MAGIC)
    625 		panic("ffs_statfs");
    626 	sbp->f_bsize = fs->fs_fsize;
    627 	sbp->f_iosize = fs->fs_bsize;
    628 	sbp->f_blocks = fs->fs_dsize;
    629 	sbp->f_bfree = fs->fs_cstotal.cs_nbfree * fs->fs_frag +
    630 		fs->fs_cstotal.cs_nffree;
    631 	sbp->f_bavail = (fs->fs_dsize * (100 - fs->fs_minfree) / 100) -
    632 		(fs->fs_dsize - sbp->f_bfree);
    633 	sbp->f_files =  fs->fs_ncg * fs->fs_ipg - ROOTINO;
    634 	sbp->f_ffree = fs->fs_cstotal.cs_nifree;
    635 	if (sbp != &mp->mnt_stat) {
    636 		sbp->f_type = mp->mnt_vfc->vfc_typenum;
    637 		bcopy((caddr_t)mp->mnt_stat.f_mntonname,
    638 			(caddr_t)&sbp->f_mntonname[0], MNAMELEN);
    639 		bcopy((caddr_t)mp->mnt_stat.f_mntfromname,
    640 			(caddr_t)&sbp->f_mntfromname[0], MNAMELEN);
    641 	}
    642 	return (0);
    643 }
    644 
    645 /*
    646  * Go through the disk queues to initiate sandbagged IO;
    647  * go through the inodes to write those that have been modified;
    648  * initiate the writing of the super block if it has been modified.
    649  *
    650  * Note: we are always called with the filesystem marked `MPBUSY'.
    651  */
    652 int
    653 ffs_sync(mp, waitfor, cred, p)
    654 	struct mount *mp;
    655 	int waitfor;
    656 	struct ucred *cred;
    657 	struct proc *p;
    658 {
    659 	struct vnode *nvp, *vp;
    660 	struct inode *ip;
    661 	struct ufsmount *ump = VFSTOUFS(mp);
    662 	struct fs *fs;
    663 	int error, allerror = 0;
    664 
    665 	fs = ump->um_fs;
    666 	if (fs->fs_fmod != 0 && fs->fs_ronly != 0) {		/* XXX */
    667 		printf("fs = %s\n", fs->fs_fsmnt);
    668 		panic("update: rofs mod");
    669 	}
    670 	/*
    671 	 * Write back each (modified) inode.
    672 	 */
    673 	simple_lock(&mntvnode_slock);
    674 loop:
    675 	for (vp = mp->mnt_vnodelist.lh_first;
    676 	     vp != NULL;
    677 	     vp = nvp) {
    678 		/*
    679 		 * If the vnode that we are about to sync is no longer
    680 		 * associated with this mount point, start over.
    681 		 */
    682 		if (vp->v_mount != mp)
    683 			goto loop;
    684 		simple_lock(&vp->v_interlock);
    685 		nvp = vp->v_mntvnodes.le_next;
    686 		ip = VTOI(vp);
    687 		if ((ip->i_flag &
    688 		    (IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE)) == 0 &&
    689 		    vp->v_dirtyblkhd.lh_first == NULL) {
    690 			simple_unlock(&vp->v_interlock);
    691 			continue;
    692 		}
    693 		simple_unlock(&mntvnode_slock);
    694 		error = vget(vp, LK_EXCLUSIVE | LK_NOWAIT | LK_INTERLOCK, p);
    695 		if (error) {
    696 			simple_lock(&mntvnode_slock);
    697 			if (error == ENOENT)
    698 				goto loop;
    699 			continue;
    700 		}
    701 		if (error = VOP_FSYNC(vp, cred, waitfor, p))
    702 			allerror = error;
    703 		VOP_UNLOCK(vp, 0, p);
    704 		vrele(vp);
    705 		simple_lock(&mntvnode_slock);
    706 	}
    707 	simple_unlock(&mntvnode_slock);
    708 	/*
    709 	 * Force stale file system control information to be flushed.
    710 	 */
    711 	if (error = VOP_FSYNC(ump->um_devvp, cred, waitfor, p))
    712 		allerror = error;
    713 #ifdef QUOTA
    714 	qsync(mp);
    715 #endif
    716 	/*
    717 	 * Write back modified superblock.
    718 	 */
    719 	if (fs->fs_fmod != 0) {
    720 		fs->fs_fmod = 0;
    721 		fs->fs_time = time.tv_sec;
    722 		if (error = ffs_sbupdate(ump, waitfor))
    723 			allerror = error;
    724 	}
    725 	return (allerror);
    726 }
    727 
    728 /*
    729  * Look up a FFS dinode number to find its incore vnode, otherwise read it
    730  * in from disk.  If it is in core, wait for the lock bit to clear, then
    731  * return the inode locked.  Detection and handling of mount points must be
    732  * done by the calling routine.
    733  */
    734 int
    735 ffs_vget(mp, ino, vpp)
    736 	struct mount *mp;
    737 	ino_t ino;
    738 	struct vnode **vpp;
    739 {
    740 	struct proc *p = curproc;		/* XXX */
    741 	struct fs *fs;
    742 	struct inode *ip;
    743 	struct ufsmount *ump;
    744 	struct buf *bp;
    745 	struct vnode *vp;
    746 	dev_t dev;
    747 	int i, type, error;
    748 
    749 	ump = VFSTOUFS(mp);
    750 	dev = ump->um_dev;
    751 	if ((*vpp = ufs_ihashget(dev, ino)) != NULL)
    752 		return (0);
    753 
    754 	/* Allocate a new vnode/inode. */
    755 	if (error = getnewvnode(VT_UFS, mp, ffs_vnodeop_p, &vp)) {
    756 		*vpp = NULL;
    757 		return (error);
    758 	}
    759 	type = ump->um_devvp->v_tag == VT_MFS ? M_MFSNODE : M_FFSNODE; /* XXX */
    760 	MALLOC(ip, struct inode *, sizeof(struct inode), type, M_WAITOK);
    761 	bzero((caddr_t)ip, sizeof(struct inode));
    762 	lockinit(&ip->i_lock, PINOD, "inode", 0, 0);
    763 	vp->v_data = ip;
    764 	ip->i_vnode = vp;
    765 	ip->i_fs = fs = ump->um_fs;
    766 	ip->i_dev = dev;
    767 	ip->i_number = ino;
    768 #ifdef QUOTA
    769 	for (i = 0; i < MAXQUOTAS; i++)
    770 		ip->i_dquot[i] = NODQUOT;
    771 #endif
    772 	/*
    773 	 * Put it onto its hash chain and lock it so that other requests for
    774 	 * this inode will block if they arrive while we are sleeping waiting
    775 	 * for old data structures to be purged or for the contents of the
    776 	 * disk portion of this inode to be read.
    777 	 */
    778 	ufs_ihashins(ip);
    779 
    780 	/* Read in the disk contents for the inode, copy into the inode. */
    781 	if (error = bread(ump->um_devvp, fsbtodb(fs, ino_to_fsba(fs, ino)),
    782 	    (int)fs->fs_bsize, NOCRED, &bp)) {
    783 		/*
    784 		 * The inode does not contain anything useful, so it would
    785 		 * be misleading to leave it on its hash chain. With mode
    786 		 * still zero, it will be unlinked and returned to the free
    787 		 * list by vput().
    788 		 */
    789 		vput(vp);
    790 		brelse(bp);
    791 		*vpp = NULL;
    792 		return (error);
    793 	}
    794 	ip->i_din = *((struct dinode *)bp->b_data + ino_to_fsbo(fs, ino));
    795 	brelse(bp);
    796 
    797 	/*
    798 	 * Initialize the vnode from the inode, check for aliases.
    799 	 * Note that the underlying vnode may have changed.
    800 	 */
    801 	if (error = ufs_vinit(mp, ffs_specop_p, FFS_FIFOOPS, &vp)) {
    802 		vput(vp);
    803 		*vpp = NULL;
    804 		return (error);
    805 	}
    806 	/*
    807 	 * Finish inode initialization now that aliasing has been resolved.
    808 	 */
    809 	ip->i_devvp = ump->um_devvp;
    810 	VREF(ip->i_devvp);
    811 	/*
    812 	 * Set up a generation number for this inode if it does not
    813 	 * already have one. This should only happen on old filesystems.
    814 	 */
    815 	if (ip->i_gen == 0) {
    816 		if (++nextgennumber < (u_long)time.tv_sec)
    817 			nextgennumber = time.tv_sec;
    818 		ip->i_gen = nextgennumber;
    819 		if ((vp->v_mount->mnt_flag & MNT_RDONLY) == 0)
    820 			ip->i_flag |= IN_MODIFIED;
    821 	}
    822 	/*
    823 	 * Ensure that uid and gid are correct. This is a temporary
    824 	 * fix until fsck has been changed to do the update.
    825 	 */
    826 	if (fs->fs_inodefmt < FS_44INODEFMT) {		/* XXX */
    827 		ip->i_uid = ip->i_din.di_ouid;		/* XXX */
    828 		ip->i_gid = ip->i_din.di_ogid;		/* XXX */
    829 	}						/* XXX */
    830 
    831 	*vpp = vp;
    832 	return (0);
    833 }
    834 
    835 /*
    836  * File handle to vnode
    837  *
    838  * Have to be really careful about stale file handles:
    839  * - check that the inode number is valid
    840  * - call ffs_vget() to get the locked inode
    841  * - check for an unallocated inode (i_mode == 0)
    842  * - check that the given client host has export rights and return
    843  *   those rights via. exflagsp and credanonp
    844  */
    845 int
    846 ffs_fhtovp(mp, fhp, nam, vpp, exflagsp, credanonp)
    847 	register struct mount *mp;
    848 	struct fid *fhp;
    849 	struct mbuf *nam;
    850 	struct vnode **vpp;
    851 	int *exflagsp;
    852 	struct ucred **credanonp;
    853 {
    854 	register struct ufid *ufhp;
    855 	struct fs *fs;
    856 
    857 	ufhp = (struct ufid *)fhp;
    858 	fs = VFSTOUFS(mp)->um_fs;
    859 	if (ufhp->ufid_ino < ROOTINO ||
    860 	    ufhp->ufid_ino >= fs->fs_ncg * fs->fs_ipg)
    861 		return (ESTALE);
    862 	return (ufs_check_export(mp, ufhp, nam, vpp, exflagsp, credanonp));
    863 }
    864 
    865 /*
    866  * Vnode pointer to File handle
    867  */
    868 /* ARGSUSED */
    869 ffs_vptofh(vp, fhp)
    870 	struct vnode *vp;
    871 	struct fid *fhp;
    872 {
    873 	register struct inode *ip;
    874 	register struct ufid *ufhp;
    875 
    876 	ip = VTOI(vp);
    877 	ufhp = (struct ufid *)fhp;
    878 	ufhp->ufid_len = sizeof(struct ufid);
    879 	ufhp->ufid_ino = ip->i_number;
    880 	ufhp->ufid_gen = ip->i_gen;
    881 	return (0);
    882 }
    883 
    884 /*
    885  * Initialize the filesystem; just use ufs_init.
    886  */
    887 int
    888 ffs_init(vfsp)
    889 	struct vfsconf *vfsp;
    890 {
    891 
    892 	return (ufs_init(vfsp));
    893 }
    894 
    895 /*
    896  * fast filesystem related variables.
    897  */
    898 ffs_sysctl(name, namelen, oldp, oldlenp, newp, newlen, p)
    899 	int *name;
    900 	u_int namelen;
    901 	void *oldp;
    902 	size_t *oldlenp;
    903 	void *newp;
    904 	size_t newlen;
    905 	struct proc *p;
    906 {
    907 	extern int doclusterread, doclusterwrite, doreallocblks, doasyncfree;
    908 
    909 	/* all sysctl names at this level are terminal */
    910 	if (namelen != 1)
    911 		return (ENOTDIR);		/* overloaded */
    912 
    913 	switch (name[0]) {
    914 	case FFS_CLUSTERREAD:
    915 		return (sysctl_int(oldp, oldlenp, newp, newlen,
    916 		    &doclusterread));
    917 	case FFS_CLUSTERWRITE:
    918 		return (sysctl_int(oldp, oldlenp, newp, newlen,
    919 		    &doclusterwrite));
    920 	case FFS_REALLOCBLKS:
    921 		return (sysctl_int(oldp, oldlenp, newp, newlen,
    922 		    &doreallocblks));
    923 	case FFS_ASYNCFREE:
    924 		return (sysctl_int(oldp, oldlenp, newp, newlen, &doasyncfree));
    925 	default:
    926 		return (EOPNOTSUPP);
    927 	}
    928 	/* NOTREACHED */
    929 }
    930 
    931 /*
    932  * Write a superblock and associated information back to disk.
    933  */
    934 int
    935 ffs_sbupdate(mp, waitfor)
    936 	struct ufsmount *mp;
    937 	int waitfor;
    938 {
    939 	register struct fs *dfs, *fs = mp->um_fs;
    940 	register struct buf *bp;
    941 	int blks;
    942 	caddr_t space;
    943 	int i, size, error, allerror = 0;
    944 
    945 	/*
    946 	 * First write back the summary information.
    947 	 */
    948 	blks = howmany(fs->fs_cssize, fs->fs_fsize);
    949 	space = (caddr_t)fs->fs_csp[0];
    950 	for (i = 0; i < blks; i += fs->fs_frag) {
    951 		size = fs->fs_bsize;
    952 		if (i + fs->fs_frag > blks)
    953 			size = (blks - i) * fs->fs_fsize;
    954 		bp = getblk(mp->um_devvp, fsbtodb(fs, fs->fs_csaddr + i),
    955 		    size, 0, 0);
    956 		bcopy(space, bp->b_data, (u_int)size);
    957 		space += size;
    958 		if (waitfor != MNT_WAIT)
    959 			bawrite(bp);
    960 		else if (error = bwrite(bp))
    961 			allerror = error;
    962 	}
    963 	/*
    964 	 * Now write back the superblock itself. If any errors occurred
    965 	 * up to this point, then fail so that the superblock avoids
    966 	 * being written out as clean.
    967 	 */
    968 	if (allerror)
    969 		return (allerror);
    970 	bp = getblk(mp->um_devvp, SBLOCK, (int)fs->fs_sbsize, 0, 0);
    971 	bcopy((caddr_t)fs, bp->b_data, (u_int)fs->fs_sbsize);
    972 	/* Restore compatibility to old file systems.		   XXX */
    973 	dfs = (struct fs *)bp->b_data;				/* XXX */
    974 	if (fs->fs_postblformat == FS_42POSTBLFMT)		/* XXX */
    975 		dfs->fs_nrpos = -1;				/* XXX */
    976 	if (fs->fs_inodefmt < FS_44INODEFMT) {			/* XXX */
    977 		int32_t *lp, tmp;				/* XXX */
    978 								/* XXX */
    979 		lp = (int32_t *)&dfs->fs_qbmask;		/* XXX */
    980 		tmp = lp[4];					/* XXX */
    981 		for (i = 4; i > 0; i--)				/* XXX */
    982 			lp[i] = lp[i-1];			/* XXX */
    983 		lp[0] = tmp;					/* XXX */
    984 	}							/* XXX */
    985 	dfs->fs_maxfilesize = mp->um_savedmaxfilesize;		/* XXX */
    986 	if (waitfor != MNT_WAIT)
    987 		bawrite(bp);
    988 	else if (error = bwrite(bp))
    989 		allerror = error;
    990 	return (allerror);
    991 }
    992