Home | History | Annotate | Line # | Download | only in ffs
ffs_vfsops.c revision 1.23
      1 /*	$NetBSD: ffs_vfsops.c,v 1.23 1997/01/31 03:05:31 thorpej 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/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 
     55 #include <miscfs/specfs/specdev.h>
     56 
     57 #include <ufs/ufs/quota.h>
     58 #include <ufs/ufs/ufsmount.h>
     59 #include <ufs/ufs/inode.h>
     60 #include <ufs/ufs/ufs_extern.h>
     61 
     62 #include <ufs/ffs/fs.h>
     63 #include <ufs/ffs/ffs_extern.h>
     64 
     65 int ffs_sbupdate __P((struct ufsmount *, int));
     66 
     67 struct vfsops ffs_vfsops = {
     68 	MOUNT_FFS,
     69 	ffs_mount,
     70 	ufs_start,
     71 	ffs_unmount,
     72 	ufs_root,
     73 	ufs_quotactl,
     74 	ffs_statfs,
     75 	ffs_sync,
     76 	ffs_vget,
     77 	ffs_fhtovp,
     78 	ffs_vptofh,
     79 	ffs_init,
     80 	ffs_mountroot,
     81 };
     82 
     83 extern u_long nextgennumber;
     84 
     85 /*
     86  * Called by main() when ufs is going to be mounted as root.
     87  *
     88  * Name is updated by mount(8) after booting.
     89  */
     90 #define ROOTNAME	"root_device"
     91 
     92 int
     93 ffs_mountroot()
     94 {
     95 	extern struct vnode *rootvp;
     96 	register struct fs *fs;
     97 	register struct mount *mp;
     98 	struct proc *p = curproc;	/* XXX */
     99 	struct ufsmount *ump;
    100 	size_t size;
    101 	int error;
    102 
    103 	if (root_device->dv_class != DV_DISK)
    104 		return (ENODEV);
    105 
    106 	/*
    107 	 * Get vnodes for swapdev and rootdev.
    108 	 */
    109 	if (bdevvp(swapdev, &swapdev_vp) || bdevvp(rootdev, &rootvp))
    110 		panic("ffs_mountroot: can't setup bdevvp's");
    111 
    112 	mp = malloc((u_long)sizeof(struct mount), M_MOUNT, M_WAITOK);
    113 	bzero((char *)mp, (u_long)sizeof(struct mount));
    114 	mp->mnt_op = &ffs_vfsops;
    115 	mp->mnt_flag = MNT_RDONLY;
    116 	if ((error = ffs_mountfs(rootvp, mp, p)) != 0) {
    117 		free(mp, M_MOUNT);
    118 		return (error);
    119 	}
    120 	if ((error = vfs_lock(mp)) != 0) {
    121 		(void)ffs_unmount(mp, 0, p);
    122 		free(mp, M_MOUNT);
    123 		return (error);
    124 	}
    125 	CIRCLEQ_INSERT_TAIL(&mountlist, mp, mnt_list);
    126 	mp->mnt_vnodecovered = NULLVP;
    127 	ump = VFSTOUFS(mp);
    128 	fs = ump->um_fs;
    129 	bzero(fs->fs_fsmnt, sizeof(fs->fs_fsmnt));
    130 	fs->fs_fsmnt[0] = '/';
    131 	bcopy(fs->fs_fsmnt, mp->mnt_stat.f_mntonname, MNAMELEN);
    132 	(void) copystr(ROOTNAME, mp->mnt_stat.f_mntfromname, MNAMELEN - 1,
    133 	    &size);
    134 	bzero(mp->mnt_stat.f_mntfromname + size, MNAMELEN - size);
    135 	(void)ffs_statfs(mp, &mp->mnt_stat, p);
    136 	vfs_unlock(mp);
    137 	inittodr(fs->fs_time);
    138 	return (0);
    139 }
    140 
    141 /*
    142  * VFS Operations.
    143  *
    144  * mount system call
    145  */
    146 int
    147 ffs_mount(mp, path, data, ndp, p)
    148 	register struct mount *mp;
    149 	const char *path;
    150 	void *data;
    151 	struct nameidata *ndp;
    152 	struct proc *p;
    153 {
    154 	struct vnode *devvp;
    155 	struct ufs_args args;
    156 	struct ufsmount *ump = NULL;
    157 	register struct fs *fs;
    158 	size_t size;
    159 	int error, flags;
    160 	mode_t accessmode;
    161 
    162 	error = copyin(data, (caddr_t)&args, sizeof (struct ufs_args));
    163 	if (error)
    164 		return (error);
    165 	/*
    166 	 * If updating, check whether changing from read-only to
    167 	 * read/write; if there is no device name, that's all we do.
    168 	 */
    169 	if (mp->mnt_flag & MNT_UPDATE) {
    170 		ump = VFSTOUFS(mp);
    171 		fs = ump->um_fs;
    172 		if (fs->fs_ronly == 0 && (mp->mnt_flag & MNT_RDONLY)) {
    173 			flags = WRITECLOSE;
    174 			if (mp->mnt_flag & MNT_FORCE)
    175 				flags |= FORCECLOSE;
    176 			if (vfs_busy(mp))
    177 				return (EBUSY);
    178 			error = ffs_flushfiles(mp, flags, p);
    179 			if (error == 0 &&
    180 			    ffs_cgupdate(ump, MNT_WAIT) == 0 &&
    181 			    fs->fs_clean & FS_WASCLEAN) {
    182 				fs->fs_clean = FS_ISCLEAN;
    183 				(void) ffs_sbupdate(ump, MNT_WAIT);
    184 			}
    185 			vfs_unbusy(mp);
    186 			if (error)
    187 				return (error);
    188 			fs->fs_ronly = 1;
    189 		}
    190 		if (mp->mnt_flag & MNT_RELOAD) {
    191 			error = ffs_reload(mp, ndp->ni_cnd.cn_cred, p);
    192 			if (error)
    193 				return (error);
    194 		}
    195 		if (fs->fs_ronly && (mp->mnt_flag & MNT_WANTRDWR)) {
    196 			/*
    197 			 * If upgrade to read-write by non-root, then verify
    198 			 * that user has necessary permissions on the device.
    199 			 */
    200 			if (p->p_ucred->cr_uid != 0) {
    201 				devvp = ump->um_devvp;
    202 				VOP_LOCK(devvp);
    203 				error = VOP_ACCESS(devvp, VREAD | VWRITE,
    204 						   p->p_ucred, p);
    205 				if (error) {
    206 					VOP_UNLOCK(devvp);
    207 					return (error);
    208 				}
    209 				VOP_UNLOCK(devvp);
    210 			}
    211 			fs->fs_ronly = 0;
    212 			fs->fs_clean <<= 1;
    213 			fs->fs_fmod = 1;
    214 		}
    215 		if (args.fspec == 0) {
    216 			/*
    217 			 * Process export requests.
    218 			 */
    219 			return (vfs_export(mp, &ump->um_export, &args.export));
    220 		}
    221 	}
    222 	/*
    223 	 * Not an update, or updating the name: look up the name
    224 	 * and verify that it refers to a sensible block device.
    225 	 */
    226 	NDINIT(ndp, LOOKUP, FOLLOW, UIO_USERSPACE, args.fspec, p);
    227 	if ((error = namei(ndp)) != 0)
    228 		return (error);
    229 	devvp = ndp->ni_vp;
    230 
    231 	if (devvp->v_type != VBLK) {
    232 		vrele(devvp);
    233 		return (ENOTBLK);
    234 	}
    235 	if (major(devvp->v_rdev) >= nblkdev) {
    236 		vrele(devvp);
    237 		return (ENXIO);
    238 	}
    239 	/*
    240 	 * If mount by non-root, then verify that user has necessary
    241 	 * permissions on the device.
    242 	 */
    243 	if (p->p_ucred->cr_uid != 0) {
    244 		accessmode = VREAD;
    245 		if ((mp->mnt_flag & MNT_RDONLY) == 0)
    246 			accessmode |= VWRITE;
    247 		VOP_LOCK(devvp);
    248 		error = VOP_ACCESS(devvp, accessmode, p->p_ucred, p);
    249 		if (error) {
    250 			vput(devvp);
    251 			return (error);
    252 		}
    253 		VOP_UNLOCK(devvp);
    254 	}
    255 	if ((mp->mnt_flag & MNT_UPDATE) == 0)
    256 		error = ffs_mountfs(devvp, mp, p);
    257 	else {
    258 		if (devvp != ump->um_devvp)
    259 			error = EINVAL;	/* needs translation */
    260 		else
    261 			vrele(devvp);
    262 	}
    263 	if (error) {
    264 		vrele(devvp);
    265 		return (error);
    266 	}
    267 	ump = VFSTOUFS(mp);
    268 	fs = ump->um_fs;
    269 	(void) copyinstr(path, fs->fs_fsmnt, sizeof(fs->fs_fsmnt) - 1, &size);
    270 	bzero(fs->fs_fsmnt + size, sizeof(fs->fs_fsmnt) - size);
    271 	bcopy(fs->fs_fsmnt, mp->mnt_stat.f_mntonname, MNAMELEN);
    272 	(void) copyinstr(args.fspec, mp->mnt_stat.f_mntfromname, MNAMELEN - 1,
    273 	    &size);
    274 	bzero(mp->mnt_stat.f_mntfromname + size, MNAMELEN - size);
    275 	if (fs->fs_fmod != 0) {	/* XXX */
    276 		fs->fs_fmod = 0;
    277 		if (fs->fs_clean & FS_WASCLEAN)
    278 			fs->fs_time = time.tv_sec;
    279 		else
    280 			printf("%s: file system not clean; please fsck(8)\n",
    281 			    mp->mnt_stat.f_mntfromname);
    282 		(void) ffs_cgupdate(ump, MNT_WAIT);
    283 	}
    284 	return (0);
    285 }
    286 
    287 /*
    288  * Reload all incore data for a filesystem (used after running fsck on
    289  * the root filesystem and finding things to fix). The filesystem must
    290  * be mounted read-only.
    291  *
    292  * Things to do to update the mount:
    293  *	1) invalidate all cached meta-data.
    294  *	2) re-read superblock from disk.
    295  *	3) re-read summary information from disk.
    296  *	4) invalidate all inactive vnodes.
    297  *	5) invalidate all cached file data.
    298  *	6) re-read inode data for all active vnodes.
    299  */
    300 int
    301 ffs_reload(mountp, cred, p)
    302 	register struct mount *mountp;
    303 	struct ucred *cred;
    304 	struct proc *p;
    305 {
    306 	register struct vnode *vp, *nvp, *devvp;
    307 	struct inode *ip;
    308 	struct csum *space;
    309 	struct buf *bp;
    310 	struct fs *fs, *newfs;
    311 	struct partinfo dpart;
    312 	int i, blks, size, error;
    313 	int32_t *lp;
    314 
    315 	if ((mountp->mnt_flag & MNT_RDONLY) == 0)
    316 		return (EINVAL);
    317 	/*
    318 	 * Step 1: invalidate all cached meta-data.
    319 	 */
    320 	devvp = VFSTOUFS(mountp)->um_devvp;
    321 	if (vinvalbuf(devvp, 0, cred, p, 0, 0))
    322 		panic("ffs_reload: dirty1");
    323 	/*
    324 	 * Step 2: re-read superblock from disk.
    325 	 */
    326 	if (VOP_IOCTL(devvp, DIOCGPART, (caddr_t)&dpart, FREAD, NOCRED, p) != 0)
    327 		size = DEV_BSIZE;
    328 	else
    329 		size = dpart.disklab->d_secsize;
    330 	error = bread(devvp, (daddr_t)(SBOFF / size), SBSIZE, NOCRED, &bp);
    331 	if (error)
    332 		return (error);
    333 	newfs = (struct fs *)bp->b_data;
    334 	if (newfs->fs_magic != FS_MAGIC || newfs->fs_bsize > MAXBSIZE ||
    335 	    newfs->fs_bsize < sizeof(struct fs)) {
    336 		brelse(bp);
    337 		return (EIO);		/* XXX needs translation */
    338 	}
    339 	fs = VFSTOUFS(mountp)->um_fs;
    340 	/*
    341 	 * Copy pointer fields back into superblock before copying in	XXX
    342 	 * new superblock. These should really be in the ufsmount.	XXX
    343 	 * Note that important parameters (eg fs_ncg) are unchanged.
    344 	 */
    345 	bcopy(&fs->fs_csp[0], &newfs->fs_csp[0], sizeof(fs->fs_csp));
    346 	newfs->fs_maxcluster = fs->fs_maxcluster;
    347 	bcopy(newfs, fs, (u_int)fs->fs_sbsize);
    348 	if (fs->fs_sbsize < SBSIZE)
    349 		bp->b_flags |= B_INVAL;
    350 	brelse(bp);
    351 	mountp->mnt_maxsymlinklen = fs->fs_maxsymlinklen;
    352 	ffs_oldfscompat(fs);
    353 	/*
    354 	 * Step 3: re-read summary information from disk.
    355 	 */
    356 	blks = howmany(fs->fs_cssize, fs->fs_fsize);
    357 	space = fs->fs_csp[0];
    358 	for (i = 0; i < blks; i += fs->fs_frag) {
    359 		size = fs->fs_bsize;
    360 		if (i + fs->fs_frag > blks)
    361 			size = (blks - i) * fs->fs_fsize;
    362 		error = bread(devvp, fsbtodb(fs, fs->fs_csaddr + i), size,
    363 			      NOCRED, &bp);
    364 		if (error)
    365 			return (error);
    366 		bcopy(bp->b_data, fs->fs_csp[fragstoblks(fs, i)], (u_int)size);
    367 		brelse(bp);
    368 	}
    369 	/*
    370 	 * We no longer know anything about clusters per cylinder group.
    371 	 */
    372 	if (fs->fs_contigsumsize > 0) {
    373 		lp = fs->fs_maxcluster;
    374 		for (i = 0; i < fs->fs_ncg; i++)
    375 			*lp++ = fs->fs_contigsumsize;
    376 	}
    377 
    378 loop:
    379 	for (vp = mountp->mnt_vnodelist.lh_first; vp != NULL; vp = nvp) {
    380 		nvp = vp->v_mntvnodes.le_next;
    381 		/*
    382 		 * Step 4: invalidate all inactive vnodes.
    383 		 */
    384 		if (vp->v_usecount == 0) {
    385 			vgone(vp);
    386 			continue;
    387 		}
    388 		/*
    389 		 * Step 5: invalidate all cached file data.
    390 		 */
    391 		if (vget(vp, 1))
    392 			goto loop;
    393 		if (vinvalbuf(vp, 0, cred, p, 0, 0))
    394 			panic("ffs_reload: dirty2");
    395 		/*
    396 		 * Step 6: re-read inode data for all active vnodes.
    397 		 */
    398 		ip = VTOI(vp);
    399 		error = bread(devvp, fsbtodb(fs, ino_to_fsba(fs, ip->i_number)),
    400 			      (int)fs->fs_bsize, NOCRED, &bp);
    401 		if (error) {
    402 			vput(vp);
    403 			return (error);
    404 		}
    405 		ip->i_din = *((struct dinode *)bp->b_data +
    406 		    ino_to_fsbo(fs, ip->i_number));
    407 		brelse(bp);
    408 		vput(vp);
    409 		if (vp->v_mount != mountp)
    410 			goto loop;
    411 	}
    412 	return (0);
    413 }
    414 
    415 /*
    416  * Common code for mount and mountroot
    417  */
    418 int
    419 ffs_mountfs(devvp, mp, p)
    420 	register struct vnode *devvp;
    421 	struct mount *mp;
    422 	struct proc *p;
    423 {
    424 	register struct ufsmount *ump;
    425 	struct buf *bp;
    426 	register struct fs *fs;
    427 	dev_t dev;
    428 	struct partinfo dpart;
    429 	caddr_t base, space;
    430 	int blks;
    431 	int error, i, size, ronly;
    432 	int32_t *lp;
    433 	struct ucred *cred;
    434 	extern struct vnode *rootvp;
    435 	u_int64_t maxfilesize;					/* XXX */
    436 
    437 	dev = devvp->v_rdev;
    438 	cred = p ? p->p_ucred : NOCRED;
    439 	/*
    440 	 * Disallow multiple mounts of the same device.
    441 	 * Disallow mounting of a device that is currently in use
    442 	 * (except for root, which might share swap device for miniroot).
    443 	 * Flush out any old buffers remaining from a previous use.
    444 	 */
    445 	if ((error = vfs_mountedon(devvp)) != 0)
    446 		return (error);
    447 	if (vcount(devvp) > 1 && devvp != rootvp)
    448 		return (EBUSY);
    449 	if ((error = vinvalbuf(devvp, V_SAVE, cred, p, 0, 0)) != 0)
    450 		return (error);
    451 
    452 	ronly = (mp->mnt_flag & MNT_RDONLY) != 0;
    453 	error = VOP_OPEN(devvp, ronly ? FREAD : FREAD|FWRITE, FSCRED, p);
    454 	if (error)
    455 		return (error);
    456 	if (VOP_IOCTL(devvp, DIOCGPART, (caddr_t)&dpart, FREAD, cred, p) != 0)
    457 		size = DEV_BSIZE;
    458 	else
    459 		size = dpart.disklab->d_secsize;
    460 
    461 	bp = NULL;
    462 	ump = NULL;
    463 	error = bread(devvp, (daddr_t)(SBOFF / size), SBSIZE, cred, &bp);
    464 	if (error)
    465 		goto out;
    466 	fs = (struct fs *)bp->b_data;
    467 	if (fs->fs_magic != FS_MAGIC || fs->fs_bsize > MAXBSIZE ||
    468 	    fs->fs_bsize < sizeof(struct fs)) {
    469 		error = EINVAL;		/* XXX needs translation */
    470 		goto out;
    471 	}
    472 	/* XXX updating 4.2 FFS superblocks trashes rotational layout tables */
    473 	if (fs->fs_postblformat == FS_42POSTBLFMT && !ronly) {
    474 		error = EROFS;		/* XXX what should be returned? */
    475 		goto out;
    476 	}
    477 	ump = malloc(sizeof *ump, M_UFSMNT, M_WAITOK);
    478 	bzero((caddr_t)ump, sizeof *ump);
    479 	ump->um_fs = malloc((u_long)fs->fs_sbsize, M_UFSMNT,
    480 	    M_WAITOK);
    481 	bcopy(bp->b_data, ump->um_fs, (u_int)fs->fs_sbsize);
    482 	if (fs->fs_sbsize < SBSIZE)
    483 		bp->b_flags |= B_INVAL;
    484 	brelse(bp);
    485 	bp = NULL;
    486 	fs = ump->um_fs;
    487 	fs->fs_ronly = ronly;
    488 	if (ronly == 0) {
    489 		fs->fs_clean <<= 1;
    490 		fs->fs_fmod = 1;
    491 	}
    492 	size = fs->fs_cssize;
    493 	blks = howmany(size, fs->fs_fsize);
    494 	if (fs->fs_contigsumsize > 0)
    495 		size += fs->fs_ncg * sizeof(int32_t);
    496 	base = space = malloc((u_long)size, M_UFSMNT, M_WAITOK);
    497 	for (i = 0; i < blks; i += fs->fs_frag) {
    498 		size = fs->fs_bsize;
    499 		if (i + fs->fs_frag > blks)
    500 			size = (blks - i) * fs->fs_fsize;
    501 		error = bread(devvp, fsbtodb(fs, fs->fs_csaddr + i), size,
    502 			      cred, &bp);
    503 		if (error) {
    504 			free(base, M_UFSMNT);
    505 			goto out;
    506 		}
    507 		bcopy(bp->b_data, space, (u_int)size);
    508 		fs->fs_csp[fragstoblks(fs, i)] = (struct csum *)space;
    509 		space += size;
    510 		brelse(bp);
    511 		bp = NULL;
    512 	}
    513 	if (fs->fs_contigsumsize > 0) {
    514 		fs->fs_maxcluster = lp = (int32_t *)space;
    515 		for (i = 0; i < fs->fs_ncg; i++)
    516 			*lp++ = fs->fs_contigsumsize;
    517 	}
    518 	mp->mnt_data = (qaddr_t)ump;
    519 	mp->mnt_stat.f_fsid.val[0] = (long)dev;
    520 	mp->mnt_stat.f_fsid.val[1] = makefstype(MOUNT_FFS);
    521 	mp->mnt_maxsymlinklen = fs->fs_maxsymlinklen;
    522 	mp->mnt_flag |= MNT_LOCAL;
    523 	ump->um_mountp = mp;
    524 	ump->um_dev = dev;
    525 	ump->um_devvp = devvp;
    526 	ump->um_nindir = fs->fs_nindir;
    527 	ump->um_bptrtodb = fs->fs_fsbtodb;
    528 	ump->um_seqinc = fs->fs_frag;
    529 	for (i = 0; i < MAXQUOTAS; i++)
    530 		ump->um_quotas[i] = NULLVP;
    531 	devvp->v_specflags |= SI_MOUNTEDON;
    532 	ffs_oldfscompat(fs);
    533 	ump->um_savedmaxfilesize = fs->fs_maxfilesize;		/* XXX */
    534 	maxfilesize = (u_int64_t)0x80000000 * fs->fs_bsize - 1;	/* XXX */
    535 	if (fs->fs_maxfilesize > maxfilesize)			/* XXX */
    536 		fs->fs_maxfilesize = maxfilesize;		/* XXX */
    537 	return (0);
    538 out:
    539 	if (bp)
    540 		brelse(bp);
    541 	(void)VOP_CLOSE(devvp, ronly ? FREAD : FREAD|FWRITE, cred, p);
    542 	if (ump) {
    543 		free(ump->um_fs, M_UFSMNT);
    544 		free(ump, M_UFSMNT);
    545 		mp->mnt_data = (qaddr_t)0;
    546 	}
    547 	return (error);
    548 }
    549 
    550 /*
    551  * Sanity checks for old file systems.
    552  *
    553  * XXX - goes away some day.
    554  */
    555 int
    556 ffs_oldfscompat(fs)
    557 	struct fs *fs;
    558 {
    559 	int i;
    560 
    561 	fs->fs_npsect = max(fs->fs_npsect, fs->fs_nsect);	/* XXX */
    562 	fs->fs_interleave = max(fs->fs_interleave, 1);		/* XXX */
    563 	if (fs->fs_postblformat == FS_42POSTBLFMT)		/* XXX */
    564 		fs->fs_nrpos = 8;				/* XXX */
    565 	if (fs->fs_inodefmt < FS_44INODEFMT) {			/* XXX */
    566 		u_int64_t sizepb = fs->fs_bsize;		/* XXX */
    567 								/* XXX */
    568 		fs->fs_maxfilesize = fs->fs_bsize * NDADDR - 1;	/* XXX */
    569 		for (i = 0; i < NIADDR; i++) {			/* XXX */
    570 			sizepb *= NINDIR(fs);			/* XXX */
    571 			fs->fs_maxfilesize += sizepb;		/* XXX */
    572 		}						/* XXX */
    573 		fs->fs_qbmask = ~fs->fs_bmask;			/* XXX */
    574 		fs->fs_qfmask = ~fs->fs_fmask;			/* XXX */
    575 	}							/* XXX */
    576 	return (0);
    577 }
    578 
    579 /*
    580  * unmount system call
    581  */
    582 int
    583 ffs_unmount(mp, mntflags, p)
    584 	struct mount *mp;
    585 	int mntflags;
    586 	struct proc *p;
    587 {
    588 	register struct ufsmount *ump;
    589 	register struct fs *fs;
    590 	int error, flags;
    591 
    592 	flags = 0;
    593 	if (mntflags & MNT_FORCE)
    594 		flags |= FORCECLOSE;
    595 	if ((error = ffs_flushfiles(mp, flags, p)) != 0)
    596 		return (error);
    597 	ump = VFSTOUFS(mp);
    598 	fs = ump->um_fs;
    599 	if (fs->fs_ronly == 0 &&
    600 	    ffs_cgupdate(ump, MNT_WAIT) == 0 &&
    601 	    fs->fs_clean & FS_WASCLEAN) {
    602 		fs->fs_clean = FS_ISCLEAN;
    603 		(void) ffs_sbupdate(ump, MNT_WAIT);
    604 	}
    605 	ump->um_devvp->v_specflags &= ~SI_MOUNTEDON;
    606 	error = VOP_CLOSE(ump->um_devvp, fs->fs_ronly ? FREAD : FREAD|FWRITE,
    607 		NOCRED, p);
    608 	vrele(ump->um_devvp);
    609 	free(fs->fs_csp[0], M_UFSMNT);
    610 	free(fs, M_UFSMNT);
    611 	free(ump, M_UFSMNT);
    612 	mp->mnt_data = (qaddr_t)0;
    613 	mp->mnt_flag &= ~MNT_LOCAL;
    614 	return (error);
    615 }
    616 
    617 /*
    618  * Flush out all the files in a filesystem.
    619  */
    620 int
    621 ffs_flushfiles(mp, flags, p)
    622 	register struct mount *mp;
    623 	int flags;
    624 	struct proc *p;
    625 {
    626 	extern int doforce;
    627 	register struct ufsmount *ump;
    628 	int error;
    629 
    630 	if (!doforce)
    631 		flags &= ~FORCECLOSE;
    632 	ump = VFSTOUFS(mp);
    633 #ifdef QUOTA
    634 	if (mp->mnt_flag & MNT_QUOTA) {
    635 		int i;
    636 		if ((error = vflush(mp, NULLVP, SKIPSYSTEM|flags)) != 0)
    637 			return (error);
    638 		for (i = 0; i < MAXQUOTAS; i++) {
    639 			if (ump->um_quotas[i] == NULLVP)
    640 				continue;
    641 			quotaoff(p, mp, i);
    642 		}
    643 		/*
    644 		 * Here we fall through to vflush again to ensure
    645 		 * that we have gotten rid of all the system vnodes.
    646 		 */
    647 	}
    648 #endif
    649 	error = vflush(mp, NULLVP, flags);
    650 	return (error);
    651 }
    652 
    653 /*
    654  * Get file system statistics.
    655  */
    656 int
    657 ffs_statfs(mp, sbp, p)
    658 	struct mount *mp;
    659 	register struct statfs *sbp;
    660 	struct proc *p;
    661 {
    662 	register struct ufsmount *ump;
    663 	register struct fs *fs;
    664 
    665 	ump = VFSTOUFS(mp);
    666 	fs = ump->um_fs;
    667 	if (fs->fs_magic != FS_MAGIC)
    668 		panic("ffs_statfs");
    669 #ifdef COMPAT_09
    670 	sbp->f_type = 1;
    671 #else
    672 	sbp->f_type = 0;
    673 #endif
    674 	sbp->f_bsize = fs->fs_fsize;
    675 	sbp->f_iosize = fs->fs_bsize;
    676 	sbp->f_blocks = fs->fs_dsize;
    677 	sbp->f_bfree = fs->fs_cstotal.cs_nbfree * fs->fs_frag +
    678 		fs->fs_cstotal.cs_nffree;
    679 	sbp->f_bavail = (fs->fs_dsize * (100 - fs->fs_minfree) / 100) -
    680 		(fs->fs_dsize - sbp->f_bfree);
    681 	sbp->f_files =  fs->fs_ncg * fs->fs_ipg - ROOTINO;
    682 	sbp->f_ffree = fs->fs_cstotal.cs_nifree;
    683 	if (sbp != &mp->mnt_stat) {
    684 		bcopy(mp->mnt_stat.f_mntonname, sbp->f_mntonname, MNAMELEN);
    685 		bcopy(mp->mnt_stat.f_mntfromname, sbp->f_mntfromname, MNAMELEN);
    686 	}
    687 	strncpy(sbp->f_fstypename, mp->mnt_op->vfs_name, MFSNAMELEN);
    688 	return (0);
    689 }
    690 
    691 /*
    692  * Go through the disk queues to initiate sandbagged IO;
    693  * go through the inodes to write those that have been modified;
    694  * initiate the writing of the super block if it has been modified.
    695  *
    696  * Note: we are always called with the filesystem marked `MPBUSY'.
    697  */
    698 int
    699 ffs_sync(mp, waitfor, cred, p)
    700 	struct mount *mp;
    701 	int waitfor;
    702 	struct ucred *cred;
    703 	struct proc *p;
    704 {
    705 	register struct vnode *vp;
    706 	register struct inode *ip;
    707 	register struct ufsmount *ump = VFSTOUFS(mp);
    708 	register struct fs *fs;
    709 	int error, allerror = 0;
    710 
    711 	fs = ump->um_fs;
    712 	/*
    713 	 * Write back modified superblock.
    714 	 * Consistency check that the superblock
    715 	 * is still in the buffer cache.
    716 	 */
    717 	if (fs->fs_fmod != 0) {
    718 		if (fs->fs_ronly != 0) {		/* XXX */
    719 			printf("fs = %s\n", fs->fs_fsmnt);
    720 			panic("update: rofs mod");
    721 		}
    722 		fs->fs_fmod = 0;
    723 		fs->fs_time = time.tv_sec;
    724 		allerror = ffs_cgupdate(ump, waitfor);
    725 	}
    726 	/*
    727 	 * Write back each (modified) inode.
    728 	 */
    729 loop:
    730 	for (vp = mp->mnt_vnodelist.lh_first;
    731 	     vp != NULL;
    732 	     vp = vp->v_mntvnodes.le_next) {
    733 		/*
    734 		 * If the vnode that we are about to sync is no longer
    735 		 * associated with this mount point, start over.
    736 		 */
    737 		if (vp->v_mount != mp)
    738 			goto loop;
    739 		if (VOP_ISLOCKED(vp))
    740 			continue;
    741 		ip = VTOI(vp);
    742 		if ((ip->i_flag &
    743 		    (IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE)) == 0 &&
    744 		    vp->v_dirtyblkhd.lh_first == NULL)
    745 			continue;
    746 		if (vget(vp, 1))
    747 			goto loop;
    748 		if ((error = VOP_FSYNC(vp, cred, waitfor, p)) != 0)
    749 			allerror = error;
    750 		vput(vp);
    751 	}
    752 	/*
    753 	 * Force stale file system control information to be flushed.
    754 	 */
    755 	if ((error = VOP_FSYNC(ump->um_devvp, cred, waitfor, p)) != 0)
    756 		allerror = error;
    757 #ifdef QUOTA
    758 	qsync(mp);
    759 #endif
    760 	return (allerror);
    761 }
    762 
    763 /*
    764  * Look up a FFS dinode number to find its incore vnode, otherwise read it
    765  * in from disk.  If it is in core, wait for the lock bit to clear, then
    766  * return the inode locked.  Detection and handling of mount points must be
    767  * done by the calling routine.
    768  */
    769 int
    770 ffs_vget(mp, ino, vpp)
    771 	struct mount *mp;
    772 	ino_t ino;
    773 	struct vnode **vpp;
    774 {
    775 	register struct fs *fs;
    776 	register struct inode *ip;
    777 	struct ufsmount *ump;
    778 	struct buf *bp;
    779 	struct vnode *vp;
    780 	dev_t dev;
    781 	int type, error;
    782 
    783 	ump = VFSTOUFS(mp);
    784 	dev = ump->um_dev;
    785 	if ((*vpp = ufs_ihashget(dev, ino)) != NULL)
    786 		return (0);
    787 
    788 	/* Allocate a new vnode/inode. */
    789 	if ((error = getnewvnode(VT_UFS, mp, ffs_vnodeop_p, &vp)) != 0) {
    790 		*vpp = NULL;
    791 		return (error);
    792 	}
    793 	type = ump->um_devvp->v_tag == VT_MFS ? M_MFSNODE : M_FFSNODE; /* XXX */
    794 	MALLOC(ip, struct inode *, sizeof(struct inode), type, M_WAITOK);
    795 	bzero((caddr_t)ip, sizeof(struct inode));
    796 	vp->v_data = ip;
    797 	ip->i_vnode = vp;
    798 	ip->i_fs = fs = ump->um_fs;
    799 	ip->i_dev = dev;
    800 	ip->i_number = ino;
    801 #ifdef QUOTA
    802 	{
    803 		int i;
    804 
    805 		for (i = 0; i < MAXQUOTAS; i++)
    806 			ip->i_dquot[i] = NODQUOT;
    807 	}
    808 #endif
    809 	/*
    810 	 * Put it onto its hash chain and lock it so that other requests for
    811 	 * this inode will block if they arrive while we are sleeping waiting
    812 	 * for old data structures to be purged or for the contents of the
    813 	 * disk portion of this inode to be read.
    814 	 */
    815 	ufs_ihashins(ip);
    816 
    817 	/* Read in the disk contents for the inode, copy into the inode. */
    818 	error = bread(ump->um_devvp, fsbtodb(fs, ino_to_fsba(fs, ino)),
    819 		      (int)fs->fs_bsize, NOCRED, &bp);
    820 	if (error) {
    821 		/*
    822 		 * The inode does not contain anything useful, so it would
    823 		 * be misleading to leave it on its hash chain. With mode
    824 		 * still zero, it will be unlinked and returned to the free
    825 		 * list by vput().
    826 		 */
    827 		vput(vp);
    828 		brelse(bp);
    829 		*vpp = NULL;
    830 		return (error);
    831 	}
    832 	ip->i_din = *((struct dinode *)bp->b_data + ino_to_fsbo(fs, ino));
    833 	brelse(bp);
    834 
    835 	/*
    836 	 * Initialize the vnode from the inode, check for aliases.
    837 	 * Note that the underlying vnode may have changed.
    838 	 */
    839 	error = ufs_vinit(mp, ffs_specop_p, FFS_FIFOOPS, &vp);
    840 	if (error) {
    841 		vput(vp);
    842 		*vpp = NULL;
    843 		return (error);
    844 	}
    845 	/*
    846 	 * Finish inode initialization now that aliasing has been resolved.
    847 	 */
    848 	ip->i_devvp = ump->um_devvp;
    849 	VREF(ip->i_devvp);
    850 	/*
    851 	 * Set up a generation number for this inode if it does not
    852 	 * already have one. This should only happen on old filesystems.
    853 	 */
    854 	if (ip->i_gen == 0) {
    855 		if (++nextgennumber < (u_long)time.tv_sec)
    856 			nextgennumber = time.tv_sec;
    857 		ip->i_gen = nextgennumber;
    858 		if ((vp->v_mount->mnt_flag & MNT_RDONLY) == 0)
    859 			ip->i_flag |= IN_MODIFIED;
    860 	}
    861 	/*
    862 	 * Ensure that uid and gid are correct. This is a temporary
    863 	 * fix until fsck has been changed to do the update.
    864 	 */
    865 	if (fs->fs_inodefmt < FS_44INODEFMT) {		/* XXX */
    866 		ip->i_uid = ip->i_din.di_ouid;		/* XXX */
    867 		ip->i_gid = ip->i_din.di_ogid;		/* XXX */
    868 	}						/* XXX */
    869 
    870 	*vpp = vp;
    871 	return (0);
    872 }
    873 
    874 /*
    875  * File handle to vnode
    876  *
    877  * Have to be really careful about stale file handles:
    878  * - check that the inode number is valid
    879  * - call ffs_vget() to get the locked inode
    880  * - check for an unallocated inode (i_mode == 0)
    881  * - check that the given client host has export rights and return
    882  *   those rights via. exflagsp and credanonp
    883  */
    884 int
    885 ffs_fhtovp(mp, fhp, nam, vpp, exflagsp, credanonp)
    886 	register struct mount *mp;
    887 	struct fid *fhp;
    888 	struct mbuf *nam;
    889 	struct vnode **vpp;
    890 	int *exflagsp;
    891 	struct ucred **credanonp;
    892 {
    893 	register struct ufid *ufhp;
    894 	struct fs *fs;
    895 
    896 	ufhp = (struct ufid *)fhp;
    897 	fs = VFSTOUFS(mp)->um_fs;
    898 	if (ufhp->ufid_ino < ROOTINO ||
    899 	    ufhp->ufid_ino >= fs->fs_ncg * fs->fs_ipg)
    900 		return (ESTALE);
    901 	return (ufs_check_export(mp, ufhp, nam, vpp, exflagsp, credanonp));
    902 }
    903 
    904 /*
    905  * Vnode pointer to File handle
    906  */
    907 /* ARGSUSED */
    908 int
    909 ffs_vptofh(vp, fhp)
    910 	struct vnode *vp;
    911 	struct fid *fhp;
    912 {
    913 	register struct inode *ip;
    914 	register struct ufid *ufhp;
    915 
    916 	ip = VTOI(vp);
    917 	ufhp = (struct ufid *)fhp;
    918 	ufhp->ufid_len = sizeof(struct ufid);
    919 	ufhp->ufid_ino = ip->i_number;
    920 	ufhp->ufid_gen = ip->i_gen;
    921 	return (0);
    922 }
    923 
    924 /*
    925  * Write a superblock and associated information back to disk.
    926  */
    927 int
    928 ffs_sbupdate(mp, waitfor)
    929 	struct ufsmount *mp;
    930 	int waitfor;
    931 {
    932 	register struct fs *dfs, *fs = mp->um_fs;
    933 	register struct buf *bp;
    934 	int i, error = 0;
    935 
    936 	bp = getblk(mp->um_devvp, SBOFF >> (fs->fs_fshift - fs->fs_fsbtodb),
    937 	    (int)fs->fs_sbsize, 0, 0);
    938 	bcopy((caddr_t)fs, bp->b_data, (u_int)fs->fs_sbsize);
    939 	/* Restore compatibility to old file systems.		   XXX */
    940 	dfs = (struct fs *)bp->b_data;				/* XXX */
    941 	if (fs->fs_postblformat == FS_42POSTBLFMT)		/* XXX */
    942 		dfs->fs_nrpos = -1;				/* XXX */
    943 	if (fs->fs_inodefmt < FS_44INODEFMT) {			/* XXX */
    944 		int32_t *lp, tmp;				/* XXX */
    945 								/* XXX */
    946 		lp = (int32_t *)&dfs->fs_qbmask;		/* XXX */
    947 		tmp = lp[4];					/* XXX */
    948 		for (i = 4; i > 0; i--)				/* XXX */
    949 			lp[i] = lp[i-1];			/* XXX */
    950 		lp[0] = tmp;					/* XXX */
    951 	}							/* XXX */
    952 	dfs->fs_maxfilesize = mp->um_savedmaxfilesize;		/* XXX */
    953 	if (waitfor == MNT_WAIT)
    954 		error = bwrite(bp);
    955 	else
    956 		bawrite(bp);
    957 	return (error);
    958 }
    959 
    960 int
    961 ffs_cgupdate(mp, waitfor)
    962 	struct ufsmount *mp;
    963 	int waitfor;
    964 {
    965 	register struct fs *fs = mp->um_fs;
    966 	register struct buf *bp;
    967 	int blks;
    968 	caddr_t space;
    969 	int i, size, error = 0, allerror = 0;
    970 
    971 	allerror = ffs_sbupdate(mp, waitfor);
    972 	blks = howmany(fs->fs_cssize, fs->fs_fsize);
    973 	space = (caddr_t)fs->fs_csp[0];
    974 	for (i = 0; i < blks; i += fs->fs_frag) {
    975 		size = fs->fs_bsize;
    976 		if (i + fs->fs_frag > blks)
    977 			size = (blks - i) * fs->fs_fsize;
    978 		bp = getblk(mp->um_devvp, fsbtodb(fs, fs->fs_csaddr + i),
    979 		    size, 0, 0);
    980 		bcopy(space, bp->b_data, (u_int)size);
    981 		space += size;
    982 		if (waitfor == MNT_WAIT)
    983 			error = bwrite(bp);
    984 		else
    985 			bawrite(bp);
    986 	}
    987 	if (!allerror && error)
    988 		allerror = error;
    989 	return (allerror);
    990 }
    991